River flow prediction is a challenging problem due to highly nonlinear hydrological processes and high spatio-temporal variability. Here we present a hybrid network of convolutional neural network (CNN) and long short-term memory (LSTM) network for river flow prediction. The hybridization enables accurate identification of the spatial and temporal features in precipitation. A shortcut layer is used as an additional channel of passing input features through the deep network to increase feature diversity. The flows in Hun River Basin, China are predicted using the trained hybrid network and are compared with the results from the Soil and Water Assessment Tool (SWAT) model. The results demonstrate the learning efficiency of the hybrid network is greatly affected by its structure and parameters, including the number of convolutional layers and LSTM cell layers, the step size of pooling and training data size. Further, the shortcut layer can effectively solve the diversity reduction problem in a deep network. The hybrid network is shown to have a similar predictive performance to SWAT but is superior in wet seasons due to its nonlinear learning ability. This study shows that the hybrid network has great promise in learning nonlinear and high spatio-temporal variability in river flow forecasting.

This paper took the Foziling Reservoir in the Pi River Basin as an example, used an improved genetic algorithm to optimize the flood control dispatching during the flood process, and compared the results with the traditional genetic algorithm and the dispatching plan in the 2020 large-scale reservoir flood control operation plan. The results showed that, compared with the traditional genetic algorithm, the improved genetic algorithm saved the time for the model to determine the penalty coefficients and made the model application more convenient. At the same time, the design of the original scheduling scheme also has certain limitations. The scheduling results obtained by improving the genetic algorithm could occupy a small flood control capacity as much as possible under the premise of ensuring the safety of the reservoir itself and the downstream area.

To maximize the benefits of an interbasin water diversion system (IBWD), an aggregation model and improved NSGA-II algorithm were constructed to establish operation strategies for water diversion and water supply. An aggregation model was constructed to aggregate multiple reservoirs into an equivalent reservoir which made the operation strategies for water diversion and supply easy to define. Then an improved nondominated sorting genetic algorithm II (NSGA-II) algorithm was developed by considering variables’ sensitivity to improve its search efficiency. Finally, the IBWD subsystem for water supply in China’s Shenzhen city was taken as a case study, and its operation strategies for different water supply scenarios were optimized by the improved algorithm. Simulation results demonstrated that the equivalent reservoir is an effective method to reduce the complexity of operational strategies, and improved NSGA-II algorithm has a higher search efficiency.

This paper presents a new decision-making strategy for hydropower operations to handle uncertainty of forecasting precipitation. This strategy takes into account three basic components: uncertainty of precipitation, operation policies and a risk-evaluation model. In real-time operation, precipitations with different probabilities at different forecasting levels are obtained, and these precipitations are applied to forecast inflows using a hydrological forecasting model. Based on the forecasting inflows, the operation policies and risks with different probabilities are obtained. This study implements China’s Huanren reservoir and medium-term precipitation forecasts from the Global Forecast System to study the efficiency and stability of this strategy.