Research On Forecasting Methods Of Water Resources Informatization Early Warning And Forecasting System

Deterioration of the water environment is one of the main survival challenges facing mankind at present,and it is urgent to treat and repair the water environment that is polluted by economic and social development.With the promotion of policies and the increase of industry attention,the speed of domestic water resources monitoring system construction is also accelerating.Analysing and mining the accompanying massive monitoring data to explore its laws and mechanisms,and then predict the future trends of various indicators in the water area.which has guiding significance for the action planning and decision-making in water resources management.With the development of water quality prediction theory,new prediction methods have been put forward continuously.However,due to the complexity of water environment and the incomplete monitoring information,the existing water quality prediction methods are difficult to take into account both the high precision of experimental test and the operability of engineering application,which hinders the practical advancement of water quality prediction.Based on the actual water conservancy monitoring project,aiming at the business demand of water quality prediction in water resources information system,this paper designs a prediction method suitable for common water quality indicators based on time series analysis and machine learning theory.First of all,the theoretical analysis of the prediction method is carried out.Based on the analysis of the water quality change characteristics of the actual water area,the differential autoregressive moving average model and the support vector machine regression model are preliminarily selected as the prediction model according to the linear and nonlinear prediction demand of the water quality data,and the feasibility of the model is analyzed from the theoretical point of view.Secondly,a combination model prediction method is constructed,and the support vector machine regression model is used to estimate the nonlinear residuals generated by the prediction of the differential autoregressive moving average model.In the process,the particle swarm algorithm is used to optimize the parameters of the support vector machine.The p,d,and q orders of the autoregressive moving average model are determined by the ADF test and the BIC criterion,respectively.In order to improve the accuracy of long-term prediction,a sliding time window-based modeling method is used to optimize the combined model.Third,the performance of the model was verified experimentally.The HP filter was used to decompose the experimental data,analyze the difference in characteristics between different water quality index data,and select different water quality data to conduct comparative experiments on the combined model and the single model.In addition,the effect of the sliding time window method was tested.Experiments show that the combined model’s predicted root mean square error of p H and dissolved oxygen in Chaohu Basin reaches 0.20 and 0.61,respectively,and the average absolute percentage error reaches 2.0% and 6.6%,which is significantly lower than 0.22 and 0.73,2.3% and 8.5% of the differential autoregressive moving average model.Similar results are obtained for the prediction of the Taihu Lake Basin.The combined model has higher accuracy,and the overall error of dissolved oxygen prediction is higher than p H,indicating that it is more difficult to predict indicators with significant seasonal fluctuations.The combined model under sliding time window optimization has lower errors in long-term prediction than the unoptimized state model.Finally,based on business needs,a water resources information system was built,and the prediction algorithm was embedded in the system’s early warning and forecast module.The system was tested,the system was stable and reliable,and the functions of each module met predetermined design requirements.The algorithm model,mathematical method optimization and system-level business optimization scheme designed in this paper have a good effect on water quality prediction,and it can be promoted and applied in other water monitoring projects to provide business support for regional water environment management.