Study On The Modeling And Algorithm Of Red Tide Forecast Based On Artificial Neural Network

Red tide is ecological abnormal phenomenon in the coastal area whereseawater of blue color becomes that of red color by the fertile breeding andgrowth of phytoplankton, protozoa and bacteria in certain environmentalcondition. At present, red tide is one of very seriously environmental problemsalong the coastal nations of the world; it is also a global hazard of sea.Consequently, it is very important to prevent red tide disaster that we studyoccurrence mechanism of red tide and predict red tide exactly. In this dissertation, in order to predict red tide, we analyze red tideoccurrence characteristic, formation mechanism and summarize the occurrencestandard of red tide and the means of red tide forecast in common use. Based onthat, we dissertate the basic principle and characteristic of artificial neuralnetwork and study the modeling and arithmetic of noctiluca scintillans red tideforecast set up by artificial neural network. After analyzing feasibility of red tide forecast by BP neural network, thisthesis presents concrete arithmetic and realization process of red tide forecastmodeling based on BP neural network. Because the weights of global errorfunction of BP neural network comprise hypersurface of multi-extremum point,and when the training of BP neural network starts, the parameters ofconfiguration are given randomly, BP neural network in learning and training iseasy to get into a small-localized region. Aiming at the deficiency of BP neuralnetwork, we apply genetic algorithms to improve on the BP neural network forsearching the global optimal connection weights, getting over the deficiency andimproving the stability and veracity of red tide forecast. Through MATLAB simulation software, we carry out the simulationresearch of red tide forecast based on artificial neural network. The experimenthas shown that red tide forecast using artificial neural network can improveveracity and steady performance.