Research On The Technique Of Algae Concentrations Prediction In Lakes And Reservoirs And Its Application

Recently, the occurrence of algal blooms caused by water eutrophication in domestic and foreign is very frequent. Occurrence of algal blooms in freshwater bodies, especially lakes and reservoirs seriously impacts on the safety of human drinking water and has caused serious damage on ecological environment. Monitoring Algal blooms in lakes and reservoirs of water and an early warning is important to ensure the safety of human drinking water and water pollution control and governance.This work aims to establish an effective prediction model of algal concentrations for application demonstration sites. In this paper, through studying the process of domestic and international algal bloom prediction technology of lakes and reservoirs, combining with the research of the algae growth situation, a competitive algae concentrations prediction model based on multiple neural network is proposed, follows are the contents and innovations:(1) Firstly, analyze the growth and variation law of algae in one Reservoir of Dongguan city. Secondly, analyze the related environmental factors which affect the growth of algae in the reservoir. Meanwhile, make a comparison with the algal growth law in the current studies and finally, get the growth mechanism of algae in this reservoir from ecological mechanism.(2) Introduce the Independent Component Analysis algorithm to study the potential impact factors that affect the growth of algae. The correlation analysis on the calculated independent components and algae concentrations shows that the independent component factors have a greater correlation with algae concentration than that of environmental factors. Therefore, after analysis by ICA the independent component factors determine the model input nodes of timing BP neural network, the input dimension is reduced, and the structure of neural network is optimized. (3) By studying the relationship between the algae concentrations and the structural changes in the dominant algae of algae populations, the growth and change law of dominant algae is analyzed. The degree of dominance is introduced as the algae population structure evaluation factor. The factor and independent factor analyzed by independent component analysis are the factor of the input nodes of the neural network. Then the prediction model can be further optimized.(4) By analyzing the algae growth law of this Reservoir, a competitive algae concentrations prediction model based on multiple neural networks is proposed. In different seasons, with the changes of algae species structure, algae concentration also have different characteristics and variation law. Dividing the algae growth by stage has an important impact on the accurate prediction of algae concentration. So firstly, use the structure of prediction model determined by previous study. Secondly get four training models trained by four seasons sample data of spring, summer, autumn and winter. Then use a second neural network to compete for getting the mode matching with the current growth status and to select the optimal model to predict the algae concentration. Finally, algal bloom prediction model based on competition of multiple neural networks is got. This method has some reference significance for the algal bloom forecast for lakes and reservoirs. According to the defect of the influence by the unknown parameters regardless of the mechanism model or intelligent model caused by the complexity of algae growth, a new idea and its realization method of combing the mechanism model and intelligent model is proposed.