Working Condition Analysis and State Trend Prediction of Hydraulic Turbine Units

摘要

Based on the pressure parameters of the turbine, the operation condition of the turbine is determined. Based on the input data, the operation condition of the turbine is predicted by the long short-term memory network. Firstly, the identification model of BP neural network method is established to identify the specific working conditions by using the historical values obtained in the practical engineering application. Then, according to the correlation between the measuring points, the multiple time series long short-term memory network prediction model (LSTM) is constructed, and the state trend of the hydraulic turbine unit under this condition is predicted. The corresponding punishment factors are calculated by using the prediction data of each measuring point and the threshold value of the prediction band, which are mapped into the radar chart. Finally, an anomaly early warning system with flexible early warning rules based on equipment deviation index is proposed. Through the experimental analysis, the validity of the long short-term memory network prediction model and the radar graph model for calculating the deviation degree of the equipment is verified, and the advanced warning for the abnormal state of different acquisition points under different working conditions is realized, which provides a new method for the abnormal prediction and fault diagnosis of the hydraulic turbine.