A study on cable current detection for new energy charging stations based on TMR sensor arrays Authors: Jicheng Yu, Siyuan Liang, Changxi Yue, Yinghui He, Kai Xiao Published: 2026 ABSTRACT Accurately measuring current in charging station cables is pivotal to the design, manufacturing, and operation of AC charging stations, as it directly dictates their performance, economic viability, and safety. Addressing this need, this paper designed a tunnel magneto-resistance (TMR) sensor array for the non-contact, real-time monitoring of current in new energy vehicle charging cables. The experimental results validate the superior performance of the designed TMR sensor array over a single TMR sensor, with optimal detection accuracy achieved at a distance of 3 cm from the cable. Furthermore, the sensor successfully monitored real-time current across ten distinct brands and models of new energy vehicles. These findings demonstrate that the proposed array offers a reliable solution for real-time cable current detection with its high precision, stability, and non-contact nature. Introduction The escalating global climate crisis and its adverse effects on human health are attributed to the excessive emission of greenhouse gases (GHGs) from fossil fuel consumption. According to the U.S. Environmental Protection Agency (EPA), CO₂ alone represents about 65% of these emissions, with a significant portion originating from the exhaust of fuel-powered vehicles1,2. Consequently, the development of sustainable, clean, and pollution-free renewable energy was imperative to achieve the dual carbon goals (carbon peak and carbon neutrality), representing a crucial strategy to address global climate change. Given this context, leading new energy automobile manufacturers, including Tesla, BYD, and Xiaopeng Motors, have rapidly expanded their market presence. Electric vehicles (EVs) are progressively replacing internal combustion engine vehicles as a primary mode of daily transportation. Meanwhile, governments worldwide are implementing policies to facilitate this transition; an illustrative example is the UK’s mandate to halt sales of new gasoline and hybrid vehicles by 2030 and 2035, respectively3. Such initiatives are crucial for mitigating environmental impact and stimulating electric vehicle industry development. This impetus was reflected in market data, which shows that pure EVs have constituted 66% of the global EV market share over the past five years1. Consequently, the proliferation of EVs has promoted a growing demand for charging infrastructure. While home charging suffices for a segment of owners, the situation highlights the gap in urban areas, where most residents must depend on public charging networks4. Despite the rapid adoption of EVs and the expansion of charging infrastructure, significant safety concerns persist beyond emerging cyber risks3,5. Traditional electrical hazards remain a critical issue. Public charging stations, for instance, are frequently subject to external environmental factors; continuous solar radiation and moisture can compromise cable insulation, potentially resulting in electrical faults. Furthermore, internal defects, aging, wear, or improper grounding within the equipment may lead to insulation failure, electric shock, or fire. Given the diverse and complex risks in new energy charging stations, implementing robust safety protocols for the charging stations was critical. Real-time current monitoring of charging cables emerges as a simple yet powerful measure, given that current is a vital parameter for assessing operational health. While various sensors exist—such as fiber optic, shunt, and Hall-effect—novel current sensors based on the magneto-resistive (MR) effect are particularly promising. Notably, MR-based sensors offer distinct advantages such as compact size, non-contact operation, high sensitivity, and strong anti-interference capabilities, positioning them as a promising solution for this application6. The operational principle of MR sensors is based on the modulation of a material’s electrical resistance by an external magnetic field, which primarily includes anisotropic magneto-resistance (AMR)7, giant magneto-resistance (GMR)8, and tunnel magneto-resistance (TMR)9. A notable advantage of TMR sensors over AMR and GMR counterparts stems from the preservation of electron spin properties following transmission through the tunnel barrier. Consequently, TMR sensors demonstrate enhanced performance metrics, including superior magnetoresistance ratio, sensitivity, operational stability, signal-to-noise ratio, linear detection range, and reduced power consumption10. Notably, Xu et al.11 applied a ring-shaped TMR array sensor to current monitoring in railway traction network cables, with simulation and experimental results showing detection errors controlled within 0.5%. In another advancement, Hu et al.12 proposed a multi-level TMR current sensor whose structure was optimized by an improved algorithm (ACGWO-BP-NSGA-II). Their experiments demonstrated that the finalized sensor offered a 60 mA measurement range, high sensitivity (minimum 85 μA AC / 50 μA DC), and effective immunity to external electromagnetic interference. Significant research efforts have been devoted to enhancing the safety of new energy charging piles. For instance, Jiao et al.13 utilized a recursive damped least squares method to solve for the error coefficients of charging gun meters, enabling real-time monitoring and assessment of metrological deviations. In a study focused on fault diagnosis, Deng et al.14 developed a digital arc simulation system to measure fault currents under point contact and carbonization path conditions. The experimental data were used to train a support vector machine model optimized by particle swarm