New Model for Weather Stations Integrated to Intelligent Meteorological Forecasts in Brasília

This paper presents a new model for low-cost solar-powered Automatic Weather Stations, based on ESP-32 microcontroller, modern sensors and intelligent forecasts for Brasilia. The proposed system relies on compact, multifunctional sensors and features an open-source firmware project and open-circuit board design. It includes a BME688, AS7331, VEML7700, AS3935 for termo-hygro-barometry (plus air quality), ultraviolet irradiance, luximetry, and fulminology, besides having a rainfall gauge and an anemometer. Powered by photovoltaic panels and batteries, it operates uninterruptedly under variable weather conditions, with data collected being sent via WiFi to a Web API that adapts the MZDN-HF (Meteorological Zone Delimited Neural Network - Hourly Forecaster) model to produce accurate 24 h multivariate forecasts. Installed in Brasilia's University, it demonstrates robust hardware performance and strong correlation with INMET's A001 data, suitable for climate monitoring, precision agriculture, and environmental research.