The Science of Macro-Scale Yield Forecasting

A solar asset is only as bankable as the forecast behind it. Before we install a single module, we predict — at high confidence — how many kilowatt-hours it will produce in year one, year ten and year twenty-five. That prediction is what separates a hopeful gesture from a financeable asset, and producing it is one of the most demanding disciplines in modern energy engineering.

From photons to probability

Forecasting begins with the resource itself. We draw on long-term satellite irradiance records and ground-station data spanning decades, capturing not just average sunshine but its variance — the cloudy winters, the heatwave summers, the year-to-year swing that determines downside risk. A single sunny year tells you almost nothing; thirty years tells you what to expect.

Onto that resource we layer the physics of the specific site: panel tilt and azimuth, inter-row shading, soiling accumulation, inverter efficiency curves, temperature derating and the slow march of module degradation. Our engine runs thousands of hourly simulations across the asset's life, producing not a single number but a probability distribution.

Why the distribution matters

Lenders do not finance averages; they finance certainty. By expressing yield as a P50 (the expected case) and a P90 (the conservative case exceeded nine years in ten), we give capital partners the risk-adjusted figures they need. The tighter we can draw that band, the cheaper the financing and the stronger the project economics.

Forecasting that never stops

Modelling does not end at commissioning. Once energised, every asset streams live generation data back to our operations centre, where actual output is benchmarked continuously against forecast. Divergence flags soiling, faults or shading before they erode returns — and every data point sharpens the next project's model. The forecast is a living instrument, not a one-time document.

This is what harnessing raw stellar power at scale actually requires: not optimism, but evidence — measured, modelled and proven across decades.