This paper applies reinforcement learning to entropy generation minimization of energy systems, with examples of heat exchangers and wind turbines. Reinforcement learning is a branch of machine learning where an agent learns to make sequences of decisions by interacting with an environment to achieve a specific goal. This study uses Deep Q Learning as a Reinforcement Learning technique for optimization of two sample energy systems. For a heat exchanger problem, the selected optimization variable is the aspect ratio for a minimum entropy generation. For a wind turbine problem, it is the tip speed ratio for maximizing the power coefficient. The algorithm presents a dynamic exploration rate parameter (ε) to optimize the learning process in terms of exploration and exploitation, by tailoring discrete actions and rewards. The results of the training process followed expected behavior of two objective functions related to both applications. For the heat exchanger case, the range of optima for the aspect ratio zone with minimum entropy generation was between 1,800 and 3,500. For the wind turbine problem, the current study considered multivariable optimization cases for design and real-time operational cases. For the one variable case, the maximum energy generation zone used an aspect ratio of 7.5 to 9.5. For the multivariable case, the maximum power coefficient was achieved with a combination of a tip speed ratio of 8.12 and a pitch angle of 0. The findings of this study provide useful indications that reinforcement learning can be effectively applied to the optimization of multivariable energy system problems. The study presents and discusses results for efficiency improvement in both heat exchangers and wind turbine power generation.