遗传粒子群混合算法在电厂机组负荷组合优化中的应用

Application of hybrid algorithm in unit commitment optimization

粒子群优化算法应用于火电厂机组组合问题中存在早熟收敛等现象,提出3方面改进的遗传粒子群混合算法:改进粒子群初始化方法,提出粒子初始化机组运行状态组合合理性判据,并初始化一定比例的粒子使其机组负荷随机在对应机组负荷上限附近赋值;采用部分解除约束结合惩罚函数的约束处理方法,对粒子进行机组负荷平衡操作,使大部分粒子满足约束条件;通过引入遗传算法中的交叉和变异操作增加了粒子的多样性,减小了算法陷入局部极值的可能性。采用改进的遗传粒子群混合算法对3机及5机火电厂机组负荷组合进行优化,仿真结果表明,优化成功率能达到100%。

Premature convergence may occur when PSO（Particle Swarm Optimization）algorithm is applied to unit commitment optimization of power plant. A hybrid algorithm of PSO and GA（Genetic Algorithm） is presented with three improvements：suggestion of criteria to judge the rationality of unit state group for PSO initialization with a certain percentage of particles initialized near the upper load limit;combination of partial constraint release and penalty function to make most particles satisfied with the constraints and the unit load balanced ;introduction of GA crossover and mutation operations to increase the particle diversity and reduce the premature convergence possibility. The hybrid algorithm is applied respectively to the unit commitment optimization of a 3-unit power plant and a 5-unit power plant,and the simulative results show that the optimization success rate reaches 100 %.