Co-Authors: Hansen Chan & Dominique Verhulst
The rise of artificial intelligence (AI) and machine learning (ML) is generating excitement and revolutionizing the way enterprises work. Power utilities have long recognized that AI and ML have the potential to help them overcome the traditional complexities of grid management. There have been extensive research efforts to explore AI neural network models for grid applications such as state estimation, proactive asset management and load forecasting.
Recent advancements in large language models (LLMs) and training methodologies such as mixture of experts (MoE) and distillation have renewed utilities’ interest in embracing AI technologies. This return to AI provides a fresh opportunity for utilities to accelerate innovation and significantly the efficiency and reliability of their grid operations.
AI brings benefits and challenges for utilities
As utilities work to decarbonize their grids, they face power generation dynamics and uncertainties that complicate system management. Phase measurement units (PMUs), Advanced Metering Infrastructure (AMI) and digital fault recorders are collecting greater volumes of multidimensional data. Utilities need smarter and more powerful software systems that can help operators classify and characterize grid events to ensure a quick response.
Traditional data analytics often fall short in addressing the complexities of multidimensional data. As a result, utilities are urgently seeking innovative AI-powered solutions that can process high volumes of time-series data in real time and adapt to rapidly changing line and weather conditions.
But processing power and adaptability aren’t everything. Utilities also need to be able understand and explain how their AI applications interpret data and produce output aimed at maintaining grid reliability. This explainability is essential for verifying, validating and eventually trusting that recommendations from AI won’t inadvertently compromise the integrity of the grid.
Smaller AI models help utilities unlock new AI opportunities
As discussed during a recent episode of Energy Central’s Power Perspectives podcast with Dominique Verhulst (Global Energy Practice Leader, Nokia) and Hansen Chan (IP Enterprise Solutions Marketing, Nokia), the arrival of open-source, modest-size AI models such as DeepSeek has disrupted the AI and utility landscapes.
With this emergence of open, accessible and resource-efficient models, utilities can deploy AI in a private operational technology (OT) cloud hosted in on-premises data centers. With a high-performance IP/MPLS wide area network (WAN) and substation local area network (LAN), utilities can take advantage of these models to enable new AI-based edge computing applications in hardened, graphics processing unit (GPU)-accelerated compute in substations.
Last but not least, these new and more efficient AI models are great news for utilities working to find the right balance between CAPEX and OPEX (i.e., acquiring versus renting physical hardware). They allow utilities to get the most from their investments in computing capacity by handling more tasks or queries at scale with the same number of GPUs.
Why utilities should act on AI now
The time is right for utilities to embrace AI. With an abundance of digital grid data and high-performance IP/MPLS WANs at their disposal, they will be able to use AI to drive innovation that ushers in a new era of efficient and reliable power delivery.
While utilities will face challenges as they embrace AI, there are great opportunities to implement impactful applications in the near term and long term. By starting now, utilities will be able to build maturity in AI and ML integration so they can get the most from these powerful tools as they continue to evolve.
Let’s talk about AI and the new energy future
Want to learn how AI integration can help you build a smarter, more efficient grid? Or how to transform your network so you can embrace AI at an accelerated pace, in a way that’s tailored to meet your specific operational environment and needs? Join experts from Nokia and other key players at DISTRIBUTECH® International 2025 (25-27 March in Dallas, TX). Let’s talk about how AI and ML can empower you for the new energy future.