Battery connection fault diagnosis method and system based on magnetic field asymmetrical state decoupling

Abstract

The application relates to the technical field of battery fault detection, and discloses a battery connection fault diagnosis method and system based on magnetic field asymmetric state decoupling. Local magnetic field responses are obtained at symmetrical positions on both sides of a target connection node, and correction processing, node-level magnetic field asymmetric analysis and health connection state magnetic field reference response comparison are combined, so that current path deviation caused by bolt loosening, weld cracking, connection sheet cracking or uneven contact pressure at the connection node can be identified earlier. Compared with a detection method that only relies on temperature rise or pressure drop, early abnormalities such as local current deviation and flow deviation can be found before obvious heat accumulation or pressure drop anomalies occur; meanwhile, by comparing the current magnetic field observation result with the health connection state magnetic field reference response under the corresponding operating state, misjudgment caused by fluctuations due to normal factors can be reduced, so that the timeliness, accuracy and reliability of battery external connection node fault diagnosis are improved.

Description

Technical Field

[0001] This invention relates to the field of battery fault detection technology, specifically to a battery connection fault diagnosis method and system based on magnetic field asymmetry state decoupling. Background Technology

[0002] With the development of new energy vehicles and energy storage power stations towards larger capacity and higher integration, large-capacity batteries are widely used. Battery packs are typically composed of a large number of individual cells connected in series and parallel through busbars, terminals, welds, bolts, and other connection nodes. These connection nodes are subject to long-term effects such as vehicle vibration, thermal expansion and contraction during charging and discharging, cyclic expansion of the cells, and changes in clamping stress, making them prone to deterioration such as loosening, cracking, oxidation, or uneven contact pressure. Due to the large current in the main circuit, the increased minute contact resistance at the connection nodes can also generate Joule heating, accelerating connection deterioration and potentially leading to safety risks such as localized overheating, arcing, and ablation.

[0003] Existing battery connection fault detection methods employ contact voltage drop detection or temperature detection. Contact voltage drop detection judges the contact status by the voltage change at the connection point, but in high-voltage battery packs, it suffers from complex wiring, high insulation requirements, and difficulty in resisting interference. For wide busbars, multiple solder joints, or parallel branches, fault current may bypass the connection, making it difficult to identify and locate local faults in a timely manner. Temperature detection suffers from thermal conduction hysteresis; by the time abnormal temperature rise is detected, the connection node may have already experienced significant heat accumulation or material degradation, failing to meet early warning requirements. While magnetic field-based detection methods can indirectly determine the connection status through magnetic field changes, the magnetic field near the connection node is also affected by cell cyclic expansion, changes in clamping status, installation micro-displacement, sensor gaps, and changes in the position of adjacent conductors. Existing magnetic field detection methods often use a single magnetic field amplitude or fixed threshold for judgment, making it difficult to distinguish between magnetic field drift under healthy conditions and current path anomalies caused by actual connection faults, easily leading to false alarms. Summary of the Invention

[0004] To address the aforementioned problems, this invention proposes a battery connection fault diagnosis method and system based on magnetic field asymmetry decoupling. By utilizing the local magnetic field spatial asymmetry coefficient of the battery pack connection nodes and combining it with multi-dimensional operating state parameters of the battery, the method achieves early warning of connection faults at the battery pack connection nodes.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: One or more embodiments provide a battery connection fault diagnosis method based on magnetic field asymmetry state decoupling, including the following steps: Acquire local magnetic field signals at symmetrical positions on both sides of the target connection node of the battery to be diagnosed, as well as battery operating status parameters; Zero-bias correction and direction correction are performed on the local magnetic field signals on both sides of the target connection node to obtain the equivalent magnetic field response on both sides of the target connection node. Then, the current magnetic field observation characteristics of the target connection node and the local magnetic field spatial asymmetry coefficient on both sides of the target connection node are calculated. For each target connection node, a multi-dimensional state magnetic field reference model is constructed to characterize the relationship between the battery operating state and the magnetic field of the battery target connection node. The current battery operating state parameters are input into the multi-dimensional state magnetic field reference model to obtain the healthy connection state magnetic field reference response of the target connection node under the current operating state. Based on the current magnetic field observation characteristics of the target connection node and the magnetic field reference response of the healthy connection state, the normalized reference residual is calculated. Based on the node-level local magnetic field spatial asymmetry coefficient and the normalized reference residual, connection fault diagnosis is performed on the target connection node to obtain the diagnosis results. One or more embodiments provide a battery connection fault diagnosis system based on magnetic field asymmetry state decoupling, including: The data acquisition module is configured to acquire local magnetic field signals at symmetrical positions on both sides of the target connection node of the battery to be diagnosed, as well as battery operating status parameters. The magnetic field asymmetry identification module is configured to perform zero-bias correction and direction correction on the local magnetic field signals on both sides of the target connection node, obtain the equivalent magnetic field response on both sides of the target connection node, and then calculate the current magnetic field observation characteristics of the target connection node and the local magnetic field spatial asymmetry coefficient on both sides of the target connection node. The health reference construction module is configured to build a multi-dimensional state magnetic field reference model for each target connection node, which characterizes the relationship between the battery operating state and the magnetic field of the battery target connection node. The current battery operating state parameters are input into the multi-dimensional state magnetic field reference model to obtain the health connection state magnetic field reference response of the target connection node in the current operating state. The normalized reference residual calculation module is configured to calculate the normalized reference residual based on the current magnetic field observation characteristics of the target connection node and the magnetic field reference response of the healthy connection state. The fault diagnosis module is configured to perform connection fault diagnosis on the target connection node based on the node-level local magnetic field spatial asymmetry coefficient and the normalized reference residual, and obtain the diagnosis result.

[0006] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention acquires local magnetic field responses at symmetrical locations on both sides of the target connection node, and combines this with correction processing, node-level magnetic field asymmetry analysis, and comparison with the magnetic field reference response of a healthy connection state. This allows for earlier identification of current path deviations at the connection node caused by loose bolts, cracked welds, cracked connecting plates, or uneven contact pressure. Compared to detection methods that rely solely on temperature rise or voltage drop, this invention can detect early anomalies such as local current deviation and current skewing before significant heat accumulation or abnormal voltage drop occurs. Furthermore, by comparing the current magnetic field observation results with the magnetic field reference response of a healthy connection state under the corresponding operating conditions, it reduces misjudgments caused by normal factors such as current changes, temperature changes, SOC changes, cell expansion, and minor installation position offsets. This improves the timeliness, accuracy, and reliability of fault diagnosis for external battery connection nodes.

[0007] The advantages of the present invention, as well as its additional advantages, will be described in detail in the following specific embodiments. Attached Figure Description

[0008] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute a limitation thereof.

[0009] Figure 1 This is a flowchart of a battery connection fault diagnosis method based on magnetic field asymmetry state decoupling provided in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the symmetrical layout of the battery connection and the smart sensing busbar dual sensors in an application example of Embodiment 1 of the present invention; Figure 3 This is a schematic diagram illustrating that the current magnetic field observation characteristics are within the magnetic field reference interval in a healthy connection state, as shown in the application example of Embodiment 1 of the present invention. Figure 4 This is a schematic diagram illustrating the deviation of the current magnetic field observation characteristics from the magnetic field reference range of a healthy connection state in an application example of Embodiment 1 of the present invention; The components include: 1. Battery cell; 2. First busbar; 3. Sensing assembly; 4. Left magnetic sensor; 5. Right magnetic sensor; 6. Battery module insulating cover; 7. First reference magnetic sensor; 8. BMS and diagnostic terminal. Detailed Implementation

[0010] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0011] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0012] It should be noted that the terminology used herein is for describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof. It should be noted that, without conflict, the various embodiments and features within those embodiments can be combined with each other. The embodiments will now be described in detail with reference to the accompanying drawings.

[0013] Example 1 In one or more of the technical solutions disclosed in the embodiments, such as Figures 1 to 3 As shown, a battery connection fault diagnosis method based on magnetic field asymmetry decoupling includes the following steps: Step S1: Obtain the local magnetic field signals at the symmetrical positions on both sides of the target connection node of the battery to be diagnosed, as well as the battery operating status parameters; Step S2: Perform zero-bias correction and direction correction on the local magnetic field signals on both sides of the target connection node to obtain the equivalent magnetic field response on both sides of the target connection node, and then calculate the current magnetic field observation characteristics of the target connection node and the local magnetic field spatial asymmetry coefficient on both sides of the target connection node. Step S3: For each target connection node, construct a multi-dimensional state magnetic field reference model to characterize the relationship between the battery operating state and the magnetic field of the battery target connection node. Input the current battery operating state parameters into the multi-dimensional state magnetic field reference model to obtain the healthy connection state magnetic field reference response of the target connection node in the current operating state. Step S4: Calculate the normalized reference residual based on the current magnetic field observation characteristics of the target connection node and the magnetic field reference response of the healthy connection state; Step S5: Based on the node-level local magnetic field spatial asymmetry coefficient and the normalized reference residual, perform connection fault diagnosis on the target connection node and obtain the diagnosis result; In this embodiment, by acquiring local magnetic field responses at symmetrical positions on both sides of the target connection node, and combining this with correction processing, node-level magnetic field asymmetry analysis, and comparison with the magnetic field reference response of a healthy connection state, current path deviations caused by loose bolts, weld cracks, connecting plate cracks, or uneven contact pressure at the connection node can be identified earlier. Compared to detection methods that rely solely on temperature rise or voltage drop, this approach can detect early anomalies such as local current deviation and current skewing before significant heat accumulation or abnormal voltage drop occurs. Furthermore, by comparing the current magnetic field observation results with the magnetic field reference response of a healthy connection state under the corresponding operating conditions, misjudgments caused by normal factors such as current changes, temperature changes, SOC changes, cell expansion, and minor installation position offsets can be reduced, thereby improving the timeliness, accuracy, and reliability of fault diagnosis for external battery connection nodes.

[0014] Unlike existing battery detection methods based on magnetic field imaging, magnetic field gradient distribution, or current distribution inversion, the detection target of this embodiment is not the internal defects of the cell body, the consistency of individual cells, or the overall current distribution of the module, but the external connection nodes of the battery pack such as busbars, terminals, bolts, and welds. A diagnostic framework is constructed to extract local magnetic field features, decouple multi-dimensional state magnetic field reference response, and fuse multi-source information to achieve early identification of battery connection faults.

[0015] In step S1, optionally, the target connection node can be an electrical connection node at the busbar, terminal post, bolt connection, welded connection or connection piece in the battery pack. Specifically, the target connection node can be the connection node between the busbar and the cell terminal in the battery module, the welding node between the connecting piece and the terminal, the bolt crimping node, or the lap joint node between adjacent conductive connectors. It is possible to measure the local magnetic field signals on both sides of the target connection node using magnetic sensors installed on both sides of the target connection node. Pairs of magnetic sensors are symmetrically arranged on both sides of each target connection node to collect the local magnetic field response on both sides of the target connection node. In one specific arrangement, to facilitate explanation of the sensor placement and detection direction, a local coordinate system for the area where the target connection node is located can be defined: the origin is the geometric center of the target connection node, the center of the solder joint, the center of the bolt, or the center of the contact area between the connecting piece and the pole post; the x-axis is the direction of the main current flow in the busbar or connecting piece; the z-axis is the direction perpendicular to and away from the busbar surface; and the y-axis direction is determined according to a right-hand coordinate system, corresponding to the width direction of the busbar or connecting piece. Magnetic sensors are respectively arranged on both sides of the target connection node along the y-axis to collect the local magnetic field signals on both sides of the target connection node.

