A system and method for adaptive tracking and compensation of baseline drift in a magnetic sensor

Abstract

The application discloses a kind of adaptive tracking compensation systems and methods of magnetic sensor baseline drift, it is related to current measurement technical field, the system includes magnetic sensor array, conservation constraint operation unit, baseline estimation unit and compensation injection unit;Conservation constraint operation unit is divided into conservation channel group to the magnetic measurement channel satisfying the determined linear conservation relation corresponding to the current to be measured, and conservation residual quantity is obtained by the weighted summation of the magnetic field response quantity in group according to the combination weight coefficient determined by geometric coupling coefficient;Baseline estimation unit updates baseline state quantity with conservation residual quantity as observation innovation, and constructs overdetermined observation equation group to solve the baseline state quantity of each magnetic measurement channel from multiple working condition points in sliding time window;Compensation injection unit deducts baseline state quantity from magnetic field response quantity to obtain compensation output quantity.The application has the advantages of online tracking and compensating baseline drift under the condition that the current to be measured exists continuously.

Description

Technical Field

[0001] This application relates to the field of current measurement technology, and in particular to an adaptive tracking compensation system and method for baseline drift of a magnetic sensor. Background Technology

[0002] In applications such as electricity metering, power gateway metering, current detection in photovoltaic and energy storage converters, and current sampling of power battery packs, continuous measurement of the current in current-carrying conductors is required. Magnetic sensors collect the magnetic field around the current-carrying conductor and inversely deduce the current based on geometric relationships. This eliminates the need to disconnect the circuit and connect the sampling element in series, and also facilitates the arrangement of multiple measurement points on the same cross-section. It is suitable for locations with limited space and requiring long-term continuous operation.

[0003] In addition to the component originating from the measured current, the output of a magnetic sensor also includes a zero-point component independent of the input, namely the baseline drift. This component is determined by packaging stress, magnetic domain relaxation, remanence of the core magnetization, and temperature changes. It shifts slowly over time and may undergo a step shift after a large current surge without recovering on its own. Since both energy accumulation and state of charge estimation are unidirectionally accumulated over time, the baseline drift exhibits a deviation that increases monotonically with time and cannot be traced retrospectively.

[0004] Existing suppression methods can be broadly categorized into four types. The first type involves periodically alternating the excitation and pickup directions of the sensing element, modulating the element's own offset to a high frequency for filtering, thus reducing the usable bandwidth to less than one-tenth of the alternation frequency, and eliminating the residual magnetism of the magnetic core outside the element from the alternation link. The second type involves calibrating the temperature curve at the factory and subtracting it back during operation, with temperature as the only independent variable, and the device characteristics being altered by packaging stress after mounting. The third type uses a feedback coil to pull the magnetic core back to the zero flux operating point, resulting in a size and static power consumption several times that of the open-loop scheme, while the zero flux reference is still determined by the drifting sensing element. The fourth type involves periodically cutting off the measured current or applying a demagnetizing pulse to reacquire the baseline, eliminating effective data within the zeroing window and preventing interference with drift between two zeroing points.

[0005] The common problem with the above methods is that the baseline drift is located in the low-frequency to quasi-DC range, while the measured current itself also contains a DC component. Since the two overlap in the frequency domain, frequency division on a single measurement channel cannot separate them. Therefore, the baseline can only be obtained under the premise that the input is known to be zero at a certain moment. However, the continuously operating current-carrying circuit does not provide this premise. Thus, by exchanging bandwidth, the static model at the time of manufacture, volume power consumption, and data continuity, there is no means to separate the baseline drift from the measured current component under the condition that the measured current is present normally and no new magnetic circuit hardware is added. Summary of the Invention

[0006] In order to separate the baseline drift from the measured current component under the condition that the measured current is continuously present and without adding magnetic circuit hardware, this application provides an adaptive tracking compensation system and method for the baseline drift of a magnetic sensor.

[0007] Firstly, this application provides an adaptive tracking compensation system for baseline drift of a magnetic sensor, employing the following technical solution: An adaptive tracking compensation system for baseline drift of a magnetic sensor, comprising: A magnetic sensing array includes multiple magnetic sensing units respectively disposed at multiple current-carrying conductors, each of the magnetic sensing units constituting a magnetic measurement channel and outputting a magnetic field response quantity; A conservation constraint operation unit, connected to the magnetic sensing array signal, is configured to divide several magnetic measurement channels whose corresponding measured currents satisfy a defined linear conservation relationship into conservation channel groups, determine a combination weight coefficient according to the geometric coupling coefficient of each magnetic measurement channel to each current-carrying conductor, and perform a weighted summation on each magnetic field response quantity within the conservation channel group according to the combination weight coefficient to obtain a conservation residual quantity. The combination weight coefficient causes the measured current component in each magnetic field response quantity to cancel each other in the weighted summation. The conservation residual quantity characterizes the linear combination of the baseline drift quantities of each magnetic measurement channel. The baseline estimation unit is signal-connected to the conservation constraint operation unit. It is configured to characterize the baseline drift of each magnetic measurement channel as a baseline state quantity, perform baseline state update using the conservation residual as observation information, construct an overdetermined observation equation set using the magnetic field response quantities of multiple operating points within a sliding time window, and perform overdetermined solution on the overdetermined observation equation set in conjunction with the geometric coupling coefficient to obtain the baseline state quantity of each magnetic measurement channel. The compensation injection unit is signal-connected to the magnetic sensing array and the baseline estimation unit, respectively, and is configured to subtract the corresponding baseline state quantity from each of the magnetic field response quantities to obtain the compensation output quantity.

[0008] By adopting the above technical solution, the conservation constraint operation unit utilizes the inherent linear conservation relationship between the measured currents, and uses the weighted summation of the magnetic field response quantities of the same group according to the combination weight coefficient determined by the geometric coupling coefficient to make the measured current components cancel each other out. Under the condition of not disconnecting the current-carrying loop and not adding magnetic circuit hardware, it obtains the observation quantity that is independent of the measured current. The conservation residual quantity only represents the linear combination of the baseline drift quantities of each magnetic measurement channel. The baseline estimation unit uses it as the observation information to drive the baseline state update, and constructs an overdetermined observation equation set with multiple operating points within the sliding time window. The linear combination is decomposed into the baseline state quantities of each channel, and the compensation injection unit deducts them channel by channel, so that baseline tracking and compensation are continuously performed during the period when the measured current exists.

[0009] Optionally, the baseline estimation unit includes a margin assessment subunit and a state switching subunit; the margin assessment subunit is configured to determine an identification margin based on the resolvability of each of the baseline state quantities according to the overdetermined observation equations; the state switching subunit is configured to: in response to the identification margin being higher than a preset margin threshold, place the baseline estimation unit in a tracking state, perform the overdetermined solution at the tracking convergence rate in the tracking state, and update the baseline temperature slope based on the correspondence between the solution results of each of the baseline state quantities and the temperature at which the magnetic sensing unit is located; in response to the identification margin being lower than the preset margin threshold, place the baseline estimation unit in a frozen state, stop the baseline state update in the frozen state, extrapolate each of the baseline state quantities according to the baseline temperature slope and the change in temperature at which the magnetic sensing unit is located, and continuously calculate and retain the conserved residual.

[0010] By adopting the above technical solution, the margin measurement is used to determine whether the excitation of the operating condition is sufficient to distinguish the baseline of each channel and convert it into a quantity that can be evaluated online. When the excitation is insufficient, it is converted into a frozen state and extrapolated by the baseline temperature slope, avoiding the estimation drifting along the null space of the equation system. The slope is taken from the solution of the tracking state and is updated online during operation.

[0011] Optionally, the baseline estimation unit further includes a backtracking correction subunit, which is configured to: switch the baseline estimation unit from the frozen state back to the tracking state in response to the identification margin recovering from below the preset margin threshold to above the preset margin threshold, and tune the tracking convergence rate according to the duration of the frozen state; use the baseline state quantities calculated in the tracking state before the start time and after the end time of the frozen state as the two boundary values, perform forward and reverse two-way smoothing on each of the conserved residual quantities remaining in the frozen state, and recalculate the baseline state quantities at each time during the frozen state; and generate correction quantities and attach correction identifiers to each of the compensation output quantities output during the frozen state based on the recalculated baseline state quantities.

[0012] By adopting the above technical solution, the conserved residuals retained during the frozen state participate in two-way smoothing together with the boundary values ​​at both ends after the tracking is restored, so that the baseline state quantities at each time of the frozen interval are recalculated, and the output compensation quantities are accompanied by correction quantities and correction labels. The deviations during the frozen period are traced and processed after the tracking is restored.

[0013] Optionally, the conservation constraint calculation unit is further configured to divide the plurality of magnetic measurement channels into a first conservation channel group and a second conservation channel group and perform the weighted summation respectively to obtain the corresponding conservation residuals; the combination weight coefficients of the first conservation channel group are configured in such a way that the environmental common-mode magnetic field components commonly experienced by each magnetic measurement channel cancel each other out in the weighted summation; the combination weight coefficients of the second conservation channel group are configured with unequal amplitudes so that the same-direction remanent magnetization components generated by the current impact of the same current-carrying conductor in each magnetic measurement channel retain non-zero projections in the weighted summation; the baseline estimation unit further includes a jump discrimination subunit, which is configured to: before and after the current impact occurs, pair the magnetic field response quantities of each magnetic measurement channel whose measured current is at the same operating point. The response quantity is composed of the channel-by-channel difference between the paired response quantities, forming a remanent magnetization increment vector. In response to the magnitude of the remanent magnetization increment vector exceeding a preset jump threshold, the remanent magnetization increment vector is determined to originate from the remanent magnetization jump of the magnetic sensing unit. The remanent magnetization direction vector corresponding to each current-carrying conductor is determined according to the geometric coupling coefficient of each magnetic measurement channel to each current-carrying conductor, resulting in a set of remanent magnetization direction vectors with the same number of roots as the current-carrying conductors. The remanent magnetization increment vector is projected onto each of the remanent magnetization direction vectors, and the current-carrying conductor corresponding to the remanent magnetization direction vector with the largest projection is determined as the current-carrying conductor experiencing the current surge. The baseline state quantities of the corresponding magnetic measurement channels are corrected using each component of the remanent magnetization increment vector, and the baseline state quantities are re-converged at a convergence rate greater than the tracking convergence rate.

[0014] By adopting the above technical solution, the unequal amplitude configuration of the second conserved channel group allows the remanent magnetization component to retain a non-zero projection in the weighted summation, thus bringing the remanent magnetization jump into the observable range; the difference between the paired response quantities at the same operating point before and after the impact distinguishes the remanent magnetization jump from the change of the measured current itself, and the projection of the remanent magnetization increment vector on a finite number of remanent magnetization direction vectors locates the current-carrying conductor that was impacted, and re-converges at an accelerated convergence rate.

[0015] Optionally, the conservation constraint calculation unit is further configured to divide the conservation channel groups according to at least two independent defined linear conservation relationships, and at least one magnetic measurement channel simultaneously belongs to two conservation channel groups and constitutes a common magnetic measurement channel; the baseline estimation unit further includes a premise verification subunit, which is configured to calculate the baseline state quantity of the common magnetic measurement channel using the conservation residual of each conservation channel group, and determine the difference between the calculation results as a conservation failure indication quantity; in response to the conservation failure indication quantity exceeding a preset failure threshold, it is determined that the defined linear conservation relationship has been violated, the baseline state update is stopped and a premise failure flag is output.

[0016] By adopting the above technical solution, two independent conservation channel groups each provide baseline state quantities for the same common magnetic measurement channel. The difference between the two constitutes a conservation failure indication quantity, which keeps the premise of determining the linear conservation relationship under monitoring. When the premise is violated, the update stops and an indicator is output to prevent erroneous residuals from continuing to drive the baseline state update.

