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 § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 21, 23 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
Claim 21, 23 recites, “the long-term predicting module determines the first portion of the predicted energy load profile further based on Internet of Things (IoT) data”. The specification states “The long-term predicting module can be designed to capture the long-term relation between weather, IoT data, and energy consumption, while the short-term predicting module can be configured to capture the recent trend in energy consumption at the facility. System 310 can use this machine learning model to determine the predicted energy load profile by: (a) determining, via the long-term predicting module, a first portion of the predicted energy load profile based at least in part on the weather forecast data and the sensor data for the facility;” pp 0034. The specification does not disclose or describe that the first portion of the predicted energy load profile is further based on Internet of Things (IoT) data. The specification fails to provide support for the claimed limitation.
Claim 21, 23 recites, “the short-term prediction module determines the remaining portion of the predicted energy load profile further based on a recent trend of energy consumption at the facility” The specification recites “while the short-term predicting module can be configured to capture the recent trend in energy consumption at the facility. System 310 can use this machine learning model to determine the predicted energy load profile by: (a) determining, via the long-term predicting module, a first portion of the predicted energy load profile based at least in part on the weather forecast data and the sensor data for the facility; and then (b) after determining the first portion, determining, via the short-term predicting module, a remaining portion of the predicted energy load profile based on residuals from the long- term predicting module” pp 034. The specification does not disclose or describe that the remaining portion of the predicted energy load profile further based on a recent trend of energy consumption at the facility. The specification fails to provide support for the claimed limitation.
Claim Rejections - 35 USC § 102
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 the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claim(s) 1, 6-10, 11, 16-20 is/are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Lee et al (US20200073342A1) hereinafter Lee.
As per Claim 1, 11, Lee discloses, A system comprising: one or more processors; and one or more non-transitory computer-readable media storing computing instructions that, when run on the one or more processors, cause the one or more processors to perform: determining, via a machine learning model, a predicted energy load profile for a facility based at least in part on weather forecast data and sensor data for the facility, wherein: “The system includes a local building system configured to collect building data and communicate the building data to a cloud platform and the cloud platform configured to receive the building data from the local building system via the network, determine whether to retrain a trained load prediction model based on at least some of the building data, retrain the trained load prediction model based on at least some of the building data in response to a determination to retrain the trained load prediction model, determine a load prediction for the building based on the retrained load prediction model, and cause the local building system to operate” (abstract) “wherein the current building data includes at least one of a weather forecast, a current occupancy level, an actual previous load, a previous load prediction, a time of day, or a day of week” (0006)
the sensor data are received from one or more energy monitoring sensors for one or more devices in the facility; “BMS 606 can be configured to monitor conditions within a controlled building or building zone. For example, BMS 606 may receive input from various sensors (e.g., temperature sensors, humidity sensors, airflow sensors, voltage sensors, etc.) distributed throughout the building and may report building conditions to energy storage controller 506. Building conditions may include, for example, a temperature of the building or a zone of the building, a power consumption (e.g., electric load) of the building” (0133)
determining one or more demand shedding time slots based at least in part on peak periods and the predicted energy load profile; “controller 506 may receive notice of a synchronous reserve event from an IBDR program which requires energy storage system 500 to shed a predetermined amount of power. Controller 506 may determine that it is optimal to participate in the IBDR program if cold thermal energy storage 532 has enough capacity to provide cooling for building 502 while the load on chiller subplant 522 is reduced in order to shed the predetermined amount of power.” (0119)(0076-0077)
determining one or more demand shedding events for the one or more devices to be scheduled during the one or more demand shedding time slots; “Controller 552 can be configured to predict the thermal energy loads (e.g., heating loads, cooling loads, etc.) of the building for plurality of time steps in an optimization period (e.g., using weather forecasts from a weather service 604). Controller 552 may generate control decisions that optimize the economic value of operating system 550 over the duration of the optimization period subject to constraints on the optimization process (e.g., energy balance constraints, load satisfaction constraints, etc.). The optimization process performed by controller 552 is described in greater detail below.” (0159)
