To improve the model accuracy for using in whole
condition, the cross validation has been implemented by
using the independent data set from six conditions. The three
cases of voltage variation and three cases of current variation
has been experimented and modeled. The DC input and AC
output power waveform of six data sets are cascaded for
represent the whole steady state operating condition of
inverter as shown in Fig 7 and Fig.8 . Then the system
identification process has been executed. It is found that the
accuracy of model and experimental of the FVLA condition
because of the operating condition of FVLA is operate with
power lower than 10% of rated which contain the high
influence of nonlinear behavior. With this reason, the simple
and no complex of deadzone estimator contain in the model
in not adequate for this range. Nevertheless the overall
accuracy of the model is satisfied. The voltage model of the
inverter has accuracy 97.53% and the current waveform has
accuracy 88.67% and the power waveform model has
81.89% as shown in Figure 9.