<![CDATA[Small-scale photovoltaic (PV) installations generally run without adequate performance monitoring, so a drop in output is noticed only after it has persisted for some time. Most Internet of Things (IoT) monitoring systems forward every sensor sample to a server untouched, and this pattern loads the network while raising the module current draw. This work moves part of the computation onto the device itself (edge computing) using an ESP32-S3 microcontroller. Panel current and voltage are read through an INA226 module, temperature through a BME280, and irradiance through a calibrated LDR. The ESP32-S3 smooths the samples with a moving average and screens out outliers before publishing the result to an MQTT broker for display on a web dashboard. Testing ran for 14 days on a single 100 Wp module at a 5-second sampling interval, with the scheme without edge processing serving as the baseline. Transmitted data fell from 184.6 KB/hour to 46.3 KB/hour, mean latency from 312 ms to 87 ms, and module current draw from 142 mA to 128 mA. Power measurement accuracy rose from 96.2% to 97.8%, because local filtering suppresses sensor reading noise. The architecture is suitable for PV installations at sites where bandwidth and power supply are limited.]]>
Copyrights © 2026
