Rectifier transformer cooperative voltage regulation management method and rectifier transformer for aluminum foil formation

Abstract

The application provides a rectifier transformer cooperative voltage regulation management method and a rectifier transformer for aluminum foil formation, relates to the fields of SCADA distributed power grid control and power conversion technology, and the method acquires multi-modal slow change characteristic data to perform dissipation evaluation calculation, generates a global fatigue damping coefficient representing cumulative attenuation; meanwhile, based on photovoltaic side electrical transient fluctuation and node load capacity, local transient voltage compensation scalar is calculated; the two are input to the joint mapping logic to output a dynamic compensation weight matrix, and the transient cooperative control instruction for indicating the bottom converter to execute trigger angle bias is generated according to the matrix; and the damping attenuation degradation reconstruction strategy triggered by communication timeout is configured. The application realizes accurate voltage stabilization and group control load balancing in AC / DC conversion network, effectively overcomes the limitation of traditional single-point anti-disturbance which does not take into account the physical attenuation of equipment, and ensures the absolute stability and high robustness of DC power supply under the condition of input power fluctuation.

Description

Technical Field

[0001] This invention relates to the field of SCADA distributed power grid control and power conversion technology, specifically to a method for coordinated voltage regulation and control of rectifier transformers and a rectifier transformer for aluminum foil forming. Background Technology

[0002] With the evolution of new energy grid connection technologies, AC / DC conversion and power supply systems (under the SCADA distributed grid control architecture) are facing a new technological environment with drastic fluctuations in input power. In scenarios where "direct photovoltaic power" drives highly sensitive dynamic loads (such as aluminum foil formation production lines), how to smooth high-frequency electrical fluctuations and ensure the ultimate stability of DC output power has become a technical challenge that urgently needs to be addressed in the field of power conversion and regulation.

[0003] To address the aforementioned input disturbances, existing technologies have explored certain approaches. For example, patent document CN121193110A proposes a composite disturbance rejection and voltage regulation control method for the rectifier stage of a solid-state transformer. This method constructs a discretized full-order voltage model and uses an observer to estimate the total disturbance online, combining this with state feedback gain to form a composite disturbance rejection control law. However, such single-node disturbance rejection strategies still have significant limitations when facing clustered applications: First, their control domain is limited to a single rectifier stage, lacking a transient mutual assistance and group control load balancing mechanism among multiple transformer nodes. When dealing with spatially uneven photovoltaic transient drops, the overall collaborative gain needs improvement. Second, existing control laws highly focus on tracking errors of fast electrical variables, failing to incorporate the slow physical evolution within the transformer (e.g., hotspot temperature distribution and mechanical wear) into the control boundary. This leads to an easy acceleration of the physical structure degradation of the equipment during extreme transient compensation, lacking a spatiotemporal decoupling mechanism between electrical regulation and hardware lifespan. Summary of the Invention

[0004] The purpose of this invention is to provide a method for coordinated voltage regulation and control of rectifier transformers and a rectifier transformer for aluminum foil forming, constructing a joint mapping topology with physical dissipation assessment as the damping boundary and local transient margin as the compensation driver. By abstracting slowly varying physical states into negative exponential constraints of control weights, and introducing an autonomous topology reconstruction strategy when underlying communication is limited, a technological paradigm shift from passive single-point disturbance rejection to distributed intelligent cooperative decision-making is achieved. This addresses the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] The method for coordinated voltage regulation and control of rectifier transformers includes the following specific steps:

[0007] S1: Obtain multimodal slow-varying characteristic data that characterizes the physical evolution state of each transformer node in the multi-rectifier transformer cluster, and perform dissipation evaluation calculation based on the multimodal slow-varying characteristic data to generate a global fatigue damping coefficient. The global fatigue damping coefficient characterizes the cumulative physical structure attenuation of the transformer node.

[0008] S2: Obtain electrical transient fluctuation data characterizing the high-frequency fluctuation state of the photovoltaic power input terminal; based on the electrical transient fluctuation data and the current load capacity of each transformer node, calculate the local transient voltage compensation scalar, the local transient voltage compensation scalar characterizing the local energy margin of a single transformer node in response to the current voltage drop;

[0009] S3: Taking the global fatigue damping coefficient and the local transient voltage compensation scalar as input, the negative exponential mapping and scalar multiplication operation are performed through the joint compensation mapping logic to output a dynamic compensation weight matrix. The dynamic compensation weight matrix represents the distribution ratio of the transient mutual assistance current within the multi-rectifier transformer cluster.

[0010] S4: Based on the dynamic compensation weight matrix, generate transient cooperative control commands to instruct the underlying converter to perform trigger angle biasing actions;

[0011] S5: Output the transient collaborative control command, and when the data update cycle of the global fatigue damping coefficient exceeds the preset communication time threshold, trigger the damping attenuation degradation strategy, reduce the global fatigue damping coefficient according to the preset attenuation function and reconstruct the dynamic compensation weight matrix.

[0012] A rectifier transformer for aluminum foil forming includes a transformer body and an edge computing gateway communicatively connected to the transformer body; the edge computing gateway contains a memory and a microprocessor, and the memory stores a computer program; when the microprocessor executes the computer program, it implements the rectifier transformer collaborative voltage regulation and control method.

[0013] Compared with the prior art, the beneficial effects of the present invention are:

[0014] This invention extracts local microscopic dissipation scalars by acquiring multimodal slowly varying characteristic data (hotspot temperature and mechanical switching frequency) and generates a global fatigue damping coefficient characterizing the degree of physical attenuation, which is then used as the core constraint input of the joint compensation mapping logic. Mechanistically, this treats slowly varying physical fatigue as a rigid boundary for the distribution of fast-changing transient currents, effectively preventing overload thermal breakdown or mechanical damage at high-fatigue nodes during ultimate voltage regulation.

[0015] By calculating the absolute difference between the negative voltage gradient of the photovoltaic side bus and the current load capacity of the node in real time, a local transient voltage compensation scalar is extracted. Combined with the fatigue damping coefficient, a dynamic compensation weight matrix is ​​output, which in turn generates the firing angle bias control command for the underlying converter. A distributed differential gradient algorithm is used to achieve optimal dynamic allocation of transient mutual assistance current within a multi-rectifier transformer cluster, enhancing the response sensitivity of the group-controlled load and the DC bus voltage regulation accuracy when dealing with extremely high-frequency and severe fluctuations at the input.

[0016] A damping attenuation degradation strategy is introduced. When the update cycle of the global fatigue damping coefficient data exceeds a preset communication time threshold, a local state overwrite process is automatically activated. This mechanism successfully mitigates the risk of paralysis in distributed networks that heavily rely on SCADA backbone communication when faced with physical link failures. By autonomously and continuously reducing fatigue damping according to a preset exponential time constant and reconstructing the backup topology, it ensures a smooth transition of control to the local voltage compensation scalar distribution, improving the bottom-line supply capability and overall robustness of this method under harsh communication environments. Attached Figure Description

[0017] Figure 1 This is a schematic diagram illustrating the core technical route and execution object of the rectifier transformer coordinated voltage regulation and control method of the present invention;

[0018] Figure 2 A schematic diagram of the technical route for generating the dynamic compensation weight matrix for steps S1 to S3;

[0019] Figure 3 A schematic diagram of the technical route for generating transient cooperative control commands and damping attenuation degradation strategies for steps S4 to S5.

[0020] Figure 4 This is a schematic diagram showing the numerical verification results of the transient collaborative control algorithm under multiple operating conditions. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0022] It is understood that the terms “first,” “second,” etc., used in this application may be used herein to describe various elements, but unless otherwise stated, these elements are not limited by these terms. These terms are used only to distinguish one element from another.

[0023] Example 1:

[0024] Please see Figures 1 to 4 The present invention provides a technical solution:

[0025] A method for coordinated voltage regulation and control of rectifier transformers, executed by a coordinated voltage regulation and control system, includes the following steps:

[0026] S1: Obtain multimodal slow-varying characteristic data that characterizes the physical evolution state of each transformer node in the multi-rectifier transformer cluster, and perform dissipation evaluation calculation based on the multimodal slow-varying characteristic data to generate a global fatigue damping coefficient. The global fatigue damping coefficient characterizes the cumulative physical structure attenuation of the transformer node.

[0027] S2: Obtain electrical transient fluctuation data characterizing the high-frequency fluctuation state of the photovoltaic power input terminal; based on the electrical transient fluctuation data and the current load capacity of each transformer node, calculate the local transient voltage compensation scalar, the local transient voltage compensation scalar characterizing the local energy margin of a single transformer node in response to the current voltage drop;

[0028] S3: Taking the global fatigue damping coefficient and the local transient voltage compensation scalar as input, the negative exponential mapping and scalar multiplication operation are performed through the joint compensation mapping logic to output a dynamic compensation weight matrix. The dynamic compensation weight matrix represents the distribution ratio of the transient mutual assistance current within the multi-rectifier transformer cluster.

[0029] S4: Based on the dynamic compensation weight matrix, generate transient cooperative control commands to instruct the underlying converter to perform trigger angle biasing actions;

[0030] S5: Output the transient collaborative control command, and when the data update cycle of the global fatigue damping coefficient exceeds the preset communication time threshold, trigger the damping attenuation degradation strategy, reduce the global fatigue damping coefficient according to the preset attenuation function and reconstruct the dynamic compensation weight matrix.

[0031] For step S1, the multimodal slow-changing characteristic data includes a first state sub-feature characterizing the hot spot temperature distribution of the transformer winding and a second state sub-feature characterizing the mechanical switching frequency of the tap changer.

[0032] Based on the aforementioned multimodal slowly varying characteristic data, a dissipation assessment calculation is performed to generate a global fatigue damping coefficient, including:

[0033] Based on a preset time integration window, a weighted integral calculation is performed on the first state sub-feature and the second state sub-feature to extract local microscopic dissipation scalars;

[0034] The local microscopic dissipation scalar is input into the Sigmoid boundary constraint function to perform nonlinear normalization mapping processing, and the global fatigue damping coefficient is output.

[0035] Further specifying, the weighted integral calculation of the first state sub-feature and the second state sub-feature is performed based on a preset time integration window to extract local microscopic dissipation scalars, specifically including the following steps:

[0036] Within the preset time integration window, environmental background run parameters characterizing the baseline of ambient temperature fluctuations are obtained;

[0037] Obtain the preset initial thermal stress conversion coefficient and initial mechanical equivalent normalization factor;

[0038] Based on the environmental background run parameters, a nonlinear compensation calculation is performed on the initial thermal stress conversion coefficient to generate a dynamic thermal stress weighting factor.

[0039] Using the dynamic thermal stress weighting factor and the initial mechanical equivalent normalization factor, dimensional unification multiplication operations are performed on the first state sub-feature and the second state feature, respectively;

[0040] Obtain the corresponding discrete sampling time step parameter;

[0041] The first state feature within the preset time integration window, after undergoing dimensional unification multiplication, is algebraically accumulated, and the accumulation result is multiplied by the discrete sampling time step parameter to obtain the heat dissipation integral.

[0042] The mechanical wear amount is obtained by performing algebraic accumulation on the second state sub-feature within the preset time integration window after the dimension-unified multiplication operation;

[0043] The integral of heat dissipation is added to the mechanical wear amount to extract the local microscopic dissipation scalar.

[0044] Further specifying, the nonlinear compensation calculation of the initial thermal stress conversion coefficient based on the environmental background run parameters is performed to generate a dynamic thermal stress weighting factor, specifically including the following steps:

[0045] Calculate the absolute deviation scalar between the environmental background run parameters and the preset standard reference environmental temperature threshold;

[0046] Determine whether the absolute deviation scalar is greater than a preset drift tolerance threshold;

[0047] If the judgment result is greater than the drift tolerance threshold, a preset exponential amplification base is obtained, the absolute deviation scalar is used as the exponent, and the exponential operation is performed on the exponential amplification base to obtain the nonlinear compensation term.

[0048] The nonlinear compensation term is multiplied by the initial thermal stress conversion coefficient to output the dynamic thermal stress weighting factor.

[0049] For step S2, the electrical transient fluctuation data includes the DC bus voltage transient drop amplitude and the converter output current change rate.

[0050] Based on the electrical transient fluctuation data and the current load capacity of each transformer node, a local transient voltage compensation scalar is calculated, including:

[0051] The negative pressure gradient of the bus voltage is calculated based on the transient drop amplitude of the DC bus voltage and the rate of change of the converter output current.

[0052] Calculate the absolute difference between the current load capacity and the negative pressure gradient of the bus voltage, and extract the local transient voltage compensation scalar.

[0053] The local transient voltage compensation scalar is broadcast to adjacent transformer nodes in the physical topology via a single-hop horizontal communication link.

[0054] Further specifying step S3, the negative exponential mapping and scalar multiplication operations are performed through the joint compensation mapping logic to output a dynamic compensation weight matrix, including:

[0055] Calculate the attenuation mapping value with the base of the natural logarithm as the base and the negative global fatigue damping coefficient as the exponent;

[0056] The effective compensation output value of the node is extracted by performing Hamiltonian scalar product operation on the local transient voltage compensation scalar and the attenuation mapping value.

[0057] Calculate the differential gradient matrix of the effective compensation output value of the transformer node between the current transformer node and the adjacent transformer nodes in the topology.

[0058] Further limiting the difference gradient matrix in step S3, after extracting the difference gradient matrix, a distributed safety reset judgment logic is executed:

[0059] Obtain the preset full-network minimum potential energy conservation threshold; obtain the rated apparent capacity of this transformer node and the total installed capacity of the entire network, calculate the ratio of the rated apparent capacity to the total installed capacity of the entire network, and obtain the capacity weight ratio coefficient;

[0060] Multiply the network-wide minimum potential energy conservation threshold by the capacity weight ratio coefficient to obtain the node-level equivalent potential energy quota.

[0061] The sum of all positive pressure difference values ​​in the differential gradient matrix is ​​calculated to obtain the total local leakage potential energy.

[0062] If the total local leakage potential energy is less than the node-level equivalent potential energy quota, the absolute value of the local global fatigue damping coefficient is reduced by a preset reduction ratio, and the total local leakage potential energy is recalculated iteratively.

