Weather, Season,
and Derate

A good annual model can still produce a bad capacity decision. Weather changes the load, the available output, and the auxiliaries that keep equipment inside its operating envelope. The owner therefore needs more than an energy forecast. The owner needs a dated, site-specific test of the hours in which demand is high and capability is low.

Section 01The average is not a capacity condition

Annual energy answers how much electricity or fuel moves through a system over time. Capacity answers whether the system can carry a particular load in a particular interval. The two are related, but they are not substitutes. A site can have a favorable annual balance and still fail during a hot afternoon, a cold start, a cloudy evening, or a period when essential loads and maintenance overlap.

Weather matters on both sides of the balance. It can raise building cooling, refrigeration, pumping, ventilation, and process loads. At the same time, it can reduce the net output of solar modules, air-breathing generators, fuel-cell auxiliaries, batteries, transformers, and cooling systems. Smoke, wind, humidity, rain, snow, and drought can introduce other limitations. The decisive condition is often coincidence, not a single extreme measured in isolation.

California’s system operator applies the same principle at grid scale. Its May 2026 summer assessment models hourly combinations of load, solar, wind, and resource outages, while warning that coincident extreme heat, wildfire, drought, and other disruptions remain outside a standard planning result.¹ A site study should be equally explicit about what its scenarios include and what they do not.

Section 02Build a weather envelope, not one weather year

Start with three distinct records. First, use recent site operations to show what the facility actually did. Second, use long-run weather statistics to describe the recurring climate. Third, add selected stress periods that expose conditions an average cannot represent. Each record has a different purpose and should retain its own dates and provenance.

NOAA defines a climate normal as a 30-year average. The current official U.S. normals use 1991–2020 observations and include annual, seasonal, monthly, daily, and hourly statistics.² Those sourced dates matter. Normals are useful for typical seasonal shape and for checking whether an operating year was unusual. By definition, however, a normal is not the worst hour and does not prove a design condition.

The weather envelope should align dry-bulb temperature, humidity or wet-bulb condition, solar irradiance, wind, precipitation, and any site-relevant smoke or icing record to the same clock as the electric interval data. Elevation and sensor location also matter. An airport station can be a defensible regional source, but it is not a roof, turbine inlet, battery enclosure, or shaded process yard. The distance and physical difference should be stated.

Do not force one event to represent every risk. A hot, clear afternoon may be difficult for cooling load and equipment heat rejection while still favorable for solar irradiance. A smoky or overcast event may reduce solar without producing the annual temperature maximum. A cold morning may increase heating or starting load while changing battery capability. Separate scenarios preserve those mechanisms.

Section 03Derate is a curve, not a haircut

A derate converts a stated rating into expected capability under stated conditions. It should be a function of the variables that actually govern the equipment: temperature, elevation or pressure, humidity, fuel quality and pressure, state of charge, operating history, degradation, cooling method, and auxiliary demand. A single percentage copied across all hours hides the relationship the study needs to test.

The boundary must be net output at the point the site uses it. Gross stack, engine, or array production is not the same as net alternating-current power after pumps, blowers, compressors, inverters, heating, cooling, controls, fuel treatment, and transformers. On the difficult day, some auxiliaries work harder. Reporting gross output while modeling the site load at the revenue meter creates an invisible gap.

Vendor curves should therefore be collected with their rating conditions, revision date, fuel basis, permitted operating range, and treatment of auxiliaries. A curve at standard conditions is documented evidence of the equipment’s stated behavior, not a promise of site output. Until the supplier confirms the exact configuration and boundary, the result remains a feasibility estimate.

As an explicitly illustrative example, suppose a unit with 100 units of nameplate capability is expected to provide 88 under the chosen stress condition. If five units are reserved for a maintenance or uncertainty allowance, the modeled contribution is 83. The arithmetic is exact. Every value is illustrative, and no value describes a real product or site.

Section 04Solar and storage have different weather jobs

Solar output is shaped by irradiance, module temperature, orientation, shading, soiling, clipping, and availability. The U.S. Department of Energy notes that solar cells generally perform better at lower temperatures; higher temperature produces a small current increase but a larger voltage decrease.³ DOE also notes that field output varies with season, time of day, clouds, dust, haze, obstructions, rain, snow, and dirt.⁴

The honest case for solar is strong daytime energy with low operating burden and a well-established modeling base. The honest case against it is equally clear: high annual production does not establish output during the site’s capacity hour. The analysis must use coincident irradiance and temperature, not an annual capacity factor, and it must show the result after inverter and site losses.

Storage solves a different problem. It can move energy from a favorable period to a difficult one, control a short peak, support power quality, or bridge a defined interruption. It cannot create energy, and its power rating does not reveal its usable duration. DOE distinguishes storage power capacity from energy capacity and notes that conversion and retrieval always involve losses.⁴

Temperature adds another layer. Sandia’s published operations guide states that stationary lithium-ion systems are generally designed around 20°C to 25°C and that operation materially above or below the designed range can reduce charge and discharge efficiency and shorten life. It also explains that enclosure heating and cooling are operating loads, especially in harsh climates.⁵ A battery schedule must therefore include thermal auxiliaries, state of charge before the event, reserve policy, degradation, and a recharge path after discharge.

Section 05Firm generation is not immune to weather

Combustion turbines and microturbines breathe ambient air. Hotter air is less dense, reducing mass flow, while compression work can rise relative to output. EPA’s technology catalog identifies ambient temperature and pressure as material drivers of both output and efficiency for these machines.⁶ Inlet cooling may recover capability, but it adds capital, water or electrical use, maintenance, and another system that must perform during the same stress event.

