A common problem arising in production planning is the fact that the designed plan is often disrupted by several uncontrollable factors. As a result, companies are often unable to meet the promised lead time of goods. This issue often appears when main plan is constructed due to the certainty assumption in parameters of problem. So, it is essential to take into account uncertainty in production planning. One of the important approaches in this issue is constructing robust production plan. . So, this schedule remains valid. It is obvious if we want to use this method in a factory, it is better to be applicable to real conditions. For this reason, by considering simultaneously uncertainty in demands, some methods have been offered in construction of robust production planning. In this research, the focus is on the medium-term production planning and especially on single-level lot sizing decisions. Uncertainty is one of the critical problems that most of problems like production planning problem encounter. This research deals with the single level capacitated lot sizing problem (CLSP), with variations in demands. The demand variations construct a set of scenarios, and one of our objectives is to find the least conflicting solution among scenarios along with minimizing costs. To this purpose in this dissertation, two robustness criteria for CLSP, is defined; also lot sizing problem with different sizes have been solved. Since the problem is NP-hard, we have used GA to solve the problem. Results show that criteria has less cost that the other criteria. But this criteria needs more solution time than criteria.