A network-configuration type converter virtual impedance compensation control method and system

Abstract

This invention relates to the field of parallel control technology for grid-connected converters. The invention discloses a virtual impedance compensation control method and system for grid-connected converters. It acquires the state frames of the synchronous branches of each parallel converter, generates an initial compensation request based on the capacity-normalized current sharing error, and removes the common voltage component according to the branch incremental admittance. When the previously submitted compensation set is currently feasible, a final compensation set is formed using the minimum allowable ratio of the entire group. This set is executed at the common effective time after confirmation by all members and a consensus multicast. If verification fails, the coordinated compensation is revoked and the uncompensated voltage reference is limited. The system consists of state frame generation, compensation frame generation, compensation execution, and power supply control modules. This invention can maintain the weighted sum of the admittances of the target compensation voltage set at zero under conditions of asymmetric feeder admittance and modulation margin, reducing first-order additional disturbances on the common bus and suppressing the effects of capacity sharing errors, internal circulating currents, and asynchronous switching.

Description

Technical Field

[0001] This invention relates to the field of parallel power supply control technology for grid-type converters, specifically to a virtual impedance compensation control method and system for grid-type converters. Background Technology

[0002] Grid-type converters can establish an AC voltage reference through droop control, virtual synchronous machine control, or virtual oscillator control, and then connect to a common AC bus in parallel via an AC filter and feeders. In parallel power supply scenarios, existing technologies typically collect the local current, power, and voltage status of each converter, adjust the virtual impedance or voltage reference based on the local deviation, and have each converter perform limiting and modulation control separately to improve current or power distribution among parallel power sources and maintain bus power supply.

[0003] When the rated capacity, feeder impedance, and available modulation voltage margin of each converter are inconsistent, and communication delays exist, the control methods relying on independent adjustment, independent limiting, and time-sharing activation of the converter's deviation are insufficient to simultaneously handle the coupling relationship between branch electrical parameters, real-time voltage boundaries, and compensation switching timing. The compensation values ​​of each converter, after being mapped through different feeders, may generate non-zero composite current excitation on the common bus side. If a converter reaches its voltage boundary first and then limits its current, it will alter the original coordination relationship. Furthermore, the activation of old and new compensation values ​​at different control times will generate additional branch voltage differences and internal circulating currents. Therefore, it is necessary to address how to simultaneously consider capacity sharing, common bus side excitation suppression, overall voltage feasibility, and synchronous execution under conditions of asymmetrical branch parameters, asymmetrical voltage constraints, and bounded communication delays. Summary of the Invention

[0004] In a first aspect, the present invention provides a virtual impedance compensation control method for a grid-type converter, applicable to multiple grid-type converters connected in parallel to a common AC bus via feeders, comprising: S1. Obtain the synchronization branch status frame, the previous submitted compensation frame, and the previous submitted compensation voltage set. S2. Generate an original compensation request based on the branch current and rated current in the synchronous branch status frame. Remove the common voltage component of the original compensation request according to the branch incremental admittance matrix bound to the synchronous branch status frame to obtain a target compensation voltage set with a weighted sum of zero admittance. Verify whether the previously submitted compensation voltage set belongs to the entire feasible set composed of the Cartesian product of the member constraint sets. If the verification passes, update the target compensation voltage set according to the minimum allowable proportion of the entire group to obtain the final compensation voltage set. After confirmation by all members and consistent multicast delivery, a virtual impedance compensation frame containing the final compensation voltage set and the common effective time is formed. S3. During unified multicast delivery, at the common effective time, the virtual impedance compensation frame is activated to form an effective compensation frame. When unified multicast is terminated, the previous submitted compensation frame is taken as the effective compensation frame. The voltage reference formation includes: when the feasibility check of the entire set of the previous submitted compensation voltages passes, the local compensation voltage in the effective compensation frame is transformed to the local coordinate system and subtracted from the local uncompensated voltage reference to form a compensated voltage reference value; when the feasibility check of the entire set of the previous submitted compensation voltages fails, the coordinated compensation is canceled in the next control cycle, the local uncompensated voltage reference is limited, and the compensated voltage reference value is formed. S4. Control the converter output according to the compensated voltage reference value and update the synchronous branch status frame.

[0005] Optionally, obtaining the synchronization branch status frame, the previous submitted compensation frame, and the previous submitted compensation voltage set specifically includes: The positive-sequence fundamental component of the branch current is extracted from the same fundamental observation window of each effective member, and the uncompensated voltage reference at the same sampling time is obtained. A two-dimensional rotation matrix is ​​determined based on the difference between the phase angle of the local coordinate system and the phase angle of the common coordinate system. The two-dimensional rotation matrix is ​​then used to transform the positive sequence fundamental component and the uncompensated voltage reference, respectively. The transformation results of each valid member, the sampling timestamp, the member set version, the branch incremental admittance matrix version, and the rated current and voltage constraint parameters are written into the same state frame to form the synchronous branch state frame; the currently submitted compensation frame is determined as the previous submitted compensation frame, and the previous submitted compensation voltage set is read from it.

[0006] Optionally, the step of generating an original compensation request based on the branch current and rated current in the synchronization branch state frame, and removing the common voltage component of the original compensation request according to the branch incremental admittance matrix bound to the synchronization branch state frame to obtain a target compensation voltage set with a weighted admittance sum of zero, specifically includes: Using the rated current, the branch current is normalized to form a normalized current sharing error, and the original compensation request is determined based on the compensation gain matrix and the normalized current sharing error. Based on the branch incremental admittance matrix bound to the synchronous branch state frame and the original compensation request, the common voltage component is determined, and the target compensation voltage set is formed using the common voltage component.

[0007] Optionally, the step of determining the common voltage component based on the branch incremental admittance matrix bound to the synchronization branch state frame and the original compensation request, and using the common voltage component to form the target compensation voltage set, specifically includes: set up For the number of valid members, For valid member serial numbers, For valid member serial numbers, and The average value is to According to the rated current Calculate capacity weights Based on the branch currents in the common coordinate system Calculate the normalized current of the capacity Capacity-weighted average And capacity normalized current sharing error , and according to The original compensation request is formed, wherein, This is the compensation gain matrix with the dimension of voltage; exist When reversible, the original compensation request is used according to... The common voltage component is formed, and according to... Forming the target compensation voltage ,make ,in, The dimensionless quantity of the first is conductance. The branch incremental admittance matrix of each effective member is determined by the target compensation voltage. The target compensation voltage set is formed.

[0008] Optionally, when the verification passes, the target compensation voltage set is updated according to the minimum allowable proportion of the entire group to obtain the final compensation voltage set, specifically including: set up The valid member number; voltage constraint parameters include the maximum permissible fundamental modulation voltage in the dimension of voltage. and the upper limit of compensation voltage amplitude The upper limit of the compensated voltage change rate, with dimensions of voltage change rate. and the coordinate period with the dimension of time ; Using the uncompensated voltage in the state frame of the synchronous branch as a reference Conservative margin with dimensions of voltage The previously submitted compensation voltage Determine the first The set of membership constraints for each valid member. , and according to Form the complete feasible set, wherein, For the first The target compensation voltage for each effective member; If the previously submitted set of compensation voltages belongs to Then seek to make right The largest established ,Pick , and according to This forms the final set of compensation voltages.

[0009] Optionally, the final set of compensation voltages is calculated by the compensation coordinator using the synchronization branch state frame as input. The virtual impedance compensation frame is generated by the compensation coordinator, which generates a preparation frame, collects confirmations from all valid members, and forms a submission certificate. The submission certificate is then verified through the redundant consistent multicast service to obtain the consistent multicast delivery, and then the frame is formed.

[0010] Optionally, the final compensation voltage set is calculated independently by each effective member according to the same member sorting, fixed-point number format, multiplication and addition order, rounding rules and serialization rules, and the calculation summaries of each effective member are compared to form a summary consistency result; The virtual impedance compensation frame is formed when the digest consensus result is consistent. The rotating proposal member proposes the frame number and common effective time without modifying the final compensation voltage set, collects the confirmation of all valid members and forms a submission certificate, and then verifies the submission certificate through the consensus multicast service and obtains the consensus multicast delivery.

[0011] Optionally, the consistent multicast delivery is determined when each valid member confirms a preparation frame containing the ordered member set, the final compensation voltage set, the previous frame summary, and the provisional co-effective time, collects all confirmations to form a commit certificate, and the consistent multicast service delivers the same commit certificate to all valid members in the ordered member set before the guardian deadline of the co-effective time, and returns a full delivery decision to all valid members. The consensus multicast termination is determined when any of the aforementioned conditions are not met, and the termination decision is returned to all valid members, so that all valid members retain the previously submitted compensation frame and end the current round of submission. In the next coordination cycle, the previously submitted compensation voltage set is re-verified to see if it belongs to the feasible set of the whole group. If the verification fails, the coordinated compensation is cancelled in the next local control cycle, and the uncompensated voltage reference of the local unit is limited.

[0012] Optionally, when the verification passes, the local compensated voltage in the effective compensation frame is transformed to the local coordinate system and subtracted from the local uncompensated voltage reference to form a compensated voltage reference value, specifically including: set up The effective member number is denoted as and the local compensation voltage expressed in the common coordinate system in the effective compensation frame is recorded as . ;according to ,Will Transformed into local compensated voltage in local coordinate system ,in, The phase angle is in the common coordinate system. The phase angle of the local coordinate system. It is a two-dimensional rotation matrix; The local compensation voltage is used as the only new execution quantity, and the equivalent resistance or equivalent reactance calculated from the local compensation voltage is only used for monitoring. Perform a base virtual impedance on-premises to form an on-premises uncompensated voltage reference. , and according to The compensated voltage reference value is then formed.

