Data-driven power grid equipment digital twin dynamic updating method and system

Abstract

The present disclosure provides a data-driven power grid equipment digital twin dynamic updating method and system, which comprises establishing a reclosing event time scale chain and demarcating the residual magnetism boundary of each iron core column before reclosing; dividing the model parameters of the digital twin into a steady-state parameter layer and a historical magnetic state layer, arranging particle elements and setting a search boundary; generating an initial particle swarm, calculating the fitness of each particle, iteratively updating the particle position until the termination condition is met, and outputting the optimal residual magnetism state combination and residual magnetism decay trajectory; loading into the historical magnetic state layer and performing reclosing simulation, respectively verifying the transient response characteristics and steady-state recovery characteristics, advancing the residual magnetism decay according to the residual magnetism decay trajectory within the outage interval, and through nonlinear excitation calculation after closing, gradually exiting the one-way magnetic flux bias component, and when the updating consistency meets the requirements, the updated results are solidified. The method realizes continuous and accurate description of the transient and steady-state characteristics of the digital twin during the whole process of transformer reclosing.

Description

Technical Field

[0001] This disclosure belongs to the field of power system automation and digital twin technology, and more specifically, relates to a data-driven method and system for dynamic updating of digital twins of power grid equipment. Background Technology

[0002] Digital twins continuously receive equipment operating parameters and update their internal models to achieve state monitoring and operational simulation of core equipment such as transformers. Traditional update methods mainly correct conventional parameters such as heat dissipation coefficient, loss coefficient, or insulation degradation coefficient based on the current response. However, when a transformer recloses after a short-term power outage due to an external fault, residual magnetism remains inside the core, which is closely related to the time of power outage, the switching phase, and the direction of magnetic flux before the fault, and has a clear directionality. Under the same power outage duration and closing voltage, residual magnetism in different directions may lead to severe inrush current, slight inrush current, or local magnetic saturation, respectively. Existing technical solutions fail to distinguish the essential physical differences between residual magnetism and conventional model parameters, easily absorbing transient responses caused by residual magnetism into winding losses or excitation parameters. When the residual magnetism naturally dissipates over time, the incorrectly modified parameters cause significant deviations in the normal operation model, resulting in the same digital twin being unable to accurately describe both the reclosing transient and subsequent steady-state processes simultaneously.

[0003] Therefore, there is an urgent need for a dynamic update method that separates and independently updates the historical state of residual magnetism in the iron core from the steady-state model parameters, so as to achieve a continuous and accurate description of the transient and steady-state characteristics of the digital twin throughout the entire reclosing process of the transformer. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to resolve the aforementioned deficiencies and propose a data-driven method and system for dynamic updating of digital twins of power grid equipment.

[0005] The present invention adopts the following technical solution.

[0006] The first aspect of this invention discloses a data-driven method for dynamically updating a digital twin of power grid equipment, the method comprising:

[0007] S1. Establish a reclosing event time-stamped chain, obtain the estimated magnetic flux value, magnetic flux direction and magnetic flux calculation trajectory of each iron core column at the moment of opening, delineate the residual magnetic boundary of each iron core column before reclosing, and form a magnetic state history sequence and a reclosing actual response sequence.

[0008] S2, the model parameters of the digital twin are divided into a steady-state parameter layer and a historical magnetic state layer. The particle elements are arranged according to the signed remanence, hysteresis position and remanence decay degree of each iron core column and the search boundary is set. Fitness evaluation rules are established and particle swarm execution parameters are set to form a particle swarm identification configuration.

[0009] S3, generate an initial particle swarm, calculate the fitness of each particle according to the fitness evaluation rule, iteratively update the particle position until the termination condition is met, and output the optimal combination of remanent magnetization states and the remanent magnetization decay trajectory.

[0010] S4, load the optimal residual magnetism state combination into the historical magnetic state layer and perform reclosing simulation to verify the transient response characteristics and steady-state recovery characteristics respectively. When both transient and steady-state verifications are qualified, encapsulate the residual magnetism dynamic update package.

[0011] S5, bind the residual magnetism dynamic update package to the historical magnetic state layer, advance the residual magnetism decay according to the residual magnetism decay trajectory during the power outage interval, and after closing the circuit, use nonlinear excitation calculation to make the unidirectional magnetic flux bias component gradually withdraw, and solidify the update result when the update consistency meets the requirements.

[0012] Optionally, S1 includes:

[0013] S1.1 According to the time of external fault occurrence, the time of each phase tripping, the time of all winding de-energization, the time of each phase closing, and the time of inrush current attenuation end, the reclosing process is divided into continuous event stages, a reclosing event time stamp chain is established, and a state table of the remaining circuits of the windings during the power outage is formed.

[0014] S1.2, Select the starting interval of integration before the external fault occurs, and obtain the estimated value of magnetic flux, magnetic flux direction and magnetic flux calculation trajectory of each iron core column at the moment of opening through numerical integration;

[0015] S1.3, Based on the remaining circuit status table of the winding during the power outage and the estimated magnetic flux at the moment of opening, determine the degree of residual magnetic decay of each iron core column before closing according to the residual magnetic decay law during the power outage, and delineate the upper and lower boundaries of residual magnetic flux of each iron core column.

[0016] S1.4, according to the order of operation before the fault, occurrence of external fault, tripping of each phase, power outage interval, closing of each phase and inrush current attenuation, connect the magnetic flux calculation trajectory and residual magnetism boundary into a complete time sequence to form a magnetic state history sequence; extract the current response characteristics after reclosing to form the actual reclosing response sequence.

[0017] Optionally, S2 includes:

[0018] S2.1, Based on the magnetic state history sequence, the model parameters of the digital twin are divided into a steady-state parameter layer and a historical magnetic state layer. The steady-state parameter layer contains electromagnetic parameters and structural parameters that reflect normal operation characteristics and is locked as read-only during the remanence identification period. The historical magnetic state layer contains the signed remanence, hysteresis position and remanence decay degree of each core column and is allowed to be updated during the identification period.

[0019] S2.2, Based on the historical magnetic state layer, and using the magnetic flux calculation trajectory and the upper and lower boundaries of remanence, arrange the signed remanence, hysteresis position and remanence decay degree of each core column as particle elements according to the core column number, and set the search boundary of each particle element.

[0020] S2.3, Establish particle fitness evaluation rules. The rules calculate fitness based on the deviation between the current response of the reclosing simulation and the actual reclosing response sequence, and impose penalties on particles that exceed the search boundary.

[0021] S2.4, based on the results of S2.1 to S2.3, sets the execution parameters of the particle swarm algorithm, including the number of particles, inertia weight, learning coefficient and maximum number of iterations, to form the particle swarm identification configuration.

[0022] Optionally, S3 includes:

[0023] S3.1, Generate an initial particle swarm according to the arrangement order of the iron core columns, so that each particle element is evenly distributed in its respective search interval, record the steady-state parameters to lock the snapshot, and form a reclosing event driving sequence.

[0024] S3.2, read each particle one by one, write the particle position into the historical magnetic state layer and restore the steady-state parameter layer to the steady-state parameter locking snapshot, execute the simulation according to the reclosing event driven sequence, calculate the fitness of each particle according to the fitness evaluation rule, register the initial position of each particle as its historical best position, and register the position of the particle with the smallest fitness as the initial global best position;

[0025] S3.3 Iteratively update particle velocity and position. After each update, re-execute the simulation and calculate the fitness of each particle. Based on this, update the historical best position of each particle and the global best position of the particle swarm until the termination condition is met.

[0026] S3.4 After terminating the iteration, output the optimal combination of remanent magnetization states and the remanent magnetization decay trajectory.

[0027] Optionally, S3.3 includes:

[0028] For each particle element, the particle velocity is iteratively updated according to the velocity update method jointly guided by the inertia weight, individual learning coefficient and group learning coefficient, and the updated velocity is constrained to prevent the particle from crossing the search boundary.

[0029] The particle positions are iteratively updated according to the position update method, and the updated positions are clipped to within the preset search boundary. After each round of position update, the updated particles are written into the historical magnetic state layer and the steady-state parameter layer is restored to the steady-state parameter locking snapshot. The reclosing simulation is then re-executed and the fitness is calculated.

[0030] Update the historical best position of each particle and the global best position of the particle swarm based on the fitness of each particle.

[0031] The termination of the iteration is determined by a convergence index, which reflects both the degree of improvement in the global optimal fitness and the degree of dispersion of all particles around the global optimal position.

[0032] Optionally, S4 includes:

[0033] S4.1 Load the historical magnetic state layer from the optimal residual magnetism state to the opening time, lock the steady-state parameter layer to read-only state, advance the core magnetic state change within the power outage interval according to the residual magnetism decay trajectory, execute the closing operation at the actual closing time to generate the reclosing simulation state sequence, and divide the transient verification interval and steady-state verification interval with the end time of inrush current decay as the boundary.

[0034] S4.2 Extract the transient response features of the reclosing simulation state sequence within the transient verification interval, compare them with the actual reclosing response sequence, calculate the transient continuity index, and form a transient verification qualified conclusion when the threshold is met.

[0035] S4.3 After the transient verification is qualified, the reclosing simulation state sequence is verified from four dimensions within the steady-state verification interval: current recovery, magnetic flux period consistency, historical state exit, and steady-state parameter invariance. The steady-state recovery index is calculated. When the threshold is met and the steady-state parameter remains locked, the steady-state verification is qualified. The unidirectional magnetic flux bias decay time constant and steady-state parameter locking snapshot after closing are determined.

