With the increasing penetration of DERs and flexible resources in distribution systems, power flow patterns have become more complex, making quantitative power contribution analysis increasingly important for system operation and planning. Conventiona...
With the increasing penetration of DERs and flexible resources in distribution systems, power flow patterns have become more complex, making quantitative power contribution analysis increasingly important for system operation and planning. Conventional power flow and power tracing methods provide high physical accuracy but require repetitive calculations for various scenarios, resulting in significant computational burden. To address this limitation, this study proposes a data-driven approach using a machine learning model trained on datasets generated through power tracing under various DER output conditions and connection configurations. In addition, a system impact-based screening technique is applied to reduce the training dataset size. To validate the proposed method, scenarios with a single DER connected to multiple buses were considered. The screening-based model was compared with a full-scale data-based model and conventional power tracing results. The results show that the proposed model maintains comparable prediction accuracy while significantly reducing computational time. Therefore, the proposed approach combines physics-based analysis and data-driven modeling, enabling accurate and efficient power contribution estimation for distribution system operation and flexible resource management.