[0013] Secondly, the present invention provides a virtual impedance compensation control system for a grid-type converter, used to implement the above method, comprising: The status frame forming module is used to obtain the status frame of the synchronization branch, the previous submitted compensation frame, and the previous submitted compensation voltage set. The compensation frame forming module is used to generate an original compensation request based on the branch current and rated current in the synchronous branch status frame, remove the common voltage component of the original compensation request according to the branch incremental admittance matrix bound to the synchronous branch status frame, obtain a target compensation voltage set with a weighted sum of zero admittance, verify whether the previously submitted compensation voltage set belongs to the entire feasible set composed of the Cartesian product of the member constraint sets, and if the verification passes, update the target compensation voltage set according to the minimum allowable proportion of the entire group to obtain the final compensation voltage set. After confirmation by all members and consistent multicast delivery, a virtual impedance compensation frame containing the final compensation voltage set and the common effective time is formed. The compensation execution module is used to activate the virtual impedance compensation frame during the common effective time when consistent multicast delivery occurs, forming an effective compensation frame. When consistent multicast is terminated, the previous submitted compensation frame is used as the effective compensation frame. The voltage reference formation includes: when the feasibility check of the entire set of the previous submitted compensation voltages passes, transforming the local compensation voltage in the effective compensation frame to the local coordinate system and subtracting it from the local uncompensated voltage reference to form a compensated voltage reference value; when the feasibility check of the entire set of the previous submitted compensation voltages fails, canceling the coordinated compensation in the next control cycle, limiting the local uncompensated voltage reference, and forming the compensated voltage reference value. The power supply control module is used to control the converter output according to the compensated voltage reference value and update the synchronous branch status frame.

[0014] Beneficial effects The technical solution provided by this invention has the following beneficial effects: 1. For parallel-connected grid converters with different branch incremental admittances, the original compensation request formed by the capacity normalized current sharing error is adjusted by removing the common voltage component from the branch incremental admittance. This makes the admittance weighted sum of the target compensation voltage set zero while retaining the relative difference between the requests of each branch. Therefore, within the linearized operating point and coordination bandwidth corresponding to the admittance parameter version, the compensation voltage does not form a first-order synthetic current excitation on the common bus side, while still adjusting along the negative feedback direction of the capacity proportional sharing error. This helps reduce the incidental disturbance of the compensation update to the bus voltage and suppresses internal circulating currents.

[0015] 2. In cases where the real-time voltage boundaries of each member are inconsistent, first verify that the current feasibility of the previously submitted compensation voltage set is met, and then take the minimum allowable proportion of the entire group along the line segment from it to the target set, so that all members use the same interpolation factor to meet the constraints of modulation voltage, compensation amplitude and rate of change, while maintaining the admittance weighting relationship; when the starting point fails, cancel the coordinated compensation and limit the uncompensated voltage reference to avoid continuing interpolation from an infeasible starting point.

[0016] 3. Within the normal operating domain where bounded communication delay and common time reference are established, through full confirmation, certificate submission, and consistent multicast results, all valid members can enable the same compensation frame at the common effective time, or jointly maintain the previously submitted compensation frame; combined with the transformation from the common coordinate system to the local coordinate system, the branch voltage difference and update circulating current caused by the mixing of old and new compensation values ​​can be reduced. Attached Figure Description

[0017] Figure 1 A schematic diagram of a multi-grid converter parallel power supply deployment scenario provided by the present invention; Figure 2 A schematic diagram illustrating the steps of a virtual impedance compensation control method for a grid-type converter provided by the present invention; Figure 3 This invention provides an overall flowchart of a virtual impedance compensation control method for a grid-type converter. Figure 4 This is a schematic diagram of the virtual impedance compensation frame formation steps provided by the present invention; Figure 5 This is a schematic diagram illustrating the timing interaction between compensation frame submission and joint activation provided by the present invention; Figure 6 This invention provides a schematic diagram of the structure of a virtual impedance compensation control system for a grid-type converter. Figure 7 The feasible region response diagram for the entire group common interpolation provided by this invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] like Figure 1 As shown, this embodiment applies to at least two grid-type converters connected in parallel to a common AC bus via AC filters and feeders. Taking three grid-type energy storage converters as an example, each converter consists of a DC terminal, a power semiconductor bridge, an AC filter, and a feeder. The feeder ends are connected to the same common AC bus and supply power to the AC load. The rated capacity, feeder impedance, and available modulation voltage margin of the three converters can be different. Each grid controller reads its own power setpoint, measured active power, reactive power, and voltage status. It first calculates the difference between the setpoint and the measured value, and then uses the state update relationship corresponding to droop control, virtual synchronous machine control, or virtual oscillator control to calculate the phase angle and voltage amplitude of the converter, generating the local AC voltage reference. Within this voltage reference channel, the local basic virtual impedance is first executed.

[0020] The output current of each converter enters the network controller via a current sensor, isolation conditioning circuit, anti-aliasing filter, and analog-to-digital converter. The AC output voltage and DC bus voltage are acquired in the same manner. A digital signal processor, field-programmable gate array (FPGA), or controller with equivalent real-time computing capabilities can all serve as the network controller. The fast voltage and current control program is stored in the local non-volatile memory, while the sampled values, rotating coordinates, and control state for the current cycle are stored in the running memory. Taking a system with a rated frequency of 50Hz as an example, the sampling frequency can be 20kHz, the positive-sequence fundamental frequency observation window can be 20ms, and the coordination period can be 20ms. In a 60Hz system, the observation window is correspondingly one fundamental frequency cycle, and the coordination period is not shorter than this observation window.

[0021] Each network controller connects to the deterministic communication network via a coordination communication interface. The compensation coordination node sequentially encodes, sends, and verifies synchronization tributary status frames, preparation frames, member acknowledgments, and certificate submissions. The deterministic communication network can employ dual-redundant time-sensitive Ethernet with a common time base and bounded group delivery delay, or it can use an intra-site bus capable of providing the same all-member delivery or all-member abort semantics. The compensation coordination node performs member and parameter version management, centralized compensation calculation, and group submission. When using distributed computing, each valid member reads the same synchronization tributary status frame and independently calculates the compensation value after reading the member status data in the order of the member identifier. The rotating proposal member only proposes the frame sequence number and common effective time after every byte of all digests is equal to the byte corresponding to the first member's digest.

[0022] like Figure 2 As shown, this embodiment provides a virtual impedance compensation control method for a grid-type converter, applied to multiple grid-type converters connected in parallel to a common AC bus via feeders, including: S1. Obtain the synchronization branch status frame, the previous submitted compensation frame, and the previous submitted compensation voltage set. S2. Generate an original compensation request based on the branch current and rated current in the synchronous branch status frame. Remove the common voltage component of the original compensation request according to the branch incremental admittance matrix bound to the synchronous branch status frame to obtain a target compensation voltage set with a weighted sum of zero admittance. Verify whether the previously submitted compensation voltage set belongs to the entire feasible set composed of the Cartesian product of the member constraint sets. If the verification passes, update the target compensation voltage set according to the minimum allowable proportion of the entire group to obtain the final compensation voltage set. After confirmation by all members and consistent multicast delivery, a virtual impedance compensation frame containing the final compensation voltage set and the common effective time is formed. S3. During unified multicast delivery, at the common effective time, the virtual impedance compensation frame is activated to form an effective compensation frame. When unified multicast is terminated, the previous submitted compensation frame is taken as the effective compensation frame. The voltage reference formation includes: when the feasibility check of the entire set of the previous submitted compensation voltages passes, the local compensation voltage in the effective compensation frame is transformed to the local coordinate system and subtracted from the local uncompensated voltage reference to form a compensated voltage reference value; when the feasibility check of the entire set of the previous submitted compensation voltages fails, the coordinated compensation is canceled in the next control cycle, the local uncompensated voltage reference is limited, and the compensated voltage reference value is formed. S4. Control the converter output according to the compensated voltage reference value and update the synchronous branch status frame.

[0023] For S1, specifically, each network controller latches the sampling time within the same coordination period. The corresponding three-phase branch currents, AC output voltage, DC bus voltage, uncompensated voltage reference, and modulation state are recorded. The three-phase branch currents are first calibrated for amplitude and zero-bias correction, and then transformed into the local dq coordinate system via synchronous rotating coordinate transformation. The controller synchronously averages the local dq current within a sliding observation window of at least one fundamental frequency cycle, removing DC bias and non-fundamental components, and outputs a two-dimensional vector of positive-sequence fundamental current. In case of sampling phase loss, analog-to-digital conversion saturation, timestamp jump, or insufficient effective samples in the observation window, the controller writes the current cycle's operating state as invalid and stops generating state frames that can participate in normal coordination.

[0024] The uncompensated voltage reference for this machine is taken from the two-dimensional voltage command after the basic virtual impedance and before the coordinated compensation subtraction node. The maximum permissible fundamental modulation voltage is calculated by multiplying the current DC bus voltage and the linear modulation coefficient allowed by the modulation method, and then converting it to the fundamental voltage range. The protection margin of this machine is based on the conservative margin described later. Individual deduction; the upper limit of the compensation amplitude is read from the voltage permission table of the coordinated function, which matches the member identifier and parameter version in volts. The upper limit of the compensation change rate is obtained by dividing the allowed single-cycle voltage step by the coordinated cycle and is verified by the voltage loop bandwidth. The member set version, branch incremental admittance matrix version, and compensation gain version are loaded by the member and parameter version management module; if any version is missing, mismatched, or invalid, the status frame of this cycle is only used for diagnosis and does not enter S2.