and causing a respective performance of each of the one or more demand shedding events by the one or more devices during the one or more demand shedding time slots. “Each of subplants 520 may include equipment that can be controlled by controller 552 to optimize the performance of system 550. Subplant equipment may include, for example, heating devices, chillers, heat recovery heat exchangers, cooling towers, pumps, valves, and/or other devices of subplants 520. Individual devices of generator subplants 520 can be turned on or off to adjust the resource production of each generator subplant. In some embodiments, individual devices of generator subplants 520 can be operated at variable capacities (e.g., operating a chiller at 10% capacity or 60% capacity) according to an operating setpoint received from controller 552.” (0129)
wherein: the machine learning model comprises an ensemble of a long-term predicting module and a short- term predicting module “In some embodiments, load/rate predictor 622 may use, but is not limited to, various supervised or unsupervised machine learning models. The models may be a generalized additive model (GAM), a Random Forecast (RF), a support vector machine (SVM) model, a generalized linear model (GLM), a nonlinear autoregressive exogenous (NARX), a neural network (e.g., a recurrent neural network (RNN) model (e.g., a long short term memory (LSTM) model), a convolutional neural network (CNN) model, etc.), and/or any other type of suitable prediction model.” (0288) “Energy storage controller 506 can be configured to predict the thermal energy loads (e.g., heating loads, cooling loads, etc.) of the building for plurality of time steps in an optimization period (e.g., using weather forecasts from a weather service 604). Energy storage controller 506 may also predict the revenue generation potential of IBDR programs using an incentive event history (e.g., past clearing prices, mileage ratios, event probabilities, etc.) from incentive programs 602. Energy storage controller 506 may generate control decisions that optimize the economic value of operating energy storage system 500 over the duration of the optimization period subject to constraints on the optimization process” (0135) “Demand charge periods can include relatively long periods (e.g., monthly, seasonal, annual, etc.) or relatively short periods (e.g., days, hours, etc.). Each of these periods can be divided into several sub-periods including off-peak, partial-peak, and/or on-peak. Some demand charge periods are continuous (e.g., beginning Jan. 1, 2017 and ending Jan. 31, 2017), whereas other demand charge periods are non-continuous (e.g., from 11:00 AM-1:00 PM each day of the month).” (0197)
“wherein determining, via the machine learning model, the predicted energy load profile further comprises: determining, via the long-term predicting module, a first portion of the predicted energy load profile based at least in part on the weather forecast data and the sensor data for the facility; and after determining the first portion, determining, via the short-term predicting module, a remaining portion of the predicted energy load profile based on residuals from the long-term predicting module. “a single source of electricity (e.g., an electric grid) is considered with multiple demand charges applicable to the electricity source (i.e., q=1 . . . N, where N is the total number of demand charges). The system includes a battery asset which can be allocated over the optimization period by charging or discharging the battery during various time steps. Charging the battery increases the amount of electricity purchased from the electric grid, whereas discharging the battery decreases the amount of electricity purchased from the electric grid.Demand charge module 910 can modify the cost function J(x) to account for the N demand charges as shown in the following equation: [see reference for equation] where the term −Pbat i +eLoadi represents the total amount of electricity purchased from the electric grid during time step i (i.e., the total electric load eLoadi minus the power discharged from the battery Pbat i ). Each demand charge q=1 . . . N can be accounted for separately in the cost function J(x) by including a separate max( ) function for each of the N demand charges. The parameter rd q indicates the demand charge rate associated with the qth demand charge (e.g., $/kW) and the weighting factor wd q indicates the weight applied to the qth demand charge. (0202, 0203)
As per Claim 6, 16, Lee discloses, wherein determining, via the machine learning model, the predicted energy load profile further comprises determining, via the machine learning model, the predicted energy load profile based on one or more of: a size of the facility; one or more geographic features for the facility; a footfall of the facility; one or more assets for the facility; or operating hours for the facility. “Still referring to FIG. 9, high level optimizer 632 is shown to include an asset sizing module 916. Asset sizing module 916 can be configured to determine the optimal sizes of various assets in a building, group of buildings, or a central plant. Assets can include individual pieces of equipment or groups of equipment. For example, assets can include boilers, chillers, heat recovery chillers, steam generators, electrical generators, thermal energy storage tanks, batteries, air handling units, or other types of equipment in a building or a central plant (e.g., HVAC equipment, BMS equipment, etc.). In some embodiments, assets include collections of equipment which form a subplant of a central plant (e.g., central plant 118). For example, assets can include heater subplant 521, chiller subplant 522, heat recovery chiller subplant 523, steam subplant 524, electricity subplant 525, or any other type of generator subplant 520. In some embodiments, assets include hot thermal energy storage 531 (e.g., one or more hot water storage tanks), cold thermal energy storage 532 (e.g., one or more cold thermal energy storage tanks), electrical energy storage 533 (e.g., one or more batteries), or any other type of storage subplant 530.” (0232)
