This document shows that ShowerPower® cooling technology, applied to standard motherboard power modules, is very well suited for use in wind power applications, achieving outstanding thermal properties, reducing size and weight, as well as the cost of the converters.
1. Introduction
Direct liquid cooling of power modules for wind turbine converters offers a more reliable, compact, and cost-effective solution than traditional indirect liquid cooling technologies, such as the radiators (“coldplates”) currently used in the wind industry. Traditionally, wind turbine power converters have used two types of power modules in combination with indirectly cooled coldplates. The first are baseplate-less modules, where the DCB substrates carrying the IGBT chips and diodes are mounted directly to a coldplate without a typical copper baseplate. This design eliminates the baseplate, which adds cost and an extra layer of thermal impedance to the heat stack. The second option uses standard modules with flat baseplates, and while these add additional thermal impedance, the baseplate helps reduce the overall thermal impedance and system cost by leveraging the dissipative effects of copper. This design ensures dual-source deliveries of the standard power module (an indispensable requirement in the wind industry) along with the benefits of volume sales in module cost.
Given that finned radiators are not welcome in the wind energy industry due to their custom design and dual-source requirements, there are basically three other options:
• Indirect liquid cooling using cold plates with power modules without a base plate;
• Indirect liquid cooling using cold plates with standard power modules;
• Direct liquid cooling with standard power modules.
The thermal stacks were analyzed and compared. All three solutions will be subjected to a typical mission profile (describing the load profile for the entire service life) derived from a wind energy application. The reliability implications for all three solutions will also be discussed.
Since a direct liquid coolant offers superior performance compared to a closed coolant, which requires a thermal interface material (TIM) layer, it is worth considering as a means to overcome the challenges of leakage risks and other issues associated with open liquid coolant designs.
2. Power Module Failure Mechanisms Under Load
The rationale for addressing power module cooling stems from the fact that power modules inevitably fail over time; it's simply a matter of when. The failure mechanisms depend on the loads the modules endure over time, as well as the temperature: cooler is better. Therefore, efficient cooling is a cornerstone of cost-effective, high-reliability converter designs.
Power modules consist of assemblies of materials with varying material properties, such as coefficient of thermal expansion (CTE), modulus of elasticity, and strength. The term "thermal stack" is used to describe this thermomechanical combination of materials that define the power module. Differences in material properties across the stack cause thermomechanical movements as the temperature changes, inducing stresses and strains that can lead to fatigue failure. The dominant failure modes are bond wire fatigue and weld joint degradation.
Apart from the “dry-out” and “pump-out” effects related to the TIM layer, bonding cable fatigue is the dominant failure mode in power modules for wind turbine converters. Due to the CTE mismatch between the silicon chip and the aluminum bonding cable, which, along with cyclical temperature changes (power-off and power-on), leads to failure, the bonding cable will fail, resulting in catastrophic failure. Several models exist for evaluating bonding cable life as a function of loads over time (the mission profile) and thermal conditions [1, 2].
3. Wind Applications
The main current converter systems can be reduced, from a feed cycle perspective, to three basic topologies: Doubly Fed Induction Generator (DFIG), Full-Converter with Gearing (FC), and gearless, known as Direct Drive (DD). These different topologies are characterized by their dominant operating frequencies. The grid power profile differs because it operates at the nominal frequency. Furthermore, depending on the control strategies and generator design, the different topologies will result in different switching patterns between diodes and IGBTs, thereby affecting the resulting fatigue effects.
The mission profiles of DFIG and FC wind turbines differ in the frequency distribution of their thermal cycles. While DFIG results in lower overall power loss, the slow temperature cycles that dominate its mission profile lead to greater damage to the power modules. Direct liquid cooling helps reduce both the average temperature in the modules and the temperature rise, both of which are necessary to meet typical specifications. Power modules in FC wind turbines have to cope with higher average power losses. Therefore, lowering average temperatures is essential. This can be perfectly achieved through direct liquid cooling of the power modules, leveraging the flexibility of ShowerPower® cooling designs.
4. ShowerPower®
ShowerPower® is a direct liquid cooling concept developed by Danfoss. Its main features are:
• Ability to homogeneously cool large power modules on flat base plates and module systems, eliminating temperature gradients, improving lifespan, and facilitating the parallel placement of many power chips.
• No pump-out or dry-out effects related to TIM
. • Very low pressure drop differential
. • Enables 3D design of compact, lightweight, and high-grade designs.
• Low cost: metal-to-plastic conversion into simple plastic parts.
The key element of the concept is the “ShowerPower® turbolator,” which guides the coolant along the module's base plate into parallel cooling cells, ensuring uniform temperatures across the module.
