DETAILED ACTION
This Office Action is in response to communications filed on February 13th, 2026 for Application No. 18/177,052, in which claims 1-15, 18-22 and 24-26 are presented for examination. The amendments filed February 13th, 2026 have been entered, where claims 1, 6-7, 10-11, 14, and 22 are amended, claims 16-17 and 23 are canceled, and claims 24-26 are added.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-2, 8-11, 14, 22, and 24-26 are rejected under 35 U.S.C. 103 as being unpatentable over Mimaroglu et al. (hereinafter Mimaroglu) (Patent Pub. No. US 2021/0158225 A1) in view of Hoffmann et al. (hereinafter Hoffmann) (“Automated Detection of Electric Vehicles in Hourly Smart Meter Data”) and Sahani et al. (hereinafter Sahani) (“Moving horizon-based optimal scheduling of EV charging: A power system-cognizant approach”).
Regarding Claim 1, Mimaroglu teaches a method, comprising (Para. [0003], “The embodiments of the present disclosure are generally directed to systems and methods for non-intrusive load monitoring using ensemble machine learning techniques”):
receiving, by one or more processors, first electricity usage data values for a training population of a plurality of user locations (Fig. 3; Para. [0048], “In some embodiments, prediction module 306 can be a machine learning module (e.g., neural network) that is trained by training data 308. For example, training data 308 can include labeled data, such as energy usage data values from a plurality of source locations (e.g., source locations 102 and 106 from FIG. 1)”, where the “plurality of source locations” are a plurality of user locations and, in the context of “electric vehicles” and “electric grid planning”, the “energy usage” is “electric”, see Para. [0022]- [0023], “Accurate disaggregation via NILM provides many benefits including energy savings opportunities, personalization, improved electric grid planning, and more . . . of many energy consuming devices, such as large household appliances and electric vehicles”; see generally Para. [0108], “In some embodiments, the functionality of FIGS. 7-11 can be implemented by software stored in memory or other computer-readable or tangible medium, and executed by a processor”, where, while the “processor” is specifically discussed in regard to “the functionality of FIGS. 7-11” a person of ordinary skill in the art would understand a processor to be required to execute the claimed functionality disclosed in the other figures)
over a first . . . time period, the first electricity usage data values having a temporal resolution of multiple intervals per day (Para. [0068], Table 1; Para. [0069], “This example includes a granularity of 15 minutes , but other suitable granularities can similarly be implemented (e.g., 1 min, 5 mins, 15 mins, 30 mins, 1 hour, hours, and the like). In some embodiments, processing the energy usage data (e.g., to generate training data 308 ) can include reducing a granularity of the data, for example so that it can be used to generate a training corpus with a consistent granularity ( e.g. , 1 hour)”, where any of the example “granularities/resolutions” are at the temporal length of multiple intervals per day; see also Para. [0024], “training data can be used to train learning models designed to effectively learn in these challenging conditions. Input to the learning models can be provided by AMI along with other types of inputs. Embodiments can accurately predict electric device energy usage in high and low granularities/resolutions (e.g. , at 1 min , 5 mins , 15 mins , 30 mins , 1 hour, or more)”);
receiving, by the one or more processors, for a plurality of the training population's user locations a corresponding label indicating whether or not an electrical vehicle (EV) uses the user location for charging (Para. [0018], “Embodiments train a machine learning model using labeled energy usage data . . . Energy usage data from multiple source locations (e.g., households) can be obtained, where the energy usage data can be labeled with device specific energy usage. For example, the household energy usage values can cover a period of time, and within that period of time individual device energy usage values (e.g., appliance 1, electric vehicle 1, appliance 2, and the like) can be labeled”, where the “training . . . data”, which “can be labeled”, is received from “multiple source locations”; see also Fig. 7; Para. [0068], Table 1; and Para. [0109], “the energy usage data can be similar to the data illustrated in Table 1 above. In some embodiments, the received data can include a timestamp, overall energy usage (which includes energy used by a plurality of devices) at a source location (e.g., household), and labeled device specific energy usage for one or multiple of the target and non-target devices”, where labels “EV” with corresponding values other than “0.0” indicates that a EV uses the “source location” for charging, whereas a value of “0.0” indicates it is not; see generally Para. [0108], “In some embodiments, the functionality of FIGS. 7-11 can be implemented by software stored in memory or other computer-readable or tangible medium, and executed by a processor”);
selecting by the one or more processors, from user locations having a corresponding label indicating that an EV does not use the user location for charging, a first subset of the training population (Para. [0077], “processing device specific labeled energy usage data for a source location can include replacing null values (or any other place holder value) with zero values. For example, when it is determined that a particular device is to be used in the training techniques for a given implementation of prediction module 306 and that portions of the energy usage data lack labeled device specific energy usage for the particular device (at one or more source locations), the labeled energy usage values for the particular device that are missing can be replaced with zero values”, where “data for a source location” for “training” can be selected as a subset of the training population where the “location” does has a corresponding label indicating an EV does not use the location for charging, such as “null” for a “device” in the context of “electric vehicles”, see Para. [0032], “Implementations and results demonstrate improved disaggregation predictions for multiple energy consuming devices (e.g., large household appliances and/or electric vehicles)”, see Para. [0068], Table 1, where the subset can be every row in the table where “EV” is “0.0”, or a subset of rows where “EV” is “0.0”, such as rows 6-9; see generally Para. [0076], where the replacement process is discussed for a specific, but not exhaustive, example; see generally Para. [0108], “In some embodiments, the functionality of FIGS. 7-11 can be implemented by software stored in memory or other computer-readable or tangible medium, and executed by a processor”);
extracting features, including time series data features, by the one or more processors from the first electricity usage data values for the training population (Para. [0112], “At 706, the energy usage data can be processed to generate training data. For example, the energy usage data can be processed based on the target device and a set of other devices (e.g., one or more non-target devices) to generate training data. Processing can be based on a correspondence between the energy usage data (e.g., the availability of labeled device specific energy usage for various devices within the energy usage data) and the set of other devices”, where the “energy usage data” is “processed” to be used as “training data”; Pgs. 7-9, Tables 1-6 and Para. [0109], “the energy usage data can be similar to the data illustrated in Table 1 above. In some embodiments, the received data can include a timestamp, overall energy usage (which includes energy used by a plurality of devices) at a source location (e.g., household), and labeled device specific energy usage for one or multiple of the target and non-target devices”, where the processing includes extracting features for the training population, such as time series data features like “timestamp[s]” associated with “energy usage” values; see generally Fig. 3 and Para. [0047], “FIG. 3 illustrates a system for using a machine learning model to disaggregate energy usage associated with a target device according to an example embodiment. System 300 includes input data 302, processing module 304, prediction module 306, training data 308, and output data 310 . . . processing module 304 can process input data 302 to generate features based on the input data”, where, while discussed in regard to “generat[ing] features based on the input data”, the features from the “energy usage data”, discussed above, must similarly be extracted as the “training data” and inputted to the “prediction model” for “training”; see generally Para. [0108], “In some embodiments, the functionality of FIGS. 7-11 can be implemented by software stored in memory or other computer-readable or tangible medium, and executed by a processor”);
training a machine learning model on the one or more processors using the extracted features from the first electricity usage data values and the corresponding labels of a combination of the first subset of the training population and of the user locations of the training population having a label indicating that an EV uses the user location (Para. [0048], “In some embodiments, prediction module 306 can be a machine learning module (e.g., neural network) that is trained by training data 308. For example, training data 308 can include labeled data, such as energy usage data values from a plurality of source locations (e.g., source locations 102 and 106 from FIG. 1) that include labeled device specific energy usage data values”, where the extracted features, “data”, and the corresponding “labels” are used for “train[ing]”, which includes the “0.0” values, converted from “null”, discussed above, Para. [0068], Table 1, where the subset can be every row in the table where “EV” is “0.0”, or a subset of rows where “EV” is “0.0”, such as rows 6-9; Para. [0068], Table 1; and Para. [0109], “the energy usage data can be similar to the data illustrated in Table 1 above. In some embodiments, the received data can include a timestamp, overall energy usage (which includes energy used by a plurality of devices) at a source location (e.g., household), and labeled device specific energy usage for one or multiple of the target and non-target devices”, where labels “EV” with corresponding values other than “0.0” indicates that a EV uses the “source location” for charging, where the subset of the training population and of the user locations of the training population having a label indicating that an EV uses the user location are combined as “training data 308”, where the combination comprises the entire training data if the subset is all “0.0” values, or a subset of the combination if it is a subset of all “0.0” values; see generally Para. [0108], “In some embodiments, the functionality of FIGS. 7-11 can be implemented by software stored in memory or other computer-readable or tangible medium, and executed by a processor”);
