Utilities and machine learning – the use cases

Machine learning is becoming increasingly important in day to day utility operations.
Published: Fri 12 Jan 2018

Machine learning is finding growing application in the energy sector as analytics and computational capabilities advance, with the potential to meet the changing requirements of an increasingly complex energy system.

Machine learning use cases

According to a new study from Navigant Research, machine learning - which supports artificial intelligence - is best suited for a handful of specific analytical processes, including clustering, regression and classification.

On this basis, some use cases for utilities in which machine learning has advantages over existing analytics techniques, include customer segmentation, pricing forecasting, anomaly detection, fraud detection and predictive maintenance.

“The utilities industry is already using self-learning algorithms, particularly in the field of asset monitoring and predictive maintenance, and several reasons suggest the use of machine learning will expand to many more use cases and its adoption will accelerate,” comments Stuart Ravens, Principal Research Analyst with Navigant Research.

“During the past decade, it has become easier for companies to deploy machine learning thanks to falling costs, new technological advancements, a softening of conservative attitudes and a fresh approach to analytics procurement.”

Utility examples

Some applications of machine learning previously covered on Engerati include weather forecasting for renewables dispatch implemented by Vermont Electric Power (Velco) and energy trading.

As an example of its capability, Chandu Visweswariah, CEO of Utopus Insights, told Engerati that while the algorithms and forecasting had taken about two years of development, a hyperlocal forecasting capability could now be set up for almost anywhere in the world within a week.

Another utility example is Pacific Gas & Electric, which has employed machine learning to increase the accuracy of load reduction forecasts for demand response.

Another application for machine learning is cybersecurity. Israeli company CyActive – subsequently acquired by PayPal – exploited the technique to generate and detect malware variants.

An example of anomaly detections comes from San Diego Gas & Electric, which has used machine learning to detect power system issues hidden in large data sets.

Machine learning for Internet of Things

Statistics on machine learning in utilities are limited but according to a 2016 study by zpryme and analytics provider SAS, almost one-third of utilities in North America were then using machine learning for metering and meter data management.

Machine learning and artificial intelligence are also fundamental to the Internet of Things. According to Capgemini’s 2017 World Energy Markets Observatory, out of almost 2,900 companies participating in North America’s ‘IoT revolution’, 401 are in the artificial intelligence and machine learning category.

And out of the region’s over $73bn investment in IoT, over $12bn is in machine learning and artificial intelligence.

Artificial intelligence is envisaged as providing benefits on both the grid and customer sides of the business – think, for example, of the growing use of digital assistants for customer engagement. But the use cases highlighted above indicate, the benefits of machine learning are primarily grid-oriented.

Nevertheless, the zpryme survey, utility executives named better customer service as one of the top three benefits envisaged from machine learning, along with increased cybersecurity and improved data driven decision making.

The zpryme study offers several recommendations for machine learning adoption. First, these need to be part of wider strategy so that they can be leveraged across the organisation.

Their implementation will require change, and learnings can be gained from connections with both other industries and other utilities.

Related Webinar