A unified framework for siting and sizing of distributed energy resources in power distribution grids using data-driven machine learning optimization

• Machine learning applied for optimal siting and sizing of DG and ESS in power grids. • Decision tree and random forest algorithms improve planning accuracy. • Voltage profile improved by 11.92 %, power losses reduced by 36.88 %. • Unserved energy reduced by 59.13 %, recovery time shortened by 30.48 %. • Validation conducted on the IEEE 33-bus test system. This paper deals with the unified framework of siting and sizing (UFSS) for real-time simultaneous placement and capacity determination of distributed generation (DG) and energy storage systems (ESS) in the power distribution grid. The presented method uses the decision tree-based optimum bus selection along with Random Forest algorithms for precise capacity estimation to improve computational efficiency, adaptability, and scalability comparing to the conventional heuristic methods. The proposed model is investigated on the IEEE 33-bus test system, where results indicate that it indeed adeptly addresses augmented reliability, operational efficiency, and economic sustainability. More specifically, the framework achieves a reduction of 36.88 % in active power losses, improves voltage profiles by 11.92 %, decreases unserved energy levels by 59.13 %, and saves at least 30.48 % the recovery time for the system, and reduces operational costs by 16.7 %, all leading to a much better, more resilient, and cost-effective power distribution grid. By mastering the intricacies of non-linear complexities and uncertainties, UFSS enables the integration of DG and ESS to optimize power distribution, cost efficiency, and system reliability. Results demonstrate that UFSS is a scalable, intelligent, and adaptive decision-making model that advances the development of autonomous, self-optimizing, and resilient smart grids, significantly enhancing the overall safety and efficiency of modern power grids.