A comparison of discrete-time models for model predictive control of induction motor drives

Abstract

System modeling and variables estimation are two important task in direct control strategies. In those schemes, the system performance is strongly related with the accuracy of the discrete-time model, especially for time-varying systems, e.g., high-performance motor drive applications. This investigation presents a numerical comparison between several discrete-time models of the induction machine (IM) used in model predictive control (MPC) schemes. The accuracy of each model is evaluated numerically for different speed operation points. The results serve as a preliminary validation to select an accurate discretization method for MPC drive applications.

Non‐linear MPC for winding loss optimised torque control of anisotropic PMSM

“…The equations for the matrices describing this affine PMSM model are the basis for the discretisation step. As discussed later, simulations showed that discretisation with Euler approximation is not sufficient, which is in accordance with [11, 12]. The system was discretised with a third‐order Picard Iteration [13] to provide remedy to that problem.…”

Novel Lexicographic MPC for Loss Optimized Torque Control of Nonlinear PMSM

“…2) Discretization: In accordance with [31] and [32] simulations show that a discretization with Euler Approximation is not reasonable for higher angular velocities ω el . The system was discretized with a third order Picard Iteration [33].…”

Adaptive finite‐control‐set model predictive current control for IPMSM drives with inductance variation

“…where the superscript 'P' denotes the predicted quantities and T s is the sampling interval. Although the zero-order-hold discretisation approximated by matrix factorisation [5,24] or high-order Taylor series has a better precision than the Euler method [25], we adopt the Euler method due to its merit of simplicity and sufficient precision at high sampling rate. At the kth instant, i d(k) , i q(k) and ω (k) are the sampled values, R, L d , L q , l f are the system parameters.…”

“…To predict the future states of the system, forward Euler discretisation (2) is applied to (1) to obtain the discrete predictive model (3) around the k th instant as followswhere the superscript ‘P’ denotes the predicted quantities and T s is the sampling interval. Although the zero‐order‐hold discretisation approximated by matrix factorisation [5, 24] or high‐order Taylor series has a better precision than the Euler method [25], we adopt the Euler method due to its merit of simplicity and sufficient precision at high sampling rate.…”