optimization (PSO-SVM), which ultimately achieved a 98.75% accuracy in identifying arc fault currents. The application of TMR sensors for current monitoring in charging station cables is an underexplored research area. To the best of the author’s knowledge, most established detection techniques require integrating additional hardware within the charging station, potentially compromising its performance. In addition, the achievable accuracy of current detection remains insufficient and demands further improvement. Despite significant advancements in DC fast-charging infrastructure, AC slow-charging stations retain an irreplaceable role, owing to their economic advantages and importance in electric vehicle battery maintenance. To enable non-intrusive monitoring in this context, this paper presents a TMR sensor array in this study. Its efficacy was experimentally validated and subsequently applied to monitor current at AC charging cables. The research further evaluated the detection accuracy across diverse environmental conditions and target variables, offering valuable insights for this practical deployment and optimization of this technology. Material and method In this study, charging services are provided by a NWDD-YL-A220/07 model charging station through the China Southern Power Grid Shunyi Charge online platform. The station’s rated specifications are as follows: input voltage of 220 V, input current of 32 A, and output power of 7 kW. The vehicles tested were selected with the onboard charger (OBC) as a key criterion, covering brands such as BYD, FULWIN, and BMW. The ambient magnetic field around the charging cables is largely unaffected by external disturbances, with its fluctuations predominantly governed by the current magnitude in the cable. TMR sensor array and hardware circuitry The TMR effect is a quantum mechanical phenomenon originating from electron spin, with its core mechanism based on the magnetic tunnel junction (MTJ) structure. A typical MTJ consists of an ultra-thin insulating tunnel barrier (1–2 nm) sandwiched between two ferromagnetic layers—a reference layer and a free layer, as illustrated in Fig. 1a15. When the magnetization directions of these two layers are parallel, electrons can readily tunnel through the barrier, resulting in a higher tunneling current and a low-resistance state. Conversely, when the magnetizations are antiparallel, electrons face a higher potential barrier, drastically reducing the tunneling current and creating a high-resistance state, with its physical origin illustrated in Fig. 1b.Fig. 1Structure and working mechanism of MTJs (a) MTJ structure (b) TMR effect principle. Structure and working mechanism of MTJs (a) MTJ structure (b) TMR effect principle. This study utilized the AK09940A, a high-precision triaxial magnetic sensor for accurate cable current detection. This integrated device, which combines x/y/z-axis sensors with driver, amplification, and processing circuits, delivers outstanding sensing performance. Consequently, to maximize measurement accuracy and reliability while also boosting the signal-to-noise ratio and interference immunity, a 4 × 4 array of these sensors was fabricated on a printed circuit board (PCB). Building upon this array sensor, the STM32F405RGT6 microcontroller was selected as the core of the sensing system. This chip, built around an ARM Cortex-M4 core, operates at up to 168 MHz and is equipped with 1 MB of Flash memory and 192 KB of RAM, providing substantial computational resources. Its integrated Floating-point unit (FPU) enables efficient execution of complex algorithms. The microcontroller also offers multiple peripheral interfaces, including UART, SPI, and I2C, for seamless communication with external devices. Additionally, its on-chip 12-bit Analog-to-digital converter (ADC) was capable of accurately acquiring weak analog signals from the sensors. The need for precise perception results necessitated the development of a dedicated hardware platform capable of processing the weak magnetic signals from the TMR sensor array with advanced data fusion algorithms. The resulting system, built upon the STM32F405RGT6 microcontroller (Fig. 2), forms the core control circuit board. This board was outfitted with critical supporting circuits, including an external crystal oscillator for timing, an ultra-high-speed serial interface for data communication, a DC/DC conversion circuit for power management, and a 32-bit level conversion circuit for signal compatibility.Fig. 2Block diagram of the control circuit based on the STM32F405RGT6 chip (a) External crystal oscillation circuit (b) UART to USB conversion circuit (c) DC-DC conversion circuit providing 1.8 V voltage for TMR sensors (d) 8-bit level shifter circuit (e) I2C bus interface for the AK09940A sensor. Block diagram of the control circuit based on the STM32F405RGT6 chip (a) External crystal oscillation circuit (b) UART to USB conversion circuit (c) DC-DC conversion circuit providing 1.8 V voltage for TMR sensors (d) 8-bit level shifter circuit (e) I2C bus interface for the AK09940A sensor. A 25 MHz external crystal oscillator circuit was integrated into the STM32-based hardware system (Fig. 3a) to provide a high-precision and stable master clock. It serves three critical functions: ensuring communication reliability, supplying an accurate time-base for the Real-time clock (RTC), and forming the foundational timing reference for the entire system. The performance of this oscillator circuit is paramount, as it directly influences the overall stability, reliability, and performance of the control system, thereby enabling the STM32F405RGT6 chip to operate at its full potential.Fig. 3Hardware system integration and communication interface. Hardware system integration and communication