[0063] If the sum of the local leakage potential energy is greater than or equal to the node-level equivalent potential energy quota, the positive gradient components with values ​​greater than zero in the differential gradient matrix are extracted to generate the dynamic compensation weight matrix.

[0064] The following is a detailed explanation of the implementation of steps S1 to S3 above:

[0065] The initial thermal stress conversion coefficient is defined to characterize the benchmark mapping value of the expected fatigue degree caused by unit continuous thermal stress on the internal insulation material of the transformer under standard operating conditions.

[0066] The initial mechanical equivalent normalization factor is defined to characterize the wear equivalent value of the physical structure caused by the frequency of mechanical actions of a unit discrete circuit breaker, with the aim of mapping discrete count values ​​to dimensionless parameters.

[0067] Explanation of the determination of the initial thermal stress conversion coefficient and the initial mechanical equivalent normalization factor:

[0068] A controlled experimental baseline environment (constant 25 degrees Celsius and 50% relative humidity) was established. A constant thermal gradient was applied continuously to multiple sets of transformer insulation samples, and the polymerization degree reduction rate was simultaneously measured using a chromatograph. A constant frequency of mechanical opening and closing was applied to multiple sets of contact switch samples, and the physical wear rate of the contact surface was measured using a contact resistance meter. For the above two sets of destructive experimental data, a least squares algorithm was used to perform univariate linear regression fitting calculations, and the slope values ​​of the fitted response lines were extracted. The slope value of the fitted line between the polymerization degree reduction rate and the continuous thermal gradient was directly calibrated as the initial thermal stress conversion coefficient. The slope value of the fitted line between the physical wear rate of the contact surface and the mechanical opening and closing frequency was divided by a preset baseline wear constant and dimensionlessly processed to calibrate as the initial mechanical equivalent normalization factor. The initial thermal stress conversion coefficient and the initial mechanical equivalent normalization factor obtained after the above calibration were pre-written and stored in specific coordinate positions in the external spreadsheet file. During the initialization phase of the control flow, the external structured spreadsheet file stored locally is parsed; the values ​​at specific coordinates in the file are read through the preset address pointer; the extracted values ​​are loaded into memory variables in sequence, and the initial thermal stress conversion coefficient and the initial mechanical equivalent normalization factor are generated respectively.

[0069] In this embodiment, the preset reference wear constant value depends on the reference material properties and rated voltage level of the transformer tap changer contacts. When the physical wear rate is characterized by the increase in contact resistance, the preset reference wear constant is typically in the range of 1.0 × 10⁻⁶. -4 micro-ohms / time up to 5.0×10 -3 Between microohms per cycle; in this embodiment, the reference wear constant is exemplarily taken as 1.5 × 10⁻⁶. -3Microohms / cycle. It should be noted that the reference wear constant is not limited to the range of the above examples and should be adjusted adaptively according to the actual transformer tap changer contacts.

[0070] The environmental background run parameter is defined to characterize the continuous evolution baseline and deviation trend of the external ambient temperature of the transformer equipment within a preset time integration window. It is obtained by performing low-pass filtering and moving average extraction on the external ambient temperature sampling sequence within the current time integration window. The logic for determining the environmental background run parameter is explained in detail below:

[0071] Extract all external environmental dry-bulb temperature sampling data within the current preset time integration window to construct an original environmental temperature sequence set. Traverse the timestamp sequence of the original environmental temperature sequence set to determine if there are any missing discrete sampling values ​​(including time-series holes caused by sensor disconnection or communication packet loss). In response to the determination of missing values, execute the time-series boundary anomaly handling logic: determine the relative position of the missing value on the time axis of the original environmental temperature sequence set; if the missing value is at the beginning of the sequence, directly extract the immediately preceding valid sampling value as interpolation data; if the missing value is at the end of the sequence, directly extract the immediately preceding valid sampling value as interpolation data; if the missing value is in the middle of the sequence, extract the immediately preceding and following historical valid sampling values ​​on the time axis, perform an arithmetic mean calculation on these two values ​​to obtain the interpolation data; fill the corresponding hole position in the original environmental temperature sequence set with this interpolation data. Perform algebraic summation on all valid sampling values ​​in the original environmental temperature sequence set after self-checking and filling processing to obtain the total temperature sum of the window. The total number of valid sampled values ​​in the original ambient temperature sequence set is extracted and used as the divisor. The total temperature in the window is divided by the divisor, and the arithmetic mean is extracted. This arithmetic mean is used as the environmental background run parameter characterizing the baseline of ambient temperature fluctuations. This ensures that, even in harsh communication environments, the macroscopic low-frequency environmental baseline affecting the heat dissipation of the transformer's underlying structure can be isolated.

[0072] The first state sub-feature is defined to characterize the continuous gradient data of the temperature distribution of hot spots in the transformer windings. The determination of the first state sub-feature is explained as follows: the original oil temperature analog signal sequence is acquired through a preset fiber optic temperature measurement data interface. The external temperature sensing module providing the analog signal sequence has a stable sampling frequency of no less than 10Hz and a measurement error boundary of no more than ±0.1 degrees Celsius; this is the physical basis for ensuring the accurate capture of minute oil temperature gradients.

[0073] A preset digital low-pass filtering algorithm is applied to the original oil temperature analog signal sequence. Specifically, a fourth-order Butterworth digital filter logic with a cutoff frequency set to 1Hz is used to perform the filtering calculation, removing high-frequency noise caused by the strong alternating electromagnetic field inside the transformer, and outputting a smooth digital oil temperature sequence. Within a preset time integration window, extreme value optimization calculation is performed on the digital oil temperature sequence. All discrete temperature values ​​within the transformer tap changer contacts are iterated to extract the highest temperature extreme value; then all discrete temperature values ​​are iterated again to extract the lowest temperature extreme value; a subtraction operation is performed between the highest and lowest temperature extreme values ​​to obtain the absolute temperature difference scalar. The absolute temperature difference scalar is used as the first state sub-feature characterizing the hot spot temperature distribution of the transformer windings. This signal pipeline abstracts the noisy analog time-series waveform into a dimensionless early-stage feature quantity that purely characterizes the thermal gradient intensity.

[0074] The second state sub-feature is defined as a discrete count value representing the frequency of mechanical switching of the tap changer. The determination of the second state sub-feature is explained as follows: The level state transition signal uploaded from the auxiliary contacts of the underlying circuit breaker is continuously monitored and extracted through the digital input interface channel. When a transition edge from low to high or from high to low is detected, the state anti-jitter discrimination calculation logic is activated. A preset anti-jitter delay timer is started. In this step, an anti-jitter delay threshold parameter is set. This parameter is used to shield the transient rebound of the mechanical spring; in this embodiment, its value is set to 20 milliseconds. This covers the longest contact physical bounce time period of a conventional high-capacity vacuum circuit breaker, achieving a balance between "avoiding missed effective opening and closing" and "eliminating spurious counts." When the anti-jitter delay timer reaches 20 milliseconds, the current level state of the digital input interface channel is reread; it is determined whether the current level state is consistent with the level state that caused the transition edge; if they are consistent, a valid mechanical action pulse is output. Within a preset time integration window, an algebraic summation operation is performed on all valid mechanical action pulses output, and the absolute count value is extracted and used as the second state sub-feature characterizing the mechanical switching frequency of the tap changer.

[0075] The global fatigue damping coefficient is denoted as The cumulative physical structural decay of the transformer nodes is characterized as the physical lifespan constraint boundary for microsecond-level control. The calculation logic for the global fatigue damping coefficient is explained as follows: The local microscopic dissipation scalar, obtained after dimensional unification and algebraic summation, is extracted. A preset translation offset constant is obtained. The translation offset constant forces the initial microscopic dissipation baseline of physical health to align to the zero point of the coordinate system's central axis. The translation offset constant is stored in the device's local configuration file; during the initialization phase, the configuration file is parsed and the parameter is extracted into the running memory by executing the file reading process. In this embodiment, the preferred value of the translation offset constant is set to 0.5. It empirically corresponds to the basic fatigue dispersion of standard transformer insulation material in the middle of its normal lifespan; the local microscopic dissipation scalar is subtracted from the translation offset constant to obtain the offset characteristic value. A preset gain scaling constant is obtained. The gain scaling constant controls the steepness of the mapping curve, representing the method's sensitivity to fatigue degradation protection. This constant is retrieved and loaded from the configuration file. In this embodiment, the preferred value is set to 10.0. This ensures that when the physical dissipation magnitude slightly exceeds the aforementioned offset axis, the output global fatigue damping coefficient can quickly approach the value of one, thereby forming isolation protection. The offset characteristic value is multiplied by the gain scaling constant to obtain the scaling value. Specifically, the natural logarithm base is obtained; the negative of the scaling value (with a negative sign) is used as the mathematical exponent, and the natural logarithm base is exponentially calculated to extract the attenuation denominator factor. The attenuation denominator factor is added to the constant value one to obtain the global damping denominator value. The constant value one is used as the numerator, and it is divided by the global damping denominator value to perform a division calculation and output the quotient value. This quotient value is used as the global fatigue damping coefficient. This nonlinearly maps any physical dissipation value to the real number domain greater than zero and less than one.

[0076] The local transient voltage compensation scalar, whose parameter symbol is: This characterizes the local energy output margin of a single node in response to a current bus voltage dip. Explanation of the local transient voltage compensation scalar and dynamic compensation weight matrix:

[0077] Determine the local transient voltage compensation scalar: Obtain the transient sag amplitude of the DC bus voltage and the rate of change of the converter output current; extract a preset zero-dead-zone threshold (set to 0.001 amperes / second), and determine whether the absolute value of the rate of change of the converter output current is less than the zero-dead-zone threshold to execute the division-by-zero anomaly protection logic; in response to the judgment result that the absolute value is less than the zero-dead-zone threshold (indicating that the current is in steady state), extract a preset steady-state minimum gradient constant, and use it directly as the negative pressure gradient of the bus voltage; wherein, the physical definition of the steady-state minimum gradient constant is a non-zero reference lower limit value artificially assigned to meet the needs of control matrix operation when the coordinated voltage regulation and control system is in steady state, and its physical dimension is consistent with the pressure gradient, volts per second (V / s). In this embodiment, the preferred range of the steady-state minimum gradient constant is 0.001V / s to 0.05V / s, and the preferred value is 0.01V / s. Its value is much smaller than the transient gradient caused by the actual voltage drop, ensuring that the data stream is not interrupted by division by zero or dead zone, and without substantially interfering with the weight allocation during normal steady-state operation. In response to the judgment result that the absolute value is greater than or equal to the zero-point dead zone threshold, the transient voltage drop amplitude of the DC bus voltage is divided by the rate of change of the converter output current, and a division calculation is performed to extract the negative pressure gradient of the bus voltage within the mathematical boundary. The current load capacity value is obtained. At this point, it is determined that the physical dimension of the current load capacity value (kilovolt-ampere or ampere) is heterogeneous with the physical dimension of the aforementioned negative pressure gradient of the bus voltage (volt-second), and a preset capacity-gradient dimension conversion coefficient is obtained. This conversion coefficient is used to eliminate the heterogeneous dimension difference, and its physical meaning is defined as: linearly mapping the dynamic pressure gradient to an equivalent apparent capacity decay value. The negative pressure gradient of the bus voltage is multiplied by the capacity-gradient dimension conversion coefficient, and the equivalent capacity pressure loss value is output. At this point, its dimension has been uniformly converted to the capacity unit consistent with the current load capacity. The current load capacity value is subtracted from the equivalent capacity compression loss value. The difference is then used to extract the local transient voltage compensation scalar.

[0078] Construct a dynamic compensation weight matrix: Obtain the natural logarithm base, and perform a power operation on it with the negative global fatigue damping coefficient as the exponent to extract the attenuation mapping value; perform scalar multiplication calculation on the local transient voltage compensation scalar and the attenuation mapping value to extract the effective compensation output value of the node; based on the broadcast network, obtain the effective compensation output value of all adjacent nodes in the physical topology; calculate the algebraic difference between the output value of this transformer node and the output value of each adjacent node, and generate a differential gradient matrix containing multiple pressure difference values; perform a Boolean judgment of greater than zero on all values ​​in the differential gradient matrix; remove the dead zone gradient with a judgment result of false (value less than or equal to zero), and only extract the gradient components with a judgment result of true (representing positive pressure difference with external mutual assistance capability), and combine them to construct the initial effective gradient vector. To ensure that the transient mutual current distribution ratio is 100% in the physical topology, a weight normalization calculation is performed on the initial effective gradient vector: all positive pressure difference values ​​in the initial effective gradient vector are summed to extract the scalar of the total regional pressure difference; each individual positive pressure difference value in the initial effective gradient vector is divided by the scalar of the total regional pressure difference in turn, and a division calculation is performed to obtain a series of dimensionless weight ratio coefficients between zero and one, whose algebraic sum is always equal to one.

[0079] Furthermore, when generating the dimensionless weighting coefficients, the following Kirchhoff conservation normalization equation is followed to accurately distribute the compensation current among the discrete network nodes: ;in, Defined as this transformer node To topologically adjacent receiving nodes The dimensionless weighting ratio coefficient for allocating transient mutual assistance current is defined as the real number interval [0,1]. Defined as this transformer node With receiving node The positive pressure difference value greater than zero is calculated and extracted (characterizing the initial effective gradient component), with the dimension of kilovolt-ampere (kVA); Defined as an algebraic summation operator; Defined in the physical-communication space mapping table, with this transformer node The set of all physically adjacent nodes that have valid DC electrical connections; Defined as a set The traversal iterative index in the table represents any adjacent topological node that satisfies the positive pressure difference condition; Defined as the calculated scalar sum of regional pressure differences, characterizing the transformer node. The total mutual energy potential difference radiating outwards is preferably in the range of [10, 1000] kVA in this embodiment, with a preferred value of 350 kVA. In this embodiment, a series of dimensionless weighting coefficients are arranged in a structured manner according to the communication addressing sequence (MAC address sequence) of the underlying converter physical topology nodes, generating a dynamic compensation weight matrix with the mapping relationship of this transformer node. This dynamic compensation weight matrix directly forces the transient compensation large current to be optimally shunted along the path of the healthiest node with the least damping.

