In this paper, a new robust downlink power control solution based on worst-case performance optimization is developed. Our approach depends on explicit modeling of uncertainties in the downlink channel correlation (DCC) matrices, worst-case performance optimization and guarantees that the quality of service (QoS) constraints are satisfied for all users using minimum amount of power. The key in the iteration is the step to solve an originally non-convex problem to obtain worst-case uncertainty matrices. When the uncertainty is small enough to guarantee that the DCC matrices are positive semidefinite, we obtain a closed-form solution of this problem. When the uncertainty is large, we transform this intractable problem into a convex problem. Simulation results show that our proposed robust downlink power control using the approach of worst-case performance optimization reduces the transmission power effectively under imperfect knowledge of the channel condition.