A transformer DC bias suppression system and method

Abstract

This invention relates to the field of power transformer protection technology, and discloses a transformer DC bias suppression system and method. The method includes collecting subway operation, power grid structure, and soil geological parameters; establishing a dynamic calculation model based on these parameters to generate distribution data of rail potential, stray current, and ground potential; constructing a bias assessment model to calculate the dynamic DC current flowing through the transformer neutral point and the excitation current distortion; using a random forest algorithm for evaluation, outputting a quantitative impact level and a set of key factors; and based on this level, key factors, and transformer grounding method, formulating a suppression strategy that includes configuration parameters of a silicon carbide PiN diode active suppression device and a grounding optimization scheme. This invention achieves accurate assessment and targeted suppression of DC bias problems caused by subways.

Description

Technical Field

[0001] This invention relates to the field of power transformer protection technology, specifically to a transformer DC bias suppression system and method. Background Technology

[0002] With the increasing development of urban rail transit, especially subway systems, the problem of DC bias magnetization caused by stray currents generated by train operation on nearby power grid transformers is becoming increasingly prominent. DC bias magnetization can lead to increased transformer vibration, noise, and localized overheating, seriously threatening the safe and stable operation of the power grid.

[0003] Existing technologies mainly focus on the monitoring and generalized suppression of bias current. Monitoring methods rely on direct measurement using sensors installed at the transformer neutral point, which is a reactive approach and cannot predict or trace the source of bias current risk. Mitigation measures commonly employ capacitors connected in series at the transformer neutral point for DC blocking or reverse compensation. These devices typically operate based on fixed threshold values ​​and represent a general solution.

[0004] Existing technologies have shortcomings. They fail to dynamically couple and analyze the three major systems of subway operation, power grid structure, and soil geology, and cannot depict the complete spatial distribution path of stray currents from their source to the transformer neutral point, resulting in a lack of refined data support for risk assessment. The strategies of general suppression devices are preset and fixed, unable to adapt to the dynamic and regional differences in subway traction load, power grid operation mode, and changes in ground resistivity, resulting in a lack of targeted suppression effects. In addition, existing methods lack intelligent quantitative analysis of the degree of influence of multi-source factors, making it difficult to identify the dominant cause of the bias magnetization of a specific transformer.

[0005] A method is needed that can perform detailed simulation, intelligent quantitative evaluation, and generate customized suppression strategies for DC bias induced by subways throughout the entire process. Summary of the Invention

[0006] The purpose of this invention is to provide a transformer DC bias suppression system and method to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides a method for suppressing DC bias in a transformer, the method comprising:

[0008] Collect sets of subway operation parameters, AC power grid structure parameters, and soil geological parameters for the target area;

[0009] Based on the set of subway operation parameters, the set of AC power grid structure parameters, and the set of soil geological parameters, a dynamic DC bias current calculation model is established to generate rail potential distribution data, stray current distribution data, and ground potential distribution data under subway locomotive operating conditions.

[0010] Based on the stray current distribution data, the ground potential distribution data, and the AC power grid structure parameter set, a transformer DC bias magnetization evaluation model is constructed to calculate the dynamic DC bias magnetization current data flowing through the transformer neutral point and the corresponding excitation current distortion data.

[0011] Based on the dynamic DC bias current data and the excitation current distortion data, the random forest algorithm is used to evaluate the degree of DC bias influence, and to generate a quantitative evaluation level and a set of key influencing factors of the transformer affected by DC bias.

[0012] Based on the quantitative assessment level, the set of key influencing factors, and the grounding operation mode of the target transformer, a DC bias suppression strategy is formulated. The DC bias suppression strategy includes the configuration parameters of an active suppression device based on silicon carbide PiN diodes and an optimized grounding method.

[0013] Preferably, the step of establishing a dynamic DC bias current calculation model to generate rail potential distribution data, stray current distribution data, and ground potential distribution data under metro locomotive operating conditions includes:

[0014] Based on the aforementioned set of subway operating parameters, a four-layer ground network return structure model is constructed, which includes a combination of single-car and double-car sections. The four-layer ground network return structure model includes the electrical and geometric parameters of the rail layer, track bed layer, structural steel reinforcement layer, and ground layer, as well as the field-track coupling equation set.

[0015] By solving the field-circuit coupling equations of the four-layer ground grid return structure model, the continuous variation curve of rail-to-ground potential at any locomotive position is calculated, and the rail potential distribution data is generated.

[0016] Based on the rail potential distribution data and the soil resistivity distribution in the set of soil geological parameters, a stray current leakage diffusion equation is established. By solving the stray current leakage diffusion equation, the spatial density distribution of the current leaked from the rail to the ground is obtained, and the stray current distribution data is generated.

[0017] Based on the stray current distribution data, the dynamic change field of the geopotential in the nearby three-dimensional space is calculated by solving the Poisson equation with the subway line as the source, and the geopotential distribution data is generated.

[0018] Preferably, the step of constructing a transformer DC bias evaluation model and calculating the dynamic DC bias current data flowing through the transformer neutral point and the corresponding excitation current distortion data includes:

[0019] Extract the location coordinates of the target substation, transformer grounding resistance, and the equivalent DC resistance network topology of the power grid from the set of AC power grid structural parameters;

[0020] Based on the location coordinates of the target substation, extract the dynamic change sequence of the ground potential at the location of the target substation from the ground potential distribution data;

[0021] Based on the dynamic change sequence of the earth potential, the grounding resistance of the transformer, and the equivalent DC resistance network topology of the power grid, the DC loop equation of the transformer neutral point is established. By solving the DC loop equation of the transformer neutral point, the numerical sequence of the DC current flowing through the transformer neutral point changing with time is obtained, and the dynamic DC bias current data is generated.

[0022] The dynamic DC bias current data is input into the transformer excitation characteristic mapping model to calculate the transformer core operating point offset caused by DC bias, thereby obtaining the harmonic content change and effective value increment of the excitation current and generating the excitation current distortion data.

