We investigate various methods of the Benders decoposition algorithm in its application to the simple plant location problem. We developed six variants. The master problem may be relaxed as an LP problem up to an appropriate point in time, or need not be solved to the optimality before a cut is added. Furthermore, since the subproblem is highly degenerated, we can generate more than one cuts at a time. The efficiency of the methods are examined using a sample problem. The result showed that the adding two-cut method was superior to the standard method. The LP relaxation and the non-optimization of the master program greatly improved the efficiency. Applying the LP relaxation method, we were able to reduce the computing time by two thirds of the time required by the standard method.