Health Performance Interval Prediction of Pumped Storage Unit Based on Multi-Objective Optimization

The health performance of pumped storage unit (PSU) affects the stable and safe operation of pumped storage station and even the power system. Thus, it is essential to predict the health tendency accurately and helpful to achieve predictive maintenance as well. An interval prediction model based on information fusing and multi-objective optimization is proposed to describe the randomness and uncertainty of the health performance. To effectively obtain the operating characteristic of PSU, the health state model is developed based on Gaussian process regression (GPR) and the monitoring data. Then, to comprehensively quantify the health degree, an integrate health index (IHI) is developed with entropy weight to fuse the healthy information of different objects. Taken the prediction interval coverage probability (PICP) and prediction interval normalized average width (PINAW) as objective constraints, the global optimization strategy is proposed based on kernel extreme learning machine (KELM) and multi-objective particle swarm optimization (MOPSO). In the final, the comparative experiments are conducted with the data collected from a PSU in China. The results show that the proposed model can deduce the health interval with reliability and accuracy.