Research On The Application Of ARIMA And RBF Neural Networks Model On Prediction Of Dissolved Oxygen In Water Environment

At present, the smart monitoring of water environmental quality is developing everywhere in China. However, most of them focused on alarm of outliers and fell to understand the importance of analysis and forecast of historical data. In order to effectively use historical data and predict the variation of dissolved oxygen, based on the data collected of the equipment of Yixing fishing water quality, this paper firstly constructed data warehouse of Yixing to optimize the storage, display and mining of data. This paper use Typical Correlation Analysis method to analyze the influence of pH, chlorophyll, water temperature, turbidity to dissolved oxygen. Because of the nonlinear character of dissolved oxygen, combined model was used to predict the variation. Based on the data in July 2012 from Yixing farming collected by hours, this paper used ARIMA to predict the cyclical trends and use RBF neural network to predict the nonlinear residuals. Combining both parts of prediction, the result is more precise than single model. The result showed that dissolved oxygen reaches the highest value from afternoon to evening every day and by cleaning up the pound, the content can be protect from being too high. By daily morning, dissolved oxygen is at its lowest value and it can be increased by chemical and clean-up methods.