AI algorithm spots electricity theft

Researchers have developed an algorithm that can spot electricity theft with much greater accuracy.

Electricity theft is a large and growing problem in many areas of the world. The United Kingdom Revenue Protection Association estimates that such activity causes losses of GBP 440 million a year in the UK alone. While in Brazil, Eletrobras, the country’s largest electric power utility, loses up to 22 percent of all the power it generates to fraud and theft. Now big data may offer a solution. A team of researchers at the University of Luxembourg has developed an AI algorithm that can recognise when energy use at a property is suspiciously low.

To develop their program, the researchers used a novel approach which combined automated statistical decision making with data on past meter readings from millions of households. The resulting algorithm allows companies to better target which properties they will inspect for possible fraud, saving time and money. The system is able to correctly identify electricity theft around 65 percent of the time, which outperforms other tools on the market. Smart Energy firm Choice, which participated in the research, plans to use the technology in its commercial software. Although there may be many reasons other than theft for a building to use less energy than adjacent structures, being able to pinpoint buildings at a high risk for theft may help drive down the costs associated with stealing electricity.

In addition to finding electricity thieves, we have also recently seen AI algorithms used for detecting people who are texting while driving and determining recipes from photos of food. In what other ways might AI help businesses to pinpoint fraud and theft?