MODELLING THE IMPACT OF TEMPERATURE COEFFICIENTS ON PV SYSTEM PERFORMANCE IN HOT AND HUMID CLIMATES

Abstract

This quantitative study examined how temperature coefficients shaped photovoltaic (PV) system performance in hot and humid climates through 12-month, minute-resolution monitoring and coefficient-aware modeling across three representative installations: a coastal roof-flush site, an urban roof-standoff site, and an inland rack-mounted site, including crystalline-silicon and thin-film technologies. Descriptive findings indicated persistent thermal stress, with plane-of-array irradiance averaging 715–768 W/m², ambient temperatures averaging 30.4–33.2°C, and module temperatures averaging 44.8–49.6°C. Relative humidity remained high (73.6–81.9%) and wind speeds were modest (1.6–2.4 m/s), producing sustained module heating above ambient. Under these conditions, normalized DC power averaged 0.84–0.89 p.u., confirming baseline thermal derating during productive hours. Correlation results showed that module temperature was strongly and negatively associated with voltage and power outputs, including open-circuit voltage (r = −0.71), normalized DC power (r = −0.62), and efficiency (r = −0.63), while associations with short-circuit current were weakly positive or near zero (r = 0.06), supporting voltage-dominated thermal losses. Site-level multivariate regressions controlling for irradiance, wind, humidity, and ambient temperature confirmed an independent temperature penalty on normalized DC power, with coefficients of −0.0043, −0.0037, and −0.0030 p.u./°C (all p < .001) and strong model fits (adjusted R² = 0.84–0.88; RMSE = 0.058–0.069 p.u.). Mixed-effects pooling showed crystalline silicon to be more temperature-sensitive than thin-film, consistent with more negative effective power slopes in silicon arrays. Using field-effective coefficients reduced prediction error compared with datasheet coefficients (RMSE 0.064 vs 0.091) and lowered annual yield estimates by about 4.7%. Overall, the results demonstrated that climate-calibrated temperature coefficients and accurate module-temperature modeling were essential for realistic PV yield prediction in hot and humid environments.