基于多策略协同进化差分算法的社区居民负荷优化调度

针对需求响应下负荷调度的问题，为提供满足居民利益的响应方案，并提高电网运行稳定性，综合考虑电价、激励型需求响应机制与居民用电需求，以用电成本和社区负荷方差最小化为目标，建立了多用户负荷调度高维目标优化模型。结合模型特征提出一种基于多策略的合作协同进化差分进化算法，设计了基于居民用电特征的混合编码与种群初始化策略，以提高解的质量；引入合作协同进化思想将问题变量分解，依据高维目标分组与聚合对种群进行划分，避免陷入局优；各子种群进化时采取双差分模式协同策略，并构建知识迁移个体实现种群间信息交互，最后经贪婪与随机选择结合的种群合并策略保留完整优秀解至外部档案，以提高Pareto最优集的收敛性与分布性。算例仿真表明所提方法可降低社区居民用电成本18%左右、负荷波动方差30%以上；随着居民数量增加，算法的收敛性与多样性与同领域其他算法相比优势更为明显。

Optimal scheduling of community residents load based on multi-strategy co-evolutionary differential algorithm

Aiming at the problem of load dispatching under demand response, it is necessary to provide a response scheme to meet the interests of residents and improve the stability of power grid operation. This paper considered electricity price, incentive demand response mechanism and residential electricity demand, and established a multi-user load scheduling high-dimensional objective optimization model with the goal of minimizing electricity cost and community load variance. Combined with the characteristics of the model, this paper proposed a cooperative co-evolutionary differential evolution algorithm based on multi-strategy. It designed a hybrid coding and population initialization strategy based on the characteristics of residential electricity consumption to improve the quality of the solution. It introduced the idea of cooperative co-evolution to decompose the problem variables, and divided the population according to the high-dimensional target grouping and aggregation to avoid falling into local optimum. In the evolution of each sub-population, it adopted a double differential mode coordination strategy, and constructed knowledge transfer individuals to realize information interaction between populations. Finally, it retained the complete excellent solution to the external file through the population merging strategy combining greedy and random selection to improve the convergence and distribution of the Pareto optimal set. Simulation results show that the proposed method can reduce the electricity cost of community residents by about 18% and the variance of load fluctuation by more than 30%. As the number of residents increases, the convergence and diversity of the algorithm are more obvious than other algorithms in the same field.