Automatic and Versatile Test Bench for Data Collection on Battery Cells

Received: 26 March 2025 Revised: 13 April 2025 Accepted: 28 April 2025 Published: 30 April 2025 Citation: Marsal, E.; Martínez, N.; Pérez Vega-Leal, A.; Barrero, F.; Hamdan, M.; Satué, M.G. Automatic and Versatile Test Bench for Data Collection on Battery Cells. Energies 2025,18, 2304. https://doi.org/ 10.3390/en18092304 Copyright: © 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/ licenses/by/4.0/). Article Automatic and Versatile Test Bench for Data Collection on Battery Cells Esteban Marsal 1,† , Nicolás Martínez 1,† , Alfredo Pérez Vega-Leal 2,† , Federico Barrero 2,*,† , Mohamad Hamdan 1,† and Manuel G. Satué 1,† 1Systems Engineering and Automation Department, University of Seville, 41092 Seville, Spain; [email protected] (E.M.); [email protected] (N.M.); [email protected] (M.G.S.) 2Electronic Engineering Department, University of Seville, 41092 Seville, Spain *Correspondence: fbarrer[email protected] †Current address: Escuela Técnica Superior de Ingeniería, University of Seville, Camino de los Descubrimientos s/n, 41092 Seville, Spain. Abstract: Rechargeable batteries are a key component of sustainable future systems, as their performance directly affects energy efficiency, maintenance costs, and system reliability. Assessing performance requires evaluating parameters such as the state of health (SoH) of the battery, which necessitates developing a system capable of efficiently gathering large amounts of data. This article presents a safe, simple, versatile, and automated system designed to test and characterize various types of battery cells. The system is conceived as a practical tool capable of automatically collecting the required data for analysis, thus enabling the determination of the performance parameters of a battery cell. The proposed system incorporates an innovative approach based on the concatenation of charge/discharge data, allowing for a more reliable evaluation of battery performance. Experimental tests show the interest and performance behavior of the proposed system. Keywords: LiFePO 4 ; lithium-ion; battery cycler; battery energy management; battery cyclic tester 1. Introduction The transition from fossil fuels to natural energy sources and the growing global demand for sustainable energy solutions have led battery storage systems to gain research and industrial prominence [ 1 ]. The variability of renewable energy systems and their unpredictable performance, with a high dependence on weather conditions, make battery storage systems essential to ensure the stability of modern electrical grids [ 2 ]. Among the available storage technologies, lithium ion batteries improve the overall utilization and useful life of renewable energy due to their high energy density and efficiency, as well as their low self-discharge rate [ 3 ]. Their inherent characteristics, coupled with advances in power management strategies, have solidified their reputation as a reliable and costeffective energy storage solution [ 4 ]. However, despite their benefits, these batteries degrade over time because of factors such as loss of active materials or a reduction in lithium inventory [ 5 , 6 ]. The degradation mechanisms affect their capacity and life span, and therefore new developments in battery design or management strategies must be developed to optimize battery cell performance and longevity [7]. A typical energy storage system consists of multiple battery cells that, despite manufacturers’ standardization efforts, may still operate under different conditions. This is Energies 2025,18, 2304 https://doi.org/10.3390/en18092304 Energies 2025,18, 2304 2 of 16 the case in electric batteries based on lithium-ion battery cells, where varying effects depending on their application appear during continuous charge and discharge cycles [ 8 ]. To minimize these inconsistencies, it is essential to monitor and control the charging and discharging of each cell by measuring or estimating key variables such as voltage, current, and temperature. Comparing these variations with the overall performance of the system enables a more efficient energy management, ensuring reliable energy utilization [9]. The state of health SoH is the most critical indicator of battery degradation [ 10 ]. It can be estimated through continuous monitoring of the voltage, current and temperature using different alternatives as summarized in [ 11 ]. One of the simplest approaches is direct measurement, which assesses SoH on the basis of capacity transfer during charge and discharge cycles or by calculating the resistance from voltage drops. This method is straightforward, easy to implement, and enables real-time monitoring without the need for complex models or predictive algorithms [ 12 ]. Considering the importance of data availability and quality in SoH estimation, all methods for SoH estimation benefit from extensive data