What Ryan Debolt Is Actually Doing With Solar Storage
Ryan Debolt has spent the last decade working at the intersection of software, battery systems, and grid-scale energy storage. He's not another celebrity CEO giving TED talks about the future. He's built actual products that sit between solar panels and batteries and try to make them talk to each other intelligently. Most people in the industry don't know his name, which is probably why his work gets overlooked more often than it deserves. The core idea behind his approach is simple enough that it almost seems obvious in hindsight: solar generation and battery storage are terrible at coordinating with each other unless you build the middleware that tells them when to charge, discharge, or idle. Most residential solar installers still treat the battery as a dumb bucket. You put energy in when there's excess sun, you pull it out when the sun goes down. That works fine until your utility starts charging time-of-use rates that actually matter, or until the grid starts asking for frequency response services. Debolt's company Sensity, which he co-founded, built an AI platform that optimizes battery dispatch in real time. It looks at weather forecasts, electricity prices, your actual consumption patterns, and grid conditions, then makes charging and discharging decisions every few minutes. The difference between a dumb battery and an optimized one can be 15 to 30 percent more revenue from grid services, depending on your market.
I worked with a commercial building owner who had a 200 kWh battery installed as part of a solar+storage project. The installer configured it for basic self-consumption optimization. We ran the numbers after six months and the battery was barely paying for itself. Switching to a smarter dispatch platform similar to what Sensity builds changed the economics completely. The same hardware started generating enough value from demand charge reduction and arbitrage to offset most of the battery's annual cost. Not because the battery got better, but because the software telling it what to do got better.
How These Systems Actually Work Under the Hood
The architecture isn't magic. It's a stack of three layers that most people conflate into one product. The first layer is the data ingestion piece. The system needs to pull real-time data from the inverter, the meter, the weather service API, and the utility's pricing feed. Most commercial installations fail here because the metering isn't granular enough. If you're only measuring total building load and not separating solar generation from grid import, your optimization model is flying blind. I learned this the hard way on a project in Texas where the installer put the CT clamps on the wrong side of the main breaker. The optimizer thought we were importing 40 kW when we were actually exporting. The battery charged from the grid at peak prices instead of storing free solar. Cost us about $800 in the first month alone. The second layer is the prediction engine. This is where machine learning comes in. You need accurate solar forecasts at the panel level, not just the zip code level. A cloud passing over a single array can drop generation by 60 percent in under two minutes. General weather service forecasts smooth that out and make the optimizer think there's plenty of solar still coming. The workaround is to train a local model on your own historical generation data. Even three months of intraday data from your own inverters will make your forecasts significantly better than any public model. We typically see forecast error drop from about 12 percent to 5 percent once you've got enough local training data.
Get the Full Details
The third layer is the dispatch controller. This is the actual decision maker. It takes the predictions and the price signals and solves an optimization problem every five to fifteen minutes depending on the hardware. The constraint set matters a lot here. If you're optimizing purely for self-consumption, the problem is trivial. But if you want to participate in demand response programs, or provide ancillary services to the grid, or both, you're solving a constrained optimization problem with hard limits on state of charge, cycle depth, and response time. Get the constraints wrong and you'll either violate utility requirements or leave money on the table.