receiving by the one or more processors . . . [information] for a first network element of an electrical distribution network (Para. [0034], “source location 102 can be supplied with power (e.g., electricity), and devices 110, 112, and 114 can draw from the power supplied to source location 102. In some embodiments, source location 102 is a household and the power to the household is supplied from an electric power grid, a local power source (e.g., solar panels), a combination of these, or any other suitable source”, where any of “an electric power grid, a local power source (e.g., solar panels), a combination of these, or any other suitable source” comprises a first network element, which is part of an electrical distribution network to “suppl[y] . . . power (e.g., electricity)” to the network of “source location 102” “household[s]”, which requires receiving information for the first network element; see generally Para. [0108], “In some embodiments, the functionality of FIGS. 7-11 can be implemented by software stored in memory or other computer-readable or tangible medium, and executed by a processor”);
receiving by the one or more processors second electricity usage data values for an inferencing population of a plurality of user locations supplied with electricity through the first network element over a second . . . time period (Fig. 8, where the second electricity usage data values for an inferencing population is the “INPUT DATA”, “RECEIV[ED]” at “802”; Para. [0120], “At 802, household energy usage data can be received over a period of time”, where the “energy usage data” is over a “period of time” for a “household” location; see generally Para. [0029], “Motivations for these investments include the advancement of AMI and smart grid technologies, a growing interest in energy efficiency, interest from customers for better information, and the like”, where the intended use is for a plurality of customers, and therefore, a plurality of households; see generally Para. [0108], “In some embodiments, the functionality of FIGS. 7-11 can be implemented by software stored in memory or other computer-readable or tangible medium, and executed by a processor”; Para. [0034], “source location 102 can be supplied with power (e.g., electricity), and devices 110, 112, and 114 can draw from the power supplied to source location 102. In some embodiments, source location 102 is a household and the power to the household is supplied from an electric power grid, a local power source (e.g., solar panels), a combination of these, or any other suitable source”, wherein during each time period, including the second time period, the first network element, “an electric power grid, a local power source (e.g., solar panels), a combination of these, or any other suitable source”, supplies electricity, “supplied with power (e.g., electricity)”, to the user locations, source location 102” “household[s]”);
determining, by the one or more processors using the trained model from the second electricity usage data values for the inferencing population, a label (Fig. 8; Para. [0122], “At 806, the processed data can be provided as input data to the trained machine learning model. For example, a model trained according to the functionality of FIG. 7 can be stored, and the processed data can be provided as input to the trained model. At 808, predictions can be generated by the trained machine learning model. For example, disaggregated energy usage for the target device based on the overall energy usage received can be predicted by the trained machine learning model”, where the “predictions . . . generated by the trained machine learning model” use the second electricity usage data, “input data”, and are determined labels for the inferencing population, “disaggregated energy usage for the target device”; see generally Para. [0108], “In some embodiments, the functionality of FIGS. 7-11 can be implemented by software stored in memory or other computer-readable or tangible medium, and executed by a processor”)
for each of the user locations of the inferencing population indicating whether or not an EV uses the user location for charging . . . label for each of the user locations of the inferencing population (Para. [0048], “Prediction model 306 can generate output data 310, such as disaggregated energy usage data for the input data 302. In some embodiments, input data 302 can be source location energy usage data and output data 310 can be disaggregated energy usage data for a target device (or a plurality of devices)”, where the “output data” in “disaggregated” form for a “target device” would indicate whether each user location of the inferencing population, “source location”, is used for running the “target device”; see generally Para. [0032], “Implementations and results demonstrate improved disaggregation predictions for multiple energy consuming devices (e.g., large household appliances and/or electric vehicles)”, where the “target device” can be an “electronic vehicle”, and where each user location in the inferencing population receives labeled results based on the “target device”);
generating by the one or more processors corresponding charging . . . for EVs at user locations of the inferencing population to charge through a corresponding charging apparatus through the first network element . . . (Para. [0034], “source location 102 can be supplied with power (e.g., electricity), and devices 110, 112, and 114 can draw from the power supplied to source location 102. In some embodiments, source location 102 is a household and the power to the household is supplied from an electric power grid, a local power source (e.g., solar panels), a combination of these, or any other suitable source”, where the devices, which includes “electric vehicle[s]”, see Para. [0018], “Embodiments train a machine learning model using labeled energy usage data . . . Energy usage data from multiple source locations (e.g., households) can be obtained, where the energy usage data can be labeled with device specific energy usage. For example, the household energy usage values can cover a period of time, and within that period of time individual device energy usage values (e.g., appliance 1, electric vehicle 1, appliance 2, and the like) can be labeled”, are charged at the user locations, “source location 102 is a household”, of the inferencing population, see Para. [0120], “At 802, household energy usage data can be received over a period of time”, through a corresponding charging apparatus, “power to the household is supplied”, connected to the first network element, “electric power grid, a local power source (e.g., solar panels), a combination of these, or any other suitable source”; see generally Para. [0108], “In some embodiments, the functionality of FIGS. 7-11 can be implemented by software stored in memory or other computer-readable or tangible medium, and executed by a processor”)
[generating predictions] based on the determined labels . . . for the user locations of the inferencing population and . . . the first network element (Para. [0023], “Embodiments utilize a deep learning scheme that can, based on limited training sets, accurately disaggregate power loads of many energy consuming devices, such as large household appliances and electric vehicles”, where “a deep learning scheme” is used to generate, “disaggregate power loads”, which are according to the determined labels for the user locations of the inferencing population, see Fig. 8 and Para. [0122], “At 806, the processed data can be provided as input data to the trained machine learning model. For example, a model trained according to the functionality of FIG. 7 can be stored, and the processed data can be provided as input to the trained model. At 808, predictions can be generated by the trained machine learning model. For example, disaggregated energy usage for the target device based on the overall energy usage received can be predicted by the trained machine learning model”, which, given that the labels are based on charging, are in turn according to the first network element, see Para. [0034], “source location 102 can be supplied with power (e.g., electricity), and devices 110, 112, and 114 can draw from the power supplied to source location 102. In some embodiments, source location 102 is a household and the power to the household is supplied from an electric power grid, a local power source (e.g., solar panels), a combination of these, or any other suitable source” and Para. [0018], “Embodiments train a machine learning model using labeled energy usage data . . . Energy usage data from multiple source locations (e.g., households) can be obtained, where the energy usage data can be labeled with device specific energy usage. For example, the household energy usage values can cover a period of time, and within that period of time individual device energy usage values (e.g., appliance 1, electric vehicle 1, appliance 2, and the like) can be labeled”);
and charging the EVs at user locations of the inferencing population through the corresponding charging apparatus . . . (Para. [0034], “source location 102 can be supplied with power (e.g., electricity), and devices 110, 112, and 114 can draw from the power supplied to source location 102. In some embodiments, source location 102 is a household and the power to the household is supplied from an electric power grid, a local power source (e.g., solar panels), a combination of these, or any other suitable source”, where the devices, which includes “electric vehicle[s]”, see Para. [0018], “Embodiments train a machine learning model using labeled energy usage data . . . Energy usage data from multiple source locations (e.g., households) can be obtained, where the energy usage data can be labeled with device specific energy usage. For example, the household energy usage values can cover a period of time, and within that period of time individual device energy usage values (e.g., appliance 1, electric vehicle 1, appliance 2, and the like) can be labeled”, are charged at the user locations, “source location 102 is a household”, of the inferencing population, see Para. [0120], “At 802, household energy usage data can be received over a period of time”, through a corresponding charging apparatus, “power to the household is supplied”, connected to the “electric power grid, a local power source (e.g., solar panels), a combination of these, or any other suitable source”).
Mimaroglu does not explicitly disclose . . . multi-month . . . a load rating . . . multi-month . . . and a corresponding confidence level value for each of the . . . schedules . . . and corresponding confidence level values . . . the load rating for . . . according to the generated corresponding charging schedule . . . (where the time periods are not specifically described as multi-month and load ratings, confidence level values, and schedules are not explicitly discussed).
