Yeah, I use AI.
Yeah, I used AI to help write this article.
But we’re not just spitting out AI-generated garbage for clicks. We can solve real-world problems with the help of ChatGPT and other AI tools. In this article, I’m going to show you exactly how we did that for a client. I created this video that shows the whole process.
Using AI and Solar Assistant to Solve a Generator Problem at an Off-Grid Home ⚡
Today we’re nerding out with data.
Earlier this week, a client mentioned that a competitor claimed we “just use AI for everything.” That made me laugh. It reminded me of something my ex–brother-in-law once told my nieces: “Uncle Jason’s not that smart. He just learns everything from the internet.”
I’ve retold that story a hundred times because it completely misses the point — who cares where I got the knowledge, as long as I have it and know how to use it.
AI is a beast when it comes to analyzing data. Sure, you can download system performance data into Excel and manipulate it to get the result you need, but there are much, much faster ways to process data. And, it can help you solve real-world problems in record time. Of course, you have to know the right questions to ask and how to feed AI the food it needs to regurgitate useful results.
Problem: Generator Runtime at an Off-Grid Beach House
We have an off-grid client in Naples. The system was working, but the generator kept kicking on — even when the house was unoccupied. That’s not supposed to happen. This system was designed for light use and minimal generator runtime when no one’s there.
So we had to figure out:
- Is this a solar problem or a battery problem?
- Do we need more panels? More storage?
- Or do we just need to manage the system differently?
Step 1: Data Dive with Solar Assistant + ChatGPT
We pulled data from the Solar Assistant dashboard and zoomed in on the trouble spots — especially July 1st and 2nd. The generator ran early in the morning, which shouldn’t happen during unoccupied periods.
We downloaded 30-second interval data:
- Load Power (blue line)
- PV Power (yellow line)
- Generator Runtime (red line)
We asked ChatGPT to do a simple comparison:
Load – PV = What batteries need to cover
Result: 40.53 kWh required from the batteries overnight.
This client has a 60 kWh battery, but only 80% is usable — around 48 kWh. That should be enough. The problem? On the afternoon of July 1st, solar never fully charged the batteries. So they started the night below 100%, and the generator had to step in.
Step 2: Repeat Analysis for July 5–6
Same drill. Generator ran again the morning of July 6th.
ChatGPT told us the battery had to cover ~40 kWh overnight — again, within its capacity. But same story: weak afternoon sun meant the batteries didn’t hit 100% the day before.
Step 3: Fix the Problem: Add Solar Panels
On July 9th, we added solar panels to the east and west roof sections. This gave us longer solar days:
- Early morning production to delay battery discharge
- Late afternoon generation to fully top off the batteries
Result? Night and day difference.
On July 10th and 11th:
- Batteries were at 100% by noon
- Solar offset afternoon loads
- Overnight usage dropped to ~27–37 kWh
- No generator runtime ✅
We kept feeding ChatGPT the data to compare all four datasets. The earlier periods needed ~41–42 kWh overnight. The upgraded periods needed significantly less, and that’s a big deal.
Conclusion: AI Isn’t a Shortcut. It’s a Tool. ⚙️
Yes, we use AI. Daily. Proudly. And effectively.
This isn’t about replacing expertise. It’s about enhancing it. By using AI + solar monitoring tools, we were able to:
- Identify the root cause of generator runtime
- Quantify overnight battery usage
- Make the case for more panels — in the right locations
- Prove the value of the upgrade with real data
Most importantly, we showed the client exactly why this mattered — in kilowatt-hours, not guesswork.
In off-grid systems, timing matters as much as total production. East- and west-facing panels can drastically reduce generator reliance by keeping batteries charged longer and earlier in the day.
This is how we do it. We don’t guess. We measure, model, and adapt.
If you’ve got an off-grid home (or want one) and you’re serious about performance, contact us.
