Short-Term Downstream Water Level Prediction Model for Three Gorges–Gezhouba Cascade Reservoir Operation Based on LSTM Algorithm

In this study, LSTM, long-term and short-term memory model, is applied to short-term downstream water level prediction of cascade hydropower stations. LSTM is an artificial neural network model that combines multiple regression ideas and time series ideas, so it is different from the traditional model based on physical model and empirical formula. By inputting the time series of downstream water level and physical factors that affect the downstream water level, the model completes training, and the purpose of accurately predicting the downstream water level is achieved. Taking the Three Gorges and Gezhouba cascade hydropower station as an example, this study shows that the neural network algorithm can increase the accuracy of short-term water level prediction.