In this paper, we have tried to reduce the induction motor losses by controlling the magnetic currents in different torque loads. Reinforcement learning is a method where an agent considers the environment state chooses one action among all possible actions, and the environment returns a numerical signal as a reward for that action. | Journal of Automation and Control Engineering Vol. 4, No. 1, February 2016 Decreasing Induction Motor Loss Using Reinforcement Learning Mohammad Bagher Naghibi Sistani Department of Electrical Engineering, Ferdowsi University of Mashhad, Mashhad, Iran Email: mb-naghibi@ Sadegh Hesari Bojnourd Branch, Islamic Azad University, Young Researcher and Elite Club, Bojnourd, Iran Email: assafsyria@ Abstract—In this paper, we have tried to reduce the induction motor losses by controlling the magnetic currents in different torque loads. Reinforcement learning is a method where an agent considers the environment state chooses one action among all possible actions, and the environment returns a numerical signal as a reward for that action. The agent aims at finding a policy by trial-anderror method to reach the maximum sum of rewards. The main proposed idea of this paper is implementing QLearning algorithm to find the optimal action in every state of the environment. In this method, quantized amounts of electromagnetic Torque and motor speed are considered as states, and magnetic current is considered as action. Simulation results shows that this method can reduce the power loss about 50% in comparison with the standard driver of motor (FOC) when the motor works in low loads. Index Terms—reinforcement learning, algorithm, induction motor, decreasing loss I. what to do, but it tries the possible actions and discovers the action with maximum reward [6]. One of the reinforcement learning methods with a simple implementation is Q-Learning method, which was presented in 1989 by Watkins [7]. This algorithm is used by the agent to learn through experience or training. Every repetition equals a training course. The aim of training is to create the brain of the agent, which is displayed by Q matrix. More training will lead to a better Q matrix that can be used by the agent to move in the optimal direction. This way, by having a Q matrix, the agent can choose .