sets that capture various battery charge and discharge scenarios. Reliable data collection, covering varying operating conditions, ensures accurate SoH estimates. A versatile and automated test bench can effectively tackle this challenge by replicating charging and discharging cycles while enabling precise control and measurement of electric battery performance. Moreover, it enhances safety and extends service life through continuous monitoring and early failure detection, including overvoltage, undervoltage, and high-temperature measurements. Real-time data acquisition and logging provide deeper insights into performance and degradation factors, contributing to more effective battery management strategies [ 13 ]. However, implementing such a system poses several challenges, particularly in maintaining precise control over parameters such as voltage or current [14]. Many battery testing benches already exist and have been presented in different recent research works. For example, in [ 3 ], a trial system is proposed with the capacity to monitor the critical parameters of the battery cells. The proposal provides automated charge, discharge, and relaxation cycles, although it offers a limited number of configurations. It is based on open-source control software and is cost-effective, but the system is customized with a particular microcontroller, that lacks versatility because it is not easy to reproduce the proposal and it cannot test large battery banks or handle high current and voltage. A different test bench for specific aerospace applications is detailed in [ 8 ], where the battery cell is reproduced with a hardware in the loop technique. The proposal is not versatile due to the difficulty of reproducing the initial complex configurations of the system. The authors also report bugs in the software provided. In [ 3 ], a test bench that includes a set of procedures to investigate the performance of customized battery cells is presented, where a LabVIEW interface allows test selection, voltage and output control, real-time monitoring, and cutoff adjustment, supporting safety operation of the system. However, due to the programmable resistive load used, only a limited number of tests can be performed. Then, the versatility of the test bench provided is somehow doubtful. Finally, particular test benches are detailed in [ 15 , 16 ], where the works focus on the variable temperature control of individual cells in a battery pack and on the failure mechanisms that govern the durability of lithium-ion pouch cells under vibration. Both works present specific and customized test benches, with limited versatility. This article presents an automated, controlled data acquisition system for testing lithium-ion batteries. The system incorporates an innovative approach based on the concatenation of charge/discharge data, enabling reliable evaluation of battery performance. It supports fully automated charge, discharge, and relaxation cycles, offering a high degree of configurability and versatility. The platform is easy to replicate, applicable to both individ- Energies 2025,18, 2304 3 of 16 ual battery cells and battery packs, and is not restricted to any specific application. Battery functionality is validated through continuous monitoring and recording of voltage, current, and temperature during charge-discharge cycles. This allows for accurate performance evaluation without the need for complex modeling. Users can define voltage and current limits, program automatic cycling routines, and access the collected data. Data is stored in a common format and hosted in an open-access repository, thereby contributing to ongoing research and development in the field. Overall, the proposed system provides an efficient and adaptable solution for battery monitoring and management, representing a valuable tool for future research and practical applications. The paper is structured as follows. Section 2summarizes the fundamentals of how a lithium ion battery normally works, describing the charge and discharge processes. Section 3describes the proposed battery test bench, detailing the designed hardware and application software. Section 4describes some experimental procedures performed to validate the system, including the realization of charge-discharge cycles and the analysis of the collected data for the estimation of the SoH parameter. Conclusions are summarized in the last section. 