Preliminary Study of Reliability Prediction from Regulation Intensity of Pumped Storage Units in Renewable Energy Systems

Pumped storage units (PSUs) have a higher intensity of fatigue damage while flexibly regulating renewable energy sources (RESs) which will significantly affect their operational reliability. How to predict the PSUs reliability in the context of a high integration of RESs means a lot for improving the reliability and life of PSUs, power plant benefits, and energy integration efficiency. Therefore, from the perspective of system scheduling and operation, this work proposes an idea to predict the PSUs reliability from units' regulation intensity. Firstly, the regulation intensity and reliability indexes of PSUs are analyzed and selected by combining the historical operation data of typical power plants in China, and the change rule between regulation intensity and reliability is initially quantified based on regression analysis. After the correlation between them is verified, the PSUs reliability is classified, and the machine learning prediction model of regulation intensity and reliability based on Adaboost is constructed, to explore the feasibility of the technical solution for predicting the reliability from the regulation intensity of scheduling operation. The results show that 76.46% of PSUs exhibit a negative correlation between reliability and regulation intensity; the accuracy of the PSUs reliability classification prediction model can reach 90%, which verifies the potential feasibility of the proposed scheme.