基于环路和改进遗传算法的配电网络重构优化

Optimization of Distribution Network Reconfiguration Based on Loop and Improved Genetic Algorithm

为了提高电网可靠性,在同时考虑网损和开关动作次数的基础上,以配电网电压降的限制、线路电流值的限制等为约束条件,建立了配电网络重构优化数学模型,用外部惩罚函数法将问题转化为无约束问题,并采用改进遗传算法进行求解。根据配电网的特点,在遗传算法产生初始解及交叉、变异操作时,采用基于环路的方法,避免了不可行解的产生,提高了算法的计算效率。针对遗传算法的局限性,改进操作过程,调整适应函数,改进交叉率和变异率的计算方法,结合模拟退火算法,给出了初温确定方法,改进了选择复制操作。该算法能有效地提高收敛速度,避免早熟收敛,并证明了该方法的有效性。

An optimization model of distribution network reconfiguration is established, in which the network loss and motion number of switches are considered, and the restrictions to the decline of voltage, current and the capacity of power source are taken as constraint conditions, through the outer penalty function method, the solution is transformed to an unrestrained optimization problem, and the improved genetic algorithm is used to solve the problem. Based on the features of distribution network, a loop method is adopted in the operations of initial solutions creating, cross and mutation, the methods of creating correct chromosomes and modifying wrong chromosomes are given, so the infeasible solutions are avoided, the computational efficiency of the method is enhanced. Aiming at the limitation of genetic algorithm, some operational processes are improved, expression of fitness function is adjusted, making the optimal direction of objective function same as the increasing direction of fitness function. Based on the relative capability of current population, computing methods of cross rate and mutation rate are improved, forming adaptive genetic algorithm. Meanwhile, simulated annealing algorithm is combined and the method of determining initial temperature is given, making use of relative capability of the initial population, operation of reproduction is improved based on the method of roulette wheel selection and differentiating rule of metropolis. Comparing with other optimal methods, the improved genetic algorithm can improve the speed of convergence and avoid premature convergence. Its application shows that the algorithm is efficient.