Data Analytics Platform for Power Equipment Intelligent Lifecycle Management

摘要

The paper presents a model of the data analytics platform for obtaining reliable estimates of the functional state of the power network equipment, aimed at development of effective maintenance and repair programs, based on technologies of Knowledge Discovery in Databases. In the proposed system Data Mining is carried out by gradient boosting of decision trees. Within the framework of the presented research, methodological, mathematical and algorithmic bases of the intelligent data analytics platform were developed. Validation of the proposed model is based on technical diagnostics data covering a period from 2005 to 2017 and providing functional state estimates of real power transmission network facilities of a regional power system. The developed system makes it possible to identify and generalize new correlation links and patterns of main technical parameters variation, including operation modes, functional states, and composition of technological fluids of electrical network equipment and to create new intelligent approach to its functional state assessment, providing highly-accurate and reliable decision making for the stakeholders. In this system, the results of technical state assessment are obtained using probabilistic approach to enable further analysis of technical and technological risks and, finally, to develop efficiently-scheduled maintenance and repair strategies of the grid companies. The system was probated in Jupyter programming environment in Python language using xgboost optimized distributed gradient boosting library and showed an acceptable average accuracy in equipment state identification: 91% of generated diagnosis matched the ones obtained from independent diagnostic laboratories.