A hybrid WOA-CNN-BiLSTM framework with enhanced accuracy for low-voltage shunt capacitor remaining life prediction in power systems
Ningning Li, Weiyao Xu, Qiuyu Zeng, Yanjie Ren, Wenchuan Ma, Kezhu Tan
Abstract
Low-voltage shunt capacitors, as a good reactive power compensation component, have been widely used in power systems. However, when their capacitance decays to a threshold value, causing them to fail, it will seriously affect the safe operation of the system. This paper aims to study the remaining service life of low-voltage shunt capacitors and establish a data-based prediction model considering various environmental factors. Based on the traditional long short-term memory neural network prediction, an improved bidirectional long short-term memory network method combining convolutional neural networks and whale optimization algorithm is proposed, which improves the accuracy, speed, and robustness of prediction. The root mean square error (RMSE) and mean absolute error (MAE) before and after optimization are compared based on simulation. The simulation results show that compared with the traditional LSTM model, the RMSE of the prediction results of the WOA-CNN-BiLSTM model is reduced by 0.0117, and the MAE is reduced by 0.0063.Therefore, the WOA-CNN-BiLSTM model has higher accuracy and stability, can effectively reduce the power quality decline caused by the abnormal working state of the reactive power compensation equipment, so as to improve the operating efficiency of each equipment in the power system and extend its service life. • The verification experiment of low voltage shunt capacitor is carried out. • Lifetime prediction of low voltage shunt capacitance by AI algorithm. • The traditional algorithm is reformed and a new algorithm is introduced. • Experiments show WOA-CNN-BiLSTM algorithm is superior to the traditional algorithm. • The work carried out has reduced economic losses in the industry.