Reciprocating engines are generally less sensitive to hot air than turbines, but they still derate with temperature and elevation. EPA’s dated catalog gives an approximate planning benchmark of a 1 percent reduction for each 10°F above 77°F and about 4 percent per 1,000 feet of elevation above 1,000 feet.⁶ Those are sourced catalog approximations, not guarantees. The actual project must use the selected configuration’s current certified curve, cooling arrangement, fuel, and emissions limits.

Fuel cells also have weather-dependent balance-of-plant equipment. EPA states that system output and efficiency can decline as ambient temperature or elevation rises, with air blowers and compressors driving much of the effect.⁶ The core electrochemical process does not remove the need to model air handling, fuel treatment, pumps, heat rejection, inverter performance, and the technology-specific operating envelope.

Linear generators and other newer conversion platforms deserve the same neutral test. Their modularity or control range may be valuable. Their performance evidence may also be younger or more configuration-specific. The owner should request a net-output surface across temperature, elevation, fuel condition, part load, degradation state, and auxiliary demand rather than borrowing a curve from another technology.

Section 06Thermal conditions can reverse the ranking

Combined heat and power is not one seasonal result. The electric unit may derate in heat while the useful thermal sink shrinks, grows, or changes temperature level. A winter process may value recovered heat when ambient conditions favor electric output. A summer facility may have little direct heat demand unless absorption cooling or another real sink is present. The heat balance must be coincident with the electric balance.

The case for engines, turbines, microturbines, and fuel cells improves when useful heat is continuously accepted at the required temperature. The case against improves when the thermal sink is intermittent, low value, or forces electric dispatch that the site does not need. A yearly thermal total cannot prove hourly acceptance. Nor can a nameplate heat-recovery number establish delivered useful heat after distribution losses and return-temperature constraints.

Load-side thermal assets can change the problem before supply is added. Precooling, chilled-water storage, hot-water storage, insulation, heat recovery, and control changes can move or reduce weather-sensitive electrical load. Their honest advantage is that they may attack the coincident peak directly. Their limitation is operational: comfort, product quality, sanitation, production, recovery, and control constraints can reduce the available flexibility.

Section 07Test every path against the same hours

Technology neutrality becomes concrete when every path sees the same load, weather, outage assumptions, and evidence grades. The comparison below is not a ranking. It is a list of the weather question each path must answer.

Section 08Use measured performance to correct the model

Commissioning does not end weather analysis. It creates the first opportunity to compare the predicted curve with site performance. The owner should trend ambient and enclosure conditions, gross and net output, auxiliary loads, fuel or charging input, state of charge, curtailment, alarms, and equipment availability through different seasons.

DOE’s battery evaluation method is built around long-term charge and discharge time series, with at least one year offered as an example analysis period. It defines efficiency and demonstrated capacity from metered operation and explicitly notes that efficiency can vary with temperature and charge rate.⁷ The larger principle applies across technologies: a rating is the starting point; measured net performance under coincident conditions is the operating truth.

Normalize carefully. A summer decline may reflect hotter air, hotter cooling water, a dirty filter, degraded heat transfer, soiling, a changed control setpoint, or a process shift. Weather adjustment should not become a method for explaining away equipment deterioration. Keep raw measurements, the adjustment method, and the residual difference visible.

Use the result to update dispatch limits, reserves, maintenance timing, and future sizing. A conservative curve that repeatedly understates performance may be refined. A curve that overstates performance on stressful days requires a correction and an operational response. Neither outcome should be hidden inside an annual average.

Section 09The owner’s derate release gate

Before a capacity recommendation is released, the weather basis and the performance boundary should be readable without opening the model. A practical gate includes the following:

1. Decision hourThe essential load and the interval that drive capacity are identified, with the operating event that causes them.
2. Weather envelopeTypical, recent, and stress conditions are separate, dated, sourced, and aligned to the site clock.
3. Net boundaryEvery supply path is measured after its heating, cooling, pumping, compression, inversion, controls, and other auxiliaries.
4. Configuration curveTemperature, elevation, humidity, fuel, state-of-charge, and degradation effects use current evidence for the proposed configuration.
5. CoincidenceLoad, resource, outage, and maintenance assumptions are tested in the same interval rather than in separate annual summaries.
6. Technology parityUtility service, load measures, solar, storage, firm generation, thermal assets, hybrids, and no project face the same scenarios.
7. Evidence labelEvery figure is sourced and dated or explicitly illustrative; no feasibility value is presented as firm.
8. Operational responseThe owner knows what happens if actual net capability falls below the modeled curve, without relying on a guarantee.

A missing curve does not justify a guessed derate. It creates a verification task or a sensitivity range. A missing stress case does not justify calling the average conservative. It means the recommendation has not yet tested the condition it is meant to survive.

Section 10Size the decision before sizing the equipment

The worst-day method is not an instruction to buy for the most extreme imaginable event. Unlimited conservatism can waste capital as easily as average-based sizing can create a shortfall. The owner must define the service level: which loads are essential, how long they must be carried, which interruptions are tolerable, what controls may act, and what contingency is acceptable.

That definition can produce a smaller and more resilient answer. Efficiency may reduce the critical load. Thermal storage may move cooling. Solar may cover part of a daytime event. Batteries may bridge a short deficit. Utility service or firm generation may carry the sustained balance. A hybrid may outperform a single asset, while a well-documented grid case may show that no new on-site supply is justified.

The discipline is consistent across every path: use averages to understand energy, use coincident scenarios to understand capacity, and use measured net performance to update both. Weather is not an adjustment applied after the equipment is chosen. It is part of the evidence that decides what the equipment, contract, controls, and contingency must do.