[0025] Each valid member forms a member status subframe. Each member status subframe uses a member identifier and a sampling timestamp as its unique key, and includes at least the member identifier, member set version, admittance parameter version, compensation gain version, sampling timestamp, common coordinate system identifier, positive-sequence fundamental current in the common coordinate system, uncompensated voltage reference in the common coordinate system, maximum permissible fundamental modulation voltage, rated current, upper limit of compensation amplitude, upper limit of compensation rate of change, operating status, currently submitted frame summary, and integrity check code. These fields in a normal frame are not allowed to be empty. Two-dimensional current and voltage quantities can be represented using a uniformly scaled signed fixed-point number. The branch incremental admittance matrix is ​​obtained from the admittance parameter version index and used for subsequent compensation calculations according to the corresponding numerical format. The timestamp is represented as a 64-bit integer under a common time base. The receiving end sequentially verifies the length, integrity check code, unique key, version, and synchronization window of each member status subframe. For duplicate member status subframes, only the one with the most recent sequence number and passed integrity verification is retained. The current preparation round is terminated when a member status subframe is missing or the timestamp dispersion exceeds the synchronization window.

[0026] In some implementations, obtaining the synchronization branch status frame, the previous submitted compensation frame, and the previous submitted compensation voltage set specifically includes: The positive-sequence fundamental component of the branch current is extracted from the same fundamental observation window of each effective member, and the uncompensated voltage reference at the same sampling time is obtained. A two-dimensional rotation matrix is ​​determined based on the difference between the phase angle of the local coordinate system and the phase angle of the common coordinate system. The two-dimensional rotation matrix is ​​then used to transform the positive sequence fundamental component and the uncompensated voltage reference, respectively. The transformation results of each valid member, the sampling timestamp, the member set version, the branch incremental admittance matrix version, and the rated current and voltage constraint parameters are written into the same state frame to form the synchronous branch state frame; the currently submitted compensation frame is determined as the previous submitted compensation frame, and the previous submitted compensation voltage set is read from it.

[0027] Specifically, all valid members use the same observation window start and end time and the same positive sequence extraction caliber. The phase angle of the common coordinate system is provided by a common rotating coordinate generator corresponding to the common time base, and the phase angle of the local coordinate system is provided by the network controller of the organization. The controller uses the local phase angle minus the common phase angle as the positive transformation angle, calculates the sine and cosine values ​​of this angle, and assembles a two-dimensional rotation matrix. First, the local positive sequence fundamental current is multiplied by this matrix on the left, and then the local uncompensated voltage reference at the same sampling time is multiplied by the same matrix on the left, thus obtaining a two-dimensional vector of the common coordinate system with consistent caliber. Angle calculations use a unified radian caliber, and after table lookup or coordinate rotation calculations are completed, the quantization is performed to the configured decimal places. When the angle is lost, the frame is marked as invalid.

[0028] Each member encodes the two transformed two-dimensional vectors, along with the aforementioned version, time, and constraint fields, into a member state subframe. When the compensation calculation node aggregates the member state subframes, they are arranged in ascending order by member identifier. In a usable implementation, current and voltage use Q16.16 format, admittance uses Q8.24 format, and angle uses Q3.29 format, employing two's complement, little-endian byte order, and nearest-even rounding. The currently submitted compensation frame is read from the local submission log, which is indexed by frame sequence number and stores the submission certificate digest, common effective time, and complete compensation voltage set. The controller selects the record with the largest frame sequence number that has obtained a full member delivery decision, reached the common effective time, completed the current execution area update, and is the previous submitted compensation frame, and reads the previous submitted compensation voltage set from its ordered member voltage field. When there is no valid record in the log, the all-zero compensation frame corresponding to the member set version is used as the initialization frame, and the system only enters the normal preparation state when the zero compensation of all members meets the current voltage feasibility condition.

[0029] For S2, specifically, as Figure 4 As shown, the compensation calculation node first searches for the member status subframe corresponding to each valid member in the ordered list of valid members, and removes member status subframes whose member identifier is not in the ordered list, whose version is inconsistent, whose timestamp exceeds the synchronization window, whose running status is abnormal, or whose integrity verification fails. Only when each member in the ordered list retains a valid member status subframe are the valid member status subframes encapsulated in ascending order of member identifier to form the synchronization branch status frame; otherwise, missing or conflicting member identifiers are recorded, the previously submitted compensation frame is retained, and the current round of preparation ends. Subsequently, the compensation calculation node sequentially reads the rated current, the common coordinate system branch current, the compensation gain matrix, and the branch incremental admittance matrix, and calculates the capacity normalized current sharing error, the original compensation request, the common voltage component, and the target compensation voltage set.

[0030] After obtaining the target compensation voltage set, the compensation calculation node reads the uncompensated voltage reference, maximum allowable fundamental modulation voltage, compensation amplitude upper limit, compensation change rate upper limit, coordination period, and the previously submitted compensation voltage for each member. It then checks whether the member set version and branch incremental admittance matrix version of the previously submitted compensation frame are consistent with the version corresponding to the current synchronization branch state frame. If the versions are inconsistent, the current round of common interpolation is not performed, and the process is handled according to the member or admittance version switching procedure. If the versions are consistent, a constraint set for each member is constructed, and the feasibility of the previously submitted compensation voltage set is checked. If all starting points are feasible, the maximum allowable proportion for each member is calculated along the same line segment from the previous set to the target set, and the minimum value is selected as the common interpolation factor for the entire group. A candidate compensation voltage set is then generated based on this factor.

[0031] After fixed-point quantization of the candidate compensation voltage set, the nodes calculate the admittance weighted residuals according to the same admittance version and a fixed multiply-accumulate order. The residual norm exceeds the acceptance threshold. First, calculate the common correction value. Execute on all members Then, round to zero to the configured fixed-point format, and recalculate the admittance-weighted residual and the three constraints for each member. The common correction is used to eliminate the common residual introduced by fixed-point quantization without changing the difference between any two member compensation voltages. After quantization and necessary common correction, when the admittance-weighted residual does not exceed the acceptance threshold and all member constraints are passed, the resulting compensation voltage set is determined as the final compensation voltage set.

[0032] In this embodiment, the vector residual adopts the L2 norm, and the matrix norm adopts an induced L2 norm compatible with the L2 norm. When the condition is reversible and the condition number does not exceed the current value, the upper limit of the allowable first-order bus voltage disturbance caused by compensation is divided by... The electrical upper bound of the residual acceptance threshold is obtained, where, The incremental admittance of the common bus load within the same operating point and compensation bandwidth; when the load has a varying range, an empirical upper bound for the norm of the inverse matrix is ​​determined for the set of load parameters within the version support range. If If the process is irreversible or the condition number exceeds the allowable upper limit, loading the corresponding load parameter and admittance parameter versions is prohibited. The quantization lower bound is calculated item by item, by multiplying the least significant bit of the admittance quantization by the upper limit of the compensation amplitude, the least significant bit of the compensation voltage quantization by the admittance norm, and the maximum rounding error of a single fixed-point multiplication-addition, and then summing the worst-case terms of all members. The electrical upper bound is quantized down to the residual comparison format to obtain the candidate threshold; if the candidate threshold is not less than the quantization lower bound, it is uniquely written as... Otherwise, loading this numerical format and parameter version is prohibited.

[0033] If the admittance-weighted residual or any membership constraint still fails after common correction and requantization, the node is set to... The shared interpolation factor for the entire group is rounded down to a configured fixed-point quantum, and the candidate compensation voltage set generation, quantization, common correction, and full verification are re-executed. Here, m is the recalculation sequence number, and the maximum number of recalculations is fixed in the compensation gain version; in this implementation, it is taken as 16 times. Both centralized and distributed implementations use this rollback order, the same stopping condition, and the same version number. When the recalculation limit is reached, the interpolation factor is quantized to zero and still fails, or the matrix operation overflows, the final residual, restricted members, failure constraints, and abort status code are written to the log of this round, the pending submission cache is cleared, and the unverified set is not sent to the execution end.

[0034] Further, the step of generating an original compensation request based on the branch current and rated current in the synchronization branch state frame, and removing the common voltage component of the original compensation request according to the branch incremental admittance matrix bound to the synchronization branch state frame to obtain a target compensation voltage set with a weighted admittance sum of zero, specifically includes: Using the rated current, the branch current is normalized to form a normalized current sharing error, and the original compensation request is determined based on the compensation gain matrix and the normalized current sharing error. Based on the branch incremental admittance matrix bound to the synchronous branch state frame and the original compensation request, the common voltage component is determined, and the target compensation voltage set is formed using the common voltage component. Specifically, the compensation calculation node first divides the two-dimensional branch current in the common coordinate system of each member by the rated current of that member, so that the currents of members with different capacities enter the same dimensionless caliber; then, it calculates the capacity weight according to the proportion of the rated current to the sum of the rated currents of the entire group, multiplies each capacity normalized current by the capacity weight item by item and sums them to form a capacity weighted average. The capacity normalized current sharing error is obtained by subtracting this average value from the capacity normalized current of each member. The error vector is multiplied by the compensation gain matrix of this version to form the original compensation request in volts. If any rated current is less than or equal to zero, the input current overflows the fixed-point representation range, or the gain version fails the loading check, the node records a parameter error and ends the preparation round.

[0035] Subsequently, nodes calculate and accumulate the product of the branch incremental admittance matrix and the original compensation request item by item in ascending order of member identifier, while simultaneously accumulating the incremental admittance matrices of all branches. After the sum of the admittance matrices is checked for minimum singularity and condition number, a two-dimensional matrix inversion is performed. The inverse matrix multiplied by the accumulated sum of the aforementioned request multiplications yields the common voltage component. Each original compensation request is subtracted from the same common voltage component to form the target compensation voltage for the corresponding member. If the sum of the admittance matrices is not invertible, the minimum singularity is below the loading threshold, or the inversion result overflows, no target set is generated in this cycle, the system retains the previously submitted compensation frame, and writes the admittance version status as invalid.

