火电站负荷调度混沌变步长混合遗传算法

Chaotic Variable Step Length-based Hybrid Genetic Algorithm for Load Dispatching at Power Plants

结合混沌运动的遍历性和遗传算法的群体搜索性，提出一种基于混沌变尺度梯度下降的混合遗传算法，应用于电厂负荷优化调度。算法采用梯度下降法对遗传变异获得的优良个体进行局部搜索，引导种群的进化。结合混沌优化策略产生自适应步长，在搜索初期加快寻优速度，随着搜索逐渐接近最优点，混沌产生的小步长实现在最优解所在的小范围内进行精确搜索。对电站负荷优化调度的应用发现，相比传统的遗传算法和值长分配法，该混合遗传算法对机组负荷调度优化具有更高的效率。通过该混合遗传算法进行机组负荷分配寻优后，全厂发电煤耗率相比于实际的值长制降低了0.29～1.07-（kW·h），为电厂带来可观的经济效益。

By combining the ergodic character of chaotic movement with population searching character of genetic algorithm,a hybrid genetic algorithm based on a chaotic variable dimension gradient-drop was presented,which can be used to optimize load dispatching at power plants.The algorithm adopts a gradient drop method to conduct a local search of excellent individuals obtained from genetic variation and guide the evolution of the population.In conjunction with a chaotic optimization strategy,a self-adaptive step length can be produced,quickening the optimum-searching speed at the initial period of searching.With the searching gradually approaching the optimized point,by way of the small step length produced by chaos,it is possible to realize an accurate searching within a small range where an optimized solution is located.It has been found through an application of the optimized load dispatching at power plants that compared with the traditional genetic algorithm and officer-on-duty distribution method,the genetic algorithm can achieve an even higher efficiency in load dispatching optimization for power plants.Through an optimization searching of load distribution of power plants by using the hybrid genetic algorithm,the power-generation coal consumption rate for a whole power plant can be reduced by 0.29～1.07 g/(kW·h) compared with that of the officer-on-duty system,bringing about sizable economic benefits for the power plant.