However, Hoffmann teaches [a method of electronic vehicle detection in data] (Pg.1, Col. 1, Abstract, “We use machine and deep learning methods to detect EV signatures in hourly smart meter data”) . . . [training data over a] multi-month [time period] (Pg. 1, Col. 2, Para. 4-5, “Our goal is to develop a model that can detect charging EVs . . . We use measurements from 81 households with EV submeters and collected data in the period from January 1st to December 31st in 2017”, where data from multiple months, “January 1st to December 31st in 2017”, is used to “develop a model”)
. . . [inference data over a] (Pg. 2, Col. 1-2, Para. 4-4, “Models were trained and tested
on 90% of the data, while 10% was held out for evaluation . . . After training, models accept a smart meter timeseries as input and return a probability of an EV charging at each sample. To assess classifier performance, we hold out a test set, ask the classifiers to make predictions on this set”, where data used for “evaluation” and “testing” is within the broadest reasonable interpretation of inference data because an output inference is used for “evaluation” or to “assess classifier performance”)
multi-month [time period] (Pg. 1, Col. 2, Para. 5, “To evaluate model performance, we use data from the first five months of 2018” and Pg. 2, Col. 1, Para. 1, “Eidsiva Smart Meter Data (Norway, Unlabelled) This dataset consists of hourly total power usage data from 116'679 customers for between two years and a few months. We have used data from one week in April 2017 and one week in April 2018”, which is used to “execute” the “models” during testing, see Pg. 4, Col. 1, Para. 3, “We now execute our models on unlabelled data”, and where both the evaluation data, “the first five months of 2018”, and testing data, “This dataset . . . for between two years and a few months”, are from a multi-month time period)
. . . and [outputting] a corresponding confidence level value for each of the [samples being labelled as an EV charging location by a model] . . . (Pg. 2, Col. 2, Para. 4, “After training, models accept a smart meter timeseries as input and return a probability of an EV charging at each sample”, where “probability” is a confidence level value; see also Pg. 3, Col. 1, Para. 2, “As classifiers yield probabilities, we must choose a threshold above which the probability is interpreted as a charging EV”)
. . . [generating predictions according to labels] and corresponding confidence level values (Pg. 1, col. 2, Para. 4, “Our goal is to develop a model that can detect charging EVs in the Norwegian grid”, where charge detection is binary labeling; see also Pg. 3, Col. 1, Para. 2, “As classifiers yield probabilities, we must choose a threshold above which the probability is interpreted as a charging EV. Intuitively, a higher threshold should only keep predictions in which the classifier is very confident in (high precision), but this comes at the price of potentially missing more EVs (lower recall). By varying the threshold, precision and recall can be traded-off. We can generate precision-recall curves (as well as calculate the average precision) by evaluating these metrics over a range of thresholds”).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the receiving of first electricity usage data values for a training population of user locations over a first time period, receiving second electricity usage data for an inferencing population of user locations over a second time period, using a trained model to determine a label for each of the user locations of the inferencing population indicating whether an EV used the location for charging, and generating prediction based on the determined labels for the user locations of the inferencing population and the first network element of Mimaroglu with the a method of electronic vehicle detection in data, comprising training data over a multi-month time period, inference data over a multi-month time period, outputting a corresponding confidence value for each of the samples being labelled as an EV charging location by a model, and generating predictions according to the labels and the corresponding confidence level values of Hoffmann in order to increase the versatility of the model by training its parameters and evaluating its inferences using multi-month data that represents seasonal and yearly differences in electricity usage (Hoffmann, Pg. 4, Col. 2, Para. 3, “Firstly, there are differences in electricity usage habits (Norway uses electric heating the winter, the US uses AC units in the summer)”; Hoffmann, Pg. 4, Col. 2, Para. 3, “This dataset consists of hourly total power usage data from 116'679 customers for between two years and a few months”, where a larger time period will mitigate errors caused by using data unrepresentative of other years) and to include confidence scores associated with model labeling outputs, which will allow operators to determine their ideal tradeoff level between model precision and recall (Hoffmann, Pg. 3, Col. 1, Para. 2, “As classifiers yield probabilities, we must choose a threshold above which the probability is interpreted as a charging EV. Intuitively, a higher threshold should only keep predictions in which the classifier is very confident in (high precision), but this comes at the price of potentially missing more EVs (lower recall). By varying the threshold, precision and recall can be traded-off. We can generate precision-recall curves (as well as calculate the average precision) by evaluating these metrics over a range of thresholds”) for binary labeling of charging events (Hoffmann, Pg. 1, col. 2, Para. 4, “Our goal is to develop a model that can detect charging EVs in the Norwegian grid”), which in turn results in high performance (Hoffman, Pg. 5, Col. 1, Para. 2, “Models tuned to maximize either precision (be certain about detections) or recall (find as many EVs as possible) can achieve false positive rates of 10% and successfully locate 90% of charging events”).
Additionally, Sahani teaches . . . [an electric vehicle charging scheduling method, comprising] . . . (Pg. 1, Col. 1, Abstract, “This paper formulates a real-time electric vehicle charging scheduling problem as an mixed-integer linear program (MILP). The problem is to be solved by an aggregator, that provides charging service in a residential community”)
[receiving] a load rating [for a first network element of an electrical distributed network] . . . (Pg. 3, Col. 1, Para. 1, “Power System Constraints[:] Consider a single phase radial distribution system with N nodes . . . Oftentimes there are maximum apparent power limits si based on feeder rating, transformer rating or contracted capacity”, where, “si” is a load rating, “maximum apparent power limits”, for a first network element, “node”, of an electrical distributed network, “radial distribution system” of a “Power System”, which must be received to be “known” and used as variable, see Pg. 3, Col. 1, Para. 1, “However, with known reactive power, the quadratic constraints
PNG
media_image1.png
38
174
media_image1.png
Greyscale
can be written as linear constraints
PNG
media_image2.png
54
234
media_image2.png
Greyscale
”, to ensure “power distribution system constraints for the entire feeder are satisfied”, see Pg. 2, Col. 1, Para. 1, “the aggregator is supposed to control the total real-time load of the station such that the power distribution system constraints for the entire feeder are satisfied”; see generally Pg. 3, Col. 1, Section “Power System Constraints”)
[generating charging] schedules [for EVs] . . . [based on] the load rating for [the first network element] . . . (Pg. 1, Col. 2, Para. 2, “This work proposes a novel computationally tractable algorithm for real-time PEV scheduling . . . Charging schedules are generated and revised periodically . . . The major contributions of the proposed formulation include: Power system operational constraints are met for all operational periods”, where “Charging schedules are generated” for plug-in electric vehicles, “PEV[s]”, based on the load rating for the first network element, “Power system operational constraints are met for all operational periods”, where, as discussed above, “Power system operational constraints” are in turn generated from the load rating, “si”, of the first network element, “node”, see Pg. 3, Col. 1, Para. 1, “Power System Constraints[:] Consider a single phase radial distribution system with N nodes . . . Oftentimes there are maximum apparent power limits si based on feeder rating, transformer rating or contracted capacity . . . with known reactive power, the quadratic constraints
PNG
media_image1.png
38
174
media_image1.png
Greyscale
can be written as linear constraints
PNG
media_image2.png
54
234
media_image2.png
Greyscale
”; see also Pg. 4-5, Col. 2-1, Para. 4-3, “The PEV charging station was assumed to be located at node 5 of the feeder . . . the variation of active power load of the EV charging station in response to price and power system load variations is shown . . . wherein the EV charging load is normalized with the transformer rating at node 5”; see also Pg. 5, Col. 1, Para. 4, “The developed formulation gives optimal charging scheduling of incoming PEVs, considers system constraints and also maximizes the number of PEVs charged at a real-time scenario . . . ”)
[and charging the EVs] . . . according to the generated corresponding charging schedule . . . (Pg. 2, Col. 1, Para. 2, “The main novelty of the proposed algorithm is that if a PEV arrives at the charging station and a contract is established, the scheduling process guarantees that the charging commitments are met despite of uncertain future schedule revisions”; see also Pg. 5, Col. 1, Fig. 3 and Pg. 5, Col. 1, Para. 2, “The novelty of the algorithm is the guarantee the aggregator provides in terms of charging time. The charging power distribution for a subset of PEVs over 10 time steps is shown in the Fig. 3. The shaded area for each PEV depicts its availability for charging and the percentage of charging completed in every time step. All PEVs are shown to receive 100% of their charging requirement”; see generally Pg. 1, Col. 2, Para. 2, “This work proposes a novel computationally tractable algorithm for real-time PEV scheduling based on a moving time horizon setting. Charging schedules are generated and revised periodically based on the actual number of PEVs arriving in real-time”).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the machine learning method for electric vehicle charging, comprising receiving information for a first network element of an electrical distribution network, generating predictions based on the determined labels and corresponding confidence level values for the user locations of the inferencing population and the first network element, and charging the EVs at user locations of the inferencing population through a corresponding charging apparatus through the first network element of Mimaroglu in view of Hoffmann with the electric vehicle charging scheduling method, comprising receiving a load rating for a first network element of an electrical distributed network, generating charging schedules for EVs based on the load rating for the first network element, and charging the EVs according to the generated corresponding charging schedule of Sahani in order to utilize the machine learning model for load monitoring of electric vehicles (Mimaroglu, Para. [0023], “Embodiments utilize a deep learning scheme that can, based on limited training sets, accurately disaggregate power loads of many energy consuming devices, such as large household appliances and electric vehicles”; see also Mimaroglu, Abstract, “Embodiments implement non-intrusive load monitoring using ensemble machine learning techniques”) with an ideal tradeoff level between model precision and recall (Hoffmann, Pg. 3, Col. 1, Para. 2, “As classifiers yield probabilities, we must choose a threshold above which the probability is interpreted as a charging EV. Intuitively, a higher threshold should only keep predictions in which the classifier is very confident in (high precision), but this comes at the price of potentially missing more EVs (lower recall). By varying the threshold, precision and recall can be traded-off. We can generate precision-recall curves (as well as calculate the average precision) by evaluating these metrics over a range of thresholds”) to develop optimal charging scheduling of electric vehicles (Sahani, Pg. 5, Col. 1, Para. 4, “The developed formulation gives optimal charging scheduling of incoming PEVs, considers system constraints and also maximizes the number of PEVs charged at a real-time scenario”; Mimaroglu, Para. [0048], “In some embodiments, prediction module 306 can be a machine learning module (e.g., neural network) that is trained by training data 308. For example, training data 308 can include labeled data, such as energy usage data values from a plurality of source locations (e.g., source locations 102 and 106 from FIG. 1) that include labeled device specific energy usage data values”), which will benefit both suppliers (Sahani, Pg. 1, Col. 1, Abstract, “The proposed formulation maximizes the profit of the aggregator, enhancing the utilization of available infrastructure”; Sahani, Col. 1, Para. 2, “PEV load demand increase calls for infrastructure addition on generation, transmission, and distribution systems. The upgradation of these infrastructure is often capital intensive and has a long time-lag. However, a meticulously designed charging approach could minimize the infrastructure-upgradation requirements”; and Sahani, Pg. 2, Col. 2, Para. 2, “So, the primary idea behind scheduling the PEV charging is to spread out the charging start times causing less steep peaking”, where suppliers are benefitted through “maximiz[ing] the profit” and “minimize[ing] the infrastructure-upgradation requirements”) and consumers (Sahani, Pg. 2, Col. 1, Para. 2, “The main novelty of the proposed algorithm is that if a PEV arrives at the charging station and a contract is established, the scheduling process guarantees that the charging commitments are met despite of uncertain future schedule revisions”, where consumers are benefitted through “guarantees that the charging commitments are met”).
