Distribution network load emergency group control system and method under main network N-1 fault

Abstract

The invention provides a distribution network load emergency group control system and method under a main network N-1 fault, and solves the problems of inaccurate fault influence quantization, high response delay and extensive load control in the prior art. According to the scheme, the method comprises the following steps: fault accurate mapping: through a transformer total-sequence component port compensation method, the influence of a main network fault on voltage sag of a distribution network is quantified, and the error rate is reduced to 5%; according to the dynamic safety boundary, a time period decoupling multi-parameter planning technology is adopted, power fluctuation threshold values of all feeder lines are generated in a second level, and calculation consumed time is compressed to 0.8 second; load grading group control: interruptible groups (shopping mall air conditioners) cut off loads according to weights; the adjustable group (factory production line) implements active-reactive cooperative adjustment; the key guarantee group (hospital) dynamically injects reactive support to guarantee zero interruption of power supply; and closed-loop iteration verification: based on a real-time voltage deviation self-adaptive adjustment instruction, ensuring that the voltage is recovered to + / -1% threshold within 8.3 seconds. Actual measurement shows that compared with a traditional method, the voltage recovery speed is increased by 67%, the key load loss is reduced by 100%, and reliable defense support is provided for a high-proportion new energy power grid.

Description

Technical Field

[0001] This invention relates to the field of power grids, and specifically to an emergency group control system and method for distribution network loads under N-1 fault conditions in the main grid. Background Technology

[0002] With the accelerated construction of new power systems, the high proportion of renewable energy integration and load diversification have dramatically increased the complexity of power grid operation. Key equipment in the main grid (such as transformers and transmission lines) may experience single-component failures without redundancy (referred to as "N-1 faults") due to natural disasters, equipment aging, or operational errors. Such faults can cause voltage dips (short-term voltage drops exceeding 10% of rated value) and power imbalances in the distribution network, and in severe cases, load tripping or even large-scale power outages. To address this challenge, emergency load control technology—which maintains grid stability through tiered and rapid regulation of interruptible loads, adjustable loads, and critical backup loads—has become a core means of ensuring power supply reliability.

[0003] The current technological system relies on the coordination of three key levels:

[0004] Fault modeling layer: A main network coupling model is established using the port compensation method to analyze the propagation path of faults from the main network to the distribution network;

[0005] Safety boundary layer: Calculates the safe operating range of the distribution network based on mathematical programming and determines the power fluctuation threshold;

[0006] Control execution layer: Divide clusters according to load priority and generate differentiated control instructions.

[0007] However, the random output of new energy sources exacerbates voltage fluctuations during fault periods. For example, when wind power penetration exceeds 40%, voltage fluctuations can reach ±3%, far exceeding the ±1% limit stipulated by national standards, posing a serious challenge to traditional methods.

[0008] Existing methods have significant shortcomings in terms of real-time performance, accuracy, and collaboration:

[0009] 1. Inaccurate quantification of fault impact:

[0010] Traditional port compensation methods ignore the blocking effect of different transformer connection methods (such as YNd type) on zero-sequence current, resulting in a voltage sag mapping error of 15% to 20%.

[0011] The dynamic characteristics of the load (such as the starting inrush current of the motor) are not included in the model, and static calculations cannot reflect the actual impedance changes at the moment of the fault.

[0012] 2. Inefficient calculation of safety boundaries:

[0013] Existing multi-parameter programming methods require handling all time-period constraints simultaneously, resulting in an exponential increase in computational complexity. Real-world testing shows that optimizing for 24 time periods takes over 5 minutes, failing to meet the requirement of second-level response to N-1 faults.

[0014] The load adjustment margin adopts a fixed ratio (such as setting the interruptible group cutoff amount to 30% of the total demand) without dynamically associating it with the real-time operating boundary, resulting in conservative scheduling or excessive load reduction.

[0015] 3. Group control strategies lack adaptability:

[0016] The load cluster weights are preset based on manual experience and are not linked to real-time voltage deviations. The reactive power compensation response delay for critical loads exceeds 10 seconds, exacerbating the risk of voltage instability.

[0017] The fixed step size used in the iterative update of control commands can easily lead to voltage oscillations or convergence failures. A case study of a provincial power grid shows that the fault recovery process experienced three repeated voltage fluctuations.

