Commercial solar is won or lost in the sizing model, not on the roof. A 500 kW rooftop that clips 4% annually, ignores the demand tariff, and overvolts its strings on the coldest January morning is a mediocre asset wearing a good proposal. The same roof, sized with discipline — right DC/AC ratio, honest loss stack, demand-charge integration, and sensitivity-tested economics — throws off six figures of NPV difference over the financing term. We support EPCs and developers through the PES Supply commercial desk, and this is the engineering framework we apply before a single module hits a purchase order: methodology, climate-adjusted DC/AC ratios, shading and production tooling, interconnection constraints, demand-charge math, and IRR sensitivity. The numbers below are from real project models, not marketing decks.
1. The Sizing Methodology Framework
Commercial sizing runs a different sequence than residential. Residential starts with annual kWh. Commercial starts with three parallel questions that converge on one array size:
- Energy target: What fraction of annual kWh should the array produce? (Usually 60–100%, capped by roof, interconnection limit, or net-metering rules.)
- Demand target: How much coincident peak kW can the array — with or without storage — reliably shave? (Drives inverter and storage sizing.)
- Constraint envelope: Roof area and structural capacity, service size, NEC 705.12 interconnection method, and utility export limits. (Caps what the first two answers can actually be.)
Run all three, then take the minimum. The roof says 620 kW fits; the 2,000A service with a 2,000A main and a 120%-rule busbar calculation says 480 kW backfed; the demand model says 350 kW AC is where marginal value flattens. The answer is 350–480 kW depending on whether the interconnection uses a breaker or a line-side tap — and that engineering decision, made in week one, just moved the project IRR by a full point. Our complete solar sizing guide covers the residential-to-commercial methodology bridge; this article picks up at the point where the spreadsheets get serious.
2. DC/AC Ratio Optimization by Climate Zone
Recommended DC/AC Ratios by Application (2026)
The DC/AC ratio is the highest-leverage sizing decision on a commercial job. Inverter $/W keeps falling slower than module $/W, so loading more DC onto each AC watt remains the cheapest energy you can buy — until clipping eats the gain. These are the ratios our desk models as starting points:
| Application | Typical DC/AC Ratio | Rationale |
|---|---|---|
| Residential | 1.15–1.25 | Smaller arrays, higher inverter cost per watt, less clipping tolerance |
| Commercial rooftop | 1.20–1.30 | Balanced economics; clipping of 1–2% annually is economically acceptable |
| Utility fixed-tilt | 1.25–1.35 | Higher ratios justified by lower inverter $/W and scale |
| Utility single-axis tracker | 1.35–1.45 | Trackers extend shoulder-hour production, justifying higher DC loading |
| NREL ATB default (utility) | 1.34 | Reference benchmark for utility-scale modeling |
Climate Zone Impact
Climate shifts these targets by 0.05–0.10 in either direction. Hot climates (Phoenix, the Texas Triangle) push ratios up: temperature derating keeps the array off its nameplate more of the year, so more DC fits under the same AC ceiling before clipping bites. Cold, clear climates (Denver, the high desert) push ratios down: a 40°F March day with fresh snow albedo can drive a module above STC irradiance conditions, and clipping hours stack up fast. Cloudy maritime climates (Seattle, Portland — our home turf) tolerate higher ratios than the raw irradiance suggests, because diffuse light rarely drives the array past 80% of nameplate. Model it with 8,760-hour data; do not guess from the annual average.
Clipping Loss Modeling
Rule of thumb we validate against PVsyst output: at a 1.25 ratio, expect 0.5–1.5% annual clipping on fixed-tilt commercial rooftops; at 1.35, expect 1.5–3%; past 1.40, clipping compounds faster than the marginal DC gain in most C&I rate structures. The crossover calculation: marginal DC watt costs roughly $0.60–$0.80 installed (module + racking + labor increment, inverter already bought). If clipping destroys more value than that watt earns — typically above ~2.5% clipping — the ratio is too high. Inverter selection matters here too; units with wide MPPT windows and high DC input limits preserve options, and the 100 kW-class inverter collection is where most of our C&I projects start shopping.