[0023] Preferably, the process of using the random forest algorithm to assess the degree of DC bias influence generates a quantitative assessment level and a set of key influencing factors for the transformer's DC bias influence, including:

[0024] A set of DC bias magnetic characteristic quantities is constructed, which includes the maximum and average values ​​of the dynamic DC bias magnetic current data, the odd harmonic growth rate in the excitation current distortion data, the maximum gradient value of the dynamic change sequence of the ground potential, and the operating density parameters of the metro locomotive.

[0025] The set of DC bias features is input into a pre-trained random forest regression evaluation model, which contains multiple decision trees. Each decision tree independently scores the degree of influence of DC bias based on a subset of features.

[0026] Summarize the scoring results of all decision trees, calculate the mean and variance of the DC bias comprehensive influence coefficient, and determine the quantitative evaluation level based on the preset threshold range in which the mean is located;

[0027] Analyze the Gini impurity reduction or feature importance score of each feature in the random forest regression evaluation model, select features whose importance scores exceed a set threshold, and generate the set of key influencing factors that includes the weight of subway operation density.

[0028] Preferably, the formulation of the DC bias suppression strategy includes:

[0029] Based on the grounding operation mode of the target transformer, determine whether the grounding type is direct grounding, grounding through a small resistor, or grounding through a capacitor;

[0030] If the grounding type is direct grounding or grounding through a small resistor, then based on the quantitative evaluation level and the amplitude of the dynamic DC bias current data, the withstand voltage level and current capacity of the silicon carbide PiN diodes to be connected in series are calculated, and the configuration parameters of the active suppression device based on silicon carbide PiN diodes are generated.

[0031] If the grounding type is grounded via a capacitor, then the capacitor's capacitance value optimization range is calculated based on the quantitative assessment level, so that the capacitor can meet the AC system grounding safety requirements while isolating the DC component, and the grounding method optimization scheme is generated.

[0032] By combining the weights of subway operation density in the set of key influencing factors, a dynamic switching logic for the active suppression device is generated. The dynamic switching logic is associated with the peak periods of the subway timetable and the rail potential distribution data.

[0033] The configuration parameters for generating the active suppression device based on the silicon carbide PiN diode include:

[0034] Based on the peak and effective values ​​of the dynamic DC bias current data, the continuous current thermal effect parameters and instantaneous impact current parameters that the active suppression device needs to withstand are calculated.

[0035] Based on the continuous current thermal effect parameters, the area requirements and heat sink specifications of the silicon carbide PiN diode chip are determined.

[0036] Based on the instantaneous inrush current parameters, determine the number of chips that need to be connected in series in the diode series stack to meet the reverse blocking voltage requirements;

[0037] Based on the current-carrying capacity of the transformer neutral point grounding conductor, determine the trigger current threshold of the bypass protection circuit connected in parallel with the diode stack.

[0038] Preferably, the method further includes:

[0039] Within a preset simulation period, the operation scenarios of subway trains under different timetables and different train formations are simulated;

[0040] For each operating scenario, the dynamic DC bias current calculation model and the transformer DC bias evaluation model are invoked to calculate the corresponding dynamic DC bias current data simulation set and excitation current distortion data simulation set.

[0041] The dynamic DC bias current data simulation set and the excitation current distortion data simulation set are input into the random forest regression evaluation model to generate a quantitative evaluation level distribution map under multiple scenarios.

[0042] Based on the quantitative assessment level distribution map, key operating scenario modes that cause the DC bias magnetism level of the transformer to reach the warning threshold are identified.

[0043] Preferably, the method further includes:

[0044] Establish the correlation and mapping relationship between the key operation scenario modes and subway operation scheduling parameters;

[0045] Based on the aforementioned correlation mapping relationship, auxiliary scheduling suggestions for power grid security are generated, including suggestions for adjusting the subway departure density and section operating speed.

[0046] The auxiliary scheduling suggestions are coordinated and optimized with the dynamic switching logic of the active suppression device based on silicon carbide PiN diodes to form a comprehensive governance scheme covering source-side regulation and receiver-side suppression.

[0047] Preferably, the method further includes:

[0048] A DC current monitoring device was installed at the neutral point of the target transformer to collect the actual measured data sequence of DC bias current during operation.

[0049] The measured DC bias current data sequence is compared and analyzed with the dynamic DC bias current data calculated by the transformer DC bias evaluation model within the corresponding time period, and the model prediction error rate is calculated.

[0050] Based on the model prediction error rate, an adaptive filtering algorithm is used to reverse-correct the soil resistivity distribution parameters in the four-layer ground grid backflow structure model.

[0051] The dynamic DC bias current calculation model is updated using the corrected soil resistivity distribution parameters.

[0052] Preferably, the set of subway operation parameters for the target area includes the real-time location, traction current, operating speed, and train formation information of the subway locomotives; the set of AC power grid structure parameters includes the geographical location of substations, transformer grounding methods and grounding resistance values, and DC resistance parameters of overhead power lines and cables; and the set of soil geological parameters includes soil layering structure and resistivity data obtained through geological exploration.

[0053] Preferably, the present invention also includes a transformer DC bias suppression system, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, it implements the steps of the transformer DC bias suppression method described above.

[0054] Compared with the prior art, the beneficial effects of the present invention are:

[0055] By establishing a computational model integrating subway operation, power grid structure, and soil geological parameters, spatial distribution data of rail potential, stray current, and ground potential are dynamically generated. This allows the analysis of DC bias problems to move from single-point measurement of the transformer neutral point to a quantitative simulation of the entire path of stray current generation, diffusion, and accumulation. This method provides the refined spatial data foundation needed to assess bias risk.

[0056] The random forest algorithm was applied to process dynamic bias current and excitation current distortion data, outputting a quantitative assessment level and identifying a set of key influencing factors. This transformed the assessment process from relying on empirical thresholds to data-driven intelligent judgment, not only determining the severity of bias current but also clarifying that its main causes originate from specific factors in the subway, power grid, or geology, providing a clear and targeted basis for subsequent decision-making.

[0057] Based on the quantitative levels and key factor sets output by the intelligent assessment, and combined with the specific grounding method of the transformer, the generated suppression strategy is highly customized. The strategy explicitly adopts an active suppression device based on silicon carbide PiN diodes, and directly provides its configuration parameters and grounding optimization scheme. This allows the suppression measures to accurately match the specific causes and degrees of magnetic bias, while leveraging the performance advantages of silicon carbide devices to improve the response speed and operational reliability of the suppression device. Attached Figure Description

[0058] Figure 1 This is a schematic diagram illustrating the working principle of the transformer DC bias suppression method described in this invention.