[0080] Furthermore, this method establishes a heterogeneous dimensional dynamic alignment and weight compensation mechanism based on environmental background run length. For step S1, dynamic weighting and dissipation dimensionality reduction of multimodal slowly varying features are performed: specifically, multimodal slowly varying feature data acquisition and dissipation evaluation calculation are performed. Within a preset time integration window, the first state sub-feature characterizing the hot spot temperature distribution of the transformer winding and the second state sub-feature characterizing the mechanical switching frequency of the tap changer are obtained.

[0081] In this embodiment, the preset time integration window is defined as the absolute time span boundary used by the control system to collect and evaluate the long-term macroscopic physical dissipation evolution state of the equipment; in this embodiment, its preferred range is limited to [1, 24] hours, and the preferred value is 4 hours.

[0082] The dynamic weight compensation process involves: acquiring the environmental background run parameter characterizing the baseline of ambient temperature fluctuations; calculating the absolute deviation scalar between this environmental background run parameter and the standard reference ambient temperature threshold; and before performing the calculation of the absolute deviation scalar, calling the system constant loading instruction to pre-read the standard reference ambient temperature threshold. This parameter is defined as the ideal external ambient dry-bulb temperature reference point, marked on the transformer's nameplate, used to measure the normal temperature rise and nominal aging rate of the internal insulation material. In this embodiment, its preferred range is limited to [20, 25] degrees Celsius, with a preferred absolute value of 25 degrees Celsius. This avoids logical collapse and arbitrary offset caused by the environmental background run parameter losing its "origin reference" during drift calculation.

[0083] A preset drift tolerance threshold is obtained. This threshold is retrieved from the device profile dataset on the cloud server via a network request. In this embodiment, the preferred value for this drift tolerance threshold is set to 5.0 degrees Celsius. This value precisely covers the safety fluctuation boundary of the transformer tank casing affected by the natural temperature difference between day and night, avoiding overcompensation.

[0084] Determine whether the absolute deviation scalar is greater than the drift tolerance threshold; if the determination result indicates that the absolute deviation scalar is greater than the drift tolerance threshold, obtain a preset exponential amplification base. This exponential amplification base is asynchronously downloaded and extracted from the cloud server. In this embodiment, the preferred value of the exponential amplification base is set to the base of the natural logarithm, approximately 2.718. Ensure that the thermal compensation penalty term exhibits a continuous and non-abrupt growth trend after crossing the dead zone; using the absolute deviation scalar as the mathematical exponent, perform an exponentiation operation on the exponential amplification base to extract the nonlinear compensation term.

[0085] The initial thermal stress conversion coefficients are read from a preset external spreadsheet file; the nonlinear compensation term is multiplied by the initial thermal stress conversion coefficients to output the dynamic thermal stress weighting factor.

[0086] In this preferred embodiment, based on the offline physical fatigue wear characteristics of the mechanical contacts of a standard vacuum circuit breaker, the specific value of the pre-stored initial mechanical equivalent normalization factor is set to 0.005. A multiplication operation is performed on the first state sub-feature using a dynamic thermal stress weighting factor, and a multiplication operation is performed on the second state sub-feature using the aforementioned instantiated initial mechanical equivalent normalization factor. This operation eliminates the dimensional difference between "continuous thermal stress (Celsius dimension)" and "discrete mechanical action (subdivisional dimension)," mapping them to a unified dimensionless value.

[0087] After performing dimensional unification multiplication of the corresponding features using dynamic thermal stress weighting factors and initial mechanical equivalent normalization factors, discrete-time integral calculations are performed to accurately extract dissipation:

[0088] A discrete sampling time step parameter corresponding to the first and second state sub-features during acquisition is extracted; in this embodiment, it is set to 0.1 seconds. An algebraic summation operation is performed on all first state sub-feature values ​​within a preset time integration window that have undergone dimension unification processing to obtain the sum of thermal stress features. The sum of thermal stress features is then multiplied by the discrete sampling time step parameter to realize the discrete Riemann integration process of continuous thermal stress, thereby extracting the absolute heat dissipation integral.

[0089] Specifically, when the discrete Riemann integration process is executed by the microprocessor built into the edge computing gateway, the absolute heat dissipation integral is extracted according to the following calculation logic: ;

[0090] in, Defined as the absolute heat dissipation integral within a preset time integration window, with dimensions being a unified dimensionless scalar. This represents the summation mathematical operator performed on all discrete sampling points k within a preset time integration window; N is defined as the total number of discrete sampling points contained within the preset time integration window, and in this embodiment, the preferred range is [100, 10000], with a preferred value of 1000; Defined as the time step index number of the discrete-time sampling point, with a value range of positive integers from 1 to N; Defined as the dynamic thermal stress weighting factor extracted at the k-th sampling time; Defined as the first state sub-feature characterizing the hot spot temperature distribution of the transformer winding, acquired at the k-th sampling time; Defined as the discrete sampling time step parameter, it represents the absolute physical time interval between two consecutive feature acquisition actions, with the unit being seconds (s). In this embodiment, the preferred range is [0.01, 1.0] seconds, and the preferred value is 0.1 seconds.

[0091] For the second state sub-feature belonging to the discrete pulse attribute, the result after unifying its dimensions is directly algebraically accumulated within the window to obtain the absolute mechanical wear amount.

[0092] The absolute heat dissipation integral and the absolute mechanical wear are added together to extract a local microscopic dissipation scalar with absolute physical meaning, while eliminating hardware sampling frequency interference. A sigmoid boundary constraint function is then called to perform nonlinear mapping, forcing the calculation results to converge to a real number interval greater than zero and less than one, outputting the global fatigue damping coefficient.

[0093] Furthermore, when the Sigmoid boundary constraint function is invoked, the processing is specifically based on the following rigorous exponential mathematical mapping equation: ;in, Defined as the global fatigue damping coefficient, it is used as the physical life constraint boundary for subsequent control, and its value range is strictly defined as a dimensionless real number in (0,1). Defined as the local microscopic dissipation scalar input to this function after dimension alignment; e is the base of the natural logarithm, approximately 2.718; G is defined as the gain scaling constant, characterizing the steepness and sensitivity of the device degradation perception curve, and in this embodiment, the preferred range is [5.0, 20.0], with a preferred value of 10.0; Defined as a translation bias constant, it represents the critical dissipation reference coordinate point at which a substantial transition occurs between the physical health state and the deterioration state. In this embodiment, the preferred range is [0.1, 1.0], and the preferred value is 0.5. This embodiment is not limited to the specific values ​​mentioned above. In practical applications, the above gain and bias constants can be flexibly configured according to the factory material aging reference calibration values ​​of a specific model of rectifier transformer.

[0094] For step S2, spatial dimensionality reduction and rapid broadcasting of high-frequency electrical transient fluctuations are performed: electrical transient fluctuation data characterizing the high-frequency fluctuation state of the photovoltaic power input terminal are acquired in the control flow, specifically extracting the DC bus voltage transient sag amplitude and the converter output current change rate; based on the acquired sag amplitude and current change rate, waveform slope analysis calculation is performed to extract the bus voltage negative pressure gradient characterizing the degree of process steady-state disruption caused by the current sag curve. In this embodiment, for the step of extracting the DC bus voltage transient sag amplitude, the following relative difference calculation is used to eliminate measurement ambiguity: the DC bus voltage sequence within a fixed historical period (5 power frequency cycles) before the current moment is extracted, and an average calculation is performed on it to extract the steady-state rated voltage reference; the instantaneous voltage sampling value under sudden sag conditions is acquired through a high-frequency analog-to-digital conversion interface; the steady-state rated voltage reference is subtracted from the instantaneous voltage sampling value, and a subtraction calculation is performed to extract the relative voltage difference scalar, and this relative voltage difference scalar is accurately marked as the DC bus voltage transient sag amplitude. The transient voltage drop amplitude of the DC bus is defined as the absolute depth of the instantaneous sampled voltage on the DC side deviating from its original dynamic steady-state reference at the moment of an extreme electrical event such as sudden photovoltaic shading. Its dimension is volts (V). In this embodiment, its preferred effective range depends on the ultimate withstand voltage difference of the aluminum foil formation process, and in this embodiment, it is [10, 150]. This embodiment is not limited to the above-mentioned steady-state reference extraction method based on dynamic moving average. In a pure grid-connected rigid DC system, the static nominal process voltage is directly used as the subtraction reference. This avoids the ambiguity of misinterpreting the absolute voltage reading collected by the ADC as the "degree of drop," thereby providing a feature input source that purely characterizes the "severity of the fault" for the subsequent spatiotemporal calculation of the compression gradient.

[0095] Obtain the current load capacity of each transformer node device; perform subtraction calculation, subtract the bus voltage negative pressure gradient and the remaining capacity safety threshold from the current load capacity, and extract the single-dimensional local transient voltage compensation scalar.

[0096] Before performing the subtraction calculation, this embodiment extracts and loads the remaining capacity safety threshold from the device configuration data table via a local interface. The remaining capacity safety threshold is defined as the underlying apparent power buffer baseline that the transformer node must absolutely prohibit from outputting external support when facing high-frequency voltage drops in order to maintain its basic physical heat dissipation and prevent the internal insulation structure from being broken down. In this embodiment, its preferred range is strictly limited to [10%, 30%] of the current load capacity rating, with a preferred value of 15% of the current load capacity rating. This embodiment is not limited to the above specific percentage values. In practical applications, this parameter is customized and fine-tuned according to the overload tolerance curve corresponding to the cooling method of a specific rectifier transformer. This avoids cascading overload burnout of healthy nodes due to "uncontrolled output" under extreme transient mutual assistance conditions, and establishes an absolute safety physical baseline for the micro-autonomy between physically adjacent nodes.

[0097] A subtraction calculation is performed, subtracting the bus voltage negative pressure gradient and the remaining capacity safety threshold from the current load capacity to extract a one-dimensional local transient voltage compensation scalar. After extracting the local transient voltage compensation scalar, spatial topology analysis logic is executed:

[0098] A preset physical-communication space mapping table is read from local memory. This mapping table stores the correspondence between the network layer communication identifiers (MAC addresses) of this transformer node and all surrounding nodes and the physical layer bus connection order (e.g., the upstream and downstream node numbers of physical cable connections). In a preferred embodiment, the physical-communication space mapping table read by the processor from local memory (e.g., non-volatile flash memory EEPROM) is structured as a two-dimensional static data array containing multiple record entries. Each record entry contains at least the following three associated configuration data fields:

[0099] Field 1 [Node Physical Sequence Number]: Defined as the absolute spatial order of the rectifier transformer cluster on the DC bus topology, the data type is configured as a 16-bit unsigned integer (uint16), and in this embodiment, the preferred value is a positive integer that increases in the order of process layout;

[0100] Field 2 [Network Addressing Identifier]: Defined as the hardware-level access address of the underlying communication network card of the corresponding node. The data type is configured as a 48-bit standard Media Access Control (MAC) address format, and the preferred value is a hexadecimal character array;

[0101] Field 3 [Electrical Connection Impedance Weight]: Defined as the equivalent physical ohmic impedance conversion ratio of the DC large-section cable between this transformer node and the corresponding record entry. The dimension is a dimensionless normalized value. In this embodiment, the preferred range is [0.01, 1.0], and the preferred value is 0.15. This embodiment is not limited to the specific data length or encoding format mentioned above. In actual field implementation, adaptive adjustments should be made according to the differences in the industrial Ethernet protocol used.

[0102] In the physical-communication space mapping table, the transformer node number that is directly adjacent to this transformer node in the physical layer bus connection sequence is retrieved, and the target network layer communication identifier corresponding to this transformer node number is extracted. Using the extracted target network layer communication identifier as the addressing filtering condition for broadcast frames, a message containing a compensation scalar is sent to the transformer node via a single-hop horizontal communication link.

[0103] For step S3, the joint compensation mapping logic is executed with the global fatigue damping coefficient and the local transient voltage compensation scalar as inputs. The mutual restraint potential energy of the sub-healthy device is forcibly truncated and shielded using exponential attenuation, specifically obtaining the base of the natural logarithm; the global fatigue damping coefficient is extracted and negativeed; using the negative global fatigue damping coefficient as the exponent, a power operation is performed on the base of the natural logarithm to extract the attenuation mapping value. The Hamiltonian scalar product operation is performed between the local transient voltage compensation scalar extracted in step S2 and the attenuation mapping value to extract the effective compensation output value of the node.

[0104] Specifically, when the edge gateway (based on a DSP architecture microprocessor) of the collaborative voltage regulation and control system performs calculations, the effective compensation output value of the node is extracted based on the following physical mapping logic: ;

[0105] in, Defined as this transformer node The effective compensation output value used for transient mutual assistance is equivalent in kilovolt-ampere (kVA) or per unit (pu); Defined as this transformer node The calculated local transient voltage compensation scalar has a preferred range of [0, 500] kVA in this embodiment, with a preferred value of 150 kVA; e is defined as the base of the natural logarithm, a constant, with an approximate value of 2.718; Defined as the global fatigue damping coefficient generated before the i-th local transformer node, it is a dimensionless real number in the interval (0,1). This embodiment is not limited to the specific values ​​mentioned above. In actual field implementation, the reference unit of the compensation scalar is linearly scaled according to the rated power level of the underlying converter.

[0106] Before generating the differential gradient matrix: receive messages from external nodes through the network interface and extract the message source address; input the source address into the physical-communication space mapping table for reverse verification; if the verification result indicates that the physical node corresponding to the source address is not adjacent to the transformer node on the bus, discard the message data directly; only perform Hamiltonian scalar product on the message data of physically adjacent nodes that pass the verification and then calculate the voltage difference between the effective compensation output value of the transformer node itself and the effective compensation output value of each topologically adjacent transformer node to generate the differential gradient matrix, thereby ensuring that the differential gradient matrix in the digital space can perfectly mirror the underlying physical electrical interconnection topology.

