Sampling averaging for inverter control

Deadbeat controllers provide a fast response approach to control of inverters. This has been extended to the control of high order filters for grid connected inverters. These controllers use high gains on the errors measured at one sample to control the errors to zero ideally by the next sample. One difficulty with this controller is that the high gains can make the control very sensitive to any errors or noise in the measurements. This paper examines the use of reset integrators in the sampling to reduce the noise present in the samples. These effectively average the measurements over the sampling intervals. The derivations in the paper show how the controller gains can be adjusted for the effective delay in the measurement. The corrections are applied to a 240V 50 Hz inverter connected to the grid via a third order filter. The T-filter provides a better tradeoff between reactive power injection and switch frequency ripple than a standard inductor filter. The experimental results show a dramatic improvement in tracking of all system states even in the presence of high measurement noise.