多维度的水厂日供水量规律性评价方法OA

[Objective]In the process of urban modernization,the complexity and dynamic characteristics of water supply systems are becoming increasingly evident,and the daily water supply volume of water treatment plants(WTPs),as an important indicator of the water supply system,can effectively reflect the overall characteristics.[Methods]This paper proposed a multi-dimensional quantitative evaluation method from temporal,frequency,and complexity domains to assess both regularity and irregularities in daily water supply patterns of WTPs,while simultaneously enhancing the transparency in model selection and development for daily water supply prediction through systematic regularity quantification.In terms of regularity quantification,it included four indices in time domain analysis:the mean absolute value of autocorrelation coefficients(MAAC),the complement of variability(1-CV),and two indices based on seasonal-trend decomposition procedure using LOESS(STL)method:the trend strength and seasonality strength(TS+SS),and the ratio of trend seasonality strength to noise strength(R0).For irregularity quantification,it included four indices:noise strength(NS)based on STL in time domain analysis,normalized spectral entropy(Hnorm)based on power spectrum in frequency domain analysis,and two indices in complexity analysis:normalized sample entropy(NSE)and normalized Hurst exponent(HE,norm).[Results]Subsequently,taking the daily water supply volumes of 10 WTPs in the Puxi area of Shanghai's central urban district as the research object,based on the analysis of water supply operation regularity and related research,the relevant parameters for autocorrelation analysis,STL decomposition,and sample entropy were set.Regularity and irregularity indices were calculated,and then the total scores for regularity and irregularity were obtained through arithmetic mean,conducting a quantification evaluation of the regularity and irregularity of daily water supply volumes.The evaluation results demonstrated that the regularity and irregularity scores of 10 WTPs assessed by the method proposed in this study for different water treatment plants can mutually corroborate each other.[Conclusion]Based on these evaluation outcomes,GARCH and XGBoost models are subsequently employed to conduct daily water demand prediction and regularity analysis.For WTPs categorized into four types-high regularity,medium regularity,low regularity,and complexity-specific water demand prediction models are recommended.This method ological framework can effectively enhance the interpretability of water demand forecasting.