interface. In addition, the control system necessitates real-time transmission of substantial sensing data to a host computer. Given this requirement, a high-performance, highly integrated USB-to-serial bridge chip (CH343P) was employed as the core of the protocol conversion circuit to facilitate stable, bidirectional data communication and ensure seamless conversion between USB and UART protocols. The designed UART-to-USB conversion circuit is shown in Fig. 3b. This design provides full hardware flow control pins for the system and supports a data transmission rate of up to 6 Mbps. The various microdevices and modules within the control system require different operating voltages; for example, the CPU core typically needs 1.2 V, I/O ports 3.3 V, and the core TMR sensor 1.8 V. To address this, a DC-DC conversion circuit based on the TPS62293DRVR chip was designed (Fig. 3c). This circuit efficiently converts a higher input voltage to a stable 1.8 V output, critical for sensor operation, while maintaining high conversion efficiency. Its primary role is thus to provide a stable and efficient 1.8 V power rail dedicated to the TMR sensor array. A level-shifting interface is necessary because the STM32F405RGT6 microcontroller operates at 3.3 V, while the TMR sensor requires a 1.8 V interface. Directly connecting them would expose the sensor to damaging voltage levels. To resolve this mismatch, a 32-bit level conversion circuit was implemented using four TXB0108RGYR 8-bit bidirectional level shifter chips (Fig. 3d). Each pair of these 8-bit circuits interfaces with an FPC-05F-20PH20 connector, which in turn links the STM32 and the TMR sensor array. This design ensures accurate and reliable bidirectional signal transmission across the entire system. Given that the control system integrates 16 TMR magnetic sensors, the I2C communication protocol was adopted for multi-device interconnection. This choice simplifies PCB layout and facilitates the expansion of the sensor array. Consequently, all digital communication between the control chip and the sensors complies with the I2C bus protocol. The sensor data, pre-digitized by each TMR sensor’s internal ADC, is transmitted via this bus. The I2C interface of the AK09940A sensor is shown in Fig. 3e. This protocol effectively simplifies the overall system design and data reading process. Design and implementation of the current sensing device The fabricated TMR sensor array, realized based on the design outlined in Sect. "TMR sensor array and hardware circuitry", features sixteen sensors arranged in a uniform 4 × 4 matrix at the center of the PCB, with the entire assembly connected to the control system via an FPC-05F-20PH20 connector. The TMR sensing system interfaces with a host computer via the UART-to-USB circuit detailed in Sect. "TMR sensor array and hardware circuitry". The current sensing process for the charging station cable is depicted in Fig. 4. The procedure is initiated when the host computer sends a command to the sensing system. Upon receiving the command, the TMR sensor array acquires the magnetic field data from strategic positions adjacent to the cable. Then, this raw data undergoes preliminary processing within the hardware system before being transmitted to the host computer. Finally, dedicated algorithms on the host computer analyze this data to calculate the precise magnitude of the current flowing through the cable.Fig. 4Current sensing process for charging pile cables. Current sensing process for charging pile cables. Several non-invasive techniques are available for detecting cable current. Among these, TMR sensors are one option, alongside other commercially available sensor types. Therefore, this study benchmarks the TMR sensor against three alternative technologies under identical conditions to assess its comprehensive performance: a closed-loop Hall with core (ACS712), a Rogowski coil (DR5000M), and a current transformer (LMZ1D-hxd1). The measurement distance critically determines accuracy: excessive proximity risks sensor saturation from an overly strong magnetic field, while excessive distance yields a weak signal and suboptimal accuracy. Therefore, to balance the trade-off between magnetic saturation (from proximity < 2 cm) and signal weakness (from distance > 7 cm), the optimal sensor-to-cable distance was evaluated. Using a consistent test setup, the sensor array was positioned at 1 cm intervals from 2 to 7 cm from the cable outer layer at a marked detection point on the energized cable. The following protocol was adopted to ensure data accuracy and validity. At each predetermined distance, the TMR sensor performed 20 sequential magnetic field measurements, with a 5-min interval between successive readings to mitigate transient effects, the experimental process is shown in Fig. 5. Furthermore, the entire distance-determination test was repeated three times, and the average value was taken as the test result to ensure the reproducibility and statistical reliability of the results.Fig. 5In-situ current monitoring of charging pile cable. In-situ current monitoring of charging pile cable. Within a twisted cable, the live and neutral wires carry opposing currents, which, in principle, lead to the cancellation of their respective magnetic fields. However, in practice, factors such as uneven stranding, insulation layer inconsistencies, asymmetric current distribution, and variations in conductor spacing can result in a minute magnetic dipole moment within energized cables, which consequently generates measurable leakage magnetic fields. A key challenge is that leakage magnetic fields arise from numerous complex factors, making direct calculation of the underlying residual magnetic dipole moments difficult in actual production. To address this, the present