An interpretable physics-informed artificial intelligence model for daily hydropower generation prediction in cascaded reservoirs incorporating spatiotemporal correlations

Accurately predicting the daily hydropower generation of cascaded reservoirs is essential for efficient hydropower resource allocation. To address the issue that existing forecasting methods for cascaded hydropower generation often insufficiently consider the spatiotemporal correlations among individual reservoirs and their physical constraints, this study proposes the physics-informed method, grounded in the physical principles of maximum output constraints and ramping rate constraints for daily generation. It integrates a spatiotemporal graph convolutional network for topological modeling, gated recurrent units for temporal prediction, and a knowledge reasoning module for enhanced feature representation. Throughout this process, physical constraints are enforced on the data-driven model via a composite loss function combining a base loss and a physics-informed loss. Furthermore, SHapley Additive exPlanations theory is introduced to quantify the contribution of input features to the model's predictions, thereby enhancing its interpretability. Application results from the real cascaded hydropower system in the Jinxi River Basin, Fujian, China, demonstrate the method's effectiveness. For a one-day forecast period, the proposed method achieves a correlation coefficient exceeding 0.93 and a Nash-Sutcliffe efficiency above 0.85, with root mean square error and mean absolute error values lower than those of baseline methods. Shapley additive explanations analysis reveals that hydraulic features such as hydropower generation discharge and outflow dominate the model's predictions. In summary, this paper proposes an interpretable method for daily hydropower generation prediction in cascaded hydropower systems that incorporates physical constraints and spatiotemporal correlations. It provides valuable insights for power system dispatch operations and the rational development and utilization of water resources.