[0107] After extracting the differential gradient matrix, a distributed security reset decision logic is executed to solve the technical problem that the global state cannot be directly obtained under single-hop local communication.

[0108] The overall network baseline protection threshold is dynamically determined by the SCADA master station or local main control program through cross-domain baseline calculation logic. Specifically, it includes the following steps to extract critical process parameters: obtaining the total number of aluminum foil formation tanks currently in online active state, and reading the preset process polarization maintenance baseline parameters, which include the minimum maintenance DC voltage and minimum maintenance DC current required to prevent polarization collapse of the electrolyte in the individual formation tank; performing a multiplication operation on the minimum maintenance DC voltage and minimum maintenance DC current to obtain the baseline active power of a single tank; performing a multiplication operation on the baseline active power of a single tank and the total number of formation tanks in operation to extract the total limit baseline active power requirement of the entire network DC side;

[0109] To obtain the overall system power conversion efficiency and overall system rated power factor of the rectifier transformer cluster, in this embodiment, for an aluminum foil formation production line using a phase-controlled rectifier topology, the overall system power conversion efficiency is preferably configured between 0.92 and 0.96, with a typical value of 0.95; the overall system rated power factor is affected by the reference steady-state firing angle, and is preferably configured between 0.85 and 0.92, with a typical value of 0.90; the total minimum active power requirement is divided sequentially by the overall system power conversion efficiency and the overall system rated power factor, and continuous division is performed to extract the corresponding AC-side equivalent apparent power lower limit; the extracted AC-side equivalent apparent power lower limit is directly assigned as the whole network minimum potential energy conservation threshold, which is sent by the SCADA master station through a message and cached in the local memory of this transformer node, representing the total minimum effective apparent output baseline required by the entire transformer cluster to maintain the formation process without crashing.

[0110] Obtain the total network minimum potential energy conservation threshold that is pre-issued by the SCADA master station and cached locally; the total network minimum potential energy conservation threshold represents the minimum total effective output baseline required for the entire transformer cluster to maintain the formation process without crashing.

[0111] The rated apparent capacity of this transformer node is obtained, and the preset total installed capacity of the entire network is obtained by parsing the system file configuration table that has been synchronized to the local file system. The rated apparent capacity of this transformer node is divided by the total installed capacity of the entire network to extract the capacity weight ratio coefficient. The network-wide minimum potential energy conservation threshold is multiplied by the capacity weight ratio coefficient, thereby reducing the unknown global constraints of the entire network to a node-level equivalent potential energy quota that can be independently observed by the local node.

[0112] Furthermore, a summation operation is performed on all positive pressure difference values ​​in the locally generated differential gradient matrix to extract the total local leakage potential energy. The system performs a judgment calculation to determine whether the total local leakage potential energy is lower than the node-level equivalent potential energy quota. If the judgment result is lower than the node-level equivalent potential energy quota, it indicates that the current node and its local neighbors are too old and cannot bear the assigned responsibility of backup mutual assistance. In this case, the safety reset logic is triggered, and the absolute value of the local global fatigue damping coefficient is reduced by a preset reduction ratio. The system then performs a dead loop prevention judgment: it determines whether the cumulative number of iterations of the current safety reset has reached the preset maximum iteration protection threshold (in this embodiment, the maximum iteration protection threshold is preferably set to 10 times). In response to the judgment result that the maximum iteration protection threshold has not been reached, the total local leakage potential energy is recalculated iteratively. In response to the judgment result that the maximum iteration protection threshold has been reached but the equivalent potential energy quota has not been met, the iteration process is forcibly terminated, the dynamic compensation weight component generated by this transformer node is directly set to zero to implement physical isolation, and an abnormal hardware interrupt signal of local mutual assistance capacity depletion is triggered to the upper-layer bus. In this embodiment, the preset reduction ratio is physically defined as the forced relaxation range of the equipment life decay assessment in a single safety reset iteration, and its dimension is a dimensionless pure number. In this embodiment, the preferred range for the preset reduction ratio is 0.80 to 0.95, with a preferred value of 0.90. This ensures the smoothness of the iterative convergence process and avoids control output oscillations caused by excessively large single reduction amplitudes, thus guaranteeing that the effective output of local nodes meets the minimum requirements. If the determination result is greater than or equal to the equivalent potential energy quota at that node level, indicating that the equivalent conservation condition is met, then all values ​​less than or equal to zero are filtered out from the difference gradient matrix, and only the positive gradient components with values ​​greater than zero are extracted, thereby constructing a dynamic compensation weight matrix. In the face of sudden photovoltaic drops, this weight matrix directly forces the transient compensation large current to be optimally diverted along the healthy node path with the least damping.

[0113] When the photovoltaic direct power supply side encounters a sudden large-scale cloud obstruction, the transient voltage drop amplitude of the DC bus surges instantaneously. This method bypasses the vertical aggregation of the backbone communication network, directly calculates the local transient voltage compensation scalar at the edge side, and disseminates it to neighboring nodes via single-hop broadcast. At this time, operators based on the differential gradient matrix calculate healthy nodes with positive voltage differentials and generate a dynamic compensation weight matrix to complete the transient current feedback support before the power of the aluminum foil forming tank is exhausted.

[0114] Regarding step S4, based on the dynamic compensation weight matrix, a transient cooperative control command is generated to instruct the underlying converter to perform a trigger angle bias action, including:

[0115] Obtain the reference steady-state firing angle; extract the target weight components corresponding to the target underlying converter from the dynamic compensation weight matrix;

[0116] The target weight components are input into the trigger angle mapping function to calculate the phase angle offset compensation amount;

[0117] The transient target firing angle is extracted by subtracting the phase angle offset compensation from the reference steady-state firing angle.

[0118] The transient cooperative control command is generated based on the transient target trigger angle. The transient cooperative control command is configured to output a high-frequency pulse sequence to drive the physical switching devices of the target underlying converter.

[0119] Further defining step S4, the target weight component is input into the trigger angle mapping function to calculate the phase angle offset compensation amount, specifically including:

[0120] Obtain the preset commutation failure safety margin parameter; perform a subtraction calculation based on the reference steady-state trigger angle and the commutation failure safety margin parameter to extract the maximum allowable forward phase angle limit;

[0121] Obtain a preset nonlinear mapping scaling factor; use the nonlinear mapping scaling factor to perform a multiplication operation on the target weight components to obtain a normalized intermediate weight quantity with unified dimensions;

[0122] Using the normalized weight intermediate as an independent variable, the inverse cosine mapping conversion logic is executed to obtain the theoretical phase angle offset.

[0123] Determine whether the theoretical phase angle offset is greater than the maximum allowable forward phase angle limit;

[0124] If the judgment result is greater than a certain value, the maximum allowable forward phase angle limit will be output as the phase angle offset compensation amount.

[0125] If the judgment result is less than or equal to the theoretical phase angle offset, the theoretical phase angle offset is output as the phase angle offset compensation.

[0126] Further defining step S5, when the data update cycle of the global fatigue damping coefficient exceeds a preset communication time threshold, a damping attenuation degradation strategy is triggered, reducing the global fatigue damping coefficient according to a preset attenuation function and reconstructing the dynamic compensation weight matrix, including:

[0127] Read the status parameters of the local communication watchdog timer; when the status parameters indicate that the data update period exceeds a preset communication time threshold, generate an interruption alarm signal representing a physical abnormality of the upper-layer data link;

[0128] In response to the interruption alarm signal, the local state overwrite process is activated, so that the current value of the global fatigue damping coefficient is continuously reduced to the lower limit of zero according to the preset exponential time constant.

[0129] The global fatigue damping coefficient during the decreasing process is input into the joint compensation mapping logic in real time and iteratively, forcing the dynamic compensation weight matrix to degenerate into a first backup topology controlled by the local transient voltage compensation scalar distribution.

[0130] Further defining step S5, the local state overwrite process is activated, causing the current value of the global fatigue damping coefficient to continuously decrease to the lower limit of zero according to a preset exponential time constant. Specifically, this includes:

[0131] Before generating the interruption alarm signal, the last validly updated global fatigue damping coefficient and the corresponding global fatigue damping coefficient value of the previous historical period are extracted from the local historical cache stack.

[0132] Perform a difference calculation between the global fatigue damping coefficient of the last effective update and the global fatigue damping coefficient value of the previous historical cycle to extract the characteristic value of damping degradation rate.

[0133] Obtain the basic configuration time constant; obtain the preset sensitivity adjustment factor; the sensitivity adjustment factor is configured to eliminate the dimensional differences in the characteristic value of the damping degradation rate;

[0134] The damping degradation rate characteristic value is multiplied by the sensitivity adjustment factor to extract the dimensionless attenuation multiplier; the dimensionless attenuation multiplier is added to the natural constant to generate an adaptive denominator operator; the basic configuration time constant is divided by the adaptive denominator operator, and the quotient is extracted as the dynamic degradation time constant.

[0135] The dynamic degradation time constant is used as the preset exponential time constant.

[0136] Further specified, within the data update cycle, the internal digital decay integral logic is initiated with the discrete control step size as the time iteration unit;

[0137] Obtain the discrete control step size and the natural logarithm base constant; divide the discrete control step size by the dynamic degradation time constant to extract the single-step decay exponent amplitude; use the negative of the single-step decay exponent amplitude as the exponent and perform a power operation on the natural logarithm base constant to extract the absolute discrete decay multiplier constant.

[0138] At the triggering time of each discrete control step, the historical retention value of the global fatigue damping coefficient is extracted; the historical retention value is multiplied by the absolute discrete attenuation multiplier constant to obtain the current value after attenuation;

[0139] The decayed current value is overwritten as the historical retention value for the next iteration; the multiplication and overwrite operations are performed repeatedly until it is determined that the decayed current value is less than or equal to a preset lower limit of zero.

[0140] The following are specific implementation instructions for steps S4 to S5 above:

[0141] The target weight component is denoted as The dimensionless allocation ratio, accurately routed from the dynamic compensation weight matrix to the current underlying converter node, determines the output power quota that the node should undertake in the network-wide transient mutual assistance. Based on the current node's underlying communication addressing sequence (MAC address abstract identifier), a traversal matching is performed in the dynamic compensation weight matrix to extract the matrix element values ​​at the corresponding coordinates. The logic for determining the target weight components specifically includes the following topology addressing and signal parsing steps: obtaining the dynamic compensation weight matrix broadcast via the high-speed industrial Ethernet bus; declaring that the matrix's data structure in memory is fixed as a one-dimensional array containing multiple addressing record entries, each record entry containing a key-value pair of network addressing feature code and dimensionless weight value; extracting the local reference anchor point; reading the target underlying converter's built-in network card memory to be executed for the current transient cooperative control command, and extracting the local Media Access Control (MAC) address feature string that is unique across the entire network. The system executes array traversal comparison logic; it uses the extracted local media access control address feature string as a static comparison benchmark and initiates a loop addressing process; it reads the network addressing feature code row by row in the array of dynamic compensation weight matrices and performs an XOR consistency matching operation with the static comparison benchmark. It captures and extracts the action; in response to a trigger condition that the matching operation result is true, it immediately terminates the loop addressing process and extracts the dimensionless weight value bound to the current matching record entry. The extracted dimensionless weight value is established as the target weight component assigned to this transformer node.

[0142] The commutation failure safety margin parameter is denoted as This refers to the absolute electrical phase angle protection line maintained by the underlying converter's physical switching devices when subjected to transient mutual current to prevent excessive thyristor commutation overlap angle from causing a bridge arm shoot-through short circuit.

[0143] The established logic is as follows: Construct an equivalent hardware calibration platform; acquire the target converter under test (DUT) and physically connect its output to an equivalent pure resistive-inductive load box simulating the dynamic process impedance characteristics of aluminum foil forming grooves; connect its AC input to a high-power simulated grid power supply with programmable transient sag generation capability. Set initial steady-state operating conditions; control the high-power simulated grid power supply to output a constant rated bus voltage and configure the DUT to operate at the nominal firing angle, recording the reference output current at this time. Execute a boundary limit test cycle; control the high-power simulated grid power supply to trigger a deep sag waveform with a sag amplitude of 50% of the rated bus voltage within a 10-millisecond window; during the voltage sag maintenance period, gradually reduce the actual operating firing angle of the DUT's DUT in absolute fixed decreasing steps of 0.1 electrical angles. During the decreasing process, simultaneously extract the transient current waveform signals of the upper and lower bridge arms of the DUT's DUT through a high-frequency Hall current sensing interface at a stable sampling frequency of not less than 100kHz. Zero-crossing and overlap interval width analyses are performed on the real-time acquired transient current waveform signals. At the instant when it is determined that the currents of both upper and lower bridge arms are greater than zero within the same absolute clock cycle (indicating a physical short circuit), the operating trigger angle value causing the short circuit is immediately frozen and extracted, and marked as the critical collapse phase angle. The previously recorded nominal trigger angle is obtained, and a subtraction calculation is performed between the nominal trigger angle and the critical collapse phase angle to extract the absolute value of the phase angle difference. A preset engineering redundancy multiplier is obtained; the absolute value of the phase angle difference is multiplied by the engineering redundancy multiplier, and the final product is used as the commutation failure safety margin parameter.

[0144] In this embodiment, the engineering redundancy multiplier is set to 1.15, and the preferred value for the commutation failure safety margin parameter calculated by the above process is set to 12.5 electrical degrees. This achieves an objective physical balance between "ensuring maximum usable electrical energy through extreme transient mutual assistance" and "covering the manufacturing discreteness of different batches of thyristors to ensure equipment survival".

