Our worst sin against our fellow man
is not hating people, but indifference.
Environmental Stories with Small Actions
Stories of nature practice together
Multi-objective optimization of orifice-shaped cathode flow field design in polymer electrolyte membrane fuel cells
Multi-objective optimization of orifice-shaped cathode flow field design in polymer electrolyte membrane fuel cells
ABSTRACT
It is well known that the orifice-shaped flow field design of the Polymer Electrolyte Membrane (PEM) fuel cell has advantages in oxygen supply and water removal. This study introduces research that optimizes the key design variables of the orifice-shaped flow field using the previously developed multi-scale multi-phase PEM fuel cell model. Key design variables include reduced channel depth, width, length in the orifice channel region, and the orifice-to-total flow field ratio, optimized to maximize cell voltage (), and minimize pressure drop () and the standard deviation of the current density distribution (). Data for these variables were generated through three-dimensional PEM fuel cell simulations and used to train an Artificial Intelligence-based Multi-layer Perceptron (MLP) model. The trained MLP model, in conjunction with the Non-dominated Sorting Genetic Algorithm-II (NSGA-II), analyzed the impact of design variables to determine the optimal points. Through multi-objective optimization study, it has been successfully demonstrated that compared to the baseline orifice-shaped flow field design, at a current density of 2.5 A/cm2, can be increased by 13 mV, can be reduced by 47.85 Pa/cm, and can be decreased by approximately 0.119 A/cm2, indicating that further performance improvement is possible.
Graphical abstract
The optimization was performed by applying a multi-objective optimization algorithm with voltage (), pressure drop (), and standard deviation of the current density distribution () as objective functions, which are the main elements of the orifice-type cathode flow field used in high-power PEM fuel cells. Based on each objective function, the optimal design points can be analyzed and applied to the PEM fuel cell design to improve high power and stability.
The abnormal climate phenomena resulting from global warming, such as the melting of polar ice in the Arctic and Antarctic that leads to rising sea levels, pose significant challenges for humanity. Additionally, severe droughts and heavy rains in certain regions further exacerbate these social and economic challenges. In response, there is an active global pursuit of policies and research initiatives for environmental conservation. According to the Net Zero scenario outlined by the
NUMERICAL MODEL
In this study, the multidimensional, multiscale 3D two-phase PEM fuel cell model was utilized to rigorously investigate the impact of diverse cathode flow field architectures. This model, developed through various prior studies, accurately captures multidimensional transport mechanisms and electrochemical phenomena, encompassing all key components within a single cell, including the cathode and anode catalyst layers (CLs), micro-porous layers (MPLs), gas diffusion layers (GDLs), gas channels
Surrogate modeling and multi-objective optimization
A surrogate model, also referred to as a meta-model, is designed to infer outputs from limited inputs. These models fall into two primary categories: regression and interpolation. Regression models, like the MLP, Support Vector Regression (SVR), and Polynomial Response Surface (PRS), approximate trends without necessarily intersecting each experimental datum, which helps to filter out noise. This makes them well-suited for datasets with inherent random errors. Conversely, interpolation models,
Evaluation of MLP-based surrogate model prediction performance
Upon formulating the optimization problem and constructing the MLP-based surrogate models as detailed in Section 3, we proceeded to evaluate their prediction performance. The MLP, as employed in this study, yields reliable results for highly nonlinear data relative to other surrogate models. We then assessed the trained MLP's prediction performance via error analysis at , which is illustrated in Fig. 3. The training dataset obtained an adjusted R2 value of 0.9713 and an RMSE of 4.0 mV for ,
CONCLUSIONS
In this study, a multi-objective optimization of the orifice-shaped cathode flow field design was conducted focusing on the performance (), pressure drop (), and uniformity of current density distribution () as objective functions for four key design variables that determine the reduced channel area () and the reduction ratio. The MLP algorithm was applied for surrogate modeling, and the NSGA-II was used to find the multi-objective optimal points for , , and . MLP-based surrogate models were
The authors report there are no competing interests to declare.
Funding
This study was supported by the Korea Evaluation Institute of Industrial Technology (KEIT) and the Ministry of Trade, Industry, & Energy (MOTIE), Republic of Korea (No. 20012121). We thank TAE SUNG S&E, Inc., Korea, for providing technical support with the Ansys Fluent software.
Declaration of Competing Interest
☒ The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Hyunchul ju reports financial support was provided by Korea Evaluation Institute of Industrial Technology. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
대화 참여하기