3. Shading Analysis Tools: PVsyst vs. HelioScope
Every commercial roof has HVAC units, parapets, and vents. The question is never whether you have shade — it is whether your tool models it honestly. The two industry standards serve different project classes:
| Criterion | PVsyst | HelioScope |
|---|---|---|
| Best for | Utility-scale >5 MW, lender-specified reports | C&I 100 kW–5 MW, fast iteration |
| Platform | Desktop (Windows) | Cloud (browser) |
| Simulation time | Minutes to hours | 30–60 seconds |
| Shading method | 3D scene + 8,760-hour calculation | 3D design + real-time as you draw |
| Bankability | Industry standard for utility-scale | DNV-validated within 1% of PVsyst; accepted by most C&I lenders |
| Component library | Extensive, user-extensible | 45,000+ components |
| Financial analysis | Limited (requires external tools) | Built-in calculator |
Our desk's practical split: HelioScope for everything under 2 MW unless the lender's term sheet names PVsyst, then PVsyst with a HelioScope cross-check. When the two disagree by more than 2% on annual yield, the shading scene is wrong in one of them — usually a parapet height or a nearby tree line entered carelessly. Garbage scenes in, bankable-looking garbage out.
4. NREL PVWatts for Quick Feasibility
Before either paid tool, PVWatts earns its keep in the first client meeting. Fifteen minutes with the address, roof tilt, and azimuth produces a defensible annual kWh range that kills bad projects early — which is a service to the client, not a lost sale. Set system losses to 14–16% for C&I rooftops (higher than ground-mount utility work; rooftop heat and soiling are real), set the DC/AC to your target, and export the hourly file. That hourly file feeds the demand-charge analysis in Section 6, where the annual total alone would mislead you. For fast client-facing estimates, our solar ROI calculator runs the same logic with tariff inputs.
5. Interconnection Impact Studies
Interconnection is where commercial schedules go to die. Utilities study anything above their fast-track thresholds — often 25 kW to 1 MW depending on the feeder — for thermal loading, voltage rise, protection coordination, and fault-current contribution. A "study required" determination adds 6–18 months and can add five or six figures of upgrade cost. Screen the feeder early: ask the utility for the hosting capacity map if one exists, and size the AC export to sit under the screen thresholds when the marginal economics allow it.
IEEE 1547-2018 Requirements
Modern interconnection agreements invoke IEEE 1547-2018, which means your inverters must provide voltage regulation (volt-var, volt-watt), frequency ride-through, and — in Category B — specific performance under abnormal grid conditions. Specify UL 1741 SB-certified inverters on the BOM; discovering at witness-test time that the specified unit only carries the older SA certification is an expensive re-order. Rule 21 in California and similar tariffs elsewhere add smart-inverter function requirements on top. Our NEC compliance guide covers the code side; the utility's interconnection handbook covers the rest, and it wins every argument.
6. Demand Charge Reduction Analysis
For most C&I customers, demand charges are 30–50% of the bill, and solar alone addresses them poorly — a single cloudy 15-minute interval at 2 p.m. sets the month's demand charge at full value. This is the analysis that separates engineered projects from panel-selling:
| Rate Structure Scenario | Solar-Only Demand Reduction | Solar + Storage Demand Reduction | Annual Value (100 kW peak shave @ $18/kW) |
|---|---|---|---|
| Coincident afternoon peak, non-ratcheting | 60–80% reliable | 95%+ reliable | $12,960–$21,600 |
| Evening peak (5–9 p.m.) | 5–15% reliable | 90%+ reliable | $1,080–$21,600 |
| Morning peak (winter, 7–9 a.m.) | ~0% | 90%+ reliable | $0–$21,600 |
| Ratcheting tariff (12-month lock) | Unreliable — one miss locks the year | High, with conservative dispatch | Binary: full value or near zero |
The table explains the storage boom in one glance: solar without storage captures demand value only when the peak is coincident and the tariff forgiving. Adding a 100–200 kWh C&I battery block converts weather risk into a dispatch problem. Size the battery kW to the shave target and the kWh to the peak window length — 2 to 4 hours covers most C&I tariffs — and our battery cost guide provides the $/kWh benchmarks for the capital side of that model.
