Conjugate gradient back-propagation based artificial neural network for real time power quality assessment

• Novel method is proposed for real time power quality assessment.
• ANN is used to detect power quality disturbances.
• LabVIEW based hardware experimental setup is implemented for verification.
• Highly suitable, robust and adaptable for monitoring power quality issues.

This paper proposes conjugate gradient back-propagation based artificial neural network for real time power quality assessment. The novel real time voltage sag and swell detection, classification scheme using artificial neural network is presented. The ANN is trained in MATLAB using voltage sampled signal data and corresponding parameters of the neural network are utilized to implement in LabVIEW in real time sag swell detection and classification. The Levenberg Marquardt, the resilient back-propagation and conjugate back-propagation algorithm performance is evaluated in MATLAB to find the best neural network for real time application. Among these three back-propagation algorithms, conjugate gradient algorithm has better performance for real time power quality monitoring. The mathematical model of Conjugate gradient back-propagation neural network is implemented in LabVIEW. Real time voltage signals of sag and swell for different time duration and magnitude, intensity are acquired from hardware experimental setup. Hardware setup mainly consists of single phase 230 V voltage source, microcontroller, dimmerstat and solid state relays. Voltage signals of sag and swell are sensed using high precision voltage sensor. Data acquisition system is used to acquire the signal from voltage sensor. The output of data acquisition system is given to the personal computer with LabVIEW. The proposed monitoring system also detects odd and even harmonic components in the voltage signal acquired using FFT. Real time hardware results obtained using proposed power quality monitoring system for detection of voltage sag, swell and harmonics claims the suitability, robustness and adaptability to monitor power quality issues.