针对遗传算法存在求解精度与收敛速度间的矛盾,提出一种新的自适应对称调和遗传新算法.该算法中交叉率(crossover rate)和变异率(mutation rate)随着染色体(chromosome)的适应度函数值(fitness value)动态改变,同时在产生子代(child)新种群(group)的过程中来源不是单一的父代(parent)种群,而是由三部分组成,即父代种群中的最优个体直接进入子代种群、通过选择交叉操作产生部分新的个体、投放部分特殊个体.建立了新算法与人工神经网络结合的电力负荷预测耦合模型,并以四川电网负荷实例进行验证.结果表明,自适应对称调和遗传算法的耦合模型避免了网络寻优的盲目性,达到了最优的拟合效果,有效地提高了预测精度和速度,为区域电力负荷预测问题提供了新的分析方法,开辟了建立电力负荷中长期预测模型的新途径.
In order to solve the conflict between accuracy and convergence rate in genetic algorithm, a novel genetic algorithm for adaptation, symmetry and congruity is proposed. In that algorithm, crossover rate and mutation rate change dynamically with the fitness value of chromosome; child group comes not only from one single parent group but from three parts: optimal individuals of parent group, new individuals derived from crossover selection, new individuals generated randomly. A coupling model for power load prediction is established; by using combination of the new algorithm and ANNS and applied to Sichuan Grid. The result demonstrates that the new coupling model avoids blindness of network seeking optimization, so that to achieve the best fitting result, improve effectively forecasting accuracy and speed, and to provide new analytical method for forecast research of regional power load.