Multiobjective Optimization of Energy and Power per Unit Mass and Area for LiFePO4-Graphite Li-Ion Battery Cells

Abstract

Lithium-ion battery has been considered as a promising solution to the current unsustainable fossil fuels based energy economy and the CO2 emission problem in metropolitan areas due to its high specific energy, high efficiency and long life. To study the properties and optimal design of Lithium-ion batteries, simulation models and optimization problems have been constructed and heavily studied by researchers. Most researchers focused on single objective oriented studies when they tried to come up with optimal designs of Lithium-ion battery cells. In a previous paper from the authors, a multi-objective optimization problem was constructed for optimal design of lithium-ion batteries. By employing a reaction zone simulation model and Genetic Algorithm, the Pareto front of the constructed problem was explored qualitatively. It was also shown that the problem constructed by authors had a wide applicability and its solution can be used to handle different design problems that designers may encounter in the industry.This paper presents a procedure for optimizing the design of battery cell considering multiple performance measures with a tuned compact differential-algebraic equation (DAE) model for the LiFePO4-graphite battery cell. The employed simulation model shows a good agreement with the experimental data at discharge/charge rates up to 4C and allow more design variables to be tuned at the same time.Firstly, two performance measures are considered as the objectives of the optimization problem, maximizing the energy for a constant discharge rate per unit separator area and minimizing the mass per unit separator area. Particle size, active material volume fraction and electrolyte volume fraction of positive and negative electrodes are selected as design variables. With these objectives and design variables picked up, a multi-objective optimization problem is constructed. This optimization is again widely applicable in different cases.To estimate the Pareto front of the constructed optimization problem, a Genetic Algorithm is implemented. Parallel computing is used to better utilize the available computation power. Due to the good validity of the simulation model employed in this paper, the Pareto front obtained is quantitatively meaningful and considered valid for design uses. As a whole Pareto optimal set is available, the proposed procedure will offer more flexibility to the product designers so that they are allowed to pick up the most fitted designs for different target applications with one single run of the optimization.Then, another objective used to reflect the high power performance is introduced to the optimization problem and the solution to the newly constructed problem is given by Genetic Algorithm as well. The Pareto front for the problem becomes a surface in a three dimensional response space and is expected to be able to handle even more design problems.The proposed procedure is expected to be useful for design optimization problems for all kinds of products. The results of this paper are expected to offer references to product development or design of experiments in the industry.