Weak Cell Analysis and Verification for EV System with Dynamic Load

Weak Cell Analysis and Verification for EV System with Dynamic Load

Dr. S. Kumar,

(Assistant Professor), EEE, JNTUH University College of Engineering Manthani

Y. Gouthami, M. Pavani, K. Tharun T. Vishnu Kumar Chary

B. Tech Student, EEE, JNTUH University College of Engineering Manthani

Abstract: Lithium-ion batteries (LIBs) are essential parts of electric vehicle (EV) systems, however internal resistance fluctuations, cell degeneration, and heat issues affect their dependability. Early detection of weak cells under dynamic load conditions is crucial to ensuring safety and extending battery life. This paper proposes a two-stage approach to identify weak cells during time-varying load operation. Thermal validation is used to locate the damaged cell within the battery string after voltage variation and threshold-based analysis are used to identify abnormal cell behaviour. Strip voltage response under dynamic loading isempirically validated and mathematically analysed. The proposed WAVES framework enhances BMS performance by providing a scalable and computationally efficient solution for real-time problem diagnosis in EV without requiring complex hardware or a significant amount of training data.

Key words: Battery Management System (BMS), Electric Vehicles (EVs), Internal Resistance (IR), Li ion battery pack (LIB), Neither Charging nor Discharging (NCND), Open Circuit Voltage (OCV), Passive Cell Balancing, Solid Electrolyte Interface (SEI), State of Charge (SoC)

1. INTRODUCTION

   The high energy density, extended lifespan, and high efficiency of lithium-ion batteries (LIBs) make them popular in contemporary energy storage applications [1]. Reliability and safety are crucial issues due to their extensive use in consumer devices, electric cars, and grid storage [2]. However, problems such internal short circuits, thermal runaway, and capacity deterioration have a big impact on operational stability [3]. Therefore, improving safety, performance, and lifetime requires early detection of such defects [4].

   Although they are useful for broad supervision, conventional monitoring techniques based on voltage, current, and temperature data are unable to precisely locate weak cells or identify early-stage problems [5]. Data-driven

   approaches, especially machine learning, have been extensively investigated for fault detection and State of Health estimate in order to overcome these constraints [6]. In order to detect weak cells more accurately, identify degradation patterns, and classify abnormal circumstances, machine learning models can examine historical data [7]. Neural networks, Random Forest, Support Vector Machines, and optimization techniques like Particle Swarm Optimization and Genetic Algorithm have demonstrated encouraging outcomes [8]-[9]. Nevertheless, these methods necessitate extensive labelled datasets and a variety of fault scenarios, which are hard to come by in real-world battery systems.

   The challenge of identifying localized weak cells is further complicated in series-parallel battery configurations, where labelling individual cell faults is not feasible. Although Battery Management Systems provide voltage and current monitoring along with cell balancing, they are limited by sparse temperature sensing and the inability to distinguish individual cells within parallel groups [10]. Advanced sensing techniques such as infrared thermography are constrained by cost and accessibility.

   Because weak cells have lower voltage and higher internal resistance, they put more strain on healthy cells by drawing extra charge, which can cause overheating and hastened degradation. Thermal runaway could occur if this is not addressed [11]-[12]. In order to provide precise and localized fault detection with little reliance on sensors and fewer computing limitations, an effective solution is needed so proposed WAVES under NCND condition [13].

   The rest of this document is structured as follows: Section II examines current developments and the body of extant literature; The problem formulation and mathematical analysis are presented in Section III; the suggested approach is described in Section IV;

2. LITERATURE REVIEW

   Lithium-ion battery (LIB) defect detection in electric vehicles (EVs) has made significant strides recently, but there are still a number of issues. Precise fault isolation is made possible by model-based techniques, such as switched designs that use comparable cell modelling. These techniques are effective, but they are not suitable for real-time applications due to their high implementation complexity and substantial processing overhead [14].