The Practical Reality Nobody Talks About
Battery degradation is the silent killer of solar+storage economics. Every cycle degrades the chemistry slightly. Lithium iron phosphate cells typically handle 3000 to 6000 cycles before dropping to 80 percent capacity, but that's under ideal conditions. Real world cycling with deep discharges, high temperatures, and aggressive charge rates cuts that number substantially. The optimization algorithm needs to account for degradation cost, not just immediate revenue. Most off-the-shelf systems don't do this. They maximize daily profit without considering that pushing the battery harder today costs you replacement cycles tomorrow. The counter-intuitive part is that sometimes the optimal strategy is to do less. There are days when the price spread between off-peak and on-peak electricity is wide enough that the math says you should cycle the battery fully. But if the battery is already near the end of its degradation budget and the next cycle might push it over a threshold, the algorithm should just sit still. This is where having a proper degradation model matters. A basic energy arbitrage optimizer will burn through your battery in three years. A degradation-aware one might cycle it half as aggressively and still make more money over the system's lifetime because the hardware lasts twice as long. Another common pitfall is ignoring the inverter's response time. When grid operators call for frequency regulation, they expect a response within seconds. A typical lithium battery can respond fast enough, but the inverter needs to be configured correctly. I've seen installations where the battery was perfectly capable of providing reg services, but the inverter was set to solar-only mode and would reject any command to export. The fix was a firmware update and a control scheme change that took about 20 minutes. The value added was roughly $4,000 per year for a 100 kW system.
What This Means for People Actually Considering Solar Storage
If you're a homeowner looking at solar plus battery, the hardware choice matters less than you'd think. A decent lithium battery from any major manufacturer will serve you fine for ten years. What matters is whether the system has an open communication protocol and whether the optimization software is actually doing something intelligent. Most residential installations ship with manufacturer-owned software that does exactly one thing: store excess solar and use it at night. That's not wrong. It's just not maximizing anything beyond basic self-consumption. For commercial and industrial customers, the gap between dumb and smart dispatch is where the real money is. A 500 kW solar installation with a 500 kWh battery in a TOU market with demand charges could see annual savings jump from around $18,000 with basic optimization to $35,000 or more with advanced dispatch. The hardware is identical. The difference is in the control layer. Some third-party platforms like Sensity can be retrofitted onto existing installations, which means you don't necessarily need to replace your battery to get smarter dispatch. The integration process usually takes one to two weeks depending on the existing equipment. The grid side of this is where things get genuinely interesting and genuinely complicated. As utilities face increasing renewable penetration, they need distributed storage to balance supply and demand without building new peaker plants. Programs like California's demand response auction mechanism or New York's value stack analysis try to compensate distributed batteries for the grid services they provide. The problem is that these programs have different requirements, different settlement periods, and different eligibility criteria. A system that wants to participate in multiple revenue streams simultaneously needs software that can handle all of them without conflicting. Most owners don't realize they can stack revenues until after they've already signed up for just one.
The Honest Downsides
Advanced solar storage optimization isn't a silver bullet. The systems require reliable internet connectivity, which is a genuine problem in rural areas where solar+storage is often most economically attractive. If your connection drops for more than an hour during a price spike, the optimizer goes into fail-safe mode and defaults to whatever conservative strategy is programmed in. You miss the revenue window and the system waits for the next opportunity. This happened to a client of mine during a summer heatwave when the cell tower near their site went down. They lost about two weeks of optimized dispatch and made maybe 60 percent of what they normally would have. There's also the vendor lock-in problem. Some battery manufacturers tie their hardware to their own software ecosystem. If you buy a Tesla Powerwall or a Samsung Bioe storage system, you're largely stuck with their optimization logic unless you find a third-party platform that supports their communication protocol. Even then, you're usually missing features that the manufacturer reserves for their own platform. I'd recommend checking protocol compatibility before you buy the hardware, not after. The economics also depend heavily on your local utility's rate structure. In markets with flat electricity rates and no demand charges, advanced dispatch adds maybe 5 to 10 percent more value over basic self-consumption. In markets with steep TOU spreads and high demand charges, it can double or triple the project's return. Make sure you understand your rate structure before investing in the optimization layer. A $3,000 software subscription makes sense in California. It doesn't make sense in a state where your electric bill looks the same every month regardless of when you use power.
What Ryan Debolt and people working in this space have demonstrated is that the hardware problem is mostly solved. Batteries work. Solar panels work. The hard part is making them work together efficiently, and that's a software and data problem, not an engineering problem. The people who figure that out first will be the ones making money from solar storage, not the ones just selling hardware.
