Chlorophyll-a predicting based on artificial neural network for marine cage fish farming area in dapeng cove in Daya Bay, South China Sea

Base on the 201 groups of data that accepted in the last ten years, a 3 layer (3,8,1) BP artificial neural network model on quickly predicting chlorophyll-a concentration in marine cage fish farming area was established. The model was established in software MATLAB7.1 (MATTrix LABoratory) using BP network. Three field accurate measurement parameters (water temperature, pH, dissolved oxygen) was as the input variable and chlorophyll-a was the output in our model. In most condition the forecast results was closely to the actual data when using this model. Its prediction accuracy was significantly higher than the linear regression equation. For the reason that the data used in building model which has some question and the complexity of predicting chlorophyll-a content, there existed some error between forecast value and actual value when using this model in several sets of data. This article put forward the methods to consummate the model in the next step.