Predictive Control usually use estimator to model the mismatch between the real and nominal value. Why do these parameter change in the filter of three phase voltage source converter change?
Variations of component values with respect to their nominal values are related to fabrication tollerances. Such variations are estimated by an estimator and then corrected so that to avoid the calibration of the controller for each single manufactured unit. Moreover, long term changes and drift of the same parameters are also corrected using this approach.
You are right but my question why parameters change and why do we only consider that all three phase inductance are equal but this is an ideal case. In real, inductance may vary.
You mean: why parameter change with respect to their nominal value or why do they undergo long-term drifts?
The answer to the first question is rather obvious, and is related to the fact that in a production environment only certain degree of accuracy can be achieved because of variability of material characteristics, mechanical tollerances in manufacturing, etc.. a 10%-20% spread is typical for power components.
Long-term drifts, on the other hand, are mainly related to component ageing, i.e. the irreversible changes of material properties (e.g. magnetic materials in a core, resistivity, characteristics of the electrolyte in electrolytic capacitors...) due to thermal cycles, varying environmental conditions (e.g. humidity, pressure), presence of chemicals, wears in moving parts and so on. The effect of all this can be neglected in the short term but can lead to meaningful changes after months or years and such changes need to be accounted for by the controller.