A comparative investigation on different randomness schemes in the particle-swarm-based repetitive controller for the sine-wave inverter

Abstract

In this paper different randomness scenarios for the recently developed direct particle swarm controller for repetitive processes are investigated and compared. The proposed controller employs the particle swarm optimizer (PSO) to solve in on-line mode the dynamic optimization problem (DOP) designed to shape the control signal in the constant-amplitude constant-frequency (CACF) voltage source inverter (VSI) with an LC output filter. The controller is of a stochastic nature. The DOP at hand is of αD dimensionality, where α denotes the number of control signal samples per each period of the reference voltage signal. Originally, the PSO requires pseudorandom number generators (PRNG) to be run throughout the iterative search to get a new set of numbers in each sample time. In DOP scenarios the swarm has to be kept alive during the operation of the inverter. This in turn implies that the pseudorandom numbers are to be generated in real-time using digital signal controller (DSC) resources. Four different randomization schemes have been tested: the dimension-and-particle-wise one, the dimension-wise one, the particle-wise one and an almost-deterministic one (also known as a listbased PSO). The last approach does not employ any PRNG in real time and as such establishes an appealing alternative in terms of its low computational burden. The effectiveness of such a scheme, when applied to the direct swarm controller, has been studied numerically in the paper.