The mission of this group is to bring together utility professionals in the power industry who are in the thick of the digital utility transformation. 

Post

Advances in artificial intelligence (AI) and machine learning coupled with smart water metering

image credit: Dreamstime.com
Dana Haasz's picture
Senior Director, Customer Success WaterSmart Software

Dana has almost 20 years experience in water conservation, utility program implementation, water resources and integrated planning, demand modeling, compliance assessment and policy. She...

  • Member since 2019
  • 1 items added with 2,174 views
  • Nov 27, 2019 1:00 pm GMT
  • 2175 views

This item is part of the Special Issue - 2019-11 - Artificial Intelligence, click here for more

Advances in artificial intelligence (AI) and machine learning coupled with smart water metering technology is bringing about a revolution in water utilities’ data collection capabilities that have enabled new ways to dynamically detect leaks and gain other critical insights into customer water consumption patterns. According to the EPA, household leaks can waste up to 1 trillion gallons of water annually.  Leaks costs everybody time and money. Customer’s water bills can increase dramatically, aside from any property damage caused by the leak. Utilities incur costs through increased customer service call volume, possible chargebacks, and field staff being dispatched to help customers find and resolve leaks.

Your access to Member Features is limited.

Data that WaterSmart Software has collected from over 4 million households indicates that as many as 50% of households will experience some type of water leak within a given year. And more than 10% of households have leaks that waste at least 90 gallons per day. In addition to the frequency that leaks occur, they can be quite expensive. The U.S. insurance industry pays out about $2.5 billion each year in homeowner insurance claims due to water damage from leaks. That’s nearly $7,000 per household which is the number two home insurance claim annually.

Water damage happens for a variety of reasons and some of these causes are unavoidable. Catastrophic weather events that lead to floods or broken pipes and leaky roofs can result in damage that is often expensive to repair. However, most leaks are more mundane and, if not entirely avoidable, easily addressed if caught quickly before extensive damage occurs. Common issues such as leaky toilets, worn faucet gaskets, frayed connection tubes, and irrigation system are easily missed by the naked eye but also simply fixed.

The trick is to catch the issue early and have a process by which to identify the likely source of the irregular water use. This is where data and technology come in to play. By actively monitoring water consumption through a modern data analytics platform, water customers can be automatically notified when changes in water use indicate a leak. This gives the end-user the opportunity to address the underlying problem before it has the opportunity to develop into a messy and costly hassle. The automated leak alerts WaterSmart has sent for its utility clients since 2015 have saved over 700 million gallons of water and unknown amounts to the customer in avoided property damage.

Watersmart’s unique dataset of over 326K alerted leaks and corresponding response data allowed for the development of a machine learning (ML) daily leak model using the XGBoost model in python. This Gradient Boosting model leverages that dataset along with each account’s hourly consumption and confirmation of leaks to develop a pattern-based model for Daily leak detection.

For all types of ML models in the Water Utility sector, keeping a robust training dataset is essential. Areas ripe for exploration are meter health and call volume data. These datasets have proven harder for utilities to export and tabulate, but data organization needs to be a stronger focus to leverage these AI and ML tools.

Given that most leaks are metered and paid for by the customer, it could be argued that water utilities don’t have an incentive to invest in helping their customers quickly identify leaks. But supporting customers costs water utilities a lot of money every year and Bill related issues are shown to account for up to 70% of all customer support calls. Every call to customer support costs between $3-7 (roughly $1 / minute), and unresolved calls, call escalations, and truck rolls for onsite service requests can drive these costs much, much higher.  Let alone the goodwill generated in helping customers avoid costs due to leaks and providing the type of customer support customers have come to expect from other service providers like their credit card or cell phone bill.  For both the customer and the utility, the value is clear.

The emerging use of smart meter technology is leading to a sea change in the way that water utilities manage customer-side leaks and communicate with end-users about those issues. By looking at regular interval of water consumption that is automatically streamed from the customer water meter, modern data analytics platforms can identify irregular usage patterns that could indicate a possible leak. These systems can then send automated messages to utility staff and customers that notify them of the change in water use. In its customer sentiment analysis, WaterSmart found that 97% of customers responded to the leak alerts they received from their utility as “positive”.  The most advanced of these leak notification systems also offer leak resolution wizards that walk customer through a step-by-step process to self-resolve issues, thus reducing calls to customer support services, and avoiding high bill surprises.  And as customers register the cause of their leaks, the utility can use this information to inform their water resources programs.

Leak detection has emerged as the killer application for smart water networks. While smart meters benefit utilities by reducing the costs of reading meters, they also provide end-use customers with detailed information on their water use (including leaks) that can help them better manage their spend and protect their property from costly water damage. These same data and communications techniques have been extended to other predictive features such as potential high bills and immediate notifications of unusual use, both which can be set my either the utility or the customer. The universe of support capabilities offered by the combination of recording technology AI and machine learning coupled with enhanced communication technologies offers an entire new relationship between water utilities and their customers and new tools in helping to manage a vital natural resource and ensure reliability and public health for the foreseeable future.

 

Dana Haasz's picture
Thank Dana for the Post!
Energy Central contributors share their experience and insights for the benefit of other Members (like you). Please show them your appreciation by leaving a comment, 'liking' this post, or following this Member.
More posts from this member
Discussions
Spell checking: Press the CTRL or COMMAND key then click on the underlined misspelled word.

No discussions yet. Start a discussion below.

Get Published - Build a Following

The Energy Central Power Industry Network is based on one core idea - power industry professionals helping each other and advancing the industry by sharing and learning from each other.

If you have an experience or insight to share or have learned something from a conference or seminar, your peers and colleagues on Energy Central want to hear about it. It's also easy to share a link to an article you've liked or an industry resource that you think would be helpful.

                 Learn more about posting on Energy Central »