Combination of methods for change-point detection in operating of power generating equipment

The article discusses the issue of the reliability of the power generation system from the point of view of the cyber-physical control of the system. Companies that generate electricity must supply this resource without interruption and monitor the generation process to identify and correct all causes of possible malfunctions in the process. The authors present a hybrid method for detecting change-point in the operation of cyber-physical power generation systems based on data from the process of power generation by gas turbine plants, provided that they are in the «generation» operating mode. The hybrid approach to a problem is a sequence (or pipeline) of steps that improve the results of the basic approach using the n-sigma rule by comparing real generation data with a performance standard. The proposed hybrid method is based on the following methods: search for optimal parameters (the indicators of precision, recall and F1-measure of the developed method for selecting the optimal parameters were 0.7, 0.7778, 0.7369, respectively); identifying outliers; detecting change-point using heuristic rules. As methods for detecting outliers, the authors use the DBSCAN algorithm and the n-sigma rule. The hybrid method using the DBSCAN algorithm identified outliers without false positives compared to the baseline approach. Advanced heuristics for change-points detection allow cyber-physical system experts to quickly identify the cause of the change-point using information about the time of the failure and the sensors on which the failure occurs. Prompt identification of the change-point allows for more accurate and timely monitoring of the performance of individual units and the entire system as a whole, develop a strategy of actions for repairing equipment in the shortest possible time and with minimal intervention in the process (until the system reaches a critical state), which can significantly reduce costs for Maintenance. Application examples demonstrate the advantages of the proposed method for both synthetic and real data.

1. Glotov A.V., Cheremisinov S.V., Shcherbakov M.V. Ontologicheskaya model' risk-orientirovannogo upravleniya tekhnicheskim sostoyaniem tekhnologicheskogo oborudovaniya. Energiya Edinoi Seti. 2019;3(45):76-85.

2. Jin R., Deng X., Chen X., Zhu L., Zhang J. Dynamic quality-process model in consideration of equipment degradation. Journal of Quality Technology. 2019;51(3), 217-229. DOI:10.1080/00224065.2018.1541379

3. Kazakov I.D., Shcherbakova N.L., Brebels A., Shcherbakov M.V. Accelerometer Data Based Cyber-Physical System for Training Intensity Estimation. Cyber-Physical Systems: Advances in Design & Modelling. Studies in Systems, Decision and Control. 2020;259:325-335. DOI:https://doi.org/10.1007/978-3-030-32579-4_26

4. Shcherbakov M., Brebels A., Shcherbakova N., Kamaev V., Gerget O., Devyatykh D. Outlier detection and classification in sensor data streams for proactive decision support systems. Conference on Information Technologies in Business and Industry 2016, Journal of Physics: Conference Series. 2016;803(1). DOI:http://dx.doi.org/10.1088/1742-6596/803/1/012143

5. Shen X., Fu X., Zhou C. A Combined Algorithm for Cleaning Abnormal Data of Wind Turbine Power Curve Based on Change Point Grouping Algorithm and Quartile Algorithm. IEEE Transactions on Sustainable Energy. 2019;10(1):46-54. DOI:10.1109/tste.2018.2822682

6. Letzgus S. Change-point detection in wind turbine SCADA data for robust condition monitoring with normal behaviour models. Wind Energy Science Discussions. 2020. DOI:https://doi.org/10.5194/wes-2020-38

7. Han S., Qiao Y., Yan P., Yan J., Liu Y., Li L. Wind turbine power curve modeling based on interval extreme probability density for the integration of renewable energies and electric vehicles. Renewable Energy. 2020;157:190-203. DOI:https://doi.org/10.1016/j.renene.2020.04.097

8. Celik M., Dadaser-Celik F., Dokuz A. S. Anomaly detection in temperature data using DBSCAN algorithm. 2011 International Symposium on Innovations in Intelligent Systems and Applications. 2011. DOI:https://doi.org/10.1109/INISTA.2011.5946052

9. Sheridan K., Puranik T.G., Mangortey E., Pinon-Fischer O.J., Kirby M., Mavris D.N. An Application of DBSCAN Clustering for Flight Anomaly Detection During the Approach Phase. AIAA Scitech 2020 Forum. 2020. DOI: https://doi.org/10.2514/6.2020-1851

10. Wang P., Govindarasu M. Anomaly Detection for Power System Generation Control based on Hierarchical DBSCAN. 2018 North American Power Symposium (NAPS). 2018. DOI://doi.org/10.1109/NAPS.2018.8600616