USE OF NEURAL NETWORKS IN ADAPTIVE ELECTRIC CAR CONTROL

Abstract

To overcome the lack of information about the parameters of the driving cycle of the electric car, neural networks are used, which provide adaptive control that allows you to adapt. electric car to external operating conditions, as well as to compensate for inaccuracies in mathematical models. Use of iterative optimization of parameters allows to adjust optimum work of power plant of the electric car (PEC) in the course of its movement. This method allows you to use a single approach to study different processes, regardless of the parametric features of electric vehicles. To accelerate adaptation, the neurocontroller and neural network model are trained using a reference control model, which is either an optimal strategy or a strategy based on logical rules of choice, obtained by methodical programming for a given driving cycle. Based on the results of the research, an adaptation algorithm is proposed. The expressions given in the article allow to carry out adaptation of the power plant on the basis of hybrid to the current driving cycle on the basis of the concept of training of the neuro-fuzzy controller with reinforcement. The expressions given in the article allow to carry out adaptation of the power plant on the basis of hybrid to the current driving cycle on the basis of the concept of training of the neuro-fuzzy controller with reinforcement. The purpose of training the neuro-fuzzy controller is the formation of such control effects of the power plant, which would reduce the quadratic value of the assessment of the quality of management.