Overflow Capacity Prediction of Pumping Station Based on Data Drive

In recent years, the information requirements of pumping stations have become higher and higher. The prediction overflow capacity can provide important reference for flood carrying capacity, water resource scheduling safety. order to improve accuracy, stability generalization ability model, a BiGRU–ARIMA data-driven method based on self-attention mechanism is proposed predict flow pump station. Bidirectional gated recurrent unit (BiGRU), variant cyclic neural network (RNN), not only deal with nonlinear components well, but also problem insufficient dependence over long distances has simple structure. Autoregressive integrated moving average (ARIMA) advantage being sensitive linear components. Firstly, characteristics pre-processed station data are selected screened through Pearson correlation coefficient mechanism. Then, bi-directional (BiGRU) used process data, dropout layer added avoid overfitting phenomena. We extract features obtained error terms using ARIMA model use them as correction items correct results BiGRU model. Finally, we obtain level. variation overdischarge analyzed by relation this paper, actual production Grade 9 Miyun Reservoir taken an example verify validity Model performance evaluated according mean absolute (MAE), percentage (MAPE) regression (R2). experimental show that, compared single ARIMAX, BP network, SA–BiGRU–ARIMA hybrid better effect than other models.