An Improved Decomposition Framework for Accelerating LSF and BD Based Methods for Network-Constrained UC Problems

This paper presents an improved decomposition framework to accelerate the convergence of linear sensitivity factor (LSF) and Benders decomposition (BD) based methods for network-constrained unit commitment (NCUC) problems. Classical LSF and BD methods solve NCUC problems by an iterative procedure between a master UC problem and transmission network evaluation subproblems. Subproblems evaluate the transmission network security of a master UC solution, and feedback violated network constraints or Benders feasibility cuts to the UC problem for seeking a feasible UC solution that could mitigate transmission violations. However, classical methods often converge very slowly and introduce major computational bottlenecks. This paper presents an improved decomposition framework by embedding network evaluation subproblems into the branch-and-bound (BAB)/branch-and-cut (BAC) procedure of the master UC problem. Thus, instead of iteratively solving the UC master problem and network evaluation subproblems, the NCUC problem is solved in one single integrated procedure by performing network evaluations at each BAB/BAC node. The proposed methods reduce the number of BAB/BAC nodes to be solved in the UC problem at the cost of additional network evaluations. Numerical tests demonstrate the efficiency of the proposed decomposition framework. The benefit of the proposed decomposition framework would be significant when a system is heavily congested and a large number of iterations is expected for the classical methods. The proposed decomposition strategy can be extended to other applications for solving large-scale optimization problems in power systems operation, maintenance, and planning.