Optimizing Operation Rules of Sluices in River Networks Based on Knowledge-driven and Data-driven Mechanism

Many river networks are controlled by sluices, especially in plain area. To prevent potential floods, maintain water level, and improve water environment in the inner river, the water transfer of river networks is needed and executed often in terms of optimized operation rules of sluices planned in advancing. To guarantee maintaining the optimized operation status, the provision of appropriate operating framework of sluices in river networks is necessary and presented in this study based on the knowledge-driven and data-driven mechanism. The general framework is formed by River Networks Mathematical Model (RNMM), Artificial Neural Networks (ANN) and Genetic Algorithms (GA), in which, ANN is used to build a rapid simulation of the flow variables in river networks, RNMM is used to train the ANN model, and GA method, whose fitness function is constructed by the ANN model, is used to optimize the operation rules of sluices. As a demonstration, the framework was applied to water transfer project of the tidal river networks locating in Pudong New Area of Shanghai in mainland China. Firstly, RNMM of Pudong was built and validated according to observed data during water transfer tests. Then, the Backward Propagation Neural Networks (BPNN) model was established as the fast simulation tool of flow variables of river networks through the numerical experiments with RNMM. The Generalized Genetic Algorithms (GGA) was recommended as optimization algorithm of sluices operation rules. Through comparing the optimization results with the RNMM simulation outputs under eight cases, it is verified that the framework can offer sub-optimal operation rules of sluices in river networks and present excellent speediness, robustness and flexibility. It is encouraged to be applied to more complicated, practical problems.