Extended Kalman Filter Method based on Backpropagation Neural Network in Current Sensor Online Calibration

The signal collected by current sensor will contain various noises, which will have a negative impact on the its accuracy and calibration. In order to remove unwanted noises, an extended Kalman filter (EKF) method based on backpropagation (BP) neural network is proposed in this paper. BP neural netw...

Full description

Saved in:
Bibliographic Details
Published inIOP conference series. Materials Science and Engineering Vol. 631; no. 5; pp. 52038 - 52044
Main Authors Zhao, Shuang, Zhang, Hanzhu, Liu, Jun, Li, Yansong
Format Journal Article
LanguageEnglish
Published Bristol IOP Publishing 01.10.2019
Subjects
Accuracy
Back propagation
Back propagation networks
Calibration
Extended Kalman filter
Neural networks
Nonlinear systems
Sensors
Signal processing
System effectiveness
Online AccessGet full text

Cover

More Information
Summary:The signal collected by current sensor will contain various noises, which will have a negative impact on the its accuracy and calibration. In order to remove unwanted noises, an extended Kalman filter (EKF) method based on backpropagation (BP) neural network is proposed in this paper. BP neural network has good adaptive ability and non-linear mapping ability. EKF can effectively filter noise and improve the calibration accuracy for non-linear systems. The signal collected by current sensor is processed by EKF and is used as the input signal of BP neural network. The trained neural network can modify the output signal of EKF, so as to improve the calibration accuracy. The angle difference and ratio difference of current sensor calibration are below 0.1, which meets the national standard and shows the effectiveness of this method.
ISSN:1757-8981
1757-899X
DOI:10.1088/1757-899X/631/5/052038