Elsevier

Energy

Volume 214, 1 January 2021, 118866
Energy

An Intelligent Fault Diagnosis Method for Lithium Battery Systems Based on Grid Search Support Vector Machine

https://doi.org/10.1016/j.energy.2020.118866Get rights and content

Highlights

  • A novel method of fault diagnosis based on SVM for the battery pack is proposed.

  • The amplitude frequency characteristics of the vibration noise can be gained.

  • A new filtering method based on white noise characteristics is proposed.

  • The modified covariance matrix is firstly proposed as the condition indicator.

  • Experimental results verify the feasibility and advantages of the proposed method.

Abstract

For the safe operation of the electric vehicle, it is critical to quickly detect the safety state and accurately identify the fault degree in battery packs. This article proposes a novel intelligent fault diagnosis method for Lithium-ion batteries based on the support vector machine, which can identify the fault state and degree timely and efficiently. Due to the noise signal’s existence, firstly, the discrete cosine filtering method is adopted, and the truncated frequency is optimized based on the characteristic of white noise to achieve reasonable denoising. Secondly, since the covariance matrix (CM) of filtered data is sensitive to the current fluctuation, a modified covariance matrix (MCM) is proposed to reduce the influence of current variation on the condition indicators. Thirdly, to ensure the accuracy and robustness of Support Vector Machine (SVM), the grid search method is proposed to optimize the kernel function parameter and penalty factor. Finally, the MCM and CM are respectively introduced into the model as the condition indicators, and the results show that the former has high accuracy and timeliness. In summary, the proposed intelligent fault diagnosis method is feasible. It provides the theoretical basis for future fault hierarchical management strategy of the battery system.

Introduction

In the transportation field, electric vehicles (Evs) have been widely recognized as an imperative option for environmental issues, which has attracted full attention [[1], [2], [3]]. The battery pack is the critical component in Evs that can provide high energy density to meet the vehicle driving range requirements and offer high power density to fit the acceleration and hill-climbing scenarios. Due to the limitation of the single cell’s voltage and capacity, the battery pack is typically composed of many cells in various series-parallel modes. In recent years, a series of traffic accidents have been reported many times in the complex operating environment of Evs, exposing lots of security risks of the battery pack, which should hinder the further promotion and application of Evs. So the safety of the battery pack has become a crucial issue that should be urgently investigated [[4], [5], [6]].

Ensuring the safe operation of Evs has become the core task for the battery management system (BMS). The BMS can predict the current working state of the battery by monitoring the voltage, current and temperature to maintain the greater security diagnosis accuracy [7,8]. Till now, many efforts have been devoted to developing various reliable BMS to improve the efficiency of battery condition identification [[9], [10], [11]] and fault diagnosis [[12], [13], [14]]. However, most of the researches focus on the estimation of the state of charge (SOC) [9,10], state of health (SOH) [15,16], state of energy [17,18] and remaining useful life (RUL) [19] of the batteries. Researches on the reliability and application technology for battery pack are relatively scarce and not well-rounded.

In the practical application of power battery, only three parameters, voltage, current, and temperature, can be directly collected. However, the current data acquisition is difficult, especially the parallel branch’s value, which is affected by electromagnetic interference, so many battery fault diagnosis and prognostics approach based on the variation of voltage and temperature parameters have been proposed. Simultaneously, according to different data processing methods, research methods can be classified into three categories: thermal-model-based, voltage-model-based, and data-driven-based methods.

