Redefining Load Forecasting: How AI Is Making Smart Grids Smarter
By Naija Enquirer Staff
With legacy load forecasting models increasingly challenged by unpredictable events, artificial intelligence (AI) is emerging as a powerful solution for grid operators navigating a rapidly changing electricity demand landscape.
Although AI itself is energy-intensive and adds pressure to power systems, its ability to forecast electricity demand with greater accuracy offers significant efficiency gains. Industry experts say the technology could help grids operate more reliably and prevent large-scale blackouts.
Across the energy sector, grid operators are already experimenting with AI-powered tools. AES operates an AI-enabled Smart Operation Centre to integrate grid data for improved source management, while E.ON’s Intelligent Grid Platform consolidates grid data to enhance operational efficiency.
National Grid has partnered with Emerald.AI to explore AI’s dual role as both a major electricity load and a flexible grid manager. Meanwhile, Hydro-Québec has reported success using AI-based load forecasting models to improve system reliability.
Smart grids—electricity networks that rely on digital communication, sensors and automation—already allow real-time monitoring and optimisation of electricity generation, transmission and consumption. However, experts say AI could take these systems to a new level, delivering unprecedented efficiency and resilience.
“Load forecasting has always been a critical function,” said Aroon Vijaykar, commercial business lead at Emerald.AI. “AI complicates forecasting because of the demand it generates, but it also provides an opportunity to forecast better. At best, it can help prevent catastrophic blackouts.”
Legacy Models Meet a Changing Energy Landscape
Traditional load forecasting relies on mathematical models calibrated using years of historical consumption data. These models work well under normal conditions but struggle during unprecedented events such as extreme weather, sudden industrial shutdowns, or rapid shifts in consumer behaviour.
Sylvain Clermont, lead author of a United Nations Economic Commission for Europe (UNECE) case study on Hydro-Québec’s AI deployment, explained that legacy models are limited by their dependence on historical patterns.
“When something completely new happens—like the COVID-19 pandemic—there is no historical curve to rely on, and the model fails,” he said.
Other complicating factors include the rapid growth of renewable energy, the rise of rooftop solar systems, and the expansion of AI data centres, all of which introduce variability and uncertainty into power demand and supply.
According to David Adkins, Head of Network Architecture and Innovation at National Grid, AI enables real-time analysis of complex datasets and supports dynamic grid management, making it essential for integrating intermittent energy sources such as wind and solar.
Hydro-Québec’s AI Breakthrough
Hydro-Québec, one of the world’s largest hydropower producers, began using AI daily for load forecasting in 2024 after several years of research and development.
The utility now applies AI for short-term forecasting within a 36-hour window, medium-term forecasting up to 12 days, and longer-term projections extending to 42 days. AI models operate alongside legacy systems, allowing operators to compare results and intervene when necessary.
During a heatwave in May 2024, Hydro-Québec reported that a legacy model failed to anticipate an unusual load pattern, requiring manual corrections of about 1,500 megawatts. The AI model, however, accurately predicted the anomaly.
“We are not moving to AI because legacy models are bad,” Clermont said. “We are moving because the few days when models fail are becoming more frequent and more severe.”
Managing AI as Both Load and Solution
While AI improves forecasting accuracy, it also creates significant electricity demand, particularly during model training. This dual role makes AI both part of the problem and part of the solution.
Emerald.AI is working with National Grid to explore how AI-driven data centres can adjust their power consumption in response to grid conditions. Through intelligent workload management, AI systems can reduce or shift electricity use during peak periods.
National Grid conducted a trial in December 2025, using AI to orchestrate non-critical data centre workloads in real time. The results are currently under review.
The Road Ahead
Experts agree that AI is here to stay and will play a defining role in the future of electricity systems. When integrated responsibly, AI promises enhanced forecasting accuracy, reduced operational costs, and improved grid resilience.
“AI will help facilitate autonomous grid operations, optimise energy flows and support the energy transition,” Adkins said. “This will ensure power systems remain robust, flexible and sustainable in the years ahead.”
