Ensemble Learning Based Convex Approximation of Three-Phase Power Flow

Though the convex optimization has been widely used in power systems, it still cannot guarantee to yield a tight (accurate) solution some problems. To mitigate this issue, paper proposes an ensemble learning based approximation for alternating current (AC) flow equations that differs from existing relaxations. The proposed approach is on three-phase quadratic rectangular coordinates. develop data-driven of flows, polynomial regression (PR) first deployed as basic learner fit relationships between independent and dependent variables. Then, algorithms such gradient boosting (GB) bagging are introduced combine learners boost model performance. Based learned flow, optimal (OPF) formulated programming problem. simulation results IEEE standard cases both balanced unbalanced systems show that, context solving OPF, outperforms conventional semi-definite (SDP) relaxation accuracy computational efficiency, especially SDP fails.