基于混合回归与分类的数据驱动潮流线性化方法以精确满足网络约束

This letter proposes a novel power flow (PF) linearization method for accurately enforcing network constraints in optimal dispatch problems. Unlike conventional linearization methods that focus on reducing PF solution errors, the proposed method aims to enhance the decision feasibility of network-constrained dispatch problems modeled with linear PF equations. A data-driven framework based on hybrid regression and classification is developed to determine coefficients of the linear PF equation. This problem is equivalent to minimizing a weighted sum of the root-mean-square error and hinge loss, which compels the linear PF model to enforce network constraints accurately. Simulations with various system scales verify that the proposed PF linearization method outperforms existing ones in terms of decision feasibility and optimality.