Study On The Model Of Water Quality Prediction Based On ARIMA-DBN

“Make sure water quality is fit for raising fish if you want high-quality fish”-because the standard of cultivation environment is directly related to the growth and development of farmed species.Aquaculture water which is living environment of farmed species determines its quality and output.In recent years,with the continuous increasing of aquaculture scale and intensive,aquaculture species and cultivation density has raised markedly,water quality of aquaculture environment has deteriorated gradually and disease rate has become much higher,which has easily caused the problem of the quality of the aquaculture product.Therefore,building a monitoring system about aquaculture water quality environment is an urgent need.Thus we know the water quality conditions of farmed fish living environment timely,so we can take effective measures to control water quality and ensure the quality of aquatic products.Combined with the actual situation,we cannot far beyond meeting the demand using the existing water quality monitoring and forecasting method.because it lack of timely、reliability and high cost etc.Large-scale,scientific farming regards as the main way the future trend of the development of aquaculture,which can achieve the goal of the high survival rate and high quality aquatic products.According to the above problems,this paper designed the monitoring and forecasting system which can automatically monitor environmental parameters of the aquaculture water quality,aquatic products circulation and sales and completed the data collection,storage,handling and forwarding of environment data of the aquaculture water quality,aquatic products circulation,sales and monitoring location information,and monitoring of real-time water quality parameters based on the sensor to predict early warning.In this paper,the main work summarized as follows:(1)With a fixed period data of water quality parameter obtained by the sensor,due to possible data quality problems in aquaculture water quality parameters,this paper puts forward the anomaly detection algorithm of water quality parameters based on sliding window,and removes abnormal points out on the original data.(2)Based on the analysis of the characteristics of ARIMA model and depth belief network(DBN)model,a combination forecasting model ARIMA-DBN which is suitable for water quality parameter prediction is established.Taking dissolved oxygen as an example,a single ARIMA model and DBN are compared and analyzed respectively to verify the accuracy and effectiveness of the combined forecasting model ARIMA-DBN.(3)On the basis of research of the above combination prediction models,this paper designs and implements the automatic monitoring and warning system for the environmental parameters of aquaculture,aquatic products circulation and sales.The system can effectively monitor the main environmental factors of aquaculture and circulation,and use the combined forecasting model ARIMA-DBN to predict the key environmental factors.There will be early warning if the prediction result is abnormal.