The horizontal motion of the balloon is governed by the winds at the float altitude. In order to navigate, and change the direction of balloon flight, knowledge of the wind environment around the balloon is needed. The real time navigation and control of zero-pressure balloons is a challenging task as there are no sensors that can be used onboard the balloon to provide real knowledge of the wind environment. Further, their is no active actuation possible and the resources available for passive actuation are limited. These constraints makes the balloon flight difficult and inflexible. In this paper, a solution to this problem of balloon navigation, and its path planning is presented by using data from ECMWF in combination with reinforcement learning. Data from ECMWF gives an overview of almost real-time environment and a reinforcement learning algorithm help in optimizing the passive actuation resources.