study determines cable currents by measuring the actual leakage magnetic field, rather than attempting to model all contributing factors theoretically. To account for multiple influencing factors, the experimental setup employed a 100 cm cable connected to a vehicle’s 3.3 kW OBC. With test points marked at 2 cm intervals, the leakage magnetic field strength was measured by sensors placed 3 cm from the cable surface after energization. Each measurement was repeated three times, with the average value taken as the final experimental result. In addition, the magnitude of the magnetic field near a cable under ideal conditions was investigated through finite element simulation using COMSOL Multiphysics. Upon identifying the selected sensor and the optimal sensor-to-cable distance, the experimental procedure was as follows: One charging station and ten new energy vehicles of different brands and models (Table 1) were used. The sensor executed 50 magnetic field measurements on the charging cable at 5-min intervals for each vehicle. The measurements were repeated three times (three test groups), and the average value of these replicates was taken as the final test result. All experimental data were subjected to statistical analysis using SPSS Statistics 27 (IBM, Chicago, IL, USA), MATLAB 2020a (MathWorks, California, USA), and Origin 2021 (Origin Lab Corporation, Northampton, USA).Table 1New energy vehicle brands/models evaluated.Vehicle typeNumberVehicle brands/modelsOBC (kW)HEV1BYD / Sea lion 053.3 2Formula Leopard / Leopard 56.6 3JETOUR / L7 Plus6.6 4FULWIN / X3L6.6 5BYD / Song L3.3 EV6Geely Galaxy / Panda3.3 7Wuling / NG MINIEV3.3 8BMW / i37 9BYD / Seal7 10DENZA / Z9GT7 Table 1New energy vehicle brands/models evaluated.Vehicle typeNumberVehicle brands/modelsOBC (kW)HEV1BYD / Sea lion 053.3 2Formula Leopard / Leopard 56.6 3JETOUR / L7 Plus6.6 4FULWIN / X3L6.6 5BYD / Song L3.3 EV6Geely Galaxy / Panda3.3 7Wuling / NG MINIEV3.3 8BMW / i37 9BYD / Seal7 10DENZA / Z9GT7 New energy vehicle brands/models evaluated. Data processing and analysis The output signal from a single MTJ in TMR sensors is inherently weak. Consequently, a Wheatstone bridge configuration comprising four MTJs is typically employed in practical applications to amplify the output signal and improve the common-mode rejection ratio. The AK09940A TMR sensor used in this study adopts a “differential dual-bridge” architecture. This design incorporates two full measurement bridges for each of the three orthogonal magnetic field components (x, y, z), with each full bridge integrating four MTJ units. When operating in push–pull mode, the output voltage (Vout) of the bridge was given by Eq. (1)16:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${V}_{out}= {V}_{bias} \times \left(\frac{R+\Delta R}{\left(R+\Delta R\right)+\left(R-\Delta R\right)}-\frac{R-\Delta R}{\left(R+\Delta R\right)+(R-\Delta R)} \right)$$\end{document}where Vbias denotes the bridge circuit’s bias voltage (V), R is the nominal resistance (Ω) of an MTJ unit, and ΔR represents its resistance change (Ω) induced by an external magnetic field. Equation (1) could be simplified to obtain Eq. (2):2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${V}_{out} = {V}_{bias} \times \left(\frac{2\Delta R}{2R}\right) = {V}_{bias} \times \frac{\Delta R}{R}$$\end{document} Given that the magnetic flux density BL detected by the sensor originates from leakage flux and was converted linearly into a voltage signal, the sensor output characteristic could be expressed as:3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${V}_{out} = {V}_{0} + S \times {B}_{L}$$\end{document}where V₀ is the zero-voltage level when no detection occurs (V), and S is the sensitivity (V/T). Equation (3) could be transformed to obtain Eq. (4):4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S = \frac{{V}_{out} - {V}_{0}}{{B}_{L}}$$\end{document} The resistance of the TMR sensor is minimized when the magnetizations of the reference and free layers are parallel, and maximized when they are antiparallel. Consequently, the magneto-resistance ratio, defined by the ratio of these two resistance states, serves as a crucial figure of merit for sensor performance, quantified by Eq. (5)17:5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$MR = \frac{{R}_{B}-{R}_{0}}{{R}_{0}} \times 100\%$$\end{document}where RB is the sensor resistance under an applied magnetic field (Ω), and R0 is the sensor resistance in the absence of a field (Ω). A higher magneto-resistance ratio value corresponds to a larger relative change in resistance, thereby indicating superior sensor sensitivity. Consequently, the performance of each MTJ unit is critical to the TMR sensor’s magnetic field sensitivity. The external magnetic field’s strength and direction could be deduced by applying the right-hand rule to correlate the field orientation with the measured change in MTJ resistance. The magnetic field (B) generated by a current (I) in a charging cable could be modeled as the vector superposition of contributions from all infinitesimal current elements along the cable. This fundamental principle is quantitatively described by the Biot-Savart law, as expressed in Eq. (6)18:6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$dB = \frac{{\mu}_{0}Idl\times {e}_{\mathbf{r}}}{4\pi {r}^{2}}$$\end{document}where μ0 is the vacuum permeability (4π × 10−7 N/A2), dl is the infinitesimal current element vector (m), er is the unit vector pointing from Idl to point P, r is the magnitude of the displacement vector er, and dB is the resulting infinitesimal magnetic flux density vector (T). Given that the test environment is free of magnetic interference, the magnetic field around the cables is governed solely by the current they carry. Furthermore, the cable length