[0018] 4. Difficulty in implementing intelligent algorithms:

[0019] Reinforcement learning-based control strategies require massive amounts of historical data for training, and the control accuracy drops by more than 30% when migrated to a new power grid topology.

[0020] The data processing latency of edge computing nodes exceeds 24 milliseconds, which is difficult to meet the millisecond-level control requirements.

[0021] Chinese patent CN202211076452.9 discloses an N-1 security analysis method for distribution networks based on a minimum isolation area information matrix. Its technical solution includes: establishing a minimum isolation area information matrix (describing switch and transformer information); establishing a minimum isolation area real-time load information matrix (determining the total load to be transferred); establishing a load transfer path switch information matrix and switch carrying capacity information matrix (describing the transfer path margin); and establishing a minimum margin information matrix (compared with the load to be transferred). However, this method has significant limitations: it only considers the internal transfer capacity of the distribution network, without coupling the analysis of the main grid voltage support capacity; when a main grid fault causes a drop in the distribution network boundary voltage, the margin calculation becomes distorted; the matrix construction relies on static equipment parameters and does not correlate with real-time margin changes caused by fluctuations in new energy output; it only identifies weak links and does not generate active control strategies.

[0022] Chinese patent CN201910038858.X discloses an N-1 security assessment method for active distribution networks considering transmission and distribution coordination. The main steps include: establishing an N-1 load recovery model for the active distribution network; establishing a transmission and distribution coordination calculation model based on the master-slave splitting method; and establishing an N-1 security assessment index system for the active distribution network. However, this method has the following shortcomings: the master-slave splitting method uses a fixed iteration step size, resulting in slow convergence speed (over 5 minutes), which cannot meet the requirements for second-level response; the security assessment index does not quantify the priority of load cluster regulation, leading to a lack of control weight for critical loads (such as hospitals); and no closed-loop control mechanism is designed, only risk assessment is implemented without generating group control commands. Summary of the Invention

[0023] To address this, this solution employs a three-layer collaborative optimization approach—fault modeling, boundary calculation, and group control execution—to overcome the shortcomings of existing patents in areas such as voltage constraint embedding, dynamic boundary generation, and differentiated cluster control.

[0024] On the one hand, this application proposes an emergency group control system for distribution network loads under N-1 fault conditions in the main grid, comprising the following components:

[0025] Fault detection module: Real-time monitoring of the operating status of key equipment in the main network. When a single component without redundancy is detected in the main network, an emergency group control command is triggered.

[0026] Main and Distribution Network Coupling Analysis Module: Based on the port compensation method, establish the mapping relationship between main network faults and distribution network voltage sags, quantify the impact of main network faults on the voltage and power of distribution network boundary nodes, and generate fault equivalent models;

[0027] Operation boundary generation module: Combining distribution network topology and real-time load data, it calculates the safe operation boundary of the distribution network under fault conditions through a multi-parameter planning algorithm, and determines the maximum allowable power fluctuation range and voltage stability critical domain of each feeder;

[0028] Load clustering module: Based on the load characteristics, priority and geographical location of distribution network users, the load is divided into interruptible groups, adjustable groups and critical guarantee groups, and weight factors are assigned to each cluster;

[0029] Group control strategy generation module: Based on the fault equivalence model and safe operation boundary, dynamically generates a load group control instruction set, including:

[0030] Perform graded resection operations on interruptible groups;

[0031] Implement coordinated regulation of active and reactive power in the adjustable group;

[0032] Maintain power supply to critical backup groups and optimize reactive power compensation;

[0033] Command execution and feedback module: Sends group control commands to distribution network intelligent terminal equipment through wide area communication network, collects voltage, current and power data after load adjustment in real time, and verifies control effect;

[0034] Adaptive optimization module: When the feedback data deviates from the safe operating boundary, the load adjustment amount is recalculated and the group control instructions are iteratively updated until the distribution network voltage recovers to within the preset fluctuation threshold and the power is balanced.

[0035] On the other hand, this application proposes an emergency group control method for distribution network load under N-1 fault of the main network, including the following steps:

[0036] Step S1: Monitor the operating status of key equipment in the main network in real time. When a single component without redundancy is detected in the main network, trigger an emergency group control command.

[0037] Step S2: Establish the mapping relationship between main grid faults and distribution network voltage sags based on the port compensation method, quantify the impact of main grid faults on the voltage and power of distribution network boundary nodes, and generate fault equivalent models.

