Study On Predict Model For Maximum Biomass Of Algal Blooms Based On BP Neural Network

Algal blooms are abnormal ecological phenomenon with explosive proliferation of phytoplankton caused by the change of water environment. The outbreak of algal blooms will not only destroy the normal ecosystem of water bodies, it also can produce algal toxins that hazard the health of human body. This paper is committed to explore the outbreak mechanism of algal blooms to find the key control factors in the rapid proliferation of Phytoplankton and proper prediction models were established to predict the time, range and extend of the outbreak of algal blooms precisely, to providing technical support for the forecasting of the breakout of algal blooms in the source of the drinking water, so that the water plants can take coping measures to reduce the harm that eutrophication caused.It obtains the correlation between algal biomass and different nutritional factors, such as total nitrogen, total phosphorus, nitrate nitrogen, nitrate nitrogen and microelement such as iron, manganese and zinc through the correlation analysis, mainly show that the correlation coefficients for ammonia nitrogen is larger than that of nitrate nitrogen, the correlation between microelements and algal biomass is more obvious in high concentration of initial nitrogen and phosphorus, the correlation between the ratios of nutritive elements and algal biomass is larger than the correlation between nutrient elements and algal biomass. The correlation also show that algae are very sensitive to the change of zinc concentration and the microelements such as iron, manganese and zinc have a great influence on algae growth just like macro-elements phosphorus. Further research was taken in this paper to determine the nutritional factors as the input of the model, such as total nitrogen, total phosphorus, nitrate nitrogen, nitrate nitrogen and microelements such as iron, manganese, zinc and so on. Then it determined the maximum algae as the predictive factor of the model through summarizing the former research conclusions about the prediction for outbreak of algal blooms.Selected different number of hidden layer neurons and different transfer function of hidden layer for network training of the model, finally confirmed the optimal number of neurons and the optimal transfer function by comparing the network performance such as convergence and mean squared error, and then obtained the model which has the best network performance. Macronutrients concentration-Algal biomass Model with best network performance which converges rapidly and mean square error (mse) is under10-7.Micronutrients concentration-Algal biomass Model with best network performance which converges rapidly and mean square error (mse) is under10-3. To verify the accuracy of the model by compare the experiment values and predict values, Finally built predict model for maximum biomass of algal blooms based on BP artificial neural network through the MATLAB neural network toolbox.It builds Macronutrients concentration-Algal biomass Model and Micronutrients concentration-Algal biomass Model through BP artificial neural network. For the Macronutrients concentration-Algal biomass Model, hidden layer neuron number is10, transfer function is tansig, accuracy reached92.37%; For the Micronutrient-Algal biomass Model, hidden layer neuron number is6, transfer function is tansig, accuracy reached90.39%.