Regarding Claim 2, Mimaroglu in view of Hoffmann and Sahani teach the method of claim 1, wherein the first electricity usage data and second electricity usage data is advanced metering infrastructure (AMI) data (Mimaroglu, Para. [0029], “Embodiments can use data (e.g., training and/or input) from any suitable meter (e.g., AMI or others)”, where both first electricity usage data, “training”, and second electricity usage data, “input”, can be “from . . . AMI”).
Regarding Claim 8, Mimaroglu in view of Hoffmann and Sahani teach the method of claim 1, wherein the first multi-month time period and the second multi-month time period both are at least a year (Hoffmann, Pg. 1, Col. 2, Para. 4-5, “Our goal is to develop a model that can detect charging EVs . . . We use measurements from 81 households with EV submeters and collected data in the period from January 1st to December 31st in 2017”, where the first time period is at least a year; Hoffmann, Pg. 2, Col. 1, Para. 1, “Eidsiva Smart Meter Data (Norway, Unlabelled) This dataset consists of hourly total power usage data from 116'679 customers for between two years and a few months”, and where the second time period is at least a year).
The reasons for obviousness were discussed in regard to the rejection of claim 1 above, and remain applicable here.
Regarding Claim 9, Mimaroglu in view of Hoffmann and Sahani teach the method of claim 1, wherein the first electricity usage data values have a temporal resolution of intervals of an hour or less (Mimaroglu, Para. [0068], Table 1; Mimaroglu, Para. [0069], “This example includes a granularity of 15 minutes , but other suitable granularities can similarly be implemented (e.g., 1 min, 5 mins, 15 mins, 30 mins, 1 hour, hours, and the like). In some embodiments, processing the energy usage data (e.g., to generate training data 308 ) can include reducing a granularity of the data, for example so that it can be used to generate a training corpus with a consistent granularity ( e.g. , 1 hour)”, where any of the example “granularities/resolutions”, other than “hours”, have intervals of an hour or less; see also Mimaroglu, Para. [0024], “training data can be used to train learning models designed to effectively learn in these challenging conditions. Input to the learning models can be provided by AMI along with other types of inputs. Embodiments can accurately predict electric device energy usage in high and low granularities/resolutions (e.g. , at 1 min , 5 mins , 15 mins , 30 mins , 1 hour, or more)”).
Regarding Claim 10, Mimaroglu in view of Hoffmann and Sahani teach the method of claim 1, where the corresponding determined labels indicating whether an EV uses the user location for charging (Mimaroglu, Para. [0018], “Embodiments train a machine learning model using labeled energy usage data . . . Energy usage data from multiple source locations (e.g., households) can be obtained, where the energy usage data can be labeled with device specific energy usage. For example, the household energy usage values can cover a period of time, and within that period of time individual device energy usage values (e.g., appliance 1, electric vehicle 1, appliance 2, and the like) can be labeled”, where the “training . . . data”, which “can be labeled”, is determined and received from “multiple source locations”; see also Mimaroglu, Fig. 7; Mimaroglu, Para. [0068], Table 1; and Mimaroglu, Para. [0109], “the energy usage data can be similar to the data illustrated in Table 1 above. In some embodiments, the received data can include a timestamp, overall energy usage (which includes energy used by a plurality of devices) at a source location (e.g., household), and labeled device specific energy usage for one or multiple of the target and non-target devices”, where labels “EV” with corresponding values other than “0.0” indicates that a EV uses the “source location” for charging)
include a first value, indicating that an EV uses the user location for charging, and a second value, indicating either that an EV does not use the user location for charging or that it is unknown whether an EV uses the user location for charging (Hoffmann, Pg. 1, Col. 2, Para. 5, “Our goal is to develop a model that can detect charging EVs . . . we use two different datasets – (a) the Pecan Street Dataport set of labelled consumption data . . . We use measurements from 81 households with EV submeters and collected data in the period from January 1st to December 31st in 2017. Ten of these households had no EV charging, while the rest had between 20 and 712 such events”, where, in view of Hoffmann, the “labelled consumption data” includes a first value indicating charging at a user location, charging “events”, and a second value indicating absence or unknown status of charging at the user location, “no EV charging”, which is required for the binary “detect[ion of] charging EVs” using a tradeoff of precision and recall during training, discussed above).
The reasons for obviousness were discussed in regard to the rejection of claim 1 above, and remain applicable here.
Regarding Claim 11, Mimaroglu in view of Hoffmann and Sahani teach the method of claim 1, wherein the inferencing population’s user locations include households (Mimaroglu, Para. [0119], “a machine learning model that is trained based on the functionality of FIG. 7 can be used to perform the functionality of FIG. 8”, where “FIG. 7” uses the first electricity usage data for “train[ing]” and “FIG. 8” uses the second electricity usage data for “perfom[ing] the functionality”, both of which, as discussed below, include households as user locations, and where the data used for “perfom[ing] the functionality” is the inferencing population; Mimaroglu, Fig. 8; Mimaroglu, Para. [0120], “At 802, household energy usage data can be received over a period of time”; Mimaroglu, Fig. 7; Mimaroglu, Para. [0109], “At 702, energy usage data including energy usage by a target device and one or more non-target devices at a plurality of source locations can be received. For example . . . a source location (e.g., household)”).
Regarding Claim 14, Mimaroglu in view of Hoffmann and Sahani teach the method of claim 1, further comprising: providing the determined labels for the user locations of the inferencing population and corresponding confidence level values to a utility (Mimaroglu, Para. [0029], “The disaggregated energy usage predictions for a target device can be useful for many reasons: providing energy savings opportunities for utilities and their customers . . . For example, electric utilities can invest in techniques for disaggregating energy usage from large appliances or devices”, where the “energy usage predictions for a target device” are the predicted labels, which must be “provid[ed] . . . [to] utilities”, directly or indirectly, in order to allow for “energy savings opportunities for utilities”, and where, in view of Hoffmann, the predictions include the confidence values, see Hoffmann, Pg. 2, Col. 2, Para. 4, “After training, models accept a smart meter timeseries as input and return a probability of an EV charging at each sample”, where “probability” is a confidence level value; see also Hoffmann, Pg. 3, Col. 1, Para. 2, “As classifiers yield probabilities, we must choose a threshold above which the probability is interpreted as a charging EV”)
supplying the electricity through the first network element of the electrical distribution network (Mimaroglu, Para. [0034], “source location 102 can be supplied with power (e.g., electricity), and devices 110, 112, and 114 can draw from the power supplied to source location 102. In some embodiments, source location 102 is a household and the power to the household is supplied from an electric power grid, a local power source (e.g., solar panels), a combination of these, or any other suitable source”, where the utility supplies electricity, “supplied with power (e.g., electricity)”, through the first network element of the electrical distribution network, “from an electric power grid, a local power source (e.g., solar panels), a combination of these, or any other suitable source”).
The remaining limitations are substantially the same as limitations of Claim 1, therefore it is rejected under the same rationale.
Regarding Claim 22, Mimaroglu teaches a system, comprising: a plurality of charging apparatuses each configured to charge a corresponding electrical vehicle (EV) . . . one or more interfaces configured to: . . . one or more processors connected to the one or more interfaces and configured to: . . . (Para. [0003], “The embodiments of the present disclosure are generally directed to systems and methods for non-intrusive load monitoring using ensemble machine learning techniques”, which includes one or more processors, “Processor”, connected to one or more interfaces, “Liquid Crystal Display”, see Fig. 2 and Para. [0043], “Processor 222 is further coupled via bus 212 to a display 224, such as a Liquid Crystal Display (“LCD”). A keyboard 226 and a cursor control device 228, such as a computer mouse, are further coupled to communication device 212 to enable a user to interface with system 200”; see also Para. [0034], “source location 102 can be supplied with power (e.g., electricity), and devices 110, 112, and 114 can draw from the power supplied to source location 102. In some embodiments, source location 102 is a household and the power to the household is supplied from an electric power grid, a local power source (e.g., solar panels), a combination of these, or any other suitable source”, where the devices, which includes “electric vehicle[s]”, see Para. [0018], “Embodiments train a machine learning model using labeled energy usage data . . . Energy usage data from multiple source locations (e.g., households) can be obtained, where the energy usage data can be labeled with device specific energy usage. For example, the household energy usage values can cover a period of time, and within that period of time individual device energy usage values (e.g., appliance 1, electric vehicle 1, appliance 2, and the like) can be labeled”, are charged at the user locations, “source location 102 is a household”, of the inferencing population, see Para. [0120], “At 802, household energy usage data can be received over a period of time”, through corresponding charging apparatus, “power to the household is supplied”, connected to the “electric power grid, a local power source (e.g., solar panels), a combination of these, or any other suitable source”, such that the system includes the plurality of charging apparatuses).
