In a notable step forward for the sector, Hydro-Québec is using artificial intelligence (AI) technologies to improve its load forecasting process. Supported by an equally novel approach to cost-effectively deliver grid reliability and prevent blackouts, this new approach prioritizes operational efficiency. By 2028, Hydro-Québec will use AI to improve short-term load forecasting at more than 350 substations. This announcement indicates a tremendously important change in how utilities are going to respond to demands for energy.
AI improvements in load forecasting are increasingly important. It provides better predictions too, as we face increasing pressures from the greater need for energy and the complexity of more frequent extreme weather events from climate change. Hydro-Québec incorporates AI to deliver accurate 12-day hourly forecasts, powered by the latest meteorological data. They’re able to provide longer-term outlooks due to access to reams of historical weather averages. This innovative implementation of technology holds the potential to transform how energy providers react to changing demands for power.
AI Enhances Forecasting Accuracy
Hydro-Québec’s AI models are particularly well-suited to managing high-dimensional, non-linear data. They do a great job using constraints pulled from the ENLSIP algorithm. Combine that daunting with a base of thousands of functions and hundreds of variables that modelers tune, and these models evolve constantly to become new and more precise.
None other than the utility’s spokesperson reminded lawmakers that coordinating and troubleshooting data from more than four million smart meters dated. As they noted, “Handling data from over 4 million smart meters is a different ball game.” This massive amount of information allows the AI to identify trends and correlations that classical models simply can’t. Most importantly, it works miracles when demand unexpectedly spikes, such as during a heatwave.
AI also consistently predicted the atypical demand spike during the recent heatwave – a prediction that proved accurate. This is showing its power to increase reliability where flaky forecasting can’t. Traditional models are being asked to do more and failing to provide results that are right. This deepening crisis points to a critical need for new solutions including AI.
“Load forecasting has always been an important function. AI is both a challenge and an opportunity on both sides of that equation. It is complicating load forecasting because of the demand it generates; however, we now have this opportunity to use this new tool to do better load forecasting.” – Vijaykar
A Shift in Utility Management
The move toward AI-driven long-term forecasting points to a larger trend occurring in the utility sector. Providers appear to be increasingly wary of moving too quickly ahead of new technology. Vijaykar underlined this feeling with the next statement, “Utilities are rightfully conservative about the system. Whereas the tech industry is able to move faster and take risks, one mistake for utilities could have a truly catastrophic outcome.”
Hydro-Québec’s path toward the use of AI started with a 2018 proof of concept, which used a basic neural network model. Since then, the utility has taken a bottom-up, regional approach to implement AI in a way that maximizes effectiveness. Clermont added that “it’s all a question of training; the AI needs to be trained. On the first day, it might seem awful. A year later, it may break your expectations. This continual learning process is so important because it gives the AI the ability to fine-tune its predictions continually as it receives more information.
If confidence in the AI systems increases, Hydro-Québec might abolish traditional forecasting methods entirely. Clermont remarked, “When they get confident, Hydro-Québec will stop using the old model,” indicating a future where AI becomes integral to load management.
The Future of Energy Management
The possible advantages from incorporating AI into load forecasting are huge. By increasing accuracy and efficiency, AI can assist utilities with proactively managing their grids and lowering their operational costs. Vijaykar emphasized that at a minimum, “AI can help to run systems more efficiently. At a maximum, it will avoid catastrophic blackouts.”
This transition has not been without its challenges. As renewable energy sources are being adopted and penetrated the grid, they increase variability in operations, making it more difficult with traditional forecasting methods. Adkins pointed out that the rapid adoption of renewables has introduced variability and uncertainty to grid operations. Consequently, linear forecasting and control methods fail to work.
AI should not and will not supplant current grid planning and forecasting. Rather, it supercharges these processes by providing new and improved analytical capabilities. As Adkins explained, “AI-driven orchestration is most effective when used in concert with grid forecasting, market signals and human operational supervision. It’s not a magic bullet, stand-alone solution, but it is one more piece of a broader flexibility and resilience toolkit.”