[0017] Optionally, the conservation constraint calculation unit further includes a hierarchical identification subunit, which is configured to: select time periods within the sliding time window where the amplitude of the measured current change is higher than the upper amplitude threshold to form a coupling identification sample set; select time periods where the amplitude of the measured current change is lower than the lower amplitude threshold to form a baseline identification sample set, wherein the upper amplitude threshold is higher than the lower amplitude threshold; use the baseline state variables currently calculated from the baseline identification sample set as known quantities, and inversely solve the geometric coupling coefficient based on the correlation between the magnetic field response variables within the coupling identification sample set; then, using the inversely solved geometric coupling coefficient as known quantities, call the baseline estimation unit to recalculate the values ​​from the baseline identification sample set. The baseline state quantities are described, and the process is repeated alternately. The change in the geometric coupling coefficient between two adjacent rounds of the alternating execution is determined as the coupling convergence quantity, and the change in each of the baseline state quantities between two adjacent rounds of the alternating execution is determined as the baseline convergence quantity. In response to the simultaneous occurrence of the coupling convergence quantity and the baseline convergence quantity within their respective preset convergence thresholds, the alternating execution is terminated, and the combined weight coefficients are updated with the geometric coupling coefficients obtained from the inverse solution. In response to the occurrence of only one of the coupling convergence quantity and the baseline convergence quantity within the corresponding preset convergence threshold, it is determined that the geometric coupling coefficient and the baseline state quantity are indistinguishable, and the upper amplitude threshold is increased and the lower amplitude threshold is decreased before the alternating execution is restarted.

[0018] By adopting the above technical solution, the high and low amplitude of the measured current change constitute the coupling identification sample set and the baseline identification sample set, respectively. The two types of unknowns are solved alternately on mutually exclusive samples and are known quantities to each other. The geometric coupling coefficient is obtained by the inverse solution of the channel output. The two types of convergent quantities must fall into the threshold at the same time to terminate. If only one converges, it is determined to be indistinguishable and the threshold is expanded and recalculated to avoid the baseline being mistakenly absorbed into the coupling coefficient.

[0019] Optionally, the conservation constraint calculation unit may assign the same magnetic measurement channel to two or more conservation channel groups simultaneously, and each conservation channel group may output one conservation residual, and the conservation residuals of each channel may form a residual vector and be used as the observation information.

[0020] By adopting the above technical solution, the same magnetic measurement channel is constrained by multiple conserved channel groups, and the residual vector composed of multiple conserved residuals is updated as a whole, thereby increasing the number of available constraints.

[0021] Optionally, the baseline estimation unit divides the baseline state update into three response levels according to the time scale of the change of the baseline state quantity. When the change time scale is shorter than the sampling time scale of the magnetic field response quantity, the baseline state quantities are kept unchanged. When the change time scale is longer than the sampling time scale and changes continuously with time, the baseline state quantity is continuously updated according to the conserved residual quantity. When the baseline state quantity undergoes a step change, it re-converges at an accelerated convergence rate.

[0022] By adopting the above technical solution, the baseline state update is divided into categories according to the time scale of the change, with no action for fast changes, continuous tracking for slow changes, accelerated reconvergence for step changes, and slow drift tracking and step response handled separately.

[0023] Optionally, the baseline estimation unit allocates the forgetting window length of each magnetic measurement channel according to the resolvability of each magnetic measurement channel, with the magnetic measurement channel with higher resolvability corresponding to a shorter forgetting window length.

[0024] By adopting the above technical solution, the length of the forget window is allocated according to the resolvability of each magnetic measurement channel. The channel with higher resolvability takes a shorter window length, and the global window length does not need to be set according to the weakest channel.

[0025] Optionally, the identification margin is mapped to a confidence flag of the compensation output and output along with the compensation output.

[0026] By adopting the above technical solution, the confidence flag of the margin measurement mapping to the compensation output is identified and output together, and the downstream power accumulation and protection criteria can be used to distinguish the data of the freeze period and the tracking period.

[0027] Optionally, the compensation injection unit sets an upper limit on the change range of the baseline state quantity within a unit time, and injects according to the upper limit when the change range exceeds the upper limit.

[0028] By adopting the above technical solution, the change range of the baseline state quantity per unit time is constrained by an upper limit, and the estimated jump does not directly enter the compensation output quantity in the form of a step.

[0029] Secondly, this application provides an adaptive tracking compensation method for baseline drift of a magnetic sensor, which adopts the following technical solution: An adaptive tracking compensation method for baseline drift of a magnetic sensor includes the following steps, with S1 to S5 being repeated at each sampling time as the magnetic field response is acquired: S1. The magnetic field response is collected by multiple magnetic sensitive units respectively set at multiple current-carrying conductors, and each magnetic sensitive unit corresponds to a magnetic measurement channel; S2. Divide the magnetic measurement channels whose corresponding measured currents satisfy the determined linear conservation relationship into a conservation channel group, and determine the combination weight coefficient according to the geometric coupling coefficient of each magnetic measurement channel to each current-carrying conductor. The combination weight coefficient makes the measured current component in each magnetic field response cancel each other out in the weighted summation. S3. Perform the weighted summation on each of the magnetic field responses in the conserved channel group according to the combined weighting coefficients to obtain the conserved residual, which represents the linear combination of the baseline drift of each of the magnetic measurement channels; S4. The baseline drift of each magnetic measurement channel is characterized as a baseline state quantity. The baseline state is updated using the conserved residual as the observation information. An overdetermined observation equation set is constructed using the magnetic field response quantities of multiple operating points within the sliding time window. The overdetermined observation equation set is solved in conjunction with the geometric coupling coefficient to obtain the baseline state quantity of each magnetic measurement channel. S5. Subtract the corresponding baseline state quantity from each of the magnetic field response quantities to obtain the compensation output quantity.

[0030] By adopting the above technical solution, S1 to S5 are repeatedly executed at each sampling time as the magnetic field response is collected. The zero-input observation constructed by the conservation relationship and the overdetermined solution within the sliding time window are connected in the same cycle. The baseline state quantity is continuously tracked as the measured current continues to exist, and the compensation output quantity is deducted from the baseline state quantity at the current time channel by channel.

[0031] Optionally, S4 includes sub-steps S41-S43: S41. Determine the identification margin based on the resolvability of each baseline state quantity according to the overdetermined observation equation set; S42. In response to the identification margin being higher than a preset margin threshold, enter the tracking state, perform the overdetermined solution at the tracking convergence rate in the tracking state, and update the baseline temperature slope with the correspondence between the solution results of each baseline state quantity and the temperature of the magnetic sensing unit. S43. In response to the identification margin being lower than the preset margin threshold, enter a frozen state, stop the baseline state update in the frozen state, extrapolate each of the baseline state quantities according to the baseline temperature slope and the change in temperature of the magnetic sensing unit, and continuously calculate and retain the conserved residual quantity.

[0032] By adopting the above technical solution, S41 quantifies the distinguishability into a distinguishability margin, and S42 and S43 switch between the tracking state and the frozen state accordingly. When the excitation is insufficient, the baseline temperature slope is extrapolated and the conserved residual is continuously retained to preserve data for subsequent backtracking.

[0033] Optionally, step S4 further includes sub-steps S44-S46: S44. In response to the recognition margin recovering from below the preset margin threshold to above the preset margin threshold, switch from the frozen state back to the tracking state, and adjust the tracking convergence rate according to the duration of the frozen state; S45. Using the baseline state quantities calculated in the tracking state before the start time and after the end time of the frozen state as the two boundary values, perform forward and reverse two-way smoothing on each of the conserved residual quantities remaining in the frozen state, and recalculate the baseline state quantities at each time during the frozen state. S46. Based on the recalculated baseline state quantities, generate correction values ​​for each of the compensation output quantities that were output during the frozen state and attach correction identifiers.

[0034] By adopting the above technical solution, S44 to S46 use the tracking state solution results at both ends of the frozen interval as boundary values ​​to perform two-way smoothing on the remaining residuals, recalculate the baseline state quantities at each time of the frozen period and generate correction quantities, so that the data output during the frozen period can be retrospectively corrected.

[0035] Optionally, S4 further includes a sub-step S4a parallel to S41 to S46, wherein S4a includes sub-steps S4a1-S4a5: S4a1. Before and after the current surge of the measured current, the magnetic field response of each magnetic measurement channel at the same operating point of the measured current is taken to form a paired response quantity, and the channel-by-channel difference between the paired response quantities forms a residual magnetism increment vector. S4a2. In response to the magnitude of the remanent magnetization increment vector exceeding a preset jump threshold, the remanent magnetization increment vector is determined to be a remanent magnetization jump originating from the magnetic sensing unit; S4a3. Determine the remanent magnetization direction vector corresponding to each current-carrying conductor according to the geometric coupling coefficient of each magnetic measurement channel to each current-carrying conductor, and obtain a set of remanent magnetization direction vectors with the same number as the number of current-carrying conductors; S4a4. Project the residual magnetization increment vector onto each of the residual magnetization direction vectors respectively, and determine the current-carrying conductor corresponding to the residual magnetization direction vector with the largest projection as the current-carrying conductor that will experience the current surge; S4a5. The baseline state quantities of the corresponding magnetic measurement channels are corrected by each component of the residual magnetism increment vector, and the baseline state quantities are reconverged at a convergence rate greater than the tracking convergence rate.

[0036] By adopting the above technical solution, S4a1 to S4a5 form a residual magnetization increment vector with the pairing difference at the same operating point and determine the residual magnetization jump. The current-carrying conductor that is impacted is projected and located on a finite number of residual magnetization direction vectors. Then, the baseline state quantity of the corresponding channel is corrected and the re-convergence is accelerated.

[0037] Optionally, in S2, the plurality of magnetic measurement channels are divided into a first conserved channel group and a second conserved channel group, wherein the combination weight coefficients of the first conserved channel group are configured with equal amplitude, and the combination weight coefficients of the second conserved channel group are configured with unequal amplitude.

[0038] By adopting the above technical solution, the method also uses two sets of weighting coefficients to share the cancellation of the measured current component and the retention of the same-direction remanent magnetization component, providing an observable path for remanent magnetization identification.

[0039] Optionally, in S3, the rate of change of the measured current between adjacent time periods is used as the gate, and the conserved residual obtained in the time period where the rate of change is higher than the preset gate threshold does not participate in the baseline state update.

[0040] By adopting the above technical solution, the conserved residual obtained during periods of high rate of change of the measured current will not be included in the update, and the sampling time misalignment between each magnetic measurement channel will not be mistakenly recorded as the baseline.

[0041] In summary, this application includes at least one of the following beneficial technical effects: 1. Construct a combination weighting coefficient based on the established linear conservation relationship between the measured currents, so that the measured current components cancel each other out in the weighted summation to obtain the conservation residual. Then, the baseline state quantities of each magnetic measurement channel are decomposed by the overdetermined solution of multiple operating points within the sliding time window and deducted channel by channel. The entire process does not disconnect the current-carrying loop and does not add magnetic circuit hardware.

[0042] 2. Use the identification margin to determine online whether the excitation of the operating condition is sufficient to distinguish the baseline of each channel. If the excitation is insufficient, freeze the update and extrapolate the baseline temperature slope based on the online update. After the tracking is restored, use the boundary values ​​at both ends to perform double-pass smoothing on the remaining residual and backtrack to correct the compensation output during the freeze period.

