PV System Loss Factors: Where the Nameplate Watts Actually Go

A 400 W module does not put 400 W into your battery. Even on a clear day at solar noon with the array pointed correctly, the number on the nameplate is a laboratory rating — measured at Standard Test Conditions, a fixed irradiance and cell temperature you will rarely see in the field — and a chain of smaller, real losses sits between that rating and the energy that actually reaches your loads.

PVWatts, NREL’s widely used PV production estimator, bakes ten of these losses into a single default “system losses” percentage so a first-pass estimate does not require modeling every one individually. Knowing what is inside that number, and how the ten factors combine, matters even if you never touch PVWatts yourself, because the same physical losses apply to any PV system regardless of what software estimates them.

A sourcing note before the numbers. NREL’s own PVWatts and technical-report domains were unreachable while this page was researched, so the values below are confirmed through pvlib’s documentation, a widely used open-source solar modeling library, which in turn cites the original NREL PVWatts manual. Treat the individual percentages as secondary-sourced — corroborated but not independently verified against NREL’s own text — while the underlying physical description of each loss is standard PV engineering, not dependent on any one source.

The ten default loss factors

Loss factor	Default	What it physically is
Soiling	2%	Dust, pollen, and grime on the module surface blocking incident light
Shading	3%	Partial shadowing from trees, vent pipes, chimneys, or adjacent rows
Snow	0%	Snow cover blocking the array; defaults to zero because it is highly site- and climate-specific
Mismatch	2%	Small manufacturing variances between modules in a string force the whole string to operate at the current of its weakest module
Wiring (DC)	2%	Resistive loss (I²R) in the DC conductors between modules and the inverter or charge controller
Connections	0.5%	Resistive loss at connectors, splices, and terminals — every junction adds a small amount
Light-induced degradation (LID)	1.5%	A one-time efficiency drop in crystalline silicon cells during their first hours of light exposure, from boron-oxygen defect formation
Nameplate rating	1%	Manufacturers’ actual shipped modules run slightly under their stated rating on average, within their tolerance band
Age	0%	Default assumes a brand-new system; real modules degrade roughly 0.5%/year and this factor should be raised for an aged system
Availability	3%	Downtime from inverter faults, breaker trips, maintenance, or other outages during otherwise-productive hours

Two of these — soiling and shading — describe environmental conditions rather than fixed physical constants, and PVWatts’ 2% and 3% defaults are generic placeholders. A dusty, low-tilt array in an arid climate can soil faster than 2% predicts; a heavily treed lot can shade far more than 3%.

Why the losses multiply, not add

The instinctive way to combine ten percentages is to add them: 2 + 3 + 0 + 2 + 2 + 0.5 + 1.5 + 1 + 0 + 3 = 15%. That arithmetic is wrong, and it is wrong for a specific reason: each loss factor acts on whatever energy is left after the previous losses have already been removed, not on the original nameplate figure.

The correct method multiplies the ten retained fractions together:

L_total = 100 × [1 − (1 − L1/100) × (1 − L2/100) × ... × (1 − L10/100)]

Walking the ten factors through in sequence, starting from a full 1.000000 (100%) and applying each retained fraction in turn:

Step	Factor applied	Running output fraction
Start	—	1.000000
Soiling	× 0.98	0.980000
Shading	× 0.97	0.950600
Snow	× 1.00	0.950600
Mismatch	× 0.98	0.931588
Wiring (DC)	× 0.98	0.912956
Connections	× 0.995	0.908391
LID	× 0.985	0.894766
Nameplate rating	× 0.99	0.885818
Age	× 1.00	0.885818
Availability	× 0.97	0.859243

The final retained fraction is 0.859243, meaning 85.92% of nameplate energy survives all ten factors — a combined loss of 14.08%, which PVWatts’ interface rounds and displays as 14%. Compare that to the naive additive answer of 15%: a full percentage point of overstated loss, purely from using the wrong operation.

A worked example makes the gap concrete. A system with a 6,000 W DC nameplate rating:

Additive (wrong): 6000 × (1 − 0.15) = 5,100 W
Multiplicative (correct): 6000 × 0.859243 = 5,155.5 W

The multiplicative method always yields a smaller total loss than the additive one whenever all the factors are positive, because each subsequent percentage is being taken of a progressively smaller base. The gap grows as the individual loss percentages grow — with ten small factors in the 0–3% range it is only about a point, but the same error compounds badly if someone tries to stack a handful of larger derates (inverter efficiency, temperature loss, a conservative design margin) by simple subtraction instead of multiplication.

Which losses you control, and which you do not

Factor	Builder influence
Soiling	High — cleaning schedule, tilt angle (steeper sheds dust faster), local dust/pollen exposure
Shading	High — site selection and layout, the one factor worth the most design effort to avoid entirely
Snow	Moderate — tilt angle and site choice affect how long snow lingers, but not whether it falls
Wiring (DC)	Moderate — proper conductor sizing per our DC wire ampacity and voltage drop references pushes this well under the 2% default
Connections	Moderate — quality terminations and fewer unnecessary splices reduce this below default
Mismatch	Low — buy modules from the same production batch and bin; otherwise largely fixed by manufacturing tolerance
Nameplate rating	None — this is the manufacturer’s shipped-average tolerance, disclosed but not adjustable after purchase
LID	None — an inherent property of crystalline silicon cell chemistry, front-loaded into the first hours of operation
Age	None (for a given year) — degradation happens regardless of design choices; the rate can vary slightly by cell technology
Availability	Moderate — quality inverter/controller selection and prompt fault response reduce downtime, but cannot eliminate it

The practical takeaway is not to try to force every default toward zero. Soiling, shading, and wiring are worth real design attention because a builder’s choices measurably move them. LID, nameplate tolerance, and cell degradation are physical properties of the hardware — accept them as part of the system’s real output ceiling rather than treating the loss estimate as pessimistic.

Using this in a sizing calculation

When estimating a system’s expected AC output, apply the combined derate factor to the DC nameplate capacity, not to an already-adjusted number, and do it before applying inverter efficiency or peak-sun-hour production separately — those are distinct steps, not part of the ten factors above. See Peak Sun Hours for how to combine a derated array rating with your site’s actual insolation to estimate daily energy production.

If any individual factor on your system is known to differ substantially from the default — a heavily shaded roof, a site with real snow accumulation, an aging system — replace that one factor with a better estimate and re-run the multiplication. Do not average a guess into the total loss percentage; adjust the specific factor it belongs to, so the combined number stays traceable to a cause.