As Per Claim 7, 17, Lee discloses wherein determining the one or more demand shedding time slots further comprises: determining a cut-off threshold based on historical peak energy consumption data of historical energy load profiles; and determining the one or more demand shedding time slots based on the predicted energy load profile, the cut-off threshold, and the peak periods. “For example, controller 506 may receive notice of a synchronous reserve event from an IBDR program which requires energy storage system 500 to shed a predetermined amount of power. Controller 506 may determine that it is optimal to participate in the IBDR program if cold thermal energy storage 532 has enough capacity to provide cooling for building 502 while the load on chiller subplant 522 is reduced in order to shed the predetermined amount of power.” (0119) “High level optimizer 632 can reduce PLC costs by shifting the customer's load to non-peak times or shaving the customer's peak load. This can be done, for example, by precooling the building during non-peak times, using thermal energy storage, and/or using electrical energy storage such as a battery asset.” (0222) “High level optimizer 632 may determine an optimal set of control decisions for each time step in the prediction window in order to optimize (e.g., maximize) the value of operating energy storage system 500. Control decisions made by high level optimizer 632 may include, for example, load setpoints for each of generator subplants 520, resource purchase amounts for each type of resource purchased from utilities 510, and/or an amount of each resource sold to energy purchasers 504. In other words, the control decisions may define resource allocation at each time step.” (0160)
As per claim 8, 18, Lee discloses, wherein determining the one or more demand shedding time slots further comprises: determining whether a demand-shedding-deferring event exists in the weather forecast data; and upon determining that the demand-shedding-deferring event exists, excluding at least one deferrable time slot in a deferring time period from the one or more demand shedding time slots. “such as weather data from weather service 604 in some embodiments. Based on the received inputs, load/rate predictor 622 is configured to predict a load and/or rate for a future time and/or times and send the predictions to high level optimizer 632, in some embodiments. High level optimizer 632 is described with greater detail in FIG. 6 and elsewhere herein. Based on the predicted loads and/or rates, high level optimizer 632 can be configured to cause the ESS 1026 to charge and/or discharge energy by generating charging and/or discharging schedules via an optimization with an objective function, various constrains, and/or the predicted loads and/or rates.” (0251) “it may be desirable to use the last several weeks of weather data because weather is likely steady (e.g., environmental conditions are within a predetermined range). However, in early spring, when temperatures may vary substantially, it may be desirable to use only weather data from the past few days. Data selector 1110 may also select data based on occupancy. On Christmas Day, using occupancy data from Christmas Day of a previous year and/or years may be desirable. Using a full week of data may predict an occupancy that is too high. Yet, during the month of February, when there are no holidays, an average of occupancy during business weeks for the past several business weeks may be desirable.” (0270) “The load and rate models 1128 and 1130 may be used to predict a load and/or rate for some time, t (or multiple times into the future), via predictor 1112. Predictor 1112 can be configured to send rate prediction 1118 and/or load prediction 1120 to demand response optimizer 630” (0279)
As per Claim 9, 19, Lee discloses, wherein :determining whether the demand-shedding-deferring event exists further comprises comparing a first portion and a second portion of the weather forecast data; the first portion is associated with a time period of the predicted energy load profile; and the second portion is associated with a subsequent time period immediately following the time period. “For example, rule selector 1114 can identify that ambient temperature has dropped by a predefined amount over a previous week. One of the rules of the look-up table may indicate to select the two previous weeks of historical data of the historical database 1108 to train the models 1126. Rule selector 1114 can provide the identified rule of the look-up table to rule executor 1112. Rule executor 1112 can select the two previous week of data from historical database 1108 can provide the selected data to model manager 1122. As another example, one of the rules of the look-up table may be a particular date (e.g., July 1st). For a school, occupancy may fall during summer months due to the absence of students, this may correspond to a lower load. In response to determining that a current data is July 1st, rule selector 1114 may select the July 1st rule of the look-up table which may have a specific condition (e.g., use the data from June and/or from one or more previous summers) to train the models 1126. Based on the July 1st rule selected by rule selector 1114, rule executor 1112 can be configured to select data for use in training models 1126 based on the condition of the July 1st rule. (0284-0285)