7. IRR Modeling with Sensitivity Tables
A single-point IRR is a sales tool. A sensitivity table is an engineering tool. The two variables that dominate commercial outcomes are installed cost ($/W) and the DC/AC-driven energy yield; rate escalation and degradation matter, but they move the answer less than people assume.
Sensitivity Table Structure
Hold the tariff and degradation constant. Vary installed cost across the columns and DC/AC ratio (a proxy for yield design choices) down the rows. Read the diagonal: the design point sits where your actual quoted cost meets your modeled ratio.
IRR Sensitivity: DC/AC Ratio vs. Installed Cost (500 kW Case Study)
| DC/AC Ratio ↓ / $/W → | $1.80/W | $2.00/W | $2.20/W | $2.40/W |
|---|---|---|---|---|
| 1.15 | 14.2% | 12.1% | 10.3% | 8.7% |
| 1.20 | 15.1% | 12.9% | 11.0% | 9.4% |
| 1.25 | 15.6% | 13.4% | 11.5% | 9.8% |
| 1.30 | 15.8% | 13.6% | 11.6% | 9.9% |
| 1.35 | 15.5% | 13.3% | 11.4% | 9.7% |
Two lessons fall out of this matrix. First, cost discipline beats design cleverness: moving from $2.20/W to $2.00/W is worth about two IRR points, more than any ratio tuning. Second, the optimum is a plateau, not a peak — anywhere from 1.20 to 1.35 performs within 0.3 IRR points at a given cost. Stop optimizing the ratio past the second decimal and go negotiate the EPC contract instead.
8. Case Study: 500 kW Commercial Rooftop
Site Profile
A distribution warehouse in the Southeast: 65,000 sq ft of usable roof, 480 V three-phase 2,000A service, 685 kWh/day average consumption, and a demand tariff at $15/kW with afternoon coincident peaks. Fixed tilt at 10°, ballasted racking, 1,136 × 440 W modules, 5 × 100 kW string inverters at a 1.25 design ratio (final 1.22 after string rounding). String voltage checked at the record low of −5°C: 22 modules per string × ~49.5 V cold Voc ≈ 1,089 V, inside the 1,100 V inverter ceiling with margin to spare — and NEC 690.7 satisfied on paper, not at inspection.
Economic Analysis
| Parameter | Value |
|---|---|
| Installed cost | $1,050,000 ($2.10/W) |
| Federal ITC (30%) | $315,000 |
| Net installed cost | $735,000 |
| Annual energy savings | $95,900 (at $0.14/kWh) |
| Annual demand charge reduction | $32,400 (180 kW × $15/kW × 12 months) |
| Total annual savings | $128,300 |
| O&M cost (first year) | $5,000 |
| Simple payback | 5.7 years (5.96 net of O&M) |
| 25-year NPV (8% discount) | $612,000 |
| IRR | 13.4% |
Design Decisions
Three calls made the returns. First, the line-side tap: the 120% rule on the 2,000A busbar would have capped backfeed at 400A (about 330 kW at 480 V three-phase), so a supply-side connection under NEC 705.12(A) unlocked the full 500 kW. Second, 180 kW of the demand shave was contracted as "firm" only with a 100 kW / 215 kWh storage addition in year two — phased to match the customer's capital calendar. Third, module choice favored temperature coefficient over nameplate efficiency: at this site's roof temperatures, a −0.29%/°C module out-produces a −0.35%/°C module of identical STC rating by roughly 2% annually. Browse current C&I module options in the commercial panel collection, and pair them with racking hardware kits rated for the site's wind exposure category.
