Quick answer
Size commercial solar by comparing generation with electricity use in each time interval, then applying the site’s actual import tariff, export compensation and interconnection limits. Compare at least two system sizes. The larger option should justify its additional cost through additional usable electricity or eligible export credits—not simply produce more annual kWh.
Start with: a full year of interval consumption, current electricity bills, a documented generation model and written confirmation of the proposed export arrangement.
This guide uses California’s PG&E service territory to illustrate the utility checks. The sizing method is transferable, but PG&E’s billing and interconnection provisions should not be applied to another utility.
1. Treat available space as a ceiling, not a sizing target
A roof survey answers how much solar might physically fit. It does not establish how much a business should buy.
The U.S. Department of Energy distinguishes rooftop technical potential from economic potential: roof size, shading, orientation and construction affect what can be installed, but physical potential alone does not account for cost or commercial viability. EPA’s Site Considerations guidance also identifies roof age and condition as important feasibility checks.
Begin with two separate questions:
- What is physically feasible? Check usable space, shading, roof condition, structural suitability and electrical infrastructure.
- What is commercially justified? Compare generation with consumption, eligible export value and the cost of each design.
Ask bidders to identify the site’s physical maximum separately from their recommended capacity. That makes it easier to see whether a proposal is driven by available space or by a documented electricity-use analysis.
A smaller system is not automatically better, either. The decision should come from comparing alternatives—not from a rule to fill the roof or avoid every export.
2. Put kW, kWh and AC/DC capacity on the same basis
These terms describe different things:
| Term | Meaning | Why it matters |
|---|---|---|
| kW | Power: the rate of producing or consuming electricity | Used for equipment capacity and electrical demand |
| kWh | Energy accumulated over time | Used for electricity consumption and solar production |
| kW DC | The photovoltaic array’s rated direct-current capacity | Common basis for describing panel capacity |
| kW AC | The inverter system’s rated alternating-current output capacity | Important when evaluating output limits and comparing designs |
PVWatts documentation distinguishes array nameplate capacity from inverter AC capacity and models inverter output limits. PG&E’s commercial tariffs likewise distinguish energy charges in kWh from demand charges in kW. (samrepo.nlr.gov)
For example, a hypothetical 100 kW DC array with 80 kW AC of inverter capacity has a DC-to-AC ratio of:
100 ÷ 80 = 1.25
That does not mean the system continuously supplies 80 kW. Its modeled output varies with weather, orientation, losses and inverter operation. Clipping occurs when available array output exceeds what the inverter can deliver; PVWatts represents this by limiting inverter output to its AC rating. (samrepo.nlr.gov)
Require every proposal to state both capacities. Comparing one bidder’s DC capacity with another bidder’s AC capacity creates a misleading comparison.
3. Build a consumption profile before choosing capacity
Annual consumption is a useful starting point, but it cannot show when electricity is needed. For interval-based solar billing, timing matters: PG&E’s Net Billing Tariff specifies 15-minute billing intervals for nonresidential customers. Its published energy-usage guidance also says electric usage data is available in 15-minute intervals. (pge.com)
Request a full year of interval data, then annotate it with operating information:
- Opening hours and production shifts.
- Weekends, holidays and seasonal shutdowns.
- Major equipment additions or removals.
- Planned efficiency improvements.
- Confirmed electrification projects and their expected schedules.
Before modeling, check that the dataset covers the correct meter and dates. Look for missing intervals, duplicate timestamps and inconsistent time-zone handling. Reconcile monthly totals against bills, investigating material differences.
If interval consumption is expressed in kWh, divide by the interval duration in hours to calculate average kW. For example, 15 kWh consumed over 15 minutes represents an average load of 60 kW.
For an existing solar site, distinguish total building consumption from electricity imported through the utility meter. PG&E’s NBT meters measure imports and exports; those readings should not be mistaken for gross building load. (pge.com)
Keep projected future consumption separate from measured consumption. Ask the modeler to run both, rather than quietly building uncertain expansion into the base case.
4. Verify export payment and export permission separately
Two documents answer two different questions:
- The billing tariff: What value can eligible exported electricity receive?
- The interconnection arrangement: How may the system operate and export?
In PG&E territory, the business Solar Billing Plan rollout began in March 2026. PG&E identifies customers who applied for interconnection after April 14, 2023, and customers whose 20-year NEM legacy period has expired; its guidance explains that affected existing business customers transition at their True-Up date. Confirm the actual account status rather than assuming all solar customers receive the same treatment. (pge.com)
Under that plan, export credit values vary by time of day, day of the week and season. Do not value every exported kWh at the electricity purchase rate. PG&E’s NBT also states that the ACC Plus export adder is unavailable to nonresidential customers. (pge.com)
Before approving a design, obtain confirmation of:
- The applicable import tariff and export-credit schedule.
- Any generation-provider provisions if another entity supplies electricity.
- The permitted export capacity and where it is measured.
- Required controls, protection and metering.
- The applicable interconnection process and operating authorization.
