Active power filter with enhanced echo state network in radar power supply system

Abstract

Abstract There have been concerns raised on load non‐linearity as more and more load types are being introduced into the radar power supply system. The non‐linear load produces harmonic current while it seriously threatens the normal and safe operation environment of other system equipment. It can be difficult to design shunt active power filter (APF) to eliminate the harmonic effect while achieving stability and reliability. This paper designs a hybrid optimization shunt active power filter, which adopts enhanced echo state network (ESN) to predict harmonics and reduce the delay problem in traditional control. The prediction performance of enhanced ESN is verified by simulation. The modelling and simulation of the hybrid optimization shunt APF are completed in Simulink, and the parameters of APF are reset according to the radar power supply system. Furthermore, a prototype based on DSP TMS320F28335 was developed to test whether the theory is feasible and accurate. The experimental results demonstrate that the APF with harmonic prediction has improved steady‐state response performance and rapid dynamic response performance.

Keywords

• active filters
• harmonic analysis
• neural nets