Using a neural network to obtain an output voltage control method in an interconnection inverter

Abstract

Generated power savings result when the harmonic current in the output of an inverter that connects solar cells and fuel cells to ac lines is suppressed. This suppression is done by canceling the harmonic voltage in the ac line by generating the same harmonic voltage in the inverter output. A method to control the output voltage of an interconnection inverter is proposed that uses a neural network that has learned a sample of the ac line voltage, including its harmonic components. Simulation and experimental results show that this method sufficiently suppresses the total harmonic distortion of the inverter output current. It also is shown that this method is adaptable to various types of distortion.There are two control methods for inverters. In one method, the inverter output voltage coincides with a reference voltage. In the other method, the inverter output current coincides with a reference current. In an interconnection inverter, the latter method is usually used. However, to supply power from the inverter when power failure in the ac line occurs, the current-control method must be changed to the voltagecontrol method. However, it is difficult both to change from an inverter output current control method to an inverter output voltage control method and to modify the control gain. On the other hand, with the voltage-control method, the reference voltage simply needs to be switched to an internal oscillator in the case of a power failure. Therefore, we use the voltage control method. In the voltage control method, canceling the harmonic voltage in the ac line suppresses the inverter harmonic current by generating the same harmonic voltage in the inverter output. Although there have been no reports on how to implement this algorithm, there are two simple methods. In the first, a harmonic voltage in the ac line is obtained by using a bandpass filter; in the second, a harmoNc voltage is obtained by using the fast Fourier transform (FFT). However, both methods involve a complicated control structure. network; harmonic distortion.