The Research On Urban Short-term Water Demand Prediction And Error Correction Method Based On GA-ELM

With the development of economy,large numbers of people flooded from the countryside into cities.Some problems such as urban water shortages and imbalances in water supply and demand have become apparent.Therefore,under the premise of ensuring the comfort of urban residents,saving water supply costs and water consumption is an inevitable trend in the development of urban water supply systems in the future.In the future,in order to realize the real-time and automated water supply dispatching method of the city,a high-precision and fast water demand forecasting method is required as a theoretical basis.The water demand forecasting method plays an important role not only in the design,planning,management and operation of urban water supply systems,but also an important reference for urban water supply strategies,operation scheduling,and optimal design.Traditional prediction methods have obvious periodicity and trend stability requirements for historical data,and there are problems such as easy to fall into local optimum,slow calculation speed,and low generalization performance.Therefore,it is rarely used alone as a method to study water demand forecasting.The combined forecasting model,which combines data,influencing factors and neural network algorithms,is favored by most scholars today and has been well developed because it combines the advantages of the two and makes up for the shortcomings.The traditional neural network prediction model has the disadvantages of long training time,easy to fall into local optimal prediction results,and insufficient prediction accuracy in the case of a small number of data samples.A prediction model for urban short-term water demand is proposed to solve these problems.First of all,the model,added influencing factors as input,of the short-term urban water demand prediction was conducted based on extreme learning machine,which was optimized by a genetic algorithm so as to achieve a higher prediction accuracy.Secondly,a method of using Markov chain to correct the prediction results of GA-ELM model is proposed.The test results show that the accuracy of the prediction results of the final combined model is higher than that of the traditional prediction model,which has more excellent practical value and provides a theoretical basis for future real-time dispatch of water resources.