The remaining limitations are substantially the same as the limitations of claim 1, therefore it is rejected under the same rationale.
Regarding Claim 24, Mimaroglu in view of Hoffmann and Sahani teach the method of claim 14, further comprising: subsequently managing the utility according to the determined labels for the user locations of the inferencing population and corresponding confidence level values (Mimaroglu, Para. [0001], “The embodiments of the present disclosure generally relate to utility metering devices, and more particularly to non-intrusive load monitoring using utility metering devices” and Mimaroglu, Para. [0029], “The disaggregated energy usage predictions for a target device can be useful for many reasons: providing energy savings opportunities for utilities and . . . enabling better grid planning including peak time demand management”, where, managing of the “utility[y]”, such as “load monitoring”, “energy savings”, or “grid planning”, is subsequently performed according to “deep learning” “disaggregated energy usage predictions”, see Mimaroglu, Para. [0023], “Embodiments utilize a deep learning scheme that can, based on limited training sets, accurately disaggregate power loads of many energy consuming devices, such as large household appliances and electric vehicles”, where “a deep learning scheme” is used to generate, “disaggregate power loads”, which are according to the determined labels for the user locations of the inferencing population, see Mimaroglu, Fig. 8 and Mimaroglu, Para. [0122], “At 806, the processed data can be provided as input data to the trained machine learning model. For example, a model trained according to the functionality of FIG. 7 can be stored, and the processed data can be provided as input to the trained model. At 808, predictions can be generated by the trained machine learning model. For example, disaggregated energy usage for the target device based on the overall energy usage received can be predicted by the trained machine learning model”, and in view of Hoffmann, the corresponding confidence values, see Hoffmann, Pg. 2, Col. 2, Para. 4, “After training, models accept a smart meter timeseries as input and return a probability of an EV charging at each sample”, where “probability” is a confidence level value; see also Hoffmann, Pg. 3, Col. 1, Para. 2, “As classifiers yield probabilities, we must choose a threshold above which the probability is interpreted as a charging EV”).
The remaining limitations are substantially the same as the limitations of claim 1, therefore it is rejected under the same rationale.
Regarding Claim 25, the claim limitations are substantially the same as the limitations of Claim 24, which is dependent on claim 14. Therefore, it is rejected under the same rationale.
Regarding Claim 26, the claim limitations are substantially the same as the limitations of Claim 24, which is dependent on claims 1 and 14. Therefore, it is rejected under the same rationale.
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Mimaroglu in view of Hoffmann, Sahani, and Shyr et al. (hereinafter Shyr) (Pat. Pub. No. US 2018/0176033 A1).
Regarding Claim 3, Mimaroglu in view of Hoffmann and Sahani teach the method of claim 1, wherein one or more of the corresponding labels for the training population’s user locations is received (Mimaroglu, Para. [0018], “Embodiments train a machine learning model using labeled energy usage data . . . Energy usage data from multiple source locations (e.g., households) can be obtained, where the energy usage data can be labeled with device specific energy usage. For example, the household energy usage values can cover a period of time, and within that period of time individual device energy usage values (e.g., appliance 1, electric vehicle 1, appliance 2, and the like) can be labeled”, where the “training . . . data”, which “can be labeled”, is received from “multiple source locations”; see also Mimaroglu, Fig. 7; Mimaroglu, Para. [0068], Table 1; and Mimaroglu, Para. [0109], “the energy usage data can be similar to the data illustrated in Table 1 above. In some embodiments, the received data can include a timestamp, overall energy usage (which includes energy used by a plurality of devices) at a source location (e.g., household), and labeled device specific energy usage for one or multiple of the target and non-target devices”, where labels “EV” with corresponding values other than “0.0” indicates that a EV uses the “source location” for charging)
Mimaroglu in view of Hoffmann and Sahani do not explicitly disclose . . . from a utility.
However, Shyr teaches . . . [receive training data and labels] from a utility (Fig. 24; Para. [0108] - [0109], “The final per-sample disaggregation may be further be used as input labels to large-scale supervised techniques . . . Consumption data is received at 2405. Note that while other data may be used in both disaggregation and itemization (such as but not limited to training data, weather data, sunrise/sunset data, home data, neighborhood/area demographics, tax records, etc.), FIG. 24 shows the treatment of consumption data, as may be received from an AMI device or from a utility”).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the receiving of labels for the training population’s user locations of Mimaroglu in view of Hoffmann and Sahani with the receive training data and labels from a utility of Shyr in order to utilize data that is already collected in a suitable format (Shyr, Para. [0005], “utility companies collect data usage, this is typically performed for validation of billing cycles, and is generally collected at a fifteen (15) minute or one hour interval”, which is consistent with the embodiments of Mimaroglu, see Mimaroglu, Para. [0029], “Embodiments can use data (e.g., training and/or input) from any suitable meter (e.g., AMI or others)”) by a the utility that would be motivated to supply the data for benefit of positively impacting its public sentiment (Shyr, Para. [0003], “Accurate disaggregation may enable personalized and actionable insights to be presented to a customer, which may positively influence customer engagement as well as sentiment towards the energy-providing utility”).
Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over Mimaroglu in view of Hoffmann, Sahani, and Shahriar et al. (hereinafter Shahriar) (“Prediction of EV Charging Behavior Using Machine Learning”).
Regarding Claim 4, Mimaroglu in view of Hoffmann and Sahani teach the method of claim 1, wherein one or more of the corresponding labels for the training population’s user locations (Mimaroglu, Para. [0018], “Embodiments train a machine learning model using labeled energy usage data . . . Energy usage data from multiple source locations (e.g., households) can be obtained, where the energy usage data can be labeled with device specific energy usage. For example, the household energy usage values can cover a period of time, and within that period of time individual device energy usage values (e.g., appliance 1, electric vehicle 1, appliance 2, and the like) can be labeled”, where the “training . . . data”, which “can be labeled”, is received from “multiple source locations”) . . . .
Mimaroglu in view of Hoffmann and Sahani do not explicitly disclose . . . are electric vehicle supply equipment (EVSE) data.
However, Shahriar teaches . . . [prediction of EV charging behavior using machine learning] (Pg. 111576, Col. 1, Abstract, “in this paper we propose the usage of historical charging data in conjunction with weather, traffic, and events data to predict EV session duration and energy consumption using popular machine learning algorithms”)
[, where training inputs] are electric vehicle supply equipment (EVSE) data (Pg. 111580, Col. 1, Para. 1, “The ACN [21] dataset is among the few publicly available datasets for non-residential EV charging and will be utilized in this work. The dataset contains charging records from two stations”, where “non-residential EV charging . . . stations” are within the broadest reasonable interpretation of EVSE data).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the use of training data labels corresponding to private household electric vehicle data to train a model of Mimaroglu in view of Hoffmann and Sahani with the use of electric vehicle supply equipment inputs to train a model of Shahriar in order to supplement the dataset with EVSE data and labels in order to develop solutions to electric vehicle utilization challenges, such as strain on power grid infrastructure (Shahriar, Pg. 111576, Col. 1, Abstract, “One of the key challenges, however, is the strain on power grid infrastructure that comes with large-scale EV deployment. The solution to this lies in utilization of smart scheduling algorithms to manage the growing public charging demand”), which is more significant for heavily trafficked public electric vehicle supply equipment than private households or other circumstances (Shahriar, Pg. 111580, Col. 1, Para. 1, “Scheduling of EV charging is more significant in public charging structures due to the unpredictable nature of the charging behavior, especially in places like shopping malls”).
Claims 5, 18, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Mimaroglu in view of Hoffmann, Sahani, and Shafee et al. (hereinafter Shafee) (“Detection of Lying Electrical Vehicles in Charging Coordination Using Deep Learning”).
Regarding Claim 5, Mimaroglu in view of Hoffmann and Sahani teach the method of claim 1, wherein receiving the corresponding label for the training population’s user locations comprises: receiving . . . data for the training population’s user locations; and determining the corresponding labels from the data (Mimaroglu, Para. [0018], “Embodiments train a machine learning model using labeled energy usage data . . . Energy usage data from multiple source locations (e.g., households) can be obtained, where the energy usage data can be labeled with device specific energy usage. For example, the household energy usage values can cover a period of time, and within that period of time individual device energy usage values (e.g., appliance 1, electric vehicle 1, appliance 2, and the like) can be labeled”, where the “training . . . data”, which “can be labeled”, where the labels must be determined to be used, and is received from “multiple source locations”).
Mimaroglu in view of Hoffmann and Sahani do not explicitly disclose . . . telematics . . . telematics . . . .
However, Shafee teaches . . . [machine learning model training using] telematics . . . telematics (Pg. 179400, Abstract, “an anomaly-based detector based on a deep neural network . . . To train the detector, we first create an honest dataset for the charging coordination application using real driving traces and information provided by an electric vehicle manufacturer”, where “real driving traces” are within the broadest reasonable interpretation of telematics data; see generally Pg. 179406, Col. 1, Para. 6, “To compute the SoC [state of charge] every minute for each EV, we randomly initialized the SoC value and then, for every minute, checked whether the EV was moving”, where telematics data is used to determine battery information) . . . .
