The four case studies below document real analyses of utility-scale photovoltaic plants. What they have in common: in all four, conventional monitoring systems showed nothing unusual. The performance ratio sat in a range considered “normal" in every case. Only the comparison against a plant-specific, physically modelled target value made the real losses visible.
Case 1 · Flevoland (NL) — 55 MWp
Problem: High DC/AC ratio of 1.40 (40% oversizing). Above 800 W/m² irradiance the inverters ran systematically at their power limit – substantial clipping. Finding: 1,379,605 kWh, or 7.69% below target production – depending on the electricity price, €110,000 (€0.08/kWh) to €300,000 (€0.22/kWh) of lost yield per year. Recommendation: a battery storage system to absorb the clipped energy delivered the highest ROI; alternatively additional inverter capacity.
Case 2 · Friesland (NL) — 103 MWp
Problem: A systematic MPP-tracking issue across several inverters under rapidly changing irradiance, plus mismatch losses from a sub-optimal string configuration. Finding: 9.2% deviation from target – which conventional monitoring waved through as “normal variation"; extrapolated to over €800,000 of loss per year. Recommendation: firmware update plus string reconfiguration, cost under €50,000, expected additional yield over €600,000/year – an ROI above 1,200%.
Case 3 · Jutland (DK) — 74 MWp
Problem: Three overlapping causes – 50% inverter oversizing, strings at the upper voltage limit in cold conditions, and underestimated row shading from the site topography. Finding: 11.4% total deviation from target, around €420,000 of lost yield per year. Recommendation: a staged approach – adjust inverter settings short-term, reconfigure strings medium-term, evaluate tracker retrofits or row spacing long-term.
Case 4 · Aquitaine (FR) — 23 MWp
Problem: The most striking example of the limits of conventional monitoring – irradiance and production curves running in parallel, no error messages. Only the analysis revealed that above 1,050 W/m² peak irradiance, production fell well short of target, caused by thermal losses (insufficient rear ventilation) and a degradation pattern pointing to PID. Finding: €175,000 of loss per year on just 23 MWp – the highest relative loss rate of the four plants; €3.5M of foregone revenue over 20 years. Without the target-value analysis the loss would have stayed hidden as “normal variation" or a “conservative forecast".
At a glance
| Site | Capacity | Main issue | Dev. from target | Loss / year |
|---|---|---|---|---|
| Flevoland (NL) | 55 MWp | Clipping (DC/AC 1.40) | 7.69% | ~€110,000 |
| Friesland (NL) | 103 MWp | MPP tracking + string mismatch | 9.2% | ~€800,000 |
| Jutland (DK) | 74 MWp | Oversizing + voltage + shading | 11.4% | ~€420,000 |
| Aquitaine (FR) | 23 MWp | Thermal losses + PID | highest rel. rate | ~€175,000 |
| Total | 255 MWp | — | — | ~€1.5M / year |
The shared insight
All four cases share one denominator: conventional monitoring would not have caught the losses. The problems were technically different – clipping, MPP tracking, configuration errors, thermal losses – but economically just as severe. Only the comparison against a dynamic, plant-specific target turns “running quietly" into a reliable “running optimally".
See what no monitoring shows.