Solar Energy Forecasting Using LSTM and IoT-Based ESP32 Monitoring System

The growing demand for reliable renewable energy highlights the need for accurate solar power forecasting, particularly for small-scale installations that operate under highly variable environmental conditions. This work presents an integrated IoT-based solar monitoring and prediction system that combines real-time sensor data from an ESP32 microcontroller with a data-driven forecasting model trained on publicly available weather and energy-generation datasets. After preprocessing and normalizing the Kaggle dataset, a Long Short-Term Memory (LSTM) network is developed to learn temporal dependencies between irradiance, temperature, humidity, and historical power output. The trained model predicts short-term solar energy generation for the next day, enabling improved scheduling and energy-usage planning. The ESP32 simultaneously measures panel voltage and current, providing live data that supports system validation and future on-device inference. Experimental results demonstrate that the proposed approach achieves stable performance with low prediction error and strong correlation between actual and predicted values. The system offers a low-cost, scalable, and practical solution for households, academic projects, and small solar farms seeking to optimise energy utilisation through intelligent forecasting