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the receiving of data for a training population’s user locations and determining labels from the data for use in training a machine learning model to analyze electric vehicle data of Mimaroglu in view of Hoffmann and Sahani with the use of telematics data in training a machine learning model to analyze electric vehicle data in order to supplement data so the model can incorporate electric vehicle battery information during electricity usage detection training and inference, which will enhance load predictions for energy management (Shafee, Pg. 179401, Col. 1, Para. 3, “The idea is that each EV should report its battery SoC periodically, e.g., every 30 minutes, and the reported data can then be used both for load prediction (for energy management)”).
Regarding Claim 18, Mimaroglu in view of Hoffmann, Sahani, and Shafee teach the method of claim 1, further comprising: receiving telematics data for EVs that charge at the one or more user locations of the inferencing population (Mimaroglu, Fig. 8, where the second electricity usage data values for an inferencing population is the “INPUT DATA”, “RECEIV[ED]” at “802”; Mimaroglu, Para. [0120], “At 802, household energy usage data can be received over a period of time”, where the “energy usage data” is over a “period of time” for a “household” location; see generally Mimaroglu, Para. [0029], “Motivations for these investments include the advancement of AMI and smart grid technologies, a growing interest in energy efficiency, interest from customers for better information, and the like”, where the intended use is for a plurality of customers, and therefore, a plurality of households, where, in view of Shafee, the inferencing data is supplemented with telematics data, see Shafee, Pg. 179400, Abstract, “an anomaly-based detector based on a deep neural network . . . To train the detector, we first create an honest dataset for the charging coordination application using real driving traces and information provided by an electric vehicle manufacturer”, where “real driving traces” are within the broadest reasonable interpretation of telematics data, which is also used for inference, see Shafee, Pg. 179407, Col. 2, Para. 2, “As discussed in Section V-A, an honest dataset was created with 12; 864 data samples . . . this dataset is divided into three parts: a training dataset for training the detector, a validation dataset for avoiding overfitting during the training process, and a test dataset for evaluating the model”);
and for the one or more user locations of the inferencing population for which telematics data is received, disaggregating EV electricity usage from total electricity usage (Mimaroglu, Para. [0070], “For example, the training of prediction module 306 can be configured to generate disaggregation predictions for a target device”, where, the trained model, “306”, “disaggregate[es]” inference “predictions”, and as discussed above, the inference data includes telematics data and the “target device” can be an “electric vehicle”, see Mimaroglu, Para. [0032], “Implementations and results demonstrate improved disaggregation predictions for multiple energy consuming devices (e.g., large household appliances and/or electric vehicles)”; and where the inferences are for one or more user locations of the inferencing population, see Mimaroglu, Fig. 8, where the second electricity usage data values for an inferencing population is the “INPUT DATA”, “RECEIV[ED]” at “802”; Mimaroglu, Para. [0120], “At 802, household energy usage data can be received over a period of time”, where the “energy usage data” is over a “period of time” for a “household” location; see generally Mimaroglu, Para. [0029], “Motivations for these investments include the advancement of AMI and smart grid technologies, a growing interest in energy efficiency, interest from customers for better information, and the like”, where the intended use is for a plurality of customers, and therefore, a plurality of households).
The reasons for obviousness were discussed in regard to the rejection of claim 5 above, and remain applicable here.
Regarding Claim 20, Mimaroglu in view of Hoffmann, Sahani, and Shafee teach the method of claim 1, further comprising: receiving telematics data for EVs that charge at one or more user locations of the training population, wherein the machine learning model is further trained using the telematics data (Shafee, Pg. 179400, Abstract, “an anomaly-based detector based on a deep neural network . . . To train the detector, we first create an honest dataset for the charging coordination application using real driving traces and information provided by an electric vehicle manufacturer”, where “real driving traces” are within the broadest reasonable interpretation of telematics data and therefore the “train[ing]” “energy usage data” of Mimaroglu, see Mimaroglu, Para. [0018], “Embodiments train a machine learning model using labeled energy usage data, is telematics data in the sense that it is supplemented to include telematics data, which is data received for EVs that charge at one or more user locations of the training population, see Mimaroglu, Fig. 3; Mimaroglu, Para. [0048], “In some embodiments, prediction module 306 can be a machine learning module (e.g., neural network) that is trained by training data 308. For example, training data 308 can include labeled data, such as energy usage data values from a plurality of source locations (e.g., source locations 102 and 106 from FIG. 1)”, where the “plurality of source locations” are a plurality of user locations and, in the context of “electric vehicles” and “electric grid planning”, the “energy usage” is “electric”, see Mimaroglu, Para. [0022]- [0023], “Accurate disaggregation via NILM provides many benefits including energy savings opportunities, personalization, improved electric grid planning, and more . . . of many energy consuming devices, such as large household appliances and electric vehicles”); and
disaggregating by the trained machine learning model of the EV electricity usage from total electricity usage for one or more user locations of the inferencing population (Mimaroglu, Para. [0070], “For example, the training of prediction module 306 can be configured to generate disaggregation predictions for a target device”, where, the trained model, “306”, “disaggregate[es]” inference “predictions”, and as discussed above, the “target device” can be an “electric vehicle”, see Mimaroglu, Para. [0032], “Implementations and results demonstrate improved disaggregation predictions for multiple energy consuming devices (e.g., large household appliances and/or electric vehicles)”; and where the inferences are for one or more user locations of the inferencing population, see Mimaroglu, Fig. 8, where the second electricity usage data values for an inferencing population is the “INPUT DATA”, “RECEIV[ED]” at “802”; Mimaroglu, Para. [0120], “At 802, household energy usage data can be received over a period of time”, where the “energy usage data” is over a “period of time” for a “household” location; see generally Mimaroglu, Para. [0029], “Motivations for these investments include the advancement of AMI and smart grid technologies, a growing interest in energy efficiency, interest from customers for better information, and the like”, where the intended use is for a plurality of customers, and therefore, a plurality of households).
The reasons for obviousness were discussed in regard to the rejection of claim 5 above, and remain applicable here.
Claim 6-7 are rejected under 35 U.S.C. 103 as being unpatentable over Mimaroglu in view of Hoffmann, Sahani, and Davis (“Evidence of a Homeowner-Renter Gap for Electric Vehicles”).
Regarding Claim 6, Mimaroglu in view of Hoffmann and Sahani teach the method of claim 1, wherein a number of user locations having a label indicating that an EV uses the user location for charging is . . . [sparse in] the training population of user locations (Mimaroglu, Para. [0068], “An example of energy usage data that can be processed to generate training data 308 includes: [Table 1]”, where the entries in only rows 2-5, which is within the broadest reasonable interpretation of sparse in a table of 11 rows, include a label “EV”, which in combination with a value other than “0.0”, indicate that an EV uses the user location for charging).
Mimaroglu in view of Hoffmann and Sahani do not explicitly disclose . . . less than 1% of . . . .
However, Davis teaches . . . [EV data, wherein a number of user locations having a label indicating that an EV uses the user location for charging is] less than 1% of . . . [the data] (Pg. 1-2, Para. 3-1, “I use newly-available nationally representative data from the U.S. Department of Transportation’s National Household Travel Survey . . . Nationwide, homeowners are more than three times more likely than renters to . . . own an electric vehicle. In particular, 0.87% (less than 1%) of homeowners own an
electric vehicle, compared to 0.25% (one-quarter of 1%) of renters”, where “nationally representative data” on user locations, “home[s]”, shows that “less than 1% . . . own an electric vehicle”, which is a label indicating that an EV uses the user location for charging, see Pg. 2, Para. 4, “For homeowners, it is relatively straightforward to invest in a 240 volt outlet, electric panel upgrades, and other improvements to speed up charging”).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the machine learning method, wherein, the number of user locations labeled as used by an EV for charging is sparse in the training population of user locations of Mimaroglu in view of Hoffmann and Sahani with the EV data, wherein a number of user locations having a label indicating that an EV uses the user location for charging is less than 1% of the data in order to generate training data for the machine learning model using nationally representative data (Mimaroglu, Para. [0018], “In some embodiments, this household and device specific energy usage can then be processed to generate training data for the machine learning model”; Davis, Pg. 1, Para. 1, “I use newly-available nationally representative data from the U.S. Department of Transportation’s National Household Travel Survey), which will allow for improved accuracy on models used nationally (Mimaroglu, Para. [0019], “the machine learning model can be trained . . . implementing embodiments of the training techniques (e.g., prediction generation, loss calculation, gradient propagation, accuracy improvements, and the like) for the machine learning model”) and to allow for generation of training data for specific model use cases (compare Mimaroglu, Para. [0018], “In some embodiments, this household and device specific energy usage can then be processed to generate training data for the machine learning model” and Davis, Pg. 2, Para. 1, “Nationwide, homeowners are more than three times more likely than renters to . . . own an electric vehicle. In particular, 0.87% (less than 1%) of homeowners own an electric vehicle, compared to 0.25% (one-quarter of 1%) of renters”, where specific “household” data can be selected for from the population with “less than 1% . . . of homeowners own[ing] an electric vehicle” to “generate training data” with an increased representation of positively labelled data, with Mimaroglu, Para. [0068], “An example of energy usage data that can be processed to generate training data 308 includes: [Table 1]”, where, without generation of artificial data, the training data has a floor of representation of positively labelled data that is above national trends; see also Mimaroglu, Para. [0019], “the training data can include target device specific energy usage at a number of different source locations (e.g., households), and thus the machine learning model can be trained to identify trends in the training data and predict target device energy usage”).