[0145] Nonlinear mapping scaling factor, denoted as This represents the dimension transformation operator used to map dimensionless abstract weight allocation ratios to independent variables with physical and electrical properties (electric angles). The determination of the nonlinear mapping scaling factor specifically includes the following calculation process:

[0146] The system retrieves the rated secondary line voltage, nominal short-circuit impedance percentage, and rated apparent capacity of the current controlled rectifier transformer as indicated on its nameplate. It also retrieves the pre-stored network-wide preset reference apparent capacity (set to 100MVA). A division operation is performed between the network-wide preset reference apparent capacity and the current equipment's rated apparent capacity to extract the capacity conversion ratio. A multiplication operation is performed between the nominal short-circuit impedance percentage and the capacity conversion ratio to extract the equivalent impedance per-unit value under a unified system benchmark. A multiplication operation is performed between the rated secondary line voltage and the equivalent impedance per-unit value to extract the absolute impedance voltage drop characteristic. A preset reference voltage constant (specifically, the nominal line voltage of the power grid where the controlled node is located) is retrieved. The absolute impedance voltage drop characteristic is divided by the reference voltage constant to extract the dimensionless voltage-to-impedance ratio characteristic. Finally, an absolute angle conversion reference constant is retrieved, with its value preset to pi (π). The quotient of the voltage impedance ratio is divided by 2, and its dimension is defined as radians / unit weight. The voltage impedance ratio characteristic is multiplied by the absolute angle conversion reference constant to obtain the nonlinear mapping scaling factor. In this embodiment, the nonlinear mapping scaling factor is dynamically calculated or fixedly configured as 0.85 radians / unit weight.

[0147] The normalized weight intermediate quantity is denoted as The standardized compensation scalar, after dimensional alignment and scaling, is directly applicable to the conditions of the nonlinear operator. It is obtained by multiplying the target weight components and the nonlinear mapping scaling factor.

[0148] The maximum permissible forward phase angle limit is denoted as This characterizes the limiting physical boundary at which the firing angle can be adjusted ahead of time under the current process load. The reference steady-state firing angle is obtained, and a subtraction calculation is performed between the reference steady-state firing angle and the commutation failure safety margin parameter to extract the value.

[0149] State parameter, denoted as This represents the physical connection and message refresh health between the edge computing node and the upper-layer macro-scheduling network (SCADA master station data link). The status parameters are determined by performing the following signal processing steps:

[0150] Define the external data input interface environment; this interface is configured to continuously receive periodically transmitted messages from the upper-layer macro-scheduling network (SCADA master station). Extract the latest valid timestamp feature from the most recently successfully parsed frame of periodically transmitted messages. Read the current local absolute timestamp output by the real-time clock generator inside the local microprocessor.

[0151] Perform a subtraction operation between the current local absolute timestamp and the latest valid timestamp to extract the absolute time deviation.

[0152] The system determines whether the absolute time deviation is greater than a preset dead-zone constant. If the absolute time deviation is greater than the dead-zone constant, a logic high-level characteristic value (Boolean value 1) representing "connection timeout" is generated. If the absolute time deviation is less than or equal to the dead-zone constant, a logic low-level characteristic value (Boolean value 0) representing "connection healthy" is generated. The output logic high-level or logic low-level characteristic value is established as a state parameter. In this embodiment, the dead-zone constant is set to 2.5 times the network's basic polling cycle. This not only eliminates dependence on the underlying chip hardware pins but also, through a 2.5 times redundancy design, filters out pseudo-timeouts caused by regular Ethernet random retransmissions and packet jitter, avoiding triggering erroneous actions.

[0153] The preset communication time threshold is denoted as It represents the maximum tolerable communication dead zone time span before an irreversible physical disconnection of the data link is determined.

[0154] For the logic of determining the reference steady-state firing angle, during the initialization phase of the coordinated voltage regulation and control system, the basic bias calculation process is executed, which specifically includes: obtaining the effective value of the rated AC line voltage on the secondary side of the current controlled rectifier transformer, and the target rated DC bus voltage required to be maintained by the current aluminum foil formation process line (example value is 750V); dividing the target rated DC bus voltage by the effective value of the rated AC line voltage to obtain the DC voltage ratio factor; obtaining the rectifier topology constant (for a three-phase fully controlled bridge rectifier circuit, the rectifier topology constant is configured as 1.35), dividing the DC voltage ratio factor by the rectifier topology constant to extract the theoretical steady-state cosine value; using the theoretical steady-state cosine value as the independent variable, performing an inverse cosine mathematical mapping operation to extract the absolute theoretical firing angle; obtaining the marking on the nameplate of the current controlled rectifier transformer. The maximum permissible DC voltage operating limit is defined, which characterizes the physical boundary of the equipment insulation and component tolerance. The maximum permissible DC voltage operating limit is divided by the rated AC line voltage effective value to obtain the DC voltage limit ratio factor. The limit ratio factor is then divided by the rectifier topology constant to extract the limit safety cosine value. Using the limit safety cosine value as the independent variable, an inverse cosine mathematical mapping operation is performed to extract the minimum physical safety firing angle. The absolute theoretical firing angle and the minimum physical safety firing angle are subtracted to obtain the phase angle difference between the two. This phase angle difference is directly assigned as the forward transient mutual assistance phase angle margin, and the absolute theoretical firing angle is used as the reference steady-state firing angle for output. In this embodiment, for a 10kA level process line, the reference steady-state firing angle extracted by the above dynamic calculation is 35 electrical degrees.

[0155] For the logic of determining the preset communication time threshold, a time boundary constraint extraction process based on physical network and process inertia is executed. Specifically, this includes: reading the read-only hardware description file embedded in the gateway's underlying communication module, using a memory addressing pointer to extract the maximum nominal value of the underlying physical protocol self-healing time that matches the currently configured industrial optical network protocol (PROFINET or EtherCAT) from the file, and directly assigning it as the upper limit characteristic value of the ring network self-healing convergence time; characterizing the longest physical delay from network failure to topology reconstruction, which is between 20 milliseconds and 50 milliseconds. The following steps are taken: First, obtain the lower limit characteristic value of the polarization collapse time of the electrolyte in the current formation tank. This value represents the physical limit critical time at which the process voltage irreversibly drops after the loss of control signal. Second, calculate the algebraic difference between the lower limit characteristic value of the polarization collapse time and the upper limit characteristic value of the ring network self-healing convergence time to extract the safety blind zone tolerance. Third, add a fixed proportion (50%) of the safety blind zone tolerance to the upper limit characteristic value of the ring network self-healing convergence time to generate and assign a preset communication time threshold. In this embodiment, the preferred threshold extracted by the above constraints is configured as 150 milliseconds. This threshold is ensured to always be greater than the network self-healing time to maintain logical silence, and less than the polarization collapse time to trigger a physical fallback action at the limit.

[0156] The characteristic value of the damping degradation rate is denoted as The degradation gradient of the physical attenuation state of the device on the time axis before it enters the communication dead zone. The extraction logic of the damping degradation rate feature value is specifically decomposed into the following steps:

[0157] Retrieve the local historical microsecond-level cache stack; this cache stack is configured to continuously store the calculated global fatigue damping coefficient values ​​with absolute time-series tags during steady-state operation using a first-in-first-out (FIFO) mechanism. In response to the triggering of an interrupt alarm signal, immediately freeze write operations on the local historical microsecond-level cache stack to protect the field data slices before the loss of connection.

[0158] The extreme boundary state interception logic is executed as follows: The current push depth feature of the local historical microsecond-level cache stack is read; it is determined whether the current push depth feature is greater than or equal to constant two (constant two represents the minimum number of historical samples required to perform differential calculation); in response to the determination that the current push depth feature is less than constant two, this state indicates an extreme boundary anomaly in the cold start phase or loss of historical records, triggering the default initialization protection branch: directly assigning constant zero to the damping degradation rate feature value and skipping subsequent historical addressing and differential calculation steps; when historical degradation trend reference is lost, the device degradation rate is forcibly assumed to be zero, thereby calling the most conservative basic configuration time constant for smooth degradation, ensuring that the device will not suffer sudden physical impact due to unknown states. In response to the determination that the current push depth feature is greater than or equal to constant two, the last validly updated global fatigue damping coefficient is extracted from the local historical microsecond-level cache stack according to the reverse timestamp addressing logic; following the same reverse addressing logic, the damping coefficient value of the previous historical period immediately preceding the last validly updated global fatigue damping coefficient in the time series is extracted. The initial single-step damping increment is extracted by subtracting the damping coefficient value from the damping coefficient value of the previous historical cycle from the last effective update of the global fatigue damping coefficient.

[0159] Before performing subsequent calculations, this embodiment performs non-negative saturation filtering logic on the initial single-step damping increment to prevent mathematical logic deadlock caused by underlying sampling white noise: Specifically, it extracts the absolute zero constant; determines whether the initial single-step damping increment is greater than the absolute zero constant; in response to the abnormal condition that the determination result is greater than, this condition indicates that the current damping coefficient is abnormally greater than the previous damping coefficient, indicating pure signal noise interference, triggering a forced amplitude limiting action, discarding the initial single-step damping increment, and assigning the absolute zero constant as a value and defining it as the effective single-step damping increment; in response to the normal condition that the determination result is less than or equal to (indicating normal damping degradation), it performs an absolute value operation on the initial single-step damping increment, and assigns the absolute value result as the effective single-step damping increment. It extracts the absolute time span parameter corresponding to the two feature storages on the stack. It performs a division calculation on the effective single-step damping increment and the absolute time span parameter to extract the damping degradation rate feature value that is constantly greater than or equal to zero. By using the aforementioned pre-processed non-negative saturation filtering logic, the risk of division-to-zero crash or negative exponential explosion that may be encountered when generating adaptive denominator operators is avoided.

[0160] In this embodiment, the preferred value of the basic configuration time constant is limited to 5.0 seconds. This is a compromise configuration based on material mechanics and electrical transient time. The voltage drop caused by photovoltaics can determine the life or death of the process within a few hundred milliseconds to 1 second (an extremely fast electrical change process), while the macroscopic decay of the tap changer and winding hot spot temperature in the transformer tank is a process on the order of minutes or more (a slow thermal change process). The technical purpose of setting a reference decay constant of 5.0 seconds is to forcibly define a physical safety buffer channel between "before the formation tank voltage completely collapses (rapidly releasing electrical capacity)" and "avoiding the sudden cancellation of fatigue damping causing a drastic change in the electromagnetic force inside the transformer to tear the insulation paper (preventing excessively rapid mechanical impact)".

[0161] The dynamic degradation time constant is denoted as This characterizes the smooth decay rate of the equipment's fatigue protection barrier when the upper-level macro-control is lost. The determination of the dynamic degradation time constant involves the following calculation logic, including dimensional compensation: obtaining the basic configuration time constant; simultaneously obtaining the previously extracted damping degradation rate characteristic value. A preset sensitivity adjustment factor is introduced and obtained; the dimension of this sensitivity adjustment factor is defined as the strict reciprocal of the dimension of the damping degradation rate characteristic value, its function being to eliminate dimensional differences and convert subsequent multiplications into dimensionless values. A multiplication operation is performed on the damping degradation rate characteristic value and the sensitivity adjustment factor to generate an absolutely dimensionless decay multiplier. The dimensionless decay multiplier is added to a preset digital constant to generate an adaptive denominator operator. In this embodiment, the addition operation avoids the division-by-zero crash anomaly that may be triggered when the degradation rate is zero. The basic configuration time constant is divided by the adaptive denominator operator. The quotient of this division operation is extracted and output as the dynamic degradation time constant. In this embodiment, the sensitivity adjustment factor is preset to 1000.0 seconds. This forces the microscopic rate of degradation to be amplified into the macroscopic control denominator.

[0162] In a preferred embodiment, the processor built into the edge computing gateway performs feature extraction based on the following nonlinear dimensional compensation equation: The physical properties and numerical boundaries of each parameter are strictly defined as follows: Defined as the dynamic degradation time constant, it represents the actual smooth inertial reference for the stripping of the equipment's protective barrier. Its data attribute is a positive real number, and its dimension is seconds (s). The basic configuration time constant is defined in seconds (s), and the preferred value in this embodiment is 5.0s; Defined as the damping degradation rate characteristic value output by non-negative saturation filtering logic, it characterizes the transient degradation gradient of mechanical life, and its dimension is the reciprocal of seconds. ); Defined as an introduced sensitivity adjustment factor, its core technical function is "dimension conversion reciprocal operator and sensitivity amplification", with the dimension being seconds (s), thereby increasing the product in the denominator. This becomes an absolute dimensionless value. In this embodiment, its preferred range is [500.0, 2000.0]s, and the preferred value is 1000.0s. This embodiment is not limited to the specific values ​​mentioned above. In practical applications, if the change of transformer insulating oil grade leads to a change in heat capacity, it can be modified by updating the configuration document. and Combination matching.

[0163] The first backup topology is denoted as... This characterizes the extreme physical fallback output state of a multi-rectifier transformer cluster after its damping protection is forcibly reset to zero, relying solely on local transient voltage drop differentials for brutal mutual assistance. Once the global fatigue damping coefficient approaches its lower limit of zero, the output is naturally reconstructed by the joint compensation mapping logic under undamped constraints.

[0164] For step S4, the micro-mapping mechanism of the transient mutual assistance electrical execution closed loop is implemented: After generating the dynamic compensation weight matrix, under the extreme condition of a precipitous drop in photovoltaic direct power supply, the instantaneous surge in compensation weights can lead to excessive forward shift of the phase angle, which can easily cause commutation failure of the underlying converter's physical switching devices, thereby causing more severe cascading power outages. To solve this bottleneck, this embodiment embeds a "dynamic phase angle saturation margin constraint mechanism" based on the commutation failure safety margin in the original trigger angle mapping step; specifically, from the previously generated dynamic compensation weight matrix, matching addressing is performed based on the local underlying communication addressing sequence to extract the target weight component corresponding to the current target underlying converter. Dimensional consistency verification and heterogeneous parameter conversion logic are executed: a preset nonlinear mapping scaling factor is obtained; the target weight component is multiplied using this nonlinear mapping scaling factor. The essence of this multiplication operation is to execute the dimension conversion operator, forcibly mapping the purely dimensionless network allocation ratio to a unified normalized input parameter that conforms to the domain of the inverse cosine function, generating a normalized weight intermediate quantity. Using the normalized weight intermediate quantity as an independent variable, it is input to the trigger angle mapping logic to perform an inverse cosine nonlinear mapping conversion based on series expansion, and extract the theoretical phase angle bias.

