AI for Public Power Utilities

Brian Taylor, General Manager for Tennessee public power utility CDE Lightband, and Lance Haynie, Government Affairs Director for the City of Santa Clara, Utah, presented examples of artificial intelligence uses for public power utilities and cities on June 10. They were on a panel called “Practical Applications of AI for Utilities” at APPA’s National Conference in New Orleans, La.

Taylor said his utility started by creating an AI policy and is using Google Gemini as its platform. The policy requires AI uses to be mission critical or core to the business. CDE Lightband has a dashboard to access AI programs. The utility’s communications and marketing teams have access to AI for creative tasks.

AI Applications for Utilities

Taylor included a slide in his presentation that listed ways utilities can use AI:

Knowledge Retrieval
Customer Understanding (sentiment)
Customer Segmentation
Anomaly Detection in Data
Satellite Imagery Analysis
Power Demand Forecasting

CDE Lightband has a centralized data warehouse for its AI activities. Taylor said he is constantly asking staff to research things like board approvals of the new vacation policy or the Silicon Ranch contract. CDE Lightband is scanning board minutes back to 1938 to make them available through AI. This allows the utility to research prior board actions, including approval dates and board member votes.

Another AI application at CDE Lightband involves advanced metering infrastructure. The utility is also in the broadband business and wants to target market to those customers. It is using AMI data to cluster customers into similar usage patterns and segment customers. CDE Lightband recently launched a smart thermostat program, targeting 20,000 ideal customers instead of all 86,000 customers.

Haynie said AI improves efficiency and enhances decision making, especially for small staffs. The city trains the AI on its decisions, including legislation. Haynie also mentioned a “Human in the Loop” philosophy, ensuring a person is always involved.

Load Forecasting

Haynie said AI has helped the city with load forecasting, achieving 92 percent accuracy. Because Utah has sunny weather 300 days a year, its weather is more predictable. The city's forecasting determines whether to generate power in house or buy excess power on the market.

Haynie said this is a dollars and cents issue and the city went from around $15,000 in savings to a quarter million in one month. The previous manual system involved a scheduler inputting weather data from weather.com and load data from various sources. The new system pulls all data directly and is iterative, sending alerts and fixing errors itself.

Haynie also detailed how AI can be used for administrative efficiency:

Meeting Minute Summarization: Using AI to condense lengthy meeting minutes into concise and actionable summaries.
Document Analysis: Using AI to extract key information from lengthy documents, reports, and emails.
Time Savings: AI applications can reduce hours of administrative work.
Implementation Ease: Tools like ChatGPT are low hanging fruit requiring minimal technical expertise.