Machine Learning-Based Approach to Optimizing the Performance of a Solar Laptop Charger

Abstract

The increasing reliance on portable electronic devices and the growing demand for sustainable energy solutions have underscored the need for efficient solar-powered charging systems. This study presents a machine learning-based approach to optimizing the performance of a solar laptop charger by employing a Long Short-Term Memory (LSTM) neural network. The proposed system aims to predict optimal charging voltages in real-time, adapting to fluctuations in solar irradiance and ambient temperature to enhance energy conversion efficiency. Historical solar data comprising environmental and electrical parameters was collected, pre-processed, and used to train and test the LSTM model. Simulation results demonstrated that the model accurately forecasted charging voltages, achieving a high coefficient of determination (R² = 0.976), with low prediction error rates (MAE = 0.227, RMSE = 0.285). Though the system was not physically implemented, the simulation results confirm the effectiveness of the LSTM model as an intelligent alternative to conventional Maximum Power Point Tracking (MPPT) techniques. This study highlights the potential of machine learning in enhancing the performance and adaptability of solar energy systems, offering a scalable and software-driven solution for sustainable energy applications.