Regarding Claim 7, Mimaroglu in view of Hoffmann, Sahani, and Davis teach the method of claim 6,
wherein the first subset of the training population is equal within a factor of two with the number of user locations having a label indicating that an EV uses the user location for charging (Mimaroglu, Para. [0068], Table 1, where, as discussed above, the first subset of the training population can be a subset of rows where “EV” is “0.0”, such as rows 6-9, whereas the entries in only rows 2-5, which is within the broadest reasonable interpretation of sparse in a table of 11 rows, include a label “EV”, which in combination with a value other than “0.0”, indicate that an EV uses the user location for charging, thus, the two values are equal, which is equal within a factor of 2; see also Mimaroglu, Para. [0077], “processing device specific labeled energy usage data for a source location can include replacing null values (or any other place holder value) with zero values. For example, when it is determined that a particular device is to be used in the training techniques for a given implementation of prediction module 306 and that portions of the energy usage data lack labeled device specific energy usage for the particular device (at one or more source locations), the labeled energy usage values for the particular device that are missing can be replaced with zero values”, where, as discussed above, “data for a source location” for “training” can be selected as a subset of the training population where the “location” does not have a corresponding label, such as “null” for a “device”, which, depending on the number of “null values”, may be of a similar size as the user locations with labels indicating EV charging; see also Mimaroglu, Para. [0018], “Embodiments train a machine learning model using labeled energy usage data . . . Energy usage data from multiple source locations (e.g., households) can be obtained, where the energy usage data can be labeled with device specific energy usage. For example, the household energy usage values can cover a period of time, and within that period of time individual device energy usage values (e.g., appliance 1, electric vehicle 1, appliance 2, and the like) can be labeled”, where, as discussed above the training data contains a number of user locations have a label indicating an EV uses the location for charging; see also Mimaroglu, Para. [0068], Table 1; and Mimaroglu, Para. [0109], “the energy usage data can be similar to the data illustrated in Table 1 above. In some embodiments, the received data can include a timestamp, overall energy usage (which includes energy used by a plurality of devices) at a source location (e.g., household), and labeled device specific energy usage for one or multiple of the target and non-target devices”, where labels “EV” with corresponding values other than “0.0” indicates that a EV uses the “source location” for charging).
Claim 12 is rejected under 35 U.S.C. 103 as being unpatentable over Mimaroglu in view of Hoffmann, Sahani, and C3.ai contributors (hereinafter C3.ai) (“Gradient-Boosted Decision Trees (GBDT)”).
Regarding Claim 12, Mimaroglu in view of Hoffmann and Sahani teach the method of claim 1, wherein the machine learning model . . . (Mimaroglu, Para. [0048], “In some embodiments, prediction module 306 can be a machine learning module (e.g., neural network) that is trained by training data 308”).
Mimaroglu in view of Hoffmann and Sahani do not explicitly disclose . . . is a gradient boosting type model.
However, C3.ai teaches [a machine learning model that] . . . is a gradient boosting type model (Pg. 1, Para. 1, “Gradient-boosted decision trees are a machine learning technique for optimizing the predictive value of a model through successive steps in the learning process”).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the machine learning model of Mimaroglu in view of Hoffmann and Sahani with the machine learning model that is a gradient boosted type model of C3.ai in order to utilize a machine learning model with proven accuracy and efficiency (C3.ai, Pg. 2, Para. 1, “Gradient-boosted models have proven themselves time and again in various competitions grading on both accuracy and efficiency”).
Claim 13 is rejected under 35 U.S.C. 103 as being unpatentable over Mimaroglu in view of Hoffmann, Sahani, and Rashid et al. (hereinafter Rashid) (“Revisiting Selection of Residential Consumers for Demand Response Programs”).
Regarding Claim 13, Mimaroglu in view of Hoffmann and Sahani teach the method of claim 1, wherein the time series data features include hour-of-day . . . (Mimaroglu, Para. [0068], “An example of energy usage data that can be processed to generate training data 308 includes: [Table 1]”, where the time series data features include the hour-of day, such as “Time” of “2019-06-01 00:00:00”).
Mimaroglu in view of Hoffmann and Sahani do not explicitly disclose . . . statistics.
However, Rashid teaches . . . [a method comprising, computing hour-of-day] statistics [based on hour-of-day information] (Pg. 2, Fig. 1, “Energy consumption patterns of four consumers on six consecutive days. Each wiggly line represents the hourly energy consumption of a consumer on a different day”, which is used to generate statistics like “average consumption”, see Pg. 3, Fig. 2, “Energy consumption patterns of 4 consumers on consecutive days. Red line represents the average consumption. Consistency score is above the graph”; for additional information see Pg. 3, Section “4 METHODOLOGY”).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the time series data features including hour-of-day information of Mimaroglu in view of Hoffmann and Sahani with the computing hour-of-day statistics based on hour-of-day information of Rashid in order to modify the extracted features to include hour-of-day statistics that provide information on the likelihood of customer consumption at a particular day and time (Rashid, Pg. 1, Col. 2, Para. 3, “Quantifying consistency in consumption pattern is important as it indicates the likelihood of a consumer following the historical pattern on DR day and time”), which increases predictability of energy consumption inferences (Rashid. Pg. 1, Col. 1, Abstract, “We demonstrate that measuring consistency quantitatively helps to understand predictability of consumer’s energy consumption”).
Claim 15 is rejected under 35 U.S.C. 103 as being unpatentable over Mimaroglu in view of Hoffmann, Sahani, and Desmond et al. (hereinafter Desmond) (Pat. Pub. No. US 2021/0174196 A1).
Regarding Claim 15, Mimaroglu in view of Hoffmann and Sahani teach the method of claim 1 . . . user locations of the training population . . . locations (Mimaroglu, Fig. 3; Mimaroglu, Para. [0048], “In some embodiments, prediction module 306 can be a machine learning module (e.g., neural network) that is trained by training data 308. For example, training data 308 can include labeled data, such as energy usage data values from a plurality of source locations (e.g., source locations 102 and 106 from FIG. 1)”, where the “plurality of source locations” are a plurality of user locations and, in the context of “electric vehicles” and “electric grid planning”, the “energy usage” is “electric”, see Mimaroglu, Para. [0022]- [0023], “Accurate disaggregation via NILM provides many benefits including energy savings opportunities, personalization, improved electric grid planning, and more . . . of many energy consuming devices, such as large household appliances and electric vehicles”).
Mimaroglu in view of Hoffmann and Sahani do not explicitly disclose . . . prior to training the machine learning model, making a determination of . . . for which the corresponding label is inaccurate; and removing from the training population . . . for which the corresponding label is determined inaccurate.
However, Desmond teaches . . . prior to training the machine learning model, making a determination of . . . [data] for which the corresponding label is inaccurate; and removing from the training population . . . [data] for which the corresponding label is determined inaccurate (Fig. 1; Para. [0058], “In some embodiments, a user 121 can examine the identified data inputs and modify the ground truth data by, for example, creating new classifications (i.e., relabeling some or all of the identified data inputs with new labels) or opting to remove identified data from use in training the model. According to some embodiments, the processing system 100 can automatically remove the data inputs associated with the ambiguous class structure(s) and retrain the model”, where, prior to the training of the model that occurs during “retrain[ing]”, “identified data inputs” in need of “new labels”, because of inaccuracy due to “ambigu[ity]” or “mislabel[ing]”, can be “remov[ed] . . . from use in training the model”; see generally Para. [0002], “round truth quality describes the overall utility and coherence of the training data and quality measures can include label noise (i.e., mislabeled data inputs), ambiguous class structures (i.e., data inputs that span multiple classes) and outliers (i.e., rare or unusual data inputs)”).
Before the effective filing date of the invention, it would have been obvious to one of ordinary skill in the art to combine the training populations with user locations associated with labels of Mimaroglu in view of Hoffmann and Sahani with the identifying of data with inaccurate labels and removing the data corresponding to the inaccurate labels, prior to training a machine learning model of Desmond in order to increase the ground truth quality of training data (Desmond, Para. [0001] – [0002], “The present invention generally relates to . . . improv[ing] ground truth quality for generating accurate machine learning models . . . Ground truth quality describes the overall utility and coherence of the training data and quality measures can include label noise (i.e., mislabeled data inputs), ambiguous class structures (i.e., data inputs that span multiple classes) and outliers (i.e., rare or unusual data inputs)”), which increases accuracy and reduces development time of models (Desmond, Para. [0003], “Advantages can include the development of more accurate machine learning models while reducing the development time of such models”).
Claims 19 and 21 are rejected under 35 U.S.C. 103 as being unpatentable over Mimaroglu in view of Hoffmann, Sahani, Shafee, and Shahriar.
