基于负梯度法的供水管网污染源识别

利用负梯度法研究供水管网突发污染事故时污染源识别问题.首先针对供水管网的污染源识别问题构建最优化问题数学模型,然后用EPANET作为内嵌的水质模拟计算引擎,沿着目标函数的负梯度方向进行问题的最优解搜索,并针对管网算例,对污染源侵入时间、地点和侵入过程进行计算求解.分析差分步长、梯度求解方法等因素对问题求解效率的影响,结果表明采用基于水质模拟的差分法求解梯度向量并以水质公差作为差分步长,具有最好的求解效果.

Contamination Source Identification in Water Distribution Networks Based on Negative Gradient Method

This paper discusses the problem of pollution source identification in water distribution systems. First, a mathematical optimization model on the identification problem is built. Then with the EPANET Toolkit as the embedded water quality simulator, the optimization problem is solved with a negative gradient method. The methodology is finally applied to a sample network model published within the EPANET software, and the result shows that it is able to identify the time, location and concentration of the contamination injection sources effectively. To evaluate the performance of the algorithm, impacts of factors such as difference step and gradient method were further discussed, and the results show that, with water quality tolerances as the differential step, the water quality simulation based finite difference method is an effective algorithm to solve the pollution source identification problem in water distribution system.