Reservoir Capacity Prediction for Pumped Storage Pumping Station Based on Machine Learning

The storage capacity of pumped storage pumping station is an important monitoring object for pump station operation, and predicting the storage capacity is an important operational management task. Predicting storage capacity can help staff arrange pumping and energy storage in a reasonable manner, maximizing energy efficiency and economic benefits. This paper used back propagation network (BPNN) and long short-term memory network (LSTM) for storage capacity prediction, including the storage capacities of upper and lower reservoirs. Trained and tested these two models by using historical storage capacity data. evaluated their performance in predicting future storage capacities. Used performance indicators <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$MAPE,\ RMSE$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R^{2}$</tex> to reflect the accuracy of prediction models. At the same time, draw distribution and scatter plots of the predicted values of the two models and the actual values of the test data. The experimental results showed that compared to BPNN model, the LSTM model exhibits higher accuracy and generalization ability in predicting the storage capacities of pumped storage pumping station. The predictions of the capacity for the upper and lower reservoirs showed an increase of 1.67% and 0.39% in MAPE, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$21.1840m^{3}$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$5.2908m^{3}$</tex> in RMSE, and 0.1494 and 0.0292 in <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R^{2}$</tex>, respectively.