As per Claim 10, 20, Lee discloses, wherein: the one or more demand shedding events for the one or more devices to be scheduled during the one or more demand shedding time slots further comprises one or more of:changing a respective setting for at least one adjustable device of the one or more devices during at least one of the one or more demand shedding time slots based at least in part on the predicted energy load profile; or switching a respective energy source for at least one of the one or more devices during at least one of the one or more demand shedding time slots based on operating configurations of one or more alternative energy sources for the facility. “For example, controller 506 may receive notice of a synchronous reserve event from an IBDR program which requires energy storage system 500 to shed a predetermined amount of power. Controller 506 may determine that it is optimal to participate in the IBDR program if cold thermal energy storage 532 has enough capacity to provide cooling for building 502 while the load on chiller subplant 522 is reduced in order to shed the predetermined amount of power.” (0119) “High level optimizer 632 can reduce PLC costs by shifting the customer's load to non-peak times or shaving the customer's peak load. This can be done, for example, by precooling the building during non-peak times, using thermal energy storage, and/or using electrical energy storage such as a battery asset.” (0222) “High level optimizer 632 may determine an optimal set of control decisions for each time step in the prediction window in order to optimize (e.g., maximize) the value of operating energy storage system 500. Control decisions made by high level optimizer 632 may include, for example, load setpoints for each of generator subplants 520, resource purchase amounts for each type of resource purchased from utilities 510, and/or an amount of each resource sold to energy purchasers 504. In other words, the control decisions may define resource allocation at each time step.” (0160)
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.
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 4- 5, 14-15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Lee et al (US20200073342A1) hereinafter Lee in view of Guangzhou Dayu Chuangfu Technology (CN111832225) hereinafter GDC.
As per Claim 4, 14, Lee discloses, [LSTM long short term memory training model] “The models may be a generalized additive model (GAM), a Random Forecast (RF), a support vector machine (SVM) model, a generalized linear model (GLM), a nonlinear autoregressive exogenous (NARX), a neural network (e.g., a recurrent neural network (RNN) model (e.g., a long short term memory (LSTM) model), a convolutional neural network (CNN) model, etc.), and/or any other type of suitable prediction model.” (0288)
Lee does not explicitly describe the well known aspects and function of LSTM as per the claims, however, GDC accurately discloses wherein: the long-term predicting module is trained based on a long-term training dataset; the short-term predicting module is trained based on a short-term training dataset; the long-term training dataset is determined based on historical training data for the facility in a long-term time period; and the short-term training dataset comprises historical residuals in a short-term time period from training the long-term predicting module. “The training data set construction method includes: for each fragment library, splicing all the kinematic fragments s in each fragment library to obtain the speed-time sequence Si (ie long fragment), and for the four fragment libraries, correspondingly obtain S L , SM , SH and S EH four long fragments. The above training data set is input into the long-term short-term memory neural network model (LSTM) for training, and the long-term short-term memory neural network model (LSTM) continuously learns the change characteristics of the speed-time series contained in the fragment library of different categories during the training process, Four corresponding time-speed sequence prediction models are trained to obtain a trained long-short-term memory neural network model. It should be noted that the time-speed sequence prediction model in this specification is the same model as the long-short-term memory neural network model. The long-short-term memory neural network model can complete the reconstruction of time series according to the spatiotemporal correlation of different categories of fragment libraries, rely on model training to identify and strengthen the temporal correlation characteristics, and predict different categories of kinematic fragments with representative characteristics. Since the network structure of the long-short-term memory neural network model is definite, it needs to be input as a vector of a specific dimension, and the length of the data in the training data set is different, which does not meet this condition. Therefore, the training data set needs to be pre-prepared before inputting the model. Processing, the training data set is processed into supervised learning data that can be used by the time-velocity series prediction model” (0124-0126)
One of ordinary skill in the art would have recognized that applying the known technique of Lee, namely, LSTM, with the known techniques of GDC, namely, specifics of using the well known LSTM training models, would have yielded predictable results and resulted in an improved system. Accordingly, applying the teachings of Lee with the teachings of Fromson to apply and use LSTM would have been recognized by those of ordinary skill in the art as resulting in an improved machine learning model tools using a known technique for processing and training a model “The invention utilizes the LSTM cycle neural network as the prediction core to generate the driving condition curve of the automobile, and constructs the time-speed sequence prediction model.” (0132)
