DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA . Claims 1-20 were amended and are still pending in this Application.
Response to Amendments/Remarks
Applicant’s argument/remarks, on pages 8-12, with respect to rejections to claims 1-20 under 35 USC § 103(a) have been fully considered and they are respectfully persuasive. Therefore the rejections have been withdrawn. However, upon further consideration, a new ground(s) of rejection is made, see the new rejection below.
On pages 8-9, The Applicant argues that:
“First, the claims now recite that the training dataset includes "labeled instances" of at least one of a component failure event or a non-nominal operational state "occurring at an identified component of the power system”… “Second, the claims now expressly recite that "the prediction identifying the respective component predicted to experience the future non-nominal operation or component failure…Applicant submits that neither Spalt nor McElhinney, alone or in combination, teaches or suggests the amended claim elements… Spalt' s training methodology does not involve providing labeled instances of component failure events or non-nominal operational states to the model...”. These arguments are partially persuasive.
In response to the previous arguments, Spalt and McElhinney while teaching supervised learning models which suggests labeled data, they do not explicitly state so in their disclosure. Thus, a new reference has been provided to teach this limitations of a training dataset including labeled data. The original disclosure does not provide any specific definition of “labeled data”. Thus, it has been interpreted in the broadest reasonable interpretation as raw data associated with a condition/failure.
With respect to the second limitation, the original disclosure dos not explicitly recite “the prediction identifying the respective component predicted to experience the future non-nominal operation or component failure”, however, it is obvious in the context of the invention that a prediction of data of a components identifies such components (0064 and 0066). Spalt clearly teaches this second limitation by identifying islanding or anti-islanding conditions of a generator and/or inverter (see 0110), voltages drops conditions at a specific substation (0106), a transformer (see 0055 and 0112). Also, McElhinney clearly teaches the prediction identifying the respective component predicted to experience the future non- nominal operation or component failure (see [0014] “given asset” means specific identified assets; [0012] “to define a model for predicting asset failures..”; [0117]). Also, the new reference clearly teaches this last limitations as well (see [0045], [0049], [0052], [0071] [0073]).
On pages 10-11, The Applicant further argues that:
“…The claims thus recite a specific training paradigm: the model receives labeled
failure data tied to specific components, and as a result, the model's predictions identify which component is at risk. Spalt' s model is not trained on failure data at all, and McElhinney' s failure model, while trained on historical failure data, is applied to individual assets that have already been selected, it does not take incoming sensor data from across a power system and identify which component will fail. The two references describe fundamentally different architectures and training approaches Combining them would not yield a single model trained on labeled component-specific failure data that then processes incoming sensor characteristics and outputs a prediction identifying a specific at-risk component, because neither reference teaches such a model, and combining a grid-level optimization model (Spalt) with an individual-asset health profiler (McElhinney) does not logically produce one”. These arguments are respectfully unpersuasive.
As stated previously, while Spalt-McElhinney teach supervised trained models, they do not explicitly teach labeled data. The reference Spalt clearly uses a supervised trained model to predicts conditions and faults of components within the grid as stated above (also, see 0035). Spalt is silent about the specifics of the training data for training the model. McElhinney clearly teaches a machine learning model trained with historical failure data as admitted by the Applicant in the arguments. Therefore, the combination of Spalt-McElhinney teach the argued limitation of training a model with already known faults associated to assets. Training a model involves the identification of features/labels to identify faults within the data. The combination of these two references facilitates the training of a model to predict abnormal states of specific components of a grid system. The Applicant has not provided any evidence that the two references can not be combined. The rejection clearly states that it would have been obvious to use training data with known failure data of an asset for training purposes which would have been obvious to one person of ordinary skill in the art to combine would have had a "reasonable expectation of success" in making that combination.
On page 11, the Applicant further argues that:
“This establishes a closed-loop protection paradigm in which: (1) the model is trained on labeled component-specific failure data; (2) the model outputs a prediction identifying a particular at-risk component; and (3) the control circuitry takes a specific protective action (e.g., modifying a transformer setpoint or disconnecting a protection device) targeted at the identified component to prevent the predicted failure. No single reference teaches this sequence, and the combination does not arrive at it because neither reference describes taking protective action in direct response to a component-specific predictive identification. Spalt' s control actions are optimization-driven (e.g., adjusting tap settings for VVO), and when Spalt mentions disabling a device, it is in the context of a device that "is nearing failure or has failed" (Spalt at ,i 117), a reactive measure, not a predictive one targeting a specific component identified by the model before failure occurs. McElhinney does not describe any control actions at all; it generates health profiles and failure likelihood scores”. These arguments are respectfully unpersuasive,
With respect to the second limitation, Spalt clearly teaches this second limitation by identifying islanding or anti-islanding conditions of a generator and/or inverter (see 0110), voltages drops conditions at a specific substation (0106), a transformer (see 0055 and 0112) and controlling this devices parameter to bring device in a non-nominal condition to desired nominal conditions (desired ranges) or to avoid failures. Also, McElhinney clearly teaches the prediction identifying the respective component predicted to experience the future non- nominal operation or component failure (see [0014] “given asset” means specific identified assets; [0012] “to define a model for predicting asset failures..”; [0117]). Also, the new reference clearly teaches this last limitations as well 9see [0045], [0049], [0052], [0071] [0073]). McElhinney, while not relied upon, teaches (2) the model outputs a prediction identifying a particular at-risk component; and (3) the control circuitry takes a specific protective action (see [0019]).
Claim Rejections - 35 USC § 103
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 1-3, 5, 7-11, 13, 16, 18 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Spalt et al (US 20210373518) in view of McElhinney et al (US 20160155315) and Gundel et al (US 20210373063).
