CUBE: estimate site's yield

From a certain power level upwards, accurately estimating the potential of a site before developing a power plant is essential to ensure its long-term economic viability. In fact, during the project design phase, the developer estimates deliverability based on the analysis of historical solar radiation data over several years.

However, on-site data are rarely available. As a result, the economic model is generally based solely on satellite data with a high resolution (in km2), which does not take into account factors intrinsic to the site. The result is a level of uncertainty that cannot be anticipated by the developer, and which can largely lead to over- or underestimation of the return announced to investors and banks.

P50 is the median energy production value. It is the reference and confidence estimate for investors, with a 50% chance that production will be higher.

The P90 indicates the energy production which has 90% of chances of being below or equal. This is a more conservative estimate, used as a reference value and trusted by bankers to size project debts.

Calibration is carried out by correcting historical satellite data using the on-site measurement campaign. The CUBE service, combining a measurement station and an algorithm, enables calibration of the satellite databases used to establish a typical meteorological year.

Calibration involves comparing satellite observation data with data collected on site. The systematic difference between on-site and satellite observation is qualified, then a statistical learning algorithm processes the data to correct the bias and obtain a yield ratio as close as possible to reality.

The assessment of annual irradiance is highly volatile, with a systematic error, whatever the satellite database, for sites on plains and in temperate climates. In mountainous or humid environments, the estimation bias is greater.

This bias, which cannot be anticipated by the developer prior to project design, compromises the representative annual irradiance estimate (typical irradiance year) and therefore the Yield Report for P50 and P90 production announced to investors and banks. After a 12-month measurement campaign, the systematic bias is eliminated. The calibrated satellite databases now have an uncertainty of 0.5% instead of 3% without calibration. The Yield Report is therefore more reliable and the business model more robust.