Solar + Storage Sizing: How Much Battery Do You Actually Need?

There’s no one-size-fits-all answer to battery sizing

If you’ve ever tried to spec a battery for a solar installation, you’ve probably noticed something: Google will give you a formula, vendors will give you a rule of thumb, and your customer—whether it’s a utility or a commercial developer—will give you a number that “feels right.”

I’ve been on both sides of this equation. In my current role as a quality and compliance manager at a renewable energy company, I review around 200+ system designs annually. And over the past four years, I’ve seen the same mistake repeat: people start with the inverter spec, then add a battery as an afterthought. That’s backwards.

Battery sizing isn’t about finding the “right” number—it’s about understanding what you’re optimizing for. Your answer changes depending on whether you’re trying to avoid grid penalties, ensure backup for critical loads, or maximize self-consumption. So let’s break this down by scenario.

Scenario A: You’re optimizing for self-consumption (commercial rooftop)

This is the most common use case I see in commercial and industrial projects. The goal is simple: store excess solar energy during the day and use it in the evening, reducing grid imports. The key metric here is load-match factor—not total capacity.

In our Q1 2024 quality audit, we looked at 42 commercial rooftop designs. The ones that worked best used a simple rule: size the battery to cover 70–80% of the facility’s evening load (say 6 PM to 10 PM), not the entire night. Why? Because most commercial buildings have minimal load after 10 PM. Oversizing just adds cost without benefit.

My recommendation: Start with the facility’s 15-minute interval load data (you can pull this from the utility meter). Calculate the average evening load from 6 PM to 10 PM. Multiply by 4 hours. That’s your minimum usable capacity. Add 10–15% for degradation over the warranty period. That’s your spec.

For example, if the evening load averages 50 kW, you need 200 kWh of usable capacity. With a 10% degradation buffer, spec 220 kWh. On the inverter side, make sure your SMA inverter (or any string inverter in the mix) supports DC-coupled storage—otherwise you’ll lose efficiency in the AC conversion.

When I compared designs that followed this rule vs. those that used a “just add one more battery module” approach, the difference in ROI was stark. The rule-following designs averaged a 4.2-year payback. The overspec’d ones? 6.8 years. Same inverter, same panels—just too much battery.

Scenario B: You’re optimizing for backup power (critical infrastructure)

This scenario applies to utilities, hospitals, or any site where losing power isn’t an option. The conversation shifts from “what’s the cheapest way to offset grid use” to “what’s the minimum reliable capacity to keep the lights on.”

The numbers said go with a 4-hour backup based on peak load. My gut said that’s too tight. Something felt off about the load profile data—it was from a holiday week. Turns out I was right. When I requested a full year of data, the actual peak load was 35% higher than the sample week suggested.

Here’s what you need to know: backup sizing should be based on critical load only, but always add a safety factor. In my experience managing over 50 utility-scale storage projects, the formula that holds up best is:

• Identify critical loads (essential circuits only)
• Multiply by intended backup duration (usually 2–4 hours for grid-tied, 8–24 for islanded)
• Add 25% margin for unforeseen loads or extended outages
• Factor in inverter efficiency (typically 90–95%) and depth of discharge (80–90% for lithium)

For a utility substation with 200 kW critical load and a 4-hour backup target, the math looks like this: 200 kW × 4h = 800 kWh. Add 25% margin = 1,000 kWh. Adjust for 90% DoD: 1,000 / 0.9 = 1,111 kWh. That’s your battery spec.

And here’s the gut-check moment I’ve learned to trust: if your battery-to-inverter ratio feels high (like 2:1 or more), and you’re using a standard string inverter—well, you might want to look at central or hybrid inverters instead. SMA’s Sunny Central Storage or a dedicated battery inverter can handle deeper cycling better than a standard string inverter designed for solar-only duty.

Scenario C: You’re optimizing for grid services (utility-scale or commercial aggregator)

This is where things get interesting. When your battery is designed to participate in energy markets—frequency regulation, peak shaving, time-of-use arbitrage—size is determined by the market opportunity, not the load.

Let me rephrase that: you’re not sizing the battery for the building. You’re sizing it for the revenue stream.

In our 2023 designs for a 5 MW commercial aggregator project, we ran into a fascinating conflict. The numbers said the optimal size for frequency regulation was 2 MWh (C/2.5 rate—fast discharge, high cycle count). But the site’s peak shaving requirement demanded 4 MWh to avoid demand charges. We couldn’t do both with one battery without oversizing or cycling too deep.

The solution? We split the battery into two virtual blocks: one 2 MWh block dedicated to regulation (shallow cycling, high cycles), and one 2 MWh block for peak shaving (deeper cycles, fewer events). Same physical battery, different control logic. That project saved the client $180,000 annually in avoided demand charges plus $45,000 in regulation revenue—versus $0 in revenue if we’d sized for just one use case.

If you’re designing for grid services, ignore the load profile. Look at the market signals: what’s the capacity price in your zone? What’s the typical duration of a dispatch event? In many ISO/RTO markets, a 2-hour battery captures 80–90% of the revenue opportunities. A 4-hour battery captures maybe 95% but costs nearly double. That marginal 5% isn’t worth it.

How to tell which scenario you’re in

Every time I present this breakdown to a project developer or installer, they ask: “Okay, but what if I’m a mix of two?” Fair question. Here’s how I decide:

• Primary objective: What’s the customer’s stated goal? If they say “reduce my electric bill,” you’re in Scenario A (self-consumption). If they say “I can’t afford a blackout,” you’re in B (backup). If they say “I want to monetize my battery,” you’re in C (grid services).
• Duration of need: For Scenario A, look at 4–6 hour evening load. For B, look at 2–4 hour critical load. For C, look at market dispatch duration (often 1–2 hours for regulation, 2–4 for arbitrage).
• Cycle frequency: Scenario A sees 1 cycle/day (PV → battery → load). Scenario B sees 1–3 cycles/year (worst-case). Scenario C can see 1–2 cycles/day (regulation) or 0.5–1 cycle/day (arbitrage).

And if you’re still on the fence? Start with the smallest viable battery for the primary use case. Then add capacity only if the secondary use case has a clear revenue or savings stream. I’ve seen too many projects overspec because they tried to do everything at once.”

One more thing: the data matters more than the formula

In 2022, I rejected a batch of system designs where the battery sizing was based on a “solar production profile” from a different region. The vendor claimed it was “within industry standard.” We rejected it, and they redid the analysis using actual on-site load data. Result: the battery spec dropped by 40% compared to the original design. The client saved $68,000 on hardware alone.

So before you run any formula, get the data. One year of 15-minute interval data is ideal. If that’s not available, three months of data adjusted for seasonal variation is acceptable. Anything less, and you’re guessing—and that guess can cost you.

TL;DR (but really, read the section that applies to you)

• Self-consumption: Size for 4-hour evening load + 10–15% degradation buffer
• Backup: Size for critical load × backup duration + 25% margin, adjusted for DoD and inverter losses
• Grid services: Size for market opportunity, not site load—and don’t oversize for marginal revenue
• Mixed use cases: Use virtual partitioning rather than physical oversizing
• Data first: Without actual load data, any formula is just a guess