Sensorless Estimation Based on Neural Networks Trained with the Dynamic Response Points.

Omar Rodríguez-Abreo,Crescencio Garcia Guendulain,Francisco Antonio Castillo Velásquez,Jonny Paul Zavala De Paz,José Luis Martínez Godoy

doi:10.3390/s21206719

Omar Rodríguez-Abreo, Crescencio Garcia Guendulain + Show 3 more

Open Access

https://doi.org/10.3390/s21206719

Copy DOI

Abstract

In the present work, a neuronal dynamic response prediction system is shown to estimate the response of multiple systems remotely without sensors. For this, a set of Neural Networks and the response to the step of a stable system is used. Six basic characteristics of the dynamic response were extracted and used to calculate a Transfer Function equivalent to the dynamic model. A database with 1,500,000 data points was created to train the network system with the basic characteristics of the dynamic response and the Transfer Function that causes it. The contribution of this work lies in the use of Neural Network systems to estimate the behavior of any stable system, which has multiple advantages compared to typical linear regression techniques since, although the training process is offline, the estimation can perform in real time. The results show an average 2% MSE error for the set of networks. In addition, the system was tested with physical systems to observe the performance with practical examples, achieving a precise estimation of the output with an error of less than 1% for simulated systems and high performance in real signals with the typical noise associated due to the acquisition system.

Highlights

Accepted: 31 August 2021The estimation of parameters is a widely studied problem; there are multiple works such as [1,2,3,4,5,6] in which the estimation of the parameters of the Photovoltaic Models is developed as the main object of study
It is due to the high interest in having a function that describes the dynamic behavior of the systems since its performance can be estimated as in [7], where the authors propose a sensor-less prediction system. Another option to use parameter estimation is precise control design since it is generally done theoretically based on system dynamics, as in [8,9,10,11,12] where the authors base the control design on the analysis of the dynamic model and the parameters precision gives the controller precision
A recurrent option in the control and analysis of the dynamic system is the use of the Transfer Function

Summary

Introduction

The estimation of parameters is a widely studied problem; there are multiple works such as [1,2,3,4,5,6] in which the estimation of the parameters of the Photovoltaic Models is developed as the main object of study. The investigations [19,20,21,22] use a heuristic method to estimate and control the systems based on the dynamic response Another example for the analysis with a heuristic method is the study [23] where the authors do the Transfer Function parameter estimation with the Vector Fitting technique. In the research shown in [30], the authors develop a method for parameters that vary in time Another widely studied option is the use of Neural Networks. The proposed research is not based on estimating a specific system Instead, it aims to develop a standard parametric estimation system for any openloop stable systems using the second-order standard Transfer Function.

Second-Order Transfer Function Characteristics and Its Dynamic Response

Neural Network as Parameter Estimator

Key points

ANNs Evaluating Performance with Simulated Systems

Experimental Results

Conclusions