Signature identification for power system events
Briefly, embodiments are directed to a system, method, and article for identifying power system event signatures. Input measurement data may be received from one or more data sources relating to a power grid system. The input measurement data may comprise normal system operation measurement data and power system event measurement data. A processor may perform operations during an online application phase. During the online application phase, a feature matrix may be generated for the power system event measurement data and the at least one trained auto-associative model. The feature matrix for the power system event measurement data may be processed to determine power system event residuals. Also during the online application phase, the power system event signatures may be identified based on residual statistics for normal system operation measurement data residuals and on the power system event residuals.
1. A system to identify power system event signatures, the system comprising:
a receiver to receive input measurement data from one or more data sources relating to a power grid system, wherein the input measurement data comprises normal system operation measurement data and power system event measurement data; and
a processor to:
during an online application phase:
generate a first feature matrix for the power system event measurement data,
process at least one trained auto-associative model and the first feature matrix for the power system event measurement data to determine power system event residuals, and
identify the power system event signatures based on residual statistics for normal system operation measurement data residuals and the power system event residuals.
2. The system of claim 1 , wherein the processor is to further:
during an offline training phase:
generate a second feature matrix for the normal system operation measurement data,
train the at least one auto-associative model based on the second feature matrix for the normal system operation measurement data, and
determine the residual statistics for the normal system operation measurement data residuals, wherein the normal system operation measurement data residuals are determined by applying the second feature matrix for the normal system operation measurement data to the at least one trained auto-associative model.
3. The system of claim 1 , wherein the input measurement data comprises one or more of phasor measurement unit (PMU) data, Supervisory Control and Data Acquisition (SCADA) data, weather data, dissolved gas analysis (DGA) sensors, and/or partial discharge (PD) monitor sensor data.
4. The system of claim 1 , wherein extracted features of at least one of the first and second feature matrixes are associated with at least one of: (i) principal components, (ii) statistical features, (iii) time series analysis features, (iv) frequency domain features, (v) geographic or position based features, (vi) interaction features, (vii) logical features, (viii) deep learning features, and (ix) domain specific features.
5. The system of claim 1 , wherein the processor is to further identify one or more power system events based on a comparison of more of more extracted features and one or more of the power system event signatures.
6. The system of claim 5 , wherein the one or power system events comprise one or more of: an instrument pre-failure, a transformer health index, an instrument drifting, a loose connection, or a breaker mis-operation.
7. The system of claim 2 , wherein extracted features of at least one of the first and second feature matrixes are based on calculations made over a sliding window of time-series measurements of the input measurement data.
8. The system of claim 2 , further comprising pre-processing on the input measurement data prior to generating the first and second feature matrixes.
9. The system of claim 1 , wherein the auto-associative model comprises an autoencoder neural network.
10. The system of claim 1 , wherein the auto-associative model comprises a similarity-based model.
11. The system of claim 1 , wherein the residual statistics comprise at least one of: mean, standard deviation, kurtosis, skewness, mode, median, quartile, range, and interquartile range.
12. A method to identify power system event signatures, the method comprising:
receiving input measurement data from one or more data sources relating to a power grid system;
during an online application phase:
generating a first feature matrix for the power system event measurement data,
processing at least one trained auto-associative model and the first feature matrix for the power system event measurement data to determine power system event residuals, and
identifying the power system event signatures based on residual statistics for normal system operation measurement data residuals and the power system event residuals.
13. The method of claim 12 , further comprising:
during an offline training phase:
generating a second feature matrix for the normal system operation measurement data,
training the at least one auto-associative model based on the second feature matrix for the normal system operation measurement data, and
determining the residual statistics for the normal system operation measurement data residuals, wherein the normal system operation measurement data residuals are determined by applying the second feature matrix for the normal system operation measurement data to the at least one trained auto-associative model.
14. The method of 12 , wherein the input measurement data comprises one or more of phasor measurement unit (PMU) data, Supervisory Control and Data Acquisition (SCADA) data, weather data, dissolved gas analysis (DGA) sensors, and/or partial discharge (PD) monitor sensor data.
15. The method of 13 , wherein extracted features of at least one of the first and second feature matrixes are associated with at least one of: (i) principal components, (ii) statistical features, (iii) time series analysis features, (iv) frequency domain features, (v) geographic or position based features, (vi) interaction features, (vii) logical features, (viii) deep learning features, and (ix) domain specific features.
16. An article, comprising:
a non-transitory storage medium comprising machine-readable instructions executable by one or more processors to:
during an online application phase:
generate a first feature matrix for the power system event measurement data,
process at least one trained auto-associative model and the first feature matrix for power system event measurement data to determine power system event residuals, and
identify the power system event signatures based on residual statistics for normal system operation measurement data residuals and the power system event residuals.
17. The article of claim 16 , wherein the machine-readable instructions are further executable by the one or more processors to further:
during an offline training phase:
generate a second feature matrix for the normal system operation measurement data,
train the at least one auto-associative model based on the second feature matrix for the normal system operation measurement data, and
determine the residual statistics for the normal system operation measurement data residuals, wherein the normal system operation measurement data residuals are determined by applying the second feature matrix for the normal system operation measurement data to the at least one trained auto-associative model.
18. The article of claim 16 , wherein the input measurement data comprises one or more of phasor measurement unit (PMU) data, Supervisory Control and Data Acquisition (SCADA) data, weather data, dissolved gas analysis (DGA) sensors, and/or partial discharge (PD) monitor sensor data.
19. The article of claim 16 , wherein the power system event comprises one or more of: an instrument pre-failure, a transformer health index, an instrument drifting, a loose connection, or a breaker mis-operation.
20. The article of claim 17 , wherein extracted features of at least one of the first and second feature matrixes are based on calculations made over a sliding window of time-series measurements of the input measurement data.