[0038] Step S3: Combining the distribution network topology and real-time load data, calculate the safe operation boundary of the distribution network under fault conditions using a multi-parameter programming algorithm, and determine the maximum allowable power fluctuation range and voltage stability critical domain for each feeder;

[0039] Step S4: Based on the load characteristics, priority and geographical location of distribution network users, divide the load into interruptible groups, adjustable groups and critical protection groups, and assign weight factors to each group.

[0040] Step S5: Based on the fault equivalent model and safe operation boundary, dynamically generate the load group control instruction set, including:

[0041] Perform graded resection operations on interruptible groups;

[0042] Implement coordinated regulation of active and reactive power in the adjustable group;

[0043] Maintain power supply to critical backup groups and optimize reactive power compensation;

[0044] Step S6: Send the group control command to the distribution network intelligent terminal equipment through the wide area communication network, collect the voltage, current and power data after load adjustment in real time, and verify the control effect;

[0045] Step S7: When the feedback data deviates from the safe operating boundary, recalculate the load adjustment amount and iteratively update the group control command until the distribution network voltage recovers to within the preset fluctuation threshold and the power is balanced.

[0046] Preferably, the step S2, which establishes the mapping relationship between main grid faults and distribution network voltage sags based on the port compensation method, specifically includes:

[0047] Establish a fault equivalent impedance model Z at the low-voltage side port v of the transformer in the distribution network. vf :

[0048] Z vf =f(Z) line Z load Z f )

[0049] Z line Z is the impedance matrix of the distribution network lines, representing the self-impedance and mutual impedance of the three-phase lines. load For the constant impedance load model, Z is calculated from real-time voltage and power. f The equivalent grounding branch impedance at the fault point is determined by the short-circuit type.

[0050] Using the transformer three-sequence model, Z vf Equivalent admittance Y mapped to high-voltage side port n nf :

[0051]

[0052] Where S is the symmetric component transformation matrix, and the superscripts 0, 1, and 2 represent the zero-order, positive-order, and negative-order components, respectively, transforming the phase components into zero-order, positive-order, and negative-order components. The order component matrix of the fault equivalent impedance;

[0053] Calculate the change in injected current ΔIn at the boundary nodes caused by a mains fault:

[0054] ΔI n =(Y nf -Y n0 )U n0

[0055] Where Y n0 U is the normal state equivalent admittance, and the admittance of the distribution network at the high-voltage side port n before the fault. n0 This is the boundary voltage before the fault.

[0056] Preferably, step S3, which calculates the safe operating boundary using a multi-parameter programming algorithm, specifically includes:

[0057] Construct a linear programming model with distribution network security constraints:

[0058]

[0059] Where P0 is the initial power vector in MW, representing the load / generation power before the fault, ΔP is the power adjustment in MW, and C... grid D gridThe constraint coefficient matrix is ​​generated from the distribution network power flow equation, and the dgrid is the constraint constant vector in MW, which includes limits such as line capacity.

[0060] Solving the critical region CR k and the corresponding linear mapping relationship:

[0061] CR k ={P|A k P≤b k},ΔP k =K k P+c k

[0062] Where A k ,b k K is the boundary coefficient matrix / vector of the critical region. k ,c k CR is the linear coefficient for power adjustment. k This represents the k-th critical region;

[0063] Determine the maximum allowable power fluctuation range for each feeder:

[0064]

[0065] in This represents the maximum allowable power fluctuation of feeder i, which is obtained by traversing the critical region and solving for the extreme values.

[0066] Preferably, the load cluster partitioning and weight allocation in step S4 specifically includes:

[0067] Based on user contracts and real-time monitoring data, clusters are divided according to priority:

[0068] Disruptive group: weight α I =0.1;

[0069] Adjustable group: weight α R =0.3;

[0070] Key Protection Group: Weight α C =0.6;

[0071] Calculate the cluster comprehensive weight factor β j :

[0072]

[0073] Where L j This represents the total load of cluster j.

[0074] Preferably, the dynamic generation of the load group control instruction set in step S5 specifically includes:

[0075] Interrupted group graded resection:

[0076]

[0077] Where ΔP req The total power to be reduced, The maximum cuttable quantity of an interruptible group;

[0078] Adjustable group active and reactive power coordinated regulation:

[0079]

[0080] Where K P ,K Q The gain coefficient is determined by sensitivity analysis to adjust the voltage. V is the real-time voltage. ref For reference voltage, The sensitivity of voltage to reactive power;

[0081] Key protection group reactive power compensation:

[0082]

[0083] Where X line Let V be the line reactance, and ΔV be the voltage deviation, where ΔV = V - Vref.

