水华风险不仅是水利工程规划时需要考虑的环境问题，也是水利设施运营时不能忽视的监测项目。为了提高明渠水化风险等级预测的准确率，针对水华成因的不确定性和发展的时序性，基于动态朴素贝叶斯网络分类器提出一种应用于明渠的水华风险评估模型。模型用水华风险等级结点对应藻叶绿素a（Chla）的浓度，并考虑了9项影响水藻生长的因素。并采用主成分分析法，处理专家咨询结果，进行参数的设计。在苏州河道北门桥2011年6月初到9月初观测的53例连续监测数据上，与基于朴素贝叶斯网络分类器的评估模型进行比较实验。混淆矩阵显示对中等风险情况的预测识别率提高了15.625%，单尾配对t检验表明在显著性水平0.05时，两模型预测识别率差异显著。考虑了时序特征的基于动态贝叶斯网络分类器的评估模型对明渠中等水化风险的预测识别率提高显著。

Algal bloom risk is not only an environmental issue to be considered in water conservancy project planning, but also a monitoring item that cannot be ignored in the operation of water conservancy facilities. In order to improve the prediction accuracy for algal bloom risk of open channels, a risk assessment model for algal bloom of open channels was proposed based on the dynamic Na?ve Bayes classifier, with consideration to the uncertainty of the cause of algal bloom and sequential nature of its development. The risk grade nodes of the proposed model correspond to the concentration of chlorophyll a (Chla), and take into consideration 9 factors affecting the growth of algae. Network parameters were designed according to the results of expert consultation using the principal component analysis method. Based on the 53 cases of consecutive monitoring data observed from June 2011 to September 2011 at Beimen Bridge on Suzhou River, comparison was made between the proposed model and the assessment model based on Na?ve Bayes classifier. Confusion matrix results showed that the prediction accuracy for medium risks increased by 15.625%. Single tailed paired t-test showed that the recognition rates of the two models were significantly different when the significance level was 0.05. The assessment model based on dynamic Na?ve Bayes classifier with consideration to time sequence has significantly higher prediction and recognition rates for medium algal bloom risk of open channels.

国家自然基金项目（U1304702）；河南省科技厅软科学项目（152400410480）；河南工程学院博士基金（D2015030）