A fuzzy logic based approach to leakage forecasting in water industry

Leakage and demand forecasting is a crucial part of the resource planning decision making in the UK water industry. OFWAT - the economic regulatory body - sets the annual leakage targets for 5 years in advance. However, the leakage is not constant throughout the year and depends on the leakage control policy applied, external factors and seasonal fluctuations. The leakage for each month of the year is forecasted and later adjusted every quarter as new data become available. In this paper, a new system is developed to help the expert in adjusting the leakage forecast based on the newly available data, as well as other factors used for leakage forecasting, such as Natural Rate of Rise (relates to the rate at which the leakage increases over time), efficiency measures (hours to detect the leakage) and the number of reported and detected leakages. Fuzzy clustering and a fuzzy interpolation method are applied to generate fuzzy if-then rules based on the recorded data. The fuzzy rules represent imprecise knowledge of the relationship between the above mentioned factors and leakages. The leakage forecast is then adjusted using the generated fuzzy rules. The proposed approach is verified using real-life data from one of the leading water supply companies in the UK