The thermal-model-based methods focus on detecting the cell surface temperature or predicting the internal value trend to determine the current working state. As the experimental process is relatively direct and smooth, it has been widely adopted. Viswanathan et al. [20] measured the entropy changes with the various cathode and anode materials using an electrochemical thermodynamic measurement system and presented the reversible heat generation rate was found to be a significant portion of the total heat generation rate. Fredrik et al. [21] reported that the cells’ heat release rate would increase with the rise of SOC, but the correlation between total heat release and SOC decreases gradually. Vantuch et al. [22] proposed that the changing trend of the battery surface temperature was measured by infrared thermal imager in long term monitoring and fault detection, but the accuracy and robustness decreased. These methods are simple and easy to operate, which can be identifying the rent working state of the power battery to improve the fault identification rate. However, those mentioned above have strict requirements on environmental conditions and instruments, which cannot be applied to real applications. To combine the internal reaction degree of the cell with the external parameters’ change, Panchal S et al. [23] presented a mathematical model to predict the transient temperature and voltage distributions at different discharging rates. Huo et al. [24] studied a 3D model of a lithium-ion battery’s thermal performance to analyze the dynamic thermal behavior by utilizing various drive cycles. The electrochemical model of power battery can accurately capture the intensity of the internal chemical reaction in different loading ratios, which is of great significance for fault diagnosis. Using the battery thermal model and the equivalent circuit model (ECM), Dey. S. et al. [25,26] proposed an identification scheme based on the Luenberger observer to detect and isolate three kinds of thermal faults: internal thermal resistance fault, convective cooling resistance fault, and thermal runaway. The cell surface temperature feature can represent the current internal reaction degree, but cell temperature can be affected by factors such as location, internal resistance, cooling degree of the battery packs, etc. Meanwhile, temperature change has specific hysteresis characteristics: the heat will pass through a period from the inside to the surface, leading to the low timeliness of the thermal-model-based methods.

Compared with temperature signals, voltage signals are more comfortable to acquire and more time-efficient, which many researchers favored. Therefore, more publications are based on the voltage signal to diagnose the faults. The voltage-model-based methods usually include four key steps. The first step is to establish a model that can represent the dynamic characteristics of the cells. Secondly, the parameters of the model are identified by intelligent algorithms, such as the Kalman filtering, particle filtering methods, and so on [[27], [28], [29]]. Thirdly, the parameters of cell fault characteristics are determined, and the corresponding threshold values are set [30]. Fourthly, the measured signal is compared with the threshold to determine whether the fault has occurred. Ouyang et al. [31] studied an internal short circuit detection method using recursive least squares to estimate the mean-difference model parameters, and the parameters of all cells in the pack are compared with the set threshold to detect the fault state. To avoid the occurrence of faults and accelerated cell aging, Zheng et al. [32] investigated a differential voltage analysis method to identify the battery’s current working state accurately. Xiong et al. [6] proposed a two-step equivalent circuit battery model that can provide a rapid and precise fault diagnosis result by separating voltage variation caused by polarization and a significant voltage drop caused by a large current. The accuracy of these diagnostic methods are relatively high but mainly depends on the prediction accuracy of model parameters, and the complexity of calculation increases.

Recently, the data-driven-based methods have been widely applied in the battery fault detection field due to the healthy development momentum of big data processing technology. Zhao et al. [33] proposed a machine-learning-based fault detection method to detect abnormal changes in cells efficiently and accurately. Xia et al. [4] presented a fault detection method for short circuits based on the correlating efficiency of voltage curves, which has strong robustness and anti-jamming ability. Yao et al. [34] compared three different entropy algorithms to identify the connecting faults, which achieved a relatively small amount of calculation. Wang et al. [35] proposed a fault detection method based on a modified Shannon entropy difference of each cell voltage to determine the location and time of the fault. Based on the data-driven principle, Zhao et al. [36] and Yao et al. [37] proposed a new fault diagnosis method using a recurrent neural network and convolution neural network with the dynamic characteristics of lithium-ion power battery. These methods do not need to build a battery model and complex parameter identification, which has the advantages of high efficiency and reliability.