[0043] 3. By grouping the remanent magnetization components into groups with unequal amplitudes, the remanent magnetization jump is distinguished and the impact conductor is located by matching the difference between the paired response quantities at the same working point before and after the impact with the projection of a finite number of remanent magnetization direction vectors. In addition, the conservation premise is cross-verified by two sets of independent conservation relationships, and the geometric coupling coefficient and baseline state quantity are identified alternately by mutually exclusive samples. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of the composition of an adaptive tracking compensation system for magnetic sensor baseline drift in one embodiment of this application.

[0045] Figure 2 This is a schematic diagram of the spatial arrangement of the current-carrying conductor and the magnetic sensitive unit, and the division of the conserved channel group in one embodiment of this application.

[0046] Figure 3 This is a schematic diagram of the overall process of an adaptive tracking compensation method for magnetic sensor baseline drift in one embodiment of this application.

[0047] Figure 4 As shown in one embodiment of this application Figure 3 A detailed flowchart of step S4.

[0048] Figure 5 As shown in one embodiment of this application Figure 3 A detailed flowchart of the parallel sub-step S4a of step S4. Detailed Implementation

[0049] The present application will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the application and are not intended to limit the scope of the application.

[0050] This application discloses an adaptive tracking compensation system for baseline drift of a magnetic sensor. Utilizing the inherent linear conservation relationship between the measured currents, this application constructs a set of combined weighting coefficients that cancel out the components of the measured current. Without disconnecting the current-carrying loop, it extracts the conservation residual, independent of the measured current, and uses it as the online observation of the baseline drift. First, the magnetic sensor array collects magnetic field responses at multiple current-carrying conductors. Then, the conservation constraint calculation unit performs a weighted summation according to the combined weighting coefficients to obtain the conservation residual. Next, the baseline estimation unit uses the conservation residual as observation information to update the baseline state and combines it with overdetermined solutions at multiple operating points within a sliding time window to obtain the baseline state of each magnetic measurement channel. Finally, the compensation injection unit subtracts the baseline state from each channel. Thus, it can continuously track the baseline drift during operation when the measured current is continuously present and output the compensated output after baseline subtraction.

[0051] The following provides a detailed description of the terminology, overall system composition, magnetic sensing array, conservation constraint operation unit, baseline estimation unit, compensation injection unit, and corresponding method steps. First, several technical terms involved in the embodiments of this application will be explained.

[0052] A magnetic measurement channel is an independent measurement consisting of a magnetic sensing unit and its subsequent signal conditioning, sampling, and digitization path. The magnetic sensing unit is placed near a current-carrying conductor to sense the magnetic flux density at that location and convert it into an electrical signal. One magnetic sensing unit corresponds to one magnetic measurement channel, and multiple magnetic measurement channels can operate in parallel without multiplexing each other.

[0053] The magnetic field response is the raw measured quantity output by the magnetic measurement channel. Its dimension can be magnetic induction intensity or the current indication value converted according to a preset scale. The magnetic field response consists of two superimposed parts: one part changes with the measured current, and the other part is independent of the measured current.

[0054] The measured current component is the portion of the magnetic field response that originates from the measured current in the current-carrying conductor. The magnetic field at the location of the magnetic measurement channel is contributed by all the current-carrying conductors in its vicinity. Therefore, the measured current component is the result of the superposition of the measured currents of each current-carrying conductor according to their respective contribution ratios, rather than being determined solely by the nearest current-carrying conductor.

[0055] Baseline drift is the portion of the magnetic field response that is independent of the measured current, originating from the zero point of the magnetic sensing element. When the current in each current-carrying conductor is zero, the output of the magnetic measurement channel is not zero; its residual value is the baseline drift at that time. The baseline drift moves slowly with changes in packaging stress, magnetic domain relaxation, remanence of the core magnetization, and temperature, and may step shift after a large current surge.

[0056] A defined linear conservation relationship is a linear constraint that holds true among the measured currents of several current-carrying conductors, with known coefficients and remains constant during operation. For example, in a three-phase power supply system without a neutral wire, the sum of the three-phase currents is always zero; the positive and negative currents of the DC bus are equal in value and opposite in direction; and the constraint relationship between the currents flowing into and out of each branch at circuit nodes all fall under the category of defined linear conservation relationships.

[0057] A conservation channel group is a set of magnetic measurement channels whose respective measured currents satisfy the same defined linear conservation relationship. The members of a conservation channel group are determined by the electrical connection relationship between the current-carrying conductors, and are independent of the physical arrangement order of the magnetic sensing units. This division refers to determining the members of each group separately; groups can share common magnetic measurement channels, and are not limited by mutual exclusion; the same magnetic measurement channel can belong to two conservation channel groups simultaneously.

[0058] The geometric coupling coefficient is the response ratio of a magnetic sensing channel to a current-carrying conductor, that is, the magnetic field response generated by a unit current in the current-carrying conductor at the magnetic sensing channel. The geometric coupling coefficient is determined by the relative position between the magnetic sensing unit and the current-carrying conductor, the cross-sectional shape of the conductor, and the magnetic core structure. It remains stable in the short term after installation; however, during long-term operation, it may slowly deviate from the calibrated value due to conductor thermal displacement and changes in the permeability of the magnetic core with temperature. If necessary, it should be re-determined by online inverse kinematics.

[0059] The combined weighting coefficients are a set of weighting coefficients determined by the geometric coupling coefficients, used to perform a weighted summation of the magnetic field responses within the conserved channel group. The values ​​of the combined weighting coefficients ensure that the measured current components in each magnetic field response cancel each other out during the weighted summation. Therefore, the determination of the combined weighting coefficients is based on the combined result of the geometric coupling coefficients and the determination of the linear conservation relationship.

[0060] Weighted summation is an operation process in which each magnetic field response quantity in the conserved channel group is multiplied by the combined weight coefficients and then summed. The addends are all taken from the magnetic field response quantities in the same conserved channel group.

[0061] The conserved residual is the result of a weighted summation. Under the condition that the combined weighting coefficients match the actual geometric coupling coefficients, the measured current component cancels out in the weighted summation. The conserved residual no longer carries the measured current information, but instead represents a linear combination of the baseline drift of each magnetic measurement channel within the conserved channel group. When the geometric coupling coefficients deviate from their calibration values ​​due to installation deviations or long-term operation, the cancellation is incomplete, leaving a component in the conserved residual proportional to the measured current amplitude. In this case, the geometric coupling coefficients are re-determined by the online inverse kinematics of the hierarchical identification subunit, and the combined weighting coefficients are updated accordingly.

[0062] The baseline state quantity is the state representation of the baseline drift of each magnetic measurement channel in the estimator, and it is the quantity to be estimated. Each magnetic measurement channel corresponds to a baseline state quantity, and its value is updated with changes in running time and temperature.

[0063] The observation information is the input that drives the baseline state update. In the embodiments of this application, the observation information is taken from the conserved residual, and not from any external reference current or independent current measurement.

[0064] Baseline state update is the process of recursively correcting each baseline state variable based on observed information at a preset convergence rate. Baseline state update is performed once at each sampling time, and together with the batch computation within the sliding time window, it maintains the baseline state variables.

[0065] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and are not used to describe a specific order or sequence; "comprising," "having," and any variations thereof are intended to cover non-exclusive inclusion.

[0066] The overall composition of this system is described below.

[0067] Reference Figure 1 The adaptive tracking compensation system for magnetic sensor baseline drift in this application includes a magnetic sensor array, a conservation constraint operation unit, a baseline estimation unit, and a compensation injection unit. The magnetic sensor array is responsible for acquiring the magnetic field quantity, the conservation constraint operation unit is responsible for the cancellation of the measured current component and the generation of the conservation residual quantity, the baseline estimation unit is responsible for the calculation and maintenance of the baseline state quantity, and the compensation injection unit is responsible for the deduction of the baseline state quantity and the output of the compensation output quantity.

[0068] The signal connections between the four components are as follows: The conservation constraint operation unit is signal-connected to the magnetic sensor array and receives the magnetic field response output from each magnetic measurement channel; the baseline estimation unit is signal-connected to the conservation constraint operation unit and receives the conservation residual; the compensation injection unit is signal-connected to both the magnetic sensor array and the baseline estimation unit, receiving the original magnetic field response on one side and the baseline state quantity from each magnetic measurement channel on the other; there is also a reverse connection between the baseline estimation unit and the conservation constraint operation unit, used to send the baseline state quantity back to the conservation constraint operation unit.

[0069] During system operation, the data flow is a single forward push with a feedback branch. The magnetic field responses output by each channel of the magnetic sensor array simultaneously enter the conservation constraint calculation unit and the compensation injection unit. The conservation residual output by the conservation constraint calculation unit enters the baseline estimation unit, and the baseline state quantities output by the baseline estimation unit enter the compensation injection unit. The compensation injection unit performs subtraction channel by channel and outputs the compensated output quantity. When the geometric coupling coefficient needs to be solved online, the feedback branch provides the baseline state quantities as known quantities to the conservation constraint calculation unit.

[0070] The above four operations are performed once in a loop at each sampling time. The compensation output can be directly output as a measurement result to downstream processes such as power accumulation, state of charge estimation, or protection criteria, or it can be cached locally and sent up in batches periodically.

[0071] This system can be applied to situations requiring long-term continuous measurement and where scheduling downtime is inconvenient, such as power metering devices, power gateway metering terminals, current detection in photovoltaic and energy storage converters, and current sampling in power battery packs. Taking a three-phase low-voltage busbar without a neutral line as an example, the three phase conductors are arranged side by side, with a magnetic sensing unit installed at each phase conductor. An additional magnetic sensing unit is installed on the other side of one of the phase conductors, forming four magnetic measurement channels. The sum of the three-phase currents is always zero, thus establishing a definite linear conservation relationship. The positive and negative conductors of the DC bus have equal and opposite currents, also establishing a definite linear conservation relationship. Under this relationship, the distribution ratio of the measured current is always a constant value, and there is no difference in the distribution ratio between different operating points. Therefore, it must be used in conjunction with other definite linear conservation relationships in the system, rather than as the sole conservation relationship of the system.

[0072] The magnetic sensing array is described below.

[0073] Reference Figure 2 The magnetic sensing array includes multiple magnetic sensing units respectively disposed at multiple current-carrying conductors. Each magnetic sensing unit constitutes a magnetic measurement channel and outputs a magnetic field response. The positions of the magnetic sensing units correspond one-to-one or many-to-one with each current-carrying conductor. They are installed in fixed relative positions on the same cross-section and are not moved after installation.

[0074] In some embodiments, the magnetic sensing unit employs a Hall effect sensor. The magnetic field generated by the measured current around the conductor passes through a Hall plate. Under the influence of a bias current, the Hall plate outputs a Hall voltage proportional to the magnetic induction intensity in the transverse direction. This voltage is amplified and converted from analog to digital to become the magnetic field response. In other embodiments, the magnetic sensing unit employs a magnetoresistive sensor, whose resistance changes with the applied magnetic field. The four magnetoresistive arms are connected in a bridge circuit to output a differential voltage, which, after passing through the same conditioning link, becomes the magnetic field response. Both types of magnetic sensing units maintain a linear correspondence between the output and the measured current, and are also implementations of the magnetic sensing unit of this application.

[0075] In a magnetic sensing array, all magnetic measurement channels share the same sampling clock. Each channel triggers sampling and synchronously completes analog-to-digital conversion at the same time. The purpose of sharing the clock is to ensure that the magnetic field responses participating in the same weighted summation correspond to the same physical moment. When the measured current changes rapidly, if there is a misalignment in the sampling time between channels, the measured current reflected by each channel will not be the same instantaneous value. The residual value after cancellation will be mixed with the baseline drift and enter the conserved residual value. Synchronous sampling eliminates this residual value before the weighted summation.

[0076] In a magnetic sensing array, the number of magnetic measurement channels is greater than the number of current-carrying conductors. For example... Figure 2As shown, the additional magnetic measurement channel can be formed by two magnetic sensing units arranged on different sides of the same current-carrying conductor, or by magnetic sensing units arranged between two adjacent current-carrying conductors. Such channels have non-zero geometric coupling coefficients for multiple current-carrying conductors, and their magnetic field response carries the measured current information of multiple conductors, providing additional independent equations when constructing the overdetermined observation equation set.

[0077] The following is an explanation of the conservation constraint operation unit.

[0078] The conservation constraint operation unit is connected to the magnetic sensing array signal. It divides several magnetic measurement channels whose corresponding measured currents satisfy a defined linear conservation relationship into conservation channel groups. Combination weighting coefficients are determined according to the geometric coupling coefficients of each magnetic measurement channel to each current-carrying conductor. Then, a weighted summation is performed on each magnetic field response quantity within the conservation channel group according to the combination weighting coefficients to obtain the conservation residual. The combination weighting coefficients cause the measured current components in each magnetic field response quantity to cancel each other out during the weighted summation. The conservation residual represents the linear combination of the baseline drift quantities of each magnetic measurement channel.

[0079] The division of the conserved channel group is based on the electrical connection relationship between current-carrying conductors. Figure 2 On a three-phase low-voltage busbar without a neutral wire, the sum of the three-phase currents is always zero. The three magnetic measurement channels corresponding to the three-phase conductors, along with any additional magnetic measurement channels located between or on the other side, form a conserved channel group. On a DC bus, the positive and negative currents are equal in value and opposite in direction. The two magnetic measurement channels corresponding to the positive and negative conductors form a conserved channel group. At a distribution node, the sum of the currents flowing into the node from each branch equals the sum of the currents flowing out of the node. The magnetic measurement channels corresponding to the conductors of each branch form a conserved channel group. The coefficients for these three types of relationships are all given by the circuit topology and remain constant during operation.

[0080] In some embodiments, two independent deterministic linear conservation relationships can coexist within the same system. Taking a three-phase four-wire power distribution system as an example, the three phase conductors and the neutral conductor satisfy the node constraint that the inflow current and outflow current are equal at the incoming node, constituting the first deterministic linear conservation relationship; the three phase conductors themselves satisfy the condition that the sum of the phase currents is zero on the branch section excluding the neutral conductor, constituting the second deterministic linear conservation relationship. The coefficient vectors of the two relationships are linearly independent and therefore mutually independent, allowing for the separate division of conservation channel groups. Furthermore, the magnetic measurement channels at the phase conductors simultaneously belong to both conservation channel groups, forming a common magnetic measurement channel. This arrangement provides the operational conditions for the implementation of the premise verification subunit and multiple sets of constraints.

[0081] The combined weighting coefficients are determined jointly by the geometric coupling coefficients and the deterministic linear conservation relation. Let the first... The magnetic field response of each magnetic measurement channel is It is determined by the measured current component and the baseline drift of the channel. The components of the measured current are summed, and the measured current components are the measured currents of each current-carrying conductor according to the geometric coupling coefficient of that channel to each current-carrying conductor. The weighted summation value, where This represents the serial number of the current-carrying conductor. Combined weighting coefficient. The value of must be such that the geometric coupling coefficients of each channel in the conserved channel group to the same current-carrying conductor are determined according to . The weighted sum is proportional to the coefficient of the current-carrying conductor in determining the linear conservation relationship. When all current-carrying conductors satisfy this condition, the measured current components cancel each other out in the weighted summation.

[0082] In the specific implementation, the conservation constraint operation unit calculates the conservation channel group according to the following formula. Conservation residuals : in, For the conservation channel group The conserved residual, For the first The combined weighting coefficients of each magnetic measurement channel For the first The magnetic field response of each magnetic measurement channel For the first The baseline drift of each magnetic measurement channel. The first equal sign in the above equation is the definition of weighted summation, and the second equal sign is the result after the measured current components cancel each other out. After cancellation, the conserved residual no longer changes with the measured current; its value is formed only by the linear combination of the baseline drift of each channel in the conserved channel group according to the same set of combination weight coefficients.

[0083] Taking the three magnetic measurement channels in a three-phase conserved channel group facing the three phase conductors as an example, the geometric coupling coefficients of each channel to its facing conductor are equal, and the geometric coupling coefficients to the other two conductors are also equal. In this case, taking the same value for the combination weighting coefficients of the three channels ensures that the sum of the three-phase currents is always zero. The total result of these three terms in the weighted summation is a fixed multiple of the sum of the three baseline drifts. This result remains unchanged regardless of the load level of the three-phase currents.

[0084] The conserved residual only yields a linear combination of baseline drift values, not the individual baseline drift values ​​of each channel. Taking a three-phase conserved channel group as an example, the baseline drift values ​​of each magnetic measurement channel within the group appear in the same equation, and the solution to this equation is not unique. The calculation of individual baseline state values ​​is completed by the baseline estimation unit through overdetermined solutions at multiple operating points within a sliding time window.

[0085] In some embodiments, the conservation constraint operation unit divides several magnetic measurement channels into a first conservation channel group and a second conservation channel group and performs weighted summation on each group to obtain corresponding conservation residuals. The combination weight coefficients of the first conservation channel group are configured such that the environmental common-mode magnetic field components commonly experienced by each magnetic measurement channel cancel each other out in the weighted summation; the combination weight coefficients of the second conservation channel group are configured with unequal amplitudes so that the remanent magnetization components generated by the current impact on the same current-carrying conductor in each magnetic measurement channel retain non-zero projections in the weighted summation. The remanent magnetization components here refer to the components whose remanent magnetization increments left by the current impact on each magnetic measurement channel are distributed according to the geometric coupling coefficient of each channel to the current-carrying conductor, and whose components have the same sign but different values; they are different from the environmental common-mode magnetic field components of equal amplitude and same sign in each magnetic measurement channel, and do not cancel each other out because the sum of the weight coefficients is zero. The division of labor between the two sets of weight coefficients is that the residuals of the first conserved channel group are not affected by external stray magnetic fields and are suitable for baseline observations under normal conditions; the residuals of the second conserved channel group retain the projection of the same-direction remanent magnetic component, so that the remanent magnetic transition can enter the correction path of the baseline estimation unit.

[0086] It should be noted that canceling the measured current component and canceling the ambient common-mode magnetic field component are two different constraints on the combined weighting coefficients. The former requires that the weighted geometric coupling coefficient be proportional to the conservation coefficient, while the latter requires that the sum of all weighting coefficients be zero. When the number of magnetic measurement channels equals the number of current-carrying conductors, the two constraints are generally compatible only when all weighting coefficients are zero. When the number of magnetic measurement channels exceeds the number of current-carrying conductors, the combined weighting coefficients have extra degrees of freedom after satisfying the first constraint, which can be used to add the second constraint, causing both to cancel each other out. In this case, the weighting coefficients are no longer equal in amplitude. The above configuration of the first conservation channel group is based on the premise that the number of magnetic measurement channels exceeds the number of current-carrying conductors. Whether the combined weighting coefficients of the first conservation channel group are equal in amplitude depends on whether it simultaneously undertakes the function of canceling the ambient common-mode magnetic field component: when only canceling the measured current component, an equal amplitude configuration can be taken; when simultaneously canceling the ambient common-mode magnetic field component, the configuration is based on the above additional conditions and is no longer equal in amplitude. These are two forms of weighting coefficient values ​​for the same conservation channel group in different embodiments.

[0087] In some embodiments, when the combined weight coefficients of the second conserved channel group are configured with unequal amplitude, the weight coefficients maintain a fixed proportional relationship and are written once during the installation and calibration phase and do not change during operation. Therefore, the difference between the two sets of residuals only comes from the measured environment and device state, and not from the drift of the weight coefficients themselves.

[0088] In some embodiments, the conservation constraint operation unit uses the rate of change of the measured current between adjacent time periods as a gate. Conservation residuals obtained during time periods with a rate of change exceeding a preset gate threshold are not included in the baseline state update. When the measured current changes rapidly, even if all channels are sampled synchronously, the bandwidth difference in the conditioning link may still introduce a small phase difference between channels. This phase difference manifests as an additional residual after weighted summation. Gating excludes residuals from baseline state updates during such periods, and automatically resumes adoption after the rate of change decreases.

[0089] In some embodiments, the conservation constraint operation unit further includes a hierarchical identification subunit for online inverse solving of the geometric coupling coefficients without relying on external calibration. The geometric coupling coefficients and baseline state variables are indistinguishable on the same batch of samples because both are additive or proportional unknowns in the magnetic field response; solving both simultaneously on a batch of samples results in multiple combinations producing the same output. The hierarchical identification subunit utilizes the property that the identifiable conditions of the two types of unknowns are mutually exclusive in time, selecting samples according to the mutually exclusive conditions and solving them alternately.

[0090] In its implementation, the hierarchical identification sub-unit selects time periods within the sliding time window where the amplitude of the measured current change is higher than the upper amplitude threshold to form the coupling identification sample set, and selects time periods where the amplitude of the measured current change is lower than the lower amplitude threshold to form the baseline identification sample set. The upper amplitude threshold is higher than the lower amplitude threshold. During periods with large amplitude changes in the measured current, the change in the magnetic field response is mainly caused by the change in the measured current, and the baseline drift is approximately constant during this period, making it suitable for inversely solving the geometric coupling coefficient. During periods with small amplitude changes in the measured current, the change in the magnetic field response is mainly caused by the movement of the baseline drift, making it suitable for solving the baseline state variables. The interval between the two thresholds is the transition zone, and samples falling into the transition zone do not enter either sample set.

[0091] The hierarchical identification subunit first uses the baseline state variables obtained from the current solution of the baseline identification sample set as known quantities, and then uses the correlation between the magnetic field response quantities in the coupling identification sample set to solve the geometric coupling coefficient. Next, using the geometric coupling coefficient obtained from the inverse solution as known quantities, it calls the baseline estimation unit to recalculate the baseline state variables from the baseline identification sample set, and so on, alternating between the two. In each round of alternation, the current solution of one type of unknown quantity provides known conditions for solving the other type of unknown quantity, and the accuracy of both types of unknown quantities gradually increases with each round.

[0092] The termination criterion for alternating execution is derived from the convergence of the two types of unknowns. The hierarchical identification subunit determines the change in the geometric coupling coefficient between two adjacent rounds of alternating execution as the coupling convergence value, and the change in each baseline state quantity between two adjacent rounds of alternating execution as the baseline convergence value. When both the coupling convergence value and the baseline convergence value fall within their respective preset convergence thresholds, alternating execution terminates and the combined weight coefficients are updated with the geometric coupling coefficients obtained from the inverse solution. When only one of the coupling convergence value and the baseline convergence value falls within its corresponding preset convergence threshold, it is determined that the geometric coupling coefficient and the baseline state quantity are indistinguishable. The upper amplitude threshold is increased and the lower amplitude threshold is decreased before resuming alternating execution. If only one converges while the other continues to change, it indicates that the value of one type of unknown is absorbed by the other type of unknown in this iteration. In this case, the upper amplitude threshold is increased and the lower amplitude threshold is decreased, making the operating characteristics of the two sample sets purer and the identifiable conditions of the two types of unknowns on the sample more thoroughly separated.

[0093] The inverse solution of the geometric coupling coefficients only requires relative proportions and does not require known values ​​of the measured current. The cancellation condition of the combined weighting coefficients remains unchanged after the geometric coupling coefficient matrix is ​​right-multiplied by an invertible scaling factor; the comparison of the projection of the direction vectors used to locate the impact conductor is performed after the direction vectors are normalized, and therefore also does not depend on the overall scale of the geometric coupling coefficient matrix. Therefore, the hierarchical identification of sub-units does not require external current transformers or any current measurements as a scale reference.

[0094] In some embodiments, the hierarchical identification subunit sets an upper limit on the number of rounds of alternating execution. If the upper limit is reached but the two types of convergence coefficients have not simultaneously fallen below the threshold, the geometric coupling coefficients obtained from the previous simultaneous convergence are retained and a check mark is output. Simultaneously, each of the two sample sets sets a lower limit on the sample size. If the sample size of either sample set falls below its lower limit after threshold expansion, the current round of expansion stops, and execution resumes only after sufficient samples have accumulated within the sliding time window, to prevent insufficient samples from causing the equation system to lose its overdeterminacy.

[0095] The baseline estimation unit is described below.

[0096] Combination Figure 1 The baseline estimation unit is connected to the conservation constraint operation unit. The baseline drift of each magnetic measurement channel is characterized as the baseline state quantity. The baseline state is updated using the conservation residual as the observation information. The overdetermined observation equation set is constructed using the magnetic field response quantities of multiple operating points within the sliding time window. The overdetermined observation equation set is solved in conjunction with the geometric coupling coefficient to obtain the baseline state quantity of each magnetic measurement channel.

[0097] The baseline estimation unit is solved in parallel along two tracks. One is the batch track, which constructs an overdetermined set of observation equations using multiple operating point samples accumulated within a sliding time window and solves for the baseline state variables of each channel in one go. The other is the recursive track, which incrementally corrects each baseline state variable at each sampling time using the current conserved residual as the observation information. The division of labor and fusion rules between the two tracks are explained later in this section.

[0098] First, the construction and solvability conditions of the overdetermined observation equations are explained. The conserved residuals at a single sampling moment provide only a linear combination of the baseline drifts within the conserved channel group. The number of unknowns equals the number of magnetic measurement channels within that group, and the number of equations is one; the solution at that moment is not unique. The operating point represents the state of operation where the measured current is at a certain level. The distribution ratio of the measured current differs at different operating points, resulting in different magnetic field responses for each magnetic measurement channel. Meanwhile, the baseline state quantities remain approximately constant within the window length.

[0099] It should be noted that the purpose of multiple operating points is not to add equations to the conservation residuals themselves. Once the combined weighting coefficients are determined, the result of the weighted summation is determined only by the baseline drift of each line and is independent of the measured current. Therefore, the conservation residual equations given by each operating point coincide with each other. The purpose of multiple operating points is to incorporate the original magnetic field response of each magnetic measurement channel into the solution, making the baseline state quantities of each channel and the measured current of each operating point unknowns in the same set of equations, and then providing constraints through the conservation relationships of each operating point.

[0100] In its implementation, the baseline estimation unit substitutes the magnetic field response and geometric coupling coefficient of each magnetic measurement channel at each operating point within the sliding time window to obtain a set of observation equations for that operating point. Simultaneously, it writes the constraint equations satisfied by the measured current at that operating point according to the defined linear conservation relationships. Accordingly, the unknowns are divided into two categories: one is the baseline state quantities of each magnetic measurement channel, the number of which equals the number of magnetic measurement channels; the other is the measured current of each current-carrying conductor at each operating point, the number of which equals the product of the number of operating points and the number of current-carrying conductors. The equations are also divided into two categories: one is the observation equations for each operating point, the number of which equals the product of the number of operating points and the number of magnetic measurement channels; the other is the conservation constraint equations for each operating point, the number of which equals the product of the number of operating points and the number of independent conservation relationships.

[0101] This yields three quantitative relationships.

[0102] First, the number of magnetic measurement channels must be greater than the number of current-carrying conductors. For each additional operating point, the increase in the number of equations equals the sum of the number of magnetic measurement channels and the number of independent conservation relationships, while the increase in the number of unknowns equals the number of current-carrying conductors. If the former is not greater than the latter, adding operating points does not change the incomplete state of the equation system. The number of independent conservation relationships is always less than the number of current-carrying conductors; therefore, configuring the system so that the number of magnetic measurement channels is greater than the number of current-carrying conductors ensures that the former is greater than the latter.

[0103] Secondly, the number of operating points within the sliding time window must ensure that the total number of equations exceeds the total number of unknowns. When the total number of equations exceeds the total number of unknowns, the system of equations is overdetermined, and overdetermined solutions in the least squares sense can be used to suppress sampling noise.

[0104] Third, the distribution ratio of the measured current at each operating point must be different. If the two sets of equations given for two operating points with the same ratio are linearly related, only one set of equations will be counted.

[0105] The above equations also contain an unidentifiable component independent of the number of operating points. Suppose a set of time-invariant deviations simultaneously satisfy all defined linear conservation relationships. If the measured current of each current-carrying conductor at each operating point is shifted according to this set of deviations, and the baseline state quantities of each magnetic measurement channel are simultaneously shifted in the opposite direction according to the result of mapping this set of deviations using geometric coupling coefficients, then all magnetic field responses and all conservation constraint equations remain unchanged. The equations cannot distinguish between the two sets of values ​​before and after the shift. The subspace spanned by this unidentifiable component has a dimension equal to the difference between the number of current-carrying conductors and the number of independent conservation relationships, independent of the number of operating points within the sliding time window.

[0106] The unidentifiable component is constant and does not accumulate over time. Its performance in the compensation output is a constant current zero-point bias, rather than a baseline drift that increases over time. Therefore, it does not affect the tracking of baseline drift over time in this application.

[0107] The baseline estimation unit is not updated in the unidentifiable subspace, and the components of each baseline state variable in this subspace retain their prior values. The identifiable components are updated window by window by the overdetermined solution. The prior value is taken from a single acquisition at the zero-current moment during the process of the measured current rising from zero to the rated current after the system is powered on. At this time, the current in each current-carrying conductor is zero, and the magnetic field response of each magnetic measurement channel is equal to its respective baseline drift. In the case where there is no zero-current window, the prior value is taken from the factory calibration record. Thus, the baseline state variable of each magnetic measurement channel is uniquely determined under the given prior value.

[0108] In some embodiments, when selecting operating points, the baseline estimation unit removes duplicates from candidate samples according to the distribution ratio of the measured current. Only one sample with a ratio difference lower than a preset discrimination is retained to enter the overdetermined observation equation set, so as to avoid the nominal number of operating points meeting the standard while the actual number of independent equations is insufficient.

[0109] Next, we will explain the division of labor and fusion rules between the batch track and the recursive track. The output of the batch track is the absolute value of the baseline state variables of each channel in the identifiable subspace. The cost is that a sufficient number of operating points with different distribution ratios need to be accumulated within the window, so the update time is sparse and lags behind the current time. The output of the recursive track is the incremental correction of each baseline state variable. It can be executed at every sampling time, but the conserved residual only constrains a linear combination of the baseline state variables at a single time. Relying solely on the recursive track for a long time will cause the baseline state variables to drift together along the null space of the equation system, and the absolute value will lose its reference.

[0110] The two orbitals work in a coordinated manner, with anchoring and tracking functions. Each time the batch orbital completes an overdetermined solution, it overwrites all baseline state variables in the identifiable subspace with the solution result, marking this moment as an anchor point. Between two anchor points, the recursive orbital uses the current conserved residual as the observation information to incrementally correct each baseline state variable according to the tracking convergence rate, ensuring that the baseline state variables are continuously tracked as temperature and time change between anchor points. The writing rule is batch priority; that is, at the anchor point, the batch solution result is used as the basis, overwriting the previously obtained recursive values. After the anchor point, the recursive orbital continues to correct using this result as a new starting point.

[0111] In the specific implementation, the overwriting of anchor points does not employ instantaneous replacement. Instead, it transitions linearly from the recursive result to the batch result over several sampling periods, and then resumes recursive tracking after the transition is complete. The difference between the batch result and the recursive result reflects the accumulated deviation of the recursive trajectory within the previous anchor point interval. Distributing this deviation over several sampling periods ensures that the compensation output does not exhibit abrupt changes. In some embodiments, the number of transition periods is on the order of one-tenth of the sliding time window length.

[0112] In some embodiments, the conservation constraint operation unit assigns the same magnetic measurement channel to two or more conservation channel groups simultaneously, and each conservation channel group outputs a conservation residual; the baseline estimation unit uses the residual vector formed by the conservation residuals as the overall observation information. Multiple sets of constraints act simultaneously on the same batch of baseline state variables, increasing the number of equations at a single moment from one to the number of conservation channel groups, thereby enhancing the constraint capability of the recursive trajectory between anchor points.

[0113] The margin assessment subunit and the state switching subunit are described below. In some embodiments, the baseline estimation unit further includes a margin assessment subunit and a state switching subunit, which are responsible for quantifying the distinguishability and switching the operating state, respectively.

[0114] The margin assessment subunit determines the identification margin based on the resolvability of each baseline state variable in the overdetermined observation equation set. The resolvability depends on the degree of difference in the distribution ratio of the measured current at each operating point within the sliding time window. When the load is constant over a long period, the distribution ratio of each operating point within the window tends to be consistent, and the equations tend to be linearly correlated. Although the equation set is nominally still overdetermined, the actual resolvable degrees of freedom are lower than the number of unknowns. If the solution continues to be calculated, it will take arbitrary values ​​along the null space of the equation set. The identification margin is a quantification of this situation; its value increases with the increase of the difference in the distribution ratio of the operating points within the window and decreases with the decrease of the difference. In specific implementations, the identification margin can be taken from the minimum singular value of the coefficient matrix of the overdetermined observation equation set, the reciprocal of the condition number of the coefficient matrix, or the statistical measure of the dispersion of the distribution ratio of the measured current at each operating point within the window. All three vary monotonically with the difference in the distribution ratio and are also implementation methods of the identification margin.

[0115] The state switching subunit determines the operating state of the baseline estimation unit based on the comparison between the identified margin metric and the preset margin threshold. If the identified margin metric is higher than the preset margin threshold, the baseline estimation unit is placed in the tracking state. In the tracking state, overdetermined solutions are performed at the tracking convergence rate, and the baseline temperature slope is updated based on the correspondence between the solution results of each baseline state quantity and the temperature of the magnetic sensing unit. If the identified margin metric is lower than the preset margin threshold, the baseline estimation unit is placed in the frozen state. In the frozen state, baseline state updates are stopped, and each baseline state quantity is extrapolated based on the baseline temperature slope and the change in temperature of the magnetic sensing unit. Conservation residuals are continuously calculated and retained.

[0116] The basis for updating the baseline temperature slope in tracking mode is that each overdetermined solution, while obtaining each baseline state quantity, also records the temperature of the magnetic sensing unit at the corresponding moment of the solution. Multiple solution results and corresponding temperatures constitute a set of data points, which determine the slope of each channel's baseline state quantity as a function of temperature, i.e., the baseline temperature slope. This slope is updated online during operation, independent of the factory calibration curve; therefore, slope changes caused by packaging stress and device aging after installation can be gradually absorbed into this slope.

[0117] The extrapolation under the frozen state is based on the baseline temperature slope accumulated during the tracking state. After the baseline state update stops, each baseline state quantity no longer changes with the conserved residual; its value is the sum of the baseline state quantity at the start of the frozen state and an increment, which is the product of the baseline temperature slope and the change in temperature relative to the temperature at the start of the frozen state. In the specific implementation, the temperature is taken from a temperature sensing element located near or integrated within the magnetic sensing unit, and its sampling is triggered synchronously with the sampling of the magnetic field response.

[0118] Even in the frozen state, the conserved residuals continue to be calculated and retained. At this point, the conserved residuals no longer drive baseline state updates; their purpose is to preserve the original observation records for backtracking corrections after subsequent recovery tracking. The retained conserved residuals are written to a buffer in the order of sampling time, with the buffer capacity configured according to the expected maximum freeze duration.

[0119] In some embodiments, the baseline temperature slope is taken from the initial value recorded during factory calibration when the system is first powered on and the tracking state has not yet completed its first solution. This initial value is given by the average slope measured in the temperature chamber for the same batch of devices and is only used for extrapolation during the cold start phase. After the tracking state completes its first solution and accumulates enough data points, this initial value is replaced by the slope obtained through online updates and is no longer used thereafter.

[0120] The state switching subunit can also set a recovery threshold higher than the preset margin threshold. When the identification margin metric crosses the preset margin threshold from high to low, it enters the frozen state; when it crosses the recovery threshold from low to high, it switches back to the tracking state. The two thresholds form a hysteresis band. The purpose of the hysteresis band is to prevent the baseline estimation unit from repeatedly switching between the tracking state and the frozen state and thus repeatedly triggering backtracking corrections when the identification margin metric fluctuates around the threshold. When using the hysteresis band, the switching condition of S44 and the backtracking correction subunit below is replaced accordingly with the identification margin metric crossing the recovery threshold from low to high.

[0121] The aforementioned threshold determination is categorized according to the level of the identification margin. Regarding the rate of change of the baseline state variables themselves, in some embodiments, the baseline estimation unit divides the baseline state update into three response levels based on the time scale of the baseline state variable change. When the change time scale is shorter than the sampling time scale of the magnetic field response, the baseline state variables remain unchanged; when the change time scale is longer than the sampling time scale and changes continuously over time, it is continuously updated according to the conserved residual; when the baseline state variables exhibit a step change, it reconverges at an accelerated convergence rate. These three levels correspond to three types of causes: sampling noise, slow temperature drift, and remanent magnetization jump, respectively. The same estimator thus takes into account both slow drift tracking and step response. Whether the baseline state variables exhibit a step change is determined by the jump identification subunit below, based on whether the magnitude of the remanent magnetization increment vector exceeds a preset jump threshold.

[0122] In some embodiments, the baseline estimation unit allocates the forget window length for each magnetic measurement channel according to its resolvability, with channels having higher resolvability corresponding to shorter forget window lengths. The forget window length is the effective tracing length for historical observations during baseline state updates for each magnetic measurement channel, with its upper limit being the sliding time window length. The resolvability of each magnetic measurement channel in the overdetermined observation equation set is not the same; channels directly facing a single current-carrying conductor are generally more resolvable than those located between two conductors. After allocating window lengths by channel, channels with higher resolvability achieve faster tracking speeds with shorter window lengths, without needing to uniformly lengthen the global window to accommodate the channel with the lowest resolvability.

[0123] In some embodiments, the baseline estimation unit further includes a backtracking correction subunit for post-processing the estimation and output during the freeze period after the freeze state ends.

[0124] The backtracking correction sub-unit, in response to the identification margin metric recovering from below a preset margin threshold to above a preset margin threshold, switches the baseline estimation unit from a frozen state back to a tracking state, and tunes the tracking convergence rate according to the duration of the frozen state. The longer the frozen state duration, the greater the potential accumulated deviation between the baseline state quantity extrapolated from the temperature slope during the frozen period and the actual value, requiring faster convergence to the new solution after tracking recovery. When the frozen duration is short, the extrapolated result is close to the actual value, and the convergence rate does not need to deviate from the normal value. The tuning relationship is preset by the actual engineering practice; only the tracking convergence rate is required to change with the frozen duration, without limiting the specific functional form.

[0125] The backtracking correction sub-unit uses the baseline state quantities calculated in the tracking state before and after the start and end times of the frozen state as its two boundary values. It performs forward and backward two-way smoothing on the remaining conserved residuals in the frozen state, recalculating the baseline state quantities at each time point during the frozen state. The conserved residuals within the frozen interval were not adopted at the time but were fully preserved. The two boundary values ​​are taken from the overdetermined solution results of the tracking state, while the original values ​​within the frozen interval are taken from the temperature slope extrapolation; their sources are different. The boundary values ​​at both ends constrain the two sides of the interval, and the baseline state quantities at each time point in the middle are obtained by interpolating the retained residuals between the two known endpoints. This result is simultaneously constrained by the known values ​​at both ends of the interval; its deviation does not monotonically accumulate with the interval length, while the deviation of the unilateral forward extrapolation accumulates with the extrapolation time.

[0126] In practical implementation, forward and reverse two-way smoothing can adopt a fixed interval smoothing form. The forward process iterates time-by-time from the start of the frozen state, obtaining estimates for each time step based only on information prior to that time step; the reverse process iterates time-by-time from the end of the frozen state, obtaining estimates for each time step based only on information after that time step; the results from the two directions are weighted and combined according to their respective reliability to obtain the final estimate for that time step. In other embodiments, two-way smoothing can also adopt a piecewise fitting form between the two boundary values, constraining the shape of the fitted curve with the retained conserved residual and constraining the endpoints of the fitted curve with the two boundary values. Both forms also achieve the purpose of recalculating the baseline state quantities of the frozen interval with bidirectional information.

[0127] The retrospective correction subunit generates correction values ​​and attaches correction labels to each compensated output value output during the frozen state period based on the recalculated baseline state values. The correction value is the difference between the recalculated baseline state value and the baseline state value actually deducted during the frozen period, and is given time-by-time and channel-by-channel. The correction label is output together with the correction value to indicate to downstream that the data for the corresponding time period has been recalculated. The downstream energy accumulation stage makes a difference in the accumulation result for the time period accordingly, and the state of charge estimation stage adjusts the integration starting point for the corresponding time period accordingly.

[0128] In some embodiments, the baseline estimation unit further includes a jump discrimination subunit for identifying baseline steps originating from the remanence of the magnetic sensing unit and triggering reconvergence.

[0129] Before and after a current surge, the transition discrimination subunit takes the magnetic field response quantities of each magnetic measurement channel at the same operating point of the measured current to form a paired response quantity. The channel-by-channel difference between the paired response quantities constitutes the remanent magnetization increment vector. The basis for selecting the same operating point for pairing is that the change in the measured current itself and the change caused by remanent magnetization coexist in the magnetic field response quantity at a single moment, and the two are indistinguishable. When the measured current falls back to the same level as before the surge, the measured current component returns to the value before the surge. At this time, the contribution of the measured current in the difference between the two response quantities cancels each other out, leaving only the part unrelated to the measured current. In specific implementation, whether the operating points are the same is based on whether the values ​​of each component of the measured current fall into the same preset interval. One moment before and after the surge is selected when the values ​​fall into the same interval.

[0130] The transition discrimination subunit determines that the remanent magnetization increment vector originates from the remanent magnetization jump of the magnetic sensing unit if the magnitude of the remanent magnetization increment vector exceeds a preset transition threshold. If the magnitude is below the threshold, the difference is attributed to sampling noise and temperature drift, and no subsequent action is triggered.

[0131] The transition discrimination subunit determines the remanent magnetization direction vector corresponding to each current-carrying conductor based on the geometric coupling coefficient of each magnetic measurement channel to each current-carrying conductor, resulting in a set of remanent magnetization direction vectors with the same number as the number of current-carrying conductors. When a current surge occurs in a current-carrying conductor, the magnetic field generated by that conductor at each magnetic measurement channel is distributed according to the geometric coupling coefficient of each channel to that conductor. Therefore, the remanent magnetization increment left by each channel exhibits a fixed proportional relationship in the channel dimension, and this proportional relationship constitutes the remanent magnetization direction vector corresponding to that conductor. Since the number of current-carrying conductors is finite, the number of candidate directions is also finite. The set of directions is determined by the known geometric coupling coefficients and does not require external reference quantities. In the specific implementation, each remanent magnetization direction vector is first normalized to a unit vector after being constructed.

[0132] The transition identification subunit projects the remanent magnetization increment vector onto each remanent magnetization direction vector, and identifies the current-carrying conductor corresponding to the remanent magnetization direction vector with the largest projection as the current-carrying conductor experiencing the current surge. Normalization is a prerequisite for this comparison: each remanent magnetization direction vector is directly composed of columns of the geometric coupling coefficients, and their magnitudes are not equal. The column norm corresponding to conductors with more direct channels is greater than that of conductors with less coupling. If normalization is not performed first, the magnitude of the projection will vary with the magnitude of the direction vector, and the comparison result will be biased towards the direction with the larger magnitude rather than the direction that is closest. After normalization of each direction vector, the projection only reflects the proximity between the remanent magnetization increment vector and each candidate direction, and the comparison result is independent of the overall scale of the geometric coupling coefficient matrix. In some other embodiments, the cosine of the angle between the remanent magnetization increment vector and each remanent magnetization direction vector can also be used directly as the comparison quantity, and the largest one can be taken, which is equivalent to taking the projection quantity after normalization. The positioning result can be reported along with the compensation output, allowing the operation and maintenance team to trace the location of the surge event.

[0133] The transition discrimination subunit corrects the baseline state variables of the corresponding magnetic measurement channels with each component of the remanence increment vector, and then reconverges each baseline state variable at a convergence rate greater than the tracking convergence rate. Component correction provides a one-time shift of the baseline state variables of each channel, so that the estimator does not have to gradually climb from the old value; the accelerated convergence rate allows subsequent recursion and batch solutions to quickly eliminate residual biases that the component correction could not cover. Remanence transition is a step event, and normal tracking uses a longer sliding time window and a slower convergence rate to ensure resolvability. Without the above treatment, the estimator would need a time on the same order of magnitude as the window length to keep up with the transition.

[0134] The component correction of the transition identification subunit and the state determination of the state switching subunit write the same set of baseline state quantities, and their priorities are defined as follows: Remanent magnetic transitions are deterministic events, so the component correction of the transition identification subunit takes precedence over the state determination of the state switching subunit. When a remanent magnetic transition is determined to occur in the frozen state, the component correction is applied first, then the frozen state ends and reconvergence begins at an accelerated convergence rate, without continuing extrapolation based on the temperature slope. After reconvergence, the identification margin is re-evaluated. If it is still below the preset margin threshold, the baseline estimation unit re-enters the frozen state, using the component-corrected baseline state quantity as the starting value for this freeze, and continues extrapolation based on the baseline temperature slope, while continuing to calculate and retain the conserved residual. If it has exceeded the preset margin threshold, the baseline estimation unit enters the tracking state, and this freeze terminates. The transition moment is written into the retention record of the freeze period. The backtracking correction sub-unit takes this moment as the segment point and performs two-way smoothing on the two intervals before and after the transition: the interval before the transition takes the baseline state quantity calculated in the tracking state before the start of the freeze state and the baseline state quantity after the transition moment component correction as the two boundary values; the interval after the transition takes the baseline state quantity after the transition moment component correction and the baseline state quantity calculated in the tracking state after the end of the freeze as the two boundary values.

[0135] The reconvergence following the aforementioned remanent magnetization jump is predicated on the unequal amplitude weighting coefficient configuration of the second conserved channel group. The remanent magnetization increment left by the current impulse of the same current-carrying conductor in each magnetic measurement channel, i.e., the same-direction remanent magnetization component mentioned above, is distributed according to the geometric coupling coefficient of the conductor to each magnetic measurement channel, and has a definite direction in the channel dimension. If the combined weighting coefficient of the first conserved channel group causes this direction to fall into its null space, then the remanent magnetization jump will not enter the conserved residual of the group, and the baseline translation after the jump cannot be observed; the second conserved channel group is configured with unequal amplitude, so that this direction retains a non-zero projection in the weighted summation, and the baseline translation after the jump thus enters the observation path of the baseline estimation unit. It should be noted that the identification and positioning of the jump identification subunit are taken from the channel-by-channel difference between the paired response quantities, which directly acts on the original magnetic field response quantity, without passing through the combined weighting coefficient and the conserved residual quantity, and therefore is not affected by the weighting coefficient configuration of the conserved channel group; only the reconvergence after the jump is affected by this configuration.

