遗传算法与其变型求解管径优选问题的比较

Comparasion of the least-cost design of water distribution systems using genetic algorithm and its variations

在优化配水管网的管径设计方案时,为了缩短计算时间、简化参数选择过程,需要选取速度快,参数敏感性低的算法。该文比较分析了标准遗传算法、稳态遗传算法和子群遗传算法进行管径优化计算时的求解速度和其速度对参数的敏感性。依据经典的管径优化问题框架,编写测试程序,利用3种算法对纽约隧道管网算例进行了多次管径优化计算。比较了不同的种群大小、重组率和变异率下3种遗传算法获得已知最优解时的水力计算次数和参数局部敏感性。结果表明:子群遗传算法达到已知最优解的平均水力计算次数少,不同的参数选取对其平均水力计算次数的影响较小。该算法求解速度快,参数敏感性低,推荐在管网设计时选用。

High speed algorithms that are not sensitive to the parameters selection are needed to reduce computation times and simplify parameter selection when optimizing pipe sizes in water distribution systems.A standard genetic algorithm（GA）,a steady-state GA,and a deme GA are evaluated in this paper.The test program for the three algorithms was developed within the canonical framework of GA pipe-sizing using the New York Tunnel distribution system as an example.The programs were run with different population sizes,crossover rates comparisons of the average number of hydraulic calculations required to obtain the known optimal solution and the local parameter sensitivities show that the deme GA has less calculations and its performance is less influenced by the parameters.Therefore,the deme GA is recommended for distribution network design always.