[0059] Figure 2 A flowchart for establishing a dynamic DC bias current calculation model;

[0060] Figure 3 This is a flowchart illustrating the process of evaluating the impact of DC bias using the random forest algorithm.

[0061] Figure 4 A bar chart comparing the parameters of DC bias suppression strategies for different grounding types of transformers;

[0062] Figure 5 This is a comparison chart of soil stratified resistivity correction and error rate. Detailed Implementation

[0063] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] Please see Figure 1 This invention provides a method for suppressing DC bias in transformers. The method includes: collecting a set of subway operating parameters, an AC power grid structural parameter set, and a soil geological parameter set for the target area; establishing a dynamic DC bias current calculation model based on the subway operating parameters, AC power grid structural parameter set, and soil geological parameter set, generating rail potential distribution data, stray current distribution data, and ground potential distribution data under subway locomotive operating conditions; constructing a transformer DC bias assessment model based on the stray current distribution data, ground potential distribution data, and AC power grid structural parameter set, calculating the dynamic DC bias current data flowing through the transformer neutral point and the corresponding excitation current distortion data; using the dynamic DC bias current data and excitation current distortion data, evaluating the degree of DC bias influence using a random forest algorithm, generating a quantitative assessment level of the transformer's DC bias influence and a set of key influencing factors; and formulating a DC bias suppression strategy based on the quantitative assessment level, the set of key influencing factors, and the grounding operation mode of the target transformer. The DC bias suppression strategy includes the configuration parameters of an active suppression device based on silicon carbide PiN diodes and an optimized grounding method.

[0065] Example 1: See Figure 2 A dynamic DC bias current calculation model was established to generate rail potential distribution data, stray current distribution data, and ground potential distribution data under metro locomotive operating conditions. The process is as follows: Based on the metro operating parameter set, a four-layer ground network return structure model was constructed, including combinations of single-car and double-car sections. The four-layer ground network return structure model includes the electrical and geometric parameters and field-circuit coupling equations of the rail layer, track bed layer, structural reinforcement layer, and ground layer. By solving the field-circuit coupling equations of the four-layer ground network return structure model, the continuous change curve of rail-to-ground potential at any locomotive position was calculated, generating rail potential distribution data. Based on the rail potential distribution data and the soil resistivity distribution in the soil geological parameter set, a stray current leakage and diffusion equation was established. By solving the stray current leakage and diffusion equation, the spatial density distribution of current leaking from the rail to the ground was obtained, generating stray current distribution data. Based on the stray current distribution data, by solving the Poisson equation with the metro line as the source, the dynamic change field of ground potential in the nearby three-dimensional space was calculated, generating ground potential distribution data.

[0066] A transformer DC bias evaluation model is constructed to calculate the dynamic DC bias current data flowing through the transformer neutral point and the corresponding excitation current distortion data. The process is as follows: The location coordinates of the target substation, the transformer grounding resistance, and the topology of the equivalent DC resistance network of the power grid are extracted from the AC power grid structural parameter set. Based on the location coordinates of the target substation, the dynamic change sequence of the ground potential at the target substation location is extracted from the ground potential distribution data. Based on the dynamic change sequence of the ground potential, the transformer grounding resistance, and the topology of the equivalent DC resistance network of the power grid, the DC loop equation of the transformer neutral point is established. By solving the DC loop equation of the transformer neutral point, the numerical sequence of the DC current flowing through the transformer neutral point over time is obtained, generating dynamic DC bias current data. The dynamic DC bias current data is input into the transformer excitation characteristic mapping model to calculate the offset of the transformer core operating point caused by DC bias, thereby obtaining the change in harmonic content and effective value increment of the excitation current, generating excitation current distortion data.

[0067] In the specific implementation, the process involves a concrete example scenario of a subway line and a nearby 220 kV substation. The subway operation parameter set collects data from a specific operating period of the subway line, including the real-time position sequence of locomotives, corresponding traction current waveforms, operating speed curves, and train formation information. The AC power grid structure parameter set obtains the location coordinates of the 220 kV substation, the grounding parameters of the main transformer's star-connected side grounded with a 0.5 ohm small resistor, and the DC resistance parameters of the 110 kV and 10 kV outgoing cables connected to the substation. The soil geological parameter set is derived from the geological exploration report of the area, including the soil stratification structure and resistivity data, divided into three layers with different resistivities from the surface to a depth of 50 meters.

[0068] The process of establishing a dynamic DC bias current calculation model is based on these specific parameters. Based on the set of subway operation parameters including single-car operation and dual-car tracking scenarios, a four-layer ground grid return structure model describing the current flowing from the locomotive into the rails and returning to the traction substation through multiple layers of media is constructed. The four-layer ground grid return structure model defines electrical and geometric parameters such as longitudinal resistance, ground leakage conductance, and interlayer mutual resistance coefficients for the rail layer, track bed layer, structural reinforcement layer, and ground layer. The field-circuit coupling equations describing the potential and current distribution in the four-layer ground grid return structure model are solved numerically to obtain the continuous change curve of rail-to-ground potential corresponding to the real-time locomotive position sequence, generating rail potential distribution data. Subsequently, based on the rail potential distribution data and the soil resistivity distribution in the soil geological parameter set, a stray current leakage diffusion equation is established, and the finite element method is used to solve this equation to obtain the spatial density distribution of current leaked from the rail to the ground, generating stray current distribution data. Finally, based on the stray current distribution data as a spatial current source, the dynamic change field of the ground potential in the nearby three-dimensional space is calculated by solving the Poisson equation with the subway line as the source distribution area, generating ground potential distribution data. The field-circuit coupling equations describing the potential and current distribution in the four-layer ground grid return structure model are solved numerically, and the matrix form of the field-circuit coupling equations can be expressed as:

[0069] ;

[0070] in: It is an admittance matrix composed of the self-admittance of each layer and the mutual admittance between layers. It is a column vector containing the ground potential of the rail layer, track bed layer, and structural steel reinforcement layer. It is a column vector containing the current source injected by the locomotive and the current sink at the return point of the traction substation. Solving this system of equations yields a continuous curve of rail-to-ground potential change corresponding to the real-time locomotive position sequence in the metro operating parameter set. This curve generates rail potential distribution data.

