Optimal Distributed Generation Placement and Real-Time Load Forecasting Using Machine Learning for Voltage Stability Enhancement

Modern power systems demand robust voltage stability and accurate real-time load forecasting, especially with the rising integration of Distributed Generation (DG) sources such as wind and solar energy. This paper presents a machine learning-based framework for optimal placement and sizing of DG units, designed to minimize the Voltage Stability Index (VSI) and improve grid resilience. The study evaluates three DG capacity scenarios (100%, 75% and 125%) on the IEEE 14-bus test system to examine their impact on voltage profiles. To enable real-time grid operations, a Bagged Regression model is employed, trained in both historical and simulated data for short-term load forecasting. Comparative analysis against traditional regression methods and advanced algorithms demonstrates superior accuracy, computational efficiency and reliability of the proposed framework. The main contributions include the integration of DG placement with real-time forecasting, the application of Bagged Regression for adaptive load prediction, performance benchmarking against boosting methods and extensive validation using the IEEE 14-bus system. The results confirm that the framework effectively enhances voltage profiles, reduces VSI and delivers scalable, adaptive and efficient solutions aligned with the operational needs of modern smart grids.