Hydro-Québec, Canada’s leading utility provider, is seriously impressing with their top rankings in load forecasting and management. Here are some of the ways they’re currently applying artificial intelligence (AI) to their work. The utility uses AI to improve its short-term forecasting capabilities, looking out 36 hours. It develops hourly forecasts at ranges of up to 12 days in advance, informed by short-term predictions made by on-staff meteorologists. This creative solution addresses the disconnect between forecast and actual. It asks for particular focus during critical periods, when mistakes in forecasting can be most damaging.
The company’s AI models use historical weather data to help inform longer-term outlooks, up to 42 days. Even as Hydro-Québec moves to maintain strong operational efficiency year after year. Their mission is to avert major perturbations in a grid that’s the backbone of over four million smart meters. On May 22, 2024, during a brutal heatwave, the utility received a record-breaking test. Their standard forecasting models just weren’t equipped to handle a sudden and unforeseen boom in demand, underscoring the immediate need for alternatives.
The Shift from Legacy Models
Hydro-Québec has traditionally relied on legacy models characterized by non-linear constraints based on the ENLSIP algorithm, which incorporates numerous functions and parameters adjusted regularly. These legacy models fail to capture the atypical or extreme circumstances, like sudden shifts in weather, of a new reality.
Aware of these shortcomings, Hydro-Québec started on a path to introduce AI in its forecasting operations. In 2018, the company started a proof of concept using a neural network at one substation. This was the first steps towards a long and slow technological shift to more advanced forecasting mediums.
“It is a training question; the AI needs to be trained. On the first day, it is probably not that good, but after a year, it is probably better than you.” – Clermont
The company has wasted years on R&D. This investment allows it to develop and test its deep neural networks to rigorously validated strengths before they’re launched into functional frameworks. This measured approach clearly indicates the utility’s understanding of the possible consequences of mistakes in forecasting.
Utilities typically take a risk-averse approach to new technology. Aroon Vijaykar, Hydro-Québec’s advances in AI commercial business lead, points out that “utilities are rightfully very conservative about the grid. In their line of work, one mistake can be catastrophic,” he stresses.
A Proactive Approach to Load Forecasting
As you can see, Hydro-Québec is driving deep into the AI highway. While doing that, it hopes to improve its load forecasting approach on its interconnected substation network. Starting in 2028, the utility plans to implement a bottom-up, regional strategy that will enhance forecasting for over 350 substations.
The goal is to help operators manage the grid more efficiently and mitigate risks associated with unexpected demand surges or drops. This proactive approach can help avoid costly missteps economically while ensuring a higher level of service reliability across the board.
“Load forecasting has always been an important function. AI is both a challenge and an opportunity on both sides of that equation,” – Vijaykar.
He describes how AI complicates load forecasting with increasing demand. It’s a huge opportunity to use some of these more advanced tools at our disposal to achieve much better outcomes.
Hydro-Québec’s AI-powered models are especially well-suited to the kind of outliers that are unique, atypical, and not accounted for in average patterns. Clermont continues, “Our experience gives us a rich source of historic curves.” Consequently, their mathematical models are particularly adept at identifying consistent, uniform patterns. In practice, though, he warns that unforeseen events will always create opportunities for error.
“Then comes something totally out of the box – whether extreme weather or something else that you have never experienced in the past – and your model will be off.” – Clermont
The Future of Smart Grids
Hydro-Québec’s commitment to innovation goes beyond just adopting AI technology for load forecasting. Beyond the program above, the utility has its eyes on constantly improving its operational AI as well as getting better at using smart meter data. We plan to create a renewable energy forecasting prototype between 2026 and 2027. This new effort is meant to address the complex challenges that are emerging with our increasing reliance on renewable energy.
The increased penetration of renewable energy changes the operations of the grid by adding more variability and uncertainty. Adkins points out that “the increasing penetration of renewables introduces variability and uncertainty into grid operations, making traditional forecasting and control methods less effective.”
Through the use of AI integration, Hydro-Québec expects increased forecasting accuracy, more proactive grid management, and lower operational costs. Adkins highlights the importance of this technological evolution: “AI integration promises enhanced forecasting accuracy, proactive grid management, reduced operational costs and improved capacity to adapt to evolving energy landscapes.”
“AI complements but does not replace grid planning and forecasting. AI-driven orchestration works best alongside grid forecasting, market signals and human operational oversight,” – Adkins.
Hydro-Québec is continuing to develop its AI expertise. At the same time, it is doggedly committed to proving that it can operate safely and reliably.
