March 26-29
Ecodesign of an Auxiliary Transformer for the Railway Traction Jean-François Convers Now with SNCF, Technicentre Industriel d'Oullins, Pôle Ingénierie, 25 ter, Quai Pierre Sémard, BP 85 La Mulatière, 69224 Oullins cedex, France E-mail:
[email protected] Tuan-Vu Tran, Stéphane Brisset L2EP, Ecole Centrale de Lille, Cité Scientifique, BP 48, 59651 Villeneuve d'Ascq cedex, France E-mail:
[email protected],
[email protected] Véronique Andries ALSTOM TRANSPORT, rue Jacquard, Parc d'Activités Lavoisier, 59494 Petite-Forêt, France E-mail:
[email protected] Marc Bekemans ALSTOM BELGIUM TRANSPORT, rue Cambier-Dupret 50-52, B-6001Charleroi, Belgique E-mail:
[email protected] Copyright © 2009 MC2D & MITI
Abstract: This paper proposes one approach for the ecodesign of an auxiliary transformer for the railway traction based on optimization methods. This approach is well suited to electrical devices that are very energy consuming and/or have a long life. A multiobjective problem is expressed to find the trade-off between the conflicting goals of the designer. The first one is to minimize the mass of the transformer and its cost. The second goal is to reduce the emission of gazes that contribute to green house effect. The life cycle analysis is used to build a model of the contribution to the global warming from the raw material extraction to the end of life of the product. This model is used jointly with economic and electromagnetic models in a multiobjective optimization to find the trade-off between the designer’s goals, helping him to take a decision. Keywords: Ecodesign, Multiobjective optimization, Life cycle assessment, Transformer, Railway traction, Multidisciplinary optimization. 1. Introduction The ecodesign consists in including environmental impacts in the design product procedures. This integration is based on a global approach, multiple environment criteria and taking into account every stage of the product life cycle. Life cycle assessment (LCA) is a standard rigorous methodology that evaluates environmental impacts. LCA is carried out in
four stages (manufacturing, distribution, use and end-of-life) which include goal and scope definitions, inventory, impact assessment and interpretation according to the ISO 14040 standard [1]. Indeed, it considers every stage of LCA of products in order to avoid the displacement of inherent pollutions in design procedure. Indeed, every characteristic modification of products has impacts throughout their life cycle. Moreover, the
ecodesign allows seeking a compromise between technical (feasibilities…), economical (cost…) and environmental (impacts…) constraints. A recent approach including ecodesign of electromagnetic energy converters by [2] deals to evaluate the global life cycle energy. The global energy consumption is the sum of the LCA energy (proportional to the materials’ masses) and the power losses (depending on the load profile). Unfortunately, this approach does not allow to evaluate and to display all environmental impacts in the product life cycle. In this paper, one approach for the ecodesign of electrical machines is presented. It is applied to an auxiliary transformer for the railway traction that is energy consuming and has a long life. This auxiliary transformer is used in intercity high speed train. Firstly, the calculation of impact indicators in every stage of the product life is obtained by environmental information and management explorer (EIME) software [3][4] using LCA methodology. Secondly, a model of the emission of gazes that contribute to the green house effect is built thank to the design of experiments (DoE) method and EIME software. Thirdly, the emission model is used jointly with economic and electromagnetic models. And finally, a multiobjective optimization is performed to find the trade-off between the emission of gazes that contributes to the green house effect and the mass of the device that is related to its cost.
2. Impact indicators The impact indicators are evaluated at each stage of the product life cycle. The stages are the manufacturing, the distribution, the use and the dismantling at the end of life. In our study, the distribution is not taken into account and the impacts during the dismantling stage are added with those during the manufacturing. The indicators for the initial transformer are shown in fig. 1. The displayed stages of the product life cycle are the manufacturing and dismantling together (M) and the use of the product (U). In EIME, 11 impact indicators are computed. They are Raw Material Depletion (RMD), Energy Depletion (ED), Water Depletion (WD), Global Warming Potential (GW), Stratospheric Ozone Depletion Potential (OD), Air and Water Toxicity (AT, WT), Photochemical Ozone Creation (POC), Air Acidification Potential (AA), Water Eutrophication (WE), and Hazardous Waste Production (HWP).
The impacts during the use of the product are far more considerable than the ones of the manufacturing and dismantling. The only impact that is not negligible during the manufacturing is the raw material depletion. All the impacts during the use of the product are related to the energy consumption, i.e. the losses of the transformer. Therefore, maximizing the efficiency of the transformer can highly reduce the environmental impacts.
Figure 1: Impact indicators during the manufacturing and dismantling (M) and use (U) of the auxiliary transformer. Impact during distribution is not computed.
3. Impact model In order to perform an optimization, a malleable model of the impact indicators is built. The authors focus on the global warming that expresses the emission of gazes that contribute to the green house effect. In fig. 1, it is obvious that they are produced at 99% during the use of the product. To avoid the displacement of inherent pollutions from one stage of the product life cycle to another during the optimization, the model includes also the manufacturing and dismantling stages. The global warming potential depends on quantity of each material that appears in product breakdown structure (PBS) and amount of electrical energy required for transformer losses: ⎡α 1 ⎤ ⎡ M iron ⎤ ⎥ ⎢α ⎥ ⎢ M GW = ⎢ 2 ⎥ ⋅ ⎢ copper ⎥ ⎢α 3 ⎥ ⎢ M epoxy ⎥ ⎥ ⎢ ⎥ ⎢ ⎣α 4 ⎦ ⎣ M nomex ⎦ Piron ⎤ ⎡ 1 ⎤ ⎡ + α 5 ⋅ life ⋅ ⎢ ⋅⎢ ⎥ ⎥ ⎣duty ⎦ ⎣ Pcopper + Pothers ⎦
the the the the
(1)
where M x is the mass of material x, Px is the loss in material x, life is the product life, duty is the duty cycle in %, and α x are coefficients. They are calculated by using the design of experiments (DoE) method and EIME software. As the model of GW is linear, the DoE is simple and requires only 6 numerical computation of EIME. It is assumed that the duty cycle of the auxiliary transformer is not related to the trip of the train. Indeed, this transformer is not the one used to supply the railway traction system. The product life is about 200,000 hours and the auxiliary transformer is at full load during 67% of its life and at no-load during 33%. The weight of the transformer is small and negligible compared to the traction system and it doesn’t influence the weight onboard neither the whole train energy consumption during acceleration and braking.
