机组组合问题是电力系统优化运行的一个难点，理论上难以得到最优解。提出了一种基于粒子群修正策略的解耦算法。首先采用集结投影次梯度的拉格朗日松弛算法得到机组组合的对偶解；然后依据对偶信息中的备用乘子及对偶组合状态建立粒子群优化空间；而后利用无约束的标准粒子群优化算法实现拉格朗日乘子的局部更新，通过粒子的调整和粒子间信息的传递改变机组启停，进而修正拉格朗日对偶解，最终得到机组组合问题的近似最优解。6个系统的仿真计算验证了该方法的求解速度及计算精度。

Due to the difficulty of achieving optimal solution theoretically, unit commitment (UC) is a hard task in optimal operation of power system. For this reason, in this paper a decoupling algorithm based on the modification of particle swarm optimization (PSO) is proposed. Firstly, by use of Lagrangian relaxation algorithm based on aggregative projection sub-gradient the dual solutions of UC is obtained; then according to the reserve multiplier and dual combination in dual information, the optimal space of particle swam is built; and then by use of unrestricted standard PSO algorithm the local updating of Lagrangian multiplier is implemented, by means of adjusting particles and the information transmission among particles the UC is changed, after that the Lagrangian dual solution is modified and finally the approximate optimal solution of UC is obtained. The solving speed and calculation accuracy of the proposed method are verified by simulation results of six test systems.