A feed-forward control approach using Neural Network observers for single-phase inverter

Abstract

In order to improve the power quality of output voltage, a feed-forward control approach based on neural network observers for a single-phase inverter is proposed. This approach is to build a Hopfield Neural Network parallel observer based on the principle of energy minimization and to extract harmonic current from single-phase inverter output through optimized computation with Hopfield Neural Network. By adding the voltage drop in filter into control system's inner loop, an instant harmonic estimate and compensate feedforward control system can be built. First, in the outer loop of control system, AC instantaneous fundamental voltage value should be converted into the amplitude and the phase equivalently. Then using PI controller, amplitude control loop and phase control loop are built respectively. Thus an amplitude-phase multi-loop control strategy is realized. Hence, not only inverter output voltage's amplitude error and phase error will be eliminated, but also voltage drop generated by harmonic current in the filter inductor will be eliminated. Therefore the power quality of single-phase inverter is improved. Finally, simulation and experimental results are shown to verify the proposed approach.