Stable Improved Dynamic Programming Method: An Efficient and Accurate Method for Optimization of Reservoir Flood Control Operation

The optimal algorithm to ensure computational efficiency and accuracy remains to be challenging for the development of robust operation model to solve the optimization problem of reservoir operation, particularly for applications involving flood control with complex flood hydrograph. The dynamic programming (DP) is one of the most popular methods to solve optimization problem, but it is limited the curse of dimensionality problem. The improved dynamic programming (IDP) method has been proposed to overcome this defeat of DP, remaining the convergence problem. The relaxation method based on approximate monotonic relationship shows potential to ensure convergence of IDP. In this study, the theoretical search range of the relaxation method are analyzed. A stable improved dynamic programming (SIDP) method is proposed based on relaxation method and a prediction method of schedulable storage states. The proposed SIDP overcomes the complex computational problem of DP and the convergence problem of IDP. The case study on an ideal reservoir and the Guanting reservoir, shows that SIDP can achieve an accuracy as high as DP, but with a much higher efficiency than DP. This method shows a strong solution to optimization problems of reservoir flood control operation.

Automated summary

This study proposes a stable improved dynamic programming (SIDP) method to solve the optimization problem of reservoir flood control operation. By combining relaxation method and a prediction method of schedulable storage states, SIDP overcomes the computational problem and convergence problem of dynamic programming. The case study demonstrates that SIDP achieves high accuracy and efficiency in reservoir flood control operation.