To increase an electric vehicle’s mileage, the energy efficiency of vehicles grows in importance every day. To this end, this study proposes an automotive air conditioner control system based on DQN(Deep Q-Learning), a reinforcement learning algorithm. This control system aims to reduce energy consumption with equivalent temperature error by controlling the compressor RPM and blower output. Also, the control system must improve its performance if there is a change in vehicle velocity profile. For this purpose, the state signals are designed to include thermal information and driving velocity, and the reward function is designed to reduce temperature error and power consumption and to suppress the undesirable behavior of the compressor. Training results show an improvement in energy efficiency when simulating with trained driving velocity profiles and other driving velocity profiles compared to the target A/C system.