[0165] It is necessary to further explain the specific details of the action in step S4, "execute the inverse cosine mapping conversion logic to obtain the theoretical phase angle offset." Specifically, it is broken down into the following steps:

[0166] Receive the normalized weight intermediate value output from the previous step. Perform numerical boundary forced pruning to prevent overflow; obtain the preset upper limit value (a constant positive one) and lower limit value (a constant negative one) of the mathematical domain; determine whether the normalized weight intermediate value exceeds the mathematical closed interval formed by the upper and lower limits; if it is determined to exceed, force it to be pruned to the boundary threshold closest to it on the number line, and output the valid domain independent variable.

[0167] Using the legally defined domain independent variables as the baseline input, an approximation algorithm based on polynomial expansion is executed. Specifically, a nonlinear empirical model based on axiomatic Maclaurin series expansion is adopted. In this step, an initialization operation is performed, loading a specific sequence of odd-order polynomial coefficient constants for approximation calculations from the local external read-only memory. The elements in this constant sequence are predefined as follows: the first-order approximation coefficient is extracted and its value is fixed as constant 1; the third-order approximation coefficient is extracted by dividing constant 1 by constant 6; the fifth-order approximation coefficient is extracted by dividing constant 3 by constant 40. This process is repeated to construct an ordered constant array containing the first five features, and this ordered constant array is loaded into the runtime memory as a specific sequence of odd-order polynomial coefficient constants. Using the first five odd-order terms for nonlinear approximation, the mapping error in the steady-state operating range can be controlled within the engineering allowable range (preferably within 1.5 electrical degrees) that meets the commutation safety margin, under the limited computing power of the converter DSP controller, thus achieving the best technical balance between "computational resource overhead" and "phase angle control accuracy". For the independent variables in the legal domain, incremental odd-order exponentiation operations are performed independently in sequence (calculating to the first, third, fifth, etc.), and the generated exponentiation result is multiplied one by one with the corresponding constant value in the odd-order polynomial coefficient constant sequence to generate a series of mutually independent polynomial product characteristic terms. Algebraic accumulation is performed on all the generated polynomial product characteristic terms to extract the first intermediate radian value. The numerical constant of half pi is obtained; a subtraction calculation is performed between the numerical constant of half pi and the first intermediate radian value to obtain the abstract theoretical bias radian.

[0168] The underlying microprocessor, based on its built-in floating-point unit (FPU), performs the extraction of the abstract theoretical bias radians according to the following algebraic equation: ;in, Defined as the abstract theoretical bias radian of the calculated output, it characterizes the spatial increment of the ideal control phase angle of the converter. Its data attribute is a single-precision floating-point scalar, and its dimension is radians (rad). Pi is defined as a constant of π, and in this embodiment, the preferred accuracy is 3.14159. Defined as an algebraic accumulation operator performed on a specified order within a set, executed by the microprocessor's accumulation register; m is defined as the power and coefficient index of the polynomial, whose values ​​are discretely constrained to the set of positive integers. The characterization only extracts the first three odd-order terms to achieve a balance between computing power and accuracy; Defined as the normalized weight intermediate quantity representing the legal domain independent variable input in the preceding steps, its dimension is a dimensionless pure number, and its legal physical boundary is rigidly clamped within the closed interval [-1.0, 1.0]. Defined as the sequence of constant coefficients for a specific odd-order polynomial of order m. Based on axiomatic expansion, each element is permanently stored in read-only memory, and its specific value is uniquely anchored as follows: when m=1, ;when hour, ;when hour, This embodiment is not limited to the specific values ​​mentioned above. In practical applications, the set of index m can be extended to... To further reduce the truncation error of nonlinear fitting, a series expansion equation with clear order boundaries and coefficient quantization is introduced to avoid excessive memory consumption and interpolation blind spots caused by using pure lookup table methods for transcendental functions in low-computing-power DSPs. This achieves high-fidelity output of phase angle compensation calculation and absolute controllability of computation latency within microsecond-level control cycles.

[0169] Obtain a preset radian-to-electrical-angle conversion coefficient. This conversion coefficient is defined to eliminate the dimensional difference between abstract mathematical radians and underlying physical electrical angles. Specifically, obtain the constant 180 and the constant pi, perform a division operation by dividing 180 by pi, and extract the quotient as the radian-to-electrical-angle conversion coefficient. Multiply the abstract theoretical offset radian by the radian-to-electrical-angle conversion coefficient; the resulting product is the theoretical phase angle offset. After the aforementioned validity check of the maximum permissible forward phase angle limit and the output of the phase angle offset compensation, perform a dimensional subtraction operation between the reference steady-state trigger angle and the phase angle offset compensation to extract the transient target trigger angle.

[0170] It should be further explained that: the initial forward target trigger angle is extracted by subtracting the output phase angle offset compensation from the reference steady-state trigger angle. To prevent overcompensation from causing the thyristor to enter the commutation failure dead zone and suffer physical damage under extreme transient drop conditions, an absolute boundary truncation logic is performed on the initial forward target trigger angle before executing the aforementioned mapping conversion logic from the spatial domain to the physical time domain: specifically, the previously calibrated commutation failure safety margin parameter is obtained; this parameter represents the minimum allowable trigger angle defense line that the switching device must retain to prevent the upper and lower bridge arms from shoot-through short circuits under the current hardware topology. It is determined whether the initial forward target trigger angle is less than the commutation failure safety margin parameter; in response to the abnormal condition that the determination result is less than, the saturation protection action is triggered: the initial forward target trigger angle is forcibly truncated and discarded, and the commutation failure safety margin parameter is directly assigned and defined as the final transient target trigger angle. In response to the safety condition that the determination result is greater than or equal to, the initial forward target trigger angle is directly assigned and defined as the final transient target trigger angle. By introducing a post-processed absolute boundary truncation check, the boundary vulnerability of physical hardware failure is plugged.

[0171] Furthermore, dynamic boundary blocking logic is executed to prevent electrical failure: A reference steady-state firing angle is obtained to maintain the rated base voltage of the aluminum foil formation process; the pre-defined commutation failure safety margin parameter is obtained; a subtraction calculation is performed between the reference steady-state firing angle and the commutation failure safety margin parameter to extract the maximum allowable forward phase angle limit; it is determined whether the theoretical phase angle offset is greater than the maximum allowable forward phase angle limit; in response to the condition that the result is greater than, a saturation cutoff action is triggered, discarding the theoretical phase angle offset and forcibly outputting the maximum allowable forward phase angle limit as the phase angle offset compensation; in response to the condition that the result is less than or equal to, the theoretical phase angle offset is directly output as the phase angle offset compensation. The reference steady-state firing angle is subtracted from the output phase angle offset compensation, and a subtraction calculation is performed to extract the transient target firing angle.

[0172] After extracting the transient target trigger angle, the mapping conversion logic from the spatial domain electrical angle to the physical time domain is executed: establishing a physical time reference; continuously acquiring the three-phase grid voltage at the AC input terminal of the target underlying converter through a hardware phase voltage sampling loop and inputting it into the phase-locked loop synchronization logic; extracting the zero-crossing moment; using the phase-locked loop synchronization logic, extracting the natural zero-crossing moment when the corresponding phase voltage crosses from negative to positive polarity and marking it as the absolute time zero point; obtaining the current grid frequency; using the phase-locked loop synchronization logic to track and output the real-time grid frequency characteristic value; the underlying microprocessor performs a reciprocal operation on the real-time grid frequency characteristic value to calculate the real-time grid power frequency cycle parameter (in this embodiment, for a 50Hz grid, the nominal value of this parameter is 20 milliseconds); performing reference conversion; obtaining constants. 360 (representing the total electrical angle of a complete power frequency cycle) is used to divide the real-time power grid cycle parameter by a constant 360, extracting the unit electrical angle time equivalent (in milliseconds / degrees). Time domain mapping is then performed. The transient target trigger angle is multiplied by the unit electrical angle time equivalent to obtain the absolute trigger delay time calculated from the absolute timing zero point. This absolute trigger delay time is loaded into the hardware timer comparison register of the underlying microprocessor. In response to the physical condition that the hardware timer count reaches the absolute trigger delay time, a high-frequency pulse sequence is immediately driven from the internal digital pins to control the conduction of the physical switching devices. This high-frequency pulse sequence is configured as a transient coordinated control command to instruct the target underlying converter physical switching devices to perform precise conduction actions. This achieves absolute decoupling of "microscopic transient mutual energy extraction" and "macroscopic underlying converter safety" at the microsecond-level execution layer.

[0173] It should be noted that the underlying microprocessor calculates and generates the count value used to control the pulse width based on the following hardware timer loading equation: ;in, Defined as absolute trigger delay time, it represents the absolute time span from the zero-crossing point of the AC voltage to the flip-out output drive pulse of the microprocessor digital pin, and its dimension is milliseconds (ms). Defined as the transient target trigger angle output after the preceding steps have undergone safety saturation truncation, its dimension is electrical angle ( ). In this embodiment, the preferred range is [12.5, 150.0] electrical angles; Defined as a real-time power grid frequency cycle parameter, extracted in real time by the underlying hardware phase voltage phase-locked loop (PLL), its dimension is milliseconds (ms). In this embodiment, for a 50Hz AC power grid, the preferred value is 20.0ms, but dynamic drift within the frequency fluctuation range of [19.5, 20.5]ms is allowed; "360" is defined as the equivalent geometric constant of a complete electrical cycle, its dimension is electrical angle (°). This embodiment is not limited to the specific values ​​mentioned above. In practical applications, for overseas aluminum foil forming production lines operating at 60Hz, The baseline optimal value for adaptive adaptation is 16.67ms.

[0174] For step S5, a degradation and physical fallback closed-loop mechanism is executed: This method embeds "adaptive degradation tracing logic based on historical attenuation gradients," sacrificing the deterministic cost of slow-changing mechanical lifespan in exchange for the ultimate output of the entire network's fast-changing electrical mutual support. The specific execution flow is as follows:

[0175] During the real-time control cycle, the state parameters of the communication timeout status monitoring logic representing the local watchdog mechanism are continuously read. It is determined whether the data update cycle for calculating the global fatigue damping coefficient exceeds a preset communication time threshold. In response to the state parameter indicating that the data update cycle exceeds the preset communication time threshold, an interruption alarm signal representing a physical anomaly in the upper-layer data link is immediately generated. In response to the interruption alarm signal, the local state overwrite process is immediately activated. In this process, the fixed static decay rate is abandoned, and dynamic feature tracing is performed instead.

[0176] Specifically, from the local historical microsecond-level cache stack, the last validly updated global fatigue damping coefficient and the damping coefficient value of the immediately preceding historical period are extracted in reverse chronological order. A difference calculation is performed between this last validly updated global fatigue damping coefficient and the damping coefficient value of the preceding historical period to extract the damping degradation rate characteristic value, which represents the fatigue acceleration trend in the instant before disconnection. The basic configuration time constant is obtained. A nonlinear proportional adjustment operation is performed between the basic configuration time constant and the damping degradation rate characteristic value to adaptively generate a dynamic degradation time constant based on the actual deterioration inertia of the current equipment. This dynamic degradation time constant is precisely assigned and replaces the preset exponential time constant in the original calculation model.

[0177] After obtaining the dynamic degradation time constant, within the data update cycle, the internal digital decay logic, which approximates the dynamic degradation step size, is initiated using a discrete control step size as the time iteration unit.

[0178] Specifically, the natural logarithm base constant is obtained; the discrete control step size of the current control system is extracted, and the discrete control step size is divided by the dynamic degradation time constant to extract the single-step decay exponent amplitude; the negative number (with a negative sign) of the single-step decay exponent is used as the determined mathematical exponent, and the natural logarithm base constant is exponentially calculated to extract the absolute discrete decay multiplier constant; at the trigger moment of each discrete control step, the historical retention value of the global fatigue damping coefficient is extracted, which is the initial value at the moment the overwrite process starts during the first iteration; the historical retention value is multiplied by the absolute discrete decay multiplier constant in a single operation to obtain the current value after decay; the current value after decay is immediately overwritten as the historical retention value required for the next iteration;

[0179] It should be further explained that: In this embodiment, the physical exponential law of continuous time is mapped into the discrete digital computing clock domain of the microprocessor without error. The edge computing gateway performs damping value update calculation for each interrupt cycle based on the following iterative state equation: ;in, Defined as the first The decayed value of the output of the next iteration has a dimensionless real number as its dimension and its range dynamically converges to the interval (0,1). Defined as the first The historical retained value read from the cache before the next iteration begins (when When the value is 0, this is the initial damping value for the instantaneous freeze triggered by the degradation strategy. Defined as the numerical iteration index number on the discrete time axis, it is a natural number that increments from 0, and each increment corresponds to the triggering of an underlying hardware timer interrupt; e is defined as the base constant of the natural logarithm. Defined as the discrete control step size, it represents the physical absolute time interval between two adjacent iterations, with the dimension of seconds (s). In this embodiment, it is controlled by the interrupt cycle of the microprocessor, and the preferred value is 0.01s. Defined as the dynamic degradation time constant calculated and passed in by the preceding steps, with the dimension of seconds (s). Defined as the negative of the single-step decay exponent magnitude, it is a dimensionless negative real scalar. This embodiment is not limited to the specific values ​​mentioned above; in practical applications, The value is flexible and varies between 1 millisecond and 50 milliseconds depending on the real-time main frequency of the selected industrial control chip. Through the above-mentioned iterative discrete geometric multiplication with the absolute discrete attenuation multiplier constant as the common ratio, the global fatigue damping coefficient is forced to converge step by step along an absolutely precise exponential decay trajectory under the drive of the digital clock, until it is determined that the current value after attenuation is equal to or less than the preset lower limit of zero.

[0180] During the aforementioned continuously decreasing dynamic process, the global fatigue damping coefficient of each iteration cycle is input into the joint compensation mapping logic in step S3 in real time and by force.