[0036] Optionally, the step of determining the common voltage component based on the branch incremental admittance matrix bound to the synchronization branch state frame and the original compensation request, and using the common voltage component to form the target compensation voltage set, specifically includes: set up For the number of valid members, For valid member serial numbers, For valid member serial numbers, and The average value is to According to the rated current Calculate capacity weights Based on the branch currents in the common coordinate system Calculate the normalized current of the capacity Capacity-weighted average And capacity normalized current sharing error , and according to The original compensation request is formed, wherein, This is the compensation gain matrix with the dimension of voltage; exist When reversible, the original compensation request is used according to... The common voltage component is formed, and according to... Forming the target compensation voltage ,make ,in, The dimensionless quantity of the first is conductance. The branch incremental admittance matrix of each effective member is determined by the target compensation voltage. The target compensation voltage set is formed.

[0037] Specifically, rated current Read from the same member parameter version, using amperes as the unit; branch current. Normalized current of capacity Capacity-weighted average ,error Original compensation request Common voltage component and target compensation voltage All are calculated and stored item by item according to the ordered membership list. Compensation gain matrix A positive definite scalar multiplied by an identity matrix can be used, or a general two-dimensional matrix with a positive definite symmetric component can be used. Before loading, a linearized model containing voltage loop, current loop, sample and hold, AC filter, feeder, and maximum communication calculation delay is used for frequency response verification to ensure that the compensation closed-loop crossover frequency is not higher than one-fifth of the minimum crossover frequency of the primary voltage control of each member, and the configured phase margin and damping lower limit are retained within the parameter tolerance range.

[0038] Branch incremental admittance matrix This represents a positive-sequence two-dimensional mapping within the compensation coordination bandwidth from the voltage difference increment between the member's internal voltage reference and the common bus voltage to the common bus injected current increment, satisfying... The unit is Siemens. When using the RL parameter path, first, in the positive direction of the voltage drop consistent with the S1 common dq coordinate system, add the resistance components of the AC filter, feeder, foundation output impedance, and foundation virtual impedance at the rated angular frequency. Add the corresponding inductive and reactive components together. Reconstruction:

[0039] Resistance and reactance parameters are read from the calibrated equipment parameter table, feeder length and unit length parameter table, and basic virtual impedance register; the parameter table also records the rated frequency, common bus voltage range, branch current range, compensation bandwidth, temperature range, and topology identifier. A usable setting is defined as an operating range with a rated frequency of 50Hz, a common bus voltage of 0.9 to 1.1 per unit, and branch current not exceeding the rated value; outside this range, this admittance version is not used. After generating the matrix, check the following: and The minimum singular value is not lower than the numerical resolution threshold within the version, and the condition number is not higher than the upper limit allowed for inversion implementation; loading is prohibited if any check fails.

[0040] When using a passive identification path, the internal voltage reference increment, common bus voltage increment, and common bus injected current increment aligned to the same timestamp are read from the normal operation record. The difference between the internal voltage reference increment and the common bus voltage increment is calculated. Samples with protection connections, modulation saturation, missing samples, and voltage difference increment amplitudes lower than the quantization resolution are removed. An identification sample matrix is ​​then formed by member identifier and ascending time order. The offline program uses the difference between the internal voltage reference increment and the common bus voltage increment as input and the common bus injected current increment as output to calculate a two-dimensional least squares mapping. The prediction residual is calculated using the unsolved samples. The mapping is only written as defined when the residuals of both coordinate components do not exceed the thresholds determined by the current sensing accuracy and target modeling error in the parameter version, and the singular value and condition number checks pass. Regardless of whether a parametric path or a passive identification path is used, the admittance record is bound to the feeder topology, basic virtual impedance, applicable operating range, algorithm version, and verification summary. No identification disturbances are actively injected during online coordination. If a topology change, basic virtual impedance change, operating point out of bounds, or parameter drift exceeds the version tolerance is detected, the current admittance version is immediately invalidated.

[0041] Taking a single calculation with three valid members as an example, the rated currents are taken as 100A, 150A, and 200A respectively; the d-axis currents of the common coordinate system are taken as 70A, 120A, and 170A respectively; and the q-axis currents are all taken as 0A. Then the capacity-normalized currents are 0.70, 0.80, and 0.85 respectively, the capacity-weighted average is 0.80, and the shared errors are -0.10, 0, and 0.05 respectively. When all three compensation gain matrices are multiplied by the identity matrix at 20V, the d-axis components of the original compensation request are -2V, 0V, and 1V respectively. If the three admittance matrices are multiplied by the identity matrix at 2S, 1S, and 1S respectively, then the d-axis component of the common voltage component is -0.75V, and the d-axis components of the target compensation voltage are -1.25V, 0.75V, and 1.75V respectively, with a weighted admittance sum of 0A; all three q-axis components are 0. This result also retains the difference between the original requests of any two branches.

[0042] In a preferred embodiment, when the verification passes, the target compensation voltage set is updated according to the minimum allowable proportion of the entire group to obtain the final compensation voltage set, specifically including: set up The valid member number; voltage constraint parameters include the maximum permissible fundamental modulation voltage in the dimension of voltage. and the upper limit of compensation voltage amplitude The upper limit of the compensated voltage change rate, with dimensions of voltage change rate. and the coordinate period with the dimension of time ; Using the uncompensated voltage in the state frame of the synchronous branch as a reference Conservative margin with dimensions of voltage The previously submitted compensation voltage Determine the first The set of membership constraints for each valid member. , and according to Form the complete feasible set, wherein, For the first The target compensation voltage for each effective member; If the previously submitted set of compensation voltages belongs to Then seek to make right The largest established ,Pick , and according to The final compensation voltage set is formed. The aforementioned formula characterizes the common interpolation relationship between each member compensation voltage and the target compensation voltage along the previously submitted compensation voltage, without considering fixed-point quantization errors. When using fixed-point quantization, a candidate compensation voltage set is obtained according to the common interpolation relationship. After the aforementioned fixed-point quantization, admittance-weighted residual verification, necessary common corrections, and verification of all member constraints, the verified fixed-point representation is used as the execution value of the final compensation voltage set.

[0043] Specifically, the maximum permissible fundamental modulation voltage Upper limit of compensation amplitude Upper limit of rate of change and coordination cycle All data is read from the current synchronization tributary status frame and its bound parameter version. Conservative margin is applied. Calculate, where, It is a fixed protection margin in volts, taken from the voltage gap reserved between the modulation linear region and the local protection threshold. The upper bound of the uncompensated reference rate of change in units of volts per second is taken as the larger of the locally enforced reference rate of change limit and the worst closed-loop rate of change obtained from offline load step, DC voltage fluctuation and parameter tolerance verification. The maximum allowable duration from the state frame sampling time to the time of common effectiveness is preset by the current constraint parameter version; Multiply by the maximum clock error Then, a margin is added due to voltage and time quantization errors. All the above parameters and their applicable operating ranges are written into the same constraint parameter version, and the time when they take effect in this round should meet the following conditions. If the maximum allowable duration is exceeded, the pending status is revoked and the current submission round ends. If the maximum allowable fundamental modulation voltage minus the conservative margin is less than or equal to zero, the member constraint set is empty, and the current round directly enters the safety bypass.

[0044] During the preparation and confirmation period, each member calculates the actual rate of change by dividing the uncompensated voltage reference difference between adjacent fast control cycles by the cycle interval, and compares it with the current version. The comparison process is as follows: if the value equals the upper bound, the frame is still considered valid; if it exceeds the upper bound, the timestamp is reversed, or the sampling interval is invalid, the frame to be submitted is immediately marked as unavailable and an abort event is sent. Continuing to enable the frame simply by increasing the current round's margin is not permitted. The new upper bound can only be loaded after offline verification and the formation of a new constraint parameter version.

[0045] The compensation calculation node sequentially substitutes the previously submitted compensation voltage into the three L2 constraints. If all constraints satisfy the condition of being equal to the boundary, it is still considered feasible. After the starting point is deemed feasible, the difference between the target compensation voltage and the previously submitted compensation voltage is taken as the direction of the member segment. The segment parameters are substituted into each L2 constraint, expanded into quadratic inequalities, and a continuous feasible interval including the zero point is obtained. The minimum value between the right endpoint of the three intervals and 1 is taken as the maximum allowable proportion for that member. If any inequality is not true at the zero point, the root-finding discriminant is abnormal, or the fixed-point operation overflows, normal interpolation is not continued; instead, the restricted constraints are recorded and the process is switched to a safety bypass. After all member proportions are calculated, the minimum value is selected as the shared interpolation factor for the entire group, ensuring that each member updates along its respective segment at the same proportion.

[0046] Using the aforementioned three-member example and setting all previously submitted compensation voltages to zero, the coordination period is set to 20ms. When the rate of change constraint of the first member allows a change of 0.875V in this period, and the three constraints of the second and third members all allow reaching the corresponding targets, the maximum allowable proportion of the first member is 0.70, and the maximum allowable proportion of other members is not less than 1. Therefore, the interpolation factor shared by the entire group is 0.70. The d-axis components of the final compensation voltage are -0.875V, 0.525V, and 1.225V, respectively, and the admittance weighted sum is still 0A. Each member also satisfies the modulation voltage, compensation amplitude, and rate of change constraints. This value is only a setting for easy recalculation. The constraint parameters can be adjusted according to the converter rating and control bandwidth, but the minimum allowable proportion shared by all members and the processing relationship of verifying the feasibility of the starting point remain unchanged.

[0047] In some implementations, the final set of compensation voltages is calculated by the compensation coordinator using the synchronization branch state frame as input. The virtual impedance compensation frame is generated by the compensation coordinator, which generates a preparation frame, collects confirmations from all valid members, and forms a submission certificate. The submission certificate is then verified through the redundant consistent multicast service to obtain the consistent multicast delivery, and then the frame is formed.

