Modelling of a DC-DC Buck Converter Using Long-Short-Term-Memory (LSTM)

TL;DR

This paper presents a neural network-based black-box modeling approach for a DC-DC Buck converter, demonstrating high accuracy in simulating its static and dynamic behavior through simulation and experimental validation.

Contribution

It introduces a novel LSTM-based neural network technique for accurately modeling Buck converters, validated with both simulated and real data.

Findings

High accuracy in predicting converter behavior

Effective simulation of static and dynamic responses

Validated with experimental data

Abstract

Artificial neural networks make it possible to identify black-box models. Based on a recurrent nonlinear autoregressive exogenous neural network, this research provides a technique for simulating the static and dynamic behavior of a DC-DC power converter. This approach employs an algorithm for training a neural network using the inputs and outputs (currents and voltages) of a Buck converter. The technique is validated using simulated data of a realistic Simulink-programmed nonsynchronous Buck converter model and experimental findings. The correctness of the technique is determined by comparing the predicted outputs of the neural network to the actual outputs of the system, thereby confirming the suggested strategy. Simulation findings demonstrate the practicability and precision of the proposed black-box method.