The San Francisco Public Utilities Commission (SFPUC) provides drinking water, green power, and wastewater management to the third largest city in California. The agency’s Hetch Hetchy Regional Water system provides drinking water to approximately 832,000 residents San Francisco, in addition to millions more through wholesale agreements with 26 water agencies in Alameda, Santa Clara, and San Mateo counties. Leaks can occur at any type of property throughout the system, and the US Environmental Protection Agency estimates the average household leak can account for nearly 10,000 gallons of wasted water annually.
For more than a decade, SFPUC has used advanced metering infrastructure (AMI) to collect hourly water meter readings from all retail customers. This technology has reduced the work hours previously required for meter reading and improved accuracy of water usage and billing, while also allowing SFPUC to automatically monitor water use for potential leaks. In a typical residential home, usage data drops during the middle of the night when most residents are sleeping. With hourly usage data transmitted from the meter directly to SFPUC systems, account holders for single family and small multi-family homes in San Francisco are automatically alerted of a potential leak if usage is consistently more than one cubic foot (about 7.5 gallons) every hour in a two-day period.
A few years ago, SFPUC wanted to upscale its leak detection program beyond residential and small multi-unit buildings to include approximately 10,000 multi-family homes with six or more residential units and 17,000 commercial, industrial, and institutional (CII) accounts. However, following the same guidelines as small residential units for these larger buildings that serve a myriad of functions simply did not work. SFPUC hired Woodard & Curran to help develop a system of detecting possible leaks for large multi-family buildings and CII accounts with more certainty.
Given that large multi-family buildings and CII accounts constitute 54 percent of San Francisco’s water consumption, there was a massive amount of data to analyze and determine water use trends. Unlike single family or small multi-unit dwellings, these water customers have varied, and in some cases continuous 24/7 use as a part of the normal course of business. For example, hotels make up approximately 6 percent of SFPUC’s commercial accounts, but use more than a quarter of total commercial consumption due to extensive laundry, dishwashing, cooling, and kitchen operations, amenities such as pools and restaurants, and large number of overnight guests. While constant use overnight at a single family home or small apartment building raises a flag, that’s not always the case for large accounts and the amount of water that may be lost from a running toilet is a small fraction of the water being used for regular activities.
Woodard & Curran partnered with local civil engineering firm Lotus Water to collect, analyze, and report on water use trends for SFPUC’s large retail water service customers. After identifying continuous use events, numerous additional factors were evaluated to determine what was indicative of a leak, including the duration and volume of each event, minimum/maximum hourly consumption on the first day of an event, and various ratios of regular use compared to potential leak use.
This data review and analysis helped Woodard & Curran’s water resources experts identify criteria for SFPUC to use for detecting leaks for large multi-family buildings and CII accounts. The first criterion takes the average number of hours of zero consumption from 1:00 to 4:00 a.m. over a 90-day period. This helps sort accounts into two categories – those that tend to have no consumption overnight and those with continuous use comparable to other hours of the day – for further evaluation. The second criterion is the ratio of consumption from 1:00 to 4:00 a.m. on the last two days divided by the average use in the same hours within the prior 90 days. Spikes in the ratio indicate an increased water usage during typically low-usage hours.