As per claim 1, Spalt teaches a system (see Fig. 1) comprising:
a sensor configured to determine at least one performance characteristic of a component of a power system, the sensor further configured to transmit the at least one characteristic (see Fig. 1 metering devices 118a..n or 120a sense grid characteristics and transmit the measured grid characteristics; also, see [0045] “monitoring devices 118a-118n that can be coupled through optional potential transformers 120a-120n to secondary utilization circuits 116. The monitoring or metering devices 118a-118n can detect (e.g., continuously, periodically, based on a time interval, responsive to an event or trigger) measurements and continuous voltage signals of electricity supplied to one or more electrical devices 119 connected to circuit 112 or 116 from a power source 101 coupled to bus 102”; thus, the sensor measure characteristic of components such as lines, power to vehicles…also, see [0053],[0054] and [0070] meters/sensors at inverters);
a processor operatively connected to the sensor (see Fig. 2 processing system 202, including a processing module optimizer 216 and model 214 and controller 218; and see [0014] “The system can include one or more processors. The one or more processors can identify a first plurality of signals detected by one or more devices located at one or more portions of a utility grid”; also, see [0059-0060] and [0061]), the processor configured to receive the at least one characteristic (see [0070] “The data processing system 202 can receive, and store in the utility grid data 22,…voltage…current… temperature”; also, see [0069] and [0071], [0075-0076]) and apply a power system model to the at least one characteristic to determine at least one operational state of the component (see Fig. 2 model 224 and 214; also, see [0078] “…using the received utility grid data to generate an input matrix for input into the machine learning model generator 214 or the machine learning model…”; also, see [0104] and [0105] “… For example, the circuit optimizer 216 can be configured to facilitate Volt/VAR optimization and voltage management; system-state prediction,…; fault, location, isolation, and system restoration; anti-islanding detection; also, see [0107] and [0108] “The data processing system 202 can determine when faults may occur on the utility grid 100, such as due to excess power usage during peak energy consumption (e.g., brownouts due to increased air conditioning use, or increased electric vehicle charging). The data processing system 202 can detect that such a fault is imminent based on detecting a trend in the input matrix using the machine learning model, and pre-emptively adjust signals of the utility grid 100 by generating commands to devices on the utility grid 100”; also, see [0111] “…If the real-time measured do not match the predicted signals, then the data processing system 202 can determine that the behavior of the utility grid 100 may be anomalous. The data processing system 202 can determine the anomalous behavior may be the result of a faulty device on the utility grid 100, or other fault, failure or attack on the utility grid 100…”; also, see [0109] and [0110] “…The data processing system 202, using the machine learning model and an input matrix, can detect when such islanding may occur based on predicting output values for signals such as frequency and voltage… whether to classify the predicted output variables or signals as an instance of islanding…”; also, see [0103], [0105], [0117], [0122]; the operational state of a generator or inverter such as its frequency and voltage or islanding is determined and outputted by the processor by using the model), the power system model trained, using a training dataset comprising historical data associated with at least one historical operational state of the power system, t (see [0010] “The method can include the data processing system training the machine learning model using a historical input matrix generated from historical signals and historical statistical metrics. The data processing system can apply one or more weights to a dimension of the historical input matrix to adjust an impact values of the dimension have on the machine learning model”; also, see [0019], [0040], [0094] and [0100-0101]), to detect anomalous operational states (see [0105] “0105] … For example, the circuit optimizer 216 can be configured to facilitate Volt/VAR optimization and voltage management; system-state prediction,…; fault, location, isolation, and system restoration; anti-islanding detection; also, see [0106] “…The circuit optimizer 216 can also account for general signals, such as time of day, weather, or transmission resources. Using the machine learning model, the circuit optimizer 216 can determine whether the predicted output for a signal falls within a desired range…”; also, see [0108] and [0111] “The data processing system 202 can use the machine learning model to verify behavior of the utility grid 100…If the real-time measured do not match the predicted signals, then the data processing system 202 can determine that the behavior of the utility grid 100 may be anomalous. The data processing system 202 can determine the anomalous behavior may be the result of a faulty device on the utility grid 100, or other fault, failure or attack on the utility grid 100”) and outputting a prediction indicative of a predicted future non-nominal operation or component failure for a respective component (see [0012] and see [0102] and [0106] “…The circuit optimizer 216 can also account for general signals, such as time of day, weather, or transmission resources. Using the machine learning model, the circuit optimizer 216 can determine whether the predicted output for a signal falls within a desired range….”; also, see [0110] “The data processing system 202 can perform anti-islanding detection, which can refer to a condition in which a distributed generator (e.g., a gas power generator or solar inverter device on the utility grid 100) continues to power a location even though electrical grid power is no longer present…. The data processing system 202, using the machine learning model and an input matrix, can detect when such islanding may occur based on predicting output values for signals such as frequency and voltage… For example, the data processing system 202, using the machine learning model, can include an output signal in the output matrix that corresponds to a classification of islanding (e.g., binary classification indicating whether islanding is present, or not present, or a likelihood that islanding is present, or a confidence score associated with the binary classification). The data processing system 202 can predict islanding based on predicted output signals, or an output matrix generated via the machine learning model can include an islanding classification signal or variable”; also, see [0041] “The data processing system, using the input matrix and machine learning model generator, can predict a corresponding output matrix. The model can predict a new output for a future time period, or predict an output value for a variable for which data does not exist in a current time period”; also, see [0108]), the prediction identifying the respective component predicted to experience the future non- nominal operation or component failure (see [0053], [0106], [0110], [0112], [0119] “…. Utility grid data can refer to, or include, timestamps and signals associated with various devices on the utility grid. The utility grid data can include an indication or identifier of the device from which the measurement is obtained…”, conditions of specific components of the grid are identified); and
a control circuitry operatively connected to the processor (see Fig. 2 grid controller 218 connected to control circuitry to effect changes in the grid; see [0072] and [0106]), to adjust an operational parameter of the respective component and another component of the power system based on the prediction to mitigate a detected or predicted non-nominal operation or prevent the component failure (see [0073] “Commands 226 data structure can include control commands to control devices on the utility grid 100, alerts, notifications, reports, instructions, or other actions that can be performed by the data processing system 202 responsive to, or based on, utility grid data 222. Examples of commands 226 can include a command to increase or decrease a tap setting of a voltage regulating transformer 106a on the primary circuit 112, or change a setpoint on the primary circuit 112 or secondary circuit 116. Additional actions or determinations can include determining a topology of a utility grid 100 (e.g., main branches, secondary branches, customer sites or other connections or couplings of the utility grid 10…”; also, see [0106] and [0111]-[0112] “The data processing system 202 (e.g., via grid controller 218) can generate one or more commands to control one or more of the components to achieve the desired outcome for the utility grid 100. The data processing system 202 can compare the value for the signal of the utility grid 100 predicted by the circuit optimizer 216 via the machine learning model with a threshold, and generate the command to control the component on the utility grid 100 based on the comparison (e.g., adjust tap setting to increase or decrease voltage level).”; also, see [0113] and [0115]; also, see [0073] “…Examples of commands 226 can include a command to increase or decrease a tap setting of a voltage regulating transformer 106a on the primary circuit 112, or change a setpoint on the primary circuit 112 or secondary circuit 116…”; also, see [0106] “…Thus, the circuit optimizer 216, using the model, can determine a signal at the substation that facilitates VVO, and provide the value to the grid controller 218 to generate one or more commands to instruct one or more devices on the utility grid 100 to realize the predicted value to facilitate VVO”; also, see [0107], also, see page 19 claim 11; also, see [0117] “…For example, the report or notification can indicate that a device on the utility grid 100 is nearing failure or has failed…”; and [0133-0135]), by modifying a control setpoint of a voltage regulating transformer or by either modifying a control setpoint of the voltage regulating transformer to maintain operation within nominal parameters or by disconnecting a protection device to isolate the component predicted to fail (see [0112] “The data processing system 202 can compare the value for the signal of the utility grid 100 predicted by the circuit optimizer 216 via the machine learning model with a threshold, and generate the command to control the component on the utility grid 100 based on the comparison (e.g., adjust tap setting to increase or decrease voltage level).”; also, see [0117] “…can automatically disable the failed or failing device using a control command. Types of control commands can include, for example, activating or deactivating a device, changing a parameter of the device, setting a parameter of the device, adjusting operation of the device, disabling or turning off a device…”; also, see [0111-0112] and [0133-0138]).