8.5 Reading Interval Data Like an Engineer
The interval CSV is the most underused document in commercial solar. Thirty-five thousand rows of 15-minute readings look like noise until you pivot them three ways. First pivot: monthly peak kW and the hour it occurred — this tells you whether the peak is solar-coincident. Second pivot: a heatmap of load by hour and month — this reveals seasonality that annual totals hide, like a summer-cooling peak that solar eats for free or a winter-morning peak it never touches. Third pivot: load duration curve — sort all 35,040 intervals descending and look at the shape. A flat curve at 60% of peak means baseload dominates and PV plus storage both pencil; a spiky curve with 5% of intervals carrying 30% of the demand value means storage earns everything and PV is a supporting actor.
Two data traps to check before trusting the file. Daylight-saving transitions create 23- and 25-hour days that naive spreadsheets misalign by an hour for months — and a one-hour shift is exactly the difference between "peak at 2 p.m." and "peak at 3 p.m.," which is the difference between solar-covered and solar-missed. And meter resets or estimated reads show up as zeros or repeated constants; filter intervals where kW equals zero during business hours and flag them as suspect before you average anything. We keep a standing rule on the desk: no commercial proposal goes out until a second person has re-pivoted the interval file independently. Two analysts, one CSV, thirty minutes — it has caught a daylight-saving misalignment and one utility's bizarre practice of reporting kWh-per-interval-as-kW, which doubled the apparent demand. Either error would have been a five-figure design mistake shipped with our name on it.
8.6 Conductor Sizing and Voltage Drop on Long Commercial Runs
Commercial rooftops mean long AC runs — 150 to 300 feet from inverter to switchgear is normal, and voltage drop becomes a design constraint instead of a footnote. The NEC does not mandate a voltage-drop limit (215.2(A) Informational Note suggests 3% branch / 5% total as good practice), but your inverter does: sustained high grid voltage at the point of connection pushes units toward their over-voltage trip threshold, and a 2% rise at full output on a marginal feeder is enough to start nuisance trips on hot afternoons.
Size from NEC 310.16 ampacity first (continuous load × 125% per 690.8), then check drop. Worked example: a 100 kW inverter at 480 V three-phase delivers 120.3 A full output. Continuous sizing: 120.3 × 1.25 = 150.4 A → a 150 A breaker per NEC 240.6 and 1/0 AWG copper at 75°C (150 A ampacity) as the floor. Now the drop check on a 250-foot run: 1/0 AWG copper at ~0.122 Ω per 1,000 ft gives a three-phase drop of about 1.27% at 120 A — acceptable. Had the run been 400 feet, 1/0 would push past 2% and stepping to 3/0 (0.0766 Ω/1,000 ft, 200 A) keeps both ampacity margin and drop comfortably inside targets. Conduit fill follows NEC Chapter 9, Table 1 — three 1/0 THHN-2 conductors plus a #6 EGC want a 1.5-inch EMT at 40% fill, and nobody enjoys pulling 400 feet of 1/0 through an undersized raceway they sized from a chart they half-remembered.
| Inverter AC Output (480 V 3φ) | Full-Load Amps | 125% Continuous | Breaker (NEC 240.6) | Min. Copper (75°C, 310.16) | Max Run for ≤2% Drop |
|---|---|---|---|---|---|
| 50 kW | 60.1 A | 75.2 A | 80 A | #4 AWG (85 A) | ~310 ft |
| 75 kW | 90.2 A | 112.8 A | 125 A | #2 AWG (115 A) | ~250 ft |
| 100 kW | 120.3 A | 150.4 A | 150 A | 1/0 AWG (150 A) | ~250 ft |
| 125 kW | 150.4 A | 188.0 A | 200 A | 3/0 AWG (200 A) | ~285 ft |
| 150 kW | 180.4 A | 225.5 A | 250 A | 4/0 AWG (230 A) | ~275 ft |
Run the drop calculation at full output with the conductor at operating temperature, not the handbook 20°C resistance — copper resistance rises roughly 0.4%/°C, and a rooftop conduit in August sun is not at 20°C. The electrical supplies collection stocks the THHN-2, lugs, and EMT these runs consume by the spool.
