Research And Analysis Of Non-point Source Pollution

There is only three percent of water on earth is fresh as a natural resource.According to the situation of our country,the per capita volume of water resources is small,that's why our country is one of the large country serious in short of water.Due to accelerate pace of urbanization in various regions,environmental pollution problems have gradually emerged.The problem is mainly manifested in point source pollution and non-point source pollution,the former one is for pipeline drainage,easy to control and manage.And the latter one is caused by rain erosion,the problem of re pollution caused by the loss of fertilizer in rural areas has greatly affected the purity of freshwater resources.Earlier experts put forward mature models for pollution problems,however because of the later start and lack of data,it is difficult to use the mature model for load forecasting.With the development of artificial intelligence and machine learning algorithms,these methods can use a small amount of data to get a more accurate model which provides a new way for non-point source load forecasting in China.Because Artificial Bee Colony Algorithm needs less parameters and has simple structure as well as easy to conducted,it has been paid more attention by researchers.The algorithm has been widely used in the field of numerical optimization,industrial engineering and other fields.But the original artificial bee colony algorithm still has some shortcomings such as its convergence speed is slow,the result is easy into the local minimum and etcetera.Aiming at the problems of the original algorithm,some improved method is presented: during the hire bee search pHase,the chemotaxis behavior of bacterial foraging algorithm is combined with original algorithm to enables the hired bee to enter a relatively large area during the search process,so as to improve the searching ability of the algorithm,during the follow bee follow pHase,we use the principle of particle swarm optimization to update method thus the convergence precision of the original algorithm is improved.Finally,the improved algorithm is compared with other algorithms.The results show that the performance of the improved algorithm has been greatly improved.Extreme learning machine(ELM)is a new type of single hidden layer feedforward networks which possessed higher training speed and higher precision.The hidden node parameters of ELM are randomly generated.Thus caused the consequence that the algorithm needs numerous hidden nodes in training process.A new improved artificial bee colony(ABC)optimized extreme learning machine(HABC-ELM)was proposed.In HABC-ELM,the chemotaxis behavior of bacterial foraging algorithm and the principle of particle swarm optimization are introduced into ABC algorithm to improve the slow convergence speed of it,then the improved ABC is used to optimize the hidden node parameters of ELM.Regression and classification problems were used in the experiment and simulation results show HABC-ELM performs better than other algorithms.In view of the non-point source pollution load forecasting,improved artificial bee colony algorithm are used to optimize the connection weights and neuron threshold of the extreme learning machine,and prediction of non-point source pollution load of People's Republic of China environmental protection department of the Wuxi data center section water quality data.After the original data were processed and normalized the fixed window sliding model is used to improve the training data set,finally the processed data are put into the model before and after optimization then the test data set is used for testing.The results show that the improved models is better than the original one.