While Spalt teaches training model with historical data associated with an operational state (see [0010], [0019], [0040], [0094] and [0101]), and using a supervised learning machine which by definition uses labeled data, Spalt does not explicitly teach the training dataset including labeled instances of at least one of a component failure event or a non-nominal operational state occurring at an identified component of the power system.
However, McElhinney teaches a system and method comprising a model trained with historical training dataset, wherein the historical training dataset including labeled instances of at least one of a component failure event or a non-nominal operational state occurring at an identified component of the power system (see [0127-0131] “…In practice, identifying past occurrences of the given failure may involve the data science system 404 identifying the type of operating data, such as abnormal-condition data, that indicates the given failure… After identifying the type of operating data that indicates the given failure, the data science system 404 may identify the past occurrences of the given failure in a number of manners. For instance, the data science system 404 may locate, from historical operating data stored in the databases 406, abnormal-condition data corresponding to the indicators associated with the given failure… At block 506, the data science system 404 may identify a respective set of operating data that is associated with each identified past occurrence of the given failure…”; also, see [0139] “Continuing in the particular example of defining a failure model based on a response variable, the data science system 404 may train the failure model with the historical operating data identified at block 506 and the generated response variable. Based on this training process, the data science system 404 may then define the failure model that receives as inputs various sensor data…”).
Therefore, it would have been obvious to one of ordinary skilled in the art before effective filing date of the claimed invention to which said subject matter pertains to have modified Spalt’s invention to include training a model with historical training dataset, wherein the historical training dataset including labeled instances of at least one of a component failure event or a non-nominal operational state occurring at an identified component of the power system as taught by McElhinney in order to predict and output anomalous operational states indicative of predicted future failure or non-nominal operations of the asset (see [0117] “…Based on the defined set of failures, the data science system 404 may take steps to define a model for predicting a likelihood of any of the failures occurring within a given timeframe in the future (e.g., the next two weeks).”; also, see [0118-0119]).
While Spalt teaches a supervised machine learning model wherein supervised machine models are widely used with labeled data, Spalt-McElhinney does not explicitly teach the training dataset including labeled instances of at least one of a component failure event or a non-nominal operational state occurring at an identified component of the power system (labeled data has been interpreted in the BRI as data (raw data) with correlated/tagged or associated information, which is different than unlabeled data representing raw data).
However, Gundel teaches a system and method comprising training a model with a training dataset, the training dataset including labeled instances of at least one of a component failure event or a non-nominal operational state occurring at an identified component of the power system (see [0004] “a computing system may implement supervised training to train the model(s) with a set of training instances, where each training instance includes sensor data associated with known (i.e., previously identified/labeled) respective failure event…”; see [0077] “supervised learning”; also, see [0129] “…The training data may, for example, comprise known (i.e., previously identified, also referred to as “labeled”) failure events and associated sensed data…”).
Therefore, it would have been obvious to one of ordinary skilled in the art before effective filing date of the claimed invention to which said subject matter pertains to have modified Spalt-McElhinney’s combination as taught above to include training a model with a training dataset including labeled instances of at least one of a component failure event or a non-nominal operational state occurring at an identified component of a power system as taught by Gundel in order to achieve more accurate prediction the health status of an article of equipment when applying the model to subsequent sensor data and alter the operation of a component before the failure occurs (see [0004-0005] “…achieving, in some examples, more accurate prediction the health status of an article of equipment when applying the model to subsequent sensor data… In this way, as the system receives sensor data from articles of electrical equipment operating in a work environment, the system may more quickly and accurately identify potential failure events that may affect the distribution of power throughout the power grid, or worker and/or civilian safety, to name only a few examples. By determining a health status of electrical equipment and predicting failure events before they occur, a computing system may proactively and preemptively generate notifications and/or alter the operation of a power grid before a failure event occurs”).
As per claim 2, Spalt-McElhinney-Gundel teaches the system of claim 1, Spalt further teaches wherein the sensor includes at least one of
voltage/current meters, current transformers, potential transformers, or transducers for determining voltage, current, load, or temperature;
cameras for determining visual characteristics; and
microphones for determining acoustic characteristics (see [0045] “…potential transformer…”; also, [0054] “Metering devices 118a-118n can process and sample the voltage signals such that the sampled voltage signals are sampled as a time series (e.g., uniform time series free of spectral aliases or non-uniform time series); also, see [0068] voltage and current measurements; [0070]); and
As per claim 3, Spalt-McElhinney-Gundel teaches the system of claim 1, Spalt further teaches wherein the power system model is a trained power system model trained using a training dataset including historical data of the power system (see [0010] “The method can include the data processing system training the machine learning model using a historical input matrix generated from historical signals and historical statistical metrics. The data processing system can apply one or more weights to a dimension of the historical input matrix to adjust an impact values of the dimension have on the machine learning model”; also, see [0019], [0040], and [0101]).
As per claim 5, Spalt-McElhinney-Gundel teaches the system of claim 1, Spalt further teaches the processor further configured to receive real-time data from at least one real-time data source and apply the power system model to the received real-time data to determine at least one other operational state of the component (see [0042] “The data processing system can receive or ingest real-time data from utility grid devices and prior commands issued to these devices. The data processing system can predict the change in the utility grid conditions for a given set of device changes”, the change in the utility grid conditions for a given set of device changes refers to at least one other operational state of the component; also, see [0075] and [0111], [0125] and [0126] “…. A data processing system can receive or identify signals via a real-time data feed or stream. Signals can refer to or include utility grid data, such as characteristics of electricity, device settings, device operation modes, or weather conditions. Signals can be obtained from various sources, such as from different devices at different locations on the utility grid, or an external server or other data sources.”; also, see [0133] “predict an output value”; also, see [0110] islanding is at least one other operational state of the component that is predicted based on the real time data; also, see [0117] “For example, the report or notification can indicate that a device on the utility grid 100 is nearing failure or has failed”, thus, another state is a device nearly failing or has failed; also, see [0138] “…including real-time signals and other control commands, and continue predicting the behavior of the utility grid…”).
As per claim 7, Spalt-McElhinney-Gundel teaches the system of claim 5, Spalt further teaches the control circuitry further configured to effect a change in the at least one of the component and another component of the power system in response to the at least one other operational state (see [0042] “The data processing system can receive or ingest real-time data from utility grid devices and prior commands issued to these devices. The data processing system can predict the change in the utility grid conditions for a given set of device changes”, the change in the utility grid conditions for a given set of device changes refers to at least one other operational state of the component; also, see [0075] and [0111] “…can determine the anomalous behavior may be the result of a faulty device on the utility grid 100, or other fault, failure or attack on the utility grid…, [0125] and [0126] “…. A data processing system can receive or identify signals via a real-time data feed or stream. Signals can refer to or include utility grid data, such as characteristics of electricity, device settings, device operation modes, or weather conditions. Signals can be obtained from various sources, such as from different devices at different locations on the utility grid, or an external server or other data sources.”; also, see [0133] “predict an output value”; also, see [0110] islanding is at least one other operational state of the component that is predicted based on the real time data; also, see [0117] “For example, the report or notification can indicate that a device on the utility grid 100 is nearing failure or has failed”, thus, another state is a device nearly failing or has failed; also, see [0138] “…including real-time signals and other control commands, and continue predicting the behavior of the utility grid…”; see [0106] and [0112] “The data processing system 202 (e.g., via grid controller 218) can generate one or more commands to control one or more of the components to achieve the desired outcome for the utility grid 100. The data processing system 202 can compare the value for the signal of the utility grid 100 predicted by the circuit optimizer 216 via the machine learning model with a threshold, and generate the command to control the component on the utility grid 100 based on the comparison (e.g., adjust tap setting to increase or decrease voltage level).”; also, see [0113]. Thus, the grid components are controlled/adjusted based on several states detected including islanding, change in the utility grid conditions for a given set of device changes, the anomalous behavior as a result of a faulty device on the utility grid 100, or other fault, failure or attack on the utility grid, near failure or failing or failed device (see [0117]).