[0048] Specifically, in the centralized implementation, the compensation coordinator assigns a unique computation task for the same frame number. After reading the ordered list of valid members, it loads the synchronization branch state frame encapsulated by the state subframes of each valid member, as well as the previously submitted compensation frame. It then completes projection, common feasibility, and residual verification, and encapsulates the final compensation voltage set in ascending order of member identifiers. The preparation frame must contain at least the ordered member set and its summary, the input state frame summary set, member and parameter versions, the common coordinate system compensation voltage for each member, the shared interpolation factor for the entire group, the previous frame summary, the new frame number, the provisional common effective time, the effective expiration time, and the integrity check code. Intra-frame sets are arranged in ascending order of member identifiers. All integers are encoded in a fixed byte order, and two-dimensional fixed-point numbers retain a uniform number of decimal places. Fields are separated by type and length encodings. SHA-256 is used to calculate a 32-byte summary of the normalized byte sequence, ensuring that the same input produces the same summary.

[0049] This implementation limits consistent multicast to the synchronous bounded delay communication domain: the end-to-end delay of each valid communication channel does not exceed the configured value. The error of all clocks relative to the common time base is no greater than The multicast service is carried by two deterministic service instances (primary and backup) and two physically independent channels. Within a single commit round, the fault model tolerates at most one channel experiencing frame loss or interruption, or one service instance failing to run; it does not tolerate malicious tampering, nor does it hide invalid members or network partitions as normal commits. Faults include member failure, simultaneous failure of both service instances, simultaneous interruption of both channels, or latency exceeding [a certain threshold]. All of these caused the current round to leave the normal coordination work domain and enter a halt or safe bypass.

[0050] The compensation coordinator uses an epoch number, a strictly increasing frame sequence number, and a complete set digest as the unique key for each commit round. It first writes the preparation frame and its digest to a non-volatile preparation log and persists it. Then, it sends the preparation frame to all valid members in the ordered list. Each member recalculates its three local voltage constraints and checks the member set digest, complete set digest, previous frame digest, version, frame sequence number, and provisional effective time. Only when all are consistent is the preparation record persisted, and an acknowledgment containing the member identifier, frame digest, and acknowledgment status returned. The compensation coordinator only accepts the first valid acknowledgment matching the unique key of the current round. Duplicate acknowledgments do not increment the count; denial, timeout, or digest conflict all cause the round to be aborted. After receiving acknowledgments from all valid members, the coordinator assembles the acknowledgment bitmap and each acknowledgment digest in ascending order of member identifier to form a commit certificate. It first sets the certificate status to pending delivery and persists it, then sends the certificate to the primary and backup service instances.

[0051] The primary and backup service instances respectively verify the ordered member table, complete acknowledgment relationship, previous frame digest, epoch number, frame sequence number, valid expiration time, and integrity checksum in the submitted certificate, and exchange certificate digests calculated in the same field order. When both instances are functioning correctly, only the primary instance holding the current service epoch lease sends the certificate after the digests match; the backup instance sends the certificate based on a digest no greater than [value missing]. The service instance periodically receives the primary instance's heartbeat. If no heartbeat is received for two consecutive periods and the common time exceeds the primary instance's lease expiration time, it verifies itself using the service instance identifier and service epoch number as the unique successor, and then continues the round from the persisted certificate and acknowledgment log. If an instance is determined to be stopped, the surviving instance can independently re-verify the certificate and send it; after the old primary instance recovers, it must read the higher service epoch number and return to the standby state, and cannot broadcast the old decision. The currently valid service instance sends the same submission certificate to all members through two channels. Each member uses the epoch number, frame sequence number, and certificate digest as idempotent keys: when a valid certificate is received for the first time, the certificate and the corresponding final compensation voltage are first written to the non-volatile pending decision log and persisted, and then a certificate delivery acknowledgment with member identifier, idempotent key, and reception status is returned to the primary and standby service instances through two channels; duplicate certificates are only resent for delivery acknowledgment, not rewritten; late certificates with lower epochs or lower frame sequence numbers are directly discarded; certificates with higher frame sequence numbers received in the current round that have not yet been resolved are only entered into the single successor cache without triggering execution.

[0052] The multicast service establishes a certificate delivery confirmation bitmap based on an ordered member list. A delivery confirmation with a correct digest returned by the same member on any channel is considered delivered. The certificate delivery confirmation deadline is the guardian deadline minus the maximum decision multicast latency and one times the maximum clock error. When all member delivery confirmations are received before this deadline, the currently active service instance first persists the round decision and the strictly incrementing service decision sequence number, then broadcasts a full-member delivery decision containing the epoch number, frame sequence number, certificate digest, previous frame digest, decision status, and service decision sequence number through two channels, ensuring the decision reaches members before the guardian deadline. If any confirmation is missing, times out, has a digest conflict, the service instance digest is inconsistent, or the certificate expires, the service persists and broadcasts a full-member abort decision in the same order. Members only set a frame to pending effect after receiving and verifying a full-member delivery decision that completely matches their local pending decision log before the guardian deadline, and after persisting the decision. Upon receiving a full-member abort decision, the pending effect flag is cleared, but the diagnostic log is retained, and the previously submitted compensation frame continues to be used.

[0053] The merging rules for the two channels are fixed as follows: messages with the same idempotent key and identical digests are considered duplicate copies; the maximum channel merging wait time is [value missing]. If only one channel has a valid copy when the waiting period expires, that copy is accepted; if the same epoch and frame number have different digests, or if the decision states of two channels are different, a deterministic all-member abort is selected and both original messages are recorded. Members cannot infer delivery on their own when a unanimous decision is lacking. Certificates, delivery confirmations, or decisions arriving after the guardian's deadline are only audited and do not change the final state of the current round. Therefore, within the aforementioned fault model, even if a channel loses a message or a service instance stops operating, the same decision can still be completed by another channel or a backup instance; outside this fault model, only a safe bypass is guaranteed for each detection node, and unconditional atomic commits across network partitions are not recommended.

[0054] In another embodiment, the final compensation voltage set is calculated independently by each effective member according to the same member sorting, fixed-point number format, multiplication and addition order, rounding rules and serialization rules, and the calculation summaries of each effective member are compared to form a summary consistent result; The virtual impedance compensation frame is formed when the digest consensus result is consistent. The rotating proposal member proposes the frame number and common effective time without modifying the final compensation voltage set, collects the confirmation of all valid members and forms a submission certificate, and then verifies the submission certificate through the consensus multicast service and obtains the consensus multicast delivery.

[0055] Specifically, in the deterministic distributed implementation, each valid member receives a synchronization branch state frame formed by encapsulating subframes of the same member state in ascending order of member identifier. All subframes are traversed in ascending order of member identifier, and the same calculation is performed independently using the aforementioned fixed-point format, matrix inversion implementation, nearest-even rounding, and fixed multiplication-addition order. During normalization serialization, the member and parameter versions, the ordered member table, the final compensation voltage for each member, the shared interpolation factor for the entire group, and the previous frame summary are written sequentially. Then, the set summary is calculated, and the member identifier and summary are broadcast. Each member builds a summary table based on the ordered member table. A consistent summary result is written only when all member summaries are not empty and are byte-by-byte equal; any missing, timed-out, or different summary is written as inconsistent, and the current round of proposal ends.

[0056] When the digests match, the rotating proposal member is selected cyclically according to the member ordered list and the current epoch number. Only strictly increasing frame numbers and common effective times that meet the guardian time are added. The compensation set or set digest must not be modified. After receiving the proposal, each member re-verifies the feasibility of the complete set digest, the previous frame digest, the version, and the local voltage. If the verification passes, a confirmation is returned. After the rotating member collects all confirmations, a submission certificate is formed according to the order of the ordered member list, confirmation bitmap, epoch number, frame number, previous frame digest, and certificate digest fields defined in the centralized implementation, and written to the pending delivery log. Subsequently, each member and the rotating proposal member jointly use the aforementioned certificate delivery confirmation bitmap, the all-member decision frame, the dual-channel merging rules, and the persistence order: each member persists the pending decision certificate and returns a delivery confirmation. The multicast service only forms an all-member delivery decision after collecting all delivery confirmations; otherwise, an all-member abort decision is formed. If the rotating member fails or fails to form a certificate within the proposal period, the all-member round is aborted, and the next member in the order becomes the proposal member in the next coordination cycle. Compensation sets that have been calculated but not submitted are not written to the effective area. New proposal members read the final state of the previous round from the persistent log; if the previous round was a complete abort, a new commit round can be established in the next coordination cycle; if the previous round was a complete delivery, a new commit round is established only after the corresponding compensation frame reaches the common effective time and the current execution area is updated; for rounds where the final state is unknown or where a complete delivery has been achieved but the common effective time has not yet been completed, a new compensation calculation is not initiated.

[0057] Furthermore, the consistent multicast delivery is determined when each valid member confirms a preparation frame containing the ordered member set, the final compensation voltage set, the previous frame summary, and the provisional common effective time, collects all confirmations to form a commit certificate, and the consistent multicast service delivers the same commit certificate to all valid members in the ordered member set before the guardian deadline of the common effective time, and returns a full delivery decision to all valid members. The consensus multicast termination is determined when any of the aforementioned conditions are not met, and the termination decision is returned to all valid members, so that all valid members retain the previously submitted compensation frame and end the current round of submission. In the next coordination cycle, the previously submitted compensation voltage set is re-verified to see if it belongs to the feasible set of the whole group. If the verification fails, the coordinated compensation is cancelled in the next local control cycle, and the uncompensated voltage reference of the local unit is limited.

