Prediction Of Water Quality Index In Flood Season In Jinan City Based On CNN-LSTM Model

In recent years,the economy of Shandong Province has been booming,but it has inevitably paid a great environmental price.It not only consumes a lot of resources and energy,but also causes serious pollution.Global warming caused by pollution,deteriorating water quality and energy shortages make the challenges ahead even more severe and daunting.To achieve sustainable development in Shandong Province,it is necessary to strike a balance between development and the environment,adhere to the principle that " Green mountains and clear water are equal to mountains of gold and silver",do a good job in daily monitoring of environmental assessment indicators,and take timely actions.As the capital city of Shandong Province,Jinan has an example and demonstration role in environmental protection.Jinan is known as the "City of Springwater".With the Yellow River flowing through it,Jinan contains abundant water resources,which makes the prevention and control of water pollution particularly important.Flood season is the key period of water pollution control.In the flood season rainfall concentration period,urban sewage collection and treatment capacity is relatively insufficient,rain and sewage mixed flow,domestic sewage overflow and other problems are concentrated and prominent,the increase of rainfall brings the risk of exceeding the standard of water environment assessment indicators.Therefore,it is necessary to predict the water environment index according to the rainfall and take timely measures to avoid the decline of water quality when it is possible to exceed the standard.In this paper,CNN-LSTM time series prediction model in deep learning was used to predict four important water quality indicators,namely potassium permanganate index,total nitrogen,total phosphorus and ammonia nitrogen.At the same time,particle swarm optimization algorithm was used to optimize the results of the prediction model with good effect,so as to further reduce the errors and increase the prediction accuracy.First of all,considering to increase the universality of the research,the sections of three different rivers,namely A section of the Tuhai River,B section of the Yellow River and C section of the Xiaoqing River,which are relatively representative in Jinan,were selected as the research objects,and the CNN and LSTM models were selected for time series data prediction based on the existing studies.The precipitation data combined with the historical water quality index data were used to predict the future water quality index value changes,and the prediction results had a larger error and a low degree of fitting,which proved that the simple CNN and LSTM models were not suitable for the prediction of the four indexes of water quality index,potassium permanganate index,total nitrogen,total phosphorus and ammonia nitrogen by precipitation.Secondly,considering the fluctuation of rainfall during the flood season,in order to better extract data characteristics,CNN and LSTM model are combined,CNN model is used for feature extraction,and then LSTM model is used for prediction.The effect of CNN-LSTM model is obviously better than that of simple LSTM and CNN model,but the prediction effect of ammonia nitrogen index is generally poor,and the prediction effect of total phosphorus index is good only at A section of the Tuhai River,indicating that this model is more suitable for the prediction of total nitrogen and potassium permanganate index.Finally,consider using evolutionary algorithm to optimize model results,particle swarm optimization algorithm was used for parameter optimization of indicators with good prediction effect to further reduce errors,improve the degree of fitting and prediction accuracy,and make the model results more accurate.In this paper,the machine learning model is used to predict the water quality indexes of three sections of Jinan City according to the flood season rainfall,and the prediction accuracy and model fitting degree are improved after the model is combined with the improvement,which proves that some water quality indexes are not suitable for the CNN-LSTM model prediction,and enrichs the relevant research on the prediction of water quality indexes of Jinan City according to the rainfall.The application research of CNN-LSTM model is developed,and it also has certain reference significance for the prediction model building of environmental supervision departments and the determination of the next research direction.