Analytics aren’t new to utility operations. Duke Energy saved some $130 million in avoided costs by using predictive data analytics to identify problems before they caused equipment failures. A utility in Brazil estimates savings in the range of $420,000 USD each month through better, analytics-based theft detection.
Why talk about analytics in a blog about artificial intelligence? Because, as an article published by Forbes notes, “Machine learning is a continuation of the concepts around predictive analytics, with one key difference: The AI system is able to make assumptions, test and learn autonomously.”
With these enhancements, data science will become more powerful than ever, and utilities stand to gain. Along with theft detection and the kind of predictive maintenance that’s paid off so handsomely for Duke Energy, ere are a few smart applications for AI:
Outage alerts by algorithm: Over 90% of outage reports that utilities receive come from customers that call them in. Some AMI networks push last gasp signals to the head end , but as the size of an outage grows this method of notifications becomes unreliable. However, artificial intelligence for fault detection and localization can pinpoint the outage and its extent by analyzing data from advanced metering infrastructure (AMI) systems.
More intelligent distributed automation: Eventually, utilities will employ grid-edge intelligence that can tell the difference between a momentary power interruption and a full-blown outage in that millisecond of time necessary to engage automated switching protocols, according to Frank Brooks, Vice President of Product Management for Software at Aclara.
“I think you’re going to start seeing a lot of artificial intelligence in your distribution/automation system. Instead of detecting an outage and sending data back to a head-end system that will decide when and if to roll trucks, AI will empower edge technology that instantly sees and fixes the issue, or at least divert power around it,” said Brooks, “The head-end system will then receive data on how the system took action and inform the operator of next steps required from them”.
Yield optimization: How do you design a wind farm to get the most from the wind currents around it? GE is using a digital twin of wind-power generation assets and using AI to optimize operations for maximum productivity. The company thinks its approach can increase a wind farm’s energy production some 20 percent and add $100 million in extra value over a 100-megawatt farm’s average lifetime.
Demand-side management (DSM) and control: A project conducted by Open Energi, an AI-based DSM technology provider, has been controlling more than 3,500 electric loads at some 350 sites in the U.K. The system harvests as many as 25,000 messages per second that track 30 different data points, and it performs “tens of millions of switches per year,” notes Michael Bironneau, Open Energi’s technical director, in a write-up of the project. He adds that, “In the U.K. alone, we estimate there are 6 gigawatts of demand-side flexibility which can be shifted during the evening peak without affecting end users. This is equivalent to roughly 10% of peak winter demand.”
Shoring up resiliency: The Department of Energy’s SLAC National Accelerator Laboratory is using AI on data from a dozen U.S. partners – including the Lawrence Berkeley National Lab and National Rural Electric Cooperative Association — to identify vulnerabilities in the grid and reinforce those areas with sensors and automation to speed restoration. “The eventual goal is an autonomous grid that seamlessly absorbs routine power fluctuations from clean energy sources like solar and wind and quickly responds to disruptive events – from major storms to eclipse-induced dips in solar power – with minimal intervention from humans,” said a press release issued by Stanford University, which operates SLAC for the DOE Office of Science.
Precision forecasting: Wholesale price variability that stems from widespread renewables on the grid combined with the proliferation of behind-the-meter solar and storage complicates forecasting for utilities and system operators, says Bob Champagne, a VP for Innowatts, an energy-focused AI technology company. In an Electric Light & Power article, he claims that, “Companies which combine bottoms-up analysis of AMI data with machine learning and artificial intelligence develop precision forecasts that are 40-60% more accurate than incumbent approaches.”
Given results like these, it’s no surprise that 52% of utilities called AI critical to their company’s success when surveyed by Zpryme. More than a quarter – 27 percent – said they were already using AI, and 16 percent had an AI strategy in place.
Ahead, AI is going to make the smart grid continuously smarter. This technology may carry the “artificial” moniker, but there’s nothing phony about the value it will deliver.