Integration of a finite element code in a generator sizing software for hydro applications
Guillaume Burnand – Basile KawkabaniGroup of Electrical [email protected]@epfl.ch
IntroductionThe sizing of a generator is normally performed using a single tool gatheringseveral components: mechanical, electrical, thermal, ventilation, insulation.The electrical part of this sizing software is mainly based on analyticalmethods. These methods are simple and fast and are able to compute theparameters of the machine in normal operating conditions. Nevertheless, thisapproach suffers from inaccuracies when specific electromagnetic problemsare studied or when the machine operates at very extreme conditions. Othercalculation methods are available especially the Finite Element Methods(FEM).The emphasis is on the determination of the load excitation current of asalient-pole synchronous generator. The precise determining of this current isvery important because the rotor Joule losses depend on it and hence theefficiency and the heating of the machine too. This procedure involves aniterative calculation of the excitation current and the stator currents phaseshift, with a 2D-magnetostatic FE model taking into consideration the non-linearity of magnetic materials.This master project will focus on the development of a Finite Element Tool,from an existing Alstom base, for the determination of the load excitationcurrent. This tool will be further incorporate in the Alstom main sizingsoftware.
Theory/Method/HypothesisBecause of the non-linearity of the magnetic materials, the iterative Newton-Raphson method is employed [1]. Every operating point can be defined bythe following 4 parameters :
• I : the stator current [A]• If : the rotor excitation current [A]• f : the electrical grid frequency [Hz]• α : the stator current phase shift [rad]
The combination of those 4 parameters leads to a specific stator voltage U, anactive power P, a reactive power Q and power factor cosφ. Since the statorfrequency f and current I are imposed, the iterative process will be carried outon the rotor excitation current If and stator current phase shift α.For a given set of parameters (If , α), the FE model is solved using amagnetostatic process. The total flux seen by each stator phase Ψa , Ψb and Ψcis then computed. Next, analytical computation such as the Parktransformation allows to compute the load point.
ResultsThe Finite Element Tool uses first-order triangular elements. In order tovalidate its repeatability and reliability, 80 models have been tested andthe results compared with FLUX2D second-order mesh results.
It can be noted that the results given with the FE Tool are moreconservative than FLUX2D second-order, 0.72 % on average. Themaximum error is 2.4 %. This overestimation comes from the first-orderelements where the induction field is constant in an element (see figurebelow). However, the standard deviation, namely the distribution aroundthe average value, is only 0.005 %, reflecting a very good reliability .
Conclusion/PerspectivesIn the future, the FE Tool could be upgraded to second-order elements formore accuracy and to time-harmonic simulations. It could be alsopossible to adapt the model generation to double-fed asynchronousgenerators. Perhaps one day, it could even be suited to 3D FE models.
References[1]T. Lugand, “Two-dimensional finite element electromagnetic
computations: Iterative excitation requirements calculation of a salient-pole synchronous machine using a magneto-static application,” Alstom internal report, 13.10.2015.
Acknowledgments• Dr. Thomas Lugand, Hydro R&D electrical team-leader at Alstom• Prof. Basile Kawkabani, Electrical Machinery Group EPFL
Group Of ElectricalMachines(GEM)
Recommended
