Elsevier

Journal of Energy Storage

Volume 26, December 2019, 101001
Journal of Energy Storage

Evaluation of batteries residual energy for battery pack recycling: Proposition of stack stress-coupled-AI approach

https://doi.org/10.1016/j.est.2019.101001Get rights and content

Highlights

  • By year 2025, 1 million metric tons of batteries in battery packs must be recycled.

  • The problem on recycling of battery packs in EVs is illustrated in this paper.

  • A stack stress-coupled-AI approach for predicting residual energy is proposed.

  • Experiments are designed to validate the proposed approach.

  • The findings can provide an optimized recycling strategy for spent batteries.

Abstract

It is predicted that by 2025, approximately 1 million metric tons of spent battery waste will be accumulated. How to reasonably and effectively evaluate the residual energy of the lithium-ion batteries embedded in hundreds in packs used in Electric Vehicles (EVs) grows attention in the field of battery pack recycling. The main challenges of evaluation of the residual energy come from the uncertainty of thermo-mechanical-electrochemical behavior of battery. This motivates the notion of facilitating research on establishing a model which can detect and predict the state of battery based on parameters enable to be measured, such as voltage and stack stress. Thus, the present work proposes a stack stress-coupled-artificial intelligence approach for analyzing the residual energy (remaining) in the batteries. Experiments are designed and performed to verify the fundamentals. A robust model is formulated based on artificial intelligence approach of genetic programming. The findings in the study can provide an optimized recycling strategy for spent batteries by accurately predicting the state of battery based on stack stress.

Introduction

Given with the limited amount of traditional energy, the society is working hard to weaken its dependence on fossil fuels and finding alternative energy sources. In the face of deteriorating environment, people are looking forward to other options which are environmentally-friendly and sustainable. With the emergence of lithium-ion battery powered EVs, it looks promising to solve this problem. [1], [2]. However, there are many problems in lithium-ion batteries during its practical applications. For example, the detection of battery state of health (SOH) and state of charge (SOC), battery aging mechanism investigation, and recycling or reuse strategies of spent batteries. Among these problems, battery pack recycling accounts for the major problem and a hot scenario for the next years. By 2025, approximately 1 million metric tons of spent battery waste will be accumulated. How to reasonably and effectively evaluate the residual energy of the lithium-ion batteries used in Electric Vehicles (EVs) grows attention in the field of battery pack recycling. The works on recycling of single cell battery has been paid great attention due to its diverse usage. The major works on recycling of single cell battery is to recover the materials such as cobalt, lithium, etc. by use of hydro-metallurgical methods. Till now, for the whole pack recycling, there is hardly any research being carried out in context of its automation or efficient recycling. The problem is attributed to hundreds of batteries embedded in series of parallel configurations in these packs. This problem of whole pack recycling could be very challenging because there shall be need of development of an intelligent model to sort out the batteries effectively based on the accurate prediction of residual energy. The main challenges of evaluation of the residual energy come from the uncertainty of thermo-mechanical-electrochemical behavior of battery. Thus, an empirical model which can reflect and predict the real time state of health of batteries in EVs also reflects the prediction of residual energy in the batteries.

In this perspective, literature has been conducted. Based on the state-of-the-art studies, the battery modeling methods are summarized as shown in Fig. 1 According to different kinds of battery models, the estimation technologies can be classified into three major types, including (1) Coulomb Counting Method (CCM) [3] (2) Black-box Battery Model (BBM) [4] and (3) State-space Battery Model (SBM) [5]. (1) CCM is largely relying on the accuracy of the current measurement. In practical scenarios of uncertain disturbances, this open-loop estimation method usually gives bad results with an accumulation of measurement errors resulting from ammeters [6], [7]. The estimation errors are even higher when the working temperature is too high or too low. (2) For Black-box Battery Models, Artificial intelligence (AI)-based approaches, such as Artificial Neural Networks (ANN), Fuzzy Logic (FL), Support Vector machine (SVM), are often employed to formulate the mathematical models. The black-box battery model can provide a good SOC estimation based on the nonlinear relationships established from a given data set. Zenati et al. conducted experiments on battery aging in different operating conditions and combine impedance measurements with the fuzzy logic inference for estimation of either SOC or SOH [8]. Awadallah and Venkatesh developed a SOC estimation model using Adaptive Neuro-Fuzzy Inference Systems (ANFIS) and compared it to the traditional Coulomb counting method [9]. (3) State-space battery models based on Kalman filtering are popular in the estimation of battery SOC since it is an online close-loop algorithm. A state-space model has been established by Ting et al. through mathematical derivations to simulate the complex behaviors of a battery system [10]. The results indicate that there exist four state variables relevant to the battery model.