Non-export does not mean “no interconnection paperwork.” PG&E publishes a specific Rule 21 application attachment for facilities operating in parallel without export, including protection and operating-mode information. (pge.com)
5. Model seasonal production with documented assumptions
A useful generation forecast explains its inputs, not just its annual total.
PVWatts V8 supports hourly output and inputs including location, system capacity, array type, tilt, azimuth, losses, DC-to-AC ratio and inverter efficiency. Its documentation also identifies the weather dataset used. These are relevant checks when comparing proposals. (developer.nlr.gov)
Ask for a model record containing:
- Site location and weather-data source.
- Model name and version.
- DC array and AC inverter capacities.
- Mounting configuration, tilt and orientation.
- Shading, soiling and other loss assumptions.
- Monthly and hourly generation.
- Treatment of export restrictions and curtailment.
PVWatts estimates production; it does not guarantee a particular year’s output. Its V8 documentation describes typical meteorological year weather data, so treat the result as a planning estimate rather than a forecast of next year’s weather. (developer.nlr.gov)
Compare generation and consumption by month as well as by interval. Ask the modeler to show weekends and seasonal shutdowns explicitly.
If production is modeled hourly but consumption is available every 15 minutes, document how the datasets are aligned. Hourly averages can hide short-duration imports and exports; use finer-resolution analysis where billing or export-control decisions require it. PG&E’s nonresidential NBT billing interval makes this distinction particularly relevant. (pge.com)
6. Compare two sizes against exactly the same load
The following is an illustrative arithmetic example, not measured site data or a PG&E-area production forecast.
Assume a business in PG&E territory is comparing:
- Option A: 100 kW DC / 80 kW AC.
- Option B: 150 kW DC / 120 kW AC.
- No battery.
- Four simplified one-hour intervals.
- Constant load and solar output within each example interval.
- Export permitted without a binding limit for the first comparison.
The generation values are deliberately selected to explain the calculation. They are not PVWatts results.
| One-hour interval | Building use, kWh | Option A solar, kWh | Option B solar, kWh |
|---|---|---|---|
| 1 | 60 | 20 | 30 |
| 2 | 60 | 50 | 75 |
| 3 | 70 | 80 | 120 |
| 4 | 40 | 40 | 60 |
| Total | 230 | 190 | 285 |
For each interval:
- Direct self-consumption = the smaller of building use and solar generation.
- Grid import = building use minus direct self-consumption.
- Potential export = solar generation minus direct self-consumption.
The resulting comparison is:
| Four-hour result | Option A | Option B |
|---|---|---|
| Directly consumed solar | 180 kWh | 200 kWh |
| Grid imports | 50 kWh | 30 kWh |
| Exports | 10 kWh | 85 kWh |
Option B produces 95 additional kWh, but only 20 additional kWh serve the business directly. The remaining 75 additional kWh are exported.
Now assume a hypothetical 20 kW export cap, enforced by idealized controls. Because each example interval lasts one hour and output is constant, the cap permits up to 20 kWh of export per interval.
Option B’s third interval would otherwise export 50 kWh. It therefore curtails 30 kWh, reducing total exports from 85 to 55 kWh. Direct self-consumption remains 200 kWh. Option A is unaffected.
Under hypothetical zero-export operation, the surplus would instead be curtailed: 10 kWh for A and 85 kWh for B.
These calculations show why extra generation is not the same as extra on-site benefit. They do not determine which system is financially preferable. That requires a full-year comparison, actual tariff values, approved operating limits and project costs.
7. Compare incremental value and check the proposal
Evaluate the additional capacity, not just each system’s average cost.
For the larger option, request a separate calculation of:
Additional avoided import charges + usable additional export credits + demonstrated demand-charge savings − additional lifetime costs
Demand savings require their own analysis. PG&E’s B-19 tariff, for example, includes demand charges based on maximum demand in specified periods, with demand generally averaged over 15-minute intervals for applicable customers. A reduction in annual kWh is therefore not evidence of a proportional reduction in demand charges. (pge.com)
Request sensitivity cases for lower consumption, delayed expansion, different export restrictions and weaker production. Compare storage separately rather than assuming a battery makes an oversized array economical.
Before selecting a proposal, check for these common mistakes:
| Mistake | Practical check |
|---|---|
| Sizing to annual consumption alone | Require an interval comparison |
| Mixing AC and DC capacities | Show both ratings for every option |
| Assuming exports receive retail value | Use the applicable export-credit schedule |
| Ignoring shutdowns | Include weekends and seasonal closures |
| Counting curtailed energy as savings | Separate usable output from curtailed output |
| Assuming future load is certain | Show measured and projected-load cases separately |
| Claiming demand savings from annual kWh | Recalculate demand charges under the actual tariff |
Where a design relies on future consumption, check tariff eligibility too. PG&E’s NBT includes an oversized-generating-facility attestation and annual-production limits for qualifying expected increases in usage; those provisions are not a blanket authorization to fill every roof. (pge.com)
The decision-ready deliverable is a comparison showing where each additional kWh goes, how it is valued and what it costs. Until that exists, the largest system that fits is only a physical option—not a justified commercial choice.