[0016] The magnetic sensors are arranged in pairs along the y-axis on both sides of the target connection node, or along both sides of the main current flow direction, with the two magnetic sensors symmetrically arranged relative to the target connection node or the target conductive path. The sensitive axis of the magnetic sensor is used to detect the tangential magnetic field component generated by the main current at the measuring point. With the above arrangement, when the target connection node is in a healthy connection state, the magnetic field response on both sides of the target connection node has a relatively stable correspondence. When the target connection node has uneven contact pressure, local cracks, welding defects or loose bolts, the current path flowing through the target connection node is deflected, causing the local magnetic field signals collected by the magnetic sensors on both sides to differ. This can provide a data basis for subsequent calculation of the local magnetic field spatial asymmetry coefficient.

[0017] The solution in this embodiment is not limited to the specific coordinate system setting described above; Because different battery modules may differ in busbar stacking, connector bending direction, terminal arrangement, insulation thickness, and sensor installation space, the direction of the magnetic field generated by the same main current near the target connection node may deflect. To ensure the accuracy and consistency of magnetic field detection results, the direction of the target magnetic field component to be detected can be determined based on the conductor orientation and current path of the target connection node, and the sensitive axis of the magnetic sensor can be matched with the direction of this target magnetic field component.

[0018] When using a single-axis magnetic sensor, the sensor primarily detects the magnetic flux density component along its sensitive axis. If the target magnetic field direction is at an angle to the sensitive axis direction, the sensor output signal may decrease, or even introduce directional errors. Therefore, in different module structures, the physical installation angle of the magnetic sensor can be adjusted to align its sensitive axis with the target magnetic field component. Alternatively, after sensor installation, calibration, coordinate transformation, or direction compensation algorithms can be used to correct the acquired magnetic field signal, ensuring the corrected magnetic field data corresponds to the target magnetic field component.

[0019] In one feasible implementation, the magnetic sensor at the target connection node can be set using an attached flexible circuit board structure. The magnetic sensor is set on the flexible circuit board and attached to the busbar surface or the connecting piece surface near the target connection node. An insulating isolation structure is set between the flexible circuit board and the metal busbar. Optionally, the insulating isolation structure includes one or more of the following: polyimide insulating film, insulating adhesive layer, potting layer, and insulating gasket, to meet the withstand voltage, creepage distance, and clearance requirements of the battery pack high-voltage system.

[0020] When the main flow direction of the current in the busbar is taken as the length direction and the width direction of the busbar is taken as the lateral direction, the first magnetic sensor is set in the left region of the target connection node, and the second magnetic sensor is set in the right region of the target connection node. The center line connecting the first and second magnetic sensors preferably extends along the width direction of the busbar. The center distance between the two magnetic sensors can be determined according to the busbar width, the target current magnitude, the sensor range, and the installation space, for example, set to 10mm to 50mm, preferably about 20mm; the distance between the sensitive element of the magnetic sensor and the busbar surface can be limited by the thickness of the flexible circuit board, the thickness of the insulating layer, and the thickness of the adhesive layer, for example, set to 1mm to 5mm, preferably about 3mm.

[0021] Optionally, the magnetic sensor can be one or more of the following: a TMR magnetic sensor, an AMR magnetic sensor, a GMR magnetic sensor, a Hall sensor, or a fluxgate sensor. The sensitive axes of the first and second magnetic sensors can be set to the same direction to acquire changes in the magnetic field amplitude and direction on both sides of the target connection node; alternatively, they can be set to opposite directions depending on the sensor packaging orientation and signal processing method, with polarity correction performed in the signal processing unit. The acquisition unit acquires the first magnetic field signal output by the first magnetic sensor and the second magnetic field signal output by the second magnetic sensor, respectively.

[0022] Furthermore, to improve the stability of the sensor mounting position, the flexible circuit board has a rigid reinforcement structure in the magnetic sensor mounting area. The rigid reinforcement structure can be an FR4 reinforcing plate, a ceramic reinforcing sheet, an insulating positioning sheet, or a locally thickened insulating support layer.

[0023] The rigid reinforcement structure is used to limit excessive local warping of the flexible circuit board under conditions such as vehicle vibration, thermal expansion and contraction, or adhesive layer aging, thus maintaining the stability of the initial height and lateral spacing of the magnetic sensor relative to the busbar. The flexible circuit board can also be fixed near the busbar using a temperature-resistant and flame-retardant adhesive layer, structural adhesive, pressure-sensitive adhesive, or local potting material. The adhesive material preferably has high and low temperature resistance, resistance to electrolyte volatiles, flame retardancy, and insulation properties.

[0024] In a surface-mount structure, since the magnetic sensor is placed close to the busbar or connector along with the flexible circuit board, the relative distance between the sensor and the target conductor is usually relatively stable.

[0025] In another embodiment, if the busbar surface of the battery module is not suitable for direct attachment of a flexible circuit board, or if there are structural factors such as high voltage insulation cover, sampling harness, or maintenance space limitations above the busbar, the magnetic sensor can be integrated on a rigid PCB board or insulating bracket and fixed to the insulating cover, high voltage protective cover, or inner side of the battery pack housing of the battery module; in this structure, a preset air gap, such as 5.0 mm, is maintained between the magnetic sensor and the busbar.

[0026] Optionally, the size of the air gap can be adjusted by an insulating bracket, a limiting post, a positioning groove, or a mechanical locking structure.

[0027] Through the two installation methods described above, the sensor setup in this embodiment is applicable not only to battery modules with attachable surfaces on the busbar, but also to battery packs where the sensor needs to be fixed to an insulating cover, housing, or bracket structure. Either of the above magnetic sensor setup methods can form a set of local magnetic field differential measurement units on both sides of each target connection node. The sensitive axis direction of the magnetic sensors on both sides is determined according to the target magnetic field component and can be adjusted through physical rotation installation, PCB trace direction design, or subsequent direction correction coefficients, so that the equivalent magnetic field response of the left and right measurement points under healthy connection conditions can be mapped to the same comparison direction. For bent busbars, multi-layer busbars, adjacent return paths, or irregularly shaped connectors, the direction correction coefficients of the left and right measurement points can be determined through healthy calibration conditions.

[0028] This is feasible. Each magnetic sensor is connected to the system's MCU. The system MCU acquires the output signals of each magnetic sensor through a low-voltage sampling line or an isolated sampling module. The sampling rate can be set according to the dynamic characteristics of the fault, for example, not less than 1kHz. A hardware synchronization trigger interface can be connected to the MCU to receive synchronization signals from the shunt, closed-loop Hall current sensor, or other total current sampling units in the BMS, in order to reduce the timestamp deviation between the magnetic field data frame and the total current of the battery system. For dynamic operating conditions such as pulse current, vehicle accelerated discharge, or rapid power regulation of the energy storage system, the sampling rate can be further increased, or a hardware trigger latching method can be used to ensure the synchronous calculation of the spatial asymmetry coefficient and the current normalized response characteristics.

[0029] In some embodiments, two magnetic sensors are provided at each target connection node, and the magnetic sensors of all target connection nodes constitute a magnetic sensor array. In step S1, magnetic induction intensity data collected by the magnetic sensor array arranged at the bus connection node of the battery pack is acquired in real time and synchronously. Simultaneously, battery operating status parameters provided by the battery management system (BMS) or auxiliary status input module are acquired. Battery operating status parameters include current parameter I, which characterizes the current load of the target connection node, the current state of charge (SOC) of the battery pack, and ambient temperature or cell temperature (T); it may also include auxiliary information such as the battery's state of health (SOH), historical cycle count, clamping pressure of the target connection node, temperature rise rate of the connection node, and individual cell voltage or branch voltage drop.

[0030] The current parameter I can be the total current of the main circuit of the battery pack, the target module current, the target parallel branch current, or the equivalent current flowing through the target connection node calculated based on the battery pack topology and branch current sharing relationship.

[0031] Step S2: Perform zero-bias correction and direction correction on the local magnetic field signals on both sides of the target connection node to obtain the equivalent magnetic field response on both sides of the target connection node, including the following steps: Step S21: Obtain the zero-bias reference value of the magnetic sensors on both sides of the target connection node; Specifically, the zero-bias reference values ​​of the magnetic sensors on both sides of the target connection node are used to eliminate the influence of the sensor's static background, ambient magnetic field, and installation background magnetic field on the detection results. The zero-bias reference value can be calculated based on a preset initial health calibration value combined with dynamic correction from the remote reference magnetic sensor. Optionally, the remote reference magnetic sensor is positioned far from the target connection node, largely unaffected by the local connection current magnetic field of that node, but in the same external magnetic field environment as the magnetic sensors on both sides of the connection node. It is used to collect the ambient magnetic field and common-mode magnetic field components. During dynamic correction, the reading of the remote reference magnetic sensor is used as the ambient magnetic field reference to extract the slowly changing drift component of the ambient magnetic field. Based on this, the zero bias reference value of the magnetic sensors on both sides is compensated and corrected so that the zero bias reference value can track the changes in the ambient magnetic field without introducing the local magnetic field component generated by the connection current.

[0032] To ensure the safety of zero-bias reference value updates, periodic updates of the zero-bias reference value are allowed only when preset safety update conditions are met. Optionally, the preset safety update conditions include: the main relay connected to the battery is disconnected, the current of the target connection node is zero, the battery energy storage system is left undisturbed for a preset time, and it is determined that there are no suspected fault records at present. Furthermore, when the monitored and identified battery system is in a suspected fault warning state or in the static fault auxiliary screening stage, the current zero bias reference value is locked, and the currently collected residual magnetic field signal is prohibited from being updated to a new zero bias reference value, so as to prevent the tiny magnetic field signal generated by the real fault from being erroneously canceled by the algorithm.

[0033] Step S22: Determine the orientation correction coefficients of the magnetic sensors on both sides based on the conductor orientation, current flow direction, magnetic sensor installation direction, and target magnetic field component direction at the target connection node. The direction correction coefficient is used to map the magnetic field components with opposite directions on both sides into equivalent magnetic field responses with the same direction and sign in a healthy connection state. The direction correction coefficient can take a value of +1 or... 1.

[0034] According to the Biot-Savart law, at symmetrical measurement points along an approximately straight main conductive path, the tangential magnetic fields generated by the main current are usually in opposite directions. Therefore, by setting a direction correction coefficient, positive and negative cancellation of the magnetic field signals on the left and right sides can be avoided in subsequent differential or asymmetric feature calculations.

[0035] Step S23: Based on the original sampled values ​​of the magnetic sensor, the zero bias reference value, and the direction correction coefficient, calculate the equivalent magnetic field response on both sides of the target connection node; ; in, , These are the local magnetic field signals measured by the magnetic sensors on the left and right sides of the same connection node, i.e., the original sampled values; , These are the zero-bias reference values ​​for the magnetic sensors on both sides of the same connection node, with units consistent with magnetic induction intensity. , The equivalent magnetic field response on both sides of the target connection node; , The orientation correction coefficients are the magnetic sensors on both sides of the target connection node. Furthermore, the amplitudes of the equivalent magnetic field responses on both sides of the target connection node are extracted separately, and the absolute value of the difference between the amplitudes of the equivalent magnetic field responses on both sides is calculated. The absolute value of the amplitude difference is divided by the sum of the amplitudes of the equivalent magnetic field responses on both sides to obtain the local magnetic field spatial asymmetry coefficient on both sides of the target connection node. Its mathematical definition is: ; in, To prevent the denominator from being too small under low current or zero current conditions, a magnetic field stability term constant is introduced to avoid the local magnetic field spatial asymmetry coefficient being amplified or diverged by sensor noise, and to ensure that it remains bounded and stable under low current conditions. Its unit is consistent with the magnetic induction intensity (e.g., mT or μT). These represent the amplitudes of the equivalent magnetic field responses on both sides of the target connection node; Optional, The value can be determined based on factors such as the inherent noise standard deviation of the magnetic sensor and the quantization resolution level of the system ADC. For example, it can be set to 3 to 5 times the standard deviation of the magnetic field noise of the battery system under healthy conditions.

