Production smoothing process has a significant role in reducing the costs and efficient meeting of the customers’ demand for a wide variety of products. In this study, the production smoothing has been considered in general lot sizing and scheduling problem (GLSP) in presence of non-zero setup and processing times and setup costs which also vary among the products in single machine system. The developed model in this research employs following dual objectives: first minimizing the production costs including sequence dependent setup costs and holding inventory costs and second minimizing the sum squared variation of the ideal production rate. Since now, the smoothing process has not been utilized in literature of the GLSP. The two mathematical models have been presented based on Clark and also Fleichman-Meyer models for this problem. We employed the sum weighted method and ?-constraint method for solving the mathematical model in a small scales. Two value parameter sets were used for genetic algorithm that the first set was similar to the Arroyo and Armentano research and second set was selected among the best results of executing genetic algorithm with different parameter values. Computational results show that in the small scales, the second parameter set has a more effective solution cases than the first one. In the small scales, the best obtained results are the combination of the results obtained from ?-constraint method and the sum weighted method. The performance of this comminuted result has been compared with the performance of genetic algorithm method. Computational results show that the Pareto optimal set which obtained by genetic local search algorithm is more effective than the solutions which obtained by solving the mathematical model with a shorter computational time.It is not logical to use this model for solving a large scales production smoothing in GLSP. As a result, the efficient solutions were obtained only by executing genetic local search algorithm. Genetic algorithm is executing by two parameter sets; the Pareto frontiers that were constructed at the end of 10, 30, 50, 80 and 100 iterations were compared together. Finite Pareto frontier is constructed by combining the results of first and second parameter sets in last iteration. Keywords: S cheduling, Lot sizing, Production Smoothing, GLSP, Genetic Local Search.