substantially exceeds the dimensions of the compact TMR sensor array, and the sensing distance was maintained below 10 cm. Therefore, in this application, the energized cable can be modeled as an infinitely long straight conductor. At a nearby point P, the resulting leakage magnetic flux density is linearly proportional to that of a single conductor, with a proportionality constant ε. Under this assumption, Eq. (6) simplifies to Eq. (7)19:7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${B}_{L} = \varepsilon \frac{{\mu}_{0}I}{2\pi r}$$\end{document} The operational principle of the TMR sensor array, combined with the defined relationship between current and magnetic field in charging cables, enables the real-time current in energized cables to be derived from the measured magnetic field strength. The complete process was depicted in the computational flowchart of Fig. 6.Fig. 6Flowchart for cable current calculation. Flowchart for cable current calculation. A sensing system comprising a 4 × 4 array of AK09940A TMR sensors is positioned at point P to acquire the three orthogonal magnetic field components (Bx, By, Bz). This triaxial measurement strategy mitigates the influence of magnetic field orientation on the results. The magnitude of the total magnetic field vector at point P is then computed as the square root of the sum of squares of its components, given by Eq. (8):8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${B}_{total} = \sqrt{{{B}_{x}}^{2}+ {{B}_{y}}^{2}+ {{B}_{z}}^{2}}$$\end{document} To account for the potential influence of the background geomagnetic field on sensor accuracy, the ambient magnetic field components (Bx0, By0, Bz0) are measured before each test with the cable de-energized. During testing, these baseline values are subtracted from the measured readings to isolate the magnetic field generated solely by the cable current. The calculation was defined by Eq. (9):9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${B}_{total} = \sqrt{{{B}_{x}}^{2}+ {{B}_{y}}^{2}+ {{B}_{z}}^{2}}= \sqrt{{({B}_{x1}-{B}_{x0})}^{2}+ {({B}_{y1}-{B}_{y0})}^{2}+{({B}_{z1}-{B}_{z0})}^{2}}$$\end{document} The real-time current Ie could be determined with the distance r from point P to the cable and the vacuum permeability μ0 being known constants. By substituting the calculated total magnetic flux density Btotal at point P into the rearranged form of Eq. (7), the current was derived as expressed in Eq. (10):10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${I}_{e} = \frac{2\pi r{B}_{total}}{{\varepsilon \mu }_{0}}$$\end{document} Results and discussion Detect magnetic leakage in cables Measurements of the leakage magnetic field were performed at various points along a 100 cm cable using the TMR sensor. The results are shown in Fig. 7.Fig. 7Test results for leaky magnetic fields. Test results for leaky magnetic fields. As shown in Fig. 7, the measured leakage magnetic field values ranged from 13.61 × 10⁻3 to 15.07 × 10⁻3 mT. The overall variation across different cable positions was minimal. A linear fit to the data produced an intercept of 14.38 × 10⁻3 and a slope of 4.16 × 10⁻⁷. The external magnetic flux density of the cable is therefore approximated as a constant 14.38 × 10⁻3 mT, since the fitted curve exhibits near-zero slope. By comparing this value with that of an equivalent single conductor, ε was derived to be approximately 0.16. The simulation results for the cable under various operating conditions are shown in Fig. 8. Since ambient temperature and the cable lay length have a significant impact on the leakage magnetic field, this study focuses primarily on the effects of these two factors on the leakage magnetic field. Within the same distance range, the magnetic field near the twisted conductor was substantially weaker than that near the single conductor, and the field strength diminishes with rising temperature. This was likely because an increase in temperature causes the resistivity of the wire to increase, which consequently reduces the current flowing through the conductor. Specifically, at a distance of 3 cm from a single energized conductor, the magnetic field strength is approximately 1.00 × 10⁻4 T. In contrast, at a wire operating temperature of 10 °C, the magnetic field strength measured at a distance of 3 cm is approximately 1.06 × 10⁻4 T. When the ambient temperature increases to 40 °C, the field strength diminishes to 9.38 × 10⁻5 T. In addition, the magnetic field near the twisted cable exhibits dependence on the twist pitch. When the axial spacing of the stranded cable is 10 mm, the magnetic field strength at a distance of 3 cm from it is approximately 1.24 × 10⁻⁶ T. When the spacing increases to 40 mm, the magnetic field strength also increases to 1.33 × 10⁻5 T. At the same time, the magnetic field near the cable is also influenced by factors such as cable material, ambient temperature, and external magnetic interference. These influences cause the measured magnetic flux density of twisted-pair cables to exceed ideal-condition values, meaning that the experimentally determined attenuation coefficient ε is larger than the theoretical ratio of twisted-pair to single-wire field strengths. This finding is consistent with the experimental results presented.Fig. 8The simulation results of electromagnetic field intensity for the cable under various operating conditions (a) Single cable (b) The working temperature is 10°C (c) The working temperature is 40 °C. (d) The axial spacing of the strands is 20 mm. (e) The axial spacing of the strands is 10 mm. (f) The axial spacing of the strands is 40 mm. The simulation results of electromagnetic field intensity for the cable under various operating conditions (a) Single cable (b) The working temperature is 10°C (c) The working temperature is 40 °C. (d) The axial spacing of the strands is 20 mm. (e) The axial spacing of the strands is 10 mm. (f) The axial spacing of the strands is 40 mm. To accurately characterize the leakage magnetic field of new energy vehicle charging cables under realistic operating conditions, the simulation parameters were configured as follows: ambient temperature of 30 °C, live and neutral wires twisted with a pitch of 45 mm, and equal-magnitude currents flowing in opposite directions through the two conductors. The results are shown in Fig. 9. The simulation results demonstrate that the measured leakage magnetic field with a 3 cm radius of the cable is approximately 1.56 × 10⁻5 T. Compared with the field generated by an equivalent single conductor, this corresponds to an attenuation coefficient ε of approximately 0.156. Furthermore, in real-world operating environments, cables are inevitably exposed to additional subtle factors that can induce minor fluctuations in the leakage magnetic field. This finding is consistent with the experimental results presented.Fig. 9The simulation results of the leakage magnetic field of cables under actual operating conditions. The simulation results of the leakage magnetic field of cables under actual operating conditions. Testing of sensor performance A comparative evaluation of each sensor’s overall performance was performed by examining parameters including: linearity, dynamic range, saturation margin, sensitivity drift, hysteresis, and magnetic interference resistance, as specified in their respective datasheets. The results are presented in Table 2.Table 2A comparative analysis of comprehensive performance for non-invasive current sensors.TypeModelLinearityDynamic rangeSaturation marginSensitivity driftHysteresisMagnetic interference resistancePriceTMR sensorAK09940A + 0.1% ~ 1%60 dB2 ~ 3 IR − 0.3 to − 0.1%/°CFaintModerate¥15 ~ 50Closed-loop Hall with coreACS712≈ ± 1.5%40 ~ 50 dB1.2 ~ 1.5 IR + 0.05%/°CSignificantWeak¥8 ~ 20Rogowski coilDR5000M-DT50ID ± 0.1% ~ ± 0.5%80 Db∞ < ± 0.005%/°C–Extremely strong¥500 ~ 3000 + Current transformerLMZ1D-hxd1 ± 0.2% ~ ± 0.5%20 ~ 30 dBLower0.005 ~ 0.02%/°CRelatively significantStronger¥30 ~ 150 Table 2A comparative analysis of comprehensive performance for non-invasive current sensors.TypeModelLinearityDynamic rangeSaturation marginSensitivity driftHysteresisMagnetic interference resistancePriceTMR sensorAK09940A + 0.1% ~ 1%60 dB2 ~ 3 IR − 0.3 to − 0.1%/°CFaintModerate¥15 ~ 50Closed-loop Hall with coreACS712≈ ± 1.5%40 ~ 50 dB1.2 ~ 1.5 IR + 0.05%/°CSignificantWeak¥8 ~ 20Rogowski coilDR5000M-DT50ID ± 0.1% ~ ± 0.5%80 Db∞ < ± 0.005%/°C–Extremely strong¥500 ~ 3000 + Current transformerLMZ1D-hxd1 ± 0.2% ~ ± 0.5%20 ~ 30 dBLower0.005 ~ 0.02%/°CRelatively significantStronger¥30 ~ 150 A comparative analysis of comprehensive performance for non-invasive current sensors. Although the Rogowski coil demonstrates the best overall performance in Table 2, its higher cost is a limitation. In contrast, the TMR sensor offers the most favorable balance of performance and cost-effectiveness at a comparable price point. Subsequently, current measurements were performed on new energy vehicles equipped with 3.3 kW, 6.6 kW, and 7 kW OBCs using all four non-invasive sensors, with results shown in Fig. 10.Fig. 10Test results of non-invasive sensors (a) The OBC was rated at 3.3 kW (b) The OBC was rated at 6.6 kW (c) The OBC was rated at 7 kW. Test results of non-invasive sensors (a) The OBC was rated at 3.3 kW (b) The OBC was rated at 6.6 kW (c) The OBC was rated at 7 kW. As shown in Fig. 10, all four sensors produced readings near the actual cable currents across different OBC ratings. However, significant differences in stability and accuracy were observed. The TMR sensor and Rogowski coil exhibited superior performance, with readings tightly clustered around the expected values of 15 A, 30 A, and 32 A, respectively. Conversely, the Hall sensor and current transformer showed pronounced fluctuations and larger errors. Considering both application performance and test results, the TMR sensor was the optimal choice for cable current measurement. Factors influencing current detection accuracy A comparison was conducted at a detection distance of 3 cm to assess the performance of the custom TMR sensor array against a single sensor for charging pile cable current detection. The results are shown in Fig. 11, and the ANOVA analysis evaluating the effect of the sensor array on detection accuracy is provided in Table 3.Fig. 11Detection accuracy of different sensor configurations (a) The OBC was rated at 3.3 kW (b) The OBC was rated at 6.6 kW (c) The OBC was rated at 7 kW.Table 3ANOVA for TMR sensor array accuracy under varied conditions.VariablesSum of squaresDegree of freedomF-valuep-ValueSigArray0.22318.5760.02*Distance0.12543.914 × 1013 < 0.001******Indicates that p < 0.001, ** indicates that p < 0.01, * indicates that p < 0.05. Detection accuracy of different sensor configurations (a) The OBC was rated at 3.3 kW (b) The OBC was rated at 6.6 kW (c) The OBC was rated at 7 kW. Table 3ANOVA for TMR sensor array accuracy under varied conditions.VariablesSum of squaresDegree of freedomF-valuep-ValueSigArray0.22318.5760.02*Distance0.12543.914 × 1013 < 0.001******Indicates that p < 0.001, ** indicates that p < 0.01, * indicates that p < 0.05. ANOVA for TMR sensor array accuracy under varied conditions. ***Indicates that p < 0.001, ** indicates that p < 0.01, * indicates that p < 0.05. As shown in Fig. 11, the 95% confidence intervals for current measurements were obtained using the TMR sensor at point P. For the 3.3 kW, 6.6 kW, and 7 kW OBCs, these intervals are 15.25–15.45 A, 30.12–30.31 A, and 31.73–31.86 A, respectively. These considerable error margins and fluctuations may stem from insufficient