Hydro-Québec has been at the forefront of Canada’s energy transformation. They’re using artificial intelligence (AI) to develop better load forecasting in order to make their electrical grid more reliable. Implementation of new AI technology fills the gaps of traditional forecasting models, especially during critical moments like extreme weather events. Then on May 22, 2024, a sudden and intense heatwave struck the utility with a double whammy. This event highlighted significant shortcomings in its legacy models and helped drive a transition toward AI-driven solutions to avoid future blackouts and increase grid stability.
The challenges of not just the pandemic, but the continued and growing reliance on conventional forecasting methods Continue reading The demand challenge Though AI models and traditional models may both work well in typical conditions, the differences manifest in times of atypical events. Hydro-Québec’s AI models accurately predicted the heightened demand during the recent heatwave, showcasing their potential for success in load forecasting.
Transitioning to AI-Based Load Forecasting
To meet these new complexities and challenges in load forecasting, Hydro-Québec has begun a full-scale transition to adopting AI technology. The utility’s old models have failed on more and more forecast days, requiring substantial corrective measures. On May 22, as an oppressive heatwave settled in, one legacy model failed to see the load pattern which caught many by surprise. Consequently, it required downward adjustments of 1,500 megawatts (MW).
Hydro-Québec’s AI model was able to predict this sudden increase in demand with exceptional precision. There was no way traditional approaches could compete with its efficacy. This success has created a sense of confidence in AI capabilities across the organization. As reiterated by Hydro-Québec’s spokesperson, “Having to deal with data from more than four million smart meters is another ball game.”
While the utility seems to be having success with a bottom-up, regional approach. Come 2028, they’ll use AI to forecast loads more than 350 substations. These innovations will improve all-hazards, short-term forecasting in up to 36 hours in advance. Further, they will expand forecasting capabilities to offer hourly, up to 42-day forecasts based on past weather patterns.
The Challenge and Opportunity of AI Integration
While the adoption of AI into Hydro-Québec’s forecasting processes will be burdensome, it holds significant potential. Utilities need to be careful about new technologies; a misstep can have catastrophic consequences. The promise of AI is starting to feel real. According to Vijaykar, “Utilities are rightly nervous about the system. When empowered, the tech industry can be some of the most innovative and risk-taking actors around. Unfortunately, the opposite is true for utilities—where one mistake could have catastrophic consequences.
Experts within the firm realize that AI makes load forecasting more complex due to the new demand curves it creates. They recognize that it offers new tools to improve precision and accuracy. Clermont emphasizes the importance of training AI systems: “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.”
As Hydro-Québécois builds confidence in these AI systems, they hope to move away from dependence on traditional models. Clermont further clarifies, “When they get confident, Hydro-Québec will stop using the old model.” This pledge to innovation is indicative of Hydro-Québec’s drive to improve operational efficiency and reduce risk tied to outdated forecasting methods.
Enhancing Grid Reliability through Intelligent Forecasting
Hydro-Québec has made AI the backbone of its innovation strategy. This precautionary step continues to ensure grid reliability in the face of climate-induced variability and the growing penetration of renewable energy resources. As renewable penetration continues to rise, bringing added variability and uncertainty with the addition of renewables, typical outdated approaches to forecasting forecast capabilities just won’t cut it. Adkins states, “The increasing penetration of renewables introduces variability and uncertainty into grid operations, making traditional forecasting and control methods less effective.”
Hydro-Québec’s multifaceted approach goes beyond simply using AI to optimize operational processes, understanding that AI cannot do it all. Adkins asserts that AI complements existing grid planning and forecasting methods: “AI complements but does not replace grid planning and forecasting. AI-driven orchestration works best alongside grid forecasting, market signals, and human operational oversight.”
Further, Hydro-Québec has spent many years developing research before putting deep neural networks into production. It’s their dedication that puts the systems in place to be resilient enough to handle unexpected events. Importantly, they have an eye on improving things from 2026 to 2027.
