The adoption of artificial intelligence (AI) by utility companies is on the rise, particularly in the United States, as they seek to improve operational efficiency and meet increasing demands. However, these companies are taking a cautious approach, especially when it comes to deploying AI in critical grid operations. This careful strategy is driven by the stringent safety and reliability requirements of the industry, where even minor errors can have significant consequences. By ensuring that AI systems are thoroughly vetted and implemented incrementally, utilities can harness the potential of AI without compromising safety and reliability, thus maintaining their commitment to high standards.
The Sandbox Approach: Testing AI in Controlled Environments
Utility companies like Pacific Gas and Electric Company (PG&E) are adopting a methodical approach to AI deployment to ensure safety and efficacy. Norma Grubb, PG&E’s director of enterprise AI and data science, describes this process as working in a “sandbox,” where new tools are tested in a controlled environment. This approach allows for innovation while maintaining a strong commitment to safety. AI is initially used for back-office tasks, such as processing regulatory documents and managing internal communications, to enhance efficiency without compromising direct customer-facing functions.
PG&E’s cautious strategy ensures that AI tools are thoroughly vetted before being introduced into operational areas. This iterative process helps mitigate risks and ensures that any potential issues are identified and addressed early on. By focusing on back-office functionalities first, PG&E can refine AI applications and build confidence in their reliability before moving them to more critical grid applications. This methodical process safeguards the utility’s operations and maintains high standards of safety and reliability that are necessary for maintaining trust with customers and regulatory bodies alike.
Oracle’s Human-Centered AI: Complementing Human Decision-Makers
Jason Strautman, vice president for data science and analytics engineering at Oracle’s utility division, highlights the unique challenges of deploying AI in a utility setting, where reliability is paramount and failures carry high stakes. Oracle’s AI-powered Opower platform, implemented by PG&E, generates personalized energy reports for residential customers, including a specialized version for solar users introduced in 2023. This platform provides insights on energy usage, solar production, and grid interaction, helping customers make informed energy choices.
The Opower platform also supports utilities by tracking electric vehicle (EV) growth, confirming accurate transformer mappings, and identifying overloaded transformers. These capabilities ease some operational burdens while maintaining human oversight. Oracle employs a “human-centered AI” approach, designed to complement human decision-makers rather than replace them, particularly in high-stakes grid operations. This incremental approach aims to build trust with utility clients gradually, ensuring that AI solutions are reliable and offer incremental confidence over time. By integrating human expertise with AI, Oracle ensures that utilities can leverage advanced technologies while mitigating the risks associated with AI-driven automation.
Governance-First Approach: Ensuring Data Security and Compliance
PG&E’s AI strategy emphasizes strict oversight and gradual testing of AI applications within controlled environments. This alignment with regulatory standards, especially given California’s rigorous utility regulations, ensures adherence to data security, bias prevention, and compliance standards before any AI tool is fully deployed. Casey Werth, IBM’s global energy and utilities lead, underscores the importance of data governance in this context. Good data governance is foundational for any successful AI implementation, ensuring that data is accurate, standardized, and of high quality, which is necessary for AI models to operate effectively.
Many utilities are currently in the exploratory phase of their AI journeys, often grappling with data quality and consistency issues. Werth advises focusing on low-risk AI applications that can yield valuable insights with minimal need for extensive data restructuring or transformations. For example, utilizing machine learning models for satellite imagery analysis or asset mapping can offer substantial benefits without significant risks. This approach enables utilities to unlock the potential of AI while overcoming the challenges associated with data governance and regulatory compliance, thereby ensuring that AI applications are both effective and secure.
Addressing Data Challenges: Improving Data Entry and Standardization
One significant challenge facing utilities is the backlog of poorly documented or inconsistently labeled asset failure data, which is crucial for predictive maintenance. Without clean, standardized data, developing advanced predictive analytics for asset management remains an uphill task. To address this, IBM is working with utilities to improve data entry processes, such as using AI-powered language models to document asset failures verbally, converting spoken language into structured data. These efforts to standardize and improve data quality are essential for enabling accurate and reliable AI-driven insights, thus supporting long-term operational planning and asset management.
These AI-driven insights, while not yet extending into direct grid operations, play a critical role in long-term resource and system planning, climate adaptation, and asset management. Successful AI deployment necessitates a close alignment between IT departments, which manage the technological aspects, and business units, which handle the operational processes. A collaborative approach where teams work together seamlessly to understand and integrate AI technology effectively is essential. This cooperation ensures that AI solutions are tailored to meet the specific needs of utility operations while maintaining high standards of data quality and integrity.
Incremental Adoption: Balancing Innovation with Safety
The adoption of artificial intelligence (AI) by utility companies is increasing, especially in the United States, as they aim to boost operational efficiency and cope with rising demands. However, these companies are proceeding with caution, particularly regarding the deployment of AI in critical grid operations. This cautious approach stems from the industry’s stringent safety and reliability requirements, where even minor errors can lead to significant consequences. To mitigate risks, utility companies are ensuring that AI systems undergo thorough vetting and are implemented incrementally. This methodical strategy allows utilities to leverage AI’s potential while maintaining robust safety and reliability standards. By adopting this careful and deliberate process, utility companies can enhance their operations without compromising their core commitment to safety and reliability. This ensures that AI integration supports, rather than undermines, the industry’s high standards, thereby fostering trust and confidence in the evolving landscape of utility management.
