无重访遗传算法及其在输电网络规划中的应用

Non-revisiting Genetic Algorithm and Its Application in Transmission Network Planning

将无重访的遗传算法(non-revisiting genetic algorithm,NrGA)应用于求解输电网络规划问题。NrGA通过空间二叉分割(binary space partitioning,BSP)和相应的二叉分割树(BSP tree)数据结构对遗传算法搜索过的历史位置进行记录,能够快速检测遗传操作产生的新解是否在BSP tree的历史存档中,对历史存档中已有的新解使用基于BSP的自适应变异机制进行操作,实现遗传算法的无重访搜索。此外,针对输电网络规划问题的具体特点,从编码、交叉、惩罚方法等方面对算法进行改进。最后通过一个典型算例对所提出的方法进行验证。与普通遗传算法相比,NrGA算法具有参数设定区间宽泛、收敛到最优解的概率高等多方面的优势。

A non-revisiting genetic algorithm（NrGA） was used to solve power transmission network planning problem.By advocating binary space partitioning（BSP） and employing a novel binary space partitioning tree（BSP tree） archive to store all the solutions that have been explored before,NrGA can quickly check whether there is a revisit when a new solution is generated by GA,and can mutate an unvisited solution through a novel adaptive mutation operator that based on BSP while a revisit has occurred,which achieves a non-revisiting search.Moreover,according to the characteristics of the transmission network planning problem,some measure was proposed to improve encoding,crossover operator and barrier terms.Finally,a typical example was evaluated to demonstrate the power of the proposed approach.Compared with a canonical genetic algorithm,NrGA shows a broader range of parameter setting and higher probability of convergence to the optimal solution.