[0136] In some embodiments, the baseline estimation unit further includes a premise verification subunit for monitoring whether the premise of determining the linear conservation relationship still holds.

[0137] The operation of the prerequisite verification subunit is based on the premise that the conservation constraint calculation unit divides the conservation channel groups according to at least two independent and defined linear conservation relationships, and that at least one magnetic measurement channel simultaneously belongs to two conservation channel groups and constitutes a common magnetic measurement channel. The prerequisite verification subunit calculates the baseline state quantity of the common magnetic measurement channel using the conservation residual of each conservation channel group, and determines the difference between the calculation results as the conservation failure indication quantity.

[0138] The specific solution method is as follows: for each conserved channel group, only the magnetic field response of each magnetic measurement channel within the sliding time window at each operating point and the corresponding geometric coupling coefficient of the group are taken. A set of overdetermined observation equations for that group is constructed and solved independently, from which the baseline state quantity corresponding to the common magnetic measurement channel is extracted. Therefore, each of the two conserved channel groups provides an estimate for the same channel. When both deterministic linear conservation relationships hold, the two estimates point to the same physical quantity, and their difference only reflects sampling noise and solution error, with a very small magnitude. When a conservation relationship is violated, the equations constructed based on that relationship no longer hold, and the solution deviates from the true value, resulting in a difference between the two estimates commensurate with the degree of violation.

[0139] When the precondition verification subunit responds to a conservation failure indication exceeding a preset failure threshold, it determines that the linear conservation relationship has been violated, stops baseline state updates, and outputs a precondition failure flag. Causes of conservation relationship violation include wiring changes, branch switching, ground leakage current caused by insulation faults, and conductor strand breakage. In such cases, the measured current component is re-introduced into the conservation residual. If baseline state updates continue, this component will be mistakenly recorded in the baseline state quantity and subtracted channel by channel, thus increasing the measurement deviation.

[0140] During the premise failure period, each baseline state quantity retains the value obtained from the last reliable calculation. The compensation injection unit continues to deduct the value channel by channel according to this value. The confidence flag of the compensation output quantity is set to a separate level and attached with the premise failure flag. In response to the conservation failure indication quantity falling below the preset failure threshold and remaining below it for a preset duration, the premise verification subunit resumes baseline state updates. The compensation output quantity output during the premise failure period is not backtracked and recalculated because the conservation residual quantity retained during this period has itself been mixed with the measured current component and is insufficient to support recalculation. Therefore, the source status of the data during this period is only marked according to the premise failure flag.

[0141] In some embodiments, the prerequisite failure identifier is given separately according to the conservation channel group, so that the operation and maintenance personnel can locate which conservation relationship has been broken, thereby narrowing the scope of verification.

[0142] The compensation injection unit is described below.

[0143] The compensation injection unit is connected to both the magnetic sensing array and the baseline estimation unit. It subtracts the corresponding baseline state value from each magnetic field response to obtain the compensation output. The subtraction is performed channel by channel: the magnetic field response of the first magnetic measurement channel is subtracted from the first baseline state value, the magnetic field response of the second magnetic measurement channel is subtracted from the second baseline state value, and so on for the remaining channels. The compensation output is then converted into a current reading according to a preset scale and output to the downstream circuit.

[0144] The compensation injection unit uses the baseline state value currently maintained by the baseline estimation unit at each sampling time. Therefore, in the tracking state, it uses the result jointly maintained by the recursive orbit and the batch orbit; in the frozen state, it uses the result extrapolated from the baseline temperature slope; and during the premise failure period, it uses the value obtained from the last reliable solution. The usage action does not distinguish between operating states; differences in operating states are handled internally by the baseline estimation unit.

[0145] In some embodiments, the compensation injection unit maps the identified margin measure to a confidence flag of the compensation output quantity and outputs it along with the compensation output quantity. The confidence flag is given in different levels according to the relationship between the identified margin measure and the preset margin threshold. It is high level when the margin is sufficient, low level when it is in a frozen state and includes the freeze duration, and a separate level with a prerequisite failure flag when it is in a prerequisite failure period. The downstream power accumulation and protection criterion links distinguish the source status of the data for each time period based on this. The data for the low-level time period can be confirmed later, and the settlement can be performed after the correction amount given by the backtracking correction subunit arrives.

[0146] In some embodiments, the compensation injection unit sets an upper limit on the variation amplitude of the baseline state variable per unit time. When the variation amplitude exceeds the upper limit, the injection is performed at the upper limit. The baseline estimation unit will provide large changes in the baseline state variable at times such as anchor point overwriting and remanent magnetic transition correction. The limiting ensures that this change enters the compensation output at a limited rate, thus preventing a step in the compensation output. The portion exceeding the upper limit but not injected continues to be injected at the upper limit at subsequent sampling times until all injection is completed. Therefore, the limiting only changes the temporal distribution of the injection, not the total amount injected.

[0147] In contrast, in other embodiments, the compensation injection unit may also subtract a fixed bias value pre-written into non-volatile memory from each magnetic field response quantity. This bias value is measured once at the factory and remains unchanged during operation. This implementation also belongs to the category of subtracting a baseline quantity from the magnetic field response quantity to obtain the compensated output quantity, the difference being that the subtracted quantity is not updated during operation.

[0148] The adaptive tracking compensation method for baseline drift of the magnetic sensor corresponding to the above system is described below. (Refer to...) Figure 3 This method includes steps S1 to S5, and steps S1 to S5 are repeated at each sampling time as the magnetic field response is collected.

[0149] S1. The magnetic field response is acquired using multiple magnetic sensing units located at multiple current-carrying conductors, with each magnetic sensing unit corresponding to a magnetic measurement channel. This step is performed by a magnetic sensing array, the specific implementation of which is described in the previous section on magnetic sensing arrays.

[0150] S2. Divide the magnetic measurement channels whose corresponding measured currents satisfy the determined linear conservation relationship into conservation channel groups. Determine the combination weight coefficient according to the geometric coupling coefficient of each magnetic measurement channel to each current-carrying conductor. The combination weight coefficient makes the measured current component in each magnetic field response cancel each other out in the weighted summation.

[0151] S3. Perform a weighted summation on the magnetic field response quantities within the conserved channel group according to the combination weight coefficients to obtain the conserved residual quantity. The conserved residual quantity represents the linear combination of the baseline drift quantities of each magnetic measurement channel. S2 and S3 are performed by the conserved constraint operation unit. The division basis, the determination method of the combination weight coefficients, and the operation form of the weighted summation are explained in the previous description of the conserved constraint operation unit.

[0152] S4. The baseline drift of each magnetic measurement channel is characterized as a baseline state quantity. The baseline state is updated using the conserved residual as the observation information. An overdetermined observation equation set is constructed using the magnetic field response quantities of multiple operating points within the sliding time window. The overdetermined observation equation set is solved in conjunction with the geometric coupling coefficient to obtain the baseline state quantity of each magnetic measurement channel. S4 is performed by the baseline estimation unit. The construction of the overdetermined observation equation set, the solvability conditions, and the handling of unidentifiable components are explained in the previous section on the baseline estimation unit.

[0153] S5. Subtract the corresponding baseline state quantity from each magnetic field response quantity to obtain the compensated output quantity. S5 is executed by the compensation injection unit.

[0154] In some embodiments, refer to Figure 4 S4 includes sub-steps S41 to S43.

[0155] S41. Determine the identification margin based on the resolvability of each baseline state quantity according to the overdetermined observation equation set.

[0156] S42. In response to the identification margin being higher than the preset margin threshold, enter the tracking state, perform overdetermined solution according to the tracking convergence rate in the tracking state, and update the baseline temperature slope with the correspondence between the solution results of each baseline state quantity and the temperature of the magnetic sensitive unit.

[0157] S43. In response to the identification margin being lower than the preset margin threshold, enter the freeze state. In the freeze state, stop updating the baseline state, extrapolate each baseline state quantity according to the baseline temperature slope and the change in temperature of the magnetic sensing unit, and continuously calculate and retain the conserved residual quantity. S41 to S43 correspond to the actions of the margin evaluation subunit and the state switching subunit, respectively.

[0158] In some embodiments, combined with Figure 4 S4 also includes sub-steps S44 to S46.

[0159] S44. In response to the recognition margin recovering from below the preset margin threshold to above the preset margin threshold, switch from the frozen state back to the tracking state, and adjust the tracking convergence rate according to the duration of the frozen state.

[0160] S45. Using the baseline state quantities calculated in the tracking state before the start time and after the end time of the frozen state as the two boundary values, perform forward and reverse two-way smoothing on each of the conserved residual quantities remaining in the frozen state, and recalculate the baseline state quantities at each time during the frozen state.

[0161] S46. Based on the recalculated baseline state quantities, generate correction values ​​for each compensation output quantity that was output during the frozen state and attach correction identifiers. S44 to S46 correspond to the actions of the backtracking correction subunit.

[0162] In some embodiments, refer to Figure 5 S4 also includes sub-step S4a parallel to S41 to S46, and S4a includes sub-steps S4a1 to S4a5.

[0163] S4a1. Before and after the current surge of the measured current, the magnetic field response quantities of each magnetic measurement channel at the same operating point of the measured current are used to form a paired response quantity, and the channel-by-channel difference between the paired response quantities forms the residual magnetism increment vector.

[0164] S4a2. In response to the magnitude of the remanent magnetization increment vector exceeding a preset jump threshold, the remanent magnetization increment vector is determined to be a remanent magnetization jump originating from the magnetic sensitive unit.

[0165] S4a3. Determine the remanent magnetization direction vector corresponding to each current-carrying conductor according to the geometric coupling coefficient of each magnetic measurement channel to each current-carrying conductor, and obtain a set of remanent magnetization direction vectors with the same number as the number of current-carrying conductors.

[0166] S4a4. Project the residual magnetism increment vector onto each residual magnetism direction vector respectively, and determine the current-carrying conductor corresponding to the residual magnetism direction vector with the largest projection as the current-carrying conductor that will experience current surge.

[0167] S4a5. The baseline state quantities of the corresponding magnetic measurement channels are corrected according to each component of the remanent magnetization increment vector, and the baseline state quantities are re-converged at a convergence rate greater than the tracking convergence rate. The actions of the transition discrimination subunit corresponding to S4a1 to S4a5 are executed in parallel with S41 to S46, and the two share the same set of baseline state quantities. The priority of writing is explained in the previous description of the transition discrimination subunit. In S4a3, each remanent magnetization direction vector is normalized to a unit vector after being constructed. The projection comparison of S4a4 is performed after normalization.

[0168] In some embodiments, in S2, several magnetic measurement channels are divided into a first conserved channel group and a second conserved channel group. The combined weight coefficients of the first conserved channel group are configured with equal amplitude, while the combined weight coefficients of the second conserved channel group are configured with unequal amplitude. The equal amplitude configuration here refers to the configuration when the first conserved channel group only performs the function of cancelling the measured current component. This configuration is suitable for symmetrical arrangements where the geometric coupling coefficients of each channel to its opposite conductor are equal. In this case, equal amplitude addition can cancel the measured current component. When the first conserved channel group also performs the function of cancelling the environmental common-mode magnetic field component, as described above, under the condition that the sum of each weight coefficient is zero when the number of magnetic measurement channels is greater than the number of current-carrying conductors, the weight coefficients are no longer equal in amplitude.

[0169] In some embodiments, in S3, the rate of change of the measured current between adjacent time periods is used for gating, and the conserved residual obtained in time periods with a rate of change higher than a preset gating threshold does not participate in the baseline state update.

