AI in Utilities - 6 Simple Steps for a Utility’s AI Adoption Journey

In the dynamic energy sector, artificial intelligence (AI) is revolutionizing how utilities operate, offering unprecedented opportunities for efficiency gains and enhanced service delivery. The journey to AI adoption may seem overwhelming, but it doesn’t have to be. The basic AI in utilities adoption journey can be condensed into a simple list of six steps which, together, ensure successful integration and maximize the potential benefits. Drawing from GE Vernova's insightful whitepaper "Empower Intelligent Grids with AI," this blog explores the steps to effective AI adoption in utilities.

Data is the lifeblood of AI. Ensuring data accuracy, availability, and security is essential for AI systems to function optimally. Thus, for utilities to adopt AI, they must firstly invest in developing a strong data foundation, which includes data collection, storage, management, and governance capabilities. This step also involves breaking down silos to enable seamless data flow across various departments and systems.

By far the simplest way of attaining a data foundation is to acquire a ready-to-deploy grid data fabric, like GridOS® Data Fabric. GridOS Data Fabric is purpose-built for utilities and the highly disparate, scattered, and siloed data sources they must work with, along with capabilities for accessing, governing, and utilizing data as needed for various use cases.

Once the data foundation is in place, the second step to AI adoption in utilities involves enhancing the quality and accuracy of data. GridOS Data Fabric plays a crucial role in achieving this by offering policy-driven data governance to bolster data integrity, consistency, and accuracy. It achieves this by consistently monitoring data dependencies and flagging any anomalies in upstream data sources. Additionally, it provides capabilities for data validation and translation, supported by a monitoring and management console for overseeing data integrations, data flows, and APIs.

Without IT/OT convergence, utility AI adoptions will struggle. That’s why the third step involves creating a grid modernization team. AI applications are increasingly integral to grid modernization efforts, and cross-functional collaboration, particularly between OT and IT, is essential. Forming a diverse, cross-functional grid modernization team with members from various departments provides a forum to drive and prioritize modernization activities.

The team members should be drawn from all corners of the organization, such as enterprise, operations, control room, planning, applications engineering, and power systems engineering. The best mix of professionals might vary somewhat between utilities; however, all grid modernization teams should include at least one AI Product Manager and data/machine learning engineer, if available.

The grid modernization team is tasked with developing a long-term grid modernization strategy and vision for the next 5-10 years, preparing for regulatory changes, and adapting to evolving grid requirements. It ensures flexibility for future needs, validates the business value of pilot programs, and accelerates time-to-value for new applications. The team leader should have the authority to make key decisions, particularly regarding budget setting and approval.

The fourth step in the AI in utilities adoption journey involves identifying and prioritizing the most suitable use cases. Utilities can employ one of two frameworks to prioritize these use cases effectively and easily.

One such framework is the effort-impact matrix, which helps utilities pinpoint and prioritize projects that require low effort—considering factors like cost, effort, or complexity—while promising high impact in terms of benefits, business value, or outcomes. Use cases that fall into the bottom-right quadrant of this matrix are ideal candidates for implementation.

Alternatively, utilities may opt for GE Vernova’s risk-based framework for use case prioritization. This approach is particularly beneficial given the novel nature of AI technologies, especially within the grid context. A risk-based strategy can help utilities navigate the complexities of utilities AI adoption while ensuring successful outcomes.

The fifth step in the journey is to pick the right AI model for each specific use case. This involves such considerations as:

More information on AI models for various grid use cases may be found in this whitepaper by Pacific Northwest National Laboratory.

The sixth step involves choosing whether to develop AI models internally or collaborate with or purchase from a utility-specific vendor. With the rapid expansion of AI models, GE Vernova advises utilities to partner with established industry vendors. This approach offers access to utility-specific AI models that are pre-trained on extensive and varied electric industry datasets, along with product lifecycle support for emerging AI models. However, certain use cases may necessitate a certain amount of in-house AI model development tailored to specific utility needs. Successfully achieving this requires leveraging vendor-ready AI models and utilizing a streamlined MLOps platform supported by a robust data fabric foundation, which is essential for effective utilities AI development and deployment.

Conclusion

By fulfilling the above steps, utilities can effectively navigate their AI adoption journeys, unlocking transformative potential and driving innovation in the energy sector. As AI continues to shape the future of intelligent grids, utilities that organize their adoption journeys in this way will be well-positioned to lead the charge towards sustainable and efficient energy management.

For more information about the AI in utilities and their adoption journey, and the benefits AI holds for the modern grid, check out GE Vernova Grid Software’s new whitepaper, Empower Intelligent Grids with AI.