IoT rig tracks perovskite solar windows through a real winter

Two semi-transparent perovskite solar window modules have spent a full autumn and winter on the roof of a building-envelope prototype in Peshawar, Pakistan, streaming their electrical vital signs to a low-cost Internet of Things logger every day. The field campaign, described in Results in Engineering, ran from October 2025 to January 2026 in the city’s semi-arid climate and captured something laboratory solar simulators cannot: the simultaneous evolution of two differently built perovskite modules under the same real weather sequence, from the strong sun of October to the compressed, humid days of December.

The study, led by Shahzad Aman and colleagues at USPCAS-E, UET Peshawar, together with It Ee Lee, addressed a persistent gap in perovskite photovoltaics. Buildings account for a substantial share of energy use and associated emissions, and building-integrated photovoltaics, or BIPV, fold generation directly into façades, roofs and windows. Metal-halide perovskites are attractive for this role because their optoelectronic properties can be tuned and semi-transparent architectures can be designed for glazing. Yet reviews of the field have repeatedly flagged the same weakness: published outdoor stability datasets for perovskites remain limited in duration and geographical coverage, and semi-transparent variants carry additional caveats around device architecture, optical transparency and operational stability.

The Peshawar experiment was designed around that gap. The team fabricated carbon-based perovskite modules starting from laser-scribed fluorine-doped tin oxide glass, depositing compact and mesoporous titanium dioxide electron transport layers, a mesoporous zirconium dioxide spacer, silver contacts and a carbon black–graphite electrode that doubles as cathode and hole transport material. The perovskite precursor was drop-cast into the porous mesoscopic stack and annealed to crystallize the absorber. Each module was then sealed in a double-glass window configuration with rubber spacers and an aluminum frame, the kind of package intended for architectural glazing that must generate electricity while still transmitting daylight.

Two configurations were installed side by side in the same prototype. Module A was a segmented, interconnected sub-cell design in a 3S-4P arrangement with a nominal active area of 300 square centimeters, wired to a 200-ohm resistive load and a green LED indicator. Module B was a strip-cell, continuous-area design described as monolithic, with a nominal active area of 500 square centimeters, running into a 1-kilohm load and a red LED. Independent INA219 sensors recorded each module’s loaded terminal voltage and current, while an SHT31 sensor tracked ambient temperature and relative humidity, a TSL2561 sensor monitored incident light, a DS3231 real-time clock stamped every reading, and an SD card stored the data locally.

The authors are explicit that these are load-dependent field indicators, not standardized module ratings. There was no solar simulator, no reference pyranometer validation, no standardized current–voltage sweeps and no maximum-power-point tracking. The TSL2561 measures illuminance in lux rather than irradiance in watts per square meter, so its output was converted dimensionally and treated as an irradiance-equivalent estimate rather than a calibrated measurement; high outdoor illuminance can also saturate the sensor. Temperature was recorded at ambient level, not at the module surface. Those constraints shape every number that follows.

The seasonal signal was unmistakable. The mean irradiance-equivalent fell from 531.2 watts per square meter in October to 425.4 in November and 335.3 in December, a 36.9 percent reduction in the available solar resource, before recovering 7.1 percent to 359.2 in January. Absolute output tracked that curve. Module A’s mean load power dropped from 686.7 milliwatts in October to 404.5 in November and 232.4 in December, then rose 5.0 percent to 244.1 in January. Module B fell from 1916.7 to 1447.4 and 1042.8 milliwatts over the same months, then climbed 5.9 percent to 1103.9. Current followed suit: Module A’s mean load current fell roughly 42.5 percent from October to December, while Module B’s fell in two steps of 20.95 and 23.13 percent before a 4.65 percent January rebound that still left it 36.41 percent below its October level.

The more revealing story sits in the normalized numbers. Because efficiency was calculated pointwise from measured terminal power against the sensor-derived irradiance input, it partially cancels the seasonal dimming. Module A’s mean load-based operating efficiency fell from 4.29 percent in October to 3.17, 2.31 and finally 2.26 percent in January, a 47.2 percent reduction relative to its October baseline. Module B declined from 7.21 percent to 6.80, 6.22 and 6.15 percent, a 14.8 percent reduction. The combined, area-weighted efficiency of the 800-square-centimeter prototype slipped from 6.12 percent in October to 4.69 percent in January. Crucially, January’s partial recovery in solar resource did not restore Module A’s normalized performance, which the authors read as consistent with persistent module-level change rather than mere seasonal fluctuation.

The physical evidence pointed the same way. By the end of the campaign, Module A had developed strong discoloration, its dark active layer shifting to green and yellow regions with non-uniform surface morphology and localized patch formation, observations the authors suggest are consistent with moisture- or oxygen-induced interfacial instability, though the images alone cannot distinguish among ingress through imperfect edge sealing, absorber decomposition or contact instability, and no post-mortem chemical analysis was performed. Module B, by contrast, showed no significant visible structural discoloration and retained a comparatively uniform surface. The authors attribute Module B’s greater stability to its monolithic structure and encapsulation integrity, while cautioning that active area, architecture and load resistance all varied together, so the contrast cannot be assigned to architecture alone.

Energy accounting reinforced the divergence. Combined monthly energy yield, obtained by summing recorded daily values, fell from 812,883 milliwatt-hours in October to 511,289 in November, 353,288 in December and 307,394 in January. Module B’s share of the total rose from 73.69 percent in October to 81.89 percent in January, while Module A’s fell from 26.31 to 18.11 percent, accumulating 445,640 milliwatt-hours against Module B’s 1,539,215 over the campaign. Because the modules were never electrically interconnected, the system total is an analytical sum of two autonomous outputs, not a measured bus-level yield.

What the dataset ultimately delivers is a methodological template rather than a certified degradation rate. The synchronized, time-stamped logging of electrical and environmental parameters from two electrically independent subsystems gives researchers a way to separate seasonal solar-resource effects from persistent device-level change, and the authors confine their conclusions to the measured field behavior of these specific configurations. They call for co-located calibrated pyranometry, dedicated module-temperature sensing, documented sensor calibration, uncertainty characterization and standardized current–voltage or maximum-power-point measurements in future work. For a technology whose commercial promise rests on surviving years of rain, heat and ultraviolet exposure behind a window pane, that kind of honest, instrumented field record is exactly where the evidence has to start.

Subject of Research: Outdoor IoT monitoring of semi-transparent perovskite building-integrated photovoltaic modules under real climatic conditions

Article Title: IoT-enabled real-time outdoor monitoring of perovskite BIPV system under real climatic conditions

Article References: Aman, S., Khan, A. D., Sadaqat, F., Khan, A. D., Wali, Q., & Lee, I. E. (2026). IoT-enabled real-time outdoor monitoring of perovskite BIPV system under real climatic conditions. Results in Engineering, 32, Article 113915. https://doi.org/10.1016/j.rineng.2026.113915

Image Credits: AI Generated

DOI: 10.1016/j.rineng.2026.113915

Keywords: perovskite solar cells, building-integrated photovoltaics, BIPV, IoT monitoring, outdoor stability, semi-transparent solar windows, Peshawar field campaign, load-based efficiency, seasonal irradiance, module degradation, encapsulation, energy yield