Short-term Water Demand Forecasting Method Based On GASA-SVR

With the acceleration of urbanization in China,the problems of arranging water resources,constructing water supply networks and optimizing water supply dispatching are becoming more and more complicated.The analysis and evaluation of water resources utilization is of strategic significance in social and economic development.However,water demand forecasting is the basis and premise of optimal dispatching and water resources allocation.Therefore,this paper takes campus water as an example,analyzes water use trends and influencing factors,discusses the problems and improvement measures in traditional water demand forecasting,and constructs a new type of short-term water demand forecasting model.Firstly,this paper studies the data loss and outliers,then proposes a data correction method.Through the vertical and horizontal analysis of the data,the threshold of the data anomaly is defined.The data is corrected according to the calculation formula.Secondly,With the studies of weather conditions,the number of students in school,holidays and other factors,this paper analyzes the changes of the daily water consumption.This paper determines the input of the prediction model by using statistical methods to analyze the main factors of daily water consumption.Thirdly,aiming at the limitations of genetic algorithm optimization parameters,this paper introduces a GASA hybrid optimization algorithm.The performance is tested by two typical standard functions.The results show that the GASA hybrid algorithm has faster convergence speed and higher precision.The prediction result can be fitted to 0.927.Finally,this paper established a short-term water demand prediction model based on GASA-SVR and error correction,by analyzing the previous work and the prediction error.The traditional prediction method is extended to continuous multi-step prediction.According quantitative analysis by examples,the results show that the improved error correction prediction method increases the RMSE value by 15.03%,the MSE value by 16.75%,and the MAE value by 17.99%.