Electricity theft and fraud represent significant economic losses to utility players. To combat these, many are turning to IoT, big data, and machine learning for new, more effective detection methods.

According to a recently published Northeast Group, LLC study, non-technical losses – theft, fraud, billing errors, and other related issues – total $ 96 billion per year globally. Some estimates that a utility can lose as much as 10 to 30 percent in revenues each year to theft. In worst cases, such losses threaten the financial stability of electric utilities.

Read also our guide to efficient power grid operations for the digital age. 

Smart Meter Data and New Levels of Insight

Smart meters, or advanced metering infrastructure (AMI), can be used for far more than automated metering and billing processes. Smart meters enable two-way communication between the consumer and the utility’s control center. These devices generate massive amounts of data for various analytics-based applications and increase the possibility of improving asset management and business processes. Today, smart meters are essential smart grid components.

Smart meters are also widely recognized as useful tools to detect fraud and other non-technical losses. With data generated in near real-time, smart meters enable utilities to gain a clearer picture of losses in their system, abnormal or sudden drops in energy consumption, and to perform consumer profile analyses. As the number of smart meters in a utility’s network grows, so does the amount of data generated. As the amount of data generated increases, the more precise utility infrastructure operations can be.  

Machine Learning Discovers Anomalies

Often referred to as a subset of AI, machine learning provides systems with the ability to learn automatically and improve from experience without being explicitly programmed. Machine learning algorithms can learn from data, make predictions upon that data, and identify unusual behavior in big data sets.

Machine learning is recognized as a successful measure for fraud detection. If you are involved in utility operations, this is good news. You can stop relying on traditional, outdated methods for fraud and theft detection, such as customers reporting suspected energy theft, manual investigation, and the monthly visits of utility meter readers. Instead, you can identify and aggregate data from everyone connected to an electrical substation in near real-time and detect anomalies by comparing consumption with substation loads and predictions based on machine learning.

Smart Meter Data and Machine Learning for Other Utilities

Although much of the discourse about smart meters centers around electric utilities, smart meters provide similar advantages to other utilities, such as water.

Water leakage due to aging infrastructure is a significant problem in the water industry. Estimates suggest that on average 25-30 % of a utility’s water is lost in the network. Smart meters provide snapshots of how much water is being used, by which customers, and when. Comparing this to the forecasted flow of water and other upstream metering, utilities can detect leakage or other anomalies such as theft and fraud in their network. 

IoT, machine learning, and big data have indeed made its way to the utility sector. As more and more sensors and connected devices are installed in utility infrastructure, the insight, intelligence, and situational awareness increases – both for infrastructure operations in general and fraud detection specifically.

If you would like to read more about the inherent opportunities of IoT and machine learning for utilities, I recommend you download our free e-book “The Next Generation Utility Infrastructure Operations: 7 Key Drivers of Change” by clicking the button below.