[0084] Preferably, the control effect verification in step S6 specifically includes:

[0085] Data after adjustment was collected using a wide-area measurement system.

[0086] ΔV actual =V post -V ref ΔP actual =P post -P0

[0087] Among them, V post To control the subsequent voltage, P post To control the power output, ΔV actual The actual voltage deviation is the difference between the controlled voltage and the reference voltage.

[0088] Check if the safety boundary is met:

[0089] |V actual |≤δ V ,||ΔP actual ||≤ΔP max

[0090] Where δV is the voltage fluctuation threshold, taken as ±1%V. ref ΔPmax is the upper limit of power fluctuation, ΔP actualThis is the actual power adjustment amount, the difference between the power after control and the initial power.

[0091] Preferably, the iterative update of the group control command in step S7 specifically includes: if |ΔVactual|>δ V Recalculate the power adjustment amount:

[0092]

[0093] Where γ is the iteration step size, which is taken from 0.1 to 0.5 to ensure convergence. ΔPold represents the voltage sensitivity to active power, ΔPold represents the power adjustment amount in the previous iteration, and ΔPnew represents the updated power adjustment amount.

[0094] Update and issue cluster control commands until the following conditions are met:

[0095] |ΔV actual |≤δ V And ||ΔP actual ||≤0.95ΔP max

[0096] Among them, 0.95ΔP max The safety margin coefficient is dimensionless to avoid boundary oscillations.

[0097] Preferably, in the three-sequence transformer model, the zero-sequence network of the connected transformer is open-circuited, and the positive / negative sequence networks satisfy:

[0098]

[0099] Where Sv is the phase rotation matrix. For the positive / negative sequence impedance of the transformer, This represents the positive / negative sequence impedance of the transformer.

[0100] Preferably, the critical region C Rk The solution employs a time-decoupling strategy:

[0101] Single-time-period constraints are independently represented as:

[0102] CR k,t ={P t |A k,t P t ≤b k,t}

[0103] Among them, P t Let A be the power variable over time period t. k,t ,b k,t The boundary coefficient for time period t;

[0104] The global critical region is the Cartesian product of the time-limited subdomains:

[0105]

[0106] Here, ∏ represents the Cartesian product mathematical operator, which combines subdomains of different time periods.

[0107] The beneficial effects of systematically improving this solution in light of the above-mentioned invention are as follows:

[0108] 1. Fault Modeling Layer: Embed the transformer full-sequence component coupling mechanism in the port compensation method to accurately characterize the impact of different wiring methods (such as YNd type) on fault propagation;

[0109] 2. Boundary Calculation Layer: Decomposes the global security boundary into independent sub-problems in a single time period, reducing the solution latency to the second level;

[0110] 3. Group control execution layer: Design a dynamic linkage mechanism between load weight and voltage deviation to improve reactive power compensation response speed to within 2 seconds;

[0111] 4. Iterative Mechanism: Based on voltage sensitivity, the control step size is adaptively adjusted to eliminate the risk of oscillation. Through actual grid measurements, this scheme reduces the time for the voltage to recover to within ±1% of the threshold after a fault to 8.3 seconds, which is 67% better than the traditional method, effectively avoiding the risks of excessive load reduction and secondary over-limit. Attached Figure Description

[0112] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0113] The present invention will be further described below with reference to embodiments and accompanying drawings:

[0114] like Figure 1 As shown, this solution provides an emergency group control method for distribution network loads under N-1 fault conditions in the main grid. Through a four-level collaborative mechanism—accurate fault modeling, dynamic safety boundary calculation, load cluster hierarchical control, and closed-loop verification iteration—it achieves rapid voltage recovery to within ±1% of the threshold after a fault. The specific process is as follows:

[0115] Step 1: Real-time fault monitoring and triggering. This step involves real-time monitoring of the operating status of critical main grid equipment (such as transformers and transmission lines). When a single component experiences a non-redundant backup fault (i.e., an N-1 fault), an emergency group control command is immediately triggered. This step utilizes a Wide Area Measurement System (WAMS) to achieve millisecond-level fault diagnosis, ensuring a response latency of less than 50 milliseconds.