The concept allows for tailored cooling if hotspots require additional attention; this is achieved simply by individually designing the cooling channels.
The ShowerPower® turbolator is used in conjunction with a manifold gasket and to distribute coolant under a standard IGBT base plate.
For further information, please refer to [3-5].
5. Indirect vs. Direct Cooling
: Direct liquid cooling eliminates the TIM layer traditionally required between the back of the power module and the cold plate. Because the TIM layer comprises 30%–50% of the Rth (reverse thermal bond), eliminating it results in a better thermal environment for the power module. Since temperature-driven failure mechanisms are the dominant cause of power module failures, this leads to increased reliability.
The transient thermal impedance has been compared for the three thermal stacks:
• Module without cold-plate base plate
• Module with cold-plate base plate
• Module with ShowerPower® base plate
The graph shows that the three solutions behave very differently: Due to the thermal time constant of the baseplate, the Zth of the module without a baseplate begins to increase at 60-70 ms compared to the modules with a baseplate; the impedance at 1 s is twice that of the modules with a baseplate, indicating that a 1 Hz power cycle would result in twice the temperature increase for the module without a baseplate. The steady-state Rth for the module without a baseplate is also twice that of the ShowerPower® configuration.
At ~1.5 s, the Zth curve for the module with a coldplate begins to diverge from the ShowerPower® curve, showing that the temperature increase for frequencies below ~0.6 Hz will be greater for the coldplate solution than for the ShowerPower® solution.
6. ShowerPower® for Direct Liquid Cooling of a Wind Turbine Converter:
The wind turbine manufacturer knows the energy loads throughout the turbine's lifetime. Knowledge of the thermal stack, including cooler efficiency and power loss over time, leads to time-dependent device temperatures. Taking this into account, the reliability of the power module can be determined. If the lifetime does not meet the requirement, an improved thermal design will be necessary. This iterative process is repeated until a design is found that meets the lifetime requirement with a reasonable margin.
Thermal analyses were performed using FloTherm®, a 3D CFD software that allows modeling not only conduction through the solid elements of the heat stack, but also the effects of fluid flow through the cold plate or the ShowerPower® module.
Two analyses were performed, comparing Rth for the three cooling principles using 50%/50% glycol/water.
A power dissipation constant was applied to the single-leg IGBT chips of a P3 module, and the maximum IGBT chip temperature was recorded.
Figure 6 shows that the module without a baseplate recorded a TJ of 138°C. The standard module with a baseplate and coldplate improved upon the solution without a baseplate, achieving a TJ of 120°C. Finally, the ShowerPower® cooler improved both records, achieving a TJ of 97°C.
Next, a transient analysis of the cooling technologies was investigated (again using glycol/water 50%/50%) using a standard wind turbine profile at frequencies below 1 Hz.
In this particular example, the diode losses are slightly higher (generator mode), and therefore the diode temperatures are also slightly higher. Both the IGBT and chip diode temperatures were recorded, and the temperature rises and mean temperature (Tmean) are listed below.
It can be observed that the solution without a base plate performed worse in terms of both mean temperature and Tmean, as well as temperature swing (DT), with values of 75°C and 43°C, respectively, for the diode.
The standard baseplate with coldplate solution outperformed the solution without a baseplate, achieving a Tmean of 70°C and a DT of 24°C. Finally, the ShowerPower® solution again outperformed the other two technologies, achieving a Tmean of 60°C and a DT of 23°C. The same trends were observed for the IGBT.
'
Benefits of ShowerPower®
Increased Dissipation Power
Based on the thermal analysis in Figure 6, the table below shows how much the steady-state dissipation power can be increased while maintaining the same junction temperature of the power semiconductors.
ShowerPower® allows you to double the dissipation power in the coldplate solution without a base plate.
Reduced silicon requirement.
Furthermore, ShowerPower® allows you to reduce the amount of silicon required by half compared to the baseplate-less solution.
Useful life
Since lifetime is closely dependent on temperature (DT and Tmean), Tables 1 and 2 clearly show that a baseplate will drastically increase lifetime compared to the module without a baseplate, as both DT and Tmean are reduced. ShowerPower® further increases lifetime because Tmean decreases. Again, this can be used to offer the same lifetime as both baseplate-less and coldplate-based solutions but with significantly less silicon, resulting in substantial cost reductions.