[0181] As the global fatigue damping coefficient continuously approaches the lower limit of zero, based on the negative exponentiation principle within the joint compensation operator, the arithmetic result of exponentiation with the natural logarithm base using a negative zero value inevitably remains constant at the value of one. This process eliminates the original fatigue attenuation constraint multiplier from the underlying mathematical logic, forcing all transformer nodes trapped in communication islands across the entire network to completely degenerate their dynamic compensation weight matrix into a first backup topology controlled only by the local transient voltage compensation scalar space distribution. This ensures that the DC bus voltage does not collapse under catastrophic conditions of communication failure coupled with photovoltaic voltage drop.

[0182] The following is an in-depth analysis of the mechanism of action of the core output variables and input parameters: In this embodiment, the core pivot parameter of the transient cooperative control command solution flow is the target weight component in the dynamic compensation weight matrix, which is directly controlled by the global fatigue damping coefficient output by the algorithm. The range of the global fatigue damping coefficient is clamped by the boundary constraint function within the real number interval greater than zero and less than one.

[0183] When the global fatigue damping coefficient approaches 0, it indicates a lower level of accumulated thermal stress and mechanical wear at the transformer node, signifying a healthier physical structure. In the joint compensation mapping logic, this value, approaching 0, serves as a negative exponent, making the attenuation mapping value closer to a constant. This grants the node the fuller capacity for mutual energy output when facing bus voltage dips, ensuring the highest efficiency in transient voltage compensation.

[0184] As the global fatigue damping coefficient approaches 1, it indicates that the transformer winding insulation or tap changer contacts are approaching the critical boundary of physical fatigue collapse. The negative exponential mapping forces the attenuation mapping value to collapse exponentially closer to zero, thereby blocking the node's mutual output request in the physical topology. This trend requires constructing a hardware-based forced isolation mechanism to prevent burnout, but this comes at the cost of sacrificing single-node transient support capabilities to avoid the risk of cascading damage to the underlying converter group.

[0185] The qualitative relationship between the absolute deviation scalar and the dynamic thermal stress weighting factor is nonlinear and positively correlated. In nature, the heat dissipation efficiency of transformer insulating oil decreases nonlinearly with increasing ambient temperature. By using a preset exponential amplification base, the absolute deviation scalar exceeding the drift tolerance threshold is transformed into an exponential operation exponent, thus reproducing the physical law that high temperature hinders heat dissipation, leading to the exponential accumulation of internal hot spot temperature.

[0186] This positive correlation design ensures that when the external meteorological environment deteriorates, the sensitivity of this method to thermal state characteristics is amplified exponentially, thus converting environmental risks into fatigue damping in advance and achieving advanced deterioration prevention under extreme conditions.

[0187] The qualitative relationship between the negative voltage sag gradient of the bus and the local transient voltage compensation scalar is positively correlated. Under the premise of a constant rate of change of the converter output current, the larger the amplitude of the transient voltage sag of the DC bus, the higher the negative sag gradient, indicating a more severe transient energy deficit faced by the bus. The compensation scalar is extracted by calculating the absolute difference between the current load capacity and this gradient, conforming to Kirchhoff's law of conservation of energy. This design ensures that the strength of the mutual assistance response matches the depth of the grid sag in real time, eliminating the response lag defect caused by fixed-step compensation.

[0188] The qualitative relationship between the damping degradation rate characteristic value and the dynamic degradation time constant is an inverse proportional negative correlation. Within the blind zone of communication interruption, devices with faster historical degradation rates have a higher probability of sudden failure in the near future. This method multiplies this value by a sensitivity adjustment factor and places it in the denominator, causing the dynamic degradation time constant to decrease sharply as the degradation rate increases. This inverse proportional relationship establishes an "adaptive fallback degradation" strategy under communication loss conditions. Nodes with severe historical degradation will eliminate damping protection at a faster rate, squeezing out remaining installed capacity before the formation tank voltage completely collapses, thus achieving a dynamic trade-off between physical lifecycle and bus electrical survival rate.

[0189] Furthermore, this embodiment is configured in a digital twin monitoring scenario that includes multiple large-capacity rectifier transformers. This scenario is designed for the photovoltaic direct power supply aluminum foil formation process, and its control model uses the ambient dry-bulb temperature collected by the external weather station, the communication handshake level issued by the upper-level dispatch system, and the bus drop amplitude collected by the lower-level high-frequency voltage transformer as the pre-input sources for the joint compensation logic of this method.

[0190] The data flow logic is established as follows: the multimodal slow-varying characteristic data acquisition channel receives the analog temperature signal and the discrete action frequency signal, and inputs them into the microprocessor core; the microprocessor calls the floating-point arithmetic unit to sequentially execute the weighted integration based on the time integration window and the nonlinear normalization mapping, and updates the global fatigue damping coefficient in the internal high-speed cache; at the same time, when the state monitoring logic detects that the communication watchdog level flips from low to high, this method immediately freezes the external network communication stack, and the internal dynamic compensation weight matrix is ​​taken over by the discrete equal-proportional decay integration logic and performs high-frequency cyclic overwriting.

[0191] Based on the established calculation model, six sets of typical system steady-state and transient calculation response example data were extracted, as shown in the table below.

[0192] Table 1: Examples of Transient Cooperative Control System Response Calculation under Different Operating Conditions and Disturbances

[0193] Scene number and objective status description Absolute deviation scalar Bus voltage negative pressure gradient Communication status parameters Local microscopic dissipative scalar Global fatigue damping coefficient Node mutual potential energy conversion rate Dynamic degradation time constant Transient target trigger angle 1: Standard Environment Steady-State Perturbation 1 0.05 0 0.15 0.029 0.971 N / A 34.8 2: Deterioration of steady state in high-temperature environment 8.5 0.05 0 0.65 0.817 0.441 N / A 34.9 3: High fatigue node encounters deep drop 8.5 12 0 0.68 0.858 0.424 N / A 34.5 4: Health nodes experience a deep drop 2 12 0 0.25 0.075 0.927 N / A 12.5 5: Physical communication failure (node ​​degradation) 1.5 15 1 0.18 0.04 0.96 4.85 12.5 6: Physical communication lost (degrading status) 7 15 1 0.6 0.731 0.481 1.25 28

[0194] The potential energy conversion rate of node mutual assistance is denoted as Specifically, this is achieved by obtaining the natural logarithm base constant; extracting the calculated global fatigue damping coefficient and performing an inverse operation on it to obtain a negative real exponent; using this negative real exponent as the exponent for the exponent calculation, performing an exponentiation calculation on the natural logarithm base constant, and extracting the dimensionless result of this calculation as the node mutual potential energy conversion rate. This parameter quantitatively characterizes the proportion of the original local transient electrical energy margin of the transformer node that can be truly and safely converted into the effective compensation output of the system after being constrained by its own physical, mechanical, and thermal fatigue. The lower this value, the deeper the physical isolation constraint on the underlying equipment.

[0195] The calculated response data quantitatively reveals the progress of this invention in achieving a dynamic balance between "transient mutual energy recovery" and "underlying equipment physical protection" compared to traditional static control technology. Comparing Scenario 1 and Scenario 2, when the absolute deviation scalar (degrees Celsius) exceeds the drift tolerance threshold to 8.5 degrees Celsius due to high temperatures, the nonlinear compensation term is activated, causing the global fatigue damping coefficient to nonlinearly jump from 0.029 to 0.817. Consequently, the nodal mutual energy conversion rate decreases from 0.971 to 0.441. Compared to standard operating conditions, the potential energy conversion rate decreases by 54.58% under high-temperature conditions. This demonstrates that this method can automatically reduce the converter's mutual energy authorization under harsh environmental conditions, verifying the design effectiveness of "implementing active physical protection through multimodal dissipation assessment."

[0196] Comparing Scenario 3 and Scenario 4, when faced with the same DC bus transient drop (negative pressure gradient of 12.00 volts / second), this method exhibits differentiated underlying driving behaviors. In Scenario 4 (healthy node), due to its low global fatigue damping coefficient of 0.075 and node mutual energy conversion rate of 0.927, this method authorizes it to provide full-load support, and the transient target firing angle is shifted forward to touch the preset commutation failure safety margin parameter (truncated to the ultimate energy extraction angle of 12.5 electrical angle). Conversely, in Scenario 3 (high fatigue node), due to the accumulated global fatigue damping coefficient of 0.858, this method forcibly removes its mutual energy authorization. Despite the external load facing energy deficit, its transient target firing angle is locked at 34.5 electrical angle to maintain it near the safety benchmark. This comparison theoretically confirms that the joint compensation mapping logic successfully achieves the technical effect of "safely separating sub-healthy physical devices from the electrical mutual energy high-voltage network."

[0197] Comparing scenarios 5 and 6 under abnormal communication conditions, this method activates the local state overwrite process when the communication state parameter (logic level) transitions to logic level 1. In scenario 6, due to its higher damping degradation rate characteristic value before the connection loss, the dynamic degradation time constant is compressed to 1.25 seconds after the dimension elimination of the sensitivity adjustment factor and the inverse proportional mapping calculation of the adaptive denominator operator. Compared to the 4.85-second dynamic degradation time constant allocated to low-deterioration nodes in scenario 5, the time constant in scenario 6 is reduced by 74.23%. This data demonstrates the technical effectiveness of this method's "adaptive fallback degradation based on historical deterioration inertia": the more severely deteriorated a node, the faster this method deprives it of damping protection at a discrete exponential rate. Thus, under the catastrophic condition of communication islanding superimposed with electrical collapse, it reconstructs the first backup topology at the deterministic cost of sacrificing mechanical lifespan. Specifically, in scenario 6, the trigger angle begins to break free from the locked state and moves towards 28.0 degrees.

[0198] This method evaluates the evolution state of physical nodes using a continuously varying "global fatigue damping coefficient." This invention is based on a damping natural exponential decay mapping function (…). Based on the underlying curvature evolution characteristics of (where x is the global fatigue damping coefficient), combined with the objective physical limitations of power electronic hardware insulation breakdown, the following practical application range is determined through mathematical extreme value identification.

[0199] Interval 1: Lossless mutual assistance response interval, global fatigue damping coefficient range: (0, 0.22]: its boundary threshold of 0.22 is determined by extracting the mathematical characteristics (initial rate of change of the second derivative) of the shallow decay smooth region of the potential energy conversion rate mapping function. Within this interval, the mutual assistance potential energy loss caused by the exponential mapping is objectively less than 20%, indicating that the internal thermal gradient of the transformer and the wear of the tap changer are both within the optimal envelope of the nominal life. The corresponding operation of this method is to convert the calculated local transient voltage compensation scalar into the effective compensation output value of the node. The underlying control command is authorized to perform high dynamic and wide-range firing angle adjustment before touching the maximum allowable forward phase angle limit to ensure absolute voltage stability of the process power grid when facing high-frequency fluctuations.

[0200] Interval 2: Fatigue suppression transition interval, global fatigue damping coefficient range: (0.22, 0.85): its interval span corresponds to the steep slip segment with the largest curvature in the nonlinear mapping function. This objective boundary characterizes that due to severe environmental run drift or high-frequency mechanical action, the internal thermal stress has broken the dissipation equilibrium, and the logic of this method then enters a highly sensitive penalty feedback state for fatigue accumulation.

[0201] The corresponding operation is to initiate the peak-shaving and limiting logic of the Hamiltonian scalar product. This forces the output of a limited dynamic compensation weight component, compelling the transient mutual current to be diverted to other low-damped, healthy nodes in the network topology. This transformer node no longer fully responds to the external pressure gradient, but only performs reduced output to meet the network-wide minimum potential energy conservation threshold.

[0202] Interval 3: Physical isolation protection interval, global fatigue damping coefficient range: [0.85, 1.0): Its trigger boundary of 0.85 is hard-anchored by the physical collapse critical line of the transformer insulation polymerization degree decrease and the minimum convergence region at the tail of the function. This interval indicates that the physical redundancy of the equipment has been completely exhausted, facing a high risk of direct short circuit or thermal breakdown. The corresponding operation is to forcibly execute the forward-shifted over-limit saturation cutoff protection. The effective compensation output value of the node is directly reduced to zero, and the phase angle forward shifting authority is revoked. The underlying converter exits the current transient cooperative network, maintains the reference steady-state trigger angle operation, realizes microsecond-level digital-physical hard isolation, and cuts off the propagation chain of cascaded power outages from the bottom layer.

[0203] The attached diagram is described below: Figure 1 This invention demonstrates the four-stage control process from state perception to anomaly fallback, constructed using rectifier transformers and underlying converters as the hardware execution base in a direct photovoltaic power supply scenario. Figure 1 The technical roadmap is derived from the execution object and deconstructed into four unidirectional logical processing boxes. Specifically, the first processing box, "Multi-dimensional State Feature Extraction," corresponds to and executes the contents of steps S1 and S2. This method simultaneously acquires slowly varying characteristic data representing the transformer's hot spot temperature and mechanical frequency to calculate and generate the global fatigue damping coefficient. At the same time, it acquires high-frequency electrical transient fluctuation data at the photovoltaic power input terminal to extract the local transient voltage compensation scalar, realizing a low-level perception of physical structure attenuation and local energy margin. The second processing box, "Joint Compensation Weight Mapping," directly corresponds to the execution of step S3, which takes the aforementioned global fatigue damping coefficient and local transient voltage compensation scalar as input. By performing negative exponential mapping and Hamiltonian scalar multiplication, a dynamic compensation weight matrix is ​​output to characterize the proportion of transient mutual current distribution within the cluster. The third processing box, "Transient Cooperative Instruction Generation," corresponds to step S4. Based on the preceding matrix, it accurately outputs cooperative control instructions that instruct the underlying converter to perform trigger angle biasing actions, achieving spatiotemporal decoupling between fast-changing electrical regulation and slow-changing physical lifespan. The fourth processing box, "Abnormal Communication Backup Degradation," corresponds to the monitoring and feedback mechanism in step S5. When this method detects in real time that the data update cycle of the global fatigue damping coefficient exceeds the preset communication time threshold, it automatically triggers the damping attenuation degradation strategy, reducing the damping according to the preset exponential time constant and reconstructing the backup topology.

