Charging pile fault prediction method combining whale optimization algorithm and long short-term memory network

Abstract

As the world’s energy structure is gradually changing, the automotive industry is shifting its focus to new energy vehicles in an effort to improve the performance and service life of the charging pile. To solve the problem that traditional models tend to fall into locally optimal solutions (i.e., the model optimization process stays in the non-optimal regional minimum) in complex parameter space, the study innovatively proposes a hybrid prediction model that combines the whale optimization algorithm with the gated recurrent unit-long short-term memory neural network. By introducing the whale optimization mechanism to globally optimize the key parameters of the neural network, the method improved the model’s ability to model complex time series data. Moreover, the method also effectively avoided the problem of traditional methods falling into local optimal solutions, thus improving the training efficiency and generalization ability while maintaining the model accuracy. It took only 21 s to complete the training of 600 samples, and the prediction accuracy was as high as 91%. In the four classes of fault classification experiments, the proposed model performs well in classification accuracy in all classes, showing strong multi-class fault recognition capability. Therefore, the fault prediction model developed in this study can accurately and effectively identify and predict charging pile faults, and shows high performance. This not only provides a strong theoretical foundation for the application of deep learning in charging pile fault prediction, but is also of great significance in terms of reducing operation and maintenance costs, supporting energy structure transformation, and promoting green development.