Optimizing Solar Energy Production with Reinforcement Learning

[nelsonic]

https://www.jackogrady.me/reinforcement-learning-solar/research-summary https://news.ycombinator.com/item?id=31197170

[nelsonic]

discussion on HN is surprisingly insightful! Basically given that the movement of the Earth in relation to the sun is 100% predictable, the optimal position of solar panels can be computed statically so the microcontroller uses considerably less power than if it had to run an ML library. If you know the coordinates of the panel array and can do basic math, you don’t need anything advanced to make this work.

One of the top comments is that the materials and energy cost to move panels outweighs the energy produced. But I have serious doubts about this. Something we will have to rest empirically. 💭☀️