Research On Forecasting Method Of Urban Monthly Water Consumption Based On Signal Decomposition And Optimized Extreme Learning Machine

In recent years,because of the ceaseless development of science and technology,a huge progress has been made in the construction of urban informatization and industrialization,and the demand for water resources in urban has shown explosive growth.At the same time,the greenhouse effect caused by the rapid pace of urban construction leads to global warming,which leads to the serious reduction of water resources on the earth.Therefore,it is necessary to use the limited water resources reasonably.Water distribution system is the decisive factor of sensible utilization of water resources,and water consumption forecasting directly affects the effectiveness and availability of the water resources dispatch and distribution pipe network layout in water distribution system,so it is very necessary to make scientific and accurate forecasting of urban water consumption.In the forecasting of water consumption,in order to get accurate forecasting results,it is necessary to analyze the data of water consumption,select reasonable preprocessing methods and forecasting methods according to the different features of different data,and build an effective model.The water consumption data is a typical nonlinear time series data with high uncertainty.Therefore,a combined forecasting method based on signal decomposition and optimized Extreme Learning Machine is adopted in this thesis to forecast the urban monthly water consumption.The specific work is as follows:Firstly,aiming at the problem that the current signal decomposition algorithm is not completely decomposed and combined the forecasting model to lead to low forecasting accuracy,a signal decomposition method of preprocessing algorithm based on Empirical Mode Decomposition(EMD)and Variational Mode Decomposition(VMD)is proposed.Under the condition of not changing the characteristics of the original data,the EMD algorithm is used to decompose the data initially,and then the VMD algorithm is used to further decompose the first component decomposed by the EMD algorithm,so that the obtained data is more suitable for forecasting.The related simulation results verify that the proposed algorithm can effectively improve the forecasting accuracy by decomposing the original water consumption data reasonably.In addition,because of the single neural network model for water consumption forecasting accuracy is not enough,the method in this thesis firstly improves the Particle Swarm Optimization algorithm which is widely used in swarm intelligence algorithm,and proposes a forecasting algorithm combined improved PSO algorithm and Extreme Learning Machine.Combined with the signal decomposition algorithm proposed in the previous thesis,this thesis further proposes a forecasting method based on signal decomposition algorithm and Extreme Learning Machine optimized by improved PSO algorithm.This method firstly by using the adaptive weight to improve the PSO algorithm,and then the improved adaptive weight PSO algorithm to optimize the initial weights and biases of ELM,secondly using the optimized ELM algorithm respectively to forecast the decomposition of each component,and each component of the predicted results into the final forecasting results.The simulation results show that this algorithm can significantly reduce the error of water consumption forecasting and improve the forecasting accuracy.