As renewable energy has become more financially feasible, it has led to a surge in self-generation, as we stick solar panels on the roof of our house or a wind turbine in the garden.  I’m sure we hope that such installations will reduce our energy bills, whilst perhaps at the same time helping the planet and all of that.

Such assumptions fall down if we can’t secure the electricity that’s generated, and electricity theft is a surprisingly big deal.  The Revenue Protection Association believe that such theft costs around £440 million a year in the UK alone.  In Brazil, Electrobras claim that a whopping 22% of their electricity is lost to a combination of fraud and theft.

AI to the rescue

It was in Brazil that a team of researchers from the University of Luxembourg attempted to improve matters.  A recently published paper documents the AI system they developed to help cut down on energy theft in the country.

The researchers analyzed the electricity usage of over 3.5 million Brazilian households over a five year period, with over 200 million monthly readings made by professional meter readers.

They developed an AI algorithm that was able to detect when the readings were suspiciously low.  They used data from previous inspections to check whether the conclusions made were correct or not.

The tool was capable of accurately identifying problem cases approximately 65% of the time, which the team believe is significantly better than other tools on the market today.

The team hope that their tool will allow for more targeted inspections of properties, whilst also warding off people from meddling with what is incredibly dangerous equipment.

The tool will be implemented by the smart energy company CHOICE Technologies, who also contributed to the research itself.  The software will be rolled out across Latin America.

The team believe that the system could also be deployed in western countries, and a number of UK utility companies have agreed to share data with the researchers to help identify thieves.