[0036] S4.4 When both transient and steady-state verifications are qualified, the optimal combination of residual magnetism states, residual magnetism attenuation trajectory, unidirectional magnetic flux bias attenuation time constant after closing, and steady-state parameter lock snapshot are encapsulated into a residual magnetism dynamic update package; when any verification fails, the corresponding boundary is corrected and the process returns to step S3 for re-identification.

[0037] Optionally, S5 includes:

[0038] S5.1, Read data from the residual magnetism dynamic update package and bind it to the historical magnetic state layer;

[0039] S5.2, at the moment of each phase tripping, the signed residual magnetism and hysteresis position in the optimal residual magnetism state combination are written into the historical magnetic state layer of the corresponding iron core column. During the power outage interval, the residual magnetism amplitude of each iron core column is advanced hourly according to the residual magnetism decay trajectory, while keeping the residual magnetism direction unchanged.

[0040] S5.3, At the closing time of each phase, the closing operation is performed. The residual magnetic state at the moment before closing is used as the initial condition to start the nonlinear excitation calculation, so that the core magnetic flux after closing is composed of steady-state periodic magnetic flux component and unidirectional magnetic flux bias component. The unidirectional magnetic flux bias component gradually exits as the excitation inrush current decays, generating the magnetic state evolution process.

[0041] S5.4 After the closing operation is completed, the consistency index is updated from four dimensions: current recovery, magnetic flux cycle consistency, historical state exit and steady-state parameter unchanged within several consecutive power frequency cycles. When the threshold is met and the steady-state parameter is locked, the current residual magnetism state, magnetic state evolution process and steady-state parameter are solidified to form the dynamic update result of the digital twin.

[0042] Secondly, this disclosure provides a system applied to the method of the first aspect, the system comprising:

[0043] The magnetic state evolution module is used to establish a reclosing event time-stamped chain, obtain the estimated magnetic flux value, magnetic flux direction and magnetic flux calculation trajectory of each iron core column at the moment of opening, delineate the residual magnetic boundary of each iron core column before reclosing, and form a magnetic state history sequence and a reclosing actual response sequence.

[0044] The updated structure building module is used to divide the model parameters of the digital twin into a steady-state parameter layer and a historical magnetic state layer. The particle elements are arranged according to the signed remanence, hysteresis position and remanence decay degree of each iron core column and the search boundary is set. Fitness evaluation rules are established and particle swarm execution parameters are set to form a particle swarm identification configuration.

[0045] The particle swarm identification module is used to generate an initial particle swarm, calculate the fitness of each particle according to the fitness evaluation rules, iteratively update the particle positions until the termination condition is met, and output the optimal combination of remanent magnetization states and the remanent magnetization decay trajectory.

[0046] The verification module is used to load the optimal residual magnetism state combination into the historical magnetic state layer and perform reclosing simulation to verify the transient response characteristics and steady-state recovery characteristics respectively. When the verification is qualified, the residual magnetism dynamic update package is encapsulated.

[0047] The dynamic update module is used to bind the residual magnetism dynamic update package to the historical magnetic state layer. During the power outage interval, the residual magnetism decay is promoted according to the residual magnetism decay trajectory. After the circuit is closed, the unidirectional magnetic flux bias component is gradually withdrawn through nonlinear excitation calculation. When the update consistency meets the requirements, the update result is solidified.

[0048] Thirdly, this disclosure provides an electronic device including a memory and at least one processor, the memory storing a computer program, and the processor executing the computer program to implement the method of the first aspect described above.

[0049] Fourthly, this disclosure provides a computer storage medium storing a computer program that, when executed, implements the method described in the first aspect.

[0050] The beneficial effects of the present invention are as follows: Compared with the prior art, the present invention has the following advantages:

[0051] (1) By dividing the model parameters of the digital twin into a steady-state parameter layer and a historical magnetic state layer, the steady-state parameter layer is locked as read-only during the residual magnetism identification period, and only the historical magnetic state layer is updated and identified. This physically distinguishes the residual magnetism of the iron core from the conventional model parameters, effectively avoiding the erroneous absorption of the transient response caused by residual magnetism into the winding loss, rated excitation characteristics or normal iron loss parameters during the particle swarm iteration process. This eliminates the problem of significant deviation in the normal operation model caused by the erroneously modified steady-state parameters after the natural decay of residual magnetism, enabling the same digital twin to continuously and accurately describe the reclosing transient and subsequent steady-state processes.

[0052] (2) By constructing a short-term power outage reclosing magnetic state evolution sequence and defining the residual magnetism physical boundary of each core column, the search space of the particle swarm optimization algorithm is constrained within a physically reasonable magnetic flux range. Multi-objective optimization is then performed by combining fitness evaluation rules that simultaneously constrain peak current, waveform bias, three-phase asymmetry, attenuation duration, and steady-state recovery deviation. This ensures that the identified residual magnetism state can accurately reproduce the inrush current impact characteristics during the reclosing transient process and that the digital twin can fully recover to the normal steady-state operation mode after the inrush current subsides. Based on this, by performing dual verification of the transient and steady-state continuity of the identification results, the simulation accuracy and physical consistency of the updated digital twin throughout the entire reclosing process are guaranteed, providing reliable technical support for the condition monitoring and operation simulation of power grid equipment. Attached Figure Description

[0053] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0054] Figure 1 A flowchart of a data-driven method for dynamically updating digital twins of power grid equipment, as provided in an embodiment of this disclosure, is shown.

[0055] Figure 2 This illustration shows a flowchart of the transient and steady-state dual verification and update package encapsulation provided in an embodiment of this disclosure;

[0056] Figure 3 A schematic diagram of the structure of a data-driven digital twin dynamic update system for power grid equipment provided in an embodiment of this disclosure is shown.

[0057] The accompanying drawings have illustrated specific embodiments of this disclosure, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concepts of this disclosure to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0058] The present disclosure will be further described below with reference to the accompanying drawings. The following embodiments are only used to illustrate the technical solutions of the present disclosure more clearly, and should not be used to limit the scope of protection of the present disclosure.

[0059] The components of the embodiments of the invention described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0060] Figure 1 A flowchart of a data-driven dynamic update method for digital twins of power grid equipment provided in this disclosure is shown below. Figure 1 As shown, the method may include the following steps:

[0061] S1. Establish a reclosing event time-stamped chain, obtain the estimated magnetic flux value, magnetic flux direction and magnetic flux calculation trajectory of each core column at the moment of opening, delineate the residual magnetic boundary of each core column before reclosing, and form a magnetic state history sequence and a reclosing actual response sequence.

[0062] S1.1: Based on the time of external fault occurrence, the time of tripping of each phase, the time of de-energization of all windings, the time of closing of each phase, and the time of end of inrush current decay, the reclosing process is divided into continuous event stages, a reclosing event time stamp chain is established, and a state table of the remaining circuits of the windings during the power outage is formed.

[0063] Obtain the rated parameters of the transformer, including winding connection method, core structure, rated voltage, rated current, winding resistance, winding leakage inductance and rated magnetic flux. At the same time, obtain the voltage and current waveform data during the steady-state operation before the fault, the voltage and current waveform data during the external fault, the opening and closing times of each phase circuit breaker, the power outage interval duration, and the waveform data of each phase current after closing.

[0064] Based on the timing of external fault occurrence, phase tripping, all winding de-energization, phase closing, and inrush current decay, the entire reclosing process is divided into multiple consecutive event stages, establishing a reclosing event timescale chain. For star windings, the measured values ​​of phase voltage and phase current are directly used; for delta windings, line voltage and line current are converted into phase voltage and phase current before use. The path state of the zero-sequence current is determined according to the neutral point grounding method. The rate of change of the core magnetic flux is obtained by subtracting the winding resistance voltage drop and leakage inductance voltage drop from the winding terminal voltage, and then numerically integrating over time to obtain the magnetic flux waveform.

[0065] Based on the actual installation position of the windings on the core column, establish a mapping relationship between each phase winding and its corresponding core column. Check whether a closed path formed by a delta winding exists after the circuit breaker is tripped, whether a zero-sequence path formed by neutral grounding exists, and whether an additional circuit formed by auxiliary loads exists. Based on this, create a status table of the remaining winding circuits during the power outage. This status table records whether each phase winding still forms an electrical path through external circuits after the power outage, and the degree of impact of this path on the decay of the core's residual magnetism.

[0066] S1.2: Select the starting interval of integration before the external fault occurs, and obtain the estimated value of magnetic flux, magnetic flux direction and magnetic flux calculation trajectory of each iron core column at the moment of opening through numerical integration.

[0067] Using the established reclosing event time-stamp chain and the mapping relationship between winding phases and core columns, 5-6 power frequency cycles before the occurrence of an external fault are selected as the starting interval for flux integration. Within this starting interval, the transformer is in normal operation, and the core flux exhibits a periodic change corresponding to the voltage, thereby determining the flux value at the start of integration. For the k-th core column, the net voltage after deducting the winding resistance voltage drop and leakage inductance voltage drop from the winding terminal voltage is integrated from the start of integration to the opening time of the line phase containing that core column to obtain the estimated flux value at the moment of opening. This calculation is achieved using the following formula: ; in, This is the estimated magnetic flux value at the instant the k-th core column is opened; The magnetic flux at the start of integration is determined according to the condition that the magnetic flux at the beginning and end of the period before the fault is equal. This refers to the number of turns in the winding. This is the starting time of integration; The tripping time of the line phase containing the k-th iron core column; This refers to the phase voltage of the winding. For winding resistance; This refers to the phase current of the winding. t represents the leakage inductance of the winding; t represents time. The direction of the magnetic flux at the moment of tripping is determined by the sign of the above estimated magnetic flux value. A positive value indicates that the magnetic flux is along the reference positive direction of the iron core, and a negative value indicates that the magnetic flux is along the reference negative direction.