2. Basics of Lithium-Ion Battery Degradation and Operation SoH is defined as the ratio between the maximum capacity achieved during a charge or discharge cycle (Co) and the initial capacity of the battery (Ci), typically expressed as a percentage: SoH (%) = (Co/Ci) × 100. The development of SoH estimation methods for lithium-ion batteries has undergone significant transformation over the years, see Figure 1, driven by advances in battery technology, computational techniques, and data availability [17]. Although a common used approach is the direct and empirical method (previously referred to as direct measurement technique in the Introduction section), alternative techniques such as coulomb counting [ 18 ], internal resistance monitoring [ 19 ], and electrochemical impedance spectroscopy (EIS) [ 20 ] are used in laboratory environments as well. These methods also offered high accuracy under controlled conditions by directly measuring capacity fade or increases in internal resistance, both of which are indicative of battery aging [ 19 ]. However, despite their simplicity and physical interpretability, they proved impractical for real-time or field applications. For example, Coulomb counting requires full charge–discharge cycles and is prone to integration errors, while internal resistance is sensitive to temperature and load variations [ 21 , 22 ]. EIS, on the other hand, requires specialized equipment and stable environments. As a result, these methods were mainly limited to offline diagnostic [23]. Subsequently, the physics-based modeling technique was introduced to address the growing need for real-time SoH estimation in increasingly complex battery systems such as those found in electric vehicles and portable electronics [ 24 ]. These methods included electrochemical models and equivalent circuit models (ECM), which describe battery behavior using electrical analogs [ 25 ]. ECM became especially popular due to their simplicity and ease of implementation. Tracking changes in circuit parameters over time makes possible to estimate SoH, as stated in [ 21 ]. ECM provided more detailed information on internal degradation mechanisms, but they posed significant computational challenges and required precise parameter identification [26]. To improve robustness and adaptability, researchers then turned to stochastic estimation and filtering techniques, such as Kalman filters (extended, unscented, and others) and particle filters [ 17 , 27 ]. These methods enabled real-time estimation of battery states, including SoH, even in the presence of noisy and incomplete sensor data [ 28 ]. Stochastic filters adapt to varying operational conditions, manage measurement noise, and continuously update model parameters [ 29 ]. However, its effectiveness heavily depends on the accuracy Energies 2025,18, 2304 4 of 16 of the underlying battery model and can entail high computational costs, particularly for nonlinear or high-dimensional systems [25]. Some researchers have explored the interest of so-called curve analysis techniques, including incremental capacity and differential voltage analysis methods (ICA and DVA, respectively) [ 30 ]. These techniques analyze voltage and capacity curves, particularly their time derivatives, to detect shifts related to degradation phenomena, such as loss of lithium inventory or active material. ICA and DVA provide valuable information on the internal condition of a battery without requiring full discharges. However, they are sensitive to noise and demand high-resolution data, limiting their practicality for real-time applications [17]. An interesting approach comes with the data-driven revolution technique, marked by the rise of machine learning (ML) methods. Algorithms such as support vector machines, random forests, artificial neural networks, deep learning models such as long short-term memory or gated recurrent units are also applied to large-scale battery datasets [ 31 ]. These approaches are capable of modeling complex and nonlinear relationships between input signals (voltage, current, temperature) and SoH, achieving high predictive accuracy and enabling real-time integration within battery management systems [ 18 ]. However, they require large and diverse data sets for effective training and often lack physical interpretability [ 32 ]. Issues such as overfitting and limited generalization remain today as ongoing challenges [ 33 ]. Ongoing challenges have also shifted toward hybrid and physicsguided approaches, which aim to combine the advantages of model-based and data-driven methods. These include physics-informed machine learning [ 34 ], multimodel ensemble learning [ 28 ], and hybrid filtering techniques [ 27 ]. Likewise, combining Kalman filters with ML models yields robust and adaptive estimation systems [ 24 ]. Although many of these methods are still under development, they hold significant promise for deployment in real-world applications, offering a compelling balance of accuracy, interpretability, and computational efficiency [ 22 ]. The pros and cons summary of the SoH estimation methods is detailed in Table 1. Battery SoH Estimation Methods Direct and Empirical Methods Coulomb Counting Internal Resistance Monitoring Electrochemical Impedance Spectroscopy Physics-Based Modeling Electrochemical Model Equivalent Circuit Model Stochastic Estimation and Filters Kalman Filter Particle Filter Curve Analysis Techniques Incremental Capacity