维分割和压缩视角下均衡搜索粒子群优化算法及在电力系统调度的应用

The research goal of particle swarm optimization (PSO) is to achieve balanced search, i.e. being able to correctly locate the subregion where the optimal/near-optimal solution resides in and to precisely determine the optimal/near-optimal solution on each dimension of the search space respectively through exploration and exploitation. For many existing PSO algorithms, they cannot find the optimal/near-optimal subregion on some dimension or the final solution accuracy isn't high on some complex benchmark functions and real-world application problems. In order to achieve the balanced search goal, this project studies associating each dimension with independent algorithm parameters and independent exploration/exploitation progress, in addition this project proposes to adaptively and appropriately segment each dimension into multiple subregions and gradually shrink each dimension. By dynamically determining the particle velocity update procedure, the particle position auxiliary adjustment technique, and the algorithm parameters based on incorporating the segmentation and search status of each dimension, PSO can gradually locate the correct optimal/near-optimal subregion and finally find out the optimal/near-optimal solution. This project applies the proposed balanced search PSO algorithm developed from the perspective of dimension segmentation and shrinking to the time and space coupled large-scale constrained optimal operation of power systems, helping deriving high quality operation policies that are economically, environmentally and/or socially beneficial. This project thus exhibits important theoretical meanings and application values.

粒子群优化算法的研究目标是实现均衡搜索，即在搜索空间各维上都能够通过探测搜索正确定位最优/次优解所属的子区域且通过开采搜索精确地找出最优/次优解。现有的诸多粒子群优化算法对于某些复杂基准测试函数和实际应用问题在某些维上找不到最优/次优子区域或解的精度不高。为了达成均衡搜索这一目标，本项目研究对搜索空间各维采用独立的算法参数和独立控制探测/开采搜索进度，提出根据算法的演化状态自适应地将各维恰当地分割成多个子区域并对各维进行渐进压缩，结合各维的分割和搜索状况动态地确定粒子飞行速度的更新过程、粒子空间位置的辅助调整方法和算法所涉及参数的设置，逐步正确定位最优/次优子区域并最终找出最优/次优解。本项目将所提出的维分割和压缩视角下均衡搜索粒子群优化算法应用于解决电力系统的时空耦合大规模复杂约束优化调度，帮助得出对经济、环境和/或社会有益的高质量调度方案。本项目的研究工作具有重要的理论意义和应用价值。

粒子群优化算法的研究目标是实现均衡搜索，即在搜索空间每一维上都能通过前期探测正确定位最优/次优解所在的区域，快速从前期探测收敛到后期开采，并通过后期开采精确地找出最优/次优解。现有的诸多粒子群优化算法对于某些复杂基准测试函数和实际应用问题在某些维上定位不到最优/次优解所在的区域、收敛速度慢和/或解的精度低。为达成均衡搜索这一目标，本项目研究了对搜索空间每一维采用独立的算法参数和独立控制探测、收敛和开采进度，根据种群在每一维的进化状态自适应地将每一维恰当地分割成多个子区域，对每一维进行渐进压缩，并调整算法参数，快速和正确地定位最优/次优解所在的子区域并精确地找出最优/次优解。本项目将所提出的维分割和压缩视角下均衡搜索粒子群优化算法应用于解决了电力系统的时空耦合复杂约束优化调度，得出了对经济、环境和/或社会有益的高质量调度方案。