Study On The Prediction Of The Pond Water Temperature And The System Based On GA-BP

Aquaculture pond has been one of the major modes of aquaculture industry in China. The pond temperature is one important environmental factor that influences the qualities of aquatic products, such as the content of dissolved oxygen(DO) and ammonia nitrogen of ponds, food intaking and growth conditions for aquatic products, and diseases causing by the sudden change of water temperature leading to economic loss finally. The pond temperature in aquaculture base of Gaocheng county, Yixing city, Jiangsu province was studied in this paper and data were got through temperature sensors and the small-scale meteorological station. Specific contents are as follows:(1) Data acquisition. Six temperature sensors were laid in the aquaculture pond and one small-scale meteorological station was located on the bank in aquaculture base of Gaocheng county, Yixing city, Jiangsu province. Meteorological data collected include amount of precipitation, wind direction, wind speed, solar radiation, air temperature, air humidity, atmospheric pressure and water temperature. These data were transferred into the existed service platform of aquaculture in China Agriculture University by GPRS. Experimental data were got from this service platform. In this paper, 718 data from September 12 th to 16 th of 2014 were used as original data.(2) Data preprocessing. Some environmental factors obtained from experimental program were not critically significant for the pond temperature. Therefore, principal component Analysis was adopted and all the factors were divided into 4 kinds. Factors that influenced the pond temperature greatly were selected according to the loads of each factor in principal components. These factors were solar radiation, air temperature, amount of precipitation, wind direction and wind speed. Principal Component Analysis could decrease factors input, reduce the dimension, simplify network structure, thereby, improve the astringency and stability of the network effectively.(3) Model prediction of the pond temperature. Aiming at the nonlinear variation of the current pond temperature and the difficulty to solve small-sample, big error problem caused by the present prediction methods, this paper adopted genetic algorithm to optimize BP neural network algorithm. Initialized weighting parameters and threshold parameters in every layer achieved the best through genetic algorithm optimization. Errors generated by random initialization were reduced. Compared with the results obtained from the traditional BP algorithm, precision of the optimized BP algorithm achieved 99.21%, mean absolute error(MAE), mean absolute percentage error(MAPE) and root mean square error(RMSE) improved 36.2%, 36.8%, and 57.2% respectively.(4) Design and implementation of the forecasting and warning system based on temperature prediction model. This paper applied temperature prediction model to the existed aquaculture services platform, which had never been done in China. This system could provide users with functions of real-time monitoring data presentation and storage, historical data download and forecasting and warning water temperature. Monitoring and early warning of the pond temperature was realized and technical supports were offered.