As per claim 5, 15, Lee discloses, wherein the computing instructions, when run on the one or more processors, further cause the one or more processors to perform: detecting outlier historical data of the historical training data based on one or more of: one or more equipment-type-based rules; one or more statistics-based rules; or an outlier-detection module trained based on one or more of: historical weather data, historical sensor data, or historical energy load profiles for the facility; and removing the outlier historical data from the historical training data. Historian 1310 can be configured to add (e.g., append), the real-time data 1325 to historical data of historical database 1308 (e.g., previously received real-time data 1325). The cloud platform 1302 can generate a load prediction and/or a rate prediction and communicate the load prediction to ESS controller 1222. ESS controller 1222 can be configured to determine the charging and/or discharging schedule of ESS 1026 based on the received load prediction and/or rate prediction. Based on the historical data stored in historian database 1308, load/rate predictor 1306 can train and/or execute load model 1324. Load/rate predictor 1306 can include some and/or all of the functionality of load/rate predictor 622 as described with reference to FIG. 11 such that the load/rate predictor 1306 can be configured to determine whether load model 1324 should be retrained based on changing conditions identified in the data of historian database 1308 (e.g., changes in occupancy, changes in load, etc.) as described with reference to FIG. 11. (0312-0313)
Claim(s) 21-24 is/are rejected under 35 U.S.C. 103 as being unpatentable over Lee et al (US20200073342A1) hereinafter Lee in view of Engelstein et al (hereinafter Engel) (US Patent Pub US20200333767)
As per claim 21, 23, Engel discloses, “the long-term predicting module determines the first portion of the predicted energy load profile further based on Internet of Things (IoT) data;” a machine learning system for use with a power distribution system is provided including a data library configured to store a plurality of data samples, wherein at least a portion of the data samples are associated with one or more intelligent electronic devices (IEDs); [0013] “and the short-term prediction module determines the remaining portion of the predicted energy load profile further based on a recent trend of energy consumption at the facility.” In step 2408, a baseline of values from the last 48 hours is also extracted. In step 2410, artificial attributes of the predicted weather data 2404 are extracted, e.g., temperature trends, windspeed trends, humidity trends, etc. Next, in step 2412, the extracted attributes from the future data points 2402, the predicted weather data 2404, the last baseline from the training data 2407 and the last weather trends from the training data 2409 are then merged [0354]
One of ordinary skill in the art would have recognized that applying the known technique of Lee, with the teaching of Engel’s features of data information included in a predictive energy model, would have yielded predictable results and resulted in an improved system. Accordingly, applying the teachings of Lee with the teachings of Engel to utilize known and common data sources of energy and methods of predicting energy usage would have been recognized by those of ordinary skill in the art as resulting in an improved machine learning model tools “he machine learning module generates a recommendation as to when ideal periods within the predetermined future time interval to lower and increase energy usage to reduce costs and the action module sends a control signal to the at least one client to turn on or off loads to increase or decrease energy usage in a way that lowers costs” (0023 Engel)
As per claim 22, 24, Engel discloses, “wherein determining the predicted energy load profile includes combining the first portion of the predicted energy load profile and the remaining portion of the predicted energy load profile into a final predicted energy load profile” Once the training model is built, the training model may be employed for predicting future results. Referring to FIG. 24, a method 2400, algorithm logic, for predicting future values, e.g., energy or demand, is provided. Initially, 32 days of future datapoints 2402 is entered as a request to system 1400 and predicted weather data 2404 is received by system 1400. Additionally, the last baseline from the training data 2407 and the last weather trends from the training data 2409 are received by system 1400. It is to be appreciated the last baseline from the training data 2407 and the last weather trends from the training data 2409 are generated from the latest training model that was built. In step 2408, a baseline of values from the last 48 hours is also extracted. In step 2410, artificial attributes of the predicted weather data 2404 are extracted, e.g., temperature trends, windspeed trends, humidity trends, etc. Next, in step 2412, the extracted attributes from the future data points 2402, the predicted weather data 2404, the last baseline from the training data 2407 and the last weather trends from the training data 2409 are then merged. It is to be appreciated that cumulative values are taken partially from the training steps above, described in relation to FIG. 23, so that the method has cumulative data in addition to hourly data, so cumulative for the month or day etc. This allows for prediction to be done hourly, daily, weekly, monthly etc. Then, the merged data is sorted by day of the week, in step 2414. [0353-0354]
One of ordinary skill in the art would have recognized that applying the known technique of Lee, with the teaching of Engel’s features of data information included in a predictive energy model, would have yielded predictable results and resulted in an improved system. Accordingly, applying the teachings of Lee with the teachings of Engel to utilize known and common data sources of energy and methods of predicting energy usage would have been recognized by those of ordinary skill in the art as resulting in an improved machine learning model tools “he machine learning module generates a recommendation as to when ideal periods within the predetermined future time interval to lower and increase energy usage to reduce costs and the action module sends a control signal to the at least one client to turn on or off loads to increase or decrease energy usage in a way that lowers costs” (0023 Engel)
Response to Arguments
Applicant's arguments filed have been fully considered but they are not persuasive.