Thus, the different models discussed above mainly rely on electrical parameters. These models are imprecise because the measurement of electrical parameters in real time is hard and generate noisy data resulting from instantaneous variations occurring in complex driving conditions. In addition, these electrical parameters-based models usually require great computing efforts and do not perform well with the aging of batteries. Thus, a simpler estimation method is expected with acceptable accuracy guaranteed. A coupling between mechanical stress and chemical characteristics of lithium-ion pouch batteries is observed in the work of Cannarella and Arnold [11]. The correlation is concluded to be electrode volume change induced by chemical reactions in lithium-ion batteries. The irreversible stress produced in the charging process is related to the energy loss inside the batteries through charging-discharging cycles [12]. In addition, some recent studies [13], [14] have explored the means of establishing the fundamental and empirical relationship between the battery residual energy and the mechanical parameters such as the mechanical stress and strain when the battery is subjected to sudden compression or three-point bending tests. The studies were proved to be useful which compliments easier and accurate evaluation.

Therefore, to address these problems related to recycling purpose, this paper proposes the Stack stress-coupled-AI approach for analyzing the residual energy in the batteries. Experiments based on the static loading on the Li-ion batteries is firstly conducted and its capacity shall be measured. The current study aims to determine the relationship between stress and capacity quantitatively. Based on the measured experimental data, the AI approach of genetic programming (GP) is then applied to formulate a functional relationship of capacity as a function of design variables (stress, temperature, voltage). Conclusions are discussed in the end.

Section snippets

Research problem statement

An effective and efficient analysis of residual energy (remaining capacity) is an important problem for purpose of recycling of battery packs used in EVs. Finding of residual energy is related to SOH/SOC of the battery. The battery SOH is defined as the ratio of the current available full-charge capacity of to its original nominal capacity when it is freshly new while the SOC is the ratio of the residual capacity to the current full-charge capacity. For instance, a battery of 1 Ah nominal

Experimental details

Battery charge-discharge experimental tests are carried out to collect data for parameters such as capacity, stack stress, and other key parameters. The schematic computer layout of the experimental is shown in Fig. 3 Battery testing Equipment in laboratory (Fig. 4) can charge and discharge batteries and measure data.

Modeling method of genetic programming

Genetic programming (GP) is an artificial intelligence (AI) algorithm that can generate symbolic functional expressions based on only the given data [16]. The evolutionary system can mimic the behaviors of the targeted complicated system [17], [18]. A lot of researches related to modeling problems in energy storage system verifies the practicality of the modeling method of GP [19], [20], [21], [22], [23]. Compared to other modeling methods, GP has advantages from different aspects:

  • (1)

    GP works

Analysis of the GP model

Model No. 10 is selected as the best model among the 15 simulation runs of GP. The post model analysis is then carried out to determine the insights into experimental process.

Conclusions

A battery residual energy (remaining life detection) framework is proposed to provide a recycling strategy for spent batteries in EVs. Experiments are performed and AI method of GP was used to generate the mathematical model for predicting the cell capacity. The main results are shown as follows:

  • (1)

    GP model generates relationship with higher fitting accuracy for prediction of capacity as a function of real-time stress under varied applied initial stresses. Thus, the GP model can be used for

Declaration of Competing Interest

All Authors declares there is no conflict of interest.

Acknowledgements

This work was supported by the National Natural Science Foundation of China [grant numbers 51675196 and 51721092], the Program for HUST Academic Frontier Youth Team [grant number 2017QYTD04] and the Program for HUST Graduate Innovation and Entrepreneurship Fund[grant number 2019YGSCXCY037].

References (28)

  • Y.S. Lee et al.

    Forecasting energy consumption using a grey model improved by incorporating genetic programming

    Energy Convers. Manag.

    (2011)
  • B. Panda et al.

    Experimental and numerical modelling of mechanical properties of 3D printed honeycomb structures

    Measurement

    (2018)
  • J.H. Aylor et al.

    A battery state-of-charge indicator for electric wheelchairs

    IEEE Trans. Indust. Electron.

    (1992)
  • D. Yang et al.

    State-of-health estimation for the lithium-ion battery based on support vector regression

    Appl. Energy

    (2017)
  • Cited by (21)

    • Artificial intelligence driven hydrogen and battery technologies – A review

      2023, Fuel
      Citation Excerpt :

      A data processing paradigm influenced by the way biological nerve systems interpret data, such as the brain. ANN algorithm architecture simulates the functioning principles of the human brain in silico [101]. Each neuron carries only a portion of overall information, which is shared between neurons via their synaptic connections and sent by electrical pulses [102].

    • Current challenges and future opportunities toward recycling of spent lithium-ion batteries

      2022, Renewable and Sustainable Energy Reviews
      Citation Excerpt :

      In order to re-use the retired batteries, it is necessary to implement screening processes to ensure that the battery capacity, internal resistance and other critical parameters, meet the consistency requirements [248]. For this purpose, AI-based approaches such as generic programming have been employed to predict the cell capacity in retired batteries as a function of stress, temperature and voltage [249]. Self-organising map (SOM) neural networks were also used to sort retired batteries based on four parameters of capacity, internal resistance, voltage and temperature.

    View all citing articles on Scopus
    View full text