[0170] The same or similar parts in the various embodiments of this application may be referred to each other; the technical features of each embodiment may be combined with each other to form new embodiments without conflict.

[0171] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device / system can be divided into different functional units or modules to complete all or part of the functions described above.

[0172] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. An adaptive tracking compensation system for baseline drift of a magnetic sensor, characterized in that, include: A magnetic sensing array includes multiple magnetic sensing units respectively disposed at multiple current-carrying conductors, each of the magnetic sensing units constituting a magnetic measurement channel and outputting a magnetic field response quantity; A conservation constraint calculation unit, connected to the magnetic sensing array signal, is configured to divide several magnetic measurement channels whose corresponding measured currents satisfy a defined linear conservation relationship into conservation channel groups. It determines combination weighting coefficients based on the geometric coupling coefficients of each magnetic measurement channel to each current-carrying conductor, and performs a weighted summation on each magnetic field response quantity within the conservation channel group according to the combination weighting coefficients to obtain a conservation residual. The combination weighting coefficients cause the measured current component in each magnetic field response quantity to cancel out in the weighted summation. The conservation residual represents a linear combination of the baseline drift quantities of each magnetic measurement channel. A baseline estimation unit, connected to the... A conservation constraint operation unit is signal-connected and configured to characterize the baseline drift of each of the magnetic measurement channels as a baseline state quantity, perform baseline state update using the conservation residual as observation information, construct an overdetermined observation equation set using the magnetic field response quantities of multiple operating points within a sliding time window, and perform overdetermined solution on the overdetermined observation equation set in conjunction with the geometric coupling coefficient to obtain the baseline state quantity of each of the magnetic measurement channels; a compensation injection unit is signal-connected to the magnetic sensing array and the baseline estimation unit respectively, and is configured to subtract the corresponding baseline state quantity from each of the magnetic field response quantities to obtain a compensation output quantity.

2. The adaptive tracking compensation system for magnetic sensor baseline drift according to claim 1, characterized in that, The baseline estimation unit includes a margin assessment subunit and a state switching subunit. The margin assessment subunit is configured to determine the identification margin based on the resolvability of each baseline state quantity according to the overdetermined observation equation set. The state switching subunit is configured to: in response to the identification margin being higher than a preset margin threshold, place the baseline estimation unit in a tracking state, perform the overdetermined solution at the tracking convergence rate in the tracking state, and update the baseline temperature slope based on the correspondence between the solution results of each baseline state quantity and the temperature at which the magnetic sensing unit is located. In response to the identification margin being lower than the preset margin threshold, the baseline estimation unit is placed in a frozen state, the baseline state update is stopped in the frozen state, each baseline state quantity is extrapolated according to the baseline temperature slope and the change in temperature of the magnetic sensing unit, and the conserved residual quantity is continuously calculated and retained.

3. The adaptive tracking compensation system for magnetic sensor baseline drift according to claim 2, characterized in that, The baseline estimation unit further includes a backtracking correction subunit, which is configured to: in response to the identification margin recovering from below the preset margin threshold to above the preset margin threshold, switch the baseline estimation unit from the frozen state back to the tracking state, and adjust the tracking convergence rate according to the duration of the frozen state; Using the baseline state quantities calculated in the tracking state before and after the start and end times of the frozen state as the two boundary values, forward and reverse two-way smoothing is performed on each of the conserved residual quantities remaining in the frozen state, and the baseline state quantities at each time during the frozen state are recalculated; based on the recalculated baseline state quantities, correction quantities are generated for each of the compensation output quantities that have been output during the frozen state, and correction identifiers are attached.

4. The adaptive tracking compensation system for magnetic sensor baseline drift according to claim 3, characterized in that, The conservation constraint calculation unit is further configured to divide the plurality of magnetic measurement channels into a first conservation channel group and a second conservation channel group and perform the weighted summation respectively to obtain the corresponding conservation residuals; the combination weight coefficients of the first conservation channel group are configured in such a way that the environmental common-mode magnetic field components commonly experienced by each magnetic measurement channel cancel each other out in the weighted summation; the combination weight coefficients of the second conservation channel group are configured with unequal amplitudes so that the same-direction remanent magnetization components generated by the current impact of the same current-carrying conductor in each magnetic measurement channel retain a non-zero projection in the weighted summation; the baseline estimation unit further includes a jump discrimination subunit, which is configured to: before and after the current impact occurs, take the magnetic field response of each magnetic measurement channel with the measured current at the same operating point to form a paired response. The remanent magnetization increment vector is formed by the channel-by-channel difference between the paired response quantities. In response to the magnitude of the remanent magnetization increment vector exceeding a preset jump threshold, the remanent magnetization increment vector is determined to be a remanent magnetization jump originating from the magnetic sensing unit. The remanent magnetization direction vector corresponding to each current-carrying conductor is determined according to the geometric coupling coefficient of each magnetic measurement channel to each current-carrying conductor, resulting in a set of remanent magnetization direction vectors with the same number of roots as the current-carrying conductors. The remanent magnetization increment vector is projected onto each of the remanent magnetization direction vectors, and the current-carrying conductor corresponding to the remanent magnetization direction vector with the largest projection is determined as the current-carrying conductor experiencing the current surge. The baseline state quantities of the corresponding magnetic measurement channels are corrected using each component of the remanent magnetization increment vector, and the baseline state quantities are re-converged at a convergence rate greater than the tracking convergence rate.

5. The adaptive tracking compensation system for magnetic sensor baseline drift according to claim 1, characterized in that, The conservation constraint calculation unit is further configured to divide the conservation channel groups according to at least two independent defined linear conservation relationships, wherein at least one magnetic measurement channel belongs to two conservation channel groups and constitutes a common magnetic measurement channel; the baseline estimation unit further includes a premise verification subunit, which is configured to calculate the baseline state quantity of the common magnetic measurement channel using the conservation residual of each conservation channel group, and determine the difference between the calculation results as a conservation failure indication quantity; In response to the conservation failure indication exceeding a preset failure threshold, it is determined that the established linear conservation relationship has been violated, the baseline state update is stopped, and a prerequisite failure flag is output.

6. The adaptive tracking compensation system for magnetic sensor baseline drift according to claim 1, characterized in that, The conservation constraint calculation unit further includes a hierarchical identification subunit, which is configured to: select time periods within the sliding time window where the change amplitude of the measured current is higher than the upper amplitude threshold to form a coupling identification sample set; select time periods where the change amplitude of the measured current is lower than the lower amplitude threshold to form a baseline identification sample set, wherein the upper amplitude threshold is higher than the lower amplitude threshold; use the baseline state variables currently calculated from the baseline identification sample set as known quantities, and inversely solve the geometric coupling coefficient based on the correlation between the magnetic field response variables within the coupling identification sample set; then use the inversely solved geometric coupling coefficient as known quantities, and call the baseline estimation unit to recalculate the baseline state variables from the baseline identification sample set, and perform this process alternately. The change in the geometric coupling coefficient between two adjacent rounds of alternating execution is determined as the coupling convergence amount, and the change in each baseline state quantity between two adjacent rounds of alternating execution is determined as the baseline convergence amount; In response to the fact that the coupling convergence and the baseline convergence both fall within their respective preset convergence thresholds, the alternating execution is terminated and the combined weight coefficients are updated with the geometric coupling coefficients obtained from the inverse solution. If only one of the coupling convergence amount and the baseline convergence amount falls within the corresponding preset convergence threshold, it is determined that the geometric coupling coefficient and the baseline state quantity are indistinguishable. The alternating execution is then re-executed after the upper amplitude threshold is increased and the lower amplitude threshold is decreased.

7. An adaptive tracking compensation method for baseline drift of a magnetic sensor, characterized in that, The process includes the following steps, and S1 to S5 are repeated at each sampling time as the magnetic field response is acquired: S1. The magnetic field response is acquired using multiple magnetic sensing units respectively set at multiple current-carrying conductors, and each magnetic sensing unit corresponds to a magnetic measurement channel; S2. The magnetic measurement channels corresponding to the measured currents that satisfy a certain linear conservation relationship are divided into conservation channel groups, and the combination weight coefficient is determined according to the geometric coupling coefficient of each magnetic measurement channel to each current-carrying conductor. The combination weight coefficient makes the measured current component in each magnetic field response cancel each other out in the weighted summation; S3. Perform the weighted summation on each of the magnetic field responses in the conserved channel group according to the combined weighting coefficients to obtain the conserved residual, which represents the linear combination of the baseline drift of each of the magnetic measurement channels; S4. Characterize the baseline drift of each magnetic measurement channel as a baseline state quantity, use the conserved residual as the observation information to perform baseline state update, and construct an overdetermined observation equation set using the magnetic field response quantities of multiple operating points within the sliding time window. Combine the geometric coupling coefficient to perform overdetermined solution on the overdetermined observation equation set to obtain the baseline state quantity of each magnetic measurement channel; S5. Subtract the corresponding baseline state quantity from each magnetic field response quantity to obtain the compensation output quantity.

8. The adaptive tracking compensation method for magnetic sensor baseline drift according to claim 7, characterized in that, S4 includes the following sub-steps: S41. Determine the identification margin based on the resolvability of each baseline state quantity according to the overdetermined observation equation set; S42. In response to the identification margin being higher than a preset margin threshold, enter the tracking state, perform the overdetermined solution at the tracking convergence rate in the tracking state, and update the baseline temperature slope with the correspondence between the solution results of each baseline state quantity and the temperature of the magnetic sensing unit. S43. In response to the identification margin being lower than the preset margin threshold, enter a frozen state, stop the baseline state update in the frozen state, extrapolate each of the baseline state quantities according to the baseline temperature slope and the change in temperature of the magnetic sensing unit, and continuously calculate and retain the conserved residual quantity.

9. The adaptive tracking compensation method for magnetic sensor baseline drift according to claim 8, characterized in that, S4 further includes the following sub-steps: S44. In response to the identification margin metric recovering from below the preset margin threshold to above the preset margin threshold, switching back from the frozen state to the tracking state, and adjusting the tracking convergence rate according to the duration of the frozen state; S45. Using the baseline state quantities calculated in the tracking state before the start time and after the end time of the frozen state as the two boundary values, performing forward and reverse two-way smoothing on each of the conserved residual quantities remaining in the frozen state, and recalculating the baseline state quantities at each time during the frozen state; S46. Based on the recalculated baseline state quantities, generate correction values ​​for each of the compensation output quantities that were output during the frozen state and attach correction identifiers.

10. The adaptive tracking compensation method for magnetic sensor baseline drift according to claim 9, characterized in that, S4 further includes a sub-step S4a parallel to S41 to S46, which includes the following sub-steps: S4a1. Before and after the measured current experiences a current surge, the magnetic field response quantities of each of the magnetic measurement channels at the same operating point of the measured current are taken to form a paired response quantity, and the channel-by-channel difference between the paired response quantities forms a remanent magnetization increment vector; S4a2. In response to the magnitude of the remanent magnetization increment vector exceeding a preset jump threshold, the remanent magnetization increment vector is determined to be a remanent magnetization jump originating from the magnetic sensing unit; S4a3. According to the geometric coupling coefficient of each magnetic measurement channel to each current-carrying conductor, the remanent magnetization direction vector corresponding to each current-carrying conductor is determined, resulting in a set of remanent magnetization direction vectors with the same number as the number of current-carrying conductors; S4a4. Project the residual magnetism increment vector onto each of the residual magnetism direction vectors respectively, and determine the current-carrying conductor corresponding to the residual magnetism direction vector with the largest projection as the current-carrying conductor that experiences the current surge; S4a5. Correct the baseline state quantity of the corresponding magnetic measurement channel with each component of the residual magnetism increment vector, and re-converge each of the baseline state quantities at a convergence rate greater than the tracking convergence rate.

Images