Machine Learning Times
Machine Learning Times
Podcast: Four Things the Machine Learning Industry Must Learn from Self-Driving Cars
    Welcome to the next episode of The Machine...
A Refresher on Continuous Versus Discrete Input Variables
 How many times have I heard that the most...
Podcast: Why Deep Learning Could Expedite the Next AI Winter
  Welcome to the next episode of The Machine Learning...
PAW Preview Video: Evan Wimpey, Director of Strategic Analytics at Elder Research
 In anticipation of his upcoming presentation at Deep Learning...

9 years ago
Using Analytics to Crack Down on Electricity Theft


Electricity theft is estimated to cost the energy industry approximately $6 billion each year. Analytics can help them recoup a significant portion of those losses.

People go to all kinds of lengths to steal electricity. Some simply turn their meters upside down so they run backwards. Others break into their meters and pour glue over the dials to gum them up and slow down readings. More daring (or desperate) electricity thieves use jumper cables to bypass the meter—exposing themselves to live, high voltage electricity in the process. And those who really wish to flirt with danger tap directly into power lines.

In many cases of meter tampering and electricity theft, perpetrators risk burns, electrocution, and even death. Those motivated to take the risk include some homeowners and renters, individuals who operate energy intensive businesses like laundromats and car washes, and indoor marijuana cultivars.¹

Despite the hazards, stealing electricity ranks as the third largest form of theft in the U.S., according to utility Pepco, behind shoplifting and copper theft. Various estimates peg the cost of electricity theft to the U.S. energy industry at around $6 billion per year.² Individual utilities, including Austin Energy and Ameren Missouri, have reported losing millions of dollars in stolen electricity.

“Electricity theft counts among the top three challenges facing utilities,” says Scott Barnes, a director with Deloitte Consulting LLP’s Power & Utilities sector. “And the thieves are getting more sophisticated.”

Indeed, Creighton Oyler, vice president of utility analytics at Oracle, describes teams of “meter fixers” who come through various neighborhoods, knocking on doors and promising to reduce residents’ electric bills by, say, 30 percent per month. “For a fee, they’ll either modify residents’ meters or swap them out for tampered ones,” he says.

Do-it-yourselfers can buy stolen or tampered meters online, along with the keys and “tamper-proof” seals for those meters, according to Oyler. By applying purchasing seals to tampered meters, electricity thieves hope their illicit activity will go unnoticed.

Just as electricity theft has grown in sophistication so, too, has utilities’ ability to detect it. In the past, utilities had to rely on reports from meter readers and tips from customers. Today, more utilities are using analytics to identify consumption patterns that indicate theft.

Barnes says market forces have compelled increasing numbers of utilities to turn to analytics. Over the past three years, he notes, utilities have seen their expenses rise with the cost of electricity. Meanwhile, consumption has decreased due to conservation efforts. Because utilities are so heavily regulated, they can’t simply raise their rates to make up the difference. (They have to propose fairly elaborate “rate cases” to regulators to justify increases.) Consequently, Barnes adds, they’re turning to analytics to try to drive out costs and recoup some of the revenue they’ve historically lost to theft.

Smart meter adoption has also encouraged some utilities to use analytics to identify theft, according to David Trudel, a senior manager with Deloitte Canada who’s worked with energy industry clients on fraud detection analytics projects. Since smart meters automatically send their readings to the utility, they no longer require a technician to read them. In the absence of technicians conducting monthly checks, utilities have had to explore new ways to spot theft.

David Steiger, a senior manager with Deloitte Consulting LLP’s Power & Utilities sector, says the revenue utilities can recover by getting customers to pay for electricity they’d siphoned can cover the cost of expanding their smart meter investments.

Trudel says utilities need to bring three fundamental technology layers together to do analytics: transaction systems, like billing and customer service; a data warehouse where information on electricity consumption and payments can be aggregated and indexed; and a reporting system.

“The architecture is fairly simple, but it requires a tremendous amount of processing power,” says Trudel.

The analytical models work by combining weather patterns, billing and payment information, data on household and neighborhood consumption, and more. One utility that developed an analytics-based fraud detection application expects to collect several million dollars over the next five years. The revenue collected will more than cover the cost to develop the application.

While more utilities are using analytics to identify electricity theft, the practice is not yet widespread for a variety of reasons. Barnes says many utilities lack the information management infrastructure required to handle and compare large volumes of generation, consumption, and billing data. And the financial investment required to implement that infrastructure may be too burdensome for some utilities, adds Steiger. Siloed systems also make it hard to access the data needed to build analytical models.

Lack of analytical skills and experience building mathematical models presents another barrier. Steiger notes that developing those models requires an agile approach, where business users identify a problem they wish to solve and the data required to solve it, then work with developers in a rapid, back-and-forth style to create and refine a prototype. Steiger says the agile methodology can be difficult for utilities accustomed to the traditional “waterfall” development model to embrace.

There are also cultural issues. As utilities move from an engrained process of manual, inspection-based theft detection to a digital forensic process, trust in the answers that the data provides will not exist on Day One, according to Oyler. Therefore, a solid change management program should accompany any analytics effort. Oyler estimates that most failed analytics initiatives result from inadequate change management, rather than pure technical shortcomings.

Despite the challenges, the results some utilities have reaped using analytics to spot electricity theft have led them to explore deploying analytics more broadly across their organizations, according to Barnes. “By identifying patterns in the volumes of transmission, consumption, billing, and customer data utilities collect, analytics can pave the way for the creation of new products and services designed to help customers manage their electricity usage.”

¹ Mills, Evan. “The carbon footprint of indoor Cannabis production.” September 2011. This study estimates that U.S. marijuana grow houses consume $6 billion worth of electricity each year.

² A press release from the Energy Association of Pennsylvania cites estimates from the FBI and the International Utilities Revenue Protection Association that puts losses from energy theft at $6 billion. Bounce Energy quotes an article from Electric Light & Power magazine that says “electricity thieves have caused a loss of revenue totaling about $6 billion industry wide each year.”

By: Deloitte CIO Journal Editor
Originally published at deloitte

Leave a Reply