   Techniques based on data and machine learning have been widely used to improve the accuracy of fault detection. These methods enable automated diagnostics and take fault investigation from detection to quantification [15]. However, their scalability and real-time application are limited by their reliance on huge training datasets and high processing requirements, especially in dynamic and uncertain operational settings.

   Aging features are successfully incorporated into the diagnostic process by methods intended to handle both new and old battery cells [16]. Although it is restricted to errors linked to voltage, the weighted Euclidean distance approach

   [17] successfully detects voltage discrepancies. Although it has problems with scalability and flexibility, the diagnostic technique in [18] improves practical applicability. High accuracy is attained using machine learning approaches [19], but they demand substantial computer power and vast datasets. Automated fault analysis in charging stations is made possible by the method in [20]. However it has trouble with dynamic and unexpected fault patterns. Although segmented regression [21] is straightforward and efficient, it performs badly in a variety of operating environments. Although it requires more validation, a transferable diagnostic approach [22] enhances flexibility between cells. While [23] offers a thorough analysis but lacks innovative techniques, [24] facilitates practical diagnosis using driving data in real circumstances, and [25] detects tiny problems with higher granularity, increasing complexity.

   However, their restricted scalability and decreased efficacy in highly dynamic contexts frequently limit their performance. Voltage-based methods, such as curve analysis and distance-based methods, offer accurate detection of internal flaws and voltage irregularities. However, these approaches are usually restricted to particular fault types and fail to account for the multi-parameter interactions found in actual battery systems.

   To deal with changing battery behaviour and concept drift, statistical and adaptive methods, such as unsupervised learning techniques, have been developed. Although these techniques offer flexibility, they also increase computing complexity and make it more difficult to discern between

   noise and real fault events [26]. Similar to this, methods for identifying small irregularities and recreating malfunctioning sensor signals improve data dependability and diagnostic granularity but rely significantly on sensor accuracy and system setup [27]. Large training datasets are also needed. The technique in [28] is limited to voltage-related defects but uses partial voltage curves to detect internal short circuits with good accuracy. Although the method in [29] can identify problems with parallel battery packs, such as sensor and internal resistance problems, it is not scalable for bigger systems.

   The practical relevance of defect detection techniques has increased as a result of efforts to close the gap between laboratory diagnostics and practical applications. Scalability, data variability, and environmental adaption issues still exist, nevertheless. Although cloud-based diagnostic systems provide better scalability and centralized analysis, latency and data security issues are raised [30].

   Internal problems such micro short circuits can be found with high precision using sophisticated diagnostic methods based on electrode parameters, relaxation voltage, and incremental capacity analysis [31]. Despite their accuracy, these techniques are frequently restricted to particular fault circumstances and necessitate precise parameter estimation. Additionally, early thermal runaway detection methods improve safety but are difficult to integrate with current battery management systems (BMS). In general, real-time monitoring of critical parameters like voltage, current, and temperature is emphasized by contemporary BMS frameworks. However, because of the interconnectedness of cells and the nonlinear nature of fault propagation, weak cell identification continues to be a major difficulty. Conventional threshold-based techniques are frequently insufficient since they may produce false alarms and are unable to identify minor flaws.

   • A systematic two-stage diagnostic methodology is suggested that uses strip voltage for the first electrical examination and thermal imaging for validation under dynamic load circumstances. At a time multiple weak cells are identified.

   • MATLAB / Simulink are used to validate the suggested method, showing consistent patterns of voltage and heat behaviour.

   • According to experimental findings, aged cells with higher internal resistance show higher surface temperatures and sharper voltage drops, making it possible to identify weak cells.

   • Practical implementation in battery management systems is made possible by the method's avoidance

     of invasive measurements and expensive impedance-based diagnostic equipment.

3. METHODOLOGY

   Fig 1: Block diagram of the BMS

   For real-time weak cell detection in lithium-ion battery packs utilized in electric car applications, a dynamic load-based framework is suggested. Fig 1 explains the suggested method uses a controlled dynamic load to replicate actual operating conditions, in contrast to conventional approaches that function under Neither Charging Nor Discharging (NCND) conditions, when fault detection is limited due to minimal voltage change. This makes capacity mismatches and variations in internal resistance between cells more visible.

