This paper proposes a new formulation to speed up the security constraint unit commitment (SCUC) solution in large scale power systems. The number of variables and prevailing constraints grow drastically as the number of system buses and the credible contingencies increase. So, as the first stage, inessential variables and constraints are eliminated and necessary variables for a tighter formulation are defined. Some constraints and variables are removed and some constraints are modified based on the results of a small sub-problem at each iteration of Benders decomposition algorithm. By utilizing the proposed framework, especially in the case of large number of severe contingencies, a large number of variables and constraints will be eliminated. Up- and down-going reserves are considered at both generation- and demand-sides. The IEEE RTS1 and RTS2 systems are studied as illustrative examples. The SCUC results drawn using the proposed algorithm are compared with those obtained by a conventional MIP-based algorithm to show the effectiveness of the proposed algorithm.