Applicant argues, “Paragraph [0197] simply discloses that demand charge periods can include "relative long periods" and "relatively short periods," and Paragraph [0288] simply discloses that "various supervised or unsupervised machine learning models" can be used. These cited portions do not disclose an ensemble of a long-term predicting module and a short-term predicting module.”
Lee clearly discloses a prediction model (0288) which provides for both long and short term predictions (broadly recited term ‘ensemble’ by definition simply means a group providing a single effect) and when and how the prediction model should be retrained using long or short term prediction criteria (0065) Lee clearly teaches retraining the overall prediction model using new models based on the new or more current data, whether it be from short term (few days) to weeks of weather data (long term) and selecting and generating the update to the LSTM (0266)
Applicant argues, “During the telephone interview of August 18, 2025, to make for the deficiencies of these cited portions, the Examiner further cited Paragraph [0065] of Lee, asserting that using a long- term predicting module and a short-term predicting module to determine the predicted energy load profile in the manner recited is inherent in the teachings of Lee. Applicant respectfully disagrees. First, Applicant respectfully submits that the fact that a certain result or characteristic may occur or be present in the prior art is not sufficient to establish the inherency of that result or characteristic.”
During this interview, the Supervisory Examiner pointed to the Applicant’s lack of limitations in the claims and/or the specification with respect to “ensemble” and “residual”, which are the underpinning to their argument. The term ensemble, which the Applicant expressed was not present in Lee, is simply “a group” or number of models including in the machine learning model, and is provided no further limitation or definition by the Applicant. Lee clearly discloses this in numerous recitations throughout the specification, describing retraining and model updates, creating new short or long term models for inclusion in the prediction model, and even disclosing well known and widely used models such as LSTM which by is core nature is fundamentally a Long Short Term Memory Model, using both long memory models and short memory models to create the LSTM.
Further the Applicant incorrectly contended the “residuals from the long term predicting module” implied or included specific elements that are unique to their invention. As is clear from the claims and the specifications, the Applicant fails to provide ANY limitation or definition of ‘residuals’ and attempts to imply or construe they are more than a simply plain and ordinary meaning in the art. In fact the term is sparsely mentioned in the specification and provide no direction or implication as to what are residuals in the claimed invention. Lee clearly teaches utilizing both long and short term models in a prediction model, further clearly teaching utilizing residuals or left over data or pieces of information from the long term model to train or retrain the short term model (days vs weeks vs months from data collected and used by the long term model) (0266, 0270, 284, 311, 331, 338) and further clearly discloses data from models can be parsed or combined or compared (residual data) to create/train/generate new models based on the desire output required. (0266, 0270, 331, 338) Further, simple machine learning models and especially those mentioned in Lee “generalized additive model (GAM), a Random Forecast (RF), a support vector machine (SVM) model, a generalized linear model (GLM), a nonlinear autoregressive exogenous (NARX), a neural network (e.g., a recurrent neural network (RNN) model (e.g., a long short term memory (LSTM) model), a convolutional neural network (CNN) model, etc.), and/or any other type of suitable prediction model” (0011), inherently ‘learn’ by training and retraining each model through feedback and previous information, utilizing the residual information from models to train and retrain models as learning progresses.
Numerous attempts were made to discuss further limitations or amendments that could be made to clarify the Applicant position prior to the filing of the proposed amendment, however the Applicant did not want to discuss these suggestions during the interview. As was pointed out in the interview and interview summary, the claims and arguments appear to be focused on two broadly recited terms ‘residual’ and ‘ensemble’ that have no limitation or definition in the claims or the specification beyond that of their simplest definition.
Conclusion
THIS ACTION IS MADE FINAL. 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 KENNETH M LO whose telephone number is (571)272-9774. The examiner can normally be reached M-F 830a - 6pm.
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, John Cottingham can be reached at 571-272-9877. 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.
KENNETH M. LO
Supervisory Patent Examiner
Art Unit 2136
/KENNETH M LO/ Supervisory Patent Examiner, Art Unit 2116
Prosecution Insights