Study On Prediction Of Total Phosphorus In Yong'an River With Optimized BP Algorithm

Nowadays, with the rapid growth of economy, the important position of water resources is becoming more and more prominent. The serious damage that human beings make to ecology has led to severe water shortages and a deteriorating water quality. Water resources are essential in human activities, but now they restrict our lives, and seriously hamper the economic development of mankind, so it is our imperative trend to protect the water resources. Over the years, many scholars at home and aboard take water quality prediction as a hot spot in water pollution to study in order to prevent further deterioration of the water resources. Prediction of water quality is very important in the process of environmental management and assessment of water. Predicting the future trend of water according to the historical data of water environment, we could find the factors affecting water quality in a timely manner. It enables more rapid treatment of water pollution. Therefore, researchers used the Markov prediction method, Fuzzy Theory, Neural Network prediction method and Support Vector Machined prediction method for a variety of scenarios to predict the water quality and they have achieved some success.For some problems existed in water forecast, for example the forecast precision is not high enough and the traditional method is hard building a satisfactory nonlinear forecast system. We tried to use BP Neural Network to forecast water situation, but the forecast method had many defects. For example, the convergence speed is slow and it is extremely easy into local optima. It led to accuracy low of water forecast, so this paper aims to optimization BP Neural Network to get better of water forecast model through improving GA algorithm and PSO algorithm. GA algorithms improved the encoding, crossover and mutation; PSO algorithm improved the inertia coefficient and accelerated the learning factor of particle velocity, and it used nonlinear methods to adjust the inertial coefficient, introduces of Adaptive acceleration factor, thereby improving the accuracy of the algorithm to accelerate the convergence rate and improving the global search capability of algorithm. Using improved algorithms to optimize weights and thresholds of BP Neural Network, the prediction model of optimized in a way that overcome the drawbacks of using BP Neural Network to forecast alone, expanded the scope of the optimal solution. It will get a reasonable result of prediction of water quality.This paper aims to predict water quality of YongAn River which belongs to a more complex water environment. We used an improved Genetic Algorithm and improved Particle Swarm Optimization to optimize the weights and thresholds between BP Neural Network nodes in order to get a model of water quality prediction. Then select known data as the learning sample to study the changing trend of water quality of YongAn River. By comparing the result of BP algorithm, an improved GA algorithm optimized BP network with improved PSO algorithm optimized BP network, we realized that in the prediction results improved PSO algorithm optimized BP network model obtained the best in the three. Obtained from the predict result of water quality of YongAn River, improved PSO algorithm optimized BP network model is feasible and effective. What’s more, the convergence rate and the improvement in forecast accuracy are very good.