The Research Of Red Tide Prediction Method Based On Artiifcial Neural Networks

Red tide is an anomalously ecological phenomenon of sea water discoloration whichoccurs under certain conditions, caused by the explosive proliferation of the ocean plankton,including algae, protozoa and bacteria, etc. At present, red tide occurs more and morefrequently, and has become a worldwide ecological disaster, which has caused huge lossesto many coastal countries and regions. For these reasons, many experts and scholars carriedout researches on the methods of red tide prediction, trying to find an effective method offorecasting and early warning of red tide, in order to greatly reduce the harm caused by redtide. The traditional prediction methods of red tide are mostly based on mathematical andphysical methods, which can judge red tide happen or not according to the variation of thered tide impact factors to build the model simulations. However, due to there are so manyred tide occurrence factors, for example, the extent level of the sea water eutrophication, theinfluence of the meteorology and hydrology, the change of physical and chemical factors ofthe sea water, the foreign phytoplankton species brought by ships, and so on, thecomplexity makes it still not clear to understanding the occurrence mechanism of red tides,meanwhile, because the highly nonlinear and uncertainty between ecological system factors,the traditional forecasting methods are not really more effective. While artificial neuralnetwork as a modern technology means has showed good characteristics in dealing withnonlinear pattern recognition, and its unique processing capacity and reconciliation abilityabout information is applicable to the higher dimensional nonlinear systems whichmechanism is still not clear, and the neural network has been extensively applied in thesystem simulation, data processing and information Extraction, et al. Consequently, usingartificial neural network to predict the red tides is a good choice.First, this dissertation reviews the reasons for the formation of red tide, the harm of redtide and the happen status of red tides in recent years in China. Second, this dissertationclears up the previous prediction methods of red tides, and illustrates the three basicelements of artificial neural network, then introduces the BP neural network, the RBFneural network and GRNN, using the MATLAB toolbox of neural network to predict thesimulation model of red tides, then gets the more ideal result, and through the contrastanalyses of the experimental results, the advantages and disadvantages of the three network,I have furthermore proved that GRNN which selects proper smooth factor has higheraccuracy prediction than the BP neural network and RBF neural network in the red tideprediction, so GRNN can get more extensive application.