Performance investigation of a permanent magnet generator

Abstract

Computer simulation using the finite-element method (FEM) is an important tool for the design of highly efficient power devices. In this work combination of FEM-software for magnetic analysis and Simulink-software for non-linear parameter identification for dynamics of a permanent magnet (PM) generator is discussed. An FE model of the generator is developed and its electromagnetic torque analysis is carried out using FEM-software. Simulink has been used for analysis of rotor moment of inertia (MI) due to variation of winding resistance and winding inductance. It is shown that system MI has a significant effect on optimal winding resistance and inductance to achieve steady state operation in shortest period of time.

Introduction

PM generators have the advantages of eliminating the exciter field winding, slip rings and brushes. The ability of the permanent generator to self excite is an attractive feature that makes it a suitable choice for operation at higher power factors and efficiencies. In addition, PM machines do have the overloading capability and full torque capability at zero and at very low speeds. The understanding of the characteristics and accurate modeling of the dynamic performance of these are of fundamental importance to design engineers. Finite-element analysis has been used extensively for the design and performance prediction of various types of permanent magnet machines [1]. Usually it complements analytical techniques which can provide a rapid and reliable means of design optimization by using adaptation procedures. Equally, simulation for power systems is also widely used to understand the behavior of the machine within the system [2]. Combining these two activities is a desirable goal – the dynamic behavior of the machine affects the system and vice versa. Until recently, the electromagnetic analysis has usually been confined to static representations of the machine geometry using, for example, frequency response methods in synchronous machines [3] or slip frequency analysis in induction motors [4]. This inevitably leads to some inaccuracy in the characterization of the machine. A complete simulation with independent dynamic electromagnetic and power system analysis has been achieved [5], but generally this is too expensive computationally for everyday design use. This paper outlines improvements to the characterization of machine that can be obtained by using dynamic non-linear electromagnetic time-stepping analysis including rotation [6]. Generation of adaptive mesh based upon nodal errors is given by [7] in which next adaptive stages are performed after updating the spacing values using previous adaptive solution. Adaptive analysis with rotation has been carried out by [8] but to take into account the rotor movement, the elements of air gap and rotor have to be modified at each rotor step. A novel algorithm of adaptive mesh generation for the non-linear finite-element analysis of electric machines has also been presented in [9]. A two-dimensional adaptive meshing technique for accurate calculation of very low cogging torque of rotating machines has been presented by [10]. But the study of effect of winding parameters on the torque is missing in all these. Simulations of PM wind turbine generators have been tested for stability by [11] but the analysis of winding parameters on the steady state performance of the generator has not been reported.

Present work has been carried out in two parts. Firstly, FE model of the generator is used to study its mechanical dynamics and subsequent starting torque curve by using FEM-software COMSOL Multiphysics. A finite-element mesh of the model is constructed using first order Lagrange quadratic elements. Finally, the Simulink Parameter Identification software is used for optimization of winding resistance and winding inductance.