   Each parallel strip is observed by the architecture as a functional unit. Weak cells show more voltage drops and more heat production under dynamic loading because of higher internal resistance and lower charge retention. The system incorporates temperature sensors, voltage sensors, and a dynamic load module to continuously monitor thermal and electrical activity. Significant voltage reductions are correlated with aberrant temperature increases to identify weak cells, increasing detection accuracy and dependability.

   The contact resistances value is calculated as 0.00964m, hence R=0.00964m. However, the value of R can also be obtained from the manufacturer. Similarly, the values of VH and VF can also be user defined. The LIB configuration details will be provided by the manufacturer, thus M and N are also user defined [13].

   Table 1: Nickel Strip Specifications

   Dimension (mm)	Cell Spacing (mm)	Width (mm)	Dimension of square hole (mm)
   0.15x27x19.5	19.75	27	12.5×12.5

   The expected strip voltage in a lithium-ion battery strip with a weak cell is estimated using a straightforward yet efficient linear model. The method begins by identifying the total strip resistance R and the voltage of a healthy cell (VH) and a weak

   cell (VF), where VF is less than VH. These characteristics are then arranged into an input matrix X, where a single weak cell is positioned at different locations within a four-cell strip to indicate distinct fault states.

   In the matrix X, each row denotes a specific failure condition: Depending on where the weak cell is located, the first four columns in each row represent the voltages of individual cells, with values of either VH or VF.

   The known strip resistance R which takes into account the voltage drop brought on by resistive effects, is always found in the fifth column. By resolving the linear equation, the model ascertains the relationship mathematically:

   X · A = Y (1)

   where:

   • X is a (N+1)×(N+1) matrix consisting of voltage arrangements and resistance,

   • Y is a (N+1)×1 output vector representing expected strip voltages.

   • A is the unknown coefficient vector estimated using MATLABs backslash operator (A=X\Y), which computes the least-squares solution to the linear system.

   Avg = VF + (N 1)VH /N, d = VH avg/N (2)

   Fig 2: cells C11,C12 are weak

   When C11, C12 are weak:

   Where R1211,13-11 refers to the resistance path for the current contributed by cell C13, R11, R12 and R13 represent the IR of the cell C11 and C12 respectively.

   R13-12,13-11=R13+RP12-13+R12+RP11-12+R11 (3)

   Where R1412,14-11 refers to the resistance path for the current contributed by cell C14, R11, R12 and R14 represent the IR of the cell C11,C12 and C14 respectively. The cell C14 is the

   farthest from C11, C12 and feeds charge through path resistance.

   R14-12,14-11=R14+RP13-14+RP12-13+RP11-12

   +R11 (4)

   R13-12,13-11<R14-12,14-11(5) I13-12,13-11>I14-12,14-11 (6)

   For weak cell C:

   ,11 = ,11 (7)

   For weak cell C:

   ,12 = ,12 (8)

   Now both weak cells contribute to voltage drop:

   = (,11 + ,12) (9)

   1. Battery Pack Model: A battery pack made up of several lithium-ion cells connected in series is used to start the simulation. During operation, each cell generates voltage and temperature values.

      The extension model allows for the purposeful weakening of two or more cells in order to examine system behaviour.

   2. Dynamic Load Block: This involves applying a dynamic load to the battery pack. Dynamic load refers to a demand that varies over time, much like actual EV circumstances like braking, acceleration, and hill climbing. This makes it easier to see how each battery cell responds to actual driving circumstances.

   3. Voltage Measurement Unit: This device continually measures the voltage of every cell in the battery string. The analysis block receives these voltages.