Analysis Differential Voltage Analysis Peak Tracking Data-Driven Methods Support Vector Machines Random Forest Artificial Neural Networks Gradient Boosting Machines Hybrid and Physics-Guided Approaches Physics-Informed Neural Networks Kalman Filters + Machine Learning Multi-Model Ensemble Learning Transfer Learning / Domain Adaptation Figure 1. Classification of main SoH estimation techniques. There are several methods for charging and discharging lithium-ion battery cells, and choosing the right one greatly affects performance, lifespan, and safety. According to [ 35 ], three common charging methods are constant current-constant voltage (CC-CV), constant loss-constant voltage (CL-CV), and constant power-constant voltage (CP-CV). Among these, the CC-CV method is widely used due to its simplicity, reliability, and efficiency [ 36 ]. It helps reduce lithium plating and thermal problems, improves capacity retention, and allows accurate monitoring of the SoH of the battery. Energies 2025,18, 2304 5 of 16 The CC-CV method starts with a CC phase, where the battery charges at a fixed current until it reaches a set voltage. Then, it switches to a CV phase, where a constant voltage is applied while the current gradually decreases [ 37 ]. In comparison, the CL-CV method adjusts current based on battery impedance to control heat generation, while the CP-CV method keeps power constant before switching to CV mode. Table 1. Summary of pros and cons of SoH estimation methods. Category Representative Methods Pros Cons Direct and empirical Coulomb counting, internal resistance, electrochemical impedance spectroscopy Simple implementation; Physically interpretable; High accuracy under lab conditions Requires controlled environments; Poor performance for online estimation; Sensitive to noise and aging Physics-based models Electrochemical model, equivalent circuit model Mechanistic insight; Captures degradation mechanisms; Enables predictive modeling High computational cost; Difficult parameter identification Stochastic filters Kalman filter, particle filter Real-time estimation; Handles noisy/incomplete data; Adaptable to system changes Model-dependent; Computationally intensive; Sensitive to initialization Curve Analysis Techniques Incremental capacity analysis, differential voltage analysis, peak tracking Non-invasive; Detects specific degradation modes; Useful for cell diagnostics Requires high-resolution data; Sensitive to noise; Not suitable for real-time use Machine Learning Support vector machines, random forest, artificial neural networks, gradient boosting machines Learns complex nonlinear patterns; Suitable for real-time applications; Can integrate into BMS Needs large, labeled datasets; Risk of overfitting; Often lacks physical interpretability Hybrid & Physics-Guided Physics-informed neural networks, Kalman filters + Machine learning, multi-model ensemble learning, transfer learning/domain adaptation Combines data and physics; High accuracy and generalization; Adaptive and robust Complex to design and tune; Synchronization of models is challenging; Often still in research phase In our case, a CC-CV charging method is applied with an initial charging stage at CC. During this initial stage, the battery voltage increases as it accepts charge. When the battery reaches a predefined voltage value, the charging continues using a CV, causing a gradual reduction in the charging current as the battery approaches full capacity. This process is illustrated in Figure 2. The procedure ends when the current reaches a certain lower threshold, known as the minimum charging current point, which is set to 0.5 A. Figure 2. CC-CV method charging process: Voltage and current as a function of time. The discharging process consists of a continuous discharge at a predefined current. During discharge, the voltage remains nearly constant for most of the time before gradually Energies 2025,18, 2304 6 of 16 decreasing. The discharge stops when the voltage reaches a predefined lower threshold, referred to as the minimum discharge voltage point. This behavior is depicted in Figure 3, where the minimum voltage is set to 2.5 V. Figure 3. CC method discharging process: Voltage and current as a function of time. 3. Proposed Framework for Testing Lithium-Ion Batteries The proposed test bench is designed to perform charge and discharge cycles on individual battery cells. Although its scalability allows for the extension to a maximum number of cells, where the total charging voltage for series-connected cells can reach up to 20 V, and the maximum charging current for parallel-connected cells can reach up to 30 A, in a single battery pack. The main hardware components are the following: • Personal Computer: manages and controls the test bench. • Power Supply: Sorensen DLM20-30, AMETEK, Inc., Berwyn, PA, USA (600 W), providing the necessary testing voltage. The system can handle the voltage and current required by