基于多层全连接神经网络的白洋淀水质预测

白洋淀是华北平原最大的天然湖泊湿地,也是雄安新区的重要生态依托,对其水质进行预测,可为其水质保护与管理提供重要依据。利用1996—2015年白洋淀南刘庄、圈头、烧车淀的生化需氧量(BOD)、化学需氧量(COD)、总氮(TN)和总磷(TP)4项水质数据,通过率定与校验,构建了全连接神经网络水质预测模型,对未来3 a白洋淀水质进行了预测。结果表明,白洋淀南刘庄、圈头、烧车淀的各项水质指标都有所改善,但部分点位TN和TP仍然超标。未来应加强入淀河流与淀内污染控制,强化生态补水与水系连通。

Water quality forecasting based on multilayer fully connected neural network for Baiyangdian Lake

Baiyangdian Lake, the largest shallow lake wetland in the North China Plain, is the most important ecological support of Xiong''an New Area. Forecasting water quality can provide important information for the environmental restoration and management of Baiyangdian Lake. This study used water quality data, including BOD, COD, TN, and TP values, obtained from the sub-lakes of Nanliuzhuang, Quantou, and Shaochedian between 1996 and 2015, to develop a fully connected neural network model for water quality predication. After model calibration and validation, we used the model to predict the changes in water quality over the next three years. The results show that all the water quality indices are improving in the sub-lakes of Nanliuzhuang, Quantou, and Shaochedian, but TN and TP still exceed their standards at some monitoring sites. In the future, we should strengthen pollution control in both the flowing rivers and Baiyangdian Lake, and increase environmental flow releases, as well as the hydrological connection between the rivers and lake.