Regarding Claim 19, Mimaroglu in view of Hoffmann, Sahani, Shafee, and Shahriar teach the method of claim 18, wherein the telematics data (Shafee, Pg. 179400, Abstract, “an anomaly-based detector based on a deep neural network . . . To train the detector, we first create an honest dataset for the charging coordination application using real driving traces and information provided by an electric vehicle manufacturer”, where “real driving traces” are within the broadest reasonable interpretation of telematics data, which is also used for inference, see Shafee, Pg. 179407, Col. 2, Para. 2, “As discussed in
Section V-A, an honest dataset was created with 12; 864 data samples . . . this dataset is divided into three parts: a training dataset for training the detector, a validation dataset for avoiding overfitting
during the training process, and a test dataset for evaluating the model”, and therefore, the inference “energy usage data” of Mimaroglu, see Mimaroglu, Para. [0120], “At 802, household energy usage data can be received over a period of time”, where the “energy usage data” is over a “period of time” for a “household” location, is telematics data in the sense that it is supplemented to include telematics data)
includes electric vehicle supply equipment (EVSE) data (Shahriar, Pg. 111580, Col. 1, Para. 1, “The ACN [21] dataset is among the few publicly available datasets for non-residential EV charging and will be utilized in this work. The dataset contains charging records from two stations”, where “non-residential EV charging . . . stations” are within the broadest reasonable interpretation of EVSE data, and, in view of Shahriar, the inference data that includes telematics data, the telematics data, also includes EVSE data).
The reasons for obviousness were discussed in regard to the rejection of Claim 4, for the combination with Shahriar, and in regard to the rejection of Claim 5, for the combination with Shafee, and remain applicable here.
Regarding Claim 21, Mimaroglu in view of Hoffmann, Sahani, Shafee, and Shahriar teach the method of claim 20, wherein the telematics data (Shafee, Pg. 179400, Abstract, “an anomaly-based detector based on a deep neural network . . . To train the detector, we first create an honest dataset for the charging coordination application using real driving traces and information provided by an electric vehicle manufacturer”, where “real driving traces” are within the broadest reasonable interpretation of telematics data and therefore the “train[ing]” “energy usage data” of Mimaroglu, see Mimaroglu, Para. [0018], “Embodiments train a machine learning model using labeled energy usage data, is telematics data in the sense that it is supplemented to include telematics data)
includes electric vehicle supply equipment (EVSE) data (Shahriar, Pg. 111580, Col. 1, Para. 1, “The ACN [21] dataset is among the few publicly available datasets for non-residential EV charging and will be utilized in this work. The dataset contains charging records from two stations”, where “non-residential EV charging . . . stations” are within the broadest reasonable interpretation of EVSE data, and, in view of Shahriar, the training dataset that includes telematics data, the telematics data, also includes EVSE data).
The reasons for obviousness were discussed in regard to the rejection of Claim 4, for the combination with Shahriar, and in regard to the rejection of Claim 5, for the combination with Shafee, and remain applicable here.
Response to Arguments
Applicant's arguments filed on February 13th, 2026 have been fully considered. Each argument is addressed in detail below.
I. Applicant argues the objections to the drawings should be withdrawn (Applicant’s Remarks, 02/13/2026, Pg. 10, Section “Objection to the Drawings”).
Applicant’s amendments to the specification have overcome each and every objection to the drawings, as previously set forth in the November 14th, 2025 Office Action. As a result, these objections have been withdrawn.
II. Applicant argues the rejections to the claims, under 35 USC § 112, should be withdrawn (Applicant’s Remarks, 02/13/2026, Pg. 10, Section “Rejection Under 35 U.S.C. § 112”).
Applicant’s amendments have overcome each and every rejection to the claims, under 35 USC § 112, as previously set forth in the November 14th, 2025 Office Action. As a result, these rejections have been withdrawn.
III. Applicant argues the rejections to the claims, under 35 USC § 103, should be withdrawn (Applicant’s Remarks, 02/13/2026, Pg. 11-16, Sections “Prior Art Rejections” and “New Claims”).
In response to Applicant’s amendments, the previously communicated rejections under 35 U.S.C. § 103, have been withdrawn. However, Applicants arguments are not persuasive in light of the new grounds for rejection, under 35 U.S.C. § 103, discussed in detail above. The new grounds of rejection rely on new prior art of record to teach the new combination of elements in the amended independent claims, which were not presented in this arrangement in any of the previously presented claims. As a result, Applicant’s arguments are rendered moot.
However, for clarity of the record and in the interest of compact prosecution, Applicant arguments applicable to the new grounds of rejection are addressed below. In order to facilitate discussion, relevant MPEP and CFR excerpts are reproduced here.
According to MPEP 2111, “During patent examination, the pending claims must be given their broadest reasonable interpretation consistent with the specification” (internal quotation marks omitted) (see also Phillips v. AWH Corp., 415 F.3d 1303, 1316, 75 USPQ2d 1321, 1329 (Fed. Cir. 2005)).
Additionally, according to MPEP 2145, “Although the claims are interpreted in light of the specification, limitations from the specification are not read into the claims” (see also In re Van Geuns, 988 F.2d 1181, 26 USPQ2d 1057 (Fed. Cir. 1993))
Furthermore, according to 37 CFR 1.111(b), a proper response to an Office Action "must be reduced to a writing which distinctly and specifically points out the supposed errors in the examiner’s action”.
First, Applicant argues Hoffmann fails to disclose “a corresponding confidence level value” because “these probabilities are . . . [not] reported out as "confidence levels"” (Pg. 14. Para. 1).
Here, as discussed in detail above, Hoffmann discloses elements that are within the broadest reasonable interpretation of “a corresponding confidence level value” (see MPEP 2111).
Specifically, “probability of an EV charging at each sample” is within the broadest reasonable of a confidence level value because it is a value that indicates the confidence level associated with the output (Hoffmann, Pg. 2, Col. 2, Para. 4; Hoffmann, Pg. 3, Col. 1, Para. 2) (relevant excerpts associated with the cited paragraphs are reproduced in regard to the 103-based rejection of claim 1 above).
Also, any limitations recited in the specification, which may require the confidence level values to be reported out as confidence levels, are not recited into the claims and, therefore, are not read into the claims (see MPEP 2145).
Finally, Applicant has asserted, but not specifically pointed out why the elements of Hoffmann relied upon to teach these claims are insufficient (see 37 CFR 1.111(b)).
As a result, the argument is not persuasive.
Second, Applicant argues Mimaroglu fails to disclose “labels specifying whether or not a user location uses the location for use of an EV or other device” or “determining a label specifying whether or not an EV or other device uses the location” (Pg. 14-15, Para. 2-1).
Here, as discussed in detail above, Mimaroglu discloses elements that are within the broadest reasonable interpretation of both “labels specifying whether or not a user location uses the location for use of an EV or other device” and “determining a label specifying whether or not an EV or other device uses the location” (see MPEP 2111).
Specifically, the values, such as “null” or “zero values”, within the context of “electric vehicles”, are within the broadest reasonable interpretation of labels specifying whether or not a user location uses the location for use of an EV or other device because it is a value assigned to a location which indicates use for EV charging (see Mimaroglu, Para. [0077]; Mimaroglu, Para. [0032]) (relevant excerpts associated with the cited paragraphs are reproduced in regard to the 103-based rejection of claim 1 above).
Additionally, the “predictions . . . generated by the trained machine learning model” are within the broadest reasonable interpretation of determining a label specifying whether or not an EV or other device uses the location because it is “disaggregated energy usage for the target device” that will assign a label indicating “indicate whether each user location of the inferencing population, “source location”, is used for running the “target device”, which within the context of “electric vehicles”, is within the broadest reasonable interpretation of labels specifying whether or not a user location uses the location for use of an EV or other device (Mimaroglu, Fig. 8; Mimaroglu, Para. [0032]; Mimaroglu, Para. [0048]; and Mimaroglu, Para. [0122]) (relevant excerpts associated with the cited paragraphs are reproduced in regard to the 103-based rejection of claim 1 above).
Furthermore, any limitations recited in the specification, which may conflict with Mimaroglu, are not recited into the claims and, therefore, are not read into the claims (see MPEP 2145).
Finally, Applicant has asserted, but not specifically pointed out why the elements of Mimaroglu relied upon to teach these claims are insufficient (see 37 CFR 1.111(b)).
As a result, the argument is not persuasive.
IV. Applicant argues the rejections to the claims, under 35 USC § 101, should be withdrawn (Applicant’s Remarks, 02/13/2026, Pg. 16, Sections “Rejection Under 35 U.S.C. § 101” and “New Claims”).
Applicant’s amendments have overcome each and every rejection to the claims, under 35 USC § 101, as previously set forth in the November 14th, 2025 Office Action. As a result, these rejections have been withdrawn.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MATTHEW BRYCE GOLAN whose telephone number is (571)272-5159. The examiner can normally be reached Monday through Friday, 8:00 AM to 5:00 PM ET.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Alexey Shmatov can be reached at (571) 270-3428. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/MATTHEW BRYCE GOLAN/Examiner, Art Unit 2123
/ALEXEY SHMATOV/Supervisory Patent Examiner, Art Unit 2123
Prosecution Insights