[0071] In some embodiments, based on rail potential distribution data and the three-layer soil resistivity distribution provided in the soil geological parameter set, a stray current leakage diffusion equation describing the loss of current from the rail to the three-dimensional space of the earth is established. This equation uses the rail potential as a boundary condition. The stray current leakage diffusion equation is solved using the finite element method to obtain the spatial density vector distribution of the leakage current in the underground layer, generating stray current distribution data. Based on the spatial current source characterized by the stray current distribution data, the dynamic scalar field of the earth potential in the surrounding three-dimensional space, including the target 220 kV substation, is calculated by solving the Poisson equation with the subway line alignment as the source distribution region. The spatiotemporal data of this scalar field is then used to generate the earth potential distribution data.

[0072] The process of constructing the transformer DC bias evaluation model relies on the output of the aforementioned steps. From the AC power grid structural parameter set, the location coordinates of the target 220 kV substation, the 0.5 ohm grounding resistance of the main transformer neutral point, and the simplified equivalent DC resistance network topology of the power grid based on the power grid wiring diagram are extracted. This equivalent DC resistance network topology includes the ground DC resistance connection between the grounding neutral points of adjacent substations. Based on the precise location coordinates of the target substation, the dynamic change sequence of the ground potential at that location point, synchronized with the rail potential change, is extracted by interpolation from the three-dimensional ground potential distribution data. Based on the dynamic change sequence of the ground potential, the transformer grounding resistance value, and the equivalent DC resistance network topology of the power grid, a DC loop equation for the transformer neutral point is established, with the ground potential as the driving voltage and the transformer grounding resistance and the equivalent DC resistance network of the power grid as the loop. By solving the DC loop equation for the transformer neutral point, the numerical sequence of the DC current flowing through the neutral point of the main transformer over time during the subway operation period in the example scenario is obtained, generating dynamic DC bias current data. Optionally, dynamic DC bias current data is used as input and fed into a transformer excitation characteristic mapping model pre-fitted from transformer no-load test data. The transformer excitation characteristic mapping model calculates the shift in the transformer core operating point caused by the DC bias current. Based on the core operating point shift, the growth rate of odd harmonic content in the excitation current waveform relative to the unbiased state, as well as the increment of the effective value of the excitation current, are further calculated. These calculation results are collectively generated as excitation current distortion data.

[0073] Example 2: See Figure 3 This paper utilizes a random forest algorithm to assess the impact of DC bias on transformers, generating a quantitative assessment level and a set of key influencing factors. The process is as follows: First, a set of DC bias feature quantities is constructed, including the maximum and average values ​​of dynamic DC bias current data, the odd harmonic growth rate in excitation current distortion data, the maximum gradient of the dynamic change sequence of ground potential, and the operating density parameters of subway trains. This set of DC bias feature quantities is then input into a pre-trained random forest regression assessment model. The model contains multiple decision trees, each independently scoring the impact of DC bias based on a subset of features. The scores from all decision trees are then summarized, and the mean and variance of the comprehensive impact coefficient of DC bias are calculated. The quantitative assessment level is determined based on the preset threshold range where the mean falls. Finally, the Gini impurity reduction or feature importance score of each feature quantity in the random forest regression assessment model is analyzed. Features with importance scores exceeding a set threshold are selected to generate a set of key influencing factors with weights including subway operating density.

[0074] In the specific implementation, the random forest algorithm is used to assess the degree of DC bias influence. Taking a specific 220 kV main transformer as an example, the dynamic DC bias current data and excitation current distortion data of the main transformer have been calculated using the transformer DC bias assessment model. The dynamic DC bias current data includes a DC current value with a time series of 1440 minutes. The excitation current distortion data includes the total harmonic distortion rate, third harmonic content rate, fifth harmonic content rate, and effective value of the excitation current for the corresponding time series. Based on these data, a set of DC bias characteristic quantities is constructed. The set of DC bias characteristic quantities includes the maximum value of 15.7 amps and the average value of 3.2 amps of the dynamic DC bias current data, the odd harmonic growth rate in the excitation current distortion data, the maximum gradient value of the dynamic change sequence of the ground potential within a 5-minute time window of 0.12 volts per minute, and the average number of trains departing per hour during the assessment period extracted from the subway operation parameter set, i.e., the operation density parameter.

[0075] The set of DC bias magnetization features, including the five specific numerical characteristics mentioned above, is input into a pre-trained random forest regression evaluation model. The random forest regression evaluation model contains 100 decision trees. During training, each decision tree independently scores the degree of DC bias magnetization influence based on four feature subsets randomly selected from the DC bias magnetization feature set. The score output is a comprehensive DC bias magnetization influence coefficient between 0 and 1. The scores from all decision trees are summarized, and the mean and variance of the 100 comprehensive DC bias magnetization influence coefficients are calculated. In some embodiments, a preset threshold range is defined as follows: a mean between 0 and 0.3 corresponds to quantitative evaluation level I, a mean between 0.3 and 0.7 corresponds to quantitative evaluation level II, and a mean between 0.7 and 1.0 corresponds to quantitative evaluation level III. Based on the preset threshold range where the calculated mean of 0.65 falls, the quantitative evaluation level of the transformer in this evaluation period is determined to be level II.

[0076] This study analyzes the importance of each feature in a random forest regression evaluation model, generating a set of key influencing factors. This analysis is based on the total reduction in Gini impurity recorded during model training at all decision tree nodes for each feature variable. For the regression task, the feature importance score is used. Features can be calculated It is obtained by taking the mean of the reduction in Gini impurity across all trees, and its expression is:

[0077] ;

[0078] in: Indicates the feature importance score. It is the total number of decision trees. Represents a single decision tree. It is a decision tree Use features The set of nodes to be split. It is a node The weighted reduction in impurity of child nodes after splitting. The importance scores of the five features—maximum DC bias current, maximum ground potential gradient, metro operating density parameter, average DC bias current, and third harmonic growth rate—were calculated to be 0.32, 0.28, 0.19, 0.12, and 0.09, respectively. An importance score threshold of 0.15 was set, and the top three features with scores exceeding 0.15—maximum DC bias current, maximum ground potential gradient, and metro operating density parameter—were selected. These three features constitute the set of key influencing factors.

