Kim J R, Ko J H, Im J H, Lee S H, Kim S H, Kim C W, Park T J
Dept of Environmental Engineering, Pusan National University, Busan, 609-735, Korea.
Water Sci Technol. 2006;53(4-5):185-92. doi: 10.2166/wst.2006.123.
The information on the incoming load to wastewater treatment plants is not often available to apply modelling for evaluating the effect of control actions on a full-scale plant. In this paper, a time series model was developed to forecast flow rate, COD, NH4(+)-N and PO4(3-)-P in influent by using 250 days data of field plant operation data. The data for 150 days and 100 days were used for model development and model validation, respectively. The missing data were interpolated by the spline method and the time series model. Three different methods were proposed for model development: one model and one-step to seven-step ahead forecasting (Method 1); seven models and one-step-ahead forecasting (Method 2); and one model and one-step-ahead forecasting (Method 3). Method 3 featured only one-step-ahead forecasting that could avoid the accumulated error and give simple estimation of coefficients. Therefore, Method 3 was the reliable approach to developing the time series model for the purpose of this research.
通常无法获取进入污水处理厂的负荷信息，以便应用模型来评估控制措施对全尺寸工厂的影响。本文通过使用现场工厂运行数据的250天数据，开发了一个时间序列模型来预测进水的流量、化学需氧量（COD）、铵根离子（NH4(+) - N）和磷酸根离子（PO4(3-) - P）。150天和100天的数据分别用于模型开发和模型验证。通过样条法和时间序列模型对缺失数据进行了插值。提出了三种不同的模型开发方法：一个模型和提前一步到七步预测（方法1）；七个模型和提前一步预测（方法2）；以及一个模型和提前一步预测（方法3）。方法3的特点是仅提前一步预测，可避免累积误差并给出简单的系数估计。因此，方法3是为本研究目的开发时间序列模型的可靠方法。
