梯级水电站群并行多目标优化调度方法

Parallel multi-objective optimal operation of cascaded hydropower system

为保障梯级水电站群多目标优化调度问题的计算效率和求解精度,提出了基于Fork/Join多核并行框架的并行多目标遗传算法。该方法以多目标遗传算法为基础,引入多种群异步进化策略保证种群间个体多样性;采用迁移机制保障子种群的信息有机互馈,提升算法收敛性和解集多样性;利用并行技术实现子种群在各内核的同步求解,提高计算效率。针对问题特点,耦合个体实数串联编码方法、混沌初始化种群策略和约束Pareto占优机制等,进一步提升方法寻优性能。澜沧江流域梯级水电站群多目标优化调度结果表明,所提方法可充分利用多核资源,提升模型计算效率与求解精度,并能获得分布均匀、合理可行的调度方案集,为水电系统多目标高效决策提供科学依据。

To ensure the computational efficiency and solution quality of multi-objective optimal dispatch ofcascaded hydropower system, we proposed a novel method, called parallel multi-objective genetic algorithm（PMOGA）, based on the Fork/Join parallel computation framework. PMOGA makes full use of the featuresof multi-objective genetic algorithm（MOGA）. Moreover,in order to maintain the diversity and astringency,the whole individuals are distributed into a number of sub-populations,and the migration model is used toexchange information between neighboring populations. In addition, three different strategies are introducedto enhance the convergence and diversity of solutions,which are the real number encoding technique,cha-os initialization strategy and Pareto dominance by constraints. The proposed method is applied to the opti-mal operation of the Lancang river cascade hydropower stations. The results indicate that the method can im-prove the accuracy of the solutions with good convergence and diversity, which is feasible to address themulti-objective optimal dispatch problem of cascaded hydropower system.