[0079] Example 3: A DC bias suppression strategy is formulated, as follows: Based on the grounding operation mode of the target transformer, the grounding type is determined to be direct grounding, grounding through a small resistor, or grounding through a capacitor. If the grounding type is direct grounding or grounding through a small resistor, the withstand voltage rating and current capacity of the required series-connected silicon carbide PiN diodes are calculated based on the quantitative assessment level and the amplitude of the dynamic DC bias current data, generating configuration parameters for an active suppression device based on silicon carbide PiN diodes. If the grounding type is grounding through a capacitor, the capacitor's capacitance optimization range is calculated based on the quantitative assessment level, ensuring that the capacitor meets the AC system grounding safety requirements while isolating the DC component, generating an optimized grounding scheme. Combining the weight of the metro operating density in the set of key influencing factors, a dynamic switching logic for the active suppression device is generated. This dynamic switching logic is correlated with the peak periods of the metro timetable and rail potential distribution data.

[0080] The configuration parameters for an active suppression device based on silicon carbide (PiN) diodes are generated as follows: Based on the peak and RMS values ​​of the dynamic DC bias current data, the continuous current thermal effect parameters and instantaneous inrush current parameters that the active suppression device needs to withstand are calculated. Based on the continuous current thermal effect parameters, the area requirements of the silicon carbide PiN diode chips and the heat sink specifications are determined. Based on the instantaneous inrush current parameters, the number of chips required to be connected in series in the diode stack is determined to meet the reverse blocking voltage requirements. Based on the current carrying capacity of the transformer neutral point grounding conductor, the trigger current threshold of the bypass protection circuit connected in parallel with the diode stack is determined.

[0081] In specific implementation, a DC bias suppression strategy is formulated for a specific 110 kV substation main transformer. The main transformer's grounding operation mode is low-resistance grounding with a resistance value of 0.5 ohms, and the quantitative assessment level is Class II. The peak value of the dynamic DC bias current data is 18.5 amps, and the effective value is 4.1 amps. Based on the target transformer's grounding operation mode, the grounding type is determined to be low-resistance grounding. In some embodiments, since the grounding type is low-resistance grounding, based on the quantitative assessment level II and the peak value of 18.5 amps and the effective value of 4.1 amps of the dynamic DC bias current data, the withstand voltage rating and current capacity of the required series-connected silicon carbide PiN diodes are calculated. The withstand voltage rating calculation considers the highest voltage to ground at the transformer neutral point, and the current capacity calculation is based on the effective value of 4.1 amps and the peak value of 18.5 amps of the dynamic DC bias current data, generating configuration parameters for an active suppression device based on silicon carbide PiN diodes. Optionally, in another scenario where the grounding type is capacitor grounding, the target transformer grounding operation mode is capacitor grounding with a capacitance value of 10 microfarads. Based on the quantitative assessment level II, the capacitance value optimization range is calculated so that the capacitor can meet the AC system grounding safety requirements while isolating the DC component. The lower limit of the capacitance value optimization range is determined by the capacitive reactance required to suppress DC bias magnetization, and the upper limit is determined by the capacitance current limit when the AC system is in a single-phase grounding fault, thus generating an optimized grounding scheme.

[0082] It is understandable that, by combining the weight of subway operating density (0.19) in the set of key influencing factors, the dynamic switching logic of the suppression device is generated. The dynamic switching logic is associated with the peak periods of the subway timetable and rail potential distribution data. For example, during the morning peak period when the subway operating density is high, the dynamic switching logic is set to activate the suppression device in advance. During the night period when the operating density is low, the dynamic switching logic is set to activate or deactivate the suppression device based on the real-time monitored DC bias current data. The configuration parameters for an active suppression device based on silicon carbide (PiN) diodes are generated as follows: Based on the peak value of 18.5 amps and the effective value of 4.1 amps from the dynamic DC bias current data, the continuous current thermal effect parameters and instantaneous inrush current parameters that the suppression device needs to withstand are calculated. The continuous current thermal effect parameters are obtained by multiplying the square of the effective current by the time integral, and the instantaneous inrush current parameters are obtained by multiplying the peak current by the inrush coefficient. Based on the continuous current thermal effect parameters, the area requirements of the silicon carbide (PiN) diode chip and the heat sink specifications are determined. The chip area requirements are calculated based on the diode's on-state current density, and the heat sink specifications are calculated based on thermal resistance and power dissipation. Based on the instantaneous inrush current parameters, the number of chips required to be connected in series in the diode series stack is determined to meet the reverse blocking voltage requirements. The calculation formula is:

[0083] ;

[0084] in: Indicates the number of chips connected in series. It is the maximum reverse voltage that the suppression device needs to block. It is the rated reverse blocking voltage of a single silicon carbide PiN diode chip. This indicates rounding up. Based on the current-carrying capacity of the transformer's neutral point grounding conductor, the trigger current threshold of the bypass protection circuit connected in parallel with the diode stack is determined. The trigger current threshold is set to 1.2 times the rated current-carrying capacity of the grounding conductor. In some embodiments, for transformers with direct grounding, the process of formulating a DC bias suppression strategy is similar, but the limit of neutral point potential offset in the direct grounding system needs to be considered when calculating the silicon carbide PiN diode configuration parameters. Optionally, the implementation of the dynamic switching logic relies on a programmable logic controller (PLC). The PLC receives signals from the subway timetable and signals from the rail potential monitoring device, and outputs switching instructions for the suppression device according to a preset logic table.

[0085] See Figure 4 This is a bar chart comparing the parameters of DC bias suppression strategies for different transformer grounding types. This chart is used for selecting and analyzing DC bias suppression strategies for transformers; the parameter differences for different grounding types correspond to different suppression schemes. Direct / low-resistor grounding: requires high-voltage-rated silicon carbide PiN diodes (the main parameter is the voltage rating); capacitor grounding: requires optimizing the capacitor value; auxiliary parameters correspond to details such as current flow, heat dissipation, or capacitance limits of the suppression device. This chart allows for a quick comparison of the cost and technical difficulty of suppression strategies for different grounding types: direct / low-resistor grounding requires high-voltage-rated devices, while capacitor grounding offers more flexibility in capacitance optimization.

