Answers are grounded in this paper and cite their sources.
Outdoor photovoltaic cooling trials without an irradiance sensor often quantify the benefit by regressing module power on module temperature, a fit exposed to omitted-variable bias because irradiance drives both. This paper formalises a paired-differential estimator that contrasts each treated module with a simultaneously sampled untreated reference, and states when the common irradiance term cancels: a shared irradiance history, no module-specific shading, soiling or water film, synchronous sampling and an untreated reference. An analytical bound converts residual irradiance mismatch into bias on the open-circuit voltage coefficient. A proposed plausibility screen tests logged short-circuit current and open-circuit voltage of uninstrumented records against a clear-sky plane-of-array envelope. Both tools were applied to a seven-day campaign in Bandung, Indonesia, with four 30 W modules under no cooling, heat sinks, water spray and both (70 hourly records). The modules were canopy-shaded at every sampled hour and not cleaned, and the electrical values were hand-recorded, some entries not independently verifiable. The record fell outside the clear-sky envelope. Short-circuit current reached 3.39 times the clear-sky expectation, and the open-circuit voltage lacked the expected irradiance dependence (−0.002 ± 0.030 V per unit ln G against 0.97 to 1.25 V expected). The electrical results, including a voltage coefficient of −88.2 mV/°C and power gains below 1% reconstructed from V_oc and I_sc at a constant fill factor, are presented as an illustration of the estimator. Thermocouple temperature differences are indicative: −1.45, −1.08 and −2.12 °C for heat sink, spray and combined treatment. With theoretical pump runtime and nameplate power, the sign of the spray energy balance is not determined.
No takes yet. Share an insight, caveat, or question.
Purnama et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: