Security margin assessment for stochastic unit commitment based on parallelized holomorphic embedding method

As the penetration of renewable energy increases, the operation of power systems faces significant uncertainties. Traditional deterministic unit commitment relies on static thresholds for safety margin assessment, which cannot dynamically reflect system risks. Therefore, it is urgent to develop efficient stochastic optimization and security evaluation methods. This study proposes a security margin model for stochastic unit commitment (SUC) based on a parallelized holomorphic embedding method (HEM). The approach innovatively integrates Latin hypercube sampling (LHS) for generating stochastic scenarios and constructs a CPU-GPU hybrid architecture. Using this parallel computational framework, the computational efficiency of high-dimensional probabilistic power flow (PPF) is enhanced. Consequently, dynamic probabilistic assessment of the security margin is achieved. Verification tests on a standard bus system and an actual large-scale power grid demonstrate that the proposed parallel HEM-GPU framework improves computational speed by 37.4 % and the convergence success rate by 99.7 % compared with conventional methods. Notably, the method can accurately capture the probability distribution of voltage limit violations. After incorporating frequency security constraints, the performance advantage exceeds 40 %, and the computation time is reduced by 76.4 %. The model achieves a critical fault identification rate of 99.5 % and supports large-scale scenario simulation while controlling the risk identification error within 2 %. Thus, this framework provides an efficient and robust security assessment tool for power grids with a high penetration of renewable energy.