每月平均日計費計量水量預測模型之建立

Developing A Forecasting Model for The Daily Averaged Billed Metered Consumption of Every Month

姚榮昇(Jung-Sheng Yao)

指導教授 ： 張哲豪

國立臺北科技大學/工程學院/土木與防災研究所/碩士(2014年)

https://doi.org/10.6841/NTUT.2014.00715

摘要

2003年國際水協會發表「標準水量平衡表」統一各項水量定義，為了與國際接軌，臺北自來水事業處根據該水量定義，並為瞭解每個月包含生產、營業、財務三方面的產銷狀況，將配水量、計費水量與水費收入列為主要管理數據。其中計費水量又以「計費計量水量(BMC)」，直接顯示自來水事業的營業情形，更須要能夠即時反映當下現況。在臺北自來水事業處，因人工抄表人力限制，當月的抄表收費水量僅代表過去1至3個月期間的BMC，無法顯示當月BMC。本研究擬以時間序列分析，研發BMC之預測模型，由當月抄表收費水量來計算推估當月BMC。研究中同時結合「平均水量分配法」，根據抄表期間的用水日數重新分配，得到合理的「每月平均日BMC」。採用IBM SPSS軟體，建立「每月平均日BMC」時間序列，在參數精簡原則及觀測數量不得少於50筆之原則下，以91年8月至101年7月「每月平均日BMC」共120筆觀測值建立的ARIMA(4,1,2)時間序列模型，其R平方值最接近1且BIC準則值最小。該ARIMA(4,1,2)模型在不同時段的預測結果，經比對觀測值，準確率皆可達97.29%以上，最小偏差率更只有0.01%。

關鍵字

計費計量水量 ； 時間序列分析 ； 預測模型

並列摘要

2003 International Water Association published a "standard water balance" harmonization of the definition of water. To international standards. Taipei Water Department is defined under the water. To understand the cause of each month includes water production, sales, finance and marketing conditions in three items. The system input volume、revenue water and water income management as the main data. In revenue water , " Billed Metered Consumption（BMC）" Show directly business situation in water utilities. More need to be able to reflect the current status of real-time. In Taipei Water Department, due to labor’s limitation of manual meter reading. Water meter charges on behalf of BMC in the past month only 1-3 month period. Can not display the status of the month. This study intends to Time Series Analysis. BMC's prediction model developed. Water meter charges by the month to calculate the estimate of the month BMC. Research and Application "Average water allocation method". According to the number of days of water meter reading did average and reallocate, and got "Monthly Averaged Daily Billed Metered Consumption". Using IBM SPSS software to create "Monthly averaged Daily BMC" Time series. Under the principle of parameters that needed more fewer and the principle of not less than the number of observations of 50. Established ARIMA (4,1,2) time series model of a total of 120 "Monthly averaged Daily BMC" observations In August 2002 to July 2012. R-squared value was the closest to 1 and the number of BIC criteria was the Minimum. Analysis ARIMA (4,1,2) model of the predicted influence at different times. After comparing observations, the accuracy of the predicted value was up to 97.29% and even Minimum deviation rate was only 0.01%.

並列關鍵字

Billed Metered Consumption ； Time Series Analysis ； Forecasting Model

參考文獻

14.周海龍，劉祥熹，台北都會區生活用水量之模式建構與模擬－類神經網路模式之應用，農業經濟半年刊，71期，2002，第77-112頁。

33.臺北市政府民政局入口網站，人口統計，

34.新北市政府民政局入口網站，人口統計，

10.Zhou, S., McMahon, T., Walton, A., and Lewis, J., “Forecasting daily urban water demand: A case study of Melbourne.” Journal of Hydrology,Vol. 236, Issues 3–4,2000, pp.153-164.

12.Jasem M. Alhumoud, “Freshwater consumption in Kuwait: Analysis and forecasting.”Journal of Water Supply Research and Technology－AQUA”, Vol.57, Issue 4, 2008, pp.279-288.