      =1

      =

      Where;

      . + B.R.(10)

   4. Voltage Threshold Analysis: A predetermined threshold value is compared to each cell voltage. Abnormal or weak cell behaviour is indicated when a cell voltage falls below the

      M=no of cells connected in series N=no of cells connected in parallel Aj, B=unknown coefficients X.A=Y (11)

      Avg = VF + (N 1)VH /N (12)

      d = VH avg/ N (13)

      d=decremental setup

4. RESULTS AND DISCUSSION

The enhanced simulation model of the EV battery system used to more precisely identify weak cells is shown in Fig 2 by the Simulation Diagram . By including voltage threshold analysis and temperature monitoring for several cells under dynamic load conditions, this model expands on the base scenario.

Fig 3: Simulation diagram

threshold. Te extension model allows the system to simultaneously identify several weak cells.

1. Temperature Monitoring: Temperature of the cell is tracked once anomalous voltage behaviour is identified. Due to greater internal resistance, weak or malfunctioning cells typically exhibit higher temperatures. This stage aids in identifying the precise cell and verifying the error.

2. Output of fault detection: Temperature variation and voltage deviation results are the outputs of the system. These outputs are used to determine which battery pack cells are weak

3. Visualization of Results: The findings of the simulation reveal:

Cells falling below the threshold on the voltage graph. A temperature graph that displays elevated temperatures for weak cells.

According to the testing findings, the strip with the lowest voltage is the weakest part since lower voltage indicates greater internal resistance and degradation. According to the obtained data, the C63C64 strip records the greatest temperature among the cells while exhibiting relatively lower voltage values. C63 and C64 are the weakest cells in the pack, as evidenced by the association between voltage drop and temperature rise, which implies higher internal losses.

Voltage deviation and temperature variation helps us to detect the weak cell location.This two stage dectection makes the model more reliable for battery Management Systems(BMS) in EVs.

BASE CASE RESULTS

Table 2: Voltage Threshold Analysis

Table 3: Cell Temperature

PROPOSED RESULTS

Table 4: Voltage Threshold Analysis

20.01

Table 5: Cell Temperature

CONCLUSION

An effective technique for detecting weak cells utilizing strip voltage analysis and thermal validation is presented in this research. With the least amount of sensors and calculation, the WAVES method with dynamic load precisely locates the defective cell and determines the weakest strip. The findings verify that weak cells exhibit greater temperatures and lower voltages as a result of higher internal resistance. The technique's practical application is enhanced by its capacity to detect numerous weak cells at once and its extension to dynamic load circumstances. All things considered, it offers a dependable and affordable way to diagnose battery problems in EV systems.

FUTURE SCOPE:

Increase accuracy by incorporating SoC, SoH, and IR. For more practicality, use electrical or model-based techniques to lessen the dependence on thermal imaging. Incorporate safety features like automatic weak cell isolation and thermal runaway prevention.

REFERENCES:

1. W. Waag, S. Käbitz, and D. U. Sauer, Critical review of the methods for monitoring of lithium-ion batteries: The need for development of advanced state-of-charge estimation methods, in Journal of Power Sources, vol. 226, pp. 76-91, 2013

2. X. Hu, C. Zou, C. Zhang, and Y. Li, Technological developments in batteries: A survey on modern advances, IEEE Transactions on Industrial Informatics, vol. 15, no. 4, pp. 2330-2345, 2019.

3. Y. Xing, E. W. M. Ma, K. L. Tsui, and M. Pecht, Battery management systems in electric and hybrid vehicles, Energies, vol. 4, no. 11, pp. 1840 1857, 2011.

4. J. Zhang and J. Lee, A review on prognostics and health monitoring of Li-ion battery Journal of Power Sources, vol. 196, no. 15, pp. 6007-6014, 2011.

5. M. Berecibar, I. Gandiaga, I. Villarreal, N. Omar, J. Van Mierlo, and P. Van den Bossche Critical review of state-of-health estimation methods of Li-ion batteries for real applications, Renewable and Sustainable Energy Reviews, vol. 56, pp. 572-587, 2016.

