摘要
为验证BP模型在河流水质预测中的有效性,利用仿真技术模拟一条河流污染物的变化趋势,并得到大量的河流水质参数数据.用上游已知河段的水质数据预测该河流下游10个检测断面的水质状况.预测过程分2种情况进行:长距离预测(一次连续预测下游10个河段)和短距离预测(每次连续预测下游2个河段),并以MSE函数生成均方误差作为对2种预测方法性能的检验.结果显示,长距离预测的性能低于短距离预测,2种方法对溶解氧预测的均方误差为0.432和0.035,对生化需氧量预测的均方误差分别为0.243、0.055.
The aim of this study is to describe Error Back Propagation approach that can be used to forecast the water quality. In order to check the validity of the approach, a hypothetical field data as a case study were produced by water quality simulation of a river. ANN was trained by using the upriver data in a river to predict the water quality of lower reaches of the river, which can be predicted by two methods namely long- distance prediction and short-distance prediction. Trying to analyze the validity of two methods using the MSE function in MATLAB 7 indicated the predicted results from short-distance approach were more accurate than that of long-distance prediction. The mean square errors of Dissolved Oxygen were 0. 432 for long- distance prediction and 0. 035 for short-distance prediction while the mean square errors of Biochemical Oxygen Demand were 0. 243 and 0. 055.
出处
《环境科学学报》
CAS
CSCD
北大核心
2007年第6期1038-1042,共5页
Acta Scientiae Circumstantiae
基金
国家自然科学基金资助项目(No.50378008)~~
关键词
BP神经网络
水质预测
河流仿真
检验
BPANN
water quality prediction
river water quality simulation
testing