[0068] The above calculations preserve the influence of the initial phase at the time of external fault occurrence and the asynchronous timing of each phase tripping on the direction of the core magnetic flux, thus avoiding the uncertainty caused by inferring the direction of residual magnetism solely based on the voltage and current after tripping.

[0069] S1.3: Based on the remaining circuit status table of the winding during the power outage and the estimated magnetic flux at the moment of opening, determine the degree of residual magnetic decay of each core column before closing according to the residual magnetic decay law during the power outage, and delineate the upper and lower boundaries of residual magnetic flux of each core column.

[0070] Based on the established state table of the remaining winding circuit during the power outage, as well as the estimated magnetic flux, direction, and calculated trajectory at the moment of tripping, and according to the residual magnetism decay law during the power outage, including the power outage interval duration, the hysteresis retention characteristics of the core material, and the demagnetizing effect of the remaining winding circuit, the tripping magnetic flux is decomposed into a long-term retention component and a component that decays with power outage time. Then, the upper and lower boundaries of the residual magnetism of each core column before reclosing are calculated. This calculation is achieved through the following formula: ; in, and These are the upper boundary and lower boundary of the remanent magnetization, respectively. The long-term magnetic flux retention ratio of the iron core was determined through a no-load tripping test of the same transformer. This ratio reflects the ability of the iron core material to retain magnetic flux when there is no external power source excitation. The power outage time from opening to reclosing; The residual magnetism decay time constant is determined by combining the ratio of the equivalent inductance to the equivalent resistance of the remaining circuit of the winding with the no-load tripping test. The decay pattern index is set to 0.95 to 1.05 to accommodate the physical process of near-exponential decay while avoiding excessive alteration of the decay pattern. The boundary expansion coefficient is set to 1.05 to 1.15 to cover the fluctuation in flux estimation caused by the positioning error at the opening time and the sampling resolution. Rated magnetic flux; This refers to the absolute measurement error of the voltage. This is the rated phase voltage; This represents the absolute measurement error of the current. This is the rated phase current; To identify errors at any given time; It is the power frequency cycle.

[0071] The equivalent resistance and equivalent inductance are determined by the residual loop state table, thus obtaining a reasonable range of values ​​for the decay time constant. Upper and lower boundaries are used to cover uncertainties caused by various measurement and parameter estimation errors during the calculation process, enabling the subsequent particle swarm optimization algorithm to find the optimal remanent magnetization state within a directionally constrained search interval.

[0072] S1.4: Connect the magnetic flux calculation trajectory and residual magnetism boundary into a complete time sequence according to the order of operation before the fault, external fault occurrence, phase tripping, power outage interval, phase closing and inrush current attenuation, forming a magnetic state history sequence; extract the current response characteristics after reclosing to form the actual reclosing response sequence.

[0073] Based on the reclosing event time-stamp chain, post-reclosing current data, magnetic flux calculation trajectory, and the upper and lower boundaries of residual magnetism, residual magnetism direction, and residual magnetism attenuation degree, the magnetic flux calculation trajectory and residual magnetism boundaries are connected into a complete time sequence according to the chronological order of operation before the fault, external fault occurrence, phase tripping, power outage interval, phase closing, and inrush current attenuation, forming a magnetic state history sequence. In this sequence, the magnetic flux trajectory calculated by S1.2 is used from before the fault to the tripping time, the upper and lower boundaries of residual magnetism determined by S1.3 are used to constrain the time from the tripping time to the closing time, and the magnetic flux calculation trajectory after the closing time is used for continuation.

[0074] The current response characteristics after reclosing are extracted to form the actual reclosing response sequence. Specifically, the peak current, waveform bias, three-phase asymmetry, and attenuation duration of each phase after reclosing are extracted until the effective current value of three consecutive power frequency cycles re-enters the pre-fault steady-state fluctuation range, forming the actual reclosing response sequence. This response sequence records the characteristic quantities of the inrush current during the actual reclosing process, including the inrush current impact amplitude, waveform distortion, three-phase imbalance, and the time required for the inrush current to subside.

[0075] S2: Divide the model parameters of the digital twin into a steady-state parameter layer and a historical magnetic state layer. Arrange the particle elements according to the signed remanence, hysteresis position and remanence decay degree of each iron core column and set the search boundary. Establish fitness evaluation rules and set particle swarm execution parameters to form a particle swarm identification configuration.

[0076] S2.1: Based on the magnetic state history sequence, the model parameters of the digital twin are divided into a steady-state parameter layer and a historical magnetic state layer.

[0077] Based on the magnetic state history sequence output by S1, all model parameters of the digital twin are divided into a steady-state parameter layer and a historical magnetic state layer according to their physical properties.

[0078] The steady-state parameter layer contains electromagnetic and structural parameters reflecting normal operating characteristics. Specifically, the steady-state parameter layer includes winding resistance, winding leakage inductance, rated excitation curve, normal iron loss, normal copper loss, and core structural parameters. These parameters reflect the electromagnetic and loss characteristics of the transformer under normal steady-state operating conditions and do not change with external faults or reclosing transient processes. During residual magnetism identification, the steady-state parameter layer is set to read-only mode, and no parameters in this layer may be modified during any particle swarm iteration.

[0079] The historical magnetic state layer contains the historical remanence of each core column. Specifically, the historical magnetic state layer includes the signed remanence, remanence amplitude, hysteresis loop position, and attenuation degree during power outages for each core column. A signed remanence greater than zero indicates positive remanence, and a signed remanence less than zero indicates negative remanence; the absolute value of the signed remanence is the remanence amplitude. The hysteresis loop position records whether the opening flux is located on the rising or falling branch of the core hysteresis loop, which determines the initial magnetization state of the core when closing. The attenuation degree during power outages records the proportion of core flux that has dissipated between the opening and closing times. The historical magnetic state layer is set to an updatable state, and the parameter values ​​in this layer can be modified in each iteration of the particle swarm optimization algorithm.

[0080] When the digital twin performs reclosing calculations, it first reads the initial magnetic flux and hysteresis position from the historical magnetic state layer, and then calls the winding resistance, winding leakage inductance, and rated excitation curve from the steady-state parameter layer to calculate the excitation current, core saturation, and three-phase winding current. After the inrush current has decayed and the residual magnetism has subsided after reclosing, it only exits the transient function of the historical magnetic state layer, allowing the digital twin to return to its normal operation mode that relies solely on the steady-state parameter layer, without modifying any parameters in the steady-state parameter layer. Through the above-mentioned division of independent read and write permissions, the path of residual magnetism transients transferring to winding losses, rated excitation characteristics, and normal iron loss parameters is blocked.

[0081] S2.2: Based on the historical magnetic state layer, and using the magnetic flux calculation trajectory and the upper and lower boundaries of remanence, arrange the signed remanence, hysteresis position and remanence decay degree of each core column as particle elements according to the core column number, and set the search boundary of each particle element.

[0082] Based on the historical magnetic state layer output by S2.1, and the historical magnetic state sequence and initial remanence boundary output by S1, the signed remanence, hysteresis position, and remanence decay degree of each core column are arranged as particle elements according to the core column number, and the search boundary of each particle element is set. Specifically, the particle elements are arranged according to the core column number. Each core column has three continuous elements: signed remanence, hysteresis position, and remanence decay degree. Therefore, for a three-core-column transformer, each particle contains nine particle elements. The signed remanence simultaneously expresses the remanence direction and magnitude, avoiding abrupt changes in particle position during iteration caused by setting the direction as a discrete element. The hysteresis position indicates whether the tripping flux is located on the rising or falling branch of the core hysteresis loop, and the branch position is represented by a continuous value. The remanence decay degree expresses the proportion of flux that has dissipated during the power outage, with a value ranging from 0 to 1.

[0083] The lower and upper boundaries of remanent magnetization formed in S1.3 are directly set as the lower and upper boundaries of the search for signed remanent magnetization, ensuring that the particle swarm optimization algorithm searches within a physically reasonable flux range and does not search outside the zero flux boundary opposite to the direction of the tripping flux. The hysteresis branch traversed by the flux calculation trajectory in S1.2 is set as the search interval for the hysteresis position. The degree of remanent magnetization attenuation formed in S1.3 is set as the search center for the attenuation element, and the search interval for the attenuation degree is formed by fluctuating around this center value by 5%-8%. This fluctuation range is used to cover the changes in attenuation degree caused by temperature changes, measurement noise, and loop parameter uncertainties during demagnetization, while limiting particles from entering excessively attenuated regions or completely non-attenuated regions that are inconsistent with short-term power outage scenarios.

[0084] S2.3: Establish particle fitness evaluation rules. The rules calculate fitness based on the deviation between the current response of the reclosing simulation and the actual reclosing response sequence, and impose penalties on particles that exceed the search boundary.

