滇池水污染物浓度预测的人工神经网络模型

Artificial Neural Network Model for the Prediction of Water Pollutant Concentration in Tien Lake

本文针对滇池日益严重的水污染现状，根据云南昆明西苑隧道断面2004年-2010年的监测资料，建立了基于BP神经网络的主要污染指标预测模型，并对其进行训练检验，研究结果表明：独立样本中pH、溶解氧、氨氮、高锰酸盐浓度的预测值与监测值的线性相关系数分别为0．952、0．967、0．945、0．936。结果证明该模型预测精度满足要求，通过准确地预测湖泊水污染物可以为治理湖泊营养化和综合利用水资源、规划管理、决策提供重要的科学依据。

Tien Lake is facing increasingly serious water pollution problem. According to the monitoring data of the Xiyuan part of Kunming city in 2004 - 2010, this study established a prediction model of the main water pollution indexes based on BP neural network and inspected the model. The research results showed that the correlation coefficient of pH, Dissolved Oxygen, ammonia nitrogen, permanganate concentration between predicted and measured values were 0. 952, 0. 967, 0. 945, 0. 936 respectively. The results indicated that the precision of the model prediction met the requirement and by predicting water pollutant values accurately could provide significant scientific basis for controlling lake eutrophication, making full use of the water resources, programming, managing and deciding.