Fault prediction method of hydro turbine generator set based on CFA-Conv-LSTM

The hydraulic, mechanical and electrical factors are coupled with each other during the operation of the hydroelectric generating set, which is highly nonlinear, unstable and time-varying, and it is difficult to accurately model and predict. Aiming at the problem of hydraulic turbine generator failure prediction, this paper attempts to introduce convolution (Conv), integrate long short-term memory neural network (Long Short Term Memory, LSTM), and propose a Conv-LSTM deep learning prediction network. At the same time, in order to solve the problem of hyperparameter optimization, the chaotic firefly algorithm (CFA) was introduced, and a fault prediction method for hydro turbine generator sets based on CFA-Conv-LSTM was constructed. Using the measured data of No. 1 unit of a hydropower station in Guangxi, compared with other commonly used methods, it is proved that the method is significantly better than other methods in terms of fault prediction accuracy and process monitoring performance, and has good application prospects and promotion value.