[0085] A particle fitness evaluation rule is established. This rule calculates the fitness based on the deviation between the current response in the reclosing simulation and the measured response, and penalizes particles that exceed the search boundary. Based on the two-layer parameter structure output by S2.1, the remanent magnetization particle encoding rule and particle search boundary output by S2.2, and the actual reclosing response sequence output by S1, each particle is sequentially decoded to the historical magnetic state layer. That is, according to the particle encoding rule, the values ​​of each element of the particle are written into the signed remanence, hysteresis position, and remanence decay degree of the corresponding iron core column, while keeping the steady-state parameter layer completely unchanged. Reclosing simulation is performed on the digital twin driven by the actual closing time and the actual closing phase angle. The peak current, waveform bias, three-phase asymmetry, decay duration, and steady-state recovery deviation are extracted from the simulation results, and the fitness of each particle is calculated. This calculation is achieved through the following formula: ; in, is the dimensionless fitness of the q-th particle. The smaller the value, the more the residual magnetism of the group matches the actual reclosing process. and These are the simulated peak current and the actual peak current, respectively. and These are the simulated waveform offset and the actual waveform offset, respectively. and These are simulated three-phase asymmetry quantities and actual three-phase asymmetry quantities, respectively. and These represent the simulated decay duration and the actual decay duration, respectively. and These are the simulated steady-state recovery deviation and the actual steady-state recovery deviation, respectively. This is the rated phase current; The power frequency cycle; , , , and The weights are, in order: peak current deviation weight, waveform bias deviation weight, three-phase asymmetry deviation weight, attenuation duration deviation weight, and steady-state recovery deviation weight. All five weights are dimensionless.

[0086] Current peak deviation weight Set to 0.23 to 0.27 to control the fitting accuracy of the maximum inrush current during reclosing. Waveform offset deviation weight. Three-phase asymmetry deviation weight Weight of decay duration All values ​​are set to 0.18 to 0.22, respectively used to control the fitting accuracy of the DC bias of the current waveform, the fitting accuracy of the imbalance between the three phases, and the fitting accuracy of the inrush current from its appearance to its disappearance. Steady-state recovery deviation weight. The values ​​are set to 0.13 to 0.17 to control whether the digital twin can recover to a steady-state operating point consistent with the actual measurement after the inrush current subsides. The sum of the above five weights after normalization is 1. The above formula simultaneously constrains the initial impact characteristics of reclosing, the three-phase difference characteristics, the inrush current decay characteristics, and the steady-state recovery characteristics, avoiding erroneous identification results that do not match the actual residual magnetism state obtained by only fitting the current peak value.

[0087] S2.4: Set the execution parameters of the particle swarm algorithm, including the number of particles, inertia weight, learning coefficient and maximum number of iterations, to form the particle swarm identification configuration.

[0088] The execution parameters of the particle swarm optimization (PSO) algorithm are set, including the number of particles, inertia weight, learning coefficient, and maximum number of iterations, forming the PSO identification configuration. Specifically, the execution parameters of the PSO algorithm are set based on the two-layer parameter structure output by S2.1, the remanent magnetization particle encoding rules and particle search boundaries output by S2.2, and the particle fitness evaluation rules and fitness weights output by S2.3. The number of particles is set to 40 to 50 to cover the multi-dimensional search space composed of the remanence, hysteresis position, and attenuation degree of each core column. When the number of particles is less than 40, it is easy to miss some combinations of joint remanence states of different core columns. When the number of particles is greater than 50, it will significantly increase the number of calculations in the digital twin regenerative braking simulation. The initial inertia weight is set to 0.85 to 0.90, so that the algorithm has a strong global exploration capability in the early stage of the search; the final inertia weight is set to 0.40 to 0.45, so that the algorithm gradually converges to the local optimum in the later stage of the search. Both the individual learning coefficient and the swarm learning coefficient are set to 1.7 to 1.9, ensuring that the particle's historical best position and the swarm's global best position have similar guiding strength, preventing one direction from excessively dominating the search process. The maximum number of iterations is set to 120 to 150, ensuring that the algorithm has enough iterations to complete the entire process from large-scale exploration to local convergence.

[0089] The aforementioned steady-state parameter layer, historical magnetic state layer, remanent particle encoding rules, particle search boundary, fitness evaluation rules, and particle swarm execution parameters together constitute the complete configuration required for S3 to perform remanent state identification.

[0090] S3: Generate an initial particle swarm, calculate the fitness of each particle according to the fitness evaluation rules, iteratively update the particle positions until the termination condition is met, and output the optimal combination of remanent magnetization states and the remanent magnetization decay trajectory.

[0091] S3.1: Generate an initial particle swarm according to the arrangement order of the iron core columns, so that each particle element is evenly distributed in its respective search interval, record the steady-state parameters to lock the snapshot, and form a reclosing event driving sequence.

[0092] Based on the steady-state parameter layer, historical magnetic state layer, remanent magnetization particle encoding rules, particle search boundary, fitness evaluation rules, and particle swarm execution parameters output by S2, and the historical magnetic state sequence and actual reclosing response sequence output by S1, three particle elements—signed remanent magnetization, hysteresis position, and remanent magnetization attenuation degree—are sequentially set for each core column according to the core column arrangement order determined by S2. For a three-core-column transformer, each particle contains nine particle elements. The unit for the signed remanent magnetization element is Weber, while the hysteresis position element and the remanent magnetization attenuation degree element are dimensionless.

[0093] The search interval for each particle element is divided into sub-intervals equal to the number of particles. For each particle element, a sub-interval is randomly selected without replacement from its corresponding sub-intervals, and an initial value is randomly generated within that sub-interval. The initial values ​​selected for each particle element are combined according to the arrangement order of the iron core columns to form a complete particle. The selection order of different particle elements and the combination order between particles are rearranged to ensure that the initial particle swarm is evenly distributed throughout the search space. The number of particles is set to 40 to 50. When the number of particles is less than 40, it is easy to miss the combination of joint remanent magnetization states between different iron core columns. When the number of particles is greater than 50, it will significantly increase the number of reclosing simulations in the digital twin. Based on the magnetic flux direction at the moment of opening determined in S1.2, the search interval for signed remanent magnetization is intercepted. The lower and upper boundaries of remanent magnetization formed in S1.3 and set in S2.2 are directly used as the lower and upper boundaries of the search to ensure that the signed remanent magnetization elements of the initial particles do not exceed the physical boundaries determined in S1.3.

[0094] After generating the initial particle swarm, the specific values ​​of winding resistance, winding leakage inductance, rated excitation curve, normal iron loss, and normal copper loss are recorded to form a steady-state parameter lock snapshot. This snapshot is used to restore the steady-state parameter layer to its initial state before each subsequent particle fitness evaluation, ensuring that the fitness differences between different particles originate entirely from differences in the historical magnetic state layer. A reclosing event driving sequence is formed according to the time sequence of operation before the fault, phase tripping, power outage interval, phase closing, and post-closing decay. This sequence drives the digital twin to perform corresponding state switching operations at each event moment.

[0095] S3.2: Read each particle one by one, write the particle position into the historical magnetic state layer and restore the steady-state parameter layer to the steady-state parameter locking snapshot, execute the simulation according to the reclosing event driven sequence, calculate the fitness of each particle according to the fitness evaluation rule, register the initial position of each particle as its historical best position, and register the position of the particle with the smallest fitness as the initial global best position.

[0096] Based on the initial particle swarm, steady-state parameter locking snapshot, and reclosing event-driven sequence output by S3.1, and the fitness evaluation rules output by S2, each particle in the initial particle swarm is read one by one. The signed remanence, hysteresis position, and remanence decay degree of the current particle are written into the historical magnetic state layer, and the steady-state parameter layer is restored to the values ​​in the steady-state parameter locking snapshot to ensure that the steady-state parameters of the digital twin are exactly the same before each simulation. The digital twin is advanced according to the reclosing event-driven sequence, and the closing operation of each phase is performed under the actual closing phase angle. The peak current, waveform bias, three-phase asymmetry, decay duration, and steady-state recovery deviation are extracted from the simulation results. The fitness value of the particle is calculated according to the fitness evaluation rules formed by S2.3.

[0097] The initial position of each particle is registered as its historical best position, and the position of the particle with the lowest fitness value is registered as the initial global best position of the particle swarm. Through the above processing, an evaluation relationship between the first round of residual magnetism and the reclosing transient response is established under the condition that the steady-state parameters are completely locked, preventing subsequent particle search processes from reducing transient response deviation by changing normal loss parameters or excitation parameters.

[0098] S3.3: Iteratively update particle velocity and position. After each update, re-execute the simulation and calculate the fitness of each particle. Based on this, update the historical best position of each particle and the global best position of the particle swarm until the termination condition is met.

[0099] Based on the initial particle fitness table, the initial historical optimal position of each particle, and the initial global optimal position of the particle swarm output by S3.2, the initial particle swarm output by S3.1, the steady-state parameter locking snapshot, and the reclosing event-driven sequence, as well as the particle search boundary and particle swarm execution parameters output by S2, the particle swarm is iteratively updated.

