Izvorni znanstveni članak
https://doi.org/10.1080/00051144.2023.2203560
A new approach based on current controlled hybrid power compensator for power quality improvement using time series neural network
T. D, Raheni
; Department of Electrical and Electronics Engineering, Kumaraguru College of Technology, Coimbatore, Tamil Nadu, India
*
P. Thirumoorthi
; Department of Electrical and Electronics Engineering, Kumaraguru College of Technology, Coimbatore, Tamil Nadu, India
K. Premalatha
; Department of Electrical and Electronics Engineering, Kumaraguru College of Technology, Coimbatore, Tamil Nadu, India
* Dopisni autor.
Sažetak
In this paper, a current controlled-hybrid power compensator (CC-HPC) is presented to reduce the effect of input current harmonics on battery chargers. Passive filters have significant power loss and degrade system frequency due to excessive harmonic attenuation. The proposed system integrates the Higher Order Sliding Mode Controller (HOSMC) with a generalized form of p–q power theory and a Time Series – Artificial Neural Network (TS-ANN) is used to produce compensating reference current for a three-phase system and generates DC link inductor current. Switching pulses to Current Controlled-Active Power Compensator (CC-APC) switches are generated using a reference compensated signal. The development of CC-HPC and its control approach helps to reduce the overall harmonic distortion of the supply current used in battery chargers are the main contributions of the proposed system. HOSMC is a robust and adaptable controller that tracks reference current without causing chattering is the significant advantage of the proposed method. The control algorithm is designed in MATLAB/SIMULINK software for various load conditions and the experimental setup has been developed for rectified fed RC load using TS-ANN. The filtering process of CC-HPC can maintain the harmonic distortion of supply current within the IEEE 519-2014 standard.
Ključne riječi
Power quality; PI controller; HOSMC; CC- HPC; time series artificial neural network
Hrčak ID:
315928
URI
Datum izdavanja:
24.5.2023.
Posjeta: 443 *