【摘要】 将人工神经网络用于复杂的市政管网水质研究,通过对苏州市3个自来水厂出水及小区管网在线监测点的大量水质数据进行分析,选取余氯作为指标建立了BP神经网络和模糊神经网络,并加以验证。结果表明,BP神经网络和模糊神经网络模型都能对管网余氯进行模拟,BP神经网络训练的最大相对误差为54.5%,验证数据的均方根误差平均为0.073;而模糊神经网络训练的最大相对误差为42.6%,验证数据的均方根误差平均为0.049,均优于BP神经网络模型,能更好地模拟自来水厂余氯,满足自来水厂出厂水余氯优化控制的要求。模糊神经网络所需要考虑的因素较少,应用方便,预测精度和效率较高,在城市给水系统水质模拟预测研究中有一定的参考应用价值。更多还原

【Abstract】 Two artificial neural networks,BP neural network and Fuzzy Neural Network(FNN),were used to simulation the water quality of municipal network,with the data collected from three waterworks and on-line water quality monitoring sites in Suzhou.The content of the residual chlorine was chosen as indexes.The research showed that both of two networks can analog residual chlorine of pipe network.Relative error of BP networks is 54.5%,average error is 0.073,and relative error of FNN is 42.6%,average error is 0.049,better than the BP network.So FNN can simulate the residual chlorine of waterworks more accurately and meet the requirements of waterworks.FNN had the features of less considerable parameters convenient for application,high accuracy and effective for prediction by a genetic algorithm optimization,so it had certain reference value on the water quality simulation and prediction for urban water supply system.更多还原