A Low-Cost IoT Device for Environmental Monitoring and Embedded Solar Forecasting with On-Device Incremental Learning

Hyperlocal meteorological sensing is essential for accurate solar photovoltaic forecasting, yet professional-grade meteorological stations require investments easily exceeding $1000 USD per node, making distributed deployments economically inaccessible. This work presents a modular internet of things (IoT) device based on the ESP32 microcontroller integrating temperature, humidity, luminosity, and solar panel voltage sensing in an IP68-rated enclosure at a total hardware cost of about $65 USD when components are sourced in Germany. The enclosure-mounted temperature sensor is subject to a daytime radiative-heating bias and is not a calibrated ambient-air measurement. A hybrid architecture decouples external model training, performed on a conventional computer using the software Python and the open-source library TensorFlow, from autonomous on-device inference: every 15 min, the embedded feedforward network produces a single one-step-ahead (15 min) prediction of solar panel voltage from the most recent 96 real sensor readings; the resulting sequence of 96 such predictions, logged and assembled over a full day, forms the diurnal forecast profile reported below. This is executed via a three-layer feedforward network with 3011 parameters (11.8 KB). The network is trained offline on site-collected data and deployed on the microcontroller as static weight matrices without cloud connectivity. An on-device incremental gradient descent mechanism enables model adaptation after deployment without external retraining. The system was evaluated through two field deployments: a short period of hardware and firmware validation in Ulm, Germany, and a 115-day deployment in Zapopan, Mexico, comprising 84 days of training and 31 days of autonomous operation with zero missing records (no 15 min interval failed to log a reading); a real-time-clock fault affecting the final three validation days is addressed separately below and excluded from the reported metrics. Over a clean 28-day daytime window, the embedded model attained a coefficient of determination of 0.9165 and a mean absolute error of 0.2975 V (4.65% of the operational range), outperforming a climatology baseline (skill score 0.64) while not surpassing a 24 h persistence baseline. A frozen-weight ablation confirms that the on-device update mechanism yields a small but statistically significant accuracy gain (p=0.001), showing that autonomous incremental learning is implementable on low-cost hardware and produces a measurable effect, without cloud connectivity.