sensor stability or uncontrolled external factors, which could introduce significant errors into the collected current data. For the TMR sensor array, the 95% confidence intervals for the measured currents are significantly narrower: 15.09–15.17 A, 30.08–30.15 A, and 31.88–31.94 A, respectively. In contrast to the single sensor, both fluctuation and error margins were reduced markedly. Despite occasional deviations caused by uncontrollable factors, the array demonstrated substantially higher overall sensing accuracy. For instance, Li et al.20 reported that an array of four TMR chips offered greater practicality for measuring magnetic fields around current lines. Similarly, Luo et al.21 achieved post-calibration current reconstruction errors of below 2% using a ring-shaped array of eight TMR chips for three-phase cable monitoring. Therefore, these results confirm that TMR sensor array designs enhance overall sensing accuracy, stability, and interference resistance. The detection accuracy varies with distance due to the leakage magnetic field’s spatial decay. Consequently, the array was deployed at distances ranging from 2 to 6 cm from the cable’s outer layer to evaluate this relationship. The resulting current detection data are presented in Fig. 12, and the ANOVA analysis quantifying the impact of distance on accuracy was provided in Table 3.Fig. 12Current detection results at varied sensing distances (a) The OBC was rated at 3.3 kW (b) The OBC was rated at 6.6 kW (c) The OBC was rated at 7 kW. Current detection results at varied sensing distances (a) The OBC was rated at 3.3 kW (b) The OBC was rated at 6.6 kW (c) The OBC was rated at 7 kW. As shown in Fig. 12, the detection accuracy of the TMR sensor array depends on its distance from the cable (point P). This was because energized charging cables from different vehicles generate magnetic fields of varying strength in the surrounding space. As a result, sensor readings at a distance of 3 cm most closely match the actual current, with minimal fluctuation and error. In contrast, distances shorter or longer than 3 cm result in significantly increased fluctuations and errors. This behavior may be attributed to the influence of external magnetic fields, such as the geomagnetic field and other ambient weak magnetic disturbances. Furthermore, the magnetic field generated by the current-carrying cable does not decay in a simple proportional manner with distance. Minor variations in the relative position of individual TMR elements within the array with respect to the cable may also contribute to accuracy variations. Liu et al.22 developed a current measurement probe with 16 TMR sensors organized into four arrays (each a Wheatstone bridge), featuring intra-array and inter-array spacings of 0.5 mm and 5 mm, respectively. The probe achieved a relative linearity error of less than 0.95% for DC currents from -3000 to 3000 mA, demonstrating that proper structural design is essential for TMR sensor arrays to deliver high accuracy and efficiency. Current detection performance on different vehicle models The current-carrying capacity of charging station cables is a critical determinant of the performance, safety, and charging costs for both the infrastructure and the EVs, as the actual current directly influences power delivery and charging speed. Although each station operates at a rated power, real-time monitoring of the actual cable current remains essential. Accordingly, using the TMR sensor array developed in this study and positioning it at the optimal detection point, current detection was performed on cables during the charging processes of ten different EV models. The corresponding results are summarized in Table 4.Table 4Performance of the TMR sensor array across different vehicle models.Vehicle typeNumberVehicle brands/modelsMaximum current (A)Minimum current (A)Average current (A)HEV1BYD / Sea lion 0515.82 ± 0.07g15.32 ± 0.01g15.54 ± 0.04f2Formula Leopard / Leopard 530.55 ± 0.31e29.78 ± 0.03f30.10 ± 0.23e3JETOUR / L7 Plus30.56 ± 0.18e30.12 ± 0.95d30.35 ± 0.02d4FULWIN / X3L30.77 ± 0.05d30.01 ± 0.25e30.45 ± 0.02c5BYD / Song L15.76 ± 0.14h15.09 ± 0.52h15.42 ± 0.01gEV6Geely Galaxy / Panda15.90 ± 0.35f14.80 ± 0.01j15.34 ± 0.12i7Wuling / NG MINIEV15.67 ± 0.30i15.03 ± 0.01i15.39 ± 0.21h8BMW / i331.96 ± 0.33a31.23 ± 0.11b31.65 ± 0.01a9BYD / Seal31.90 ± 0.22b31.34 ± 0.02a31.63 ± 0.13a10DENZA / Z9GT31.85 ± 0.18c30.76 ± 0.05c31.15 ± 0.04b Table 4Performance of the TMR sensor array across different vehicle models.Vehicle typeNumberVehicle brands/modelsMaximum current (A)Minimum current (A)Average current (A)HEV1BYD / Sea lion 0515.82 ± 0.07g15.32 ± 0.01g15.54 ± 0.04f2Formula Leopard / Leopard 530.55 ± 0.31e29.78 ± 0.03f30.10 ± 0.23e3JETOUR / L7 Plus30.56 ± 0.18e30.12 ± 0.95d30.35 ± 0.02d4FULWIN / X3L30.77 ± 0.05d30.01 ± 0.25e30.45 ± 0.02c5BYD / Song L15.76 ± 0.14h15.09 ± 0.52h15.42 ± 0.01gEV6Geely Galaxy / Panda15.90 ± 0.35f14.80 ± 0.01j15.34 ± 0.12i7Wuling / NG MINIEV15.67 ± 0.30i15.03 ± 0.01i15.39 ± 0.21h8BMW / i331.96 ± 0.33a31.23 ± 0.11b31.65 ± 0.01a9BYD / Seal31.90 ± 0.22b31.34 ± 0.02a31.63 ± 0.13a10DENZA / Z9GT31.85 ± 0.18c30.76 ± 0.05c31.15 ± 0.04b Performance of the TMR sensor array across different vehicle models. As shown in Table 4, the currents detected by the TMR sensor in cables closely match the actual values across hybrid and pure electric vehicles with different OBCs. Specifically, for vehicles with 3.3 kW, 6.6 kW, and 7 kW OBCs, the measured current ranges are 15.00 –16.00 A, approximately 30.00 A, and 31.00–32.00 A, respectively. This fluctuation is characteristic of AC charging systems, where the current is dynamically regulated by the vehicle’s onboard charger and battery management system (BMS). During the constant-voltage charging phase, for instance, the