8.7 Degradation, Availability, and O&M Reserves
Three quiet numbers decide whether year-15 cash flow matches the pro forma. Module degradation: 0.4%/year for premium N-type, 0.55% for standard P-type — over 25 years that gap is ~3.5% of lifetime energy, worth $15,000–$25,000 on a 500 kW project, which is why the cheapest module per watt is not always the cheapest module per lifetime kWh. Availability: budget 99–99.5% for string-inverter plants with a monitoring service; a single 100 kW unit down for a two-week parts wait costs ~2,800 kWh plus truck rolls. O&M reserve: $8–$15/kW-year covers inspections, cleaning where soiling warrants, and vegetation or roof-membrane coordination; below that, you are deferring maintenance into somebody else's headache, and "somebody else" is usually the owner's opinion of your company at year five.
9. Common Engineering Mistakes in Commercial Sizing
- Annual-total demand analysis. Modeling demand savings off annual kWh instead of 15-minute interval data. The month is billed on its worst quarter-hour; model quarter-hours.
- Ignoring the 120% rule until the permit set. NEC 705.12 decides your interconnection method in week one, whether you check it or not.
- Nameplate string voltage. STC Voc at 25°C is not the design voltage. NEC 690.7 cold-temperature correction is mandatory; record lows, not average lows, set the multiplier.
- Zero clipping tolerance. Chasing 0.0% clipping leaves energy value on the table; 1–2% annual clipping is usually the IRR optimum.
- Single-point economics. One IRR, one escalation rate, one degradation figure. Lenders and CFOs read sensitivity tables; give them one.
- Forgetting degradation in the pro forma. 0.4–0.55%/year on modules, plus inverter replacement reserves around year 12–15 for string units.
- Roof-first design. Maxing module count to fill the roof, then discovering the load only absorbs 60% of production under a non-carryover net-billing tariff.
8.8 What the Lender Actually Reads
If the project carries third-party money, the model you built for the owner is not the model that gets financed. Lenders haircut everything: P90 production instead of P50 (typically 6–8% lower), degradation at the conservative end, O&M escalated at 2.5–3%, and a debt-service coverage ratio floor around 1.25–1.30 that effectively caps leverage. A project showing 13.4% IRR unlevered at P50 can come back from the bank as a 9% levered IRR at P90 with 60% debt — still financeable, but only if your pro forma was honest enough to survive the haircut. Build the P90 case yourself before the lender does it to you. The designers who lose deals are rarely the ones with bad engineering; they are the ones whose P50 optimism made the bank's credit committee do the conservative math in a room the designer was not invited into.
Documentation matters as much as numbers at this stage: the shading report, the interval-data analysis, the string-voltage calculations at record-low temperature, and the structural letter should all exist as standalone exhibits. When they do, diligence takes weeks. When they have to be created after the term sheet, it takes months — and months kill commercial solar deals more reliably than bad IRRs ever have.
The Engineering Bottom Line
Commercial array sizing is a constrained optimization with four levers — DC/AC ratio, interconnection method, storage pairing, and installed cost — and cost is usually the strongest one. Model with interval data, check voltage windows at record temperatures, demand UL 1741 SB on the BOM, and present sensitivity tables instead of point estimates. Do the daylight-saving pivot check on the interval file, run the voltage-drop math at August conductor temperatures, and price the P90 case before the bank does. None of this is glamorous. All of it compounds: each percentage point of avoidable loss or avoidable cost is a point of IRR, and points of IRR are what turn a warehouse roof from a feel-good capital expense into the best-performing asset on the property. When the model is done, the procurement is straightforward: modules from the solar panel collection, conversion from the inverter collection, storage from energy storage systems, and a firm number from our quote desk.