As per claim 8, Spalt-McElhinney-Gundel teaches the system of claim 1, Spalt further teaches wherein the control circuitry includes circuit breakers (see [0069] circuit breaker), switchgear (see [0069] “…switch, breakers, fuse…” these are examples of switchgear), reclosers (see [0069] recloser), disconnects (see [0069] “…switch, breakers, fuse…” these are examples of disconnect devices; also, see [0117]), interrupters (see [0069] “…switch, breakers, fuse…” these are examples of interrupters), tap changers (see [0015” tap changer”; also, see [0045], [0050], and [0055]), circuit switchers (see [0069] “…switch…”), and switches (see [0069] “…switch…”).
As per claim 9, Spalt teaches a method (see [0002], [0005] and [0125]) comprising:
receiving, from a sensor designed to determine at least one performance characteristic of a first component of a plurality of components of a power system, the at least one performance characteristic (see Fig. 1 sensing modules 118a..n or 120a sense grid characteristics and transmit the measured grid characteristics; also, see [0045] “monitoring devices 118a-118n that can be coupled through optional potential transformers 120a-120n to secondary utilization circuits 116. The monitoring or metering devices 118a-118n can detect (e.g., continuously, periodically, based on a time interval, responsive to an event or trigger) measurements and continuous voltage signals of electricity supplied to one or more electrical devices 119 connected to circuit 112 or 116 from a power source 101 coupled to bus 102”; also, see [0053],[0054]);
applying, using a processor (see Fig. 2 processing system 202, including a processing module optimizer 216 and model 214 and controller 218; and see [0014] “The system can include one or more processors. The one or more processors can identify a first plurality of signals detected by one or more devices located at one or more portions of a utility grid”; also, see [0059-0060] and [0061]), a power system model to the at least one characteristic to determine at least one operational state, (see Fig. 2 model 224 and 214; also, see [0078] “…using the received utility grid data to generate an input matrix for input into the machine learning model generator 214 or the machine learning model…”; also, see [0104] and [0105] “. For example, the circuit optimizer 216 can be configured to facilitate Volt/VAR optimization and voltage management; system-state prediction,…; fault, location, isolation, and system restoration; anti-islanding detection; also, see [0107] and [0108] “The data processing system 202 can determine when faults may occur on the utility grid 100, such as due to excess power usage during peak energy consumption (e.g., brownouts due to increased air conditioning use, or increased electric vehicle charging). The data processing system 202 can detect that such a fault is imminent based on detecting a trend in the input matrix using the machine learning model, and pre-emptively adjust signals of the utility grid 100 by generating commands to devices on the utility grid 100”; also, see [0111] “…If the real-time measured do not match the predicted signals, then the data processing system 202 can determine that the behavior of the utility grid 100 may be anomalous. The data processing system 202 can determine the anomalous behavior may be the result of a faulty device on the utility grid 100, or other fault, failure or attack on the utility grid 100…”; also, see [0109] and [0110] “…The data processing system 202, using the machine learning model and an input matrix, can detect when such islanding may occur based on predicting output values for signals such as frequency and voltage… whether to classify the predicted output variables or signals as an instance of islanding…”; also, see [0103], [0105], [0117], [0122]; the operational state of a generator or inverter such as its frequency and voltage or islanding is determined and outputted by the processor by using the model), the power system model corresponding to the power system trained, using a training dataset comprising historical data associated with at least one historical operational state of the power system, t (see [0010] “The method can include the data processing system training the machine learning model using a historical input matrix generated from historical signals and historical statistical metrics. The data processing system can apply one or more weights to a dimension of the historical input matrix to adjust an impact values of the dimension have on the machine learning model”; also, see [0019], [0040], [0094] and [0101]), to detect anomalous operational states (see [0105] “0105] … For example, the circuit optimizer 216 can be configured to facilitate Volt/VAR optimization and voltage management; system-state prediction,…; fault, location, isolation, and system restoration; anti-islanding detection; also, see [0106] “…The circuit optimizer 216 can also account for general signals, such as time of day, weather, or transmission resources. Using the machine learning model, the circuit optimizer 216 can determine whether the predicted output for a signal falls within a desired range…”; also, see [0108] and [0111] “The data processing system 202 can use the machine learning model to verify behavior of the utility grid 100…If the real-time measured do not match the predicted signals, then the data processing system 202 can determine that the behavior of the utility grid 100 may be anomalous. The data processing system 202 can determine the anomalous behavior may be the result of a faulty device on the utility grid 100, or other fault, failure or attack on the utility grid 100”) and outputting a prediction indicative of a predicted future non-nominal operation for a respective component (see [0012] and see [0102] and [0106] “…The circuit optimizer 216 can also account for general signals, such as time of day, weather, or transmission resources. Using the machine learning model, the circuit optimizer 216 can determine whether the predicted output for a signal falls within a desired range….”; also, see [0110] “The data processing system 202 can perform anti-islanding detection, which can refer to a condition in which a distributed generator (e.g., a gas power generator or solar inverter device on the utility grid 100) continues to power a location even though electrical grid power is no longer present…. The data processing system 202, using the machine learning model and an input matrix, can detect when such islanding may occur based on predicting output values for signals such as frequency and voltage… For example, the data processing system 202, using the machine learning model, can include an output signal in the output matrix that corresponds to a classification of islanding (e.g., binary classification indicating whether islanding is present, or not present, or a likelihood that islanding is present, or a confidence score associated with the binary classification). The data processing system 202 can predict islanding based on predicted output signals, or an output matrix generated via the machine learning model can include an islanding classification signal or variable”; also, see [0041] “The data processing system, using the input matrix and machine learning model generator, can predict a corresponding output matrix. The model can predict a new output for a future time period, or predict an output value for a variable for which data does not exist in a current time period”; also, see [0108]), the prediction identifying the respective component predicted to experience the future non- nominal operation or component failure (see [0053], [0106], [0110], [0112], [0119] “…. Utility grid data can refer to, or include, timestamps and signals associated with various devices on the utility grid. The utility grid data can include an indication or identifier of the device from which the measurement is obtained…”, conditions of specific components of the grid are identified); and
