How long will my battery last? The perils of prediction – By Dr. Edmund Dickinson, Head of Electrochemistry at About:Energy
Battery degradation is not an esoteric topic – most of us have owned a smartphone and learned through observation that battery capacity declines with use. This concern becomes more marked for current or prospective electric vehicle (EV) owners, and perhaps yet more so EV manufacturers who must gauge performance against warranty and budget for future claims. In a passenger car, though, more limited capacity and power performance after time is mainly an irritation; for some applications, such as emergency power supplies or aerospace, a precise assessment of aging is safety-critical.
Batteries are intrinsically chemical systems, and their aging comprises various degradation phenomena: mainly chemical reactions, with thermal and mechanical effects also playing a part. The capacity of a lithium-ion battery depends on how much lithium can be moved reversibly between the two electrodes. Chemical reactions that prevent the lithium from reacting at either electrode, or mechanical fatigue causing particles to become electrically disconnected, lead to loss of capacity.
Just as for vehicle crash testing, we cannot test degradation during a design phase exactly as it will occur in practice. Design decisions require predictions based either on an experimental surrogate or a model; in practice, both will typically be used. But even for experimental testing on individual cells, we cannot perform a 10-year longitudinal experiment before a product is released! Extrapolation may be required, or inferences may be drawn from the behaviour of similar cells in the past. Or, we can use an ‘accelerated aging’ study in which a cell is aged much more rapidly than the real-life scenario through constant, aggressive utilisation. This yields much faster insight into cell failure mechanisms, but whatever we learn must be somehow projected back to the real-world operating conditions before use.
Also, in degradation analysis, it is common to focus on the behaviour of the cells making up a battery pack. Cell performance is removed from practical system performance, however – in a pack comprising hundreds or thousands of cells, managed by a Battery Management System (BMS), the degradation of pack performance might either be less or more severe than the individual cell. The performance evolution of the integrated system is also tightly coupled with its thermal management, which, again, rests on a heat source that increases due to degradation!
Due to these limitations, almost all approaches to degradation analysis are likely to lead to excessively conservative design: a typical consequence is pack ‘oversizing’ to compensate for future capacity and power performance loss. This oversizing may also apply to balance-of-plant hardware such as thermal management systems, or may manifest through artificial power limitation from a BMS to mitigate degradation. This conservatism has negative consequences for the efficiency of a battery deployment, due to excess weight and volume; likewise, it harms sustainability, due to excess material requirements.
If no experiment will prevail, what can we do? Frequently, simulation is the answer. Of course, simulations do not describe reality exactly, but neither can controlled lab experiments correspond identically to the behaviour of a system in a field. Simulations can offer insight by means of scenario modelling: an engineer can consider that if a certain distribution of aging applies to the cells in a battery pack, then the predicted performance of the pack will change in a certain way. The simulation can potentially incorporate the coupled complexities of the cell-to-module, module-to-pack, BMS, thermal management, and even interactions with the vehicle model.
At About:Energy, we specialise in cell characterisation, and developing cell-specific models (electrical, thermal, degradation) that engineers can use in scenario modelling studies. A trusted description of the cell of interest ensures sound decision-making for cell selection and cell integration tasks. About:Energy is currently participating in two projects funded by Innovate UK as part of the Faraday Battery Challenge programme, to develop our capabilities related to cell degradation.
The Faraday Battery Challenge “Voltt” project has supported a series of aging experiments on commercially popular cells, such as the LG M50LT or Molicel P45B, using surface conductive temperature control to ensure highly consistent aging conditions, targeted at model development. We are then working with Imperial College London to perform “degradation mode analysis” on the data – an interpretation of electrical testing data on aged cells to gain insight into their evolving chemistry.
This approach aims to mitigate the problem of path-dependent aging. To handle degradation in a simulation, we need to understand first how to express the ‘state of degradation’ of a cell, and second how to predict its rate of change. While easily measured properties such as cell capacity and cell resistance provide an attractive set of metrics, they are unreliable descriptors of what the cell will do next. Two aged cells with the same extent of capacity fade (we say at the same “state-of-health”) may continue to degrade differently, depending on the underlying physicochemical changes that provoked the capacity fade in each case.
Factors including temperature, operating current magnitude, and state-of-charge (SOC) operating window will all impact aging rate. For instance, cycling at high SOC (high voltage) may degrade the cell more rapidly than in a medium SOC range. SOC dependence is especially prominent in blended electrodes, such as graphite-silicon negative electrodes, where different materials are active at high vs low SOC. Here, degradation mode analysis creates a richer chemical description of degradation that can help to unpick the issues of path-dependent and SOC-dependent aging.
In our assessment of the present state-of-the-art, degradation simulation in industry is far from providing a universally precise prediction of the fate of a battery. In saying this, we must remain aware that experimental data also fail to provide a full picture of what will happen in the field. The advantages of simulation-based approaches – in tandem with carefully measured input data – is that they are overwhelmingly faster and cheaper than forever extending a suite of aging tests. This advantage grows as simulations can be adapted to the use of the same cell in different circumstances, such as different potential pack architectures, BMS algorithms, or vehicle use cases. We project a growing scope for aging simulation in the battery industry as we extend not just model accuracy but also our ability to fairly assess the uncertainty of models and place correct value on their predictions as tools for decision-making.
READ the latest Batteries News shaping the battery market
How long will my battery last? The perils of prediction
