Development of Machine Learning based Hybrid Power Generation System for Rural Electrification

Abstract

Introduction: Electrifying rural areas in developing countries presents numerous challenges, including the limited reach of conventional power grids, high upfront infrastructure costs, and insufficient maintenance capabilities. Additionally, regulatory restrictions and logistical obstacles, such as difficulties in transportation and security concerns, further complicate efforts to expand electricity access. Overcoming these barriers necessitates the implementation of a dependable and economically viable solution that can provide a consistent and sustainable power supply, ensuring long-term benefits for rural communities.

Objectives: This study is dedicated to the development of a Hybrid Power Generation (HPG) system that combines multiple renewable energy sources to improve both reliability and efficiency. The research emphasizes the use of a machine learning-based approach to optimize power generation by implementing a real-time Maximum Power Point Tracking (MPPT) technique. This ensures optimal energy conversion, even under dynamic environmental conditions such as fluctuations in sunlight and wind speed, ultimately enhancing the system’s overall performance and sustainability.

Methods: The proposed Hybrid Power Generation (HPG) system integrates solar and wind energy sources using a dual-input DC-DC buck-boost converter, which allows both sources to function independently while maintaining simultaneous operation. This integration enhances the system’s reliability and ensures a stable power supply. To optimize the Maximum Power Point Tracking (MPPT) process, a Linear Regression-based Machine Learning Algorithm is employed, which dynamically adjusts the duty cycle of the converter through a microcontroller-driven data storage unit. This algorithm effectively predicts the optimal duty cycle, even in the presence of Partial Shading Conditions (PSC) and rapidly changing atmospheric factors. The system’s efficiency and adaptability are assessed under three different shading scenarios: zero shading, weak shading, and strong shading, demonstrating its capability to maintain consistent performance across varying environmental conditions.

Results: The developed Hybrid Power Generation (HPG) system exhibits exceptional efficiency across different shading scenarios, with recorded efficiencies of 99.91% in zero shading, 99.38% in weak shading, and 99.78% in strong shading conditions. Furthermore, the system demonstrates rapid response times, achieving convergence within 0.4 seconds under zero shading, 0.8 seconds under weak shading, and 1.4 seconds under strong shading. A comparative assessment against traditional Maximum Power Point Tracking (MPPT) algorithms underscores the enhanced performance of the Linear Regression-based Machine Learning Algorithm. By effectively optimizing power extraction, the proposed system ensures stable and efficient energy generation, making it a reliable solution for sustainable power supply.

 Conclusions: The findings of this study validate that the proposed Hybrid Power Generation (HPG) system, enhanced through a Linear Regression-based Machine Learning Algorithm, provides a highly efficient and dependable solution for rural electrification. By maintaining consistently high efficiency across diverse environmental conditions, the system proves to be a practical and sustainable alternative for power generation in developing regions. Its ability to adapt to variations in solar and wind energy availability ensures reliable energy access, making it a promising approach to addressing the challenges of rural electrification.