Document Type

Presentation

Publication Date

8-1-2014

Abstract

An approach based on time series clustering and Support Vector Regression (SVR) is proposed to characterize typical daily urban water demand patterns and use them to perform short term forecasts. Time series clustering is aimed at retrieving the hourly quantity of water supplied during each day within a specific time window (e.g. last one or two years). A suitable distance measure has been considered to capture differences in shape and volume between time series. The final goal is to group similar daily water demand curves and then summarize them in correspondent “prototypal” patterns. A post processing step is then performed to identify the relationship linking each prototypal pattern to a specific period of the year and/or type of day (e.g. week-end, holiday, working day), capturing periodic behaviors at different time level. Urban demand forecasting, in the short term (today or tomorrow), is performed in two steps: first, the prototypal pattern that is expected depending on the period of the year and type of the day is proposed as a first prediction. In a second step, the hourly water supplied up to the very early morning (e.g. 06:00 o’clock) is used to predict, more accurately, and in one time, the expected hourly urban demand for the entire day. One SVR is trained for each hour of the day and for each prototypal pattern, by using the time series in the correspondent cluster. The approach has been validated through leave one out validation, showing high prediction reliability. The proposed approach enables a deeper understanding of the periodic urban water demand variations as well as the reduction of operational costs (e.g., by optimizing caption, treatment, storage and distribution) without the introduction of time-lag such as for other techniques (e.g. ARIMA)

Comments

Session R50, Water Distribution Networks: Demand Forecasting

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.