An online data-driven model-free predictive control method with low computational burden for power converters

Abstract

Finite-set model predictive control is widely used in power converter control due to its simple structure and superior performance, but its high sensitivity to system parameters restricts its popularization. To solve this problem, many model-free control methods are emerged. However, complex control algorithms are often accompanied by high computational burden, which affects their online application on low-power processors. Motivated by this, this paper proposes a novel online data-driven model-free predictive control method with low computational burden. The prediction module adopts the data-driven form of autoregressive exogenous model, which eliminates the dependence on physical parameters by information mining through system input and output data. In order to reduce the computational burden, this paper proposes a vector optimization method based on linear programming, which avoids the traversal optimization in the conventional finite-set model predictive control method and thus improves the computational efficiency. Meanwhile, a model-free midpoint voltage balancing solution is proposed, which utilizes the small vector redundancy feature to achieve accurate balancing. Simulation and experimental results show that the proposed method achieves excellent performance and computational efficiency.