directing, a control circuitry (see Fig. 2 grid controller 218 connected to control circuitry to effect changes in the grid; see [0072] and [0106]) operatively connected to at least one of the first component and a second component of the plurality of components, to adjust an operational parameter of the respective component and another component of the power system based on the prediction to mitigate a detected or predicted non-nominal operation (see [0073] “Commands 226 data structure can include control commands to control devices on the utility grid 100, alerts, notifications, reports, instructions, or other actions that can be performed by the data processing system 202 responsive to, or based on, utility grid data 222. Examples of commands 226 can include a command to increase or decrease a tap setting of a voltage regulating transformer 106a on the primary circuit 112, or change a setpoint on the primary circuit 112 or secondary circuit 116. Additional actions or determinations can include determining a topology of a utility grid 100 (e.g., main branches, secondary branches, customer sites or other connections or couplings of the utility grid 10…”; also, see [0106] and [0111]-[0112] “The data processing system 202 (e.g., via grid controller 218) can generate one or more commands to control one or more of the components to achieve the desired outcome for the utility grid 100. The data processing system 202 can compare the value for the signal of the utility grid 100 predicted by the circuit optimizer 216 via the machine learning model with a threshold, and generate the command to control the component on the utility grid 100 based on the comparison (e.g., adjust tap setting to increase or decrease voltage level).”; also, see [0113] and [0115]; also, see [0073] “…Examples of commands 226 can include a command to increase or decrease a tap setting of a voltage regulating transformer 106a on the primary circuit 112, or change a setpoint on the primary circuit 112 or secondary circuit 116…”; also, see [0106] “…Thus, the circuit optimizer 216, using the model, can determine a signal at the substation that facilitates VVO, and provide the value to the grid controller 218 to generate one or more commands to instruct one or more devices on the utility grid 100 to realize the predicted value to facilitate VVO”; also, see [0107], also, see page 19 claim 11; also, see [0133-0135]), by modifying a control setpoint of a voltage regulating transformer or by activating a protection relay to prevent component failure (see [0112] “The data processing system 202 can compare the value for the signal of the utility grid 100 predicted by the circuit optimizer 216 via the machine learning model with a threshold, and generate the command to control the component on the utility grid 100 based on the comparison (e.g., adjust tap setting to increase or decrease voltage level).”), or by either modifying a control setpoint of the voltage regulating transformer to maintain operation within nominal parameters or by disconnecting a protection device to isolate the component predicted to fail (see [0112] “The data processing system 202 can compare the value for the signal of the utility grid 100 predicted by the circuit optimizer 216 via the machine learning model with a threshold, and generate the command to control the component on the utility grid 100 based on the comparison (e.g., adjust tap setting to increase or decrease voltage level).”; also, see [0117] “…can automatically disable the failed or failing device using a control command. Types of control commands can include, for example, activating or deactivating a device, changing a parameter of the device, setting a parameter of the device, adjusting operation of the device, disabling or turning off a device…”; also, see [0111-0112] and [0133-0138]).
W While Spalt teaches training model with historical data associated with an operational state (see [0010], [0019], [0040], [0094] and [0101]), and using a supervised learning machine which by definition uses labeled data, Spalt does not explicitly teach the training dataset including labeled instances of at least one of a component failure event or a non-nominal operational state occurring at an identified component of the power system.
However, McElhinney teaches a system and method comprising a model trained with historical training dataset, wherein the historical training dataset including labeled instances of at least one of a component failure event or a non-nominal operational state occurring at an identified component of the power system (see [0127-0131] “…In practice, identifying past occurrences of the given failure may involve the data science system 404 identifying the type of operating data, such as abnormal-condition data, that indicates the given failure… After identifying the type of operating data that indicates the given failure, the data science system 404 may identify the past occurrences of the given failure in a number of manners. For instance, the data science system 404 may locate, from historical operating data stored in the databases 406, abnormal-condition data corresponding to the indicators associated with the given failure… At block 506, the data science system 404 may identify a respective set of operating data that is associated with each identified past occurrence of the given failure…”; also, see [0139] “Continuing in the particular example of defining a failure model based on a response variable, the data science system 404 may train the failure model with the historical operating data identified at block 506 and the generated response variable. Based on this training process, the data science system 404 may then define the failure model that receives as inputs various sensor data…”).
Therefore, it would have been obvious to one of ordinary skilled in the art before effective filing date of the claimed invention to which said subject matter pertains to have modified Spalt’s invention to include training a model with historical training dataset, wherein the historical training dataset including labeled instances of at least one of a component failure event or a non-nominal operational state occurring at an identified component of the power system as taught by McElhinney in order to predict and output anomalous operational states indicative of predicted future failure or non-nominal operations of the asset (see [0117] “…Based on the defined set of failures, the data science system 404 may take steps to define a model for predicting a likelihood of any of the failures occurring within a given timeframe in the future (e.g., the next two weeks).”; also, see [0118-0119]).
While Spalt teaches a supervised machine learning model wherein supervised machine models are widely used with labeled data, Spalt-McElhinney does not explicitly teach the training dataset including labeled instances of at least one of a component failure event or a non-nominal operational state occurring at an identified component of the power system (labeled data has been interpreted in the BRI as data (raw data) with correlated/tagged or associated information, which is different than unlabeled data representing raw data).
However, Gundel teaches a system and method comprising training a model with a training dataset, the training dataset including labeled instances of at least one of a component failure event or a non-nominal operational state occurring at an identified component of the power system (see [0004] “a computing system may implement supervised training to train the model(s) with a set of training instances, where each training instance includes sensor data associated with known (i.e., previously identified/labeled) respective failure event…”; see [0077] “supervised learning”; also, see [0129] “…The training data may, for example, comprise known (i.e., previously identified, also referred to as “labeled”) failure events and associated sensed data…”).
Therefore, it would have been obvious to one of ordinary skilled in the art before effective filing date of the claimed invention to which said subject matter pertains to have modified Spalt-McElhinney’s combination as taught above to include training a model with a training dataset including labeled instances of at least one of a component failure event or a non-nominal operational state occurring at an identified component of a power system as taught by Gundel in order to achieve more accurate prediction the health status of an article of equipment when applying the model to subsequent sensor data and alter the operation of a component before the failure occurs (see [0004-0005] “…achieving, in some examples, more accurate prediction the health status of an article of equipment when applying the model to subsequent sensor data… In this way, as the system receives sensor data from articles of electrical equipment operating in a work environment, the system may more quickly and accurately identify potential failure events that may affect the distribution of power throughout the power grid, or worker and/or civilian safety, to name only a few examples. By determining a health status of electrical equipment and predicting failure events before they occur, a computing system may proactively and preemptively generate notifications and/or alter the operation of a power grid before a failure event occurs”).
As to claim 10, this claim is the method claim corresponding to the system claim 2 and is rejected for the same reasons mutatis mutandis.
As to claim 11, this claim is the method claim corresponding to the system claim 8 and is rejected for the same reasons mutatis mutandis.
