This paper proposes an improved lagrangian relaxation approach for the unit commitment problem. This method has improved the following three aspects :f irst,a heuristic sorting method is used to get better initial lagrangian multipliers; second,the total cost of solving the single-unit problem by dynamic programming is reduced to achieve a better commitment start stop decision; at last,the optimization strategy of adaptive subgradient combined with aggregative projection subgradient is applied to accelerate the convergence.Simulation examples with the number of generating units in the range of 10 to 100 show that the two strategies used alternately can alleviate the oscillation and speed up convergence.