Study On Data Processing And Forecasting Technology Of Aquaculture Water Quality

Although China is the first largest production country of aquatic farming in the world, study of aquaculture water quality analysis is considerably limited and the relevant technology lags far behind. Due to poor accuracy and timeliness, analysis on water quality of aquaculture in the traditional farming process is difficult to be applied in practical production. Aquaculture environment is a multi variable, nonlinear system, and in theory, the artificial neural network can realize arbitrary nonlinear mapping and has good self-learning ability, so artificial neural network can be applied to the prediction of water quality in aquaculture. As the most widely used artificial neural network in practice, BP neural network also has certain drawbacks and limitations, however, mind evolutionary computation can compensate for these shortcomings and optimize the function of BP neural network to some extent. This study is undertaken to make a research of prediction method of BP neural network based on mind evolutionary computation to the dissolved oxygen of water quality factor in aquaculture. Draw on experiences of the existing research findings, the main works in this paper are as follows:Firstly, preprocessed the collected data of water quality in aquaculture. In the actual process of water quality monitoring, data loss and abnormal data often occur in the actual collected data owing to the effect of environment, human interference, instrument measurement error and other factors. Linear interpolation method and average method are used to repair the data in this paper.Secondly, analyzed the characteristics of the BP neural network and mind evolutionary computation; and to solve the problem of the slow convergence speed in actual application and easily falling into local optimum of the BP neural network, BP neural network is optimized by mind evolutionary computation which has the characteristics of strong global search. Instead of giving the BP neural network initial weights and thresholds by random variable in the original, he BP neural network initial weights and thresholds are given by global search of the mind evolutionary computation, then the optimal solution is assigned to the BP neural network so that the BP neural network can run with a near optimal solution algorithm at start, which makes the BP neural network to find the optimal solution fast.Thirdly, part of the water quality data collected from the aquaculture base of Maowei Sea, Guangxi was taken as the training sample, the remaining data was taken as the test sample. The water quality factor of dissolved oxygen was simulated and predicted by the water quality prediction model of BP neural network in two groups respectively and BP neural network model was optimized by mind evolutionary computation. At last, this paper analyzed the prediction results of two models comparatively.The results show the faults of slow convergence speed and falling into local minimum value easily have been solved effectively after the BP neural network optimized by mind evolutionary computation. The BP neural network has achieved good prediction effect as well when it was applied in aquaculture water quality prediction.