[0100] For each particle element, the particle velocity is iteratively updated according to a velocity update method jointly guided by inertia weight, individual learning coefficient, and group learning coefficient. Specifically, for the j-th particle element of the q-th particle, the particle velocity and particle position for the next iteration are calculated according to the following velocity update formula: ; in, and These are the particle element velocities in the r-th and (r+1)-th iterations, respectively; and These represent the particle element positions in the two iterations, respectively. The historical best position element of the q-th particle; The element at the global optimal position of the particle swarm and These represent the lower and upper bounds of the search for the j-th particle element, respectively; r is the current iteration number. This represents the maximum number of iterations. The initial inertia weight is set to 0.85 to 0.90; The final inertia weight is set to 0.40 to 0.45; The inertial weight decay morphology coefficient is set to 2.8 to 3.2, so that the inertial weight decreases rapidly in the early stage of the iteration and smoothly approaches the final value in the later stage of the iteration. The individual learning coefficient is set to 1.7 to 1.9; The group learning coefficient is set to 1.7 to 1.9. and A random number generated independently between 0 and 1.

[0101] The inertia weight term in the above velocity update formula adopts an exponential decay form, so that the search process gradually shifts from large-scale exploration to local convergence.

[0102] The particle positions are iteratively updated according to the position update method, and the updated positions are clipped to within a preset search boundary. Specifically, the above position update formula clips the updated particle positions to within the predetermined search boundary, ensuring that signed remanent magnetic elements do not exceed the physically reasonable magnetic flux range, and that hysteresis position elements and remanent decay degree elements do not exceed their respective reasonable ranges.

[0103] The updated velocity is constrained to prevent particles from crossing the search boundary. Specifically, after each velocity update, the absolute value of the velocity is constrained, limiting it to within 18% to 22% of the width of the particle element's search interval. This constraint is used to prevent signed remanence from crossing the physical boundary due to excessively large step sizes during iterative updates, thus avoiding invalid solutions that violate the actual magnetization characteristics of the iron core.

[0104] After each round of position update, the updated particle is rewritten into the historical magnetic state layer, the steady-state parameter layer is restored to the steady-state parameter locked snapshot, reclosing simulation is performed, and the fitness of the particle is recalculated.

[0105] The historical best position of each particle and the global best position of the particle swarm are updated based on the fitness of each particle. Specifically, when the current fitness of a particle is less than its historical best fitness, the historical best position of the particle is replaced by its current position; when the minimum historical best fitness of all particles is less than the current global best fitness, the global best position of the particle swarm is replaced by the historical best position of the particle with the minimum fitness.

[0106] The termination of the iteration is determined by a convergence index, which reflects both the improvement in the global optimal fitness and the dispersion of all particles around the global optimal position. Specifically, if the current iteration number r is less than the lookback length m, the convergence index determination is skipped, and the process proceeds directly to the next iteration; if r ≥ m, the following convergence index determination steps are performed: After each iteration, the particle swarm is used to determine whether the termination condition is met using the following convergence index calculation formula: ; in, It is a dimensionless convergence index; Let be the globally optimal fitness in the r-th iteration; is the global optimal fitness in the rm-th iteration; m is the fitness lookback length, set to 8 to 10 iterations; Q is the number of particles; J is the number of elements per particle; The fitness improvement weight is set to 0.45 to 0.55, so that the degree of fitness improvement has a similar determining effect to the degree of particle aggregation; the remaining variables are defined as before. The first term of this convergence index formula represents the relative improvement of the global optimal fitness, and the second term represents the normalized dispersion of all particles around the global optimal position. The formula as a whole is dimensionless.

[0107] The convergence metric threshold is set to 0.012 to 0.015, and the transient fitness threshold is set to 0.0025 to 0.0030. The former confirms that the particle swarm has stably clustered near the global optimum, while the latter ensures that the deviations of the five main normalized responses are controlled within approximately 5%. When both the convergence metric and transient fitness simultaneously meet their respective threshold requirements, the termination type is registered as "deviation satisfied." When the number of iterations reaches the upper limit of 120 to 150 but has not yet simultaneously met both thresholds, the termination type is registered as "iteration upper limit terminated." Both termination types, along with the global optimum, global optimum fitness, simulation response corresponding to the optimal particle, and records of previous iterations, are output to S3.4.

[0108] S3.4: After terminating the iteration, output the optimal combination of remanent magnetization states and the remanent magnetization decay trajectory.

[0109] Based on the global optimal position of the terminating particle swarm, the global optimal fitness of the terminating particle swarm, the simulation response corresponding to the terminating particle, the termination type, and the particle convergence record output by S3.3, and the reclosing event-driven sequence output by S3.1, the global optimal position of the terminating particle swarm is analyzed according to the remanent magnetization particle encoding rule of S2. The remanent magnetization direction, remanent magnetization amplitude, hysteresis position, and remanent magnetization attenuation degree of each core column are extracted one by one. The remanent magnetization direction is determined by the sign of the signed remanent magnetization, and the remanent magnetization amplitude is determined by the absolute value of the signed remanent magnetization.

[0110] The residual magnetism of each core column is sequentially applied to the magnetic state time nodes corresponding to the opening moment, power outage interval, and actual closing moment, generating a residual magnetism decay trajectory along the reclosing event-driven sequence. This decay trajectory records the entire process of change from the initial magnetic flux at the moment of opening, through gradual decay during the power outage interval, to the residual magnetic flux at the moment before closing. The simulation response corresponding to the terminating particle, the global optimal fitness of the terminating particle, the termination type, the global optimal fitness change sequence in each iteration, and the particle discreteness change sequence are all written into the particle convergence record. The above optimal residual magnetism state combination, residual magnetism decay trajectory, and particle convergence record serve as inputs for S4 to verify the transient and steady-state continuity of the residual magnetism identification results.

[0111] S4: Load the optimal residual magnetism state combination into the historical magnetic state layer and perform reclosing simulation to verify the transient response characteristics and steady-state recovery characteristics respectively. When both transient and steady-state verifications are qualified, encapsulate the residual magnetism dynamic update package.

[0112] S4.1: Load the historical magnetic state layer from the optimal residual magnetism state to the opening time, lock the steady-state parameter layer to read-only state, advance the core magnetic state change within the power outage interval according to the residual magnetism decay trajectory, execute the closing operation at the actual closing time to generate the reclosing simulation state sequence, and divide the transient verification interval and steady-state verification interval with the end time of inrush current decay as the boundary.

[0113] Figure 2 The flowchart illustrating the transient and steady-state dual verification and update package encapsulation provided in this disclosure embodiment is shown, as follows: Figure 2 As shown, based on the optimal remanent magnetization state combination, remanent magnetization decay trajectory, and particle convergence record output by S3, the magnetic state history sequence and reclosing actual response sequence output by S1, and the steady-state parameter layer and historical magnetic state layer output by S2, the remanent magnetization direction, remanent magnetization amplitude, and hysteresis position of each core column in the optimal remanent magnetization state combination are loaded into the historical magnetic state layer corresponding to the opening time of each phase circuit breaker. At the same time, the steady-state parameter layer is locked in a read-only state to ensure that the steady-state parameters do not change during the replay process. The change of core magnetic state within the power outage interval is advanced according to the remanent magnetization decay trajectory, so that the signed remanent magnetization of each core column gradually decreases with the power outage time according to the exponential decay law, but the remanent magnetization direction remains consistent with the magnetic flux direction at the opening time.

[0114] At the actual closing time and actual closing phase angle, the closing operations of each phase circuit breaker are executed sequentially. The magnetic flux change, excitation current, and three-phase winding current of each core column are continuously calculated to generate a reclosing simulation state sequence. Taking the actual closing time as the starting point and the actual inrush current decay end time as the ending point, this interval is set as the transient verification interval, and the end time of this interval is denoted as . The steady-state verification interval is set as the period from the end of the actual inrush current decay to the beginning of the interval. This 8 to 10 power frequency cycles cover the periodic steady-state recovery process after the excitation inrush current exits, while avoiding interference from subsequent load adjustments or system operations on the verification results due to an excessively long verification interval.

[0115] S4.2: Extract the transient response features of the reclosing simulation state sequence within the transient verification interval, compare them with the actual reclosing response sequence, calculate the transient continuity index, and form a transient verification qualified conclusion when the threshold is met.

[0116] Based on the reclosing simulation state sequence and transient verification interval output by S4.1, and the actual reclosing response sequence output by S1, the peak current, waveform offset, three-phase asymmetry, and attenuation duration are extracted from the simulation results and measured data, respectively, and these characteristic quantities are compared. Simultaneously, the complete waveforms of the three-phase currents within the transient verification interval are compared to calculate the transient continuity index. This calculation is achieved using the following formula: ; in, It is a dimensionless transient continuity index; and These are the simulated peak current and the actual peak current, respectively. and These are the simulated waveform offset and the actual waveform offset, respectively. and These are simulated three-phase asymmetry quantities and actual three-phase asymmetry quantities, respectively. and These represent the simulated decay duration and the actual decay duration, respectively. is the rated phase current; h is the phase sequence number, with values ​​1, 2, and 3 representing phases A, B, and C respectively; and These are the simulated instantaneous current and the actual instantaneous current of phase h, respectively. This refers to the actual closing time; The time interval for transient verification ends; t is time. , , , and These are five transient weights. The first four terms in the formula are normalized by the rated current or the actual decay duration, and the fifth term is normalized by the square integral of the actual current. Therefore, the overall dimension of the formula is one.

