AI for climate: a real world test

Forecasting

The episode explores the challenges of forecasting in AI-driven climatetech, highlighting the limitations of AI-generated forecasts. Shayle Kann points out that while AI can perform tasks like multivariate regression, the quality of forecasts is crucial for decision-making 1. Seyed Madaeni emphasizes that poor forecasts can lead to suboptimal market positions, underscoring the "garbage in, garbage out" principle 1. Duncan Campbell shares his experience with ChatGPT, noting the difficulty in building a comprehensive forecasting model due to his unfamiliarity with regression techniques 2.

The quality of your forecast drives the quality of your decisions.

--- Seyed Madaeni

Despite these challenges, the episode acknowledges the potential for improvement in AI forecasting capabilities.

   

Market Complexity

The complexities of applying AI in energy markets are significant, particularly when dealing with ancillary services and market operations. Duncan Campbell admits to skipping forecasting due to the intricate nature of marrying day-ahead energy markets with ancillary services 3. Seyed Madaeni explains that ancillary services, which operate on a second-by-second basis, add a layer of complexity that is often reduced in simplified models 3. The episode also touches on the challenges of bidding energy storage assets in wholesale markets, a task complicated by the limited duration of energy storage 4.

Energy storage creates a whole host of challenges for market bidding.

--- Seyed Madaeni

These complexities highlight the need for advanced AI solutions to effectively manage energy market operations.