Cutting electricity theft in Spain - what role is data analytics playing?

Big data is helping Spanish utilities to increase revenue collection. Engerati asks Celestino Güemes of Atos Worldgrid for insights from the field
Published: Mon 28 Nov 2016

Economic recession in Spain means that demand for the country’s 102GW installed energy capacity is down but that utility non-technical losses (NTLs) remain high.

A study conducted by the University of Seville in 2011 found that energy theft accounted for 35-45% of Spain’s non-technical losses. And in 2015, electricity fraud was equivalent to electricity usage in the city of Seville and its metropolitan area during one year, according to major Spanish utility Endesa.

The nature of electricity and gas fraud in the country is surprising, says Celestino Güemes, Solutions R&D manager at Atos Worldgrid Spain, an international subsidiary of Atos, who has worked with several Spanish utilities on data-driven revenue protection initiatives. He says the reality of customers stealing power often contradicts the typical intuition-based profiling.

“People expect electricity and gas theft to happen in poor areas with high levels of crime. But data analytics is showing electricity suppliers that medium-size businesses in specific commercial areas such as the hospitality industry have a greater propensity to theft.”

Güemes says while cases of large-scale fraud are unusual, consumers are increasingly creative in the ways they steal power and damage metering equipment. For example, a case was reported of holes being drilled into a meter and filled with sugar in order to attract ants into the mechanism to block it up.

There is also a strong relationship between energy theft and organised crime, says Güemes, who is the lead architect of the Atos Codex for Smart Grid Applications platform. “We have seen people selling ‘energy saving services’ but in reality, they are tampering with meters. The customer is honest in that he isn’t trying to steal, he thinks that he is adding some kind of energy-efficiency so he is also a victim of fraud.”

Data analysis of gas revenue losses meanwhile is throwing up interesting findings with significant levels of tampering occurring among more affluent customers in Spain who “have the means and way to do the theft easier than other societal levels”, says Güemes.

Smart meters - solution to revenue protection?

This may sound surprising in a country that is comfortably on track to deploy 21.8 million smart electricity meters by 2020. Güemes however points out that although smart meters are fitted with anti-tamper mechanisms, the event signals generated tend to be quite ‘noisy’, especially during the initial rollout phase. “This creates many false alarms and means that sometimes these signals are ignored.” But Güemes also highlights how these events are useful as they can be cross-correlated with other data to enhance overall efficiency.

In the instance of fraud external to the meter, like diversions or bridging phases, smart meters could detect this type of derivation using energy balance but only if the energy company has a completely automated low-voltage network. “Energy balance methods are evolving, they are mathematically quite complex and computationally intensive. So we can say they are in initial stages,” says Güemes.

On the flip side, however, smart meter networks do provide benefits to an analytic revenue protection solution by supplying more granular data that allows for more sophisticated forms of machine learning analysis, in addition to network balancing models.

Atos Worldgrid’s cloud-based data analytics platform draws on meter data as well as external sources such as typical weather temperatures - which are so important for analysing gas consumption - and looks at customer history and related cases of fraud. Lists of candidate points for inspection are then passed to utility field staff to investigate. “The idea is we get a toolbox of methods and we combine it and apply to the specifics of the utilities.”

So it is effective? Güemes claims that utilities in Spain were seeing recoveries of 5-8% by using more reactive methods such as responding to police reports or blackout incidents. By deploying a data analytics platform, they have increased revenue collections of 35% to 40%.

Atos announced this month a collaboration with Spanish gas distribution company Madrileña Red de Gas, where a joint approach resulted in a sixfold increase in the effectiveness of inspections by field staff. And Atos says an investment in a revenue protection project is recovered in less than a year.

Customer decision trees and NTLs

The Department of Electronic Technology at the University of Seville in a research paper in 2011 - ‘New Methods to Detect Non-technical Losses on Power Utilities’ - reported good rates of revenue collection using data mining techniques. In a trial, a research team applied a methodology based on decision trees to detect energy fraud in the Spanish provinces of Catalonia and Andalusia among Spanish utility Endesa’s customer base.

The paper, which describes the methodology of detecting NTLs in Spanish energy companies as “not very advanced”, notes how the researchers used decision trees of customers with consumption patterns similar to customers with NTLs previously detected by the company. Results showed a collection rate of 20% compared to around to 2% rate of NTLs from customers from the original data sets of the company.

Data insights bring ‘clarity’

Güemes said that Atos identified revenue protection in 2011 as a data analytics use case that would have a quick return for utilities. As well as proving effective in increasing collections, however, he says there are other collateral results of a data-driven revenue collection programme. By gaining a deeper knowledge of customer behaviour and patterns of consumptions, revenue protection provides “clarity” and is a type of use case that can be the starting point for new ideas and opportunities by reusing the same platform.

Data programmes often reveal issues with data quality and at a more basic level bring administrative errors to light such as a utility not mapping its topology correctly. “In this instance, we’ve been able to supply energy providers with new information on their end points, helping them to correct issues with their operations saving money too. This is an interesting side effect.”

Non-technical losses briefing

Engerati and Atos Worldgrid have co-produced a briefing on non-technical losses. Watch 'Enhancing Revenue Protection Activities with Advanced Analytics in the Cloud' on demand now.