[0117] The five transient weights mentioned above are all set to 0.18 to 0.22 and normalized to a sum of 1, so that the discrete characteristic deviation and the complete waveform deviation have similar verification effects. The transient continuity threshold is set to 0.045 to 0.055. When the transient continuity index is not greater than the transient continuity threshold, a transient verification qualified conclusion is formed, indicating that the identified residual magnetism state can accurately reproduce the excitation inrush current impact amplitude, waveform distortion degree, three-phase imbalance characteristics, and inrush current decay rate during the reclosing transient process. When the transient continuity index is greater than the transient continuity threshold, a transient verification unqualified conclusion is formed, and the transient characteristic residual is output for subsequent correction.

[0118] S4.3: After the transient verification is qualified, the reclosing simulation state sequence is verified from four dimensions within the steady-state verification interval. The steady-state recovery index is calculated. When the threshold is met and the steady-state parameters are locked, the steady-state verification is qualified. The unidirectional magnetic flux bias attenuation time constant and steady-state parameter locking snapshot after closing are determined.

[0119] Steady-state recovery continuity verification is performed only when S4.2 outputs a transient verification pass conclusion. Based on the reclosing simulation state sequence, steady-state parameter locking snapshot, and steady-state verification interval output by S4.1, the effective values ​​of the simulated current and the actual current are extracted according to the power frequency cycle within the steady-state verification interval. It is checked whether the core flux at the same phase point in adjacent power frequency cycles has recovered to a periodically consistent change pattern, whether the flux bias added to the historical magnetic state layer has been removed, and whether the steady-state scalar parameters (including winding resistance, winding leakage inductance, normal iron loss, and normal copper loss; rated excitation curve and core structure parameters are only checked for locked values ​​and are not included in numerical comparison) remain unchanged before and after verification. The steady-state recovery index is calculated. This calculation is achieved through the following formula: ; in, is a dimensionless steady-state recovery index; M is the number of power frequency cycles contained in the steady-state verification interval; n is the power frequency cycle number; and These are the simulated effective current value and the actual effective current value for the h-th phase n-th power frequency cycle, respectively; The simulated magnetic flux of phase h in the reference phase of the nth power frequency cycle; The power frequency cycle; The rated magnetic flux of phase h; The magnetic flux bias applied to the h-th phase in the nth power frequency cycle is the historical magnetic state layer; K is the number of steady-state scalar parameters involved in the locking (i.e., the total number of winding resistance, winding leakage inductance, rated excitation curve, normal iron loss, and normal copper loss). For the first time after the verification is completed One steady-state scalar parameter; Lock the first snapshot of steady-state parameters One steady-state scalar parameter; , , and These are the current recovery weight, magnetic flux period consistency weight, historical state exit weight, and steady-state parameter invariance weight, respectively. Each term in the formula is composed of the ratio of parameters with the same dimension, and the overall dimension is one.

[0120] Current recovery weight Set to 0.28 to 0.32 to control the accuracy of the current RMS value recovering to the measured level after the inrush current subsides. Flux period consistency weight. and historical state exit weight All values ​​are set to 0.23 to 0.27, respectively, to control the degree to which the core flux recovers to a normal periodic variation and the accuracy of confirming that the additional bias of the historical magnetic state layer has completely exited and no longer affects the flux calculation. Steady-state parameter invariant weights. The values ​​are set to 0.18 to 0.22 to ensure that the steady-state parameters after verification are exactly the same as before verification. The weights of the above four items are normalized to a sum of 1. The steady-state recovery threshold is set to 0.025 to 0.030.

[0121] When the steady-state recovery index is not greater than the steady-state recovery threshold, and all steady-state parameters remain locked at their values, a steady-state verification is deemed successful. This indicates that the identified residual magnetism can fully restore the digital twin to its normal steady-state operating mode after the transient process ends. Simultaneously, the unidirectional flux bias decay time constant and steady-state parameter lock snapshots are recorded after the circuit breaker is closed. When the steady-state recovery index exceeds the steady-state recovery threshold or any steady-state parameter is changed, a steady-state verification is deemed unsuccessful. The steady-state recovery index is then output for subsequent correction.

[0122] S4.4: When both transient and steady-state verifications are qualified, the optimal combination of residual magnetism states, residual magnetism attenuation trajectory, unidirectional magnetic flux bias attenuation time constant after closing, and steady-state parameter lock snapshot are encapsulated into a residual magnetism dynamic update package; when any verification fails, the corresponding boundary is corrected and S3 is returned for re-identification.

[0123] Based on the transient continuity index, transient characteristic residual and transient verification conclusion output by S4.2, the steady-state recovery index, steady-state parameter invariance record and steady-state verification conclusion output by S4.3, and the optimal remanent magnetization state combination, remanent magnetization decay trajectory and particle convergence record output by S3, they are processed separately according to the different verification conclusions.

[0124] When both transient and steady-state verifications are successful, the optimal residual magnetism state combination, residual magnetism attenuation trajectory, actual closing phase angle, transient verification interval, steady-state verification interval, transient continuity index, steady-state recovery index, steady-state parameter locking snapshot, the unidirectional flux bias attenuation time constant after closing determined in S4.3, and the steady-state parameter locking snapshot are encapsulated into a residual magnetism dynamic update package and output to S5. This update package contains the complete residual magnetism state information that has been verified and is successful, as well as all timing parameters and attenuation parameters required for dynamic updating.

[0125] When transient verification fails, the residual magnetization direction boundary is corrected based on the polarity of the three-phase waveform bias, the residual magnetization amplitude boundary is corrected based on the current peak residual, and the residual magnetization boundary of each core column is corrected based on the three-phase asymmetry residual. Specifically, if the polarity of the simulated waveform bias of a certain phase is higher than the measured value, it indicates that the estimated residual magnetization direction of that phase is too large, and the search boundary of the residual magnetization direction of the corresponding core column is adjusted in the opposite direction; if the current peak residual is positive, it indicates that the estimated residual magnetization amplitude is too large, and the upper boundary of the amplitude search is lowered; if the three-phase asymmetry residual reflects that the inrush current of a certain phase is too large, the search boundary of the core column where that phase is located is corrected separately.

[0126] When the steady-state verification fails, the remanent magnetization attenuation range is corrected based on the residual attenuation duration and the additional magnetic flux bias of the historical magnetic state layer. If the simulated attenuation duration has a positive residual, it indicates that the identified remanent magnetization attenuation rate is too slow, and the attenuation degree search interval is shifted towards the direction of faster attenuation; if it has a negative residual, it is shifted towards the direction of slower attenuation.

[0127] The single boundary correction amount is limited to 8% to 10% of the initial search interval width formed in S1.3 to avoid losing the physical boundary constraints formed in step S1 due to excessive movement of the search interval at one time. The global maximum correction rounds are set to 3 to 5, and a global cumulative correction round counter is established. This counter is initialized to 0 when S3 is executed for the first time, and incremented by 1 each time S3 is re-executed after returning to the current sub-step and completing boundary correction. In subsequent steps, if boundary correction needs to be performed again in the current sub-step, this global cumulative correction round counter should be used continuously without repeated initialization. After each correction round, S3 is returned to re-execute particle swarm identification. If both transient and steady-state verifications are successful within the global maximum correction rounds, a remanent magnetization dynamic update package is generated and output to S5; if it is still unsuccessful after reaching the global maximum correction rounds, the remanent magnetization state combination corresponding to the minimum global optimal fitness in the most recent particle swarm identification is used as the suboptimal result, and an update package with a finite correction mark is output, triggering a manual review prompt.

[0128] S5: Bind the residual magnetism dynamic update package to the historical magnetic state layer, and advance the residual magnetism decay according to the residual magnetism decay trajectory during the power outage interval. After closing the circuit, the unidirectional magnetic flux bias component is gradually withdrawn through nonlinear excitation calculation. When the update consistency meets the requirements, the update result is solidified.

[0129] S5.1: Read data from the residual magnetism dynamic update package and bind it to the historical magnetic state layer.

[0130] Based on the residual magnetism dynamic update package output by S4, the optimal residual magnetism state combination, residual magnetism attenuation trajectory, actual closing phase angle, transient verification interval, steady-state verification interval, unidirectional magnetic flux bias attenuation time constant after closing, and steady-state parameter locking snapshot are read from the update package.

[0131] Based on the equipment identification, winding number, and core column number, the residual magnetism direction, residual magnetism amplitude, hysteresis position, and attenuation degree of each core column are bound to the corresponding storage location in the historical magnetic state layer of the digital twin. The winding resistance, winding leakage inductance, rated excitation curve, normal iron loss, and normal copper loss are checked item by item against the values ​​in the steady-state parameter locking snapshot. Once complete consistency is confirmed, a steady-state parameter locking snapshot is formed to ensure that the steady-state parameter layer always uses this locking snapshot as the standard during dynamic updates. A residual magnetism state writing sequence is formed according to the phase opening time, power outage interval duration, and phase closing time. This sequence specifies the writing order and writing nodes of the residual magnetism state of each core column on the time axis.

[0132] S5.2: At the moment of each phase trip, the signed residual magnetism and hysteresis position in the optimal residual magnetism state combination are written into the historical magnetic state layer of the corresponding iron core column. During the power outage interval, the residual magnetism amplitude of each iron core column is advanced hourly according to the residual magnetism decay trajectory, while keeping the residual magnetism direction unchanged.

[0133] Based on the device version binding record, steady-state parameter locking snapshot, and residual magnetism state writing timing output by S5.1, when the tripping time of each phase arrives, the signed residual magnetism and hysteresis position in the optimal residual magnetism state combination are written into the historical magnetic state layer of the corresponding iron core column.