A sizing model including electric, magnetic, and thermal phenomena is used jointly with the impact model. The most common way to build this kind of model is to assemble a system of equations that describe the electric, magnetic and thermal phenomena. The coupling between the thermal and electromagnetic equations is strong and leads to a significant computation time to obtain the exact solution (fig. 2). temperatures
electric and magnetic model thermal nodal model losses
temperatures_out
Some outputs of the sizing model such as the masses and losses are inputs of the impact model. All the models are aggregated to form a multidisciplinary model.
5. Multiobjective optimization The ecodesign problem of the auxiliary transformer is now expressed as a bi-objective problem: ⎡GW ⎤ min ⎢ ⎥ ⎣ Mass ⎦
thermal nodal model losses
(3)
s.t. temp _ in = temp _ out
Figure 2: Exact solution of the electromagnetic and thermal equations
Another approach is to have an inexact solution of the model. In this approach, the strong coupling between both models is suppressed and the computation time is smaller (fig. 3). One additional input of the model is the initial (guess) temperature. With this value, the copper resistivity is computed The solution is exact only if: temp _ in = temp _ out
temperatures_in
Figure 3: Inexact solution of the electromagnetic and thermal equations
4. Multidisciplinary model
electric and magnetic model
Therefore, a constraint is added in the formulation of the optimization problem. This is very effective for the convergence of the optimization and greatly reduces the computation time. At the optimal solution, the constraint is fulfilled and the solution is exact. During the optimization process, the algorithm works with inexact solutions for a faster convergence.
(2)
It can be solved by stochastic multiobjective optimization methods, e.g. NSGA-II [5] and SPEA2 [6], or transformed in a set of monoobjective optimization problems by using a weighted sum of objectives or a goal attainment method. The weighted sum of both objectives is used as a single objective and expressed in the reduced optimization problem (4). This method is effective at one condition that is checked after the optimization. The weight w takes one hundred values uniformly distributed between 0 and 1. The objective functions (i.e. Mass and GW) are normalized to avoid numerical problems.
min w ⋅ GW + (1 − w ) ⋅ Mass s.t. temp _ in = temp _ out
(4)
The set of one hundred solutions is shown in the objectives space on fig. 4. This curve is called the Pareto optimal set or Pareto front. As this curve is convex, the weighted sum method is effective and there is no need to use more sophisticated optimization techniques such as NSGA-II or adaptive weighted sum [7]. Each point of this set is Pareto optimal; this means that it is an optimal trade-off between both objectives. Due to normalization, the initial design of the transformer is plotted on the graph with a small circle at position (1,1). It is obvious that improvement of both objectives is possible. If only one of the objectives is improved, the GW can be reduced of 42.6% or the mass can be decreased of 15.6%. With the Pareto front, designers can take decision for the product taking into account economical and ecological criteria with no a priori on their relative importance.
Figure 4: Trade-off between the global warming potential and the mass of the auxiliary transformer
6. Conclusion One approach is proposed for the ecodesign of an auxiliary transformer for the railway traction. A model of environmental impacts is built with the life cycle assessment methodology and the design of experiments technique. A multidisciplinary model taking into account the electric, magnetic, thermal phenomena is used jointly with the impact model in multiobjective optimization procedure. This approach performs to find all the optimal tradeoff between economical and ecological criteria. It’s a powerful tool for decision making. Applied to the transformer, it shows that the emission of gazes that contribute to the green
house effect can be reduced of 42% or the mass can be decreased of 15%. Moreover, a reduction of both objectives is possible.
References [1] ISO 14040: 2006 Environmental Management – Life Cycle Assessment – Principles and Framework. [2] Debusschere, V., Ben Ahmed, H., and Multon, B., “Eco-design of Electromagnetic Energy Converters: The case of the electrical transformer”, proceedings of IEMDC’07: Electric Machines & Drives Conference, May 2007, pp. 1599-1604. [3] Jean, P., Coulon, R., and Timmons, D., “Building an EcoDesign toolkit for the electronics industry”, proceedings of EcoDesign'99: First International Symposium on Environmentally Conscious Design and Inverse Manufacturing, Feb. 1999, pp. 701-706. [4] EIME, indicator manual v1.9, http://www.codde.fr [5] K. Deb, “A Fast and Elitist Multiobjective Genetic Algorithm: NSGA-II”, IEEE Trans. Evol. Comp., vol. 6, no. 2, Apr. 2002. [6] E. Zitzler, M. Laumanns and L. Thiele. "Spea2: Improving the strength Pareto evolutionary algorithm", Technical Report 103, Gloriastrasse 35, CH-8092 Zurich, Switzerland, May 2001. [7] I.Y. Kim and O.L. de Weck, “Adaptive Weighted-Sum Method for Bi-objective Optimization: Pareto Front Generation”, AIAA/ASME/ASCE Conference, Palm Spring, California, Apr. 19-22, 2004.