charging current is intentionally modulated to prevent overcharging while maintaining stable voltage. Additional factors, including battery temperature variations and grid voltage fluctuations, further contribute to current dynamics. Consequently, the specific charging current profile varies across different vehicle models due to distinct system characteristics and operating conditions. The TMR sensor array developed in this study effectively captures these real-time current variations under appropriate sensing conditions, with significant differences observed between maximum, minimum, and average current measurements. Xu et al.23 developed a current sensor centered on a TMR chip for weak current detection in smart grids. Their design incorporated a highly linear interface circuit to eliminate fixed bias and integrated a magnetic flux concentrator to improve sensitivity and noise immunity. Experimental results demonstrated a sensitivity of 29.4 mV/V/mA within the ± 100 A range. In a separate study, Feng et al.24 embedded TMR sensors within SiC MOSFET modules to monitor module current and compare it with predefined thresholds, enabling rapid identification of short-circuit or overcurrent faults. Their tests confirmed that this method achieves shorter detection latency, faster response times, and reduced overall protection durations. Comprehensive performance evaluation of the TMR sensor array Experimental results confirm that the designed TMR sensor array achieves superior accuracy and stability over single-sensor configurations. Its performance is distance-dependent, requiring optimal positioning for precise cable current detection. Validation across multiple vehicle models demonstrates the array’s capability for real-time, high-precision current monitoring in charging stations. The array leverages the high magnetoresistance ratio and miniaturized form factor of TMR units, enabling sensitive, non-intrusive magnetic field sensing without circuit interference and with inherent operational safety. Furthermore, the practical deployment of wireless charging systems for new energy vehicles was hindered by two critical challenges: metallic foreign object intrusion and coil misalignment. TMR sensor arrays offer a dual-function solution to these issues. For foreign object detection, the system employs a magnetic dipole model to analyze the magnetic gradient tensor in relation to distance vectors, enabling precise localization. For misalignment detection, it monitors changes in magnetic field distribution to determine both the direction and distance of coil displacement. In wireless electric vehicle charging systems employing DD coils, a sensor array can be configured to align with the characteristic “pole-to-pole” bipolar magnetic field distribution inherent to the DD coil geometry. By matching this unique magnetic field configuration, the array enables reliable metallic object detection through monitoring of absolute voltage variations in individual TMR sensors25. For example, Niu et al.26 designed an inter-pole induction coil array consisting of 15 coil units. The array exploits its inherent compatibility with the polarization characteristics and electromagnetic field distribution of a DD coil to detect metallic objects. The experimental results demonstrate that the proposed method yields a more uniform voltage spectrum, with the standard deviation decreasing from 177.38 mV to 90.16 mV. Therefore, TMR sensor arrays can not only be used to monitor current changes in cables in real time but also to detect anomalies in wireless charging systems, playing a significant role in advancing the development of new energy vehicle charging technologies. Conclusion This study developed a TMR-based sensor array to enable non-contact real-time current measurement in EV charging cables. Based on the experimental results, the main conclusions can be drawn as follows:TMR sensors provide a superior combination of sensing performance and cost-effectiveness among non-invasive current sensors, making them a compelling choice for this application.The TMR sensor array demonstrates a distinct advantage, offering higher measurement accuracy and greater stability than a single TMR sensor.The optimal detection distance between the TMR sensor array and the cable was 3 cm. TMR sensors provide a superior combination of sensing performance and cost-effectiveness among non-invasive current sensors, making them a compelling choice for this application. TMR sensors provide a superior combination of sensing performance and cost-effectiveness among non-invasive current sensors, making them a compelling choice for this application. The TMR sensor array demonstrates a distinct advantage, offering higher measurement accuracy and greater stability than a single TMR sensor. The TMR sensor array demonstrates a distinct advantage, offering higher measurement accuracy and greater stability than a single TMR sensor. The optimal detection distance between the TMR sensor array and the cable was 3 cm. The optimal detection distance between the TMR sensor array and the cable was 3 cm. This study has successfully validated the feasibility of TMR sensor arrays for cable current detection. However, sensor performance remains susceptible to environmental factors such as cable bending, external magnetic interference, and temperature variations in practical applications, which can compromise detection accuracy. This may require the use of relevant optimization algorithms to address the issue. Furthermore, the current measurement range of the sensors warrants further optimization and expansion to accommodate broader applications. Future work will focus on promising avenues for enhancement, including structural optimization and further improvements in measurement accuracy.