[0086] Example 4: Within a preset simulation period, simulate subway train operation scenarios under different timetables and train formation numbers. For each operation scenario, call the dynamic DC bias current calculation model and the transformer DC bias current assessment model to calculate the corresponding dynamic DC bias current data simulation set and excitation current distortion data simulation set. Input the dynamic DC bias current data simulation set and the excitation current distortion data simulation set into the random forest regression assessment model to generate a quantitative assessment level distribution map under multiple scenarios. Based on the quantitative assessment level distribution map, identify the key operation scenario modes that cause the transformer DC bias current impact level to reach the warning threshold.

[0087] Establish a mapping relationship between key operational scenario modes and metro operation scheduling parameters. Based on this mapping relationship, generate auxiliary scheduling suggestions for power grid security, including suggestions for adjusting metro departure density and section operating speed. Coordinate and optimize the auxiliary scheduling suggestions with the dynamic switching logic of an active suppression device based on silicon carbide PiN diodes to form a comprehensive governance solution covering source-side control and receiver-side suppression.

[0088] In the specific implementation, the simulation period was preset to be 24 consecutive hours, simulating the operation scenarios of subway trains under different timetables and different train formation numbers. The operation scenarios included five typical scenarios: 6-car formation during off-peak hours, 8-car formation during peak hours, mixed operation of long and short routes, single-route operation at night, and temporary addition of express trains. For each operation scenario, a dynamic DC bias current calculation model and a transformer DC bias evaluation model were invoked. The corresponding subway operation parameter set for the scenario was input, and the corresponding dynamic DC bias current data simulation set and excitation current distortion data simulation set were calculated. The dynamic DC bias current data simulation set contained five time series arrays, and the excitation current distortion data simulation set contained five sets of harmonic analysis results. The dynamic DC bias current data simulation set and the excitation current distortion data simulation set were input into a pre-trained random forest regression evaluation model to evaluate the data for the entire simulation period of each scenario, generating a quantitative evaluation level distribution map for multiple scenarios. See Table 1, which presents the comprehensive evaluation results corresponding to different operation scenarios in tabular form.

[0089] Table 1: Distribution of Quantitative Assessment Levels in Multiple Scenarios

[0090]

[0091] Based on the quantitative assessment level distribution map, key operating scenario modes that cause the transformer DC bias magnetization level to reach the warning threshold III were identified. These key operating scenario modes include the 8-car train minimum interval operation mode during the morning peak and the temporary addition of express trains. In some embodiments, a correlation mapping relationship between the key operating scenario modes and metro operation scheduling parameters is established. This correlation mapping is achieved by analyzing the correlation between characteristic parameters under the key operating scenario modes and metro operation scheduling parameters. The correlation calculation formula uses the Pearson correlation coefficient:

[0092] ;

[0093] in: This represents the calculated Pearson correlation coefficient. This represents a sequence of characteristic parameters for key operational scenarios, such as the sequence of integral values ​​of traction current over a given time period. It is its mean. This represents a sequence of subway operation and scheduling parameters, such as the train interval time sequence. This is its average value. Calculation results show that the correlation coefficient between the peak DC bias current and train departure density is -0.89, and the correlation coefficient with the average number of train formations is 0.92. It can be understood that, based on the correlation mapping relationship, auxiliary scheduling suggestions for grid security are generated. These suggestions include adjustments to subway train departure density and interval operating speed. For example, during periods of grid weakness, it is suggested to temporarily reduce the departure density of scenario S2 from 24 pairs / hour to 18-20 pairs / hour, or adjust the operating speed of express trains in scenario S5 from 80 km / h to 60-65 km / h. The auxiliary scheduling suggestions are then coordinated and optimized with the dynamic switching logic of an active suppression device based on silicon carbide PiN diodes to form a comprehensive governance scheme covering source-side control and receiver-side suppression. This coordinated optimization is reflected in the fact that when the scheduling system executes the auxiliary scheduling suggestions, it simultaneously sends a signal to the control system of the suppression device, allowing the suppression device to switch to the corresponding level in advance according to the adjusted operating schedule. Optionally, the establishment of the association mapping relationship can be based on historical data mining, using clustering algorithms to classify historical operation data into different scenario patterns, and then statistically analyzing the correspondence between each pattern and the evaluation level.

[0094] In some embodiments, the comprehensive governance scheme is stored in the control system in the form of a strategy table. The strategy table defines the source-side dispatching measures and preset parameters of the receiving-end suppression devices that should be prioritized under different combinations of power grid operation modes and subway operation plans. It is understood that the update of the quantitative assessment level distribution map is periodic. When there are major adjustments to the subway operation plan or changes in the power grid structure, it is necessary to re-perform multi-scenario simulations to update the map and comprehensive governance scheme.

[0095] Example 5: A DC current monitoring device is installed at the neutral point of the target transformer to collect the measured DC bias current data sequence during actual operation. The measured DC bias current data sequence is compared and analyzed with the dynamic DC bias current data calculated by the transformer DC bias current evaluation model within the corresponding time period, and the model prediction error rate is calculated. Based on the model prediction error rate, an adaptive filtering algorithm is used to reverse-correct the soil resistivity distribution parameters in the four-layer grounding grid return structure model. The corrected soil resistivity distribution parameters are used to update the dynamic DC bias current calculation model.

[0096] The set of subway operation parameters for the target area includes the real-time location, traction current, operating speed, and train formation information of subway trains. The set of AC power grid structure parameters includes the geographical location of substations, transformer grounding methods and grounding resistance values, and DC resistance parameters of overhead lines and cables. The set of soil geological parameters includes soil layering structure and resistivity data obtained through geological exploration.

