Precise lifetime predictions for lithium‐ion cells are crucial for efficient battery development and thus enable profitable electric vehicles and a sustainable transformation towards zero‐emission mobility. However, limitations remain due to the complex degradation of lithium‐ion cells, strongly influenced by cell design as well as operating and storage conditions. To overcome them, a machine learning framework is developed based on symbolic regression via genetic programming. This evolutionary algorithm is capable of inferring physically interpretable models from cell aging data without requiring domain knowledge. This novel approach is compared against established approaches in case studies, which represent common tasks of lifetime prediction based on cycle and calendar aging data of 104 automotive lithium‐ion pouch‐cells. On average, predictive accuracy for extrapolations over storage time and energy throughput is increased by 38% and 13%, respectively. For predictions over other stress factors, error reductions of up to 77% are achieved. Furthermore, the evolutionary generated aging models meet requirements regarding applicability, generalizability, and interpretability. This highlights the potential of evolutionary algorithms to enhance cell aging predictions as well as insights.
Lifetime validation is crucial to the development of lithium-ion cells, especially in the automotive context as high regulatory and warranty requirements have to be met. As the experimental lifetime validation can get tedious and costly, lifetime modelling is a way to cut down experimental effort and predict cell aging based on accelerated aging experiments with well selected aging conditions.Using these experiments with constant aging conditions, lifetime models are widely built in a path independent way, where the aging is expressed as a function of only the state of health in capacity (SoHC) and resistance (SoHR). The SoH hereby defines the relative capacity and resistance in comparison to their values at begin of life.Following a design of experiment approach, 62 large format NMC/graphite lithium-ion pouch-cells were tested at 28 different aging conditions, coming close to realistic cyclic aging conditions. This experiment was initially set up in order to build such an empirical lifetime prediction model, capable of predicting SoHC and SoHR in dependence of factors such as the state-of-charge window of the cyclation, the temperature and the charging power.Analyzing this accelerated aging experiment, we show that the definition of a state of health based on only capacity and inner resistance is falling short.This is deduced from differential voltage analysis (DVA) and post mortem analysis from cells with comparable SoH. With differential voltage analysis, changes in the electrochemically active parts of the lithium-ion cell that arise from aging and affect the open circuit potential can be visualized and assigned to certain parts of the cell without its destruction.Hereby a published fitting algorithm is used that can estimate losses of active material (LAM) and the loss of active, cyclable lithium (LL) from slow-discharge potential curves. Following the destructive post mortem analysis of a cell, we validate the results of this algorithm with 3-electrode-measurements.Together with further material analytics entirely different internal states become apparent when LAM on anode and cathode side as well as the loss of active lithium and its distribution in the electrodes are compared.Based on these findings we discuss the implication of the underlying variation in internal states at similar SoH of lithium-ion cells to the modelling of cyclic aging with a path-independent damage accumulation approach following Miner’s rule.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.