Improvement Of QGA - BP Model And Its Application In Forecasting Total Nitrogen Content In

Currently, water resources is facing two major problems, one is the shortage of water, the other is the deterioration of water quality. Water has become a hazard of human survival, it impedes prosperity and development of human beings, so the protection of water resources has been imminent. In recent years,as one of the most important in water pollution prevention and control, many academics at home and abroad all have focused on water quality prediction. Water quality prediction is necessary for water environmental planning, assessment and management.By predicting water quality trends, we are able to understand the situation about a water quality in the next period of time, then we can promptly identify the reason that deterioration of water quality,and take measures. We can see that the current maintenance and management of water quality can not be separated on the basis of water quality prediction work. It is difficult to establish an accurate nonlinear predictive model by using conventional methods. For this reason, scholars have done a great job of predicting water quality parameters, and have achieved some results by using some methods,for example, Artificial Neural Networks, Support Vector Machines, Gray Theory.Water quality prediction method which based on artificial neural networks is very suitable for describing complex water environment trends because of its good characteristics of nonlinear mapping and self-learning ability, This method has become a hot topic in the field of water quality prediction today. However, the Neural Network has a low speed of convergence and is usually trapped to a local optimum. These shortcomings seriously affect the accuracy of water quality prediction. This forced scholars to constantly introduce a lot of intelligent algorithm, such as Quantum Genetic Algorithm(QGA), Particle Swarm Optimization(PSO), Simulated Annealing Algorithm(SA) and so on. They use intelligent algorithms to solve solutions that BP neural network prediction exists an unstable effect of prediction in practice. QGA-BP algorithm combines the QGA global search ability and local accurate searching of BP network characteristics,it becomes a kind of new method in areas of prediction. QGA-BP algorithm has strongly global search ability and convergence speed quickly; but in the late evolution, convergence speed changes slowly and chromosomes have a low effective utilization rate, although the algorithm improves the prediction accuracy to some extent. But the effect is still not satisfactory.Based on the above research, an improved QGA-BP model is constructed for predicting complex water quality of Miju river. First QGA introduces a dynamic improvement strategy and catastrophic policy. We adopt improved QGA which is seen as the operational guidelines of evolution to optimize the weights and thresholds of BP model. Then we select a group historical observation data as training samples for predicting Miju river water quality. And the comparison of simulation among BP model, QGA-BP model and improved QGA-BP model turns out that the improved QGA-BP model has improved greatly including evolution algebra, convergence speed and accuracy of forecasts. The results of Miju river shows that the improved QGA-BP model for water quality prediction is feasible, effective, and has greatly improved in forecasting accuracy than traditional methods.