供水管网的多水质组分预测建模研究

Forecast Model for Multiple Water Quality Parameters in Water Supply Network

现有的城市供水管网水质预测模型仅针对单一水质组分,不足以全面反映管网水的水质状况。为此,以天津市某高校的校园管网为研究对象,在结合水力分析的基础上,同时对余氯和浊度进行预测和模拟建模。在模型的求解上,经对多元回归、人工神经网络和支持向量机进行详细的分析后,选用了较为成熟的BP算法。结果表明,模型预测的相对误差平均为20%左右,能满足实际应用的要求。

The existing forecast models for water quality in water supply network are only for a single water quality parameter, which can not entirely reflect the water quality status of the whole water supply network. To overcome the above deficiency, taking a water supply network of an university in Tianjin as study object, a new mathematical model for simultaneous forecast and simulation of residual chlorine and turbidity was developed on the basis of the hydraulic analysis. For model solution, after a detailed analysis on the multivariate regressive method, the artificial neural network method and the support vector machine, a comparatively mature algorithm, namely BP algorithm was applied for the model. The results show that the relative forecast error is about 20% , indicating that the model is feasible.