[0134] During the power outage interval, the residual magnetism amplitude of each core column is progressively increased hourly according to the residual magnetism decay trajectory in the residual magnetism dynamic update package. Specifically, the residual magnetism amplitude at each moment is obtained by multiplying the magnetic flux at the moment of opening by an exponential decay factor that decreases over time, while maintaining the residual magnetism direction perfectly aligned with the magnetic flux direction at the moment of opening, ensuring that the directional attribute does not change due to amplitude decay. Upon reaching the closing time of each phase, the residual magnetism state of each core column at the moment before closing is extracted. The signed residual magnetism is used as the initial condition for the core flux, and the hysteresis position is used as the starting point of the core excitation curve. Throughout the entire process, the steady-state parameter layer remains in a locked snapshot state, unaffected by any update operations.

[0135] Through the above processing, the magnetic state at the moment of opening is continuously transmitted to the moment of reclosing through the complete power outage interval, avoiding the loss of residual magnetic history information by directly resetting the core magnetic flux to zero after power failure.

[0136] S5.3: At the closing time of each phase, the closing operation is performed. The residual magnetic state at the moment before closing is used as the initial condition to start the nonlinear excitation calculation, so that the core magnetic flux after closing is composed of steady-state periodic magnetic flux component and unidirectional magnetic flux bias component. The unidirectional magnetic flux bias component gradually exits as the excitation inrush current decays, generating the magnetic state evolution process.

[0137] Based on the residual magnetism state before reclosing, the initial magnetic flux of each phase during reclosing, and the historical magnetic state layer during power outages output by S5.2, as well as the steady-state parameter locking snapshot output by S5.1, the nonlinear excitation calculation of the digital twin is initiated at the moment of reclosing of each phase, using the initial magnetic flux of each phase as the starting condition. After reclosing, the core magnetic flux consists of two components: one part is the periodic magnetic flux component calculated by the steady-state parameter layer under the actual reclosing voltage, and the other part is the unidirectional magnetic flux bias component provided by the historical magnetic state layer. This unidirectional magnetic flux bias gradually withdraws as the winding resistance dissipates and the inrush current decays, so that the core magnetic flux at the initial stage of reclosing simultaneously reflects the initial bias effect of residual magnetism and the subsequent gradual dissipation of the physical process.

[0138] The magnetic flux of each core column is connected according to the following segmentation formula: ; in, Let be the magnetic flux of the k-th core column at time t; The signified magnetic flux at the moment of circuit breaker opening is taken from the final value of the magnetic flux at the moment of circuit breaker opening after S3 identification and determination. To maintain the magnetic flux ratio over a long period; This is the time of circuit breaker tripping; This is the closing time; This is the time constant for residual magnetism decay during a power outage; The attenuation morphology index during power outages is set to 0.95 to 1.05; The residual magnetism at the moment before closing the circuit breaker; The periodic magnetic flux is calculated for the steady-state parameter layer under the actual closing voltage; The inrush current attenuation process after the unidirectional magnetic flux bias decays after closing is determined by the inrush current attenuation process verified by S4. The decay characteristic index after closing is set to 0.95 to 1.05 to maintain a first-order decay characteristic close to that formed by the winding resistance and the excitation circuit. The second segment occurs at the closing moment... The calculated magnetic flux value is equal to the residual magnetic value of the first segment at the moment before closing. This ensures the continuity of the magnetic flux state at the moment of reclosing. The formula treats residual magnetism as an additional magnetic flux bias at the initial moment of closing and allows this bias to gradually decrease over time. Through the above calculations, a magnetic flux evolution sequence after reclosing, a record of historical magnetic state exit, and a record of steady-state parameter retention are generated. It should be noted that in extreme scenarios of deep saturation, the magnetic flux of each core column needs to be adjusted, but this invention does not consider this situation.

[0139] S5.4: After the closing operation is completed, the consistency index is updated from four dimensions: current recovery, magnetic flux cycle consistency, historical state exit and steady-state parameter unchanged within several consecutive power frequency cycles. When the threshold is met and the steady-state parameter is locked, the current residual magnetism state, magnetic state evolution process and steady-state parameter are solidified to form the dynamic update result of the digital twin.

[0140] Based on the flux evolution sequence after reclosing, historical magnetic state exit records, and steady-state parameter retention records output by S5.3, the steady-state parameter locking snapshot output by S5.1, and the actual reclosing response sequence output by S1, the updated consistency index is calculated over 5 to 6 consecutive power frequency cycles after reclosing. This consistency check uses the actual reclosing response sequence output by S1.4 as the measured comparison benchmark and follows the same index system as the steady-state recovery verification in S4.3. It performs a weighted fusion judgment based on four dimensions: current effective value deviation (comparison of simulated and measured values), flux consistency deviation between adjacent power frequency cycles, residual amount of additional flux bias from the historical magnetic state layer, and relative changes in each steady-state parameter before and after the update. The weight settings are consistent with S4.3: current recovery weight is set to 0.28 to 0.32, flux cycle consistency weight and historical state exit weight are both set to 0.23 to 0.27, and steady-state parameter invariance weight is set to 0.18 to 0.22. The sum of the four weights after normalization is 1. The update consistency threshold is set to 0.025 to 0.030.

[0141] When the update consistency index is not greater than the update consistency threshold, and all steady-state parameters remain locked and are completely consistent with the steady-state parameter locking snapshot, the current residual magnetism, magnetic state evolution process, and steady-state parameters are solidified to form the dynamic update result of the digital twin of the power grid equipment. This update result records the complete magnetic state history of the transformer in this reclosing event, including the initial residual magnetism at the moment of opening, the attenuation process during the power outage interval, the initial magnetic flux bias at the moment of closing, and the entire process of the magnetic flux bias gradually fading to zero after closing. All records are obtained based on a data-driven approach rather than relying on empirical assumptions.

[0142] When the update consistency index exceeds the update consistency threshold, the residuals are used to form a remanent magnetization dynamic update application residual and returned to S4. S4.4 then corrects the remanent magnetization dynamic update package based on the specific source of the residuals and executes all sub-steps of S5 again until the update consistency index meets the solidification condition and the final digital twin dynamic update result is formed. Upon returning to S4.4, the global maximum correction round and global cumulative correction round counter set in that sub-step should be used. If the global cumulative correction round has reached the global maximum correction round, no further correction is performed; instead, the current update package with a limited correction flag is output, and a manual review prompt is triggered.

[0143] This disclosure also provides a data-driven dynamic update system for digital twins of power grid equipment. Figure 3 This is a schematic diagram of a data-driven digital twin dynamic update system for power grid equipment, provided according to an embodiment of this disclosure. This system is used to run the methods described in the above embodiments. (Refer to...) Figure 3 The system may include:

[0144] The magnetic state evolution module is used to establish a reclosing event time-stamped chain, obtain the estimated magnetic flux value, magnetic flux direction and magnetic flux calculation trajectory of each iron core column at the moment of opening, delineate the residual magnetic boundary of each iron core column before reclosing, and form a magnetic state history sequence and a reclosing actual response sequence.

[0145] The updated structure building module is used to divide the model parameters of the digital twin into a steady-state parameter layer and a historical magnetic state layer. The particle elements are arranged according to the signed remanence, hysteresis position and remanence decay degree of each iron core column and the search boundary is set. Fitness evaluation rules are established and particle swarm execution parameters are set to form a particle swarm identification configuration.

[0146] The particle swarm identification module is used to generate an initial particle swarm, calculate the fitness of each particle according to the fitness evaluation rules, iteratively update the particle positions until the termination condition is met, and output the optimal combination of remanent magnetization states and the remanent magnetization decay trajectory.

[0147] The verification module is used to load the optimal residual magnetism state combination into the historical magnetic state layer and perform reclosing simulation to verify the transient response characteristics and steady-state recovery characteristics respectively. When the verification is qualified, the residual magnetism dynamic update package is encapsulated.

[0148] The dynamic update module is used to bind the residual magnetism dynamic update package to the historical magnetic state layer. During the power outage interval, the residual magnetism decay is promoted according to the residual magnetism decay trajectory. After the circuit is closed, the unidirectional magnetic flux bias component is gradually withdrawn through nonlinear excitation calculation. When the update consistency meets the requirements, the update result is solidified.

[0149] According to embodiments of this disclosure, an electronic device is also provided, which may include a processor, a communications interface, a memory, and a communication bus, wherein the processor, the communications interface, and the memory communicate with each other via the communication bus. The processor can invoke logical instructions stored in the memory to execute the methods provided in the above embodiments.

[0150] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0151] On the other hand, this disclosure also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the methods provided in the above embodiments.