the single battery pack (up to a maximum of 20 V and 30 A supplied by this electronic equipment). This is a major limitation of our system. Higher voltage or current testing values require another power supply. • Router: D-Link D-300, D-Link Corporation, Taiwan, facilitating communication between devices. • Electronic load: B&K Precision 8614, B&K Precision Corporation, Yorba Linda, CA, USA, designed to emulate battery charge and discharge conditions by dynamically adjusting voltage and current parameters. This allows accurate characterization of battery performance under various load scenarios. • I/O Device: National Instruments USB-6281, Emerson Electric Co., St. Louis, MO, USA, serving as the test bench’s monitoring platform. It enables battery temperature measurement and allows system expansion by integrating additional relays and external monitoring devices. • Thermocouple: K type, ensuring accurate thermal monitoring by measuring battery temperature during analysis. These components are interconnected, as shown in Figure 4, which presents the schematic diagram of the developed system. The test system connects the battery in parallel with the power supply and the programmable electronic load. The thermocouple is positioned at an intermediate point on the external casing of the battery to monitor temperature variations. The power supply is integrated into the system via a local area network with Ethernet connectivity. The router dynamically assigns an IP address to the Energies 2025,18, 2304 7 of 16 power supply through the Dynamic Host Configuration Protocol. A photo of the test bed, where all the components are identified, is shown in Figure 5. During charge and discharge cycles, the system captures current values using the internal sensors of the power supply (for charging) and the programmable load (for discharging). The voltage across the battery terminals is measured using the voltage sensor of the DLM20-30 power supply. Additionally, the thermocouple is connected to an analog input of the NI USB-6281 device, allowing for precise thermal data acquisition. Figure 4. Schematic of the proposed system for the testing of rechargeable batteries. Figure 5. Photograph of the designed test bench. The tested battery is placed in a fireproof box of the following dimensions: 21.5 ×15 ×17 cm. The main characteristics of the software tool developed to control the test bed are the following: • It uses a LabVIEW environment to control and monitor the test bench. The developed software package allows the programming of charge and discharge cycles, real-time tracking of battery parameters, and data storage in plain text format for further analysis. • Python (version 3.12.7) was used to develop a post-processing data tool and generate graphical representation of the results obtained. The acquisition and control system (SDAYC from now on for simplicity) was developed in the LabVIEW graphical programming environment. This system enables control of Energies 2025,18, 2304 8 of 16 both the power supply and the programmable electronic load. In addition to its control functions, SDAYC provides real-time monitoring of voltage, current, and temperature, allowing for a comprehensive analysis of the battery’s performance during testing. Figure 6 presents the SDAYC interface developed in LabVIEW. This interface not only controls the test bench but also facilitates real-time data acquisition from the battery under analysis. Figure 6. Monitoring and control environment for the battery characterization. The SDAYC includes two operating modes: manual and automatic. In manual mode, the user can access the front panel Figure 6to manually perform charging and discharging operations. In this mode, specific parameters, such as charging voltage, charging current, and discharging current, can be adjusted at any time according to the requirements of the desired experimental process. In automatic mode, the user initially sets the values for charging voltage, charging current, discharging current, minimum discharge voltage, and data export time period in seconds. The minimum charging current is fixed at a constant value of 0.5 A. Once these parameters are configured, the program continuously executes charging and discharging operations until the user decides to end the process. In this mode, the charging process automatically stops when the current drops below 0.5 A, while the discharging process ends when the voltage reaches the predefined minimum discharge value. In addition, a resting period of 20 min is incorporated between each charge and discharge cycle, ensuring that the battery cools down properly before starting a new cycle. In automatic mode, two new independent text files are generated for each completed cycle: one for charging and the other for discharging. This means that every time a new charging or discharging period starts, the system automatically generates the corresponding files