基于一种时间序列模型的河流重金属污染浓度预测研究

Forecast Study on Forecasting Pollutant Concentration of Heavy-metal Contaminants in Streams

水环境是一个充满不确定性的复杂巨系统,传统水质模型很难体现重金属污染物在河流中迁移的随机性,因此经典的时间序列模型——ARIMA模型被应用于河流重金属污染浓度的预测。实例分析证实,通过采用将获得的最新数据不断地添加到用于模型设定的样本中,并再此基础上获得最近向前一个时期预测值的动态预测方法,ARIMA模型能够获得很好的预测表现,尤其是在充分考虑模型残差统计分布特征的情况下,采用具有学生t分布的模型预测更精确。

Traditional stream water quality models are hardly able to describe stochastic behavior of heavy metal contaminants in water, due to stream environment influenced by various uncertainties. Therefore, a classic time series model, namely autoregressive integrated moving average （ARIMA） model, is used to predict pollutant concentration of heavy metal contaminants in streams. An empirical analysis evaluates the forecasting performance of two ARIMA models with different statistical distribution errors using a dynamic forecast approach. The results indicate that the two ARIMA models both perform very well, especially the one with student t distribution.