[0116] Step 2: Quantifying the impact of main and distribution network faults. Based on the distribution network topology and real-time load data, a fault equivalent impedance model is constructed at the low-voltage side port of the transformer. This model integrates line impedance, constant load impedance characteristics, and grounding branch parameters at the fault point to accurately characterize the electrical characteristics at the moment of the fault.

[0117] By using the full-sequence component model of the transformer (especially for the zero-sequence open-circuit characteristics of YNd connection), the equivalent impedance of the low-voltage side is mapped to the boundary node of the high-voltage side to generate the equivalent fault admittance.

[0118] The change in injected current at the boundary node caused by the fault is calculated. This change is determined by the product of the difference in equivalent admittance before and after the fault and the boundary voltage before the fault, thus quantifying the impact of the main grid fault on the voltage sag and power surge of the distribution network.

[0119] Step 3: Dynamically generate safe operating boundaries

[0120] Calculation of the critical region for time-sharing decoupling:

[0121] The power distribution network security constraints are broken down into independent sub-problems for each time period, and a linear programming model is constructed for each time period. By solving this model, the feasible region (called the critical region) of the power adjustment amount for each time period is determined, avoiding the exponential computational complexity caused by global optimization.

[0122] Based on the critical domain boundary coefficient, the maximum allowable power fluctuation range and voltage stability threshold for each feeder are generated.

[0123] Dynamically link real-time data:

[0124] By injecting real-time data such as fluctuations in renewable energy output and load changes into the critical domain calculation, the safety boundary is ensured to be dynamically updated according to the operating status. Field tests show that this strategy reduces the calculation latency from 5 minutes using traditional methods to less than 1 second.

[0125] Step 4: Load Cluster Partitioning and Weight Allocation

[0126] Three-level cluster partitioning:

[0127] Interruptible groups: Low-priority loads such as commercial air conditioning and discontinuous production loads are allowed to be completely cut off;

[0128] Adjustable load group: Adjustable load for industrial and commercial applications (such as flexible production lines), supporting bidirectional adjustment of active and reactive power;

[0129] Critical protection groups: Sensitive loads such as hospitals and emergency facilities require absolutely guaranteed power supply.

[0130] Dynamic weighting factor design:

[0131] Based on the real-time total load and preset priority coefficients (0.1 for interruptible groups, 0.3 for adjustable groups, and 0.6 for critical support groups), the comprehensive weight factor of each cluster is calculated to ensure that critical load adjustment has the highest priority.

[0132] Step 5: Dynamically generate group control commands

[0133] Interrupted group graded resection:

[0134] The amount of power to be cut should be allocated proportionally based on the total power to be reduced and the cluster weight, so as to avoid excessive load reduction.

[0135] Adjustable group coordinated regulation:

[0136] Active power regulation: Adjust the power output proportionally based on the voltage deviation value (the difference between the real-time voltage and the reference voltage);

[0137] Reactive power compensation: Based on the voltage sensitivity coefficient to reactive power, the reactive power compensation amount is dynamically generated to suppress voltage fluctuations.

[0138] Key support group optimization:

[0139] Based on the line reactance and voltage deviation, the required reactive power compensation value is calculated and quickly injected through SVG / SVC equipment.

[0140] Step 6: Real-time data acquisition for closed-loop verification of control effect:

[0141] The system collects voltage, current, and power data after control via a smart terminal, and calculates the actual voltage deviation and power adjustment.

[0142] Security boundary verification:

[0143] Verify that the actual voltage deviation is ≤ ±1% and that the power adjustment does not exceed the allowable fluctuation range.

[0144] Step 7: Instruction Iterative Update

[0145] Deviation-driven recalculation:

[0146] If the voltage deviation exceeds the threshold, the power regulation amount is adaptively adjusted based on the voltage sensitivity coefficient to active power (the step size changes dynamically with the deviation).

[0147] Convergence criteria:

[0148] Repeatedly issue update commands until the voltage recovers to within ±1%, and retain a 5% safety margin for power adjustment to prevent boundary oscillations.

[0149] The beneficial effects of the above steps are as follows:

[0150] 1. Faults affect the improvement of quantization accuracy:

[0151] By using a full-sequence component coupling model of the transformer, the 20% mapping error caused by zero-sequence open circuit in traditional methods is resolved, and the voltage sag prediction accuracy is improved to 95%.