[0036] In the above embodiments of this example, pairs of magnetic sensors are set on both sides of the target connection node to collect the magnetic field response of the local area of ​​the connection node. After zero bias correction and direction correction, the node-level local magnetic field spatial asymmetry coefficient is extracted. This spatial asymmetry coefficient is used to characterize the local deflection or flow around the connection node caused by factors such as loose bolts, local cracks in welds, cracks in connecting plates, shrinkage of conductive cross sections, or uneven contact pressure. Compared with the method of judging solely based on the overall magnetic field amplitude, the entire magnetic field image, or the abrupt change in the magnetic field gradient on the surface of the battery cell, the spatial asymmetry coefficient feature can more directly correspond to the spatial redistribution of local current paths at the connection node.

[0037] In a further technical solution, step S2 involves calculating the current magnetic field observation characteristics of the target connection node, including the equivalent magnetic field amplitude characteristics. and current normalized response characteristics At least one of them; Equivalent magnetic field amplitude characteristics , where is the mean of the equivalent magnetic field response amplitudes on both sides of the target connection node, used to characterize the average magnetic field response level of local measuring points on both sides of the target connection node, is given by the formula: ; in, The unit is the same as that of magnetic induction intensity, and it is a non-negative amplitude characteristic used to describe the overall magnetic field response amplitude at the target connection node.

[0038] This embodiment constructs equivalent magnetic field amplitude characteristics. This method can uniformly characterize the magnetic field response amplitudes of local measurement points on both sides of the target connection node, avoiding judgment biases caused by installation deviations, local noise, electromagnetic disturbances, or differences in measurement point positions on one side. This makes the overall magnetic field response level at the target connection node more stable and objective. Because... The calculation uses the average amplitude of the equivalent magnetic field response on both sides and is limited to a non-negative amplitude characteristic, thus directly reflecting the overall change in magnetic induction intensity at the target connection node. When there are abnormalities such as poor contact, loose connection, increased local resistance, or uneven current distribution at the connection node, the magnetic field distribution near the node will change accordingly, and the constructed... This approach effectively captures changes, thereby improving the accuracy and robustness of subsequent state recognition, anomaly detection, and fault diagnosis. Furthermore, the feature calculation method is simple, facilitating rapid extraction in real-time sampling and online monitoring scenarios, which helps reduce algorithm complexity and enhances the feasibility of system engineering applications.

[0039] Current normalized response characteristics The equivalent magnetic field amplitude characteristics at the target connection node. The normalized result relative to the equivalent current amplitude flowing through the target connection node is used to reduce the influence of equivalent current changes on the magnetic induction intensity response amplitude under different current-carrying conditions. It characterizes the average magnetic field response level of local measuring points on both sides of the target connection node under unit current. The formula is: ; in, This represents the equivalent current flowing through the target connection node; when the target connection node is located in the main circuit of the battery system, The total current of the main circuit of the battery system can be taken; when the target connection node is located in the target module, parallel branch, or local connection branch. The corresponding module current, branch current, or equivalent current calculated based on the battery pack topology and branch current sharing relationship can be used. The current stabilizing term is introduced to prevent instability in the denominator of small currents, and its unit is ampere (A). The settings can be configured according to the current sensor noise, current sampling resolution, or stability requirements under low current conditions.

[0040] This embodiment uses the equivalent magnetic field amplitude characteristics Dividing by the sum of the current amplitude and the current stability term can reduce the impact of different load currents, different charge / discharge rates, or transient current fluctuations on the magnetic field observation results, thus making... It more comprehensively reflects the structural state, contact state, or local current distribution changes at the target connection node. When the target connection node has poor contact, loose connection, abnormal local impedance, or uneven current distribution, under the same or similar current normalization conditions, It will have an offset relative to the normal connection state, which can be used as a characteristic parameter for subsequent anomaly identification, status assessment or fault diagnosis.

[0041] In step S3, for each target connection node, a multi-dimensional state magnetic field reference model is constructed to characterize the relationship between the battery operating state and the magnetic field of the battery target connection node. Specifically, the multi-dimensional state magnetic field reference model uses the battery operating state parameters as input parameters and outputs the healthy connection state magnetic field reference response or the healthy connection state magnetic field reference interval at the target connection node under the no-connection fault state. Optionally, the multidimensional state magnetic field reference model can be constructed using a physical model, a data-driven model, or a model that couples physical priors with state residuals. The physical model can include an equivalent conductor magnetic field model, a simplified finite element model, a calibration geometric factor model, or a Biot-Savart approximation model; the data-driven model can include a multidimensional lookup table model, a polynomial fitting model, a multivariate nonlinear regression model, a recursive least squares model, or a machine learning prediction model. As a preferred approach, the multidimensional state magnetic field reference model is constructed by coupling physical priors with state residuals, i.e., it is composed of physical prior reference terms. With state residual correction term Coupled together, it can be represented as: ; in, In order to match the current observation features The corresponding healthy connection status magnetic field reference response; It is a physical prior reference term used to characterize the fundamental magnetic field response generated by the equivalent current flowing through the target connection node under a healthy connection structure and a given sensor installation relationship.

[0042] When using the equivalent magnetic field amplitude characteristic In one feasible approach, when using the current magnetic field observation characteristics as a basis, under the conditions that the target connection node is in a healthy connection state and the sensor installation position remains unchanged, the equivalent current amplitude and the corresponding equivalent magnetic field amplitude characteristics under multiple current-carrying conditions are collected. and to The relationship between the magnetic field magnitude and the equivalent current magnitude is linearly fitted through the origin, and the slope of the linear fit is determined as the magnetic field proportionality coefficient. And calculate the physical prior reference terms according to the following formula: ; Where I is the equivalent current flowing through the target connection node, and |I| is the equivalent current amplitude; This is the magnetic field proportionality coefficient corresponding to the target connection node, expressed in mT / A or μT / A. It is used to comprehensively characterize the influence of the conductor structure, current path, and magnetic sensor installation relationship of the target connection node on the foundation's magnetic field response. If a linear fitting method including the intercept term is used, the intercept can be incorporated into the state residual correction term. In the constant term.

[0043] In this implementation, the physical prior reference term With equivalent magnetic field amplitude characteristics The dimensions are consistent, and the units are consistent with the magnetic induction intensity.

[0044] Physical prior references The basic part representing the change of magnetic field response with equivalent current under healthy connection conditions is not used to represent the magnetic field response deviation caused by slow changes in SOC, temperature, SOH, cycle number, clamping state or installation state, nor is it used to represent abnormal magnetic field response caused by connection failure.

[0045] In addition to the above method of obtaining the magnetic field proportionality coefficient based on the health connection state calibration, alternative physical prior reference terms are available. It can also be constructed based on the equivalent conductor magnetic field model, the finite element simplified model, the calibration geometric factor, the Biot-Savart approximation model, or a combination of the above methods.

[0046] This is a state residual correction term, used to correct the actual magnetic field response and the physical prior reference term in a healthy connection state. The deviation between [the two parameters]. This deviation refers to the deviation caused by normal operating conditions such as SOC, temperature, SOH, number of cycles, clamping pressure, changes in installation status, or changes in the position of adjacent conductors, assuming no connection faults at the connection node, but not by physical a priori reference items. Fully characterized magnetic field response changes.

[0047] During the health connection status calibration phase, based on the observed magnetic field characteristics of the health connection status collected under different operating conditions, the physical prior reference terms under the corresponding operating conditions are subtracted. The state residuals under healthy connection states are obtained, and a state residual correction term is constructed based on the state residuals. .

[0048] State residual correction term It is only used to compensate for magnetic field response deviations caused by changes in normal operating conditions under healthy connection conditions, and is not used to fit abnormal magnetic field responses caused by connection faults such as loose connection, weld crack, connecting piece crack, conductive cross-section shrinkage, or abnormal local current path.

[0049] Optional, state residual correction term It can be constructed using multidimensional lookup table interpolation, polynomial fitting, multivariate nonlinear regression, recursive least squares, machine learning prediction models, or combinations thereof.

[0050] In this embodiment, by using physical prior reference terms With state residual correction term By superposition, a healthy connection state magnetic field reference response corresponding to the current current and operating state is obtained. The current magnetic field observation characteristics will be compared with the magnetic field reference response of the healthy connectivity status. The comparison is used to characterize the degree of deviation of the magnetic field response of the current target connection node from the healthy connection state under the same operating conditions.

[0051] Physical prior references Provides a core response with physical interpretability and operating condition extrapolation capability, along with state residual correction terms. This embodiment compensates for slowly changing state factors not covered by the physical model. While models based solely on physical priors have good extrapolation capabilities under high-current conditions, they struggle to cover magnetic field response drift caused by slowly changing state factors such as SOC, temperature, SOH, cycle count, and clamping pressure. Models based solely on data fitting can compensate for these slowly changing state factors, but their extrapolation capabilities are insufficient and their physical interpretability is weak under high-current or extreme conditions not covered by the training data. This embodiment couples the two approaches, using... Provides a physically reasonable backbone response, by By correcting the slowly varying state residuals, the multidimensional state magnetic field reference model achieves high fitting accuracy within the training conditions and maintains physically reasonable externality outside the training conditions, thereby improving the normalized reference residuals. This improves the accuracy of characterizing real connection anomalies and reduces the risk of misjudging normal changes in magnetic field response under healthy connection conditions as connection failures.

[0052] Optionally, in this embodiment, the multidimensional state magnetic field reference model uses the magnetic field reference response as the output, which can be compared with the equivalent magnetic field amplitude characteristics. Correspondingly, it can also be compared with the normalized magnetic field response characteristics of the current. correspond; It should be noted that the output of the multidimensional state magnetic field reference model should maintain dimensional consistency with the currently selected observation features. When the system uses equivalent magnetic field amplitude features... When making the judgment, the multidimensional state magnetic field reference model outputs the healthy connection state magnetic field reference response. Its unit is consistent with magnetic flux density; in this case, the reference response of the magnetic field in a healthy connection state can be expressed as: When the system adopts the current normalized response characteristics When using the current magnetic field observation characteristics as a basis, the same normalization process can be applied to the healthy connection state magnetic field reference response under the equivalent magnetic field amplitude dimension, following the calculation method of the current normalized response characteristics in step S2, to obtain: ;in, To match the current normalized response characteristics The corresponding healthy connection state magnetic field reference response, the unit of which is the ratio of magnetic induction intensity to current; This is the current stabilization term introduced in step S2 to prevent instability in the denominator of small currents.

[0053] Alternatively, the current-normalized response characteristics under healthy connection conditions can be directly used. As a fitting or training object, construct with A dimensionally consistent multidimensional state magnetic field reference model.

[0054] The appropriate sampling link and operating conditions can be selected. or As observational features, both can be used for joint judgment, but in the same judgment logic, the observational features and the output of the multidimensional state magnetic field reference model should maintain the same dimensions.

[0055] In one specific implementation, the multidimensional state magnetic field reference model can be expressed as: ; in, Indicates the characteristics observed at present. The term refers to the collective reference response of the magnetic field in the corresponding healthy connection state. When using the equivalent magnetic field amplitude characteristic... At that time, the corresponding healthy connection state magnetic field reference response is denoted as ;when Using normalized response characteristics At that time, the corresponding healthy connection state magnetic field reference response is specifically denoted as: .

[0056] As one feasible solution, the state residual correction term This can be achieved using a second-order polynomial fitting method, defining the normalized state variables: ; Where I is the equivalent current flowing through the target connection node; This is a reference current greater than zero, measured in amperes. This is a reference temperature, in degrees Celsius. It is a normalized scale for temperatures greater than zero, and its units are... Consistent; All are dimensionless state variables.

[0057] State residual correction term It can be represented as: ; Where a0 to a9 represent the fitting coefficients of the state residual correction term, which can be determined by least squares regression based on calibration data collected under healthy connectivity conditions. This is because the normalized state variables... Since all variables are dimensionless, the units of the fitting coefficients a0 to a9 are the same as those of the state residual correction term. The units are consistent. When using the equivalent magnetic field amplitude characteristic... When used as a current characteristic of magnetic field observation And the units of a0 to a9 are all consistent with the magnetic induction intensity; when the response characteristics are directly normalized to current. When constructing the corresponding state residual correction model, The units of the corresponding fitting coefficients are the ratio of magnetic induction intensity to current.

