Determining leakage by analysing demand changes over time
Author: Peter van Thienen, Ilse Pieterse-Quirijns, Jan Vreeburg, Karel Vangeel and Zoran Kapelan
Leakage continues to be a problem for water companies around the world, with water lost ranging from 3% to more than 50% of distribution input. The water lost this way represents a financial loss, but is also undesirable from the environmental, social and sustainability point of view. The two principal methods to determine the amount of Non-Revenue Water (NRW) in a supply area are the top-down method and the bottom-up methods, both with considerable degree of uncertainty. Several new, alternative methods have been developed which make use of optimization based inverse type models. These model optimization based methods generally require a hydraulic model of (or information about the system in) the supply area and can be computationally demanding.
In this paper, several applications of a novel method called the Comparison of Flow Pattern Distributions (CFPD) are presented. This method allows the user to compare flow patterns of arbitrary duration for an arbitrarily sized supply area and distinguish consistent from inconsistent changes in the pattern. The so called consistent changes can be interpreted in terms of changes in demand due to changes in the population characteristics (growth or shrink in longer term, holiday periods in shorter term). The so called inconsistent changes can be interpreted in terms of new large volume customers, new types of water use and/or change in leakage. As the water companies have (access to) information about the first two changes, the method allows quantitative statements to be made about the third item, i.e. leakage. The method presented here is relatively simple, not computationally intensive, independent of any model assumptions and easily implemented. The automated application of this method on long time series, called CFPD block analysis, is used on data from three different areas. These applications illustrate, respectively, the identification and pinpointing (in time) of a small leak, the independent quantification of concurrent different types of changes with opposite signs in a supply pattern, and the difficulties of interpretation in cases where climate has an overwhelming influence on demand patterns. In each case, the power of the CFPD block analysis is illustrated: discriminative quantification and visualization result in features and trends in complicated time series becoming apparent at a glance.