[0058] Specifically, such as Figure 5As shown, the compensation computing node first encapsulates a preparation frame containing a complete set summary, version, previous frame summary, and tentative common effective time, and then sends the preparation frame to effective member 1 and effective members 2 to n respectively. After each member reviews the feasibility of its local voltage and the complete summary, it first records the preparation status, and then returns an acknowledgment with member identifier and frame summary respectively; the computing node checks and confirms each member in the ordered member list, and after collecting them, forms a commit certificate containing complete acknowledgment relationships and sends it to the consensus multicast service. Figure 5 Items 8 through 10 combine the certificate verification, certificate delivery confirmation collection, and all-member decision broadcasting performed by the multicast service into a single service interaction representation. In the program implementation, the service first sends certificates to each member. Each member persists the certificate to be decided and then returns a delivery confirmation. The service persists and broadcasts the all-member delivery decision only when it has collected all delivery confirmations in the ordered member list before the certificate delivery confirmation deadline, which is earlier than the guardian's deadline. Otherwise, it persists and broadcasts the all-member abort decision. When all members deliver, each member sets the frame to a pending state after persisting the decision. When all members abort, each member clears the pending flag for this round and continues to hold the previously submitted compensation frame.

[0059] The common effective time satisfies the following conditions: no earlier than the certificate generation time, twice the maximum group delivery delay, twice the maximum clock error, and the sum of the guard interval. The first group delivery is used for certificate transmission and delivery confirmation, and the second group delivery is used for broadcasting the decision to all members. Simultaneously, the common effective time satisfies the aforementioned conditions. The earliest co-effective time determined according to the maximum group delivery delay, maximum clock error, and guard interval is later than the limit. If this happens, the current round will not continue to achieve consensus on multicast delivery, and a decision to terminate all members will be made. Taking a maximum group delivery latency of 2ms (including one allowed service instance failover), a maximum clock error of 50μs, and a guardian interval of 0.5ms as an example, under the condition of meeting the aforementioned maximum allowed duration, the co-activation time can be selected as 5ms after certificate generation. Members will only remain in the pending activation state if they receive and verify the correct certificate and the all-member delivery decision at the guardian deadline before the co-activation time, the frame version is still valid, and the local machine is still in a normal network state; if the certificate expires, the previous frame digest does not match, the member version changes, the clock is lost, or local protection takes over, the pending activation state will be revoked and a revocation event with the epoch number, frame sequence number, and reason code will be sent.

[0060] If the full delivery decision has been persisted or the guardian deadline has expired, and a member exits the normal network state before the common effective time, that member first writes the failure reason and the inability to recover flag for that frame to its local log. Instead of waiting for group activation, it enters a safe bypass in the next local fast control cycle and sends the failure event to the service and other members via two channels. Other members who receive the failure event before the common effective time also write the inability to recover flag, cancel their pending state, and enter a safe bypass. Members that receive the event later enter a safe bypass in the next fast control cycle after receiving it. This situation exceeds the normal coordination domain, and the text does not interpret a possible communication propagation window for each member as a full group atomic switch. When the number of network partitions or failures exceeds the aforementioned model, only local safe bypass and new epoch recovery are performed.

[0061] Each member sequentially saves the epoch number, frame sequence number, previous frame digest, preparation record, pending certificate, unanimous decision, pending status, anti-recovery flag, co-effective time, and current execution frame digest in the non-volatile commit log. The writing order is fixed as follows: preparation confirmation can only be sent after the preparation frame is persisted; delivery confirmation can only be sent after the pending certificate is persisted; pending status can only be set after the unanimous decision is persisted; and the current execution frame digest is only written after the pointer swap is completed at the co-effective time. If any step fails, the process will not skip to the next state. When the process restarts, the log is replayed in ascending order of epoch number and frame sequence number: preparation records without unanimous decisions are treated as unanimous aborts; frames with unanimous abort records or anti-recovery flags retain the previous committed compensation frame; frames with unanimous deliveries but not yet at the co-effective time are only restored to pending status if the version, clock, certificate, and member status are re-verified successfully; otherwise, they enter a safe bypass. If the restart occurs after the co-effective time but the current execution frame digest has not yet been written, the certificate is not enabled late; instead, the process enters a safe bypass with zero coordination compensation and waits for the new epoch. When recovery fails, retain the log, failure steps, and reason codes, and do not write the intermediate compensation set into the execution area.

[0062] like Figure 3 As shown, when the previously submitted compensation voltage set does not belong to the current feasible set of the entire group, each member sets the coordinated compensation execution amount to zero in the next local fast control cycle, sends the local uncompensated voltage reference to the local protection limiter, and writes the operating status as safe bypass. The limiter restricts the amplitude of the uncompensated voltage reference to the non-negative range after deducting the local protection margin from the maximum allowable fundamental modulation voltage. When hardware overcurrent limiting or fault ride-through has been triggered, the higher priority local protection directly takes over. Before restoring normal coordination, it must be confirmed that zero compensation is feasible for all members, clock and communication have been restored, and multiple consecutive observation windows are valid before establishing a new epoch and starting from the all-zero compensation frame.

[0063] When a member is added or removed, the feeder topology is changed, or the admittance version is updated, the old member version first sets the target compensation set to all zeros, continues to use common interpolation, full member confirmation, and consistent multicast until all zero frames take effect, then closes the old epoch and loads the new member table and the new admittance table. If a member is lost unplanned, the network partitions, or admittance drift exceeds the version tolerance, it is not immediately recalculated in the old projection relationship; members that can detect abnormalities are switched to a safe bypass. After communication is restored, a new epoch is established to avoid direct interpolation between different admittance projection subspaces.

[0064] For S3, specifically, each member sets a current execution area and a pending area. The current execution area stores the previously submitted compensation frame, while the pending area only stores new frames that have received a full member delivery decision and whose certificates have passed verification. Within the local control interrupt when the common time counter reaches the common effectiveness time, the controller first verifies the pending frame sequence number, epoch number, certificate digest, and validity expiration time, and then switches the local common coordinate system compensation voltage from the pending area to the current execution area using a single pointer exchange. If full member delivery is not obtained or any verification fails, the pointer exchange is not performed, and the consensus multicast aborts the branch to continue reading the previously submitted compensation frame. Here, the single pointer exchange is an atomic replacement within the local control task, and cross-node switching in the abnormal communication domain is not interpreted as an unconditional atomic operation.

[0065] In each fast control cycle, the controller reads the local common coordinate system compensation voltage, common coordinate system phase angle, and local coordinate system phase angle in the current execution area. It calculates the sine and cosine values ​​corresponding to the difference between the two phase angles and assembles a two-dimensional rotation matrix to transform the compensation voltage to the local dq coordinate system. The transformation result is subtracted from the local uncompensated voltage reference and written to the compensated voltage reference register. If the verification fails, the coordinated compensation is set to zero, and the uncompensated voltage reference after local limiting is written to the same register, so that both branches form a definite voltage reference value for S4 to read.

[0066] In some implementations, when the verification passes, the local compensated voltage in the effective compensation frame is transformed to the local coordinate system and subtracted from the local uncompensated voltage reference to form a compensated voltage reference value, specifically including: set up The effective member number is denoted as and the local compensation voltage expressed in the common coordinate system in the effective compensation frame is recorded as . ;according to ,Will Transformed into local compensated voltage in local coordinate system ,in, The phase angle is in the common coordinate system. The phase angle of the local coordinate system. It is a two-dimensional rotation matrix; The local compensation voltage is used as the only new execution quantity, and the equivalent resistance or equivalent reactance calculated from the local compensation voltage is only used for monitoring. Perform a base virtual impedance on-premises to form an on-premises uncompensated voltage reference. , and according to The compensated voltage reference value is then formed.

[0067] Specifically, the common coordinate system compensates for the voltage. To submit the two-dimensional fixed-point vector corresponding to the intra-frame local member identifier, the common phase angle The local phase angle is read from the common rotating coordinate generator corresponding to the common time base. The phase status is read from the real-time phase state of the network controller. The controller performs rotation in the direction shown in the formula, making the local-to-common transformation in S1 and the common-to-local transformation in S3 inverses of each other; then the rotation result is rounded according to the local voltage reference scale and limited to the value range of the compensation execution register. If there is an overflow or phase angle loss, it enters the safety bypass.

[0068] The basic virtual impedance module first reads the local grid voltage reference and the local branch current, generates a basic virtual impedance voltage drop based on the loaded basic virtual resistance and basic virtual reactance, subtracts this voltage drop from the grid voltage reference, and writes it into the uncompensated voltage reference register. The compensation execution module then only subtracts the local compensation voltage. The equivalent resistance or equivalent reactance obtained by dividing the compensation voltage by the instantaneous current is not written into the control channel; if the instantaneous current is too small and the back calculation is unstable, the monitoring terminal marks the equivalent impedance value as invalid, which does not affect the execution of the compensation voltage. The final compensated voltage reference value... Write to the voltage regulator input register and accept the highest priority limiting or takeover from the local protector.

[0069] For S4, specifically, the voltage regulator reads the compensated voltage reference value and the sampled local output voltage in a local fast control cycle, calculates the voltage deviation, and generates a current reference through a pre-tuned proportional-integral or state feedback adjustment relationship. The current regulator reads this current reference and the local branch current to generate a modulation voltage command. The modulator divides the modulation voltage command by the current DC bus voltage to normalize it, limits it to the allowable modulation range, and converts it into the duty cycle of each bridge arm. The gate driver then drives the power semiconductor bridge accordingly, supplying power to the common AC bus through the AC filter and feeder.

