Universal Optimal Model Predictive Control of Multi-Phase AC Converters

One step ahead optimal Model Predictive Control (MPC) has been developed and demonstrated by simulation for current control in various - frequently used power converters, namely, 2 level, 3 level, matrix (MxC) and Vienna converters. The universal m * n power converter with (m) inputs and (n) outputs was modeled by switching matrix whose elements define on/off states of power devices. To allow modeling of unidirectional switching devices (diodes) where conduction (and thus switching state) depend on current polarity, additional current polarity matrix was introduced. New switching matrix for converters with both types of devices was defined as dot product (products of individual elements) of original switching matrix and current polarity matrix. The matrix elements - switching states that drive power section were selected by minimization of cost function by search algorithm. The multi - objective cost function, depends on converter topology and control objectives, in general have terms for quality of current tracking, minimization of switching frequency, balancing of dc bus voltage, reactive power minimization, power structure losses minimization etc. The different formulations of cost functions that enable penalization of individual switches for loss minimization, current tracking and DC link voltage balancing in multi-level inverters and operation of Vienna converter were analyzed and compared.