and begins recording data in them, depending on the process being carried out. Additionally, the SDAYC incorporates an interruption system that continuously monitors parameters such as temperature and cell voltage. If any of these values exceed an established range or if a manual stop button is pressed, the program is automatically interrupted to ensure system safety and protect the battery. In Figure 7, the general logic of the developed software is shown. In both operating modes, the following parameters are recorded: charging current, discharging current, battery cell voltage, and elapsed time. These data are periodically exported as text files (.txt) with an export interval predefined by the user. Energies 2025,18, 2304 9 of 16 START AUTOMATIC MODE CHARGING PROCESS YES CHARGE CURRENT< 0.5A REST 20 min VOLTAGE < MIN VOLTAGE TEMPERATURE > 45 ºC OR VOLTAGE > 4.2 V OR STOP BUTTON ACTIVATED MANUAL MODE NO NO MANUAL DISCHARGE PROCESS CREATE NEW .TXT FILES INTERRUPTION STOP PROGRAM REST 20 min YES YES NO RETURN Figure 7. Flowchart diagram of the software developed. 4. Experimental Validation To verify the correct operation of the battery test bench in conjunction with the automatic mode of the SDAYC that has been developed, automatic tests were performed. Using the data obtained from these tests, charge and discharge curves and parameters such as the SoH, which characterize the battery’s condition, were determined. Note that a strong point of our proposal is that there is no need for external data pre-processing for the reproducibility of the experiments. Our measurement system is based on the instruments utilized, which provide the voltage and current for the post-processing steps. The temperature is measured using a thermocouple, applying a first order hardware filter with a cut-off frequency of 1 Hz. The main specifications and characteristics of the battery analyzed are shown in Figure 8. Taking into account these characteristics, a minimum charging current threshold of 0.5 A was selected. The tested battery has a nominal capacity of 6 Ah. Although 0.1 C (0.6 A) is commonly accepted as the standard current cut-off value, a slightly lower value of 0.5 A was chosen according to [ 38 ] to offer a conservative margin without significantly compromising efficiency or charging time. In our study, a 2.5 V cutoff voltage was selected, where a conservative criterion was applied in the selection to guarantee the integrity and lifespan of the cell analyzed. Finally, the interruption system uses practical threshold values (temperature > 45 ◦ C, voltage > 4.2 V) also obtained from the cell specifications analyzed. Regarding overcharging, batteries can reach 4.8 V without the risk of explosion or fire, but a safety factor is applied and the threshold voltage is reduced to 4.2 V to ensure safe operation of the system. Additionally, the maximum recommended temperature ranges for most of the analyzed LiFePO 4 cells vary between 55 and 65 ◦ C. Then, the maximum system temperature in our system was limited to 45 ◦ C, which somehow considers possible Energies 2025,18, 2304 16 of 16 33. Li, Y.; Zhang, X.; Li, Z.; Li, X.; Liu, G.; Gao, W. Accurate and adaptive state of health estimation for lithium-ion battery based on patch learning framework. Measurement 2025,250, 117083. [CrossRef] 34. Chen, J.; Li, K.; Liu, W.; Yin, C.; Zhu, Q.; Tang, H. A Novel State of Charge Estimation Method for LiFePO4 Battery Based on Combined Modeling of Physical Model and Machine Learning Model. J. Energy Storage 2025,115, 115888. [CrossRef] 35. Chen, G.J.; Chung, W.H. Evaluation of Charging Methods for Lithium-Ion Batteries. Electronics 2023,12, 4095. [CrossRef] 36. Liu, H.; Naqvi, I.; Li, F.; Liu, C.; Shafiei, N.; Li, Y.; Pecht, M. An Analytical Model for the CC-CV Charge of Li-Ion Batteries with Application to Degradation Analysis. J. Energy Storage 2020,29, 101342. [CrossRef] 37. Maheswaran, V.D.; Carter, S.B.R.; Vijay, V.; Raj, S.V. Study and Implementation of Constant Current-Constant Voltage(CC-CV) Charger for Lithium-Ion(Li-Ion) Battery Using Machine Learning(ML). In Proceedings of the 2024 IEEE 4th International Conference on Sustainable Energy and Future Electric Transportation (SEFET), Hyderabad, India, 31 July–3 August 2024; pp. 1–5. [CrossRef] 38. Shen, W.; Vo, T.T.; Kapoor, A. Charging Algorithms of Lithium-Ion Batteries: An Overview. In Proceedings of the 2012 7th IEEE Conference on Industrial Electronics and Applications (ICIEA), Singapore, 18–20 July 2012; pp. 1567–1572. [CrossRef] 39. Singh, P.; Chen, C.; Tan, C.M.; Huang, S.C. Semi-Empirical Capacity Fading Model for SoH Estimation of Li-Ion Batteries. Appl. Sci. 2019,9, 3012. [CrossRef] 40. Theuerkauf, D.; Swan, L. Characteristics of Open Circuit Voltage Relaxation in Lithium-Ion Batteries for the Purpose of State of Charge and State of Health Analysis. Batteries 2022,8, 77. [CrossRef] Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.