基于动态剪枝的城市供水管网优化调度算法

城市供水管网的优化调度模型,需要反复调度水力学模型来计算目标函数和约束条件,从而产生巨大的计算和时间成本。在优化迭代过程中,为了获得更优秀的优化结果,通常在搜索空间中约束条件的边界进行搜索,产生了大量不满足约束的样本和较低的优化效率。为解决这一问题,提出了一种高效的基于约束条件的动态剪枝方法,在运行水力学模型前,利用优化计算过程中积累的数据来判断样本是否符合约束条件,从而剔除不满足约束的样本,提高优化效率。模型分别在1个案例管网和1个真实管网上进行测试,结果表明:使用朴素贝叶斯、决策树和支持向量机作为动态剪枝算法,在获得和原始优化模型几乎相同结果的同时,分别减少56.4%、58.5%和56.8%的计算次数。

OPTIMIZATION OF URBAN WATER SUPPLY NETWORK BASED ON DYNAMIC PRUNING MODEL

The optimal scheduling model for urban water supply networks often requires repeated scheduling of the hydraulics model to calculate the objective function and constraints,which leads to high computational and time cost.In the optimization iteration process,searching is usually performed at the boundaries of the constraints in the search space to obtain better optimization results,which leads to a large number of samples that do not satisfy the constraints and optimization efficiency.To solve this problem,an efficient constraint-based dynamic pruning method was proposed,which used the data accumulated during the optimization calculation to determine whether the samples met the constraints before running the hydraulics model,to eliminate the samples that do not satisfy the constraints and improve the optimization efficiency.The model was tested on a case network and a real network,and the results showed that the use of Naïve Bayes,decision trees,and support vector machines as dynamic pruning models reduced the number of computation by 56.4%,58.5%,and 56.8%,respectively,while obtaining the similar results as the original optimization model.