This paper mainly studies the intelligent diagnosis of the connection fault of a series battery pack. The fault principle has been described in detail in Ref. [34]. Due to the complex non-linear time-varying system with the absolute inconsistency of the batteries, it is difficult for the conventional methods to diagnose the cell parameters’ abnormal change and predict the time of faults occurring in the initial stage. Meanwhile, lithium-ion batteries can be affected by road conditions, driving habits, environmental temperature, and other factors making it challenging to provide a real-time, accurate, and timely diagnosis in real operated Evs. Therefore, to bridge these drawbacks, the condition indicators based on the voltage data are mined and applied to support vector machine (SVM) to diagnose the potential abnormality state and classify the severity of fault accurately. Specifically, four parts are made in this article. Firstly, an optimal filtering method is adopted using a discrete cosine filter method based on white noise characteristics. Secondly, the covariance matrix (CM) of filtered data is analyzed, and a new condition indicator is determined. Thirdly, the grid search method is adopted to optimize the kernel function parameter and penalty factor to ensure the model’s accuracy and robustness. Finally, through the training of SVM parameters by condition indicators, the method can effectively realize the intelligent fault diagnosis. Simultaneously, the MCM’s efficiency and engineering practicability are verified from the identification rate and diagnostic accuracy. The detailed test information of this paper is shown in Fig. 1.

The rest of this paper is summarized as follows. The test workbench and cell are described in Section 2. The proposed fault diagnosis method for a series battery pack is presented in Section 3. The condition indicators and the optimal filtering method based on white noise characteristics are both described in the following section. The experimental results will be discussed in Section 5, and finally, in Section 6, a summary of findings and the conclusion will be presented.

Section snippets

Test platform

The test platform shown in Fig. 2 is to implement the battery pack fault diagnosis, which contains a battery test instrument (Digtron Battery Test System: BTS-600), a vibrating test bench, an information collector, several voltage sensors and the data processor. The battery tester, which integrates the power supply, electric load, signal acquisition, and transmission, is used for battery experiments to mainly simulate the charge and discharge mode in the electric vehicle operation. The

Discrete cosine transform

With the development of technology, the Discrete Cosine Transform (DCT) has the property of the orthogonal transform, which is widely used in signal and image processing [38]. This method can quickly transform data from the time domain to the frequency domain, attempt to de-correlate the collection data, and enable each transform coefficient to encode independently after de-correlation. On the basis of preserving the authenticity of the original image to the greatest extent, it can improve the

Preliminary analysis

The collected data usually includes normal operation data signal, noise signal, and various types of fault information, which has the characteristics of high time-varying, nonlinear, and complexity. According to the principle of DCT and IDCT, this section describes the noise filtering process of the collected data in detail.

As shown in Fig. 4, in real operated Evs, due to the existence of noise, when the current is constant, many high-frequency tiny burrs appear on the voltage curve, which

SVM parameter selection

Based on the above description of the SVM principle, the penalty factor C and kernel function parameter σ are two variables that determine the accuracy of the fault classification algorithm. If the C value is too large, the classifier is easy to produce over-fitting, which affects the algorithm’s generalization ability; if the C value is too small, the classifier will not care about the identification error, resulting in low classification accuracy. While the parameter σ is an essential

Conclusions

In this study, an intelligent fault diagnosis method based on data-driven is proposed for the lithium-ion battery system. Accurate and reliable experimental voltage data is essential to determine the current working state of battery packs. However, due to the noise interference’s real-time existence, it is challenging to identify the fault state accurately in real-time. Therefore, a novel implementation of the DCT denoising technique is firstly presented for the actual voltage value, which

Credit author statement

Lei Yao: CRediT roles: Methodology; Data Curation; Writing-Original Draft; Formal analysis. Zhanpeng Fang: CRediT roles: Writing-Review & Editing; Funding acquisition; Methodology; Project administration; Resources; Supervision. Yanqiu Xiao: CRediT roles: Writing-Review & Editing; Formal analysis; Investigation; Methodology. Junjian Hou: CRediT roles: Investigation; Methodology. Zhijun Fu: CRediT roles: Formal analysis.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

The authors are grateful to the financial support of National Natural Science Foundation of China (No. 51805491), and the Key Scientific Research Projects of Higher Education Institutions in Henan Province (No. 18A460035) and the Doctor Research Foundation of Zhengzhou University of Light Industry (No. 2016BSJJ014).

References (39)

Cited by (0)

View full text