Fault diagnosis in hydropower units based on chaotic Kepler optimization algorithm-enhanced BiLSTM model

Abstract

Safe and stable operation of hydropower units is the cornerstone of the whole hydroelectric power generation system. This paper proposes a deep learning model based on the chaotic Kepler optimization algorithm (CKOA) for fault diagnosis of hydropower units. The Tent chaotic function is introduced to initialize the initial population of KOA, which accelerates the convergence speed of the KOA algorithm and improves the global search capability by improving the uniformity and uncertainty of the population. CKOA is used to optimize the hyperparameters of Bidirectional Long Short-term Memory (BiLSTM) to improve the robustness and generalization ability of the model, making it suitable for dealing with complex nonlinear fault signals of hydropower units. The experimental results show that the training accuracy and testing accuracy of the CKOA-BiLSTM model are 97.6% and 98.3%, respectively, which are better than those of the LSTM, BiLSTM, and KOA-BiLSTM models. Meanwhile, diagnosing faults from the acoustic point of view, the accuracy of impact faults in hydropower units is much higher than that of wear and tear faults. This study can serve as a valuable supplement to the existing hydropower units condition monitoring and fault diagnosis system.