Prediction Model Of Key Environmental Factors In Marine Ranch Research

Marine ranching is a modern sustainable fishery production mode.Dissolved oxygen(DO),ammonia nitrogen,water temperature,chlorophyll a,salinity,turbidity and other environmental factors have an important impact on the growth of marine organisms such as fish,shrimp,shellfish and algae in marine ranching.Among them,dissolved oxygen and ammonia nitrogen are two key environmental factors that directly affect the survival and reproduction of aquatic organisms.Therefore,the prediction of dissolved oxygen and ammonia nitrogen is of great significance for the environment of marine ranching.In this paper,Levenberg-Marquardt(LM)neural network,genetic algorithm(GA)and principal component analysis(PCA)algorithm are combined to propose a prediction model of dissolved oxygen and ammonia nitrogen in marine ranching based on GA-LM-PCA.PCA is used to determine the influencing factors,and the decoupling and dimensionality reduction of the influencing factors are realized.Genetic algorithm is used to optimize the network structure and determine the appropriate number and weight of hidden layer nodes;the neural network is trained by LM to improve the convergence speed.And a prediction model verification platform composed of perception layer,transmission layer,power module and PC monitoring and management platform is built to verify the reliability,real-time performance of data transmission and the accuracy of model prediction.The results show that the verification platform can realize the rapid and real-time acquisition of water quality and environmental information in the marine ranch.The monitoring data transmission of the marine ranch is reliable,and the monitoring system runs smoothly.The dissolved oxygen and ammonia nitrogen values predicted by the GA-LM-PCA method are in good agreement with the measured values.The average absolute error and root mean square error are 0.0047 and 1.8727×10^-4,0.0065and 9.4287×10^-4,respectively.It is suitable for the occasions with a large number of influencing factors.GA-LM-PCA can be considered as an effective prediction tool for key environmental factors in the marine ranch.Especially,it is more effective for nonlinear systems with complex influencing factors.