燃料电池流道内流速的六种实测方法与流体分配均一性研究

Investigation and equalisation of the flow distribution in a fuel cell stack

Florian Bürkle
Florine Moyon
Lukas Feierabend
Jens Wartmann
Angelika Heinzel
Jürgen Czarske
Lars Büttner

Abstract
The possibility to use fuel cells as an electrical power source makes them interesting for a wide range of applications． In this work， computational fluid dynamics （CFD） simulations and optical measurements are performed to predict the flow distribution in a flow setup resembling the parallel flow circuits in fuel cell stacks． For the first time it is shown that by an adaptation of the port sizes in the inlet manifold to the individual fuel cells， the average global deviation between the flow rates can be reduced from 10．1％ to 4．0％ by means of a model experiment． The measurements are performed with a high resolution laser Doppler velocity profile sensor （LD－PS） specifically developed for measurements in small－scale channels， in this work 4＊1mm2， allowing for a spatial resolution below 2um and relative velocity uncertainties below 0．1％， helping to resolve installation effects possibly occurring in fuel cells to improve their efficiency． The presented results can be used by manufacturers to increase the efficiency of their fuel cell stacks．

文章首先介绍了燃料电池流道内流速的测量方法，

The velocity measurements are performed within the channels of the
experimental flow model with a nominal depth of 1 mm， which sets
challenging requirements for the optical measurement system as microscale flow measurements need to be performed． Furthermore， the relative flow velocity deviation between the single channels is expected to be below 1％， such that precise velocity estimations are necessary． Conventional mass or volume flow measurement methods （e．g．

第一种方法ultrasonic，

第二种方法thermal mass or

第三种方法magnetic flow meter） are designed for bigger flow diameters and have high uncertainties （＞＝ 1％） ． The most commonly used methods to obtain micro－scale flow data are particle image／tracking velocimetry （PIV／PTV） and laser Doppler anemometry （LDA）．

第四种方法是粒子图像跟踪测速系统（适用于燃料电池）

An established method for micro－scale flow measurements is
astigmatism micro PTV （A－μPTV） which allows 3D3C measurements
with a spatial resolution below 1 μm along the optical axis with a
measurement volume length in the order of 100 μm．

该方法的缺点：However， for evaluation the measured data is binned in grids with sizes of ＞ 10 μm， worsening the spatial resolution． Furthermore， the possibility of three－dimensional and three－component （3D3C） velocity field measurements also has no advantage here as the flow has one main direction with negligible orthogonal components． Finally， the relative standard deviation of the velocity measurement of μPTV／μPIV techniques is greater than 1％， mostly around 2–3％． For the measurements， lower deviations are required as uncertainties should be one order of magnitude lower than the phenomenon that is observed as a rule of thumb．

第五种方法是激光多普勒速度和方向测定法（适用于燃料电池）

根据多普勒效应，用激光束照射运动物体，检测并记录反射、透射、散射的光与参考光调频后的信号从而获得运动物体速度的方法。

With the LDA technique， velocity measurement uncertainties of
less than 1％ are possible． However， the spatial resolution depends on
the length of the measurement volume， which can be in the order of 100μm． Further reducing the measurement volume length is
possible by decreasing the focal width of the beams． This leads to a
higher curvature of the Gaussian beam resulting in a higher gradient of the fringe spacing， which causes a higher relative velocity uncertainty． The dependence can be expressed by Δv／v∝（1／l）2 where l is the length of the measurement volume． Hence， there is a lower limit to the product of spatial and velocity uncertainty similar to Heisenberg’s uncertainty principle． To overcome the limitations of PIV／PTV and the diffraction limit of LDA， the changing curvature of the Gaussian beam can be used to build a optical measurement system based on conventional LDA．

第六种方法是激光多普勒轮廓传感器

To combine high spatial resolution with low uncertainty in velocity，
the laser Doppler profile sensor （LD－PS） was developed which was
already used to measure micro－scale flows with micrometre resolution．

