Thanks! Coding would go a bit beyond the target audience here, but I do have some examples from experience and the internet. Whenever starting on a new problem, I've found there's two steps to repeat (neither are really coding time, more so training time). The first is to run some training to see if any hyperparameters of the RL agent need to be significantly adjusted (discount factor, learning rate, etc) and the second is to actually train the best combination of agents. For initiating the training, there's very little new coding to do.
Now, if you also had to code a simulation environment for the agent to interact with, then that could be significant coding as you move to solve a new problem. Updating the state features/action space are minimal code though. Hopefully that helps clarify!
A lot of my friends and colleagues don't work on AI but are really interested in the field, and I'm frequently asked "What is AI? / Can you send me some links to good examples?" These are awesome questions, but I found myself doing this a handful of times and decided to make a page to host it all in one place so they (and anyone else on the internet) could have easier access to this info. I'd love to hear from some people that are not practicing AI about whether or not they find the page helpful!
You certainly could, but that doesn't entirely account for shading / system degradation / site-specific diffuse light opportunities (consider a huge amount of light reflecting off the side of a mountain at some time of day). Those are both really difficult and time-intensive to model for, so there's a desire to have an AI that can simply learn those things specific to the system it's optimizing without humans having to do it. I see the larger impact of RL as scaling humanity's problem solving capability. If we have to use N human hours per installation to get to 97% optimality per installation but RL can use N/10000000 per installation to get to 95%, we could free up all those N human hours for things that RL still struggles with. Just my 2 cents though, it's a very fair question
So large solar farms are usually single-axis tracking, which provides a huge energy production benefit over static panels. Consider this panel's application more for a standalone installment, where it could make sense to use a dual-axis tracking panel over a single-axis or static panel. Re: solar farms and cost, I actually learned that around ~70% of the cost of solar installation is soft costs, not the actual panels/solar cells etc. Applying RL to much larger installations would be about finding non-obvious ways to leverage the single-axis tracking when the sun is not directly overheard. A fun problem for the future :)
> around ~70% of the cost of solar installation is soft costs
This is incorrect, especially for solar farms
Panels are indeed only about 1/3 of the cost.
Additional components, labour, inverter, mounting - another 1/3.
The last third is indeed soft(ish) cost, but this includes profit (duh..), certification, survey, tax, fees, etc. This can be reduced, but it is not going to magically disappear ...
Another fun fact : newer mega farms in low-altitude deserts are considering no-axis no-mounting zero-tilt - just laying the panels on the ground ... It all comes down to cost vs yield
A nice project by the way. Did you ever compare the results to pre-calculated angles based on time/location/season?
Farmers are placing them mounted vertically, like fences, in productive fields, running north/south, far enough apart to run a tractor between. They pick up power in morning and afternoon. The crops benefit from reduced heat stress and water loss, and yield more.
Panels that work with light from either side are preferred for this use.