6. A. Farmann, W. Waag, J. Marongiu, and D. U. Sauer, Critical review of on-board capacity estimation techniques for lithium-ion batteries in electric and hybrid electric vehicles, Journal of Power Sources, vol. 281, pp. 114 130, 2015.

7. G.L. Plett, Extended Kalman filtering for battery management systems of LIB-based HEV battery packsPart 1. Background,

   Journal of Power Sources, vol. 134, no. 2, pp. 252-261, 2004

8. C. Zhang, F. Li, H. Wang, and J. Wang, State-of-health estimation for Li ion batteries by combining the incremental capacity analysis method

   with machine learning, Journal of Energy Storage, vol. 25, p. 100922, 2019.

9. H. He, R. Xiong, J. Fan, "A method for state of charge and state of health estimation for lithium-ion batteries using a dual Kalman filter," Energy, vol. 39, no. 1, pp. 310318, 2012.

10. M. Elmahallawy, T. Elfouly, A. Alouani and A. M. Massoud, "A Comprehensive Review of Lithium-Ion Batteries Modelling, and State of Health and Remaining Useful Lifetime Prediction," in IEEE Access, vol. 10, pp. 119040119070,2022,doi:10.1109/ACCESS.2022.3221137.

11. V. Jha and B. Krishnamurthy, Modelling the SEI layer formation and its growth in lithium-ion batteries (LIB) during charge-discharge cycling, Ionics, vol. 28, pp. 36613670, 2022.

12. G. Luo, Y. Zhang, and A. Tang, Capacity degradation and aging mechanisms evolution of lithium-ion batteries under different operation conditions, Energies, vol. 16, no. 10, p. 4232, 2023.

13. Devang P. Kubitkar1, Twinkle Pattnaik2,(Student Member, IEEE), Makarand S. Ballal 3, (Senior Member, IEEE)

14. Mitra, Desham, and Siddhartha Mukhopadhyay, Detection and isolation of faults in a lithium-ion battery pack using a switched architecture of equivalent cell diagnosers, Journal of Energy Storage , vol. 90, 29 Apr. 2024, pp. 111811111811.

15. Li, Jinwen, Yunhong Che, Kai Zhang, Hongao Liu, Yi Zhuang, Congzhi Liu, and Xiaosong Hu, Efficient Battery Fault Monitoring in Electric Vehicles: Advancing from Detection to Quantification, Energy, vol. 313, 8 Dec. 2024, p. 134150.

16. Sepasiahooyi, Sara, and Farzaneh Abdollahi, Fault Detection of New and Aged Lithium-Ion Battery Cells in Electric Vehicles, Green Energy and Intelligent Transportation, vol. 3, no. 3, June 2024, p. 100165.

17. Liu, Qiquan, Jian Ma, Xuan Zhao, Kai Zhang, Kang Xiangli, and Dean Meng, A novel method for fault diagnosis and type identification of cell voltage inconsistency in electric vehicles using weighted Euclidean distance evaluation and statistical analysis, Energy, vol. 293, 9 Feb. 2024, pp. 130575130575

18. Zhao, Jingyuan, Xuning Feng, Manh-Kien Tran, Michael Fowler, Minggao Ouyang, and Andrew F. Burke, Battery Safety: Fault Diagnosis from Laboratory to Real World, Journal of Power Sources,vol.598,1Apr.2024, pp. 234111234111.

19. Zhao, Jingyuan, Xuning Feng, Manh-Kien Tran, Michael Fowler, Minggao Ouyang, and Andrew F. Burke, Fault Detection and Isolation in EV Battery Using Machine Learning Approach., 2022 8th International Conference on Advanced Computing and Communication Systems (ICACCS),14 Mar. 2024, pp. 729734.

20. A. Shufian, N. Hannan, M. M. H. Emon, S. Kabir, S. A. Fattah, and M.

    S. Hossain, Automatic fault detection and analysis of electric vehicle charging station by machine learning, in Proc. IEEE Region 10 Symp. (TENSYMP), 2024, pp. 16.