[0058] Specifically, for each healthy connectivity state calibration sample, the observed magnetic field characteristics and corresponding physical prior reference terms of that sample are calculated. The difference between them is used as the fitting target for the state residual correction term; the equivalent current amplitude, SOC and temperature corresponding to the sample are used as model inputs to determine the fitting coefficients a0 to a9.

[0059] The second-order polynomial is used to describe the relationship between state residuals and current, state of charge (SOC), and temperature under healthy connection conditions. The output of the second-order polynomial is the state residual correction term. Instead of a complete healthy connection state magnetic field reference response .

[0060] It should be noted that the second-order polynomial given in this embodiment uses the equivalent current amplitude, SOC, and temperature as input variables to explain the state residual correction term. One specific construction method is as follows: When SOH, number of cycles, clamping pressure, or quantifiable installation state parameters are used as independent input variables of the state residual correction model, corresponding normalized state variables and corresponding linear, quadratic, or interaction terms can be added to the above model. Alternatively, the aforementioned multidimensional lookup table model, multivariate nonlinear regression model, recursive least squares model, or machine learning prediction model can be used for construction.

[0061] This second-order polynomial model is used only as a state residual correction term. One feasible solution. State residual correction term. It can also be constructed using multidimensional lookup table models, multivariate nonlinear regression models, recursive least squares models, or machine learning prediction models, or combinations thereof. The scope of protection of the method in this embodiment is not limited to this specialized form.

[0062] Physical prior references Other construction methods are determined according to the aforementioned equivalent conductor magnetic field model, finite element simplified model, Biot-Savart approximation model, calibration geometric factor, or a combination of the above methods.

[0063] This embodiment's multi-dimensional state magnetic field reference model describes the normal changes in the magnetic field response under healthy connection conditions caused by changes in normal operating conditions, based on the equivalent current flowing through the target connection node, SOC, temperature, and optional state parameters such as SOH, cycle count, and clamping pressure. By comparing the current observed magnetic field characteristics with the magnetic field reference range under healthy connection conditions, it distinguishes between normal changes in the magnetic field reference response under healthy connection conditions and abnormal deviations in the observed magnetic field response. Simultaneously, by combining auxiliary information such as connection node temperature rise, connection voltage drop, branch voltage drop, or individual unit voltage, it can perform a fusion judgment on contact degradation accompanied by thermal and electrical anomalies without significant changes in the current path. Therefore, the method of this embodiment can separately identify connection structure faults that alter the current path, normal changes in the magnetic field response under healthy connection conditions, and contact degradation risks without significant changes in the current path, avoiding the simplistic equating of all magnetic field amplitude changes with connection faults.

[0064] In step S4, to determine whether the current magnetic field observation characteristics deviate from the magnetic field reference response of the healthy connection state, the difference between the current magnetic field observation characteristics of the target connection node and the magnetic field reference response of the healthy connection state is calculated. Based on the offset of the difference relative to the magnetic field reference response of the healthy connection state, the normalized reference residual is calculated, as follows: ; in, Based on current magnetic field observation characteristics, it can be used for or ; The healthy connection state magnetic field reference response output by the multidimensional state magnetic field reference model under the corresponding dimensions; To prevent the stable term from diverging in normalization calculations due to an excessively small reference response, the unit and Consistent.

[0065] In step S5, based on the local magnetic field spatial asymmetry coefficient and normalized reference residual on both sides of the target connection node, a connection fault diagnosis is performed on the target connection node to obtain the diagnosis result, including the following steps: Step S51: Perform preliminary verification on the magnetic field detection data of the target connection node to eliminate external common-mode magnetic field interference, sensor failure, installation abnormality or sampling link abnormality, and determine whether the magnetic field detection data of the target connection node meets the requirements for the effectiveness of connection fault diagnosis. If it is effective, proceed to the next step. Before performing a comprehensive fault assessment and outputting a definitive fault conclusion, this step performs a pre-test for the effectiveness of the magnetic sensor array to avoid misjudging external magnetic interference, sensor saturation, zero-bias drift, or installation abnormalities as connection structure faults.

[0066] Specifically, the validity detection of the magnetic sensor detection data on the left and right sides of the target connection node is as follows: Step S511: When the local magnetic field signals on the left and right sides of the target connection node simultaneously undergo a sudden change in the same direction exceeding a set ratio, and the sudden change has a high synchronous correlation with the magnetic field change collected by the reference magnetic sensor at the far end of the battery pack, the current anomaly is determined to be external common-mode magnetic field interference, not a connection node failure. For example, external common-mode magnetic field interference can cause high-voltage relays to trip or drive motors to leak flux. As an optional implementation method, within a preset time window Within this time window, the correlation coefficient between the common-mode variation of the original magnetic field signals on both sides of the target connection node and the original magnetic field variation of the remote reference magnetic sensor is calculated. : ; in, The left and right magnetic sensors of the target node are respectively located within a preset time window. The change in the internally measured magnetic field signal, superscript This represents the original magnetic field signal (i.e., the original sampled signal without zero bias correction and direction correction). For the remote reference magnetic sensor in the same time window The change in the magnetic field signal within; the units of the above three magnetic field changes are all consistent with the magnetic induction intensity. This represents the correlation calculation function; The dimensionless correlation coefficient between the common mode variation on both sides and the variation at the far reference; The preset time window length, in seconds or milliseconds.

[0067] It should be noted that, This represents the change in the original magnetic field signal within the time window, distinct from the equivalent magnetic field response obtained after zero bias correction and direction correction in step 23. In step 23, the equivalent magnetic field response has eliminated the sensor's static background and the ambient magnetic field, and mapped it to a unified comparison direction, thus providing... The uncorrected raw signal variation is used only for external common-mode magnetic field interference identification.

[0068] when Higher than the preset correlation threshold When the common-mode change does not correspond as expected to the change in the equivalent current I flowing through the target connection node, the system determines it to be external common-mode magnetic field interference; where, The correlation threshold for determining external common-mode magnetic field interference can be set through environmental magnetic field disturbance calibration or engineering experience, and it is also a dimensionless parameter. The length of the preset time window is used for calculation. The sampling interval parameter can be set according to the typical duration of external magnetic interference, and it is not used as a calculation variable in the relevant system formula.

[0069] Step S512: When the magnetic sensor on either side of the target connection node experiences range saturation, sudden change in zero bias, excessive noise, abnormal sampling, or when the sensor detection data and the equivalent current flowing through the target connection node do not have a clear physical correspondence, it is determined that the sensor is faulty, the installation position is faulty, or the sampling link is abnormal, rather than the connection node is faulty. Among them, noise exceeding the limit means that the noise variance exceeds the noise threshold; sampling anomalies include lost sampling data and asynchronous sampling time; the mismatch between the magnetic sensor detection data and the equivalent current flowing through the target connection node means that the change in the magnetic sensor detection data and the change in the equivalent current do not have the expected physical correspondence.

[0070] It should be noted that the asymmetry in the magnetic sensor responses on both sides of the target connection node is not a basis for determining magnetic sensor failure or sampling link abnormality. When the magnetic sensor validity check passes, the asymmetry in the bilateral magnetic field response serves as the basis for calculating the local magnetic field spatial asymmetry coefficient in step S53 and participates in subsequent connection fault diagnosis.

[0071] Furthermore, when external magnetic interference, sensor failure, abnormal sensor installation, or abnormal sampling is detected, the connection fault diagnosis of the target node is not performed, and only equipment maintenance, anti-interference verification, or sensor status abnormality prompts are output; otherwise, it is determined that the magnetic field detection data of the target connection node meets the validity requirements of the connection fault diagnosis, and step 52 is executed to carry out the connection fault diagnosis process of the target connection node.

[0072] After confirming in step S51 that the magnetic sensor array is effective and external magnetic interference does not exceed the diagnostic allowable range, the spatial asymmetry coefficient obtained in step S2 is used as the basis for further steps. and observation features Multidimensional state magnetic field reference model output and normalized reference residuals It executes a comprehensive fault determination logic process that includes feature classification, suspected early warning, and multi-source confirmation.

[0073] Step S52: Normalize the reference residual The absolute value and the normal tolerance of the reference residual Compare the results to determine if the normalized reference residuals are normal. Step S52 is the one-dimensional determination of the residual, which serves as the fusion diagnostic index (CDI) in step S54 in the fusion weights. The one-dimensional degradation simplification case at time; the one-dimensional determination result of the normalized reference residual is used to characterize the degree of deviation of the current magnetic field observation response from the magnetic field reference response in the healthy connection state, and serves as the input for the fusion determination in step S54; Specifically, if the normalized reference residual is within the normal tolerance range of the reference residual, then the current magnetic field observation response is considered to match the magnetic field reference response in a healthy connection state.

[0074] like Figure 3 As shown, the current magnetic field observation characteristics lie in Determined healthy connection state magnetic field reference range Within this range, the normalized reference residual is determined to be within the normal tolerance range; if the normalized reference residual exceeds the normal tolerance range of the reference residual, it is considered that the current magnetic field observation response deviates from the magnetic field reference response in a healthy connection state; if... Figure 4 As shown, the current magnetic field observation characteristics Continuous deviation by The normalized reference residual of the defined healthy connection state magnetic field reference interval satisfies: At this point, the normalized reference residual is scaled using the reference residual normalization method. After scaling, step S54 is executed to calculate the fusion diagnostic index (CDI).

[0075] In this embodiment, the dimensionless normalized reference residual is processed by scaling. Based on the reference residual normalization scale Standardization enables residuals of different states and magnitudes to be compared under the same judgment scale.

[0076] It should be noted that, Figure 3 and Figure 4 These are schematic diagrams illustrating the normalization principle, used to explain the relationship between the current magnetic field observation characteristics and the magnetic field reference interval for a healthy connectivity state. The horizontal axis represents the normalized sampling time within the observation window, and the vertical axis represents the normalized magnetic field response after processing at the same scale. Both axes are dimensionless. Normalization is only used for illustrative purposes and does not alter the current magnetic field observation characteristics during actual diagnosis. Magnetic field reference response with healthy connection state The requirement is to maintain the same dimensions.

[0077] Optionally, the normal tolerance range of the reference residual corresponds to the magnetic field reference interval of the healthy connection state of the target connection node, which can be determined by the standard deviation, confidence interval, engineering experience limit, or model prediction error of the health status calibration data. For example, the magnetic field reference interval of the healthy connection state can be set as the ±3σ interval of the health calibration residual, or segmented tolerances can be established according to different SOC, temperature, aging stage, or installation method.

[0078] Determine the normalized reference residual During the process, you can set a normal state, a boundary observation state, and an abnormal deviation state. The specific judgment conditions are as follows: 1.1) When the absolute value of the normalized reference residual Less than or equal to the reference residual normal tolerance At that time, it is determined that the current magnetic field observation response is in a healthy connection state within the magnetic field reference interval, and the normalized reference residual is within the normal tolerance range; 1.2) When the absolute value of the normalized reference residual Greater than the reference residual normal tolerance The normalized reference residual is determined to have a single deviation, and is input into step S54 to participate in the calculation of the fusion diagnostic index; the single deviation is not used as the final diagnostic conclusion alone; 1.3) When the fusion diagnostic index meets the requirements When, according to step S54, it is determined that the target connection node is in the boundary observation state; when And the reference residual contribution satisfies When the reference residual contribution is determined to be dominant, step S552 is performed to confirm the abnormal deviation of the magnetic field observation response.

[0079] in, The reference residual normal tolerance is used to limit the normal range of variation of the normalized reference residual under healthy connection conditions; As a reference residual normalization scale, used for normalized reference residuals Scaled processing is performed to obtain the relative deviations involved in the calculation of the fusion diagnostic index (CDI). This is due to the normalized reference residuals. Since it is a dimensionless quantity, therefore, and All of these are dimensionless parameters.

[0080] In step S52, the normalized reference residual is calculated based on the current magnetic field observation characteristics and the healthy connection state magnetic field reference response output by the multidimensional state magnetic field reference model. Since the multidimensional state magnetic field reference model has incorporated the influence of normal operation factors such as current, SOC, temperature, aging state, and clamping state changes on the magnetic field response, the normalized reference residual no longer simply reflects the magnitude change of the magnetic field amplitude, but rather reflects the degree of deviation of the current target connection node's magnetic field response from the healthy connection state under the same operating conditions.

[0081] Specifically, when a battery is at different states of charge (SOC), temperatures, or aging stages, even if there is no connection fault at the target connection node, the surrounding magnetic field response may fluctuate normally due to current distribution, structural thermal expansion and contraction, changes in clamping conditions, or slight changes in the relative position of sensors. The multidimensional state magnetic field reference model, by inputting the current battery operating state parameters, outputs a healthy connection state magnetic field reference response or a healthy connection state magnetic field reference range that matches the current operating conditions. This compensates for the magnetic field response changes caused by variations in normal operating conditions. The normalized reference residual calculated in this way can weaken the interference of changes in normal operating conditions on fault judgment, allowing the residual to primarily characterize the magnetic field response offset caused by abnormal factors such as loose connections, localized weld cracks, connecting piece cracks, conductive cross-section shrinkage, or changes in local current paths.

[0082] This embodiment, through the aforementioned multidimensional state magnetic field reference response decoupling process, can distinguish between changes in the healthy connection state magnetic field reference response caused by variations in SOC, temperature, aging, clamping conditions, or sensor installation geometry, and abnormal magnetic field responses caused by connection structure anomalies. This processing method gives the normalized reference residual a clear physical meaning and ensures that the current observed characteristics are dimensionally consistent with the multidimensional state magnetic field reference model, providing a stable, comparable, and interpretable characteristic basis for subsequent comprehensive fault determination by combining node-level local magnetic field spatial asymmetry coefficients.

[0083] Step S53: Calculate the local magnetic field spatial asymmetry coefficient. Normalized scale with spatial asymmetry of local magnetic field Compare the results to determine whether the local magnetic field spatial asymmetry coefficient is normal. This step involves a one-dimensional determination of the local magnetic field spatial asymmetry coefficient, which serves as the basis for the fusion weighting of the diagnostic index CDI in step S54. The one-dimensional degenerate simplification case of time; the one-dimensional determination result of the local magnetic field spatial asymmetry coefficient is used to characterize whether there are local deflection or flow around the target connection node, and serves as the input for the fusion determination in step S54; Specifically, based on the local magnetic field spatial asymmetry coefficients on both sides of the target connection node, it is determined whether the target connection node has spatial asymmetry anomalies. The determination conditions are as follows: 2.1) Local magnetic field spatial asymmetry coefficient Not less than the local magnetic field spatial asymmetry normalization scale If the duration exceeds the set digital filtering window, it is determined that the target connection node has spatial asymmetric anomaly characteristics; the spatial asymmetric anomaly characteristics are one of the inputs for the calculation of the fusion diagnostic index CDI and the identification of the dominant source of the anomaly in step S54, and are not output as a deterministic connection structure fault conclusion alone. 2.2) Local magnetic field spatial asymmetry coefficient Smaller than the local magnetic field spatial asymmetry normalization scale ,Right now The local magnetic field spatial asymmetry coefficient is determined to be normal. Among them, the local magnetic field spatial asymmetry normalization scale It can be determined based on the statistical distribution of the local magnetic field spatial asymmetry coefficient under healthy connection conditions, the noise level of the magnetic sensor, the sensor installation deviation, or the engineering calibration results.

[0084] Step S54: Based on the normalized reference residual and the local magnetic field spatial asymmetry coefficient Calculate the Fusion Diagnostic Index (CDI), and use the CDI to determine the connection faults of the target connection nodes to obtain the initial judgment results of the connection faults. Specifically, the normalized reference residuals are completed in steps S52 and S53 respectively. and the local magnetic field spatial asymmetry coefficient After the one-dimensional determination, the two are scaled to construct fusion diagnostic features and calculate the fusion diagnostic index CDI: ; ; ; in, To define the local magnetic field spatial asymmetry normalization scale, The set reference residual normalization scale; , These are the local magnetic field spatial asymmetry coefficients. Normalized reference residuals Dimensionless deviation after each normalization scale processing; use The absolute value is calculated so that any anomalies, whether the current magnetic field observation response is higher or lower than the magnetic field reference response in a healthy connection state, can be reflected, thus avoiding the mutual cancellation of negative anomalies and positive deviations from spatial asymmetry in the fusion calculation. , These are the fusion weights for the local magnetic field spatial asymmetry characteristics and the normalized reference residual, respectively. These weights can be preset fixed weights or determined segmentally based on system current I, SOC, temperature, sampling noise, historical health calibration data, or false alarm rate requirements. They can be reduced under low current or low magnetic field signal-to-noise ratio conditions. Or extend the decision window.

[0085] Preferably, the fusion weights , are all non-negative numbers and satisfy . Under the same weight scale, the grading threshold of the first fusion diagnosis index CDI1 and the grading threshold of the second fusion diagnosis index CDI2 are calibrated, with CDI1<CDI2, which is used to realize the magnitude judgment of the fusion diagnosis index CDI; Further, when the fusion diagnosis index CDI is not less than the grading threshold of the second fusion diagnosis index CDI2, the proportion of spatial asymmetry contribution of the local magnetic field is defined as and the reference residual contribution proportion : ; ; Wherein, ε is a stabilization term for preventing unstable calculation of contribution proportion caused by excessively small CDI, and its value is far smaller than CDI2; 、 are respectively used to compare the relative contribution magnitudes of the local magnetic field spatial asymmetry feature and the normalized reference residual to the current fusion diagnosis index. Since they adopt the same denominator, their ratio can reflect the relative relationship between the contribution amounts of the two features. It should be noted that due to 、 introduces the stabilization term ε for preventing calculation instability into their denominators, the sum of the two is not required to be strictly equal to 1. The contribution proportion referred to in this embodiment is used to characterize and compare the relative contribution relationship of the two features to the fusion diagnosis index, and does not mean that the two constitute a strict percentage proportion.

[0086] To identify the dominant source of abnormality, a contribution dominant proportion is set , and . The contribution dominant proportion can be determined based on connection structure fault samples with altered current paths, magnetic field observation response abnormal deviation samples or historical operation data; as an example, can be taken as 1.5.

[0087] Connection fault judgment is performed on the target connection node based on the fusion diagnosis index CDI and the contribution proportions, so as to obtain the initial judgment result of the connection fault, and the judgment conditions are as follows: 3.1) When CDI<CDI1, it is determined that the target connection node is in a normal connection state, that is, the current magnetic field change falls within the normal magnetic field reference interval under the healthy connection state, and no connection fault alarm is output; 3.2) When CDI1≤CDI<CDI2, it is determined that the target connection node is in a boundary observation state, the sampling frequency is increased or the observation window is extended, and hysteresis judgment is performed in combination with the change trend of CDI within the observation window; 3.3) When CDI≥CDI2 and when it is determined that the local magnetic field spatial asymmetry contribution is dominant, it is preliminarily determined that the target connection node has a connection structure failure risk that changes the current path; the corresponding failures may include local bias current caused by loose bolts, flow around caused by local fatigue cracking of laser welds, cracks in connecting sheets, local deformation of busbars, local shrinkage of conductive cross-sections, or current redistribution caused by uneven contact pressure at connection nodes; 3.4) When CDI≥CDI2 and is satisfied, it is determined that the reference residual contribution is dominant, and it is preliminarily determined that the magnetic field observation response of the target connection node deviates abnormally, and the secondary confirmation process is triggered to distinguish between the geometric position change of conductive connectors, the relative installation position change of sensors, local deformation of busbars or structural looseness; 3.5) When CDI≥CDI2, and neither the above-mentioned local magnetic field spatial asymmetry contribution dominant condition nor the reference residual contribution dominant condition is satisfied, that is, and are both not satisfied, indicating that , have relatively close numerical values, and both types of features make non-negligible contributions to the current fusion diagnosis index, so it is determined as a mixed abnormality and the multi-source fusion confirmation process is entered; 3.6) When CDI does not reach the second fusion diagnosis index grading threshold CDI2, but the temperature rise of the connection node, connection voltage drop, branch voltage drop or single cell voltage is abnormal, it is determined that the target connection node has a contact degradation risk that does not significantly change the current path.

[0088] Wherein, CDI1 and CDI2 are grading thresholds of the fusion diagnosis index, which are dimensionless quantities and satisfy CDI1<CDI2; when is satisfied, the fusion diagnosis index CDI degenerates into a one-dimensional scaled result of the normalized reference residual: ; when is satisfied, the fusion diagnosis index CDI degenerates into a one-dimensional scaled result of the local magnetic field spatial asymmetry coefficient: . Under different weight configurations, CDI1 and CDI2 shall be calibrated according to healthy samples and abnormal samples under corresponding scales.

[0089] Step S55: Based on the initial judgment result of connection failure combined with multi-source physical state information, perform fault identification on the target connection node; The multi-source physical state information may include one or more of adjacent node magnetic field response, remote reference magnetic sensor status, connection node temperature, connection node temperature rise rate, connection node voltage drop, battery connection branch voltage drop, single cell voltage change, local temperature response or maintenance history data, and multi-source confirmation is performed on the target connection node; Specifically, based on the classification results of the Fusion Diagnostic Index (CDI) and the dominant relationship between the contribution ratio of spatial asymmetry and the contribution ratio of the reference residual, the target connection node is further classified into normal state, boundary observation state, suspected contact degradation risk, connection structure failure, abnormal deviation of magnetic field observation response, mixed anomaly, or pending verification state based on multi-source physical state information, and corresponding early warning information or handling suggestions are output, including the following: Step S551, Confirmation of Suspected Contact Deterioration Risk: When the Fusion Diagnostic Index (CDI) does not reach the second fusion diagnostic index grading threshold (CDI2) (i.e., it is in a normal connection state or only in a boundary observation state), but at least one of the synchronously acquired connection node temperature rise rate, connection node voltage drop, branch voltage drop, single cell voltage change, or local temperature response is abnormal, it is determined that the target connection node has a suspected contact degradation risk. Among them, contact degradation risk refers to the abnormal temperature rise or abnormal voltage drop that occurs when the target connection node has not changed its original current path, but due to increased contact resistance, decreased contact pressure or deterioration of the contact interface.

[0090] For such suspected contact degradation risks, further confirmation can be made through infrared temperature measurement, retesting of connection voltage drop, verification of fastening force, maintenance inspection, or verification of subsequent operating trends.

[0091] Step S552, Confirmation of abnormal deviation in magnetic field observation response: When the fusion diagnostic index CDI reaches or exceeds the second fusion diagnostic index grading threshold CDI2, and the reference residual contribution ratio meets the following conditions: At that time, it was determined that the reference residual contribution was dominant, and it was initially determined that there was an abnormal deviation in the magnetic field observation response of the target connection node. By combining one or more of the following: changes in the readings of the remote reference magnetic sensor, consistency of the magnetic field response of adjacent nodes, temperature stability of the connection node, voltage drop of the target connection node, voltage drop of the branch in which it is located, and maintenance history data, the cause of the abnormal deviation was confirmed, in order to determine whether the abnormal deviation originated from one or more of the following: changes in the geometric position of the conductive connector, changes in the relative installation position of the magnetic sensor, local deformation of the busbar, or structural loosening.

[0092] When multi-source confirmation indicates that the deviation originates from a non-faulty relative position drift and the sensor validity check passes, the multi-dimensional state magnetic field reference model, model parameters, or healthy connection state magnetic field reference range are updated in a controlled manner; when structural loosening or deformation risk is confirmed, an abnormal deviation warning of the magnetic field observation response is output.

[0093] Step S553, Fault confirmation of changing current path connection structure: When the fusion diagnostic index CDI reaches or exceeds the second fusion diagnostic index grading threshold CDI2, and the contribution ratio of local magnetic field spatial asymmetry meets the following conditions: when it is determined that the contribution from spatial asymmetry of the local magnetic field is dominant, it is preliminarily determined that the target connection node has a risk of connection structure fault that changes the current path. Multi-source confirmation is performed by combining one or more of temperature rise of the connection node, temperature rise rate of the connection node, voltage drop of the connection node, voltage drop of the branch where the connection node is located, magnetic field response distortion of adjacent nodes, individual voltage change or maintenance historical data, so as to confirm the fault type; connection structure fault types can include one or more of bolt looseness, weld cracking, connecting sheet fracture, busbar deformation or loss of contact pressure of the connection structure, and the severity of the connection structure fault is further evaluated in combination with the abnormal duration and current load, and suggestions for current limiting, power reduction or shutdown maintenance are output according to the evaluation result.

[0094] Step S554: When the fusion diagnosis index CDI is in the boundary observation classification (CDI1≤CDI<CDI2), the observation window is extended, and multi-source confirmation is performed by combining auxiliary features in the observation window, wherein the auxiliary features include one or more of temperature rise trend of the connection node, voltage drop trend of the connection node, magnetic field response evolution of adjacent nodes or local temperature response trend, and the multi-source confirmation is as follows: When any auxiliary feature develops from a boundary state to exceed the corresponding threshold, according to the spatial asymmetric contribution ratio, reference residual contribution ratio and multi-source confirmation result, the target connection node is upgraded and determined as connection structure fault or abnormal deviation of magnetic field observation response; When all auxiliary features in the observation window remain stable and none exceed the corresponding thresholds, the boundary observation state is cancelled, and the target connection node is restored to be determined as a normal state.

[0095] Step S555: Processing of mixed abnormality and pending recheck state: When the fusion diagnosis index CDI reaches or exceeds the second fusion diagnosis index classification threshold CDI2, and the contribution ratio of local magnetic field spatial asymmetry , When both conditions are satisfied, it indicates that the contribution ratio of local magnetic field spatial asymmetry and the contribution ratio of reference residual are relatively close, and both types of features have non-negligible contributions to the current fusion diagnosis index, so it is determined as mixed abnormality. For mixed abnormality, multi-source fusion confirmation is performed by combining one or more of adjacent node magnetic field response, remote reference magnetic sensor state, connection node temperature, connection node temperature rise rate, connection node voltage drop, branch voltage drop, individual battery voltage change or maintenance historical data. When it is still insufficient to make a definite judgment based on the above multi-source information, a pending recheck state is output, no definite connection fault conclusion is output, and one or more of extending the observation window, increasing the sampling frequency, supplementing auxiliary detection or manual maintenance recheck are triggered until the multi-source information meets the judgment conditions.

[0096] As an additional safety screening function of the connection fault diagnosis system in this embodiment, in addition to the external connection node fault diagnosis process in step S5, a static cell internal short circuit auxiliary safety screening process can also be set to provide auxiliary early warning for the risk of suspected internal micro short circuit or abnormal self-discharge of the cell when the system is in a static state. Step S5 is applicable when the battery system is in a current-carrying condition, to diagnose the connection status of external connection nodes such as busbars, terminals, bolts, and welds, in order to identify connection faults such as loose connections, contact deterioration, structural fractures, weld cracks, or busbar deformation. The static internal short circuit (ISC) auxiliary safety screening process is applicable when the battery system is in a no-current or near-no-current-carrying static condition, to assist in screening for suspected micro-short circuits or abnormal self-discharge risks inside the cell body.

[0097] In terms of diagnostic logic, the static ISC-assisted safety screening process constitutes a complementary diagnostic process to step S5 under static conditions. It is not a necessary step for step S5 to diagnose connection faults in external connection nodes, nor does it change the execution logic of steps S1 to S5.

[0098] In one implementation, the static ISC-assisted safety screening process can be performed when the battery system has not identified any external connection node faults during the previous current-carrying operation phase, and both sensor validity checks and external magnetic field interference checks have passed, and the battery system subsequently enters a resting condition that meets preset conditions. By distinguishing between the current-carrying condition and the resting condition, the interference caused by the magnetic field signal formed by the current of the external connection node and its abnormal current path on the auxiliary screening of internal cell anomalies can be reduced, thereby reducing the risk of confusion between external connection anomalies and internal cell anomalies.

[0099] As an optional implementation, the static ISC-assisted security screening process includes the following steps: Step S61: When the vehicle equipped with the battery system is turned off or the energy storage system stops, the main relay disconnects, and the total current of the battery system meets the requirements. And continue for more than the set time Furthermore, after the effects of cell polarization decay over a preset resting time, the battery system enters a static ISC auxiliary screening state. Where I is the total system current, in amperes; The current threshold for static determination is measured in amperes and is used to determine whether the battery system is in a state of no current or near no current. Minimum settling time required to proceed with static screening.

[0100] Step S62: Obtain the current static magnetic field value of the magnetic sensor near the cell body monitoring area and / or connection node, and combine it with the healthy static magnetic field reference value and the remote reference magnetic sensor signal to obtain the static residual magnetic field characteristics. The formula can be: ; in, The target magnetic sensor's current static magnetic field measurement value. This corresponds to the reference value for a healthy static magnetic field. This is the current static magnetic field measurement value from the remote reference magnetic sensor. This is the reference magnetic field value of the remote reference magnetic sensor in a healthy static state, and the units are consistent with the magnetic induction intensity. The environmental magnetic field compensation coefficient is used to characterize the coupling ratio of the remote reference magnetic field to the disturbed target measurement point. It is usually a dimensionless parameter and can be obtained through health calibration or online calibration.

[0101] Step S63: Based on the static residual magnetic field characteristics of the historical effective resting phase, generate an adaptive residual threshold. The formula is: ; in, and These are the mean and standard deviation of the static residual magnetic field characteristics during the historical effective static stage, respectively, with units consistent with magnetic induction intensity; The statistical margin coefficient is a dimensionless parameter that can be calibrated according to the false alarm rate requirements. This is the minimum residual threshold, used to avoid the threshold being too small in extremely low noise environments.

[0102] Step S64: When the characteristic amplitude of the static residual magnetic field... Exceeding the adaptive residual threshold generated based on historical health resting data And it continues beyond the set static residual magnetic field time window. When the output cell is suspected of having an internal micro-short circuit or abnormal self-discharge risk, it will trigger one or more of the following for secondary verification: static retest, voltage self-discharge analysis, temperature verification, or manual inspection. in, This is a window for the duration of static residual magnetic field anomalies, used to avoid false alarms caused by transient noise or short-term external disturbances. The auxiliary warning is not directly used as a definitive diagnostic conclusion of an internal short circuit in the battery cell.

[0103] Furthermore, after the system outputs an auxiliary warning for abnormal static residual magnetic field, it further triggers the BMS to perform a secondary verification based on one or more auxiliary characteristics such as static voltage decay, OCV recovery characteristics, static temperature rise, capacity retention rate, or AC impedance characteristics. When the abnormal static residual magnetic field and at least one auxiliary characteristic are both abnormal, the system increases the risk level of the corresponding cell and outputs an enhanced warning indicating a suspected internal micro-short circuit or abnormal self-discharge risk. When only the abnormal static residual magnetic field exists, and the static voltage decay, OCV recovery characteristics, static temperature rise, capacity retention rate, or AC impedance characteristics are not abnormal, a definitive conclusion of internal short circuit is not output. Instead, the system remains under observation and continues to verify during subsequent effective static periods.

[0104] In this embodiment, under normal circumstances, there should be no continuous abnormal local residual magnetic field changes near the target magnetic sensor when the battery system is in a state of no current or near no current. If, after excluding the effects of system operating current, ambient magnetic field disturbances, and static transients, a static residual magnetic field exceeding the threshold still appears at the target measurement point, it may indicate an abnormal self-discharge path, weak internal leakage current, or other abnormal electrochemical behavior within the battery cell. By introducing a remote reference magnetic sensor for ambient magnetic field compensation, the impact of external magnetic field changes on the target measurement point can be reduced; by introducing historical valid static data to generate an adaptive residual threshold, misjudgments caused by differences in sensor noise, installation location, and battery pack background magnetic field can be reduced; by combining static voltage decay, OCV recovery characteristics, static temperature rise, capacity retention rate, or AC impedance characteristics for secondary verification, a definitive conclusion of internal short circuit can be avoided based solely on magnetic field anomalies, thereby improving the reliability and safety of auxiliary early warning.

[0105] The battery connection fault diagnosis method in this embodiment can improve the ability to distinguish between connection structure faults and normal changes in the magnetic field reference response of a healthy connection state. Firstly, by combining the local magnetic field spatial asymmetry coefficient on both sides of the connection node, a multi-dimensional state magnetic field reference model, and direction correction logic, it distinguishes between the current deviation or circumferential characteristics caused by connection loosening, local weld cracking, connection plate cracking, or local shrinkage of the conductive cross-section, and the magnetic field reference response of a healthy connection state caused by changes in SOC, temperature, cyclic aging, clamping state changes, or changes in sensor installation geometry. Compared to alarms based solely on a single threshold value for magnetic field amplitude, this method helps reduce the risk of false alarms caused by normal structural changes or installation gap drift during long-term operation of large-capacity batteries.

[0106] For connection structure faults that alter the local current path, such as loose bolts, localized weld cracks, localized busbar deformation, and cracked connectors, the system detects localized current deviation or shunting phenomena by observing asymmetrical changes in the magnetic field response on both sides of the connection node. Compared to detection methods that rely solely on NTC temperature rise or abnormal total voltage drop at the connection point, this system can provide early warnings of connection node abnormalities in certain fault scenarios, prior to significant temperature rise or total voltage drop alarms. This provides earlier auxiliary information for BMS current limiting, power reduction, maintenance prompts, or safety protection.

[0107] Furthermore, by introducing a magnetic sensor validity check, a remote reference magnetic sensor interference identification, and a zero-bias safety update mechanism before fault determination, the impact of external common-mode magnetic fields, sensor saturation, zero-bias drift, installation abnormalities, or sampling synchronization abnormalities on diagnostic results can be reduced. By prohibiting the updating of residual magnetic fields to new zero bias during suspected fault states and static auxiliary screening stages, it is also possible to prevent real weak abnormal signals from being canceled out by algorithm errors. The sensor in this embodiment can adopt an attached FPC sensing structure, a module cover fixed sensing structure, or other equivalent sensor mounting structures, and can be synchronously fused with the current, SOC, temperature, voltage, and temperature rise information of the BMS. It does not require the additional deployment of a large number of high-voltage sampling lines at each connection node and can be used as an auxiliary safety monitoring unit in energy storage battery modules, large-capacity square battery modules, or power battery packs.

[0108] To illustrate the implementation process and effects of the method in this embodiment, this embodiment takes the application in an energy storage battery module as an example to explain how to distinguish between connection structure faults that change the current path, normal changes in magnetic field response under healthy connection conditions, and contact degradation that does not significantly change the current path by using the asymmetric features of the local magnetic field space on both sides of the connection node, a multi-dimensional state magnetic field reference model, and the fusion of auxiliary state information.

[0109] The application scenario configuration is as follows: This embodiment is applied to a 3P2S battery module in a commercial energy storage system, such as... Figure 2 As shown, three individual battery cells 1 are connected in parallel to form a parallel unit, including a left parallel unit and a right parallel unit. The left parallel unit includes battery cells Cell1 to Cell3, and the right parallel unit includes battery cells Cell4 to Cell6. The two parallel units are connected in series to form a module. Sensing components 3 are respectively set at the connection point between the two parallel units, the first busbar 2, or the connecting piece, including a left magnetic sensor 4 and a right magnetic sensor 5. A first reference magnetic sensor 7 is set at the remote end. The left magnetic sensor 4, the right magnetic sensor 5, and the first reference magnetic sensor 7 are connected to the BMS and the diagnostic terminal 8. The MCU of the diagnostic terminal executes the diagnostic process of steps S1 to S5.

[0110] Battery cell 1 is a square lithium iron phosphate battery with a single-cell capacity of 280Ah, and the equivalent capacity of the module is 840Ah. The parallel units within the module are connected in series and parallel via busbars and bolts. To reduce the influence of hysteresis or remanence of the connectors on magnetic field measurements, the series bridging nodes are preferably secured using low-permeability austenitic stainless steel connectors, non-ferromagnetic connectors, or low-magnetic connectors selected through magnetic screening.

[0111] This embodiment employs a fixed sensor structure on the battery module cover. The magnetic sensor array is integrated onto a rigid PCB board or insulating bracket and fixed to the inside of the battery module insulating cover 6 or high-voltage protective shield. A pair of left and right magnetic sensors are installed above each target connection node, symmetrically arranged along the bus width or on both sides of the target conductive path. Since the position of the sensors relative to the module cover is relatively stable in this installation method, while the cell terminals and bus below may undergo repeatable small displacements with SOC, temperature, cycle count, or clamping state, it is necessary to compensate for normal changes in the magnetic field response under healthy connection conditions using a multi-dimensional state magnetic field reference model.

[0112] The calibration and comprehensive diagnosis process is as follows: First, calibration of the multidimensional state magnetic field reference model and establishment of initial zero bias; During the factory testing phase of the energy storage module, the module was placed in an ambient temperature chamber and subjected to standard charge and discharge calibration conditions at 25℃, 45℃, and different SOC platforms from 10% to 100%. Simultaneously, the outputs of the magnetic sensors on both sides of the target connection node, the total current I of the battery system, SOC, temperature T, and relevant auxiliary status information provided by the BMS were collected.

[0113] Zero-bias correction and direction correction are performed on the magnetic sensor outputs on both sides of the target connection node to obtain the equivalent magnetic field response on both sides. , Since the left and right measuring points are located on opposite sides of the main conductive path in this embodiment, and the tangential magnetic field directions of the target are opposite under health calibration conditions, a direction correction coefficient can be set. In other module structures, the orientation correction coefficient can also be determined based on the health calibration conditions.

[0114] In this embodiment, the magnetic field stability term constant Use 0.01mT. This value is only an example calibration parameter and can be adjusted according to the magnetic sensor's noise floor, ADC resolution, or health status noise envelope.

[0115] Based on data collected by the health module under different current, SOC, and temperature conditions, a multidimensional state magnetic field reference model is established. This model can employ a lookup table model, a polynomial fitting model, a multivariate nonlinear regression model, or a machine learning prediction model. The model output maintains dimensional consistency with the currently observed characteristics. In this embodiment, the system uses equivalent magnetic field amplitude characteristics: ; And based on Establish a multidimensional state magnetic field reference model Simultaneously, a reference range for the magnetic field of the healthy connection state is set based on the health calibration residuals, for example, by using the standard deviation or empirical tolerance of the health calibration residuals to generate the reference range for the magnetic field of the healthy connection state.

[0116] Once the module is connected to the actual operating system, under the conditions that the main relay is not closed, the battery cell is stationary, and the system has not recorded any suspected faults, the ambient background magnetic field is collected as the initial zero-bias reference. Subsequently, updates to the zero-bias or multi-dimensional state magnetic field reference model are only permitted when conditions such as no suspected faults, no external strong magnetic interference, and sensor effectiveness checks are passed, to prevent early fault characteristics from being learned as part of the healthy connection state magnetic field reference response.

[0117] Second, decoupling of normal changes in magnetic field response under healthy connection conditions; In its healthy state during the early stages of its lifespan, the module operates under 1C standard discharge conditions, with a total current of approximately 840A. The main current is approximately uniformly distributed at the target bus connection node. After removing the zero bias, the target magnetic field components measured by the magnetic sensors on both the left and right sides are as follows: ; After orientation correction: ; Calculate the spatial asymmetry coefficient: ; This value is much lower than the local magnetic field spatial asymmetry normalization scale set in this embodiment. This indicates that the current paths on the left and right sides of the node are basically symmetrical, and there is no obvious local deflection or flow around the node.

[0118] As the energy storage module undergoes multiple deep charge-discharge cycles, changes in cell expansion, clamping conditions, or minute elastic displacements of the busbar may alter the effective spatial relationship between the target conductor and the magnetic sensor fixed to the top cover. Under a high SOC condition, such as T=35℃ and SOC=90%, the module still operates at 840A. At this time, the equivalent magnetic field amplitudes on the left and right sides after direction correction become: ; Compared to the initial healthy state, the overall magnetic field amplitude exhibits a repeatable drift of approximately 18%. If only a single magnetic field amplitude threshold is used for judgment, this drift may be misjudged as an abnormal system current, sensor malfunction, or deformation of the connection structure.

[0119] However, in the diagnostic logic of this embodiment, the spatial asymmetry coefficient is calculated first: ; This value is still relative to the normalization scale. The low level indicates that no current deviation has occurred at this node, altering the current path. Next, the system invokes a multi-dimensional state magnetic field reference model to calculate the healthy connection state magnetic field reference response under these current, SOC, and temperature conditions. And further calculate the normalized reference residual. .like If the value is within the normal tolerance range of the reference residual, it is classified as a normal change in the magnetic field response under healthy connection conditions, and no connection fault alarm is output.

[0120] Therefore, when changes in health status cause the magnetic field amplitudes on both sides to change approximately proportionally, the spatial asymmetry coefficient remains at a low level; and the multidimensional state magnetic field reference model can explain the overall amplitude drift. Combining these two approaches can reduce the risk of misinterpreting normal changes in the healthy connection state magnetic field reference response as connection failure, busbar deformation, or current deviation faults.

[0121] Third, connection fault determination based on fusion diagnostic index; Take the local magnetic field spatial asymmetry normalization scale Reference residual normalization scale Fusion diagnostic index grading thresholds: CDI1=0.3, CDI2=0.7, fusion weight. Stability term ε=0.02, dominant contribution ratio The above parameters are merely exemplary calibration parameters and can be adjusted according to the background noise of the magnetic sensor, the statistical characteristics of the health calibration residual, the false alarm rate requirements, or the operating condition segments. They do not constitute the only limitation on the solution of this embodiment.

[0122] To illustrate the relative relationship between the magnetic field reference response in a healthy connectivity state, the magnetic field reference interval in a healthy connectivity state, and the current magnetic field observation characteristics, Figure 3 and Figure 4 The data is plotted in a normalized form. The horizontal axis represents the normalized sampling time within the observation window, and the vertical axis represents the normalized magnetic field response after processing at the same scale. Both axes are dimensionless. The solid black line represents the current magnetic field observation characteristics. The solid blue line represents the magnetic field reference response in a healthy connection state. The two dashed lines represent the upper and lower bounds of the magnetic field reference interval for a healthy connection state, respectively.

[0123] Scenario 1: Normal changes in the magnetic field response under healthy connection conditions, such as... Figure 3 As shown; Under the aforementioned magnetic field amplitude drift condition of approximately 18%, the spatial asymmetry coefficient α ≈ 0.003, thus... Normalized reference residuals after reference residual normalization scaling: The fusion diagnostic index is: Therefore, if the target connection node is determined to be in a normal connection state, no abnormality source identification is performed. For example... Figure 3 As shown, the current magnetic field observation characteristics are within the magnetic field reference range of the healthy connection state. The magnetic field amplitude drift is decoupled by the multidimensional state magnetic field reference model into a normal change in the magnetic field response under the healthy connection state, and no connection fault alarm is output.

[0124] Scenario 2: Faulty connection structure that alters the current path, resulting in asymmetric contribution of local magnetic field. leading; Suppose the equivalent magnetic field response on both sides under the condition of a loose bolt: ;have to , ;set up ;but ,further, , ,because The system determines that the local magnetic field spatial asymmetry is the dominant factor, and initially determines that the target connection node has a connection structure fault risk that changes the current path. Then, it proceeds to step S553 for multi-source confirmation.

[0125] Scenario 3: Abnormal deviation in magnetic field observation response; Suppose a sensor is in a state of relative position change. ,but ;at the same time The calculation at the current moment yields: Because the fusion diagnostic index at the current moment satisfies Therefore, the current target connection node is in a boundary observation state. At this time, increase the sampling frequency or extend the observation window, and continuously track changes in the current magnetic field observation characteristics, normalized reference residuals, and fusion diagnostic index. When the current magnetic field observation characteristics are within subsequent observation windows... When the magnetic field deviates continuously from the reference range of the healthy connectivity state, and the fusion diagnostic index (CDI) rises to or exceeds the second fusion diagnostic index grading threshold (CDI2), the contribution ratio of local magnetic field spatial asymmetry is further calculated. and the contribution ratio of the reference residual .like Figure 4As shown, when the fusion diagnostic index CDI reaches or exceeds the second fusion diagnostic index grading threshold CDI2, and satisfies: If the reference residual contribution is dominant, it is initially determined that there is an abnormal deviation in the magnetic field observation response of the target connection node, and then proceeds to step S552 for secondary confirmation to distinguish between changes in the geometric position of the conductive connector, changes in the relative installation position of the magnetic sensor, local deformation of the busbar, or structural loosening.

[0126] Scenario 4: Risk of exposure to degradation; When the CDI does not reach the second fusion diagnostic index grading threshold CDI2, but the connection node temperature rise rate, connection node voltage drop, or individual voltage change is abnormal, proceed to step S551 and determine it as a contact degradation risk that has not significantly changed the current path but is accompanied by abnormal temperature rise or voltage drop.

[0127] Fourth, fault detection by changing the connection structure of the current path; During subsequent operation, due to long-term vibration, alternating heating and cooling, or weakening of fastening force, a locking nut at a certain target connection node loosened, causing a decrease in local contact pressure and uneven changes in the effective conductive area of ​​the busbar-terminal contact region. At this time, the main current is significantly biased to one side of the conductive area compared to the healthy state, forming a local bias current.

[0128] Under the next high-current discharge condition, after zero-bias correction and direction correction, the equivalent magnetic field amplitude measured by the left magnetic sensor increases to: ; The equivalent magnetic field amplitude measured by the magnetic sensor on the right decreases to: ; Calculate the spatial asymmetry coefficient: ; This value exceeds the scale normalized by the spatial asymmetry of the local magnetic field. After normalization, a high spatial asymmetry term is formed, and the duration of this anomaly exceeds the set filtering window, for example, 500ms. After checking the effectiveness of the sensor and eliminating external interference, the system determines that the connection node has a risk of connection structure failure that changes the current path, and outputs an early warning message "Node ID-04 has a risk of loose current-biased connection" through CAN communication or BMS diagnostic interface.

[0129] This early warning does not rely on the obvious temperature rise or total voltage drop of the connection node reaching the auxiliary alarm condition. Instead, it captures the redistribution of the current path based on the asymmetrical change of the local magnetic field response on the left and right sides. Therefore, it can be used for early monitoring of faults that change the current path, such as bolt loosening, uneven local contact pressure of the connecting piece, and local shrinkage of the conductive cross section.

[0130] Fifth, the judgment of contact degradation fusion without significant change in current path; At another target connection node, if a relatively uniform oxide film forms due to moisture intrusion, surface contamination, or aging of the contact interface, the contact resistance of that node may increase, but the macroscopic conductive path does not show a significant left-right shift. This type of fault may not cause significant local magnetic field spatial asymmetry.

[0131] At this point, the spatial asymmetry coefficients calculated by the magnetic sensors on both sides of the node after direction correction are, for example: ; This value is within the normal range. Meanwhile, the system is based on... or The calculated normalized reference residual It is also within the reference range of the magnetic field in a healthy connection state, indicating that the magnetic field characteristics do not show any deflection or circumferential flow phenomena that change the current path.

[0132] However, the system's auxiliary status input module synchronously acquires information such as connection node temperature, connection node voltage drop, branch voltage drop, or individual unit voltage provided by the BMS or independent sampling unit. It detects an abnormally high rate of temperature rise reported by the temperature sensor near the node, or that the connection voltage drop of the branch containing the node exceeds the health statistical limit. For example, the following voltage drop anomaly criterion can be used: ; in, The voltage drop across the target connection node or its branch. and These are the mean and standard deviation of voltage drop for similar connected nodes under healthy conditions or during historical operating periods. Node temperature rise rate limits, branch voltage drop thresholds, or abnormal voltage trends in individual units can also be used as supplementary criteria for judgment.

[0133] In the above situation, it is not judged as a connection structure failure that changes the current path. Instead, the diagnosis result is "the node has a risk of contact deterioration that does not significantly change the current path but is accompanied by abnormal temperature rise or voltage drop". Further confirmation is recommended through infrared temperature measurement, retesting of connection voltage drop, verification of fastening force or maintenance inspection.

[0134] As can be seen from this embodiment, the method distinguishes four types of situations: First, when the left and right magnetic fields change approximately proportionally and the normalized reference residual is within the normal tolerance range, it is classified as a normal change in the magnetic field response under healthy connection conditions; Second, when the fusion diagnostic index reaches the abnormal classification condition and the local magnetic field spatial asymmetry contribution dominates, it is classified as a connection structure failure risk that changes the current path; Third, when the fusion diagnostic index reaches the abnormal classification condition and the reference residual contribution dominates, it is classified as an abnormal deviation in the observed magnetic field response; Fourth, when the magnetic field characteristics are not obviously abnormal but the temperature rise or voltage drop is abnormal, it is classified as a contact degradation risk that does not significantly change the current path. When neither the local magnetic field spatial asymmetry contribution nor the reference residual contribution meets the dominant condition, it is judged as a mixed anomaly and enters multi-source verification. This classification logic shows that the method of this embodiment does not simply rely on the magnetic field amplitude threshold, but rather fuses and judges the local magnetic field asymmetry characteristics, the multi-dimensional state magnetic field reference model, the normalized reference residual, and BMS auxiliary information.

[0135] Example 2 Based on Embodiment 1, this embodiment provides a battery connection fault diagnosis system based on magnetic field asymmetry state decoupling, including: The data acquisition module is configured to acquire local magnetic field signals at symmetrical positions on both sides of the target connection node of the battery to be diagnosed, as well as battery operating status parameters. The magnetic field asymmetry identification module is configured to perform zero-bias correction and direction correction on the local magnetic field signals on both sides of the target connection node, obtain the equivalent magnetic field response on both sides of the target connection node, and then calculate the current magnetic field observation characteristics of the target connection node and the local magnetic field spatial asymmetry coefficient on both sides of the target connection node. The health reference construction module is configured to build a multi-dimensional state magnetic field reference model for each target connection node, which characterizes the relationship between the battery operating state and the magnetic field of the battery target connection node. The current battery operating state parameters are input into the multi-dimensional state magnetic field reference model to obtain the health connection state magnetic field reference response of the target connection node in the current operating state. The normalized reference residual calculation module is configured to calculate the normalized reference residual based on the current magnetic field observation characteristics of the target connection node and the magnetic field reference response of the healthy connection state. The fault diagnosis module is configured to perform connection fault diagnosis on the target connection node based on the node-level local magnetic field spatial asymmetry coefficient and the normalized reference residual, and obtain the diagnosis result.

[0136] It should be noted that each module in this embodiment corresponds one-to-one with each step in embodiment 1, and their specific implementation process is the same, so it will not be repeated here.

[0137] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0138] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A battery connection fault diagnosis method based on magnetic field asymmetry decoupling, characterized in that, Includes the following steps: Acquire the local magnetic field signals at symmetrical positions on both sides of the target connection node of the battery to be diagnosed, as well as the battery operating status parameters; Zero-bias correction and direction correction are performed on the local magnetic field signals on both sides of the target connection node to obtain the equivalent magnetic field response on both sides of the target connection node. Then, the current magnetic field observation characteristics of the target connection node and the local magnetic field spatial asymmetry coefficient on both sides of the target connection node are calculated. For each target connection node, a multi-dimensional state magnetic field reference model is constructed to characterize the relationship between the battery operating state and the magnetic field of the battery target connection node. The current battery operating state parameters are input into the multi-dimensional state magnetic field reference model to obtain the healthy connection state magnetic field reference response of the target connection node under the current operating state. Based on the current magnetic field observation characteristics of the target connection node and the magnetic field reference response of the healthy connection state, the normalized reference residual is calculated. Based on the node-level local magnetic field spatial asymmetry coefficient and the normalized reference residual, connection fault diagnosis is performed on the target connection node to obtain the diagnosis results. The local magnetic field signals on both sides of the target connection node are subjected to zero-bias correction and direction correction to obtain the equivalent magnetic field response on both sides of the target connection node, including the following steps: Obtain the zero-bias reference values ​​of the magnetic sensors on both sides of the target connection node; Based on the conductor orientation, current flow direction, magnetic sensor installation direction, and target magnetic field component direction at the target connection node, determine the orientation correction coefficients corresponding to the magnetic sensors on both sides. Based on the original sampled values ​​of the magnetic sensor, the zero bias reference value, and the direction correction coefficient, the equivalent magnetic field response on both sides of the target connection node is calculated.

2. The battery connection fault diagnosis method based on magnetic field asymmetry decoupling as described in claim 1, characterized in that, The local magnetic field signals on both sides of the target connection node are measured by magnetic sensors installed on both sides of the target connection node; The origin is taken as the geometric center, solder joint center, bolt center, or center of the contact area between the connecting piece and the terminal post of the target connection node. The x-axis is the direction of the main current flow in the busbar or connecting piece, and the z-axis is the direction perpendicular to the busbar surface and away from the busbar surface. The y-axis direction is determined according to the right-hand coordinate system, and the y-axis direction corresponds to the width direction of the battery busbar or connecting piece. Magnetic sensors are respectively set on both sides of the target connection node along the y-axis direction to collect the local magnetic field signals on both sides of the target connection node.

3. The battery connection fault diagnosis method based on magnetic field asymmetry state decoupling as described in claim 1, characterized in that, The magnetic sensor at the target connection node is set using an attached flexible circuit board structure. The magnetic sensor is set on the flexible circuit board and attached to the surface of the busbar or the connecting piece near the target connection node. An insulating isolation structure is set between the flexible circuit board and the metal busbar.

4. The battery connection fault diagnosis method based on magnetic field asymmetry state decoupling as described in claim 1, characterized in that, The amplitudes of the equivalent magnetic field responses on both sides of the target connection node are extracted separately, and the absolute value of the difference between the amplitudes of the equivalent magnetic field responses on both sides is calculated. The absolute value of the amplitude difference is divided by the sum of the amplitudes of the equivalent magnetic field responses on both sides to obtain the local magnetic field spatial asymmetry coefficient on both sides of the target connection node. .

5. The battery connection fault diagnosis method based on magnetic field asymmetry state decoupling as described in claim 1, characterized in that, The calculation of the current magnetic field observation characteristics of the target connection node includes the equivalent magnetic field amplitude characteristics. and current normalized response characteristics At least one of them; Equivalent magnetic field amplitude characteristics , is the average value of the equivalent magnetic field response amplitude on both sides of the target connection node, used to characterize the average magnetic field response level of local measurement points on both sides of the target connection node; Current normalized response characteristics The equivalent magnetic field amplitude characteristics at the target connection node. The normalized result relative to the equivalent current magnitude flowing through the target connection node.

6. The battery connection fault diagnosis method based on magnetic field asymmetry state decoupling as described in claim 1, characterized in that, The multidimensional state magnetic field reference model uses battery operating state parameters as input parameters and outputs the healthy connection state magnetic field reference response or the healthy connection state magnetic field reference interval at the target connection node under the state of no connection failure.

7. The battery connection fault diagnosis method based on magnetic field asymmetry state decoupling as described in claim 1, characterized in that, The difference between the current magnetic field observation characteristics of the target connection node and the magnetic field reference response in the healthy connection state is calculated. Based on the offset of the difference relative to the magnetic field reference response in the healthy connection state, the normalized reference residual is calculated.

8. The battery connection fault diagnosis method based on magnetic field asymmetry state decoupling as described in claim 1, characterized in that, Based on the local magnetic field spatial asymmetry coefficients and normalized reference residuals on both sides of the target connection node, connection fault diagnosis is performed on the target connection node to obtain the diagnosis results, including the following steps: Step S51: Perform preliminary verification on the magnetic field detection data of the target connection node to eliminate external common-mode magnetic field interference, sensor failure, installation abnormality or sampling link abnormality, and determine whether the magnetic field detection data of the target connection node meets the requirements for the effectiveness of connection fault diagnosis. If it is effective, proceed to the next step. Step S52: Normalize the reference residual The absolute value and the normal tolerance of the reference residual Compare the results to determine if the normalized reference residuals are normal. Step S53: Calculate the local magnetic field spatial asymmetry coefficient. Normalized scale with spatial asymmetry of local magnetic field Compare the results to determine whether the local magnetic field spatial asymmetry coefficient is normal. Step S54: Based on the normalized reference residual and the local magnetic field spatial asymmetry coefficient Calculate the Fusion Diagnostic Index (CDI), and use the CDI to determine the connection faults of the target connection nodes to obtain the initial judgment results of the connection faults. Step S55: Based on the initial judgment result of connection failure and combined with multi-source physical state information, identify the fault of the target connection node.

9. A battery connection fault diagnosis system based on magnetic field asymmetry decoupling, characterized in that, include: The data acquisition module is configured to acquire local magnetic field signals at symmetrical positions on both sides of the target connection node of the battery to be diagnosed, as well as battery operating status parameters. The magnetic field asymmetry identification module is configured to perform zero-bias correction and direction correction on the local magnetic field signals on both sides of the target connection node, obtain the equivalent magnetic field response on both sides of the target connection node, and then calculate the current magnetic field observation characteristics of the target connection node and the local magnetic field spatial asymmetry coefficient on both sides of the target connection node. The health reference construction module is configured to build a multi-dimensional state magnetic field reference model for each target connection node, which characterizes the relationship between the battery operating state and the magnetic field of the battery target connection node. The current battery operating state parameters are input into the multi-dimensional state magnetic field reference model to obtain the health connection state magnetic field reference response of the target connection node in the current operating state. The normalized reference residual calculation module is configured to calculate the normalized reference residual based on the current magnetic field observation characteristics of the target connection node and the magnetic field reference response of the healthy connection state. The fault diagnosis module is configured to perform connection fault diagnosis on the target connection node based on the node-level local magnetic field spatial asymmetry coefficient and the normalized reference residual, and obtain the diagnosis result. The local magnetic field signals on both sides of the target connection node are subjected to zero-bias correction and direction correction to obtain the equivalent magnetic field response on both sides of the target connection node, including the following steps: Obtain the zero-bias reference values ​​of the magnetic sensors on both sides of the target connection node; Based on the conductor orientation, current flow direction, magnetic sensor installation direction, and target magnetic field component direction at the target connection node, determine the orientation correction coefficients corresponding to the magnetic sensors on both sides. Based on the original sampled values ​​of the magnetic sensor, the zero bias reference value, and the direction correction coefficient, the equivalent magnetic field response on both sides of the target connection node is calculated.

Images