[0097] In practical implementation, a DC current monitoring device is installed at the neutral point of the target transformer. This device utilizes the Hall effect principle, has a range of ±50 amperes, and an accuracy of 0.5 class. It collects measured DC bias current data sequences during actual operation, recording the neutral point DC current value at 1-minute intervals for a complete morning peak period. The measured DC bias current data sequences are then compared with the dynamic DC bias current data calculated by the transformer DC bias assessment model for the corresponding time period. The model prediction error rate is calculated. The comparative analysis selects the same time segment where the model-calculated dynamic DC bias current value is 15.7 amperes, while the measured DC bias current data sequence value is 18.2 amperes. The model prediction error rate is calculated using the following formula:

[0098] ;

[0099] in: This represents the model prediction error rate. This represents the sequence of measured DC bias current values. The value represents the dynamic DC bias current calculated by the model. The error rate obtained by substituting the value is 15.9%.

[0100] Based on the model prediction error rate, an adaptive filtering algorithm is used to reverse-correct the soil resistivity distribution parameters in the four-layer ground grid return structure model. The adaptive filtering algorithm employs the least mean square algorithm, using the model prediction error rate as the input signal to adjust the estimated values ​​of the soil resistivity distribution parameters. It can be understood that the soil resistivity distribution parameters have a three-layer structure, with initial values ​​from the geological exploration report: 50 ohm-meters, 120 ohm-meters, and 300 ohm-meters, respectively. The correction process iteratively updates the soil resistivity values ​​of the middle layer (the second layer) based on the error rate. The corrected soil resistivity distribution parameters are used to update the dynamic DC bias current calculation model. These updated parameters are then substituted into the four-layer ground grid return structure model to recalculate the rail potential distribution data, stray current distribution data, and ground potential distribution data.

[0101] In some embodiments, the set of subway operation parameters for the target area includes the real-time location, traction current, operating speed, and train formation information of the subway locomotive. The real-time location is obtained through the latitude and longitude coordinates of the automatic train monitoring system, the traction current is obtained from the waveform data recorded by the traction substation, and the operating speed and train formation information are obtained from the train timetable. The set of AC power grid structure parameters includes the geographical location of the substation, the grounding method and grounding resistance value of the transformer, and the DC resistance parameters of the overhead power lines and cables. The geographical location of the substation is obtained through GPS coordinates, the grounding method of the transformer is grounded through a small resistor with a resistance value of 0.5 ohms, and the DC resistance parameters of the overhead power lines and cables are obtained through the line parameter manual and actual measurements. The set of soil geological parameters includes the soil layer structure and resistivity data obtained through geological exploration. The geological exploration adopts the symmetrical four-electrode sounding method, and the soil layer structure is shown as three layers: backfill soil, sandy clay, and rock. The resistivity data corresponds to the initial value before correction. Optionally, the measured data sequence of DC bias current collected by the DC current monitoring device can be stored in a database for periodic verification and training of the random forest regression evaluation model.

[0102] In some embodiments, the calculation of the model prediction error rate can be performed separately for different operating conditions, such as distinguishing between peak and off-peak periods, and calculating the average error rate for the corresponding time periods. It can be understood that the reverse correction of soil resistivity distribution parameters is a continuous process. Whenever the measured DC bias current data sequence accumulates to a certain length, such as once a week or once a month, an automatic correction process is triggered to maintain the accuracy of the dynamic DC bias current calculation model. Optionally, the step size factor in the adaptive filtering algorithm can be dynamically adjusted according to the magnitude of the model prediction error rate; when the error rate is large, the step size is increased to accelerate convergence, and when the error rate is small, the step size is decreased to improve stability.

[0103] See Figure 5 This is a chart comparing soil resistivity correction and error rates across different soil layers. This chart is used for soil parameter correction in a transformer DC bias magnetization calculation model. The resistivity differences between different soil layers are significant, and these are key parameters for calculating stray current diffusion and geopotential distribution. The initial and corrected resistivity of the sandy clay layer differs greatly, corresponding to a lower error rate, indicating a significant correction effect. The resistivity of the rock layer is not corrected, but the error rate is extremely high, suggesting further optimization of the correction method may be needed. This chart can be used to evaluate the effectiveness of soil resistivity correction: the corrected parameters can reduce the error in DC bias current calculation, improve model prediction accuracy, and provide more reliable basic data for subsequent suppression strategy formulation.

