PV Array Energy Yield

NREL's PVWatts model reduces a year of solar production to one line: $E = P_{dc} H \, PR$. Take the array's DC rating, multiply by the peak sun hours the site receives over the period, and multiply by a performance ratio that carries every loss between the glass and the meter. It is a crude-looking equation that predicts annual yield remarkably well, because the two hard parts — the resource and the losses — have both been measured to death.

The performance ratio is where all the engineering hides, and quoting one without knowing its parts is how forecasts go wrong. A typical breakdown: 8–15% for cell temperature above 25 °C, 2–3% for soiling, 2% for module mismatch and wiring, 2–4% for inverter conversion, 1–3% for shading and snow, and a further 0.5% a year of permanent degradation thereafter. A well-built fixed-tilt system in a temperate climate lands at 0.78–0.85. A hot, dusty or partly shaded site can fall below 0.70. Anything quoted above 0.90 has left something out.

Measured against real meter readings, the performance ratio becomes the most useful diagnostic a PV system produces, precisely because it divides out the weather. Tracking it month on month exposes soiling, a failed string, or an inverter fault that raw kilowatt-hours would hide behind a cloudy fortnight. A ratio calculated above 1 almost always means peak sun hours have been confused with daylight hours.

The equation is deliberately blind to when the energy arrives, and that limitation decides which systems it suits. For a grid-tied array the timing scarcely matters, since the grid absorbs the surplus and returns it later. For an off-grid system with batteries the annual total is nearly irrelevant: the December figure governs the design, and in southern Canada that can be a quarter of the June figure. Sizing an off-grid array on annual average sun hours produces a system that works beautifully all summer and fails in the third week of December.