基于自适应线性潮流模型的交流网络约束机组组合

Network-constrained unit commitment (UC) problem typically uses a linear approximation of AC constraints to guarantee solution efficiency. However, the key difficulty current linearization methods face is that the change of UC status forms wide operating conditions and AC feasibility recovery. Meanwhile, the accuracy of linear approximation greatly depends on the closeness between the operating condition and initial points. To solve this problem, this paper proposes a network-constrained UC solution based on adaptive linear power flow model. The proposed method splits the operating conditions into several regions considering unit status, load profile, and topology. In the UC model, an auxiliary binary variable is introduced to represent the adaptive selection of region and corresponding best-fit linear power flow model. In addition, to reduce the computational burden, this paper presents a model-scale reduction method via identifying the critical branches that have a major influence on UC accuracy. By only using adaptive model for critical branches, the improvement of UC solution is achieved with an acceptable computational burden. The performance of proposed method is verified in several standard test systems and a provincial power grid in China. It shows that the proposed method can reduce the linearization errors of branch active/reactive power flow and bus voltage magnitude/phase by up to 48.01%, 45.83%, 40.58%, and 51.15% compared with existing UC methods, respectively. The AC-feasible solution of the proposed method achieves a reduction in constraint violations and a saving in operation costs.