As per claim 13, Spalt- McElhinney-Gundel teaches the method of claim 9, further comprising: Spalt further teaches receiving, from a real-time data source, real-time data associated with at least one of the first component, the second component, and a third component of the plurality of components (see first component 101/power sources, second components 104, third components/loads 119,…more components include vehicles (0006), voltage regulators, inverters, circuit meters, and so on [0042] “The data processing system can receive or ingest real-time data from utility grid devices and prior commands issued to these devices. The data processing system can predict the change in the utility grid conditions for a given set of device changes”, the change in the utility grid conditions for a given set of device changes refers to at least one other operational state of the component; also, see [0075] and [0111], [0125] and [0126] “…. A data processing system can receive or identify signals via a real-time data feed or stream. Signals can refer to or include utility grid data, such as characteristics of electricity, device settings, device operation modes, or weather conditions. Signals can be obtained from various sources, such as from different devices at different locations on the utility grid, or an external server or other data sources.”; also, see [0133] “predict an output value”; also, see [0110] islanding is at least one other operational state of the component that is predicted based on the real time data; also, see [0117] “For example, the report or notification can indicate that a device on the utility grid 100 is nearing failure or has failed”, thus, another state is a device nearly failing or has failed; also, see [0138] “…including real-time signals and other control commands, and continue predicting the behavior of the utility grid…”);
applying, using the processor, the power system model to the received real-time data to determine at least one other operational state (see [0042] “The data processing system can receive or ingest real-time data from utility grid devices and prior commands issued to these devices. The data processing system can predict the change in the utility grid conditions for a given set of device changes”, the change in the utility grid conditions for a given set of device changes refers to at least one other operational state of the component; also, see [0075] and [0111], [0125] and [0126] “…. A data processing system can receive or identify signals via a real-time data feed or stream. Signals can refer to or include utility grid data, such as characteristics of electricity, device settings, device operation modes, or weather conditions. Signals can be obtained from various sources, such as from different devices at different locations on the utility grid, or an external server or other data sources.”; also, see [0133] “predict an output value”; also, see [0110] islanding is at least one other operational state of the component that is predicted based on the real time data; also, see [0117] “For example, the report or notification can indicate that a device on the utility grid 100 is nearing failure or has failed”, thus, another state is a device nearly failing or has failed; also, see [0138] “…including real-time signals and other control commands, and continue predicting the behavior of the utility grid…”);
and directing the control circuitry to effect a change in the at least one of the first, second, and third components in response to the determined at least one other operational state (see [0106] and [0112] “The data processing system 202 (e.g., via grid controller 218) can generate one or more commands to control one or more of the components to achieve the desired outcome for the utility grid 100. The data processing system 202 can compare the value for the signal of the utility grid 100 predicted by the circuit optimizer 216 via the machine learning model with a threshold, and generate the command to control the component on the utility grid 100 based on the comparison (e.g., adjust tap setting to increase or decrease voltage level).”; also, see [0113]).
As to claim 16, this claim is the non-transitory computer readable storage medium claim corresponding to the system claim 1 and/or method claim 9 and is rejected for the same reasons mutatis mutandis (see Spalt [0141] “The processes, systems and methods described herein can be implemented by the computing system 500 in response to the processor 510 executing an arrangement of instructions contained in main memory 515. Such instructions can be read into main memory 515 from another computer-readable medium, such as the storage device 525. Execution of the arrangement of instructions contained in main memory 515 causes the computing system 500 to perform the illustrative processes described herein).
As to claim 18, this claim is the non-transitory computer readable storage medium claim corresponding to the method claim 13 and is rejected for the same reasons mutatis mutandis.
As to claim 20, this claim is the non-transitory computer readable storage medium claim corresponding to the method claim 15 and is rejected for the same reasons mutatis mutandis.
Claims 4, 12, and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Spalt et al (US 20210373518) in view of McElhinney et al (US 20160155315) and Gundel et al (US 20210373063) as applied to claims 1, 9, 16, respectively above, and further in view of Koval et al (US 11734704).
As per claim 4, Spalt-McElhinney-Gundel teaches the system of claim 1, while Spalt teaches collecting voltage/current measurements and temperature measurements of one or more components of a poser system for determining a condition/state (see claim 2), Spalt does not explicitly teach wherein the at least one operational state is an overload, an underload, a short circuit, and overheating.
However, Koval teaches a system comprising determining at least one operational state of component of a power system (see the Abstract “The present disclosure is directed to a machine learning system for use with a power distribution system. The machine learning system includes a data library, a machine learning module, and an action module. The data library stores a plurality of data samples, where at least a portion of the data samples are associated with one or more intelligent electronic devices (IEDs). The machine learning module processes data samples from the data library using at least one machine learning algorithm and outputs at least one recommendation and/or prediction based on the data samples received.”; also, see Col 54 lines 3-42 predicting a fault using a machine learning algorithm and performing an action correction), wherein the at least one operational state is an overload, an underload, a short circuit, and overheating (see Col 3 lines 57-60 “…the at least one type of fault includes at least one of transients, interruption of supply voltage or load current, undervoltage, overvoltage, waveform distortion, and/or voltage fluctuations”; also, see Col 62 lines 40-58 “…short circuit…”; also, see Col 63 lines 8-15 “…short circuit…”).
Therefore, it would have been obvious to one of ordinary skilled in the art before effective filing date of the claimed invention to which said subject matter pertains to have modified Spalt’s combination as taught above to include the determining at least one operational state of component of a power system, wherein the at least one operational state is an overload, an underload, a short circuit, and overheating as taught by Koval in order to perform at least one action based on a recommendation and/or prediction, wherein the action includes outputting at least one of a communication signal and/or at least one control signal to at least one client or at least one of the one or more IEDs to mitigate the operational state (see Col 2 lines 32-49 “…perform at least one action based on the recommendation and/or prediction, wherein the action includes outputting at least one of a communication signal and/or at least one control signal to at least one client or at least one of the one or more IEDs…”; also, see Col 3 lines 1-10; also, see Col 61 lines 23-41).
As to claim 12, this claim is the method claim corresponding to the system claim 4 and is rejected for the same reasons mutatis mutandis.
As per claim 17, Spalt-McElhinney-Gundel teaches the non-transitory computer-readable storage medium of claim 16, Spalt further teaches wherein the sensing module includes at least one of
voltage/current meters, current transformers, potential transformers, or transducers for determining voltage, current, load, or temperature;
cameras for determining visual characteristics; and
microphones for determining acoustic characteristics (see [0045] “…potential transformer…”; also, [0054] “Metering devices 118a-118n can process and sample the voltage signals such that the sampled voltage signals are sampled as a time series (e.g., uniform time series free of spectral aliases or non-uniform time series); also, see [0068] voltage and current measurements; [0070]);
wherein the control module includes circuit breakers (see [0069] circuit breaker), switchgear (see [0069] “…switch, breakers, fuse…” these are examples of switchgear), reclosers (see [0069] recloser), disconnects (see [0069] “…switch, breakers, fuse…” these are examples of disconnect devices; also, see [0117]), interrupters (see [0069] “…switch, breakers, fuse…” these are examples of interrupters), tap changers (see [0015” tap changer”; also, see [0045], [0050], and [0055]), circuit switchers (see [0069] “…switch…”), and switches (see [0069] “…switch…”); and
while Spalt teaches collecting voltage/current measurements and temperature measurements of one or more components of a poser system for determining a condition/state (see claim 2), Spalt does not explicitly teach wherein the at least one operational state is an overload, an underload, a short circuit, and overheating.
