A Fast and Stable Method for a Simplified Short-Term Hydrothermal Scheduling Problem using Particle Swarm Optimization and Direct Search

Short-term hydrothermal scheduling (SHS) is a nonlinear optimization problem with a nonlinear objective function and a mixture of linear and nonlinear constraints (if there are water head changes). The nonlinear characteristics and the complicated hydraulic coupling of cascaded reservoirs make it extremely difficult to find an exact optimal SHS solution. This paper studies a simplified SHS problem in which unit commitment has been solved, the valve-point effect isn’t considered, and all the online thermal units are represented as an equivalent thermal plant. EPSODS, an enhanced method using particle swarm optimization (PSO) and direct search (DS), is proposed to solve this problem. EPSODS applies a PSO that divides particles into multiple local groups to find feasible SHS solutions. When updating the position of each particle, the PSO takes into considerations such as reference to a local best flying experience and alleviating constraint violations. A DS is used to refine each final local best feasible SHS solution obtained from the PSO. The DS iteratively modifies two to three reservoir discharge rates and adjusts affected thermal power outputs to improve total fuel cost, while preserving the feasibility of the solution. A multi-level strategy enables the DS to speed up its convergence. Numerical results on a test hydrothermal power system demonstrate that EPSODS can reliably find a feasible near-optimal SHS solution in a very short time.