A Static Neural Network for Input-Output Mapping of Power Electronic Circuits

Abstract

This paper investigates the effectiveness of a static neural network for input-output mapping of power electronic circuits. The neural network is a multilayer perceptron (MLP) that is trained to form a mapping between the inputs and outputs of a power electronic circuit. The circuit consists of a full bridge diode rectifier, together with the source inductance and the output filter. Dynamic models have been used for the rectifier diodes. The ultimate objective of the designed neural network is to provide an indication when the performance properties of one or more components in the rectifier circuit have changed. Simulation results are provided that indicate the neural network is capable of mapping the inputs and outputs of the circuit and detect operating conditions that are different from the original condition.