However, Koval teaches a system comprising determining at least one operational state of component of a power system (see the Abstract “The present disclosure is directed to a machine learning system for use with a power distribution system. The machine learning system includes a data library, a machine learning module, and an action module. The data library stores a plurality of data samples, where at least a portion of the data samples are associated with one or more intelligent electronic devices (IEDs). The machine learning module processes data samples from the data library using at least one machine learning algorithm and outputs at least one recommendation and/or prediction based on the data samples received.”; also, see Col 54 lines 3-42 predicting a fault using a machine learning algorithm and performing an action correction), wherein the at least one operational state is an overload, an underload, a short circuit, and overheating (see Col 3 lines 57-60 “…the at least one type of fault includes at least one of transients, interruption of supply voltage or load current, undervoltage, overvoltage, waveform distortion, and/or voltage fluctuations”; also, see Col 62 lines 40-58 “…short circuit…”; also, see Col 63 lines 8-15 “…short circuit…”).
Therefore, it would have been obvious to one of ordinary skilled in the art before effective filing date of the claimed invention to which said subject matter pertains to have modified Spalt’s combination as taught above to include the determining at least one operational state of component of a power system, wherein the at least one operational state is an overload, an underload, a short circuit, and overheating as taught by Koval in order to perform at least one action based on a recommendation and/or prediction, wherein the action includes outputting at least one of a communication signal and/or at least one control signal to at least one client or at least one of the one or more IEDs to mitigate the operational state (see Col 2 lines 32-49 “…perform at least one action based on the recommendation and/or prediction, wherein the action includes outputting at least one of a communication signal and/or at least one control signal to at least one client or at least one of the one or more IEDs…”; also, see Col 3 lines 1-10; also, see Col 61 lines 23-41).
Claims 6, 14, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Spalt et al (US 20210373518) in view of McElhinney et al (US 20160155315) and Gundel et al (US 20210373063) as applied to claims 5, 13, and 18, respectively above, and further in view of Prasanna et al (US 20220036199), Priyadarsini et al (US 20210382470) and Mestha et al (US 20180260561).
As per claim 6, Spalt- McElhinney-Gundel teaches the system of claim 5, Spalt further teaches wherein the at least one real-time data source includes;
weather and environmental data (see [0015] “The second plurality of signals can include at least one of current weather, forecasted weather…”; also, see [0071]);
However, Spalt does not explicitly teach wherein the at least one real-time data source includes at least one of real-time images, video, and sounds;
anecdotal/observational human reports;
and social media data (while the claim requires only of “one real-time data source”, the following references have been provided to teach the limitations).
However, Prasanna teaches a system for identifying states/anomalies of a component of a power system (see the Abstract) comprising at least one real-time data source for providing real time data, wherein the at least one real-time data source includes at least one of real-time images, video, and sounds(see [0054] “Field sensors can also include video data, e.g., from cameras. Field sensors can also include RF noise detectors that may detect cracked insulators”); weather and environmental data (see [0054] “…Field sensors can additionally measure environment and/or weather sensors that can measure temperature, relative humidity, and wind speed and direction…”).
Therefore, it would have been obvious to one of ordinary skilled in the art before effective filing date of the claimed invention to which said subject matter pertains to have modified Spalt’s combination as taught above to include at least one real-time data source for providing real time data, wherein the at least one real-time data source includes at least one of real-time images, video, and sounds; weather and environmental data as taught by Prasanna in order to detect one other operational state of a component of a power system (see [0048] and [0084] “…to determine the presence of an anomaly corresponding to foliage impingement, abnormal power flow loading, infrastructure failure, and predictive failure…”).
However, Spalt-McElhinney-Gundel-Prasanna still does not explicitly teach wherein the at least one real-time data source includes at least one of anecdotal/observational human reports; and social media data.
However, Priyadarsini teaches a system for predicting state of a plant comprising a least one real-time data source for providing real time data, wherein the at least one real-time data source includes at least one of anecdotal/observational human reports
(see Fig. 2 incident reports 246; also, see Fig. 3 320; also, see [0019] “…and field logbooks and incident reports (246)…”).
Therefore, it would have been obvious to one of ordinary skilled in the art before effective filing date of the claimed invention to which said subject matter pertains to have modified Spalt’s combination as taught above to include wherein the at least one real-time data source includes at least one of anecdotal/observational human reports and other data as taught by Priyadarsini in order to detect the state of a system plant using machine learning (see [0019] “…An embodiment processes and integrates multiple data sources to learn about the plant states and plant modes. FIG. 2 illustrates the primary data sources that are used to first train a model to determine a plant state …The primary data sources include sensor data (process parameter values) from plant process histories; …and engineering configuration data field logbooks (e.g., incident reports) from field operators.… log file data are integrated from a plant 210, an operation/control room 220, and field data 230. …field logbooks and incident reports (246). These data are first processed in the machine learning/training phase 250. This learning/training generates data mining and analysis modules/models 260; also, see [0031] and see claim 18 in page 9).
However, Spalt-McElhinney-Gundel-Prasanna-Priyadarsini still does not explicitly teach wherein the at least one real-time data source includes social media data.
However, Mestha teaches a system for determining a state of a component of an electrical component (see Fig. 1, Fig. 3 and Fig. 8 using a model to detect the state of one or more electrical components in a grid 332) comprising a least one real-time data source for providing real time data, wherein the at least one real-time data source includes social media data (see [0021] “According to some embodiments, the data source nodes 130 provide “heterogeneous” data. That is, the data may represent information from widely diverse areas, such as social media data, w… weather data (e.g., temperature data, National Oceanic and Atmospheric Administration (“NOAA”) information, etc.), IT inputs, market pricing, etc.”; also, see [0047], [0055], [0075] “…social media, weather data,…”).
Therefore, it would have been obvious to one of ordinary skilled in the art before effective filing date of the claimed invention to which said subject matter pertains to have modified Spalt’s combination as taught above to include a least one real-time data source for providing real time data, wherein the at least one real-time data source includes social media data as taught by Mestha in order to determine the state of component of a power system using machine learning (see [0022] and [0030] “At S240, the system may automatically transmit an abnormal alert signal (e.g., a notification message, etc.) based on results of the comparisons performed at S230. The abnormal state might be associated with, for example, an actuator attack, a controller attack, a data source node attack, a plant state attack, spoofing, physical damage, unit availability, a unit trip, a loss of unit life, asset damage requiring at least one new part, a stealthy attack not detectable by alarms, a load alternating attack, a topology change attack, a stability compromise attack, and/or a frequency compromise attack. According to some embodiments, one or more response actions may be performed when an abnormal alert signal is transmitted…”; also, see [0032]) and allow an operator to restore power grid to normal operations and/or avoid damages to the system (see [0025] and [0030], [0032].
