Neural Network Observer of the Inductive Current Average Value of the Buck DC Converter

This article proposes an algorithm for constructing a system of artificial neural network observers of the average value of the inductance current of a Buck DC converter. The system can be used as an alternative to Hall effect current sensors in order to improve the weight and dimensions of the device. The structure of the neural network observer's system is proposed, an algorithm for collecting and preparing data for training is described. The versatility of the algorithm makes it possible to obtain a neural network current observer for the entire family of DC-DC converters described by the same number of independent variables. The neural network observer of the average value of the inductance current was tested on a sample that differs from the training one for different control laws. The accuracy of the system in both cases is 5%. Recommendations on training and use of neural network observers are given.