A Real-Time IoT-Based System for Solar Energy Forecasting and Automated Anomaly Notifications

Energy efficiency is crucial in modern society, particularly in residential settings where photovoltaic (PV) systems enable users to generate electricity and reduce costs. However, without accurate solar energy forecasting, PV systems often operate below optimal efficiency, impacting energy management in smart grids and residential systems. In recent years, IoT and Machine Learning techniques have significantly improved energy monitoring by providing tools to track production. Despite the demonstrated efficacy of existing solutions, there is a lack of real-time systems that combine dashboard visualization with automated notifications for notifying overproduction, underproduction, or inadequate power generation due to weather conditions. This paper presents an IoT-based platform for solar energy production forecasting that integrates a) a data acquisition module to acquire real-time meteorological and production data, b) a Machine Learning module to forecast production, c) a dashboard to show the collected information, and d) a notification system to notify production anomalies (overproduction, underproduction, or potential malfunctions). Forecasting is performed using an ensemble of Support Vector Regression (SVR), Random Forest, Ridge Regression, and Kernel Ridge Regression, improving robustness against data variability and demonstrating the enhancements that ensemble Learning can bring compared to individual models. The proposed solution is validated on a real-world photovoltaic system, contributing to the overall enhancement of the research in smart energy management, and offering a scalable IoT framework for optimizing renewable energy production.