[0152] 2. Breakthrough in safety boundary calculation efficiency:

[0153] The time-period decoupling strategy reduces the optimization time for 24-hour periods from >5 minutes to <1 second, meeting the requirements for second-level response.

[0154] 3. Cluster control coordination optimization:

[0155] The dynamic weighting mechanism ensures that critical load losses such as those in hospitals are reduced by 42%, and the reactive power compensation response speed is within 2 seconds.

[0156] 4. Significantly enhanced voltage recovery capability:

[0157] Power grid measurements show that it only takes 8.3 seconds for the voltage to recover to within ±1% of the threshold after a fault, which is 67% faster than the traditional method.

[0158] The specific embodiments described above are only used to explain and illustrate the present invention, and are not intended to limit the present invention. Any changes and substitutions made to the present invention without creative effort within the scope of the inventive concept and claims shall fall within the protection scope of the present invention patent.

Claims

1. An emergency group control system for distribution network loads under N-1 fault conditions in the main grid, characterized in that, It includes the following components: Fault detection module: Real-time monitoring of the operating status of key equipment in the main network. When a single component without redundancy is detected in the main network, an emergency group control command is triggered. Main and Distribution Network Coupling Analysis Module: Based on the port compensation method, establish the mapping relationship between main network faults and distribution network voltage sags, quantify the impact of main network faults on the voltage and power of distribution network boundary nodes, and generate fault equivalent models; Operation boundary generation module: Combining distribution network topology and real-time load data, it calculates the safe operation boundary of the distribution network under fault conditions through a multi-parameter planning algorithm, and determines the maximum allowable power fluctuation range and voltage stability critical domain of each feeder; Load clustering module: Based on the load characteristics, priority and geographical location of distribution network users, the load is divided into interruptible groups, adjustable groups and critical guarantee groups, and weight factors are assigned to each cluster; Group control strategy generation module: Based on the fault equivalence model and safe operation boundary, dynamically generates a load group control instruction set, including: Perform graded resection operations on interruptible groups; Implement coordinated regulation of active and reactive power in the adjustable group; Maintain power supply to critical backup groups and optimize reactive power compensation; Command execution and feedback module: Sends group control commands to distribution network intelligent terminal equipment through wide area communication network, collects voltage, current and power data after load adjustment in real time, and verifies control effect; Adaptive optimization module: When the feedback data deviates from the safe operating boundary, the load adjustment amount is recalculated and the group control instructions are iteratively updated until the distribution network voltage recovers to within the preset fluctuation threshold and the power is balanced.

2. A method for emergency group control of distribution network loads under N-1 fault conditions in the main grid, characterized in that, Includes the following steps: Step S1: Monitor the operating status of key equipment in the main network in real time. When a single component without redundancy is detected in the main network, trigger an emergency group control command. Step S2: Establish the mapping relationship between main grid faults and distribution network voltage sags based on the port compensation method, quantify the impact of main grid faults on the voltage and power of distribution network boundary nodes, and generate fault equivalent models. Step S3: Combining the distribution network topology and real-time load data, calculate the safe operation boundary of the distribution network under fault conditions using a multi-parameter programming algorithm, and determine the maximum allowable power fluctuation range and voltage stability critical domain for each feeder; Step S4: Based on the load characteristics, priority and geographical location of distribution network users, divide the load into interruptible groups, adjustable groups and critical protection groups, and assign weight factors to each group. Step S5: Based on the fault equivalent model and safe operation boundary, dynamically generate the load group control instruction set, including: Perform graded resection operations on interruptible groups; Implement coordinated regulation of active and reactive power in the adjustable group; Maintain power supply to critical backup groups and optimize reactive power compensation; Step S6: Send the group control command to the distribution network intelligent terminal equipment through the wide area communication network, collect the voltage, current and power data after load adjustment in real time, and verify the control effect; Step S7: When the feedback data deviates from the safe operating boundary, recalculate the load adjustment amount and iteratively update the group control command until the distribution network voltage recovers to within the preset fluctuation threshold and the power is balanced.

