Neural network control of induction machines using genetic algorithm training

Direct torque control (DTC) is the simplest torque control of induction machines. The key component of DTC is the state selector. In this paper, a neural network (NN) is used to emulate the state selector of the conventional DTC. Training the neural network is achieved using a genetic algorithm (GA). Binary and floating-point GA data representations are used. GA operators used are one- and two-point crossovers, bit mutation for binary encoding and nonuniform mutation, arithmetical crossover and nonuniform arithmetic mutation in floating point encoding. This has greatly improved the fine local tuning capabilities of a genetic algorithm. Simulations have been performed using the trained state selector NN instead of the conventional DTC. The results show agreement with those of the conventional DTC. It is, also, found that using floating-point encoding algorithm gave better results than the binary encoding.