Prediction Of Zhengzhou’s Water Consumption Based On Multiple Data Sources

The prediction of water consumption in different sectors of society has important implications for modern urban planning and water resource management.This study aims to establish an accurate prediction model for water consumption in different social sectors(life,agriculture,industry,and ecology)and analyze the trend of water consumption.Two models were established in this paper: a VMD-PSO-LSTM-based time series prediction model for water consumption and a GA-SVR multiple regression prediction model based on grey relational analysis.The results show that the proposed models can accurately predict water consumption in different sectors and provide strong support and guidance for urban planning and water resource management.Additionally,the research findings can provide a basis for developing more accurate water use policies to effectively control water consumption and ensure the sustainable use of water resources.The main research contents of this paper are as follows:(1)Selection,correlation test,and prediction of the influencing factors of water consumption in different sectors in Zhengzhou city.The direct influencing factors of water consumption in different sectors in Zhengzhou were determined by studying the water consumption of each sector,including effective irrigation area,aquaculture area,livestock population,permanent population,value added of tertiary industry,and value added of the construction industry.Grey relational analysis was used to test the correlation between the influencing factors and water consumption,and all selected influencing factors passed the correlation test.The future values of the water consumption influencing factors used in this paper’s model were obtained from the "Comprehensive Plan for Water Resources in Zhengzhou City"(2020-2030).(2)Dimensionality reduction of water consumption time series data.Since the water consumption data of different sectors in Zhengzhou city available were annual scale data,the time series sample was small and contained limited information that is required for time series prediction models.To meet the high accuracy requirement of the proposed model,this paper deeply described the water use process of each sector and established a mathematical model for water consumption to achieve the purpose of dimensionality reduction of annual time series data.Models for life,agriculture,and industry were established to obtain daily scale life water consumption data,decadal scale agricultural water consumption data,and monthly scale industrial water consumption data.(3)Zhengzhou water consumption prediction based on a combination model of Variational Mode Decomposition(VMD)and Particle Swarm Optimization-Long Short-Term Memory(PSO-LSTM).The water consumption time series data of each sector after dimensionality reduction had characteristics such as nonlinearity,non-stationarity,periodicity,and volatility.Based on these characteristics,this paper proposed to decompose the data of each sector into VMD and obtain a relatively concentrated stationary sequence of the frequency distribution.The PSO method was used to automatically optimize the hyperparameters of LSTM to improve the accuracy and robustness of the model.Finally,the VDM obtained sample sequence was used to train the PSO-LSTM model to predict the future water consumption data of each sector in Zhengzhou city and compared with the simulation accuracy of other non-combination models,which demonstrated the advanced nature of the proposed model.(4)An improved support vector machine regression prediction model based on genetic optimization algorithm.Since time series prediction models are black box models,it is difficult to give different water consumption demand predictions based on the future development trends of a city.Therefore,this paper proposed a GA algorithm-based SVR model,which takes the determinants of water consumption as inputs and water consumption as outputs.The GA algorithm is used to optimize the parameters of the SVR model to improve its accuracy and robustness.The model was tested using the water consumption data of different sectors in Zhengzhou city,and the results show that it can effectively predict water consumption in different sectors.