Feature Selection Using Hybrid Machine Learning Approach for Wind Power Generation

The accurate estimation of wind power generation is crucial for advancing renewable energy systems. Efficient feature selection is critical for optimizing wind power generation systems. This study focuses on identifying the most impactful features influencing wind power generation in the Shajapur region of Madhya Pradesh, India, using advanced machine learning approaches. The methodologies employed include Decision Tree (DT), Extra Tree (ET) and a proposed hybrid model that combines the strengths of both algorithms. These approaches are specifically designed to select the most relevant features, rather than directly estimating wind power output. The Decision Tree algorithm provides interpretability by ranking features based on their contribution, while the Extra Tree algorithm enhances the process by incorporating randomized splitting, which reduces overfitting and improves robustness. The hybrid model integrates these techniques to create a more effective framework for feature selection. Key meteorological and environmental parameters, such as wind speed, air pressure, temperature, wind direction, solar irradiance and relative humidity are analyzed to identify the factors most critical to achieving maximum wind power generation. The comparative analysis evaluates the performance of DT, ET and the proposed hybrid model based on their ability to prioritize the right features for wind power optimization. Outcomes demonstrate that the hybrid model conquers individual algorithms, offering a more reliable and accurate feature selection process. This study provides a robust framework for feature selection in wind power systems, paving the way for enhanced efficiency in renewable energy projects. The findings hold significant potential for guiding stakeholders in the planning and implementation of wind energy systems in Shajapur.