[0070] After the control cycle ends, the controller writes the updated branch current, AC side output voltage, uncompensated voltage reference, maximum allowable fundamental modulation voltage, current execution frame summary, and operating status into the next observation window buffer, and uses the updated data in the buffer as the status update content of the next synchronous branch status frame. In the next coordination cycle, S1 reads the data from the buffer at the same sampling time and reassembles the next synchronous branch status frame, thus forming a closed loop from the compensated voltage reference, power stage output, branch current feedback to the next compensated frame. When modulation saturation, hardware overcurrent, sampling fault, or fault ride-through is triggered, the local protector overrides the coordinated compensation control, records the protection status, and causes subsequent status frames to exit the normal coordinated branch.

[0071] This embodiment also provides a virtual impedance compensation control system for a grid-type converter, such as... Figure 6 As shown.

[0072] The system is used to implement the aforementioned method, and the system includes: The status frame forming module is used to obtain the status frame of the synchronization branch, the previous submitted compensation frame, and the previous submitted compensation voltage set. The compensation frame forming module is used to generate an original compensation request based on the branch current and rated current in the synchronous branch status frame, remove the common voltage component of the original compensation request according to the branch incremental admittance matrix bound to the synchronous branch status frame, obtain a target compensation voltage set with a weighted sum of zero admittance, verify whether the previously submitted compensation voltage set belongs to the entire feasible set composed of the Cartesian product of the member constraint sets, and if the verification passes, update the target compensation voltage set according to the minimum allowable proportion of the entire group to obtain the final compensation voltage set. After confirmation by all members and consistent multicast delivery, a virtual impedance compensation frame containing the final compensation voltage set and the common effective time is formed. The compensation execution module is used to activate the virtual impedance compensation frame during the common effective time when consistent multicast delivery occurs, forming an effective compensation frame. When consistent multicast is terminated, the previous submitted compensation frame is used as the effective compensation frame. The voltage reference formation includes: when the feasibility check of the entire set of the previous submitted compensation voltages passes, transforming the local compensation voltage in the effective compensation frame to the local coordinate system and subtracting it from the local uncompensated voltage reference to form a compensated voltage reference value; when the feasibility check of the entire set of the previous submitted compensation voltages fails, canceling the coordinated compensation in the next control cycle, limiting the local uncompensated voltage reference, and forming the compensated voltage reference value. The power supply control module is used to control the converter output according to the compensated voltage reference value and update the synchronous branch status frame.

[0073] Specifically, the state frame formation module executes S1. Its input is connected to the sampling terminals of each grid-type converter, the local control status register, and the submission log. Its output transmits the synchronization branch state frame and the previously submitted information to the compensation frame formation module through the frame buffer. The compensation frame formation module executes S2 and includes a member and parameter version management module, a compensation calculation module, and a consensus submission module. The member and parameter version management module provides an ordered member table and admittance and gain versions. The compensation calculation module reads the state frame buffer and outputs the final compensation voltage set. The consensus submission module encapsulates the verified set into a preparation frame, collects confirmation and sends submission certificates through the coordination communication interface, and then writes the overall decision and virtual impedance compensation frame into the pending-effective area of ​​the compensation execution module.

[0074] The consistent commit module includes a preparation and confirmation manager, a certificate generator, a dual-channel multicast interface, a certificate delivery confirmation register, a full-member decision state machine, and a non-volatile commit log. The preparation and confirmation manager takes the epoch number, frame sequence number, and complete set digest as input and outputs an ordered confirmation bitmap. The certificate generator assembles this bitmap, the previous frame digest, the common effective time, and parameter versions into a commit certificate. The dual-channel multicast interface sends the certificate to the primary and backup service instances and writes the delivery confirmations returned by each member into the confirmation register addressed by member identifier. The full-member decision state machine checks the confirmation register and the guardian cutoff counter bit by bit, outputting a full-member delivery decision only when all bits are valid; other final states output a full-member abort decision, and writes the results to the commit log in the order of preparation record, certificate to be decided, full-member decision, pending effective state, and current execution digest. Therefore, the system module has the same idempotency, timeout, conflict, abort, and restart recovery interfaces as the aforementioned methods, instead of only considering the certificate sending action as the full-member delivery result.

[0075] The compensation execution module executes S3, with its inputs connected to the current execution area, the area awaiting activation, the common time counter, the local coordinate system phase angle, and the uncompensated voltage reference register. Its output is connected to the voltage regulator input register of the power supply control module. The power supply control module executes S4, with its voltage regulator, current regulator, modulator, and gate drive interface sequentially connected to the converter power stage. It also feeds back the sampled updated branch status to the status frame formation module. The local protection module connects the compensation execution module and the power supply control module. The protection takeover signal has the highest priority; when the protection signal is valid, it prevents the activation of the waiting-to-activate frame and drives the power supply control module with the limited uncompensated voltage reference or fault control command.

[0076] The aforementioned modules can be used as software functional units within the same controller, or the compensation frame forming module can be deployed on the microgrid controller, while the other modules can be deployed on the local controllers of each grid-type converter. When deployed across devices, modules transmit data via synchronous branch status frames, preparation frames, confirmations, certificate submissions, and overall decision-making. When deployed within the same device, modules transmit data via a shared memory queue protected by version numbers and frame sequence numbers. Both deployments sequentially execute status frame transmission, compensation frame verification, overall decision verification, voltage reference writing, and branch status feedback. The output of the status frame forming module directly enters the compensation frame forming module. The compensation frame forming module only writes the frame that has received overall delivery to the compensation execution module. The voltage reference output from the compensation execution module enters the power supply control module, and the feedback data is then returned to the status frame forming module.

[0077] From the perspective of the overall technical mechanism, when the rated capacity and feeder incremental admittance of parallel members are different, the independent compensation of this unit will cause the capacity sharing adjustment to generate a non-zero composite excitation on the common AC bus side after being mapped through different branches. In this implementation, the branch current in the common coordinate system is first normalized to the rated current so that the sharing error reflects the capacity ratio; then, the common voltage component is removed from the original compensation request of the entire group using the branch incremental admittance, so that the weighted sum of the admittances of the target compensation set is zero, while retaining the relative difference between the original requests of the branches. Thus, the compensation target still acts along the direction of the aggregation negative feedback of the sharing error, and at the same time, no first-order composite current excitation of the common bus directly caused by this compensation is formed within the linearized operating point and coordination bandwidth corresponding to the admittance version.

[0078] When the modulation margin, compensation authority, and allowable rate of change differ among members, each machine will use different scaling ratios for limiting, thus disrupting the aforementioned admittance weighting relationship. This implementation first confirms that the previously submitted compensation set is still within the current feasible set for the entire group. Then, it calculates the allowable ratio for each member along the line segment from the previous set to the target set, and lets the most restricted member determine the common ratio for the entire group. The same interpolation factor ensures that the algebraic relationship is maintained between the endpoints where the weighted sum of the old and new admittances is zero. If the starting point is no longer feasible, interpolation is not continued from that point, and the coordinated compensation is revoked and limiting protection is entered in the next local control cycle, thereby preventing the failure starting point from becoming a new execution command.

[0079] Communication delays can cause the same compensation set to arrive at different members sequentially. If each member activates it immediately upon receipt, old and new compensations may be mixed during parameter switching intervals. This implementation binds the complete compensation set, member and parameter versions, previous frame summary, and common effective time to the same submission certificate, and uses a certificate delivery confirmation bitmap to prove that each member has persisted the same certificate. The multicast service only persists the full-member delivery decision after receiving all delivery confirmations before the guardian deadline; otherwise, it persists the full-member abort decision. Therefore, in the normal communication domain where bounded delay, clock error, and single fault tolerance are all met, and all members are continuously in a normal network state, all members only enter the pending-effective state after obtaining and persisting the same full-member delivery decision, and replace their local compensation voltage at the common effective time. The reverse rotation from the common coordinate system to the local coordinate system ensures that the same physical compensation vector correctly enters each local voltage reference channel, thereby reducing branch voltage differences and update circulating currents caused by asynchronous switching. When a member is abnormal, the network is partitioned, or the fault model is exceeded, the conclusion of the normal domain is terminated and a safety bypass is executed, and the fault domain is not described as still having the unconditional synchronous submission property.

[0080] like Figure 7 As shown, assuming the membership set and admittance parameter versions remain unchanged and the previously submitted compensation voltage set is currently feasible, the candidate interpolation scale is used. The horizontal axis is plotted as the normalized remaining feasible margin of each member along the line segment from the previous compensation voltage to the target compensation voltage, and the vertical axis is plotted as the normalized remaining feasible margin of each member along the line segment from the previous compensation voltage to the target compensation voltage. For each member and each First press Obtain candidate points for the line segment; then calculate the remaining amount between the maximum allowable fundamental modulation voltage minus the conservative margin and the compensated voltage reference norm, the remaining amount between the upper limit of the compensation amplitude and the candidate compensation norm, and the remaining amount between the allowable change in a single cycle and the change norm of the candidate point relative to the previous compensated voltage. Divide each of the three remaining amounts by the corresponding positive upper limit. , and The minimum of the three values ​​is taken as the vertical axis value of the member; if any denominator is less than or equal to zero, the member is directly judged as infeasible and is not plotted as a normal curve. A vertical axis value greater than zero indicates that all three constraints have remainders, equal to zero indicates that at least one constraint has just reached the boundary, and less than zero indicates that at least one constraint has been violated.

[0081] Figure 7 Each curve corresponds to the constraint response of the aforementioned three-member example. As the candidate proportion increases, the residual margin of each member decreases respectively; the curve for effective member 1 first... At the zero boundary, effective members 2 and 3 still maintain positive margins at this position. Therefore, a shared factor for the entire group is selected. At this point, the most restricted member is exactly at the boundary, while the remaining members are still in the feasible region; if the candidate proportion is further increased, then effective member 1 enters the infeasible region. This response demonstrates that the minimum allowed proportion for the entire group can keep all members feasible under asymmetric constraints and avoid each member independently choosing to update the proportion.

