A method for large-area power outage fast self-recovery of distribution network based on multi-source cooperation

Abstract

This invention discloses a method for rapid self-healing recovery of large-area power outages in distribution networks based on multi-source collaboration, relating to the field of distribution network self-healing recovery control technology. The method includes: collecting equipment operating parameters at monitoring points along candidate recovery paths; dividing the time period into pre-interconnection and post-interconnection periods; extracting outward expansion, contraction, and continuous power loss; generating bidirectional drift intensity and fault core area scores; identifying stable fault core areas and the first set of disconnecting switches; constructing a recovery action graph based on these parameters; using an ant colony algorithm to search for the optimal recovery sequence; and determining online whether to regenerate the optimal recovery sequence based on drift residuals during execution. This invention can improve the accuracy of fault section identification and the stability and efficiency of large-area power outage recovery.

Description

Technical Field

[0001] This invention relates to the field of power distribution network self-healing and recovery control technology, and in particular to a method for rapid self-healing and recovery of large-area power outages in power distribution networks based on multi-source collaboration. Background Technology

[0002] In urban core area power distribution networks, underground cable ring networks, sectionalizing switches, tie switches, fault indicators, and distribution automation terminals collectively form the foundation for rapid recovery after large-scale power outages. When a power failure in the upstream supply or an underground cable fault causes a large-scale power outage, the system typically needs to continuously collect fault indication status, voltage, current, and switch status data at monitoring points along candidate recovery paths. It then needs to combine this data with changes in status before and after tie-off power transfer to determine the scope of the fault's impact before executing the first round of isolation and tie-off power transfer recovery. Because the underground cable network undergoes a continuous process of voltage maintenance, reverse power supply, and boundary reorganization after a fault occurs and before tie-off power transfer is completed, the fault boundary is prone to first expanding towards the healthy side and then retreating towards the actual fault area. This makes it difficult to identify the stable fault core area in a timely manner, and the isolation location and recovery path are also prone to fluctuations with boundary changes. Therefore, accurately distinguishing the actual continuous impact range of the fault from the boundary drift area during the recovery process in large-scale power outage scenarios, and achieving rapid and stable self-healing recovery accordingly, has become a crucial technical problem to be solved in the recovery control of this type of power distribution network.

[0003] Existing technologies for handling large-area power outage restoration typically focus on fault segment identification and tie-off control based on single fault indications, static energized states, or conventional restoration sequence generation. They lack targeted analysis of the continuous boundary changes before and after tie-off, making it difficult to simultaneously account for the differences between the boundary expansion phase, boundary retreat phase, and sustained power loss phase. This can lead to some monitoring points, even if they show fault indication fading and reverse power supply during restoration, still being mistakenly identified as fault-prone sections. It may also mask the truly affected stable sections with boundary changes, resulting in oversized isolation ranges, repeated adjustments to restoration paths, or re-triggered anomalies during restoration. Furthermore, existing restoration methods lack the ability to online verify boundary changes after restoration actions are executed. When the stable fault core area changes, it is difficult to reconstruct the restoration action diagram and search for a more reasonable restoration sequence in a timely manner, thus affecting the accuracy, stability, and overall efficiency of large-area power outage restoration. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies in the process of large-scale power outages in distribution networks, which make it difficult to accurately distinguish the actual continuous impact range of the fault from the boundary drift area, resulting in unstable fault isolation locations, easy repeated adjustments to the recovery path, and insufficient self-healing stability. Therefore, this invention proposes a method for rapid self-healing recovery of large-scale power outages in distribution networks based on multi-source collaboration.

[0005] To address the problems existing in the prior art, the present invention adopts the following technical solution: A method for rapid self-healing and recovery of large-area power outages in distribution networks based on multi-source collaboration includes: S1. Collect equipment operating parameters at monitoring points along the candidate recovery path, determine reference status quantities, and divide the period before and after the connection transfer into two phases, with the closing time of the first connection switch as the boundary. S2. Based on equipment operating parameters, the period before and after the power supply transfer, extract the outward expansion, retraction, and continuous power loss that characterize the changes in the fault boundary, and generate the bidirectional drift intensity and fault core area score of the monitoring point based on the outward expansion, retraction, and continuous power loss. S3. Set the judgment threshold based on the fault core area score, filter the target monitoring points and map them to the physical section, and extract the stable fault core area and the first round of disconnecting switches; S4. Construct a recovery action graph with stable faulty core regions as constraints, determine the comprehensive cost of candidate recovery sequences by combining bidirectional drift intensity, and search for the optimal recovery sequence in the recovery action graph using the ant colony algorithm. S5. Sequentially execute the isolation actions corresponding to the first set of disconnect switches and the communication transfer actions included in the optimal recovery sequence. After each action, update the bidirectional drift intensity and fault core area score online to obtain the drift residual. Based on the changes in the stable fault core area and the drift residual, determine whether to regenerate the optimal recovery sequence.

[0006] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention simultaneously collects fault indication status, three-phase voltage RMS value, branch current, and switch status along candidate recovery paths. It divides the period before and after power transfer by using the closing time of the first tie switch as the boundary, extracting the outward expansion, retraction, and continuous power loss respectively. Furthermore, it generates bidirectional drift intensity and fault core area scores, thereby distinguishing boundary drift areas from the actual continuously affected areas during large-area power outage recovery. This accurately extracts stable fault core areas and the first set of disconnect switches. Through this method, fault segment identification becomes more stable, reducing problems such as excessively large isolation ranges, oscillating isolation positions, and inaccurate fault core area judgments caused by the boundary first expanding and then retracting. This improves the accuracy of fault location and first-round isolation in large-area power outage scenarios, establishing a clear and reliable boundary foