Leakage 2005 - Conference Proceedings Page 1 The Importance of Real Loss Component Analysis for Determining the Correct Intervention Strategy P Fanner*, J Thornton** *Fanner & Associates Ltd, 7 Brunswick Hill, Reading, RG1 7YT, UK Paul@FannerAssociates.com **Thornton International Ltd, Rua Arueira 370, Condominium Sausalito, SP 07600-000 Mairipora, Brazil thornton@water-audit.com Keywords: Background leakage; Component analysis; Infrastructure condition factor (ICF) Introduction Practitioners have for many years been trying to decide how best to rationalize and rank water loss intervention strategies. The advent of detailed water loss component analysis in conjunction with a water balance calculation allows the practitioner to better understand not only the volume of the water loss but also the nature of the individual components. Component analysis is used for strategic and economic analysis of both apparent and real losses, however this paper focuses on the real loss aspect. A better understanding of the nature of real losses such as: Background losses Reported losses Un-reported losses Allows for not just a band aid remedy, but rather a cause and effect analysis enabling the development of a sustainable solution and effective intervention strategy. A detailed component analysis model will allow analysis of: Separation of the volumes of annual water loss related to unreported leaks and bursts which are running unchecked in the system and would be remedied by traditional leak detection and repair contracts. The most appropriate repair time for reported and un-reported leaks and bursts. The volume of background leakage which will really only be effected by either pressure management or infrastructure replacement may also be analyzed and remedied in the most economic manner. The latter should not be confused with the other two volumes and must not be included in any potential benefit analysis for traditional leak detection and repair contracts. Practitioners have estimated background losses in various ways for many years however recent analysis shows how sensitive the component analysis is to error in the estimation of the background loss volume. This paper discusses component analysis models in detail and provides practical examples of how to best estimate the infrastructure condition factor (ICF) in order to make a meaningful separation of the components of background loss. The paper also provides a sensitivity analysis of a water system s background loss using different performance indicators and discusses the use of various performance indicators for comparison and for target setting.
Leakage 2005 - Conference Proceedings Page 2 Problem Statement Using the standardized International Water Association (IWA) top-down water audit methodology provides the best practice determination of the components of the water balance and in particular the volume of real losses. The IWA performance indicators determined using this volume of real losses are extremely valuable in allowing a utility to benchmark its real loss management performance with other utilities and to compare performance over time. However, once the volume of real losses has been determined, practitioners have to select an appropriate real loss reduction strategy and seek answers to questions such as: What is the most cost effective strategy for reducing real losses? How low can I reduce real losses? What is my economic level of leakage? What is my optimum operating pressure? How often should I undertake leak surveys? What sort of leak survey? What are appropriate levels of service for repairs? Because totally different strategies are required for each type of real loss, in order to answer these and other questions, it is necessary to break down the overall volume of real losses calculated from the top-down audit into its main constituent components: Background losses Reported losses Un-reported losses It is also necessary to understand where real losses occur in order to develop the appropriate loss reduction strategy. This requires an analysis of the volumes of real losses within the different components of the transmission and distribution system, such as: Transmission mains Distribution mains Valves Hydrants Other fittings Service lines Ferrule connections (main tap) Stop taps Meters Component analysis of the Awareness, Location and Repair (ALR) time and an analysis of the Infrastructure Condition Factor (ICF) are used, in conjunction with a topdown water balance, in order to evaluate the volumes of these components of real losses. The component analysis models when calibrated become the engine for economic analysis of the various intervention options.
Leakage 2005 - Conference Proceedings Page 3 Awareness, Location and Repair Time The volume lost through a leak is determined by the flow rate (at the current operating pressure) of the leak and the time it is running. The total run time of the leak can be broken down into three components: AWARENESS TIME (the average time from the start of the leak until the water utility becomes aware that a leak exists) LOCATION TIME (the average time it takes the water utility to locate the leak once it has become aware that a leak exists) REPAIR TIME (the average time it takes the water company to repair the leak after it has been located) FLOW RATE LEAK DURATION A L R TIME Figure 15 Leak volume = ALR Time x Flowrate Figure 15 illustrates these three components of the overall leak volume for any leak at the current system operating pressure and Figure 16 illustrates the build up of the leak volumes for three different types of typical reported bursts. Although in this example the flow rate of the mains burst is three times the flow rate of the service pipe bursts, it is located and repaired quickly. On the other hand, it takes longer for the utility to become aware of a burst on the utility section of the service pipe (communication pipe) and location and repair is not given the same priority as the mains burst. This results in a total leak volume almost five times the mains burst volume. A burst on the customer section of the service pipe (supply pipe) takes much longer to repair than on the utility section of the pipe because it is necessary to ensure that the customer makes the repair. The resulting leak volume is almost three times the leak volume on the utility section of the pipe. m 3 / day 75 1.1 Days REPORTED MAINS BURST 82.5 m 3 m 3 / day 25 16 Days A L R REPORTED COMMUNICATION PIPE BURST 400 m 3 m 3 / day 25 A L R 46 Days REPORTED SUPPLY PIPE BURST 1150 m 3 Figure 16 Time makes the difference A component analysis of ALR times uses typical flow rates for each type of leak together with utility data on leak run times for each type of leak to estimate the volumes of
Leakage 2005 - Conference Proceedings Page 4 real losses attributable to reported and unreported bursts on the various components of the transmission, distribution and service network. The following tables provide an actual example of a component analysis of ALR times in Salt Lake City Public Utilities Department (SLCPUD), a US utility that does not undertake any leak detection surveys. Table 5 shows the times assumed in the analysis for the utility to become aware of a reported burst and the average times taken from the work management system for locating and repairing bursts and leaks of different types. In this example, the data held in the work management system did not make it possible to obtain separate times for location and repair, so for this utility, these two components have been evaluated together. Table 5 also includes the average system pressure, which for this system is quite high at 71.7m head, and the ICF. The ICF was not determined from measurements, but estimated based on discussions with SLCPUD staff. As this was only an estimate, the sensitivity of this estimate is explored later in the paper. Table 5 Assumed awareness times and location and repair times from work management system Reported Leaks and Bursts (A) Work Order System Waiting Time (LR) bursts 0.5 days breaks > than 300mm 2.4 days leaks 5.0 days breaks = < than 300mm 0.5 days Active service 14 days leaks>300mm 2.0 days Abandoned service 180 days leaks <300mm 5.0 days Hydrants and Valves 14 days Services SLC repaired 11.1 days Unreported Leaks and Bursts (A) Services Customer repaired 70.0 days One Leakage Survey every months Hydrants and valves 86.4 days Average awareness time = 0.0 months System variables Average awareness time = 0.0 days Average pressure 71.7 m head ICF 1.50 Table 6 illustrates the determination of average annual loss for leaks and bursts on the individual components of the distribution system, taking into account the variation in the FAVAD exponent of the pressure leakage relationship, N1 with different pipe materials. The typical flow rates used in this table were derived from work in another US utility. (Analysis undertaken using FastCalc). Table 6 Calculation of real losses from reported bursts for different distribution system components Assumed Typical Annual FAVAD N1 Average Ave. Length % of Flow Average Average Average Reported or number Frequency for this Location & Annual (km) or metal rate @ Awareness duration Ml lost Unreported of per 1000 type of repair / shut off Loss number pipes 50m (hours) (hours) per leak bursts burst or (hours) Ml/year l/sec leak <75mm Reported 3 29 0.1033 93% 0.88 0.57 12 58 70 0.27 0.8 75mm Reported 7 0.0000 100% 0.88 0.50 12 58 70 0.26 0.0 100mm Reported 44 178 0.2472 92% 2.78 0.58 12 58 70 0.86 37.8 150mm Reported 209 1036 0.2017 78% 5.80 0.72 12 58 70 1.89 394.5 200mm Reported 32 273 0.1171 65% 5.80 1.50 12 58 70 2.50 80.1 250mm Reported 35 0.0000 68% 5.80 1.50 12 58 70 2.50 0.0 300mm Reported 300 0.0000 54% 14.01 1.50 12 12 24 2.08 0.0 >300mm Reported 287 0.0000 68% 14.01 1.50 12 12 24 2.08 0.0 Hydrants Reported 39 79994 0.0005 100% 0.22 0.50 336 2074 2410 2.30 89.5 Valves Reported 14 63 0.2205 100% 0.44 0.50 336 2074 2410 4.53 63.4 Service <25mm Reported 536 95 5.6710 100% 0.44 0.50 336 266 602 1.13 606.5 Services >25mm Reported 50 1428 0.0350 100% 0.88 0.50 336 266 602 2.28 114.0
Leakage 2005 - Conference Proceedings Page 5 Table 7 illustrates the calculation of the components of background losses in the system, taking into account the average system pressure and the assumed ICF value. Background Leakage on Transmission mains Background Leakage on Distribution Background Leakage on Hydrants Background Leakage: Services Table 7 Calculation of background losses Length (km) or Number IWA values at 50m Assumed FAVAD N1 Value Infrastructure Condition Factor ICF Average pressure (m) Background Leakage Calculation m3/hr @50m m3/hr m3/day Ml/yr 287 20.0 1.50 1.50 71.7 9 15 356 129.8 1859 20.0 1.50 1.50 71.7 56 96 2304 841.0 8987 1.14 1.50 1.50 71.7 17 29 696 254.0 90662 1.25 1.50 1.50 71.7 170 293 7020 2562.5 Table 8 summarizes the results of the analysis of losses from reported bursts and background losses, using an assumed ICF of 1.5. The total estimate of real losses determined from the component analysis is compared with the estimate of real losses from the top-down water balance, with the balancing error calculated as the difference between these two estimates. In this example, the term balancing error in the table is misleading because no estimate has been made of losses from unreported bursts as there is no data on the numbers of unreported bursts in the system or their run times. For this reason, the volume of real losses calculated from a component analysis of reported bursts and background losses would be expected to be lower than the volume of real losses calculated from the top-down water balance. Table 8 ALR analysis of the components of real losses using an ICF of 1.5 Background Reported Unreported Total Percent Ml/yr Ml/yr Ml/yr Ml/yr of total Reservoirs 0 0 0 0 0.0% Trans. 130 1 0 131 2.5% Dist. 841 513 0 1354 26.2% Services 2562 720 0 3283 63.4% Hydrants & Valves 254 153 0 407 7.9% Totals 3787 1388 0 5175 100.0% WATER BALANCE BEST ESTIMATE TO +/- 47.8% 6029 Ml/year BALANCING ERROR = 854 Ml/year It may be seen in Table 8 that confidence is low in the actual volume of real losses from the water balance. This is due to the lack of confidence in some of the system input metering components. The highest estimate of real losses from the water balance is 8,915 Ml/year. The estimated volume of real losses in the component analysis is sensitive to the value used for the ICF. Table 9 summarizes the results of an analysis of the sensitivity of the volume of real losses calculated from the component analysis to the assumed value of ICF, with the volume of real losses compared with both the best and highest estimates of real loss from the water balance. Table 9 Sensitivity of volume of real losses in component analysis to assumed value of ICF
Leakage 2005 - Conference Proceedings Page 6 Assumed Value of ICF Real Loss Volume from BABE Analysis (Ml/year) Best Estimate of Real Loss Volume from Water Balance (Ml/year) Difference (Ml/year) Highest Estimate of Real Loss Volume from Water Balance (Ml/year) Difference (Ml/year) 1.00 3914 6030 2116 8915 5001 1.25 4542 6030 1488 8915 4372 1.50 5175 6030 856 8915 3740 1.75 5807 6030 223 8915 3108 2.00 6439 6030-409 8915 2476 2.25 7067 6030-1037 8915 1847 2.50 7700 6030-1669 8915 1215 2.75 8332 6030-2302 8915 583 3.00 8964 6030-2934 8915-49 From this analysis, it can be seen that, even if there were no unreported bursts in the system, which is highly unlikely, the ICF must be less than 2.0 if the best estimate of real losses from the water balance is used and less than 3.0 if the highest estimate is used. Values of ICF higher than these values result in a negative difference with the water balance estimate of real losses. It is evident from this analysis that it is important to determine the actual value of the system ICF in order to improve on the confidence of the component analysis. There was no provision within the Salt Lake City water balance project to undertake measurements to determine the average ICF of the SLCPUD system, but how could the ICF value have been estimated? There are three possible approaches for estimating the ICF: Determine from leak detection and repair work within District Metered Areas (DMA) and extrapolate for whole system Determine from N1 pressure step tests and extrapolate for whole system Estimate from ILI These three approaches are described in the following sections and their relative merits discussed. Determining the ICF from DMA This method requires all recoverable leakage to be removed from the district and a measurement made of the lowest achievable (background) leakage. The ICF is the ratio between the actual level of background leakage in a district and the calculated IWA unavoidable background leakage for a well maintained system. The IWA unavoidable background leakage is calculated for the district using the factors in Table 10.
Leakage 2005 - Conference Proceedings Page 7 Table 10 Unavoidable background leakage rates Infrastructure Component Background Leakage at ICF=1 Units 9.6 Litres per km mains per day per metre of pressure Service Connections main to property boundary 0.6 Litres per service connection per day per meter of pressure Service Connections property boundary to customer meter 16.9 Litres per km of service connection per day per meter of pressure Source: IWA Water Loss Task Force In order to determine the weighted average ICF for the whole system it is necessary to undertake this analysis in a representative sample of DMA across the system. This is a significant amount of work, particularly for utilities that currently do not use DMA. The benefit of this method is that it provides the only method of reliably determining the ICF for many utilities. Determining the ICF from N1 Pressure Step Tests This method also requires DMA or small pressure zones with closed boundaries and a single feed point. During the test, the supply pressure is reduced either by throttling a valve, or by operating a PRV, or even by controlling a suitable pump in a pumping station, near to the zone inlet point (the later two methods provide more stable results). The relationship between leakage and pressure can be summarized by the following equation: L is proportional to P N1 where L represents Leakage Rate, P represents Average System Pressure and N1 is the power law exponent for the relationship. For fixed size holes, such as bursts, N1 will be 0.5 (square root relationship) whilst for variable sized holes, such as background leakage at joints and splits in plastic pipes, N1 will be 1.5. The inflow to the zone and the pressure at the average pressure point in the zone are logged while the pressure at the inflow to the zone is reduced in a series of steps during the stable portion of the minimum night flow period when consumption is at a minimum. Each step needs to be of sufficient length, (usually around 20 to 20 minutes) for the inflow to become stable at the new inflow pressure. The data is then analyzed and estimated consumption accounted for in order to determine the effective areas of leakage in the district and to compare the change in effective areas caused by the change in pressure. From this comparison, it is possible to estimate the N1 value and to estimate the ratio of fixed size holes (bursts) and variable sized holes (assumed to be background leakage). By plotting the average zone pressure against the effective leakage area, the slope of the line gives the N1 value and the intercept on the Y axis gives the effective area of the fixed size holes (bursts). This can then be used to determine the ratio of fixed sized holes (bursts) to variable sized holes (background leakage) at the initial pressure at the start of the test.
Leakage 2005 - Conference Proceedings Page 8 The ICF can be determined from the calculated volume of background losses in the DMA. Figure 17 illustrates the calculation of the ICF in a zone from an N1 pressure step test. This method can only be used on rigid systems where the N1 of bursts is around 0.5, i.e. on systems with metallic pipe work. (other-wise variable area bursts may be confused with background leakage and the ICF value overestimated) In order to determine an overall system weighted average ICF value, it is necessary to carry out pressure step tests in a representative sample of DMAs. It should also be remembered that following a leak detection and repair exercise to remove bursts from within a DMA, the value of N1 will change and a further pressure step test should be carried out for the DMA. This method also entails significant field work, but less than the first method discussed. Estimating the ICF from the ILI It has been observed that in many systems the value of the ICF is found to be similar to the value of the ILI. Therefore in the absence of any other data, a reasonable starting assumption is to use a value for the ICF based on the ILI. However, it should be remembered that this relationship is not seen in all systems and with high values of ILI it is probably not an appropriate assumption.
Leakage 2005 - Conference Proceedings Page 9 Zone Ref. West Roxbury / Roselindale Date 24-May-02 Zone Name Boston Domestic Properties Non-Domestic Properties Total Properties Connections Length (km) Connection Density Domestic Night Use Non-Domestic Night Use Total Assessed Night Use 8199 nr 1 nr 8200 nr 8200 nr 126.491 km 65 per km 2.79 l/conn/hr 6.35 l/s 821.2 l/conn/hr 0.23 l/s 6.58 l/s Time Start Time End Average Zone Pressure Total Inflow Physical Losses Calculated Values of N1 for Physical Losses (hr) (hr) (m) (l/s) (l/s) Stage 1 Stage 2 Stage 3 Start 02:00 02:10 51.77 64.43 57.85 Step 1 02:21 02:31 49.07 61.65 55.07 0.92 Step 2 02:50 03:00 45.72 58 51.42 0.95 0.97 Step 3 03:13 03:23 40.25 55.76 49.18 0.65 0.57 0.35 Step 4 03:34 03:44 37.93 54.05 47.47 0.64 0.58 0.43 0.60 Step 5 AVERAGE N1= 0.66 Effective Area (mm2) 2000 1500 1000 500 Pressure Step Test Zone: West Roxbury / Roselindale - Boston 0 0.0 10.0 20.0 30.0 40.0 50.0 60.0 Average Zone Pressure (m) IF Intercept = 1563 mm 2 Average Zone Pressure = 51.77 m Physical Losses = 57.85 l/s THEN Bursts = 49.83 l/s Background Leakage = 8.02 l/s IWA International Lowest Achieveable Background Leakage Rates with Infrastructure Condition Factor (ICF) = 1.0 Background Leakage on Service Connections to Curb Stop 0.60 l/conn/d/m Background Leakage on 9.6 l/km/d/m Background Leakage on Serv Conn from Curb Stop to Meter 0.25 l/conn/d/m (assumes average of 15m service line) Lowest Achieveable Background Leakage Rate for this Zone (ICF=1.0) Background Leakage on Service Connections to Curb Stop 10,613 l/hr Background Leakage on Serv Conn from Curb Stop to Meter 4,484 l/hr Background Leakage on 2,619 l/hr Lowest Achieveable Background Leakage Rate Actual Background Leakage Rate for this Zone Background Leakage (from Pressure Step Test) Actual Background Leakage Rate = 17,716 l/hr 2.16 l/conn/hr 8.02 l/s 28,877 l/hr 3.52 l/conn/hr INFRASTRUCTURE CONDITION FACTOR = 3.52 l/conn/hr 2.16 l/conn/hr = 1.6 Figure 17 Determination of the ICF in a zone from an N1 pressure step test
Leakage 2005 - Conference Proceedings Page 10 IWA Performance Indicators (PI) for Real Losses The IWA PI for real losses are used to benchmark a utility performance and to set targets for improved performance. However the most commonly used indicators OP24 (volume of real losses per connection per day) and OP25 (the ILI) are designed for use with the top down water balance and consider the whole annual volume of real losses and not the individual components as discussed in this paper. In order to properly set targets for efficient and economic intervention against real losses it is important to go one step further and understand what makes up the whole. For example: Example; A utility may have an OP24 PI of 100 litres per connection per day, but what makes up the 100 litres? The following questions might be asked: Is the majority on mains, services or on reservoir leakage? If it is on mains or services what volume is potentially recoverable using leak detection methods or improved levels of service for repair? What portion of the volume is only recoverable by pressure management or infrastructure replacement? If for example the break down was 20 litres per connection per day on mains, 80 litres per connection per day on services and nothing on reservoirs then the utility certainly would not want to invest in reservoir maintenance to reduce the loss. Taking the example further if 15 litres of the 20 were background losses and 70 litres of the 80 were background losses then the utility would likely need to decide on either a pressure management program and or an infrastructure replacement program and not a leak detection and repair program. If they opted for an infrastructure repair program then clearly a service replacement program would offer better returns than the mains replacement. In Table 8 from the Salt Lake City example above, it may be seen that even with an assumed ICF of only 1.5, over 73% of the overall volume of real losses in the component analysis are due to background losses. This high level of background losses is a direct result of the high average system pressure. An ICF of 1.5 does not indicate that the infrastructure is in a poor condition, therefore an infrastructure replacement program is not the appropriate solution in this case, but background losses could be reduced significantly through the use of a pressure management program. Of the reported bursts in the Salt Lake City example, the greatest volume of real loss is from service line bursts. Examination of the build up of this component in Table 6 shows that this volume of loss could be significantly reduced by reducing the average repair time for service line bursts from the current level of 11.1 days. None of these conclusions could have reached without the benefit of a component analysis model, which should be a standard tool of the water loss management practitioner. Summary Component analysis allows the utility to properly understand why they have the PI they have and which components of the system are responsible for the loss. The OP24 volumetric indicator can be broken down to better identify where the gains lie. Once the gains have been volumetrically identified an economic business case can be designed and implemented to reduce the losses and maintain the losses at economic levels.
Leakage 2005 - Conference Proceedings Page 11 However, component analysis is sensitive to the value of ICF used in the analysis. For this reason measurements should ideally be undertaken to determine the ICF of the system infrastructure prior to undertaking a component analysis. Acknowledgements The authors would like to acknowledge SLCPUD for allowing the use of their data in this paper. References Lewis, J.M. and Fanner, P.V. (2005) Experience of using the IWA/AWWA water audit methodology in Salt Lake City Public Utilities Public Utilities Department, IWA conference Leakage 2005, Halifax, Nova Scotia. McKenzie, R and Lambert, A.O. (2004) Best Practice Performance Indicators: A Practical Approach, Water21, August 2004 Kunkel, G. (2003) Worldwide Best Management Practices in Water Loss Control, AWWA Water Loss Control Committee Report, AWWA Journal, August 2003