System and method for fault detection of components using information fusion technique
An example method comprises receiving historical sensor data of a first time period, the historical data including sensor data of a renewable energy asset, extracting features, performing a unsupervised anomaly detection technique on the historical sensor data to generate first labels associated with the historical sensor data, performing at least one dimensionality reduction technique to generate second labels, combining the first labels and the second labels to generate combined labels, generating one or more models based on supervised machine learning and the combined labels, receiving current sensor data of a second time period, the current sensor data including sensor data of the renewable energy asset, extracting features, applying the one or more models to the extracted features of the current sensor data to create a prediction of a future fault in the renewable energy asset, and generating a report including the prediction of the future fault in the energy asset.
1. A non-transitory computer readable medium comprising executable instructions, the executable instructions being executable by one or more processors to perform a method, the method comprising:
receiving historical sensor data of a first time period, the historical sensor data including sensor data from one or more sensors of a renewable energy asset;
extracting features from the historical sensor data;
performing a unsupervised anomaly detection technique on the historical sensor data to generate first labels associated with the historical sensor data;
performing at least one dimensionality reduction technique to generate second labels associated with the historical sensor data, the dimensionality reduction technique reducing a second number of dimensions of the extracted features when compared to a first number of dimensions of the extracted features analyzed using the unsupervised anomaly detection;
combining the first labels and the second labels to generate combined labels;
generating one or more models based on supervised machine learning and the combined labels;
receiving current sensor data of a second time period, the current sensor data including sensor data from at least a subset of the one or more sensors of the renewable energy asset;
extracting features from the current sensor data;
applying the one or more models to the extracted features of the current sensor data to create a prediction of a future fault in the renewable energy asset; and
generating a report including the prediction of the future fault in the renewable energy asset.
2. The non-transitory computer readable medium of claim 1 , wherein the unsupervised anomaly detection technique includes an isolation forest technique.
3. The non-transitory computer readable medium of claim 1 , wherein the at least one dimensionality reduction technique includes a principal component analysis (PCA) technique.
4. The non-transitory computer readable medium of claim 1 , wherein combining the first and second labels are combined in a complimentary manner.
5. The non-transitory computer readable medium of claim 1 , wherein generating one or more models based on supervised machine learning and the combined labels includes K-nearest neighbor algorithm or a neural network.
6. The non-transitory computer readable medium of claim 1 , wherein supervised machine learning includes k-means clustering.
7. The non-transitory computer readable medium of claim 1 , wherein performing the unsupervised anomaly detection technique on the historical sensor data to generate the first labels comprises generating an anomaly score based on an output of the unsupervised anomaly detection technique and comparing the anomaly score to a threshold to determine if at least one of the first labels should be generated.
8. The non-transitory computer readable medium of claim 1 , wherein performing the at least one dimensionality reduction technique to generate the second labels comprises generating a z-score based on an output of the at least one dimensionality reduction technique and comparing the z-score to a threshold to determine if at least one of the second labels should be generated.
9. The non-transitory computer readable medium of claim 1 , the method further comprising comparing a fault prediction against a criteria to determine significance of one or more predicted faults and generating an alert based on the comparison, the alert including generating a message identifying the one or more predicted faults.
10. A system, comprising
at least one processor; and
memory containing instructions, the instructions being executable by the at least one processor to:
receive historical sensor data of a first time period, the historical sensor data including sensor data from one or more sensors of a renewable energy asset;
extract features from the historical sensor data;
perform a unsupervised anomaly detection technique on the historical sensor data to generate first labels associated with the historical sensor data;
perform at least one dimensionality reduction technique to generate second labels associated with the historical sensor data, the dimensionality reduction technique reducing a second number of dimensions of the extracted features when compared to a first number of dimensions of the extracted features analyzed using the unsupervised anomaly detection;
combine the first labels and the second labels to generate combined labels;
generate one or more models based on supervised machine learning and the combined labels;
receive current sensor data of a second time period, the current sensor data including sensor data from at least a subset of the one or more sensors of the renewable energy asset;
extract features from the current sensor data;
apply the one or more models to the extracted features of the current sensor data to create a prediction of a future fault in the renewable energy asset; and
generate a report including the prediction of the future fault in the renewable energy asset.
11. The system of claim 10 , wherein the unsupervised anomaly detection technique includes an isolation forest technique.
12. The system of claim 10 , wherein the at least one dimensionality reduction technique includes a principal component analysis (PCA) technique.
13. The system of claim 10 , wherein combining the first and second labels are combined in a complimentary manner.
14. The system of claim 10 , wherein generating one or more models based on supervised machine learning and the combined labels includes K-nearest neighbor algorithm or a neural network.
15. The system of claim 10 , wherein supervised machine learning includes k-means clustering.
16. The system of claim 10 , wherein performing the unsupervised anomaly detection technique on the historical sensor data to generate the first labels comprises generating an anomaly score based on an output of the unsupervised anomaly detection technique and comparing the anomaly score to a threshold to determine if at least one of the first labels should be generated.
17. The system of claim 10 , wherein performing the at least one dimensionality reduction technique to generate the second labels comprises generating a z-score based on an output of the at least one dimensionality reduction technique and comparing the z-score to a threshold to determine if at least one of the second labels should be generated.
18. The system of claim 10 , the instructions being further executable by the at least one processor to compare a fault prediction against a criteria to determine significance of one or more predicted faults and generating an alert based on the comparison, the alert including generating a message identifying the one or more predicted faults.
19. A method comprising:
receiving historical sensor data of a first time period, the historical sensor data including sensor data from one or more sensors of a renewable energy asset;
extracting features from the historical sensor data;
performing a unsupervised anomaly detection technique on the historical sensor data to generate first labels associated with the historical sensor data;
performing at least one dimensionality reduction technique to generate second labels associated with the historical sensor data, the dimensionality reduction technique reducing a second number of dimensions of the extracted features when compared to a first number of dimensions of the extracted features analyzed using the unsupervised anomaly detection;
combining the first labels and the second labels to generate combined labels;
generating one or more models based on supervised machine learning and the combined labels;
receiving current sensor data of a second time period, the current sensor data including sensor data from at least a subset of the one or more sensors of the renewable energy asset;
extracting features from the current sensor data;
applying the one or more models to the extracted features of the current sensor data to create a prediction of a future fault in the renewable energy asset; and
generating a report including the prediction of the future fault in the renewable energy asset.