Field Notes from Commercial Installs
We have stood on enough flat roofs in August to know that the model and the membrane disagree more often than the drawings admit. On a 750 kW job in Central Point, the structural letter came back after design freeze: 3.5 psf allowable, not the 5 psf everyone assumed — we re-spaced the ballast trays and dropped 40 kW rather than trigger a roof reinforcement that would have killed the IRR. I have watched a $40,000 switchgear upgrade appear in an interconnection study that a ten-minute hosting-capacity check in month one would have predicted. And on a cold-storage facility, the "solar-friendly" load profile in the owner's head turned out to be a 3 a.m. defrost cycle that owned the demand peak — the design flipped from 600 kW of PV to 400 kW plus 200 kWh of storage in one interval-data session. The roof is the easy part. The tariff is the design. A last one for the road: we inherited a 400 kW design from a competitor whose model assumed 100% self-consumption on a site that closes for two weeks every July. Fourteen days of full export at avoided-cost compensation — under this utility, roughly a quarter of retail — was enough to knock the promised payback from 6.1 years to 7.4. The owner had signed nothing yet, so the fix was free: resize to 340 kW and add a modest storage block. Had the array already been on the roof, that same fix would have been a change order nobody forgets. Model the calendar, not just the clock — holidays, shutdowns, and seasonal schedules are load profile too.
Frequently Asked Questions
What DC/AC ratio should a commercial rooftop use?
1.20–1.30 for most fixed-tilt C&I rooftops, nudged up ~0.05 in hot climates and down in cold, clear ones. Expect 0.5–2% annual clipping in that band, which is the economic optimum for typical rate structures. Verify against 8,760-hour simulation, not rules of thumb, before freezing the design.
How much do demand charges change commercial solar economics?
Enormously. Demand charges run 30–50% of many C&I bills, but solar alone only captures them reliably when peaks are solar-coincident and the tariff is non-ratcheting. Adding storage converts demand savings from weather-dependent to dispatchable — often the difference between a 9% and a 13% project IRR.
Is PVsyst or HelioScope better for a 1 MW commercial project?
HelioScope for speed and iteration on 100 kW–5 MW C&I work; PVsyst when the lender specifies it or the project is utility-scale. DNV validation puts HelioScope within ~1% of PVsyst on annual yield. If the two disagree by more than 2%, your shading scene is wrong in one of them.
What does IEEE 1547-2018 require of my inverters?
Voltage regulation functions (volt-var, volt-watt), defined ride-through behavior, and Category A/B performance during grid abnormalities. Practically: specify UL 1741 SB-certified inverters and confirm the utility's interconnection handbook requirements before the BOM is final.
How do I model battery storage for demand reduction?
Size battery kW to the peak shave target and kWh to the peak window duration (typically 2–4 hours). Then stress-test against the tariff: ratcheting structures punish a single missed dispatch for 12 months, so conservative state-of-charge reserves are worth more than maximal cycling.
What installed cost should I assume for commercial rooftop solar in 2026?
$1.80–$2.40/W turnkey for most 250 kW–1 MW rooftops, with ballasted flat-roof work at the lower half and complex structural or interconnection scopes at the top. Every $0.10/W moves IRR by roughly half a point — negotiate hard.
Related Products & Collections
- Commercial solar panels — high-wattage modules for C&I rooftops
- 100 kW-class inverters — three-phase string conversion
- C&I battery blocks — 100–200 kWh demand-management storage
- Request a quote — project pricing through the PES commercial desk
