Optimization of the power transformer tank.

Parpala, Radu Constantin ; Stancu, Octavian ; Popescu, Diana 等

1. INTRODUCTION

Power transformers are designed and built according to high engineering standards to provide many years of outstanding performance and reliability (*** IEC, 2005). Electrical devices, insulation and cooling systems are subjected to advanced technologies which require safe and durable design solutions for the mechanical part, too. Experience has shown that a key technical issues can be analyzed using 3D modeling and FEM analysis (Prevost & Woodcock, 2007), but few references are found regarding the behavior of the transformer tank. The vast majority of the installed power transformer tanks were not designed using FEM.

The paper deals with optimization attempts and design improvements of the tank for a 40 MVA power transformer. Starting from an initial design an improved solution was calculated and validated. Reinforcements were changed. The tank weight decreased and the total displacements on the side walls were reduced. A new solution was found as the basis for the new conceptual design of the tank. Remarks regarding parameterization and optimization procedures were included.

2. MODEL PREPARATION

The transformer tank is a welded structure designed for strength, durability, and compact form. It accommodates the core-and-coil assembly and the oil filling, usually weighing several tones. This calls for a statically secure and oil-leak-proof design, with an optimized weight.

[FIGURE 1 OMITTED]

Tank structures of power transformers are exposed to wide variations in pressure during testing and subsequent start-up periods. The design of the wall reinforcements is a tough job as it must take into account the mounting elements, tune the static behavior with the total weight and prevent magnetic flux leakage.

An initial design (T1) of a 40 MVA power transformer tank, comprising vertical reinforcements and a horizontal band, was compared with a modified variant (T2) with only vertical ribs. The CAD models of the two tanks were built using CATIA V5, as shown in fig. 1. The assemblies comprise 212 and 310 parts respectively.

When completing the CAD models clashes and clearances were solved. New constraints were added to assure contact on surfaces, not on edges, tiny features were eliminated and all the gaps were closed. Some parts were rebuilt in order to assure a correct closure of the bottom part of the assemblies. Other defeaturing options were activated when the CAD models were imported to the solver. The volume of the two tanks was 8.027 [m.sup.3] for T1 and 7.922 for T2 corresponding to a total mass of 6302.5 kg and 6.218.7 respectively.

Model preparation for FEM analysis of large welded structures is difficult to complete because the solver requires a clean geometry and topology. Penetration and gap checks were performed with an accuracy of [10.sup.-16] mm in the contact regions (ANSYS, 2007). Because solid elements were used, the element size was chosen taking into account the wall thickness and contact control options were activated. Figure 2 shows T2 meshed model, as well as details on side walls, while Table 1 contains mesh characteristics for both tanks. The wave front decreased for the improved variant, which enabled shorter computing time. For each structure two static FEM analyses were performed corresponding to the required physical tests regarding power transformer tanks. The load cases were: vacuum and 0.5 bar internal pressure. Although the load cases presume a linear analysis, multiple contact regions due to the bolted connections and welded parts influenced not only the mesh strategy, but also the behaviour of the two tanks (Gomez, 2002).

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

[FIGURE 4 OMITTED]

3. FEM RESULTS

At the same load and close geometry characteristics, the maximum values of the total displacements are relevant regarding the static behaviour and tank stiffness (Heckman, 1998). Figures 3 and 4 show isometric views of the two tanks and the displacement maps for the second load case, which was 0.5 bar internal pressure. When processing the results the scale factor was kept similar, in order to show the improved behaviour of the new design. The total displacement decreased with 50%, from 8.6229 mm (T1) to 4.4746 mm for the T2 variant. Stress distribution was also improved because stress concentrators were eliminated and safety factors were significantly increased.

The calculations were the basis for changing the tank reinforcements for future projects. Optimization procedures, as well as sensitivity analysis were performed on reduced models of the side walls. The number of the vertical ribs, as well as the rib thickness and the rib height are now obtained using multicriteria optimization procedures, based on parameter definition using recent functionalities of the solver. Many "what is scenarios" were completed.

4. EXPERIMENTS

Power transformer tanks are special cases of pressure vessels design. Before assembling the core-and-coil and oil filling, the required experiments are performed on the tank (Declercq, & Lakhiani, 2006).

When analyzing the first design variant information from experiments was already available. The average error of the calculated displacements for the pressure test was less than 10%, which is pretty good for a large and complex structure. The measurement devices for the pressure test of the improved tank T2 were placed exactly in the areas where local and global maximum were revealed by the FEM analysis.

For a better description of the tank behavior the measurement points were increased to 10, and the devices were placed both on the vertical ribs and on the walls. Figure 5 shows a comparison between FEM results and measured displacements (MES).

All the tests, both vacuum and 0.5 bar pressure confirmed that the design and the simulation were correct. The accuracy of the computed displacement was improved as well, and there were measured points were the error decreased under 0.5% (see points 7 and 9 Fig. 5). This proved a better model preparation strategy.

[FIGURE 5 OMITTED]

5. CONCLUSION

Worldwide the actual fleet of power transformers is close to the lifetime. New technologies are in progress requiring design improvements for the mechanical part too. FEM simulation is a powerful tool for the transformer design. It offers rapid verification procedures and points out useful practical information obtained priory the final tests. A new solution was found as the basis for the new conceptual design of the tank.

Design optimization is an innovation tool, minimizing time and costs and providing innovative design solutions. Parameterization of the reinforcement elements allowed sensitivity analysis and shape optimization procedures. New attempts are in progress for combining topology and shape optimization procedures and to customize the CAD-FEM environment.

Bringing simulation and analysis into design is the basis for a reliable, cost-efficient and durable product. The paper demonstrates the efficiency of a simulation driven design.

6. REFERENCES

ANSYS (2007). ANSYS Structural Analysis Guide, ANSYS INC., Huston 2007

Declercq, J. & Lakhiani, V.K. (2006). Building reliability into design of transformers for reliability of electrical systems, Available from: www.cigre.cl/sem_inter_cigre_nov_2005/ presentaciones/crompton.pdf, Accessed:2009-01-18

Gomez, E. (2002). Welded Joint Analysis for Pressure Vessels, Available from: http://www.ansys.com/events/proceedings/ 2002/PAPERS/98.pdf. Accessed: 2009-22-02

Heckman, D. (1998). Finite Element Analysis of Pressure Vessels, MBARI

Prevost, T.A. & Woodcock, D.J. (2007). Transformer Fleet Health and Risk Assessement. IEE PES Transformers Committee Tutorial. Available from: http://www. transformerscommittee.org/meetings/f2004_LasVegas/Min utes/F04-Main.pdf, Accessed: 2009-03-12

*** IEC 60076-7. (2005). Power transformers Part 7: Loading guide for oil-immersed power transformers

PARPALA, R[adu] C[onstantin]; STANCU, O[ctavian]; POPESCU, D[iana] & PUPAZA, C[ristina] *

* Supervisor, Mentor

Tab. 1. Mesh characteristics

Design variant T1 T2

Nodes 201346 414044
Equations 600525 1231551
Wave front 831 492