An Internal Model Controller for Three-Phase APF Based on LS-Extreme Learning Machine

ABSTRACT

Aiming at the problem that the three-phase APF’s dynamic model is a multi-variable, nonlinear and strong coupling system, an internal model controller for three-phase APF based on LS-Extreme Learning Machine is studied in this paper. As a novel single hidden layer feed-forward neural networks, extreme learning machine (ELM) has several advantages: simple net structural, fast learning speed, good generalization performance and so on. In order to improve the controller’s dynamic responses, a least squares extreme learning machine for internal model control is proposed. A least squares ELM regression (LS-ELMR) model for the three-phase APFS on-line monitoring was built from external factors with in-out datum. Moreover, the relative stable error is presented to evaluate the system performance and the features for the internal model control system based on extreme learning machine, neural network, kernel ridge regress and support vector machine. The experimental results show that the LS-internal model control system based on extreme learning machine has good dynamic performance and strong filtering result.

Keywords:

Extreme learning machine (ELM), internal model control (IMC), Shunt active power filter (SAPF), Least squares (LS), Robustness.