3. The method according to claim 2, characterized in that, Step S2, which establishes the mapping relationship between main grid faults and distribution network voltage sags based on the port compensation method, specifically includes: Establish a fault equivalent impedance model Z at the low-voltage side port v of the transformer in the distribution network. vf : ； Z line Z is the impedance matrix of the distribution network lines, representing the self-impedance and mutual impedance of the three-phase lines. load For the constant impedance load model, Z is calculated from real-time voltage and power. f The equivalent grounding branch impedance at the fault point is determined by the short-circuit type. Using the transformer three-sequence model, Z vf Equivalent admittance Y mapped to high-voltage side port n nf : ； Where S is the symmetric component transformation matrix, and the superscripts 0, 1, and 2 represent the zero-order, positive-order, and negative-order components, respectively, transforming the phase components into zero-order, positive-order, and negative-order components. The order component matrix of the fault equivalent impedance; Calculate the change in injected current ΔIn at the boundary nodes caused by a mains fault: ； Where Y n0 U is the normal state equivalent admittance, and the admittance of the distribution network at the high-voltage side port n before the fault. n0 This is the boundary voltage before the fault.

4. The method according to claim 2, characterized in that, Step S3, which calculates the safe operating boundary using a multi-parameter programming algorithm, specifically includes: Construct a linear programming model with distribution network security constraints: ； Where P0 is the initial power vector in MW, representing the load / generation power before the fault, ΔP is the power adjustment in MW, and C... grid D grid The constraint coefficient matrix is ​​generated from the distribution network power flow equations. d grid is a constraint constant vector, in MW, which includes limits such as line capacity; Solving the critical region CR k and the corresponding linear mapping relationship: ； Where A k b k K is the boundary coefficient matrix / vector of the critical region. k c k CR is the linear coefficient for power adjustment. k This represents the k-th critical region; Determine the maximum allowable power fluctuation range for each feeder: ； in This represents the maximum allowable power fluctuation of feeder i, which is obtained by traversing the critical region and solving for the extreme values.

5. The method according to claim 2, characterized in that, The load cluster partitioning and weight allocation in step S4 specifically includes: Based on user contracts and real-time monitoring data, clusters are divided according to priority: Interruptible Groups: Weights ; Adjustable group: weight ; Key Protection Group: Weight ; Calculate the cluster comprehensive weight factor β j : ； Where L j This represents the total load of cluster j.

6. The method according to claim 2, characterized in that, The dynamic generation of the load group control instruction set in step S5 specifically includes: Interrupted group graded resection: ; Where ΔP req The total power to be reduced, The maximum cuttable quantity of an interruptible group; Adjustable group active and reactive power coordinated regulation: ； Where K P K Q The gain coefficient is determined by sensitivity analysis to adjust the voltage. V is the real-time voltage. ref For reference voltage, The sensitivity of voltage to reactive power; Key protection group reactive power compensation: ； Where X line Here, ΔV is the line reactance, and ΔV is the voltage deviation. .

7. The method according to claim 2, characterized in that, The control effect verification in step S6 specifically includes: Data after adjustment was collected using a wide-area measurement system. ； Among them, V post To control the subsequent voltage, P post To control the power output, Δ V actual The actual voltage deviation is the difference between the controlled voltage and the reference voltage. Check if the safety boundary is met: ； Where δV is the voltage fluctuation threshold, taken as ±1%. V ref ， ΔPmax is the upper limit of power fluctuation, Δ P actual This is the actual power adjustment amount, the difference between the power after control and the initial power.

8. The method according to claim 7, characterized in that, The iterative update of the group control command in step S7 specifically includes: if |ΔVactual|>δ v Recalculate the power adjustment amount: ； Where γ is the iteration step size, which is taken from 0.1 to 0.5 to ensure convergence. ΔPold represents the voltage sensitivity to active power, ΔPold represents the power adjustment amount in the previous iteration, and ΔPnew represents the updated power adjustment amount. Update and issue cluster control commands until the following conditions are met: ； Among them, 0.95Δ P max The safety margin coefficient is dimensionless to avoid boundary oscillations.

9. The method according to claim 3, characterized in that, In the three-sequence transformer model, the zero-sequence network of the connected transformer is open, and the positive / negative sequence networks satisfy the following: ； Where Sv is the phase rotation matrix. For the positive / negative sequence impedance of the transformer, This represents the positive / negative sequence impedance of the transformer.

10. The method according to claim 4, characterized in that, The critical region C RK The solution employs a time-decoupling strategy: Single-time-period constraints are independently represented as: ； Among them, P t Let A be the power variable over time period t. k,t b k,t The boundary coefficient for time period t; The global critical region is the Cartesian product of the time-limited subdomains: ； Here, ∏ represents the Cartesian product mathematical operator, which combines subdomains of different time periods.

Images