Long-Term Joint Operation of Cascade Reservoirs Using Enhanced Progressive Optimality Algorithm and Dynamic Programming Hybrid Approach

Abstract

Dynamic Programming (DP) is one of the most classical methods adopted for reservoir operation. It reduces the computational efforts of complex high-dimensional problems by piecewise dimensionality reduction and provides the global optimums of the problems, but it suffers the “curse of dimensionality”. Progressive Optimality Algorithm (POA) has been used repeatedly in reservoir operation studies during last decades because it alleviates the “curse of dimensionality” of DP and has good convergence and extensive applicability. Nonetheless, the POA encounters two difficulties in multireservoir operation applications. One is the transfer interrupt problem that makes the search procedure hard to achieve free allocation of water between two non-adjacent stages, and another is the dimensionality problem that leads to a low convergence rate. In order to overcome these deficiencies, this study made some enhancements to the POA and proposed a hybrid approach combining the Enhanced POA and DP (EPOA-DP) for long-term operation of cascade reservoir systems. In EPOA-DP, the EPOA is employed to improve the quality of the solutions and DP is used to reduce the computational effort of the two-stage problem solution. The proposed approach was tested using a real-world four-reservoir cascade system and a ten-reservoir benchmark test example. The results demonstrated that it outperforms the POA both in computational time and quality of the solution.