基于BP神经网络的湖南镇水库叶绿素a浓度预测模型的研究

PREDICTING MODEL OF THE CONCENTRATION OF CHLOROPHYLL-A BASED ON BP NEURAL NETWORK IN HUNANZHEN RESERVOIR

以湖南镇水库2006—2007年间的监测资料为基础,利用插值和主成分分析法,运用BP人工神经网络方法构建预测模型,探求叶绿素a浓度与总氮、总磷、溶解氧等5项环境因子之间的关系,选出最佳预测模型,并对模型进行敏感度分析。结果显示:(1)BP神经网络模型对叶绿素a浓度预测值与实测值之间拟合程度良好,相关系数达到了0.95,说明BP神经网络模型可以用于湖南镇水库水体中叶绿素a浓度的短期预测。(2)通过对构建的BP神经网络模型进行敏感度分析,阐明了湖南镇水库藻类水华的主要驱动因素,并指出有效控制水体的pH值和溶解氧的变化是湖南镇水库藻类水华防治工作的重点。

According to the monitored data in Hunanzhen Reservoir between 2006 and 2007, using inserted and principal component analysis method to create the predicting model of the concentration of chlorophyll-a based on BP neural network in Hunanzhen Reservoir. A predicting model was constructed to deal with the relation between the concentration of chloro- phyll-a and 5 environmental factors such as total nitrogen, total phosphorus, water temperature, etc with the back propagation artificial neutral network method. The best predicting model was selected t, and sensitivity analysis was performed to the model. The results showed that the forecasted value of the concentration of chlorophyll-a according to the BP neural network predicting model had a better fit with actual value of the concentration of chlorophyll-a, and the correlation coefficient achieved 0.95, it indicated the BP neural net- work predicting model can be used for short-term forecast of the algal blooms; and through carried on sensitivity analysis to the constructed BP neural network predicting model, it clarified the main driver factor of algal blooms in Hunanzhen Reservoir, and the result showed that controlling PH value and the content of oxygen is important to prevent and control the algal blooms in Hunanzhen Reservoir.