[0204] Figure 4This is a schematic diagram of the numerical verification experiment results for the transient collaborative control logic of the rectifier transformer in an embodiment of the present invention. This numerical verification aims to test the theoretical response characteristics of the present invention under preset standardized boundary conditions (the composite test vector of environmental drift, bus drop, and communication status in Table 1). In the figure, the horizontal axis represents the verification scenario number, the left vertical axis represents the calculated global fatigue damping coefficient, and the right vertical axis represents the transient target firing angle derived based on the joint compensation mapping logic of this method. The dark blue solid line and coordinate axes represent the damping coefficient characteristic curve, and the light yellow broken line and coordinate axes represent the target firing angle response curve.

[0205] like Figure 4 As shown, the numerical calculation results clearly reveal the nonlinear constraint logic of physical dissipation state on the electrical control execution boundary. The joint compensation mapping logic proposed in this method can effectively achieve dynamic decoupling between transient mutual aid and hardware protection. Specifically, in the test range of high fatigue and deep drop (corresponding to scenario 3), the calculated global fatigue damping coefficient climbs to a high level of 0.858. Constrained by this, the transient target trigger angle output by this method is forcibly clamped near the safety baseline of 34.5 electrical degrees, rejecting the mutual aid response. In contrast, in the test of a healthy node facing a deep drop (corresponding to scenario 4), the transient target trigger angle is accurately oriented by the calculation logic to the commutation safety lower limit of 12.5 electrical degrees. This directly confirms that by introducing the global fatigue damping coefficient as a negative exponential feedback feature, this method can proactively implement algorithm-level hard isolation of sub-healthy equipment from the high-voltage mutual aid network.

[0206] Example 2:

[0207] A rectifier transformer for aluminum foil forming includes a transformer body and an edge computing gateway communicatively connected to the transformer body; the edge computing gateway contains a memory and a microprocessor, and the memory stores a computer program; when the microprocessor executes the computer program, it implements the rectifier transformer collaborative voltage regulation and control method.

[0208] The computational logic involved in this application can be constructed using algorithms such as regression analysis in machine learning, establishing a mathematical model by analyzing the inherent trends and interrelationships of the collected parameters. This process can be implemented using specialized computational tools (such as Python's Scikit-learn library or the R language environment). Throughout all calculations, to eliminate the influence of different physical dimensions and ensure that data is compared and analyzed on the same scale, the input parameters in each formula are dimensionless. The dimensionless techniques used include, but are not limited to, max-min normalization or Z-score standardization.

[0209] To decouple the core algorithm from specific application strategies and ensure the configurability and ease of debugging of the technical solution, all configurable operating parameters in the specific implementation path of this invention are read through a standardized "configuration interface". The data source of this configuration interface is a "data storage module" (e.g., a non-transitory computer-readable storage medium, such as a configuration file, database entry, or cloud configuration service), which is configured to store configuration data in key-value pair format.

[0210] It should be emphasized that the foregoing embodiments are merely illustrative of preferred implementations of the present invention and are not intended to limit the scope of protection of the present invention. This application also provides a computer-readable storage medium having computer program instructions stored thereon.

Claims

1. A method for coordinated voltage regulation and control of rectifier transformers, characterized in that, The specific steps include: S1: Obtain multimodal slow-varying characteristic data that characterizes the physical evolution state of each transformer node in the multi-rectifier transformer cluster, and perform dissipation evaluation calculation based on the multimodal slow-varying characteristic data to generate a global fatigue damping coefficient. The global fatigue damping coefficient characterizes the cumulative physical structure attenuation of the transformer node. S2: Obtain electrical transient fluctuation data characterizing the high-frequency fluctuation state of the photovoltaic power input terminal; based on the electrical transient fluctuation data and the current load capacity of each transformer node, calculate the local transient voltage compensation scalar, the local transient voltage compensation scalar characterizing the local energy margin of a single transformer node in response to the current voltage drop; S3: Taking the global fatigue damping coefficient and the local transient voltage compensation scalar as input, the negative exponential mapping and scalar multiplication operation are performed through the joint compensation mapping logic to output a dynamic compensation weight matrix. The dynamic compensation weight matrix represents the distribution ratio of the transient mutual assistance current within the multi-rectifier transformer cluster. S4: Based on the dynamic compensation weight matrix, generate transient cooperative control commands to instruct the underlying converter to perform trigger angle biasing actions; S5: Output the transient collaborative control command, and when the data update cycle of the global fatigue damping coefficient exceeds the preset communication time threshold, trigger the damping attenuation degradation strategy, reduce the global fatigue damping coefficient according to the preset attenuation function and reconstruct the dynamic compensation weight matrix.

2. The rectifier transformer coordinated voltage regulation and control method according to claim 1, characterized in that: The multimodal slow-varying characteristic data includes a first state sub-feature characterizing the temperature distribution of hot spots in the transformer windings and a second state sub-feature characterizing the frequency of mechanical switching of the tap changer. Based on the aforementioned multimodal slowly varying characteristic data, a dissipation assessment calculation is performed to generate a global fatigue damping coefficient, including: Based on a preset time integration window, a weighted integral calculation is performed on the first state sub-feature and the second state sub-feature to extract local microscopic dissipation scalars; The local microscopic dissipation scalar is input into the boundary constraint function and subjected to nonlinear normalization mapping to output the global fatigue damping coefficient. Extracting local microscopic dissipation scalars specifically includes the following steps: Within the preset time integration window, environmental background run parameters characterizing the baseline of ambient temperature fluctuations are obtained; Obtain the preset initial thermal stress conversion coefficient and initial mechanical equivalent normalization factor; Based on the environmental background run parameters, a nonlinear compensation calculation is performed on the initial thermal stress conversion coefficient to generate a dynamic thermal stress weighting factor. Using the dynamic thermal stress weighting factor and the initial mechanical equivalent normalization factor, dimensional unification multiplication operations are performed on the first state sub-feature and the second state feature, respectively; Obtain the corresponding discrete sampling time step parameter; perform algebraic accumulation on the first state sub-feature within the preset time integration window after dimensional unification multiplication operation, and multiply the accumulation result with the discrete sampling time step parameter to obtain the heat dissipation integral; The mechanical wear amount is obtained by performing algebraic accumulation on the second state sub-feature within the preset time integration window after dimensional unification multiplication operation; the thermal dissipation integral is added to the mechanical wear amount to extract the local microscopic dissipation scalar.

3. The rectifier transformer coordinated voltage regulation and control method according to claim 2, characterized in that: Generating a dynamic thermal stress weighting factor specifically includes the following steps: Calculate the absolute deviation scalar between the environmental background run parameter and the preset standard reference environmental temperature threshold; determine whether the absolute deviation scalar is greater than the preset drift tolerance threshold; If the judgment result is greater than the drift tolerance threshold, a preset exponential amplification base is obtained, the absolute deviation scalar is used as the exponent, and the exponential operation is performed on the exponential amplification base to obtain the nonlinear compensation term. The nonlinear compensation term is multiplied by the initial thermal stress conversion coefficient to output the dynamic thermal stress weighting factor.

4. The rectifier transformer coordinated voltage regulation and control method according to claim 3, characterized in that: The electrical transient fluctuation data includes the DC bus voltage transient drop amplitude and the converter output current change rate; Based on the electrical transient fluctuation data and the current load capacity of each transformer node, a local transient voltage compensation scalar is calculated, including: The negative pressure gradient of the bus voltage is calculated based on the transient drop amplitude of the DC bus voltage and the rate of change of the converter output current. Calculate the absolute difference between the current load capacity and the negative pressure gradient of the bus voltage, and extract the local transient voltage compensation scalar. The local transient voltage compensation scalar is broadcast and distributed to adjacent transformer nodes in the physical topology via a single-hop horizontal communication link. The joint compensation mapping logic performs negative exponential mapping and scalar multiplication operations, outputting a dynamic compensation weight matrix, including: Calculate the attenuation mapping value with the base of the natural logarithm as the base and the negative global fatigue damping coefficient as the exponent; The effective compensation output value of the node is extracted by performing Hamiltonian scalar product operation on the local transient voltage compensation scalar and the attenuation mapping value. Calculate the differential gradient matrix of the effective compensation output value of the transformer node between the current transformer node and the adjacent transformer nodes in the topology.

5. The rectifier transformer coordinated voltage regulation and control method according to claim 4, characterized in that: After extracting the difference gradient matrix, the distributed security reset decision logic is executed: Obtain the preset threshold for the conservation of potential energy across the entire network; Obtain the rated apparent capacity of this transformer node and the total installed capacity of the entire network, calculate the ratio of the rated apparent capacity to the total installed capacity of the entire network, and obtain the capacity weighting ratio coefficient. Multiply the network-wide minimum potential energy conservation threshold by the capacity weight ratio coefficient to obtain the node-level equivalent potential energy quota. The sum of all positive pressure difference values ​​in the differential gradient matrix is ​​calculated to obtain the total local leakage potential energy. If the total local leakage potential energy is less than the node-level equivalent potential energy quota, the absolute value of the local global fatigue damping coefficient is reduced by a preset reduction ratio, and the total local leakage potential energy is recalculated iteratively. If the sum of the local leakage potential energy is greater than or equal to the node-level equivalent potential energy quota, the positive gradient components with values ​​greater than zero in the differential gradient matrix are extracted to generate the dynamic compensation weight matrix.

6. The rectifier transformer coordinated voltage regulation and control method according to claim 5, characterized in that: Based on the dynamic compensation weight matrix, a transient cooperative control command is generated to instruct the underlying converter to perform a trigger angle bias action, including: Obtain the reference steady-state firing angle; extract the target weight components corresponding to the target underlying converter from the dynamic compensation weight matrix; The target weight components are input into the trigger angle mapping function to calculate the phase angle offset compensation amount; The transient target firing angle is extracted by subtracting the phase angle offset compensation from the reference steady-state firing angle. The transient cooperative control command is generated based on the transient target trigger angle. The transient cooperative control command is configured to output a high-frequency pulse sequence to drive the physical switching devices of the target underlying converter.

7. The rectifier transformer coordinated voltage regulation and control method according to claim 6, characterized in that: The target weight component is input into the trigger angle mapping function to calculate the phase angle offset compensation amount, specifically including: Obtain the preset commutation failure safety margin parameter and nonlinear mapping scaling factor; perform subtraction calculation based on the reference steady-state trigger angle and the commutation failure safety margin parameter to extract the maximum allowable forward phase angle limit; The target weight components are multiplied using the nonlinear mapping scaling factor to obtain a normalized intermediate weight with unified dimensions. Using the normalized weight intermediate as an independent variable, the inverse cosine mapping conversion logic is executed to obtain the theoretical phase angle offset. Determine whether the theoretical phase angle offset is greater than the maximum allowable forward phase angle limit; If the judgment result is greater than a certain value, the maximum allowable forward phase angle limit will be output as the phase angle offset compensation amount. If the judgment result is less than or equal to the theoretical phase angle offset, the theoretical phase angle offset is output as the phase angle offset compensation.

8. The rectifier transformer coordinated voltage regulation and control method and the rectifier transformer for aluminum foil forming according to claim 7, characterized in that: When the data update cycle of the global fatigue damping coefficient exceeds a preset communication time threshold, a damping attenuation degradation strategy is triggered. This involves decreasing the global fatigue damping coefficient according to a preset attenuation function and reconstructing the dynamic compensation weight matrix, including: Read the status parameters of the local communication watchdog timer; when the status parameters indicate that the data update period exceeds a preset communication time threshold, generate an interruption alarm signal representing a physical abnormality of the upper-layer data link; In response to the interruption alarm signal, the local state overwrite process is activated, so that the current value of the global fatigue damping coefficient is continuously reduced to the lower limit of zero according to the preset exponential time constant. The global fatigue damping coefficient during the decreasing process is input into the joint compensation mapping logic in real time and iteratively, forcing the dynamic compensation weight matrix to degenerate into a first backup topology controlled by the local transient voltage compensation scalar distribution.

9. The rectifier transformer coordinated voltage regulation and control method and the rectifier transformer for aluminum foil forming according to claim 8, characterized in that: Activating the local state overwrite process, causing the current value of the global fatigue damping coefficient to continuously decrease to the lower limit of zero according to a preset exponential time constant, specifically includes: Before generating the interruption alarm signal, the last validly updated global fatigue damping coefficient and the corresponding global fatigue damping coefficient value of the previous historical period are extracted from the local historical cache stack. Perform a difference calculation between the global fatigue damping coefficient of the last effective update and the global fatigue damping coefficient value of the previous historical cycle to extract the characteristic value of damping degradation rate. Get the basic configuration time constant; get the preset sensitivity adjustment factor; The damping degradation rate characteristic value is multiplied by the sensitivity adjustment factor to extract a dimensionless attenuation multiplier; the dimensionless attenuation multiplier is added to the natural constant to generate an adaptive denominator operator; the basic configuration time constant is divided by the adaptive denominator operator to extract the quotient as the dynamic degradation time constant; the dynamic degradation time constant is used as the preset exponential time constant. Within the data update cycle, the internal digital decay integral logic is initiated with the discrete control step size as the time iteration unit. Obtain the discrete control step size and the natural logarithm base constant; divide the discrete control step size by the dynamic degradation time constant to extract the single-step decay exponent amplitude; use the negative of the single-step decay exponent amplitude as the exponent and perform a power operation on the natural logarithm base constant to extract the absolute discrete decay multiplier constant. At the triggering time of each discrete control step, the historical retention value of the global fatigue damping coefficient is extracted; the historical retention value is multiplied by the absolute discrete attenuation multiplier constant to obtain the current value after attenuation; The decayed current value is overwritten as the historical retention value for the next iteration; the multiplication and overwrite operations are performed repeatedly until it is determined that the decayed current value is less than or equal to a preset lower limit of zero.

10. A rectifier transformer for aluminum foil forming, characterized in that: Includes the transformer body and an edge computing gateway that is communicatively connected to the transformer body; The edge computing gateway contains a memory and a microprocessor, and the memory stores computer programs. When the microprocessor executes the computer program, it implements the rectifier transformer coordinated voltage regulation and control method as described in any one of claims 1-9.

Images