The mission of this group is to bring together utility professionals in the power industry who are in the thick of the digital utility transformation. 


Utilities engage machine learning to predict what’s next

Andrew Braeger's picture
Data Services Manager Tacoma Public Utilities

Andrew Braeger is responsible for leading Tacoma Public Utilities data management practice. He has been with the utility since 2017 and has played a key role in advancing strategic initiatives...

  • Member since 2020
  • 4 items added with 5,826 views
  • Jan 11, 2021

The 2020 COVID pandemic has significantly impacted electric utilities across North America. As the crisis began, many utilities found themselves analyzing a variety of operational datasets to enhance situational awareness. The challenge, however, is that historical datasets provide only limited insight into what might be coming next. An operational report, for example, may display where distribution system power outages have recently occurred in the service territory. In order to understand, however, where power outages are likely to occur in the future, utilities must harness the capabilities of machine learning. Moving into 2021, power utilities will more intentionally leverage predictive modeling and machine learning to proactively anticipate customer, operational, workforce, asset, and financial challenges and opportunities.


The first area of opportunity in utility predictive analytics centers on customer experience. In 2021, utilities will need to continue to support their customers on the long road of recovery. By engaging machine learning, utilities can predict which customers are likely to need bill assistance resources and offer targeted communications to build long-term customer loyalty. Digitally progressive utilities may even begin to provide predictive customer next best offer features in customer-facing digital channels.

Operations & Workforce

Next, utilities will more effectively leverage machine learning capabilities to improve operational effectiveness. During the winter and spring storm seasons, utilities may utilize vegetation imagery, outage history, and asset health datasets to predict where outages or system disruptions are likely to take place. Predictive modeling may also help with field and mobile workforce optimization as dispatch operators can position crews proactively to the most vulnerable areas of the network.


Utility SCADA platforms for generation, transmission, and distribution operations produce overwhelming amounts of data that utilities often struggle to convert into actionable insights. As utilities continue to advance their machine learning practices, data scientists can train these massive datasets to separate the signal from the noise, leading to valuable asset management takeaways around maintenance and optimization. These insights can save millions of dollars and directly benefit customers in keeping rates affordable.


Finally, as the road to financial recovery from the COVID crisis becomes clearer, utility data scientists will build predictive machine learning models around receivables, cash flow, and other essential financial indicators to help decision-makers gauge the financial state of the utility. These predictive tools become essential for future planning and effective executive decision-making.

Overall, moving into 2021, power utilities will begin to tap into the heaps of utility datasets in order to gain predictive insights. As data science technologies and patterns becoming increasingly more commonplace, power utilities will be well positioned to engage these insights to enhance the customer experience, reduce operational expenses, and prolong asset life.

Matt Chester's picture
Matt Chester on Jan 11, 2021

utility data scientists will build predictive machine learning models around receivables, cash flow, and other essential financial indicators to help decision-makers gauge the financial state of the utility

This is surprising to me that it'll be so critical, but that's likely just my lack of exposure to the bill collection / cash flow part of the industry. How much of a variation does the typical utility have to deal with in this regard that not only can data models influence the business decisions being made, but so much so that investing in ML solutions to optimize that knowledge is a bottom-line boosting endeavor? What would be the month-to-monthly type benefit the intelligent utility using this tools would expect to gain? 

ALESSANDRO CAPO on Jan 13, 2021

ML is already in use in condition monitoring and predictive maintenance policies regarding power transformers and other power devices such as cables and boards. There are many succesful cases around

Oren Ahronson's picture
Oren Ahronson on Jan 13, 2021

Very interesting article. some of it (and more) is already available. 

Happy to tell more about it. 

Matt Chester's picture
Matt Chester on Jan 13, 2021

Would love to hear more, Oren! What do you think is the most exciting or promising of these tech developments that's out there but maybe not ubiquitously? 

Oren Ahronson's picture
Oren Ahronson on Jan 25, 2021

@Evolution Energy we use AI to help utilities make their distribution grid more visible (true interactive map, phase identification, LV DER management), more resilient (load balancing, fault prediction) and save costs by taking data driven decisions on grid planning, workforce optimization, voltage optimization and more.   All this is done with no additional H/W which grants fast implementation and ROI. 

Klaar De Schepper's picture
Klaar De Schepper on Jan 13, 2021

In addition to benefits to just the utility's own bottom line, Machine Learning can serve overall public good, and help regulated utilities comply with regulation. Machine Learning can for example be particularly helpful for managing and accelerating Distributed Energy Resource interconnection and grid support. Here's a paper about this topic from South Korea, which currently largely relies on imported fossil fuel for electricity generation.

Korea Electric Power Corporation (KEPCO) is responsible for 93% of Korea's electricity generation. The South Korean government (directly and indirectly) owns a 51.11% share of KEPCO. With this clear incentive to reduce reliance on imported fossil fuel it's no wonder that:

The government intends to raise the share of renewable energy to 20% of total power generation by 2030 and at least 30% by 2040 through several government policies announced in the past few years to limit coal-fired and nuclear generation.

Source: U.S. EIA

Are utilities in the U.S. sufficiently incentivized to accelerate interconnection? Could Machine learning help? Although the DOE has put some funding muscle towards application of advanced modeling, it currently looks like different regions and utilities in the U.S. are inventing their own wheels.

Finally, Machine Learning is only as good as the data it's based on, so standardization on the back-office data exchange systems that support various Machine Learning applications will be crucial, as recognized by NAESB's new Digital Committee.


Paul Korzeniowski's picture
Paul Korzeniowski on Jan 14, 2021

Finally, Machine Learning is only as good as the data it's based on

Key point here. The challenge is collecting enough information so they can build algorithms that have enough data to provide them with real insight. The Covid pandemic has made such work more difficult. Nothing has been normal since it started. So, utilities have the dual challenge of collecting the data and then somehow deducing how it will impact their operations as they move forward. As a result, there may be a blip or two encountered before the industry fully realizes the potential that ML offers. 


ALESSANDRO CAPO on Jan 14, 2021

I would suggest to follow contributions by Ms Sruti Charaborty that is one of the pioneers in ML applications in power sector, here one example: Machine learning for transformer diagnosis

Andrew Braeger's picture
Thank Andrew for the Post!
Energy Central contributors share their experience and insights for the benefit of other Members (like you). Please show them your appreciation by leaving a comment, 'liking' this post, or following this Member.
More posts from this member

Get Published - Build a Following

The Energy Central Power Industry Network is based on one core idea - power industry professionals helping each other and advancing the industry by sharing and learning from each other.

If you have an experience or insight to share or have learned something from a conference or seminar, your peers and colleagues on Energy Central want to hear about it. It's also easy to share a link to an article you've liked or an industry resource that you think would be helpful.

                 Learn more about posting on Energy Central »