[0082] It should be understood that Figure 7 This reflects the mechanism of common interpolation within the normal coordinated working domain, and does not imply that updates can still be performed along the original line segment when the previous set is no longer feasible, member versions change, or admittance versions become invalid. For two or more network-type converters, the compensation calculation nodes still perform the same capacity normalization, admittance weighted projection, member feasible interval calculation, and minimum value selection according to the ordered member table; for off-diagonal two-dimensional admittance matrices, common voltage components are still obtained through two-dimensional matrix accumulation and inversion. When the network controller adopts droop control, virtual synchronous machine control, or virtual oscillator control, only the local phase angle, uncompensated voltage reference, and branch state need to be provided, and the subsequent common constraints and submission mechanism remain unchanged.

[0083] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A virtual impedance compensation control method for a grid-type converter, applied to multiple grid-type converters connected in parallel to a common AC bus via feeders, characterized in that... include: S1. Obtain the synchronization branch status frame, the previous submitted compensation frame, and the previous submitted compensation voltage set. S2. Generate an original compensation request based on the branch current and rated current in the synchronous branch status frame. Remove the common voltage component of the original compensation request according to the branch incremental admittance matrix bound to the synchronous branch status frame to obtain a target compensation voltage set with a weighted sum of zero admittance. Verify whether the previously submitted compensation voltage set belongs to the entire feasible set composed of the Cartesian product of the member constraint sets. If the verification passes, update the target compensation voltage set according to the minimum allowable proportion of the entire group to obtain the final compensation voltage set. After confirmation by all members and consistent multicast delivery, a virtual impedance compensation frame containing the final compensation voltage set and the common effective time is formed. S3. During unified multicast delivery, at the common effective time, the virtual impedance compensation frame is activated to form an effective compensation frame. When unified multicast is terminated, the previous submitted compensation frame is taken as the effective compensation frame. The voltage reference formation includes: when the feasibility check of the entire set of the previous submitted compensation voltages passes, the local compensation voltage in the effective compensation frame is transformed to the local coordinate system and subtracted from the local uncompensated voltage reference to form a compensated voltage reference value; when the feasibility check of the entire set of the previous submitted compensation voltages fails, the coordinated compensation is canceled in the next control cycle, the local uncompensated voltage reference is limited, and the compensated voltage reference value is formed. S4. Control the converter output according to the compensated voltage reference value and update the synchronous branch status frame.

2. The method according to claim 1, characterized in that, The acquisition of the synchronization branch status frame, the previous submitted compensation frame, and the previous submitted compensation voltage set specifically includes: The positive-sequence fundamental component of the branch current is extracted from the same fundamental observation window of each effective member, and the uncompensated voltage reference at the same sampling time is obtained. A two-dimensional rotation matrix is ​​determined based on the difference between the phase angle of the local coordinate system and the phase angle of the common coordinate system. The two-dimensional rotation matrix is ​​then used to transform the positive sequence fundamental component and the uncompensated voltage reference, respectively. The transformation results of each valid member, the sampling timestamp, the member set version, the branch incremental admittance matrix version, and the rated current and voltage constraint parameters are written into the same state frame to form the synchronous branch state frame; the currently submitted compensation frame is determined as the previous submitted compensation frame, and the previous submitted compensation voltage set is read from it.

3. The method according to claim 1, characterized in that, The step of generating an original compensation request based on the branch current and rated current in the synchronization branch state frame, and removing the common voltage component of the original compensation request according to the branch incremental admittance matrix bound to the synchronization branch state frame to obtain a target compensation voltage set with a weighted admittance sum of zero, specifically includes: Using the rated current, the branch current is normalized to form a normalized current sharing error, and the original compensation request is determined based on the compensation gain matrix and the normalized current sharing error. Based on the branch incremental admittance matrix bound to the synchronous branch state frame and the original compensation request, the common voltage component is determined, and the target compensation voltage set is formed using the common voltage component.

4. The method according to claim 3, characterized in that, The process of determining the common voltage component based on the branch incremental admittance matrix bound to the synchronous branch state frame and the original compensation request, and forming the target compensation voltage set using the common voltage component, specifically includes: set up For the number of valid members, For valid member serial numbers, For valid member serial numbers, and The average value is to According to the rated current Calculate capacity weights Based on the branch currents in the common coordinate system Calculate the normalized current of the capacity Capacity-weighted average And capacity normalized current sharing error , and according to The original compensation request is formed, wherein, This is the compensation gain matrix with the dimension of voltage; exist When reversible, the original compensation request is used according to... The common voltage component is formed, and according to... Forming the target compensation voltage ,make ,in, The dimensionless quantity of the first is conductance. The branch incremental admittance matrix of each effective member is determined by the target compensation voltage. The target compensation voltage set is formed.

5. The method according to claim 1, characterized in that, When the verification passes, the target compensation voltage set is updated according to the minimum allowable proportion of the entire group to obtain the final compensation voltage set, which specifically includes: set up The valid member number; voltage constraint parameters include the maximum permissible fundamental modulation voltage in the dimension of voltage. and the upper limit of compensation voltage amplitude The upper limit of the compensated voltage change rate, with dimensions of voltage change rate. and the coordinate period with the dimension of time ; Using the uncompensated voltage in the state frame of the synchronous branch as a reference Conservative margin with dimensions of voltage The previously submitted compensation voltage Determine the first The set of membership constraints for each valid member. , and according to Form the complete feasible set, wherein, For the first The target compensation voltage for each effective member; If the previously submitted set of compensation voltages belongs to Then seek to make right The largest established ,Pick , and according to This forms the final set of compensation voltages.

6. The method according to claim 3, characterized in that, The final set of compensation voltages is calculated by the compensation coordinator using the state frame of the synchronization branch as input. The virtual impedance compensation frame is generated by the compensation coordinator, which generates a preparation frame, collects confirmations from all valid members, and forms a submission certificate. The submission certificate is then verified through the redundant consistent multicast service to obtain the consistent multicast delivery, and then the frame is formed.

7. The method according to claim 3, characterized in that, The final compensation voltage set is calculated independently by each effective member according to the same member sorting, fixed-point number format, multiplication and addition order, rounding rules and serialization rules. The calculation summaries of each effective member are compared and a consistent summary result is formed. The virtual impedance compensation frame is formed when the digest consensus result is consistent. The rotating proposal member proposes the frame number and common effective time without modifying the final compensation voltage set, collects the confirmation of all valid members and forms a submission certificate, and then verifies the submission certificate through the consensus multicast service and obtains the consensus multicast delivery.

8. The method according to claim 1, characterized in that, The consistent multicast delivery is determined when each valid member confirms a preparation frame containing the ordered member set, the final compensation voltage set, the previous frame summary, and the provisional common effective time, collects all confirmations to form a submission certificate, and the consistent multicast service delivers the same submission certificate to all valid members in the ordered member set before the guardian deadline of the common effective time, and returns a full delivery decision to all valid members. The consensus multicast termination is determined when any of the aforementioned conditions are not met, and the termination decision is returned to all valid members, so that all valid members retain the previously submitted compensation frame and end the current round of submission. In the next coordination cycle, the previously submitted compensation voltage set is re-verified to see if it belongs to the feasible set of the whole group. If the verification fails, the coordinated compensation is cancelled in the next local control cycle, and the uncompensated voltage reference of the local unit is limited.

9. The method according to claim 1, characterized in that, When the verification passes, the local compensated voltage in the effective compensation frame is transformed to the local coordinate system and subtracted from the local uncompensated voltage reference to form the compensated voltage reference value, specifically including: set up The effective member number is denoted as and the local compensation voltage expressed in the common coordinate system in the effective compensation frame is recorded as . ;according to ,Will Transformed into local compensated voltage in local coordinate system ,in, The phase angle is in the common coordinate system. The phase angle of the local coordinate system. It is a two-dimensional rotation matrix; The local compensation voltage is used as the only new execution quantity, and the equivalent resistance or equivalent reactance calculated from the local compensation voltage is only used for monitoring. Perform a base virtual impedance on-premises to form an on-premises uncompensated voltage reference. , and according to The compensated voltage reference value is then formed.

10. A virtual impedance compensation control system for a grid-type converter, characterized in that, The system is used to implement the method of any one of claims 1 to 9, and the system comprises: The status frame forming module is used to obtain the status frame of the synchronization branch, the previous submitted compensation frame, and the previous submitted compensation voltage set. The compensation frame forming module is used to generate an original compensation request based on the branch current and rated current in the synchronous branch status frame, remove the common voltage component of the original compensation request according to the branch incremental admittance matrix bound to the synchronous branch status frame, obtain a target compensation voltage set with a weighted sum of zero admittance, verify whether the previously submitted compensation voltage set belongs to the entire feasible set composed of the Cartesian product of the member constraint sets, and if the verification passes, update the target compensation voltage set according to the minimum allowable proportion of the entire group to obtain the final compensation voltage set. After confirmation by all members and consistent multicast delivery, a virtual impedance compensation frame containing the final compensation voltage set and the common effective time is formed. The compensation execution module is used to activate the virtual impedance compensation frame during the common effective time when consistent multicast delivery occurs, forming an effective compensation frame. When consistent multicast is terminated, the previous submitted compensation frame is used as the effective compensation frame. The voltage reference formation includes: when the feasibility check of the entire set of the previous submitted compensation voltages passes, transforming the local compensation voltage in the effective compensation frame to the local coordinate system and subtracting it from the local uncompensated voltage reference to form a compensated voltage reference value; when the feasibility check of the entire set of the previous submitted compensation voltages fails, canceling the coordinated compensation in the next control cycle, limiting the local uncompensated voltage reference, and forming the compensated voltage reference value. The power supply control module is used to control the converter output according to the compensated voltage reference value and update the synchronous branch status frame.

Images