21. M. B. Kaleem, Y. Zhou, F. Jiang, Z. Liu, and H. Li, Fault detection for Li-ion batteries of electric vehicles with segmented regression method, Sci. Rep., vol. 14, no. 1, pp. Article 82960, Apr. 2024. Accessed on: May 2, 2025, DOI: 10.1038/s41598-024-82960-0.

22. A. Singh, A. Lodge, Y. Li, W. D. Widanage, and A. Barai, A new method to perform Lithium-ion battery pack fault diagnostics Part 2: Algorithm performance in real-world scenarios and cell-to-cell transferability, Energy Rep., vol. 11, pp. 304315, Nov. 2023.

    Accessed on: May 2, 2025, DOI: 10.1016/j.egyr.2023.11.057.

23. J. Hong, F. Liang, J. Yang, and S. Du, An exhaustive review of battery faults and diagnostic techniques for real-world electric vehicle safety,

    J. 14 Energy Storage, vol. 99, Art. no. 113234, Mar. 2024. Accessed on:

    May 2, 2025, DOI: 10.1016/j.est.2024.113234.

24. S. Park, B. Kang, D. Yu, M. Jeong, and Y. Hong, Fault diagnosis for electric vehicle battery pack interconnection system using real-world driving data, IEEE Trans. Ind. Electron., pp. 19, Jan. 2025. Accessed on:May2,2025,OI:10.1109/TIE.2024.3522461.

25. L. Ma, B. Duan, C. Zhang, Y. Kang, C. Li, and K. Liu, Detection and differentiation of multiple types of minor anomalies in battery packs, Energy, vol. 2025, Art. no. 135557, Apr. 2025. Accessed on:May2,2025,DOI:10.1016/j.energy.2025.135557.

26. D. H. Avinash and A. Rammohan, Integrating level shift anomaly detection for fault diagnosis of battery management system for Lithium-Ion batteries, IEEE Access, vol. 12, pp. 116071116084, 2024. Accessed on: May 2, 2025, DOI: 10.1109/ACCESS.2024.3445955.

27. S. Wang, Z. Wang, J. Pan, Z. Zhang, and X. Cheng, Data-driven fault tracing of lithium-ion batteries in electric vehicles, IEEE Trans. Power Electron., vol. 39, no. 12, pp. 1660916621, Dec. 2024. Accessed on:

    May 2, 2025, DOI: 10.1109/TPEL.2024.3441572.

28. D. Qiao, X. Wei, B. Jiang, W. Fan, H. Gong, X. Lai, Y. Zheng, and H. Dai, Data-driven fault diagnosis of internal short circuit for series-connected battery packs using partial voltage curves, IEEE Trans. Ind. Informat., vol. 20, no. 4, pp. 67516761, Apr. 2024. Accessed on: May

    2, 2025, DOI: 10.1109/TII.2024.3353872.

29. H. Jin, Z. Zhang, S. X. Ding, Z. Gao, Y. Wang, and Z. Zuo, Fault diagnosis for parallel lithium-ion battery packs with main current sensor fault and internal resistance fault, IEEE Trans. Instrum. Meas., vol. 73, pp. 110, 2024. Accessed on: May 2, 2025, DOI: 10.1109/TIM.2024.3400362.

30. A. Tang, Z. Wu, Y. Xu, K. Liu, and Q. Yu, Cloud-based Li-ion battery anomaly detection, localization and classification, IEEE Trans. Ind. Informat., pp. 19, 2024. Accessed on: May 2, 2025, DOI: 10.1109/TII.2024.3514131.

31. Y. Xu, X. Ge, R. Guo, C. Hu, and W. Shen, Electrode-parameter-based fault diagnosis and capacity estimation for lithium-ion batteries in electric vehicles, IEEE Trans. Ind. Electron., pp. 111, 2024. Accessed on: May 2, 2025, DOI: 10.1109/TIE.2024.3447749