燃料电池设计的单流道流阻模拟和实物照片

单流道内流体流动和激光多普勒轮廓传感器的实测数据，这个实测数据简直令人叹为观止。

网格粗细对仿真结果的影响和进出气口流型的不完美性。

不同流道深度的差异

流道深度的加工精度有限，但流道宽度进行不同宽度设计，使各流道内部的流动速度均一。

Conclusion and outlook

For the first time it was experimentally shown that by adjustment of
the channel inlet sizes of a HT－PEMFC stack model it is possible to
achieve a more equal flow distribution． The results of this study can be
used to equalize the flow distribution of a real operated fuel cell stack in order to increase its efficiency．

The work presented in this paper focused on the investigation and
improvement of the flow distribution in a HT－PEMFC stack with a U configuration．A reduced transparent flow model based on the fluidic
circuit design of the HT－PEMFC stack was used to allow flow investigation with measurements by a laser－Doppler velocity profile sensor and
CFD simulations． The LD－PS exhibits excellent properties for micro－scale flow measurements． With a spatial resolution of less than 2 μm and a relative velocity uncertainty of below 0．1％ it is the instrument of choice for high precision micro－flow measurements． Even with an SNR（信噪比） below 0 dB， measurements with uncertainties below 0．2％ are possible． The usage of fluorescent particles also allows measurements close to the walls of the measurement object， which will be important for measurements in micro－flows above a gas－diffusion－layer （GDL） used in fuel cells．

The fluidic stack model was designed to have similar flow characteristics as the real stack： a U－configuration with similar pressure drop
characteristics of the fuel cells represented by straight channels， similar
manifold dimensions， and equal distances between the T－junctions． Due to the material choice for the construction （PMMA） and the accompanied limited machining precision， the channels exhibited non－negligible variations in depth， which lead to variations of the flow resistances of the individual channels． These variations modified the cross section of the channels．

It was observed that the trend of the flow distribution in the stack
corresponded to the trend of the channel depth measured after the
milling process． The numerical model agrees well with the measurements， which supports the approach of utilizing CFD simulations for the investigation and design of similar fluidic circuits． A maximal global deviation of the gas flow of 45％ between the channels was calculated and measured． With these results， it has been shown that the tolerances have a strong influence on the flow distribution． Thus， this problem of mal－distribution of the reactant flow may not only concern large stacks with a high number of cells， but also stack with non－negligible manufacturing tolerances of the individual components．

The geometry of the manifold was modified to improve the flow
distribution in the stack． The port size at the inlet of the channels was
individually adjusted iteratively to allow for varying additional flow
resistances for the channels． CFD simulations with the final manifold
design showed a significant improvement of the flow distribution
leading to a maximum global deviation of the volumetric flow rates
between the channels of 7，6％．

Unfortunately， the measurements did not show the same improvement， which is assumed to be due to manufacturing tolerances for the
modified manifold geometry． However， the mean deviation could still be reduced from 10．05％ to 3．98％． Apart from the sensitivities to the
manufacturing tolerances， the applied methods prove to be accurate and effective to investigate the flow distribution in comparable fluidic circuits and consequently improve the design of the manifold system．
Furthermore the adaption of the channel inlet port sizing seems to be
effective if manufacturing tolerances can be compensated． Further investigations aim for the total equalisation of the flow as a more even
distributed flow can increase the efficiency of an operated fuel cell stack． With higher efficiency fuel cells are a more attractive alternative for other electric power sources like batteries．

This work focused on a single operating point of the fuel cell stack． In
real applications， fuel cell stacks are operated within a wider range． Thus it will be crucial to identify manifold geometries， which will work
optimal for the targeted range of operation． Furthermore， one could
think of adjustable nozzles for the channel inlet ports， however this
would add another degree of complexity to the fuel cell stack．