几种AR和ARMA模型对城市用水量预测的比较

Comparison of Several AR and ARMA Models for Urban Water Consumption Prediction

利用时间序列法中的自回归模型(AR)和自回归移动平均模型(ARMA),对城市某区域的用水量进行预测。模型参数采用最小二乘法进行估计,并将预测值与实测值进行了比较。结果显示,预测值的相对误差均在±10%之内,且仅有个别点的相对误差在±5%之外,预测结果与实测值基本一致。对预测结果的进一步分析表明,高阶AR模型预测结果的均方差(MSE)低于低阶AR模型,而从平均绝对百分比误差(MAPE)来看,ARMA模型的预测结果并不优于AR模型。

Water consumption of a certain area in a city has been predicted using AR and ARMA models, which belong to time series analysis. The parameters of the models were estimated with Least Squares Method. The predicted hourly and daily water consumption has been compared with the actual data. It shows the relative errors of the predicted data are all within ±10%, and very few out of the range ±5%. The predicted data are in good agreement with the actual water consumption.Further analysis of the predicted results indicated the MES of high order AR model prediction is less than that of low order AR model prediction, while from the MAPE ARMA model prediction is not always better than that of AR models.