[0104] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0105] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for suppressing DC bias in a transformer, characterized in that, The method includes: Collect sets of subway operation parameters, AC power grid structure parameters, and soil geological parameters for the target area; Based on the set of subway operation parameters, the set of AC power grid structure parameters, and the set of soil geological parameters, a dynamic DC bias current calculation model is established to generate rail potential distribution data, stray current distribution data, and ground potential distribution data under subway locomotive operating conditions. Based on the stray current distribution data, the ground potential distribution data, and the AC power grid structure parameter set, a transformer DC bias magnetization evaluation model is constructed to calculate the dynamic DC bias magnetization current data flowing through the transformer neutral point and the corresponding excitation current distortion data. Based on the dynamic DC bias current data and the excitation current distortion data, the random forest algorithm is used to evaluate the degree of DC bias influence, and to generate a quantitative evaluation level and a set of key influencing factors of the transformer affected by DC bias. Based on the quantitative assessment level, the set of key influencing factors, and the grounding operation mode of the target transformer, a DC bias suppression strategy is formulated. The DC bias suppression strategy includes the configuration parameters of the active suppression device based on silicon carbide PiN diodes and the grounding mode optimization scheme. The establishment of a dynamic DC bias current calculation model generates rail potential distribution data, stray current distribution data, and ground potential distribution data under metro locomotive operating conditions, including: Based on the aforementioned set of subway operating parameters, a four-layer ground network return flow structure model is constructed, comprising combinations of single-car and double-car sections. This model includes electrical and geometric parameters and a set of field-track coupling equations for the rail layer, track bed layer, structural reinforcement layer, and ground layer. By solving the field-track coupling equations of the four-layer ground network return flow structure model, the continuous change curve of the rail-to-ground potential at any locomotive position is calculated, generating the rail potential distribution data. Based on the rail potential distribution data and the soil resistivity distribution in the set of soil geological parameters, a stray current leakage and diffusion equation is established. By solving this equation, the spatial density distribution of the current leaking from the rail to the ground is obtained, generating the stray current distribution data. Based on the stray current distribution data, by solving the Poisson equation with the subway line as the source, the dynamic change field of the ground potential in the surrounding three-dimensional space is calculated, generating the ground potential distribution data. The construction of the transformer DC bias evaluation model, which calculates the dynamic DC bias current data flowing through the transformer neutral point and the corresponding excitation current distortion data, includes: Extract the location coordinates of the target substation, transformer grounding resistance, and the equivalent DC resistance network topology of the power grid from the AC power grid structure parameter set; based on the location coordinates of the target substation, extract the dynamic change sequence of the ground potential at the location of the target substation from the ground potential distribution data; based on the dynamic change sequence of the ground potential, the transformer grounding resistance, and the equivalent DC resistance network topology of the power grid, establish the DC loop equation of the transformer neutral point; by solving the DC loop equation of the transformer neutral point, obtain the numerical sequence of the DC current flowing through the transformer neutral point changing with time, and generate the dynamic DC bias current data; input the dynamic DC bias current data into the transformer excitation characteristic mapping model, calculate the offset of the transformer core operating point caused by DC bias, and then obtain the harmonic content change and effective value increment of the excitation current, and generate the excitation current distortion data; The random forest algorithm is used to assess the degree of DC bias influence, generating a quantitative assessment level of the transformer's DC bias influence and a set of key influencing factors, including: A set of DC bias magnetic characteristic quantities is constructed, which includes the maximum and average values ​​of the dynamic DC bias magnetic current data, the odd harmonic growth rate in the excitation current distortion data, the maximum gradient value of the dynamic change sequence of the ground potential, and the operating density parameters of the subway locomotive. This set of DC bias magnetic characteristic quantities is input into a pre-trained random forest regression evaluation model, which contains multiple decision trees. Each decision tree independently scores the degree of DC bias magnetic influence based on a subset of features. The scoring results of all decision trees are summarized, and the mean and variance of the comprehensive influence coefficient of DC bias magnetic influence are calculated. The quantitative evaluation level is determined based on the preset threshold range in which the mean value falls. The reduction in Gini impurity or the feature importance score of each feature quantity in the random forest regression evaluation model is analyzed. Feature quantities with importance scores exceeding a set threshold are selected to generate the set of key influencing factors, which includes the weight of the subway operating density. The formulation of the DC bias suppression strategy includes: Based on the grounding operation mode of the target transformer, the grounding type is determined to be direct grounding, grounding through a small resistor, or grounding through a capacitor. If the grounding type is direct grounding or grounding through a small resistor, the withstand voltage rating and current capacity of the required series-connected silicon carbide PiN diodes are calculated based on the quantitative assessment level and the amplitude of the dynamic DC bias current data, generating the configuration parameters of the active suppression device based on the silicon carbide PiN diodes. If the grounding type is grounding through a capacitor, the capacitance optimization range of the capacitor is calculated based on the quantitative assessment level, so that the capacitor meets the AC system grounding safety requirements while isolating the DC component, generating the optimized grounding scheme. Combining the weight of the metro operating density in the set of key influencing factors, the dynamic switching logic of the active suppression device is generated, and the dynamic switching logic is associated with the peak periods of the metro timetable and the rail potential distribution data. The configuration parameters for generating the active suppression device based on the silicon carbide PiN diode include: Based on the peak and effective values ​​of the dynamic DC bias current data, the continuous current thermal effect parameters and instantaneous inrush current parameters that the active suppression device needs to withstand are calculated; based on the continuous current thermal effect parameters, the area requirements and heat sink specifications of the silicon carbide PiN diode chip are determined; based on the instantaneous inrush current parameters, the number of chips to be connected in series in the diode series stack is determined to meet the reverse blocking voltage requirements; based on the current carrying capacity of the transformer neutral point grounding conductor, the trigger current threshold of the bypass protection circuit connected in parallel with the diode stack is determined.

2. The method for suppressing DC bias in a transformer according to claim 1, characterized in that, The method further includes: Within a preset simulation period, the operation scenarios of subway locomotives under different timetables and different train formations are simulated; For each operating scenario, the dynamic DC bias current calculation model and the transformer DC bias evaluation model are invoked to calculate the corresponding dynamic DC bias current data simulation set and excitation current distortion data simulation set. The dynamic DC bias current data simulation set and the excitation current distortion data simulation set are input into the random forest regression evaluation model to generate a quantitative evaluation level distribution map under multiple scenarios. Based on the quantitative assessment level distribution map, key operating scenario modes that cause the DC bias magnetism level of the transformer to reach the warning threshold are identified.

3. The transformer DC bias suppression method according to claim 2, characterized in that, The method further includes: Establish the correlation and mapping relationship between the key operation scenario modes and subway operation scheduling parameters; Based on the aforementioned correlation mapping relationship, auxiliary scheduling suggestions for power grid security are generated, including suggestions for adjusting the subway departure density and section operating speed. The auxiliary scheduling suggestions are coordinated and optimized with the dynamic switching logic of the active suppression device based on silicon carbide PiN diodes to form a comprehensive governance scheme covering source-side regulation and receiver-side suppression.

4. The transformer DC bias suppression method according to claim 3, characterized in that, The method further includes: A DC current monitoring device was installed at the neutral point of the target transformer to collect the actual measured data sequence of DC bias current during operation. The measured DC bias current data sequence is compared and analyzed with the dynamic DC bias current data calculated by the transformer DC bias evaluation model within the corresponding time period, and the model prediction error rate is calculated. Based on the model prediction error rate, an adaptive filtering algorithm is used to perform reverse correction on the soil resistivity distribution parameters in the four-layer ground grid backflow structure model. The dynamic DC bias current calculation model is updated using the corrected soil resistivity distribution parameters.

5. The method for suppressing DC bias in a transformer according to claim 1, characterized in that, The set of subway operation parameters for the target area includes the real-time location, traction current, operating speed, and train formation information of subway locomotives; the set of AC power grid structure parameters includes the geographical location of substations, transformer grounding methods and grounding resistance values, and DC resistance parameters of overhead power lines and cables; the set of soil geological parameters includes soil layering structure and resistivity data obtained through geological exploration.

6. A transformer DC bias suppression system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the transformer DC bias suppression method as described in any one of claims 1 to 5.

Images