7. Testing Direct Liquid Cooling Circuits
A liquid cooling system with 15-20 years of service life must be carefully designed with regard to liquid sealing, corrosion resistance, and consistent performance. Sealing is a potential weak point; however, accelerated life testing results have confirmed the reliability of the chosen solution. Another important result is obtained from various long-term performance tests. Relevant results are presented regarding coolant stability, the corrosion behavior of the materials involved, and their influence on cooling performance. Long-term Rth and sealing performance are studied using a special long-term test bench with a defined corrosion limit condition.
The test bench tubing is made exclusively of aluminum and non-metallic materials. All non-standard components, such as heaters, flow meters, pumps, and sensors, are specially designed in aluminum or high-temperature plastic. The goal was to achieve long-term operation at a coolant temperature of 105°C, a safe operating point for a 50% water/glycol mixture at a system overpressure of 1.5 bar. Heat is generated by a specially designed electric heating tube. For thermal stability and energy efficiency, the entire test bench is encapsulated within an insulated metal test platform, and data logging is performed from the cooler external environment outside the platform.
The Rth efficiency of the power modules is determined after every 1000 hours on a different test bench at 25°C. For this purpose, the power modules are electrically powered, and the dissipated power is drained by the coolant. The junction temperature of the semiconductors is measured by IR thermography.
The hot run period is interrupted every 5 days by 2 days at ambient temperature with no coolant circulation. This forces the seal (double O-ring) to move within the groove, driven by temperature differences.
The investigation begins with new, unused power modules with standard (blank) surfaces (AISiC and Ni- or Cu-plated).
During periods of high heat, the power modules are passively heated. The water/glycol coolant is a mixture of demineralized water and a standard glycol with corrosion inhibitors for aluminum and steel. The inhibitor is an anodic type that forms a passivation layer on aluminum surfaces, preventing metal oxidation. Careful dosage control is required because if the anodic inhibitors are used at too low a concentration, they can actually exacerbate pitting corrosion by forming a non-uniform layer with localized anodes [6,7].
The correct water/glycol mixture is monitored weekly by refractive index measurements. Figures 11 and 12 show the Rthjw measurement results. The AlSiC material exhibits very stable Rth behavior over time, with only minor variations initially after forming a thin protective corrosion layer. The corrosion inhibitor plays an important role in the thermal pathway, as well as in long-term corrosion stability. This is shown in the same experiment but using Ni-plated or Cu-based material, Figure 12.
The Ni plating becomes coated with a protective layer within the first 100-300 hours of operation. This layer slightly increases flow resistance, as can be seen in the refrigerant pressure drop in Figure 13.
To ensure consistent cooling performance, even during the initial period of protective layer formation on system surfaces, it is recommended to maintain a constant pump speed. This strategy results in consistent heat transfer from the power module to the coolant.
A second successful outcome of the long-term observation is that no leaks were observed in the DUTs after the long-term cycle.
The experiment was conducted for up to 8000 hours at 105°C to allow extrapolation of the results to 20 years of operation at a refrigerant temperature of only 45°C.
8. Conclusion
Direct liquid cooling using ShowerPower® has proven to offer several advantages over standard liquid cooling technologies with cold plates: increased reliability, regardless of the application (e.g., DFIG or FC), as well as reduced weight, size, and overall power stack cost.
Furthermore, accelerated life tests have demonstrated that corrosion and water tightness issues are controlled, making these solutions reliable for the 20-year service life required by the wind energy industry.
ShowerPower® will be the best global choice for future generations of liquid-cooled converters for the
wind industry.
Author: By: Klaus Olesen (
9. Literature
[1] M. Held et al. “Fast power cycling test for IGBT Modules in traction applications”, PEDS Conference Proceedings 1997, Singapore, pp. 425-430.
[2] M.-H. Poech al, “Investigations on the damage mechanisms of aluminum wire bonds used for high power applications.” Proc. EUPAC 1996, pp. 128-131.
[3] K. Olesen et al., ”ShowerPower® New Cooling Concept”, PCIM Conference Proceedings 2004, Nuremberg.
[4] F. Osterwald et al., “Innovative Kühltechnologie für Leistungsmodule”, Bauelemente der Leis-tungselektronik und ihre Anwendungen, 10.-11. Oktober 2006, Bad Nauheim.
[5] R. Bredtmann et al., “”Power under the hood” Increasing power density of inverters with a novel 3D-approach”, APE, March 25-26 2009, Paris.
[6] Dr. Dmitri Kopeliovich, “Anodic corrosion inhibitors” www.substech.com
[7] H.-D. Held, „Kühlwasser; Verfahren und Systeme der Aufbereitung, Behandlung und Kühlung von Süßwasser-Brackwasser-Meerwasser zur industriellen Nutzung“ 5. Auflage