Claims

1. A data-driven method for dynamically updating digital twins of power grid equipment, characterized in that, The method includes: S1. Establish a reclosing event time-stamped chain, obtain the estimated magnetic flux value, magnetic flux direction and magnetic flux calculation trajectory of each iron core column at the moment of opening, delineate the residual magnetic boundary of each iron core column before reclosing, and form a magnetic state history sequence and a reclosing actual response sequence. S2, the model parameters of the digital twin are divided into a steady-state parameter layer and a historical magnetic state layer. The particle elements are arranged according to the signed remanence, hysteresis position and remanence decay degree of each iron core column and the search boundary is set. Fitness evaluation rules are established and particle swarm execution parameters are set to form a particle swarm identification configuration. S3, generate an initial particle swarm, calculate the fitness of each particle according to the fitness evaluation rule, iteratively update the particle position until the termination condition is met, and output the optimal combination of remanent magnetization states and the remanent magnetization decay trajectory. S4, load the optimal residual magnetism state combination into the historical magnetic state layer and perform reclosing simulation to verify the transient response characteristics and steady-state recovery characteristics respectively. When both transient and steady-state verifications are qualified, encapsulate the residual magnetism dynamic update package. S5, bind the residual magnetism dynamic update package to the historical magnetic state layer, advance the residual magnetism decay according to the residual magnetism decay trajectory during the power outage interval, and after closing the circuit, use nonlinear excitation calculation to make the unidirectional magnetic flux bias component gradually withdraw, and solidify the update result when the update consistency meets the requirements.

2. The method according to claim 1, characterized in that, S1 includes: S1.1 According to the time of external fault occurrence, the time of each phase tripping, the time of all winding de-energization, the time of each phase closing, and the time of inrush current attenuation end, the reclosing process is divided into continuous event stages, a reclosing event time stamp chain is established, and a state table of the remaining circuits of the windings during the power outage is formed. S1.2, Select the starting interval of integration before the external fault occurs, and obtain the estimated value of magnetic flux, magnetic flux direction and magnetic flux calculation trajectory of each iron core column at the moment of opening through numerical integration; S1.3, Based on the remaining circuit status table of the winding during the power outage and the estimated magnetic flux at the moment of opening, determine the degree of residual magnetic decay of each iron core column before closing according to the residual magnetic decay law during the power outage, and delineate the upper and lower boundaries of residual magnetic flux of each iron core column. S1.4, according to the order of operation before the fault, occurrence of external fault, tripping of each phase, power outage interval, closing of each phase and inrush current attenuation, connect the magnetic flux calculation trajectory and residual magnetism boundary into a complete time sequence to form a magnetic state history sequence; extract the current response characteristics after reclosing to form the actual reclosing response sequence.

3. The method according to claim 1, characterized in that, S2 includes: S2.1, Based on the magnetic state history sequence, the model parameters of the digital twin are divided into a steady-state parameter layer and a historical magnetic state layer. The steady-state parameter layer contains electromagnetic parameters and structural parameters that reflect normal operation characteristics and is locked as read-only during the remanence identification period. The historical magnetic state layer contains the signed remanence, hysteresis position and remanence decay degree of each core column and is allowed to be updated during the identification period. S2.2, Based on the historical magnetic state layer, and using the magnetic flux calculation trajectory and the upper and lower boundaries of remanence, arrange the signed remanence, hysteresis position and remanence decay degree of each core column as particle elements according to the core column number, and set the search boundary of each particle element. S2.3, Establish particle fitness evaluation rules. The rules calculate fitness based on the deviation between the current response of the reclosing simulation and the actual reclosing response sequence, and impose penalties on particles that exceed the search boundary. S2.4, based on the results of S2.1 to S2.3, sets the execution parameters of the particle swarm algorithm, including the number of particles, inertia weight, learning coefficient and maximum number of iterations, to form the particle swarm identification configuration.

4. The method according to claim 1, characterized in that, S3 includes: S3.1, Generate an initial particle swarm according to the arrangement order of the iron core columns, so that each particle element is evenly distributed in its respective search interval, record the steady-state parameters to lock the snapshot, and form a reclosing event driving sequence. S3.2, read each particle one by one, write the particle position into the historical magnetic state layer and restore the steady-state parameter layer to the steady-state parameter locking snapshot, execute the simulation according to the reclosing event driven sequence, calculate the fitness of each particle according to the fitness evaluation rule, register the initial position of each particle as its historical best position, and register the position of the particle with the smallest fitness as the initial global best position; S3.3 Iteratively update particle velocity and position. After each update, re-execute the simulation and calculate the fitness of each particle. Based on this, update the historical best position of each particle and the global best position of the particle swarm until the termination condition is met. S3.4 After terminating the iteration, output the optimal combination of remanent magnetization states and the remanent magnetization decay trajectory.

5. The method according to claim 4, characterized in that, S3.3 includes: For each particle element, the particle velocity is iteratively updated according to the velocity update method jointly guided by the inertia weight, individual learning coefficient and group learning coefficient, and the updated velocity is constrained to prevent the particle from crossing the search boundary. The particle positions are iteratively updated according to the position update method, and the updated positions are clipped to within the preset search boundary. After each round of position update, the updated particles are written into the historical magnetic state layer and the steady-state parameter layer is restored to the steady-state parameter locking snapshot. The reclosing simulation is then re-executed and the fitness is calculated. Update the historical best position of each particle and the global best position of the particle swarm based on the fitness of each particle. The termination of the iteration is determined by a convergence index, which reflects both the degree of improvement in the global optimal fitness and the degree of dispersion of all particles around the global optimal position.

6. The method according to claim 1, characterized in that, S4 includes: S4.1 Load the historical magnetic state layer from the optimal residual magnetism state to the opening time, lock the steady-state parameter layer to read-only state, advance the core magnetic state change within the power outage interval according to the residual magnetism decay trajectory, execute the closing operation at the actual closing time to generate the reclosing simulation state sequence, and divide the transient verification interval and steady-state verification interval with the end time of inrush current decay as the boundary. S4.2 Extract the transient response features of the reclosing simulation state sequence within the transient verification interval, compare them with the actual reclosing response sequence, calculate the transient continuity index, and form a transient verification qualified conclusion when the threshold is met. S4.3 After the transient verification is qualified, the reclosing simulation state sequence is verified from four dimensions within the steady-state verification interval: current recovery, magnetic flux period consistency, historical state exit, and steady-state parameter invariance. The steady-state recovery index is calculated. When the threshold is met and the steady-state parameter remains locked, the steady-state verification is qualified. The unidirectional magnetic flux bias decay time constant and steady-state parameter locking snapshot after closing are determined. S4.4 When both transient and steady-state verifications are qualified, the optimal combination of residual magnetism states, residual magnetism attenuation trajectory, unidirectional magnetic flux bias attenuation time constant after closing, and steady-state parameter lock snapshot are encapsulated into a residual magnetism dynamic update package; when any verification fails, the corresponding boundary is corrected and the process returns to step S3 for re-identification.

7. The method according to claim 1, characterized in that, S5 includes: S5.1, Read data from the residual magnetism dynamic update package and bind it to the historical magnetic state layer; S5.2, at the moment of each phase tripping, the signed residual magnetism and hysteresis position in the optimal residual magnetism state combination are written into the historical magnetic state layer of the corresponding iron core column. During the power outage interval, the residual magnetism amplitude of each iron core column is advanced hourly according to the residual magnetism decay trajectory, while keeping the residual magnetism direction unchanged. S5.3, At the closing time of each phase, the closing operation is performed. The residual magnetic state at the moment before closing is used as the initial condition to start the nonlinear excitation calculation, so that the core magnetic flux after closing is composed of steady-state periodic magnetic flux component and unidirectional magnetic flux bias component. The unidirectional magnetic flux bias component gradually exits as the excitation inrush current decays, generating the magnetic state evolution process. S5.4 After the closing operation is completed, the consistency index is updated from four dimensions: current recovery, magnetic flux cycle consistency, historical state exit and steady-state parameter unchanged within several consecutive power frequency cycles. When the threshold is met and the steady-state parameter is locked, the current residual magnetism state, magnetic state evolution process and steady-state parameter are solidified to form the dynamic update result of the digital twin.

8. A data-driven dynamic update system for a digital twin of power grid equipment, applied to the method described in any one of claims 1-7, characterized in that, The system includes: The magnetic state evolution module is used to establish a reclosing event time-stamped chain, obtain the estimated magnetic flux value, magnetic flux direction and magnetic flux calculation trajectory of each iron core column at the moment of opening, delineate the residual magnetic boundary of each iron core column before reclosing, and form a magnetic state history sequence and a reclosing actual response sequence. The updated structure building module is used to divide the model parameters of the digital twin into a steady-state parameter layer and a historical magnetic state layer. The particle elements are arranged according to the signed remanence, hysteresis position and remanence decay degree of each iron core column and the search boundary is set. Fitness evaluation rules are established and particle swarm execution parameters are set to form a particle swarm identification configuration. The particle swarm identification module is used to generate an initial particle swarm, calculate the fitness of each particle according to the fitness evaluation rules, iteratively update the particle positions until the termination condition is met, and output the optimal combination of remanent magnetization states and the remanent magnetization decay trajectory. The verification module is used to load the optimal residual magnetism state combination into the historical magnetic state layer and perform reclosing simulation to verify the transient response characteristics and steady-state recovery characteristics respectively. When the verification is qualified, the residual magnetism dynamic update package is encapsulated. The dynamic update module is used to bind the residual magnetism dynamic update package to the historical magnetic state layer. During the power outage interval, the residual magnetism decay is promoted according to the residual magnetism decay trajectory. After the circuit is closed, the unidirectional magnetic flux bias component is gradually withdrawn through nonlinear excitation calculation. When the update consistency meets the requirements, the update result is solidified.

9. An electronic device, characterized in that, The electronic device includes a memory and at least one processor, the memory storing a computer program, and the processor executing the computer program to implement the method of any one of claims 1-7.

10. A computer storage medium, characterized in that, It stores a computer program, which, when executed, performs the method according to any one of claims 1-7.

Images