As to claim 14, this claim is the method claim corresponding to the system claim 6 and is rejected for the same reasons mutatis mutandis.
As to claim 19, this claim is the non-transitory computer readable storage medium claim corresponding to the system claim 6 and is rejected for the same reasons mutatis mutandis.
Claim 15 is rejected under 35 U.S.C. 103 as being unpatentable over Spalt et al (US 20210373518) in view of McElhinney et al (US 20160155315) and Gundel et al (US 20210373063) as applied to claim 9 above, and further in view of Smiley et al (US 20140365271).
As per claim 15, Spalt-McElhinney-Gundel teaches the method of claim 9, Spalt further teaches wherein the power system model is a trained power system model (see [0072] “…The model 224 can include a voltage power flow model generated, constructed or trained by a machine learning model generator 214…”; also, see Fig. 2 214 is a machine learning model; also, see [0101] “…Once trained, the model can be used to predict new outputs using a new input matrix.”), the method further comprising:
retrieving, from a database, a training dataset including historical data associated with at least one historical operational state of the power system (see [0101] “The machine learning model generator 214 can be configured to generate and train the initial model based on initial training data sets that may be weighted. The machine learning model generator 214 can use a historical input matrix (e.g., generated from historical signals and historical statistical metrics determined from the historical signals) to train the machine learning model. Once trained, the model can be used to predict new outputs using a new input matrix. );
applying, using the processor, the power system model to the historical data (see [0010], [0019], [0094] and [0101]);
determining, using the processor, a predicted operational state of the power system based on the historical data (see [0101] “…Once trained, the model can be used to predict new outputs using a new input matrix...”; also, see [0103]);
and updating, using the processor, the power system model (see [0034] “These models can be updated whenever new input data is available, which can produce a dynamically accurate prediction during state changes in the circuit…”; also, see [0043] and [0116])
However, Spalt does not explicitly teach updating the power system model based on a deviation between the predicted operational state and the at least one historical operational state to create the trained power system model.
However, Smiley teaches a system and non-transitory computer readable storage medium (see [0131]) comprising teaches updating the power system model based on a deviation between a predicted operational state and at least one historical operational state to create the trained power system model (see [0027] “..industrial asset is used herein to describe a piece of equipment, element thereof, and/or a group of equipment logically and/or physically assembled together to form a production unit. Examples of such industrial assets may include a transformer, bushing, circuit breaker, substation (e.g., comprising one or more transformers and/or one or more circuit breakers), transmission tower, power generator, etc.…”; also, [0002] “to analyze data pertaining to an industrial asset to generate a health profile that describes past, present, or expected future conditions of the industrial asset and/or likely cause(s) of the conditions”; the health profile suggests at least one historical operational state or past conditions and also predicted operational states; see [0007] “…to update the model based upon a comparison of the health profile with data generated during the prediction period and indicative of the industrial asset”; also, see Figs. 3-5 show predicted operational state during predicted period 310 and at least one historical operational state during assessment period 308 of assets are determined; see [0031]-[0033]; also, see Fig. 5 and see [0065] “The assessment period 308 describes a time window of data samplings used to generate the health profile and the prediction period 310 describes a forecast window of interest. Thus, based upon the data acquired during the assessment period 308, one or more predictions or forecasts can be made about a state or health of the industrial asset during the prediction period 310…”; health profile; also, see Fig. 8 and [0088]-[0089] “At 806 in the example method 800, the data generated during the prediction period and/or other acquired data is compared to a health profile of the industrial asset to identify discrepancies, if any, between the information in the health profile and actual events that occurred during the prediction period… If a maintenance report generated during the prediction period indicates that the insulation was degraded but operating reports do not indicate that the loading on the transformer was reduced after the identification of the degraded insulation (e.g., despite redundant transformers that a portion of the load could have been moved to), the likelihood that overloading was the cause of the degraded insulation may be reduced (e.g., unless during the analysis it can be determined that the operators were likely to have been either unaware of the degradation or unaware that load reduction might be a useful mitigation technique to slow the deterioration of insulation). As still another example, data generated during the prediction period and yielded from an enterprise system may be analyzed to determine whether one or more predicted impacts of a condition were felt…”; also, see [0126]; also, see page 16 claim 10).
Therefore, it would have been obvious to one of ordinary skilled in the art before effective filing date of the claimed invention to which said subject matter pertains to have modified Spalt’s combination as taught above to include updating the power system model based on a deviation between the predicted operational state and the at least one historical operational state to create the trained power system model.
as taught by Smiley in order to facilitate addressing operational and/or physical changes to an industrial asset (see [0084])) and improve the confidence of prediction of the model prediction (see [0096]).
Conclusion
The prior art made of record and not relied upon, as cited in PTO form 892, is considered pertinent to applicant's disclosure.
Currently cited references:
Zhu et al (US 20230203132) teaches supervised training is performed on the AI model, the training data that is used for training and that is in the training dataset includes a label (see 0057).
Rhoads et al (US 20160198245) teaches Sets of abnormal input may be labeled and used to train a classifier to discriminate particular classes of abnormal signatures (thus, training dataset of labeled failures).
Wages et al (US 20230213569) teaches historical change in impedance data may be labeled with the known cause of the change in impedance (e.g., tree at location XYZ) such that the machine learning model may further refine the cause of the change in impedance between two devices to a narrower location (e.g., the particular tree location)…The historical training data set may include label indicative of the known impact cause for each change in impedance (0061 and 0065). Thus, it teaches labeled data including a failure or non-nominal condition of an electrical component.
He et al (CN 110334740) teaches a training dataset including labeled instance of at least one of a component failure or non-nominal operational state (see the Abstract, see page 3 “…the training data set, and transmitting the type and position as , input to the deep convolutional neural network in training to obtain the power equipment detection neural network after training…”);
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.
Examiner respectfully requests, in response to this Office action, support be shown for language added to any original claims on amendment and any new claims. That is, indicate support for newly added claim language by specifically pointing to page(s) and line number(s) in the specification and/or drawing figure(s). This will assist Examiner in prosecuting the application.
When responding to this Office Action, Applicant is advised to clearly point out the patentable novelty which he or she thinks the claims present, in view of the state of the art disclosed by the references cited or the objections made. Applicant must also show how the amendments avoid or differentiate from such references or objections. See 37 CFR 1.111 (c).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to OLVIN LOPEZ ALVAREZ whose telephone number is (571) 270-7686 and fax (571) 270-8686. The examiner can normally be reached Monday thru Friday from 9:00 A.M. to 6:00 P.M.
If attempts to reach the examiner by telephone are unsuccessful, the examiner's supervisor, Robert Fennema, can be reached at (571) 272-2748. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from Patent Center. Status information for published applications may be obtained from Patent Center. Status information for unpublished applications is available through Patent Center for authorized users only. Should you have questions about access to Patent Center, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
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) Form at https://www.uspto.gov/patents/uspto-automated- interview-request-air-form.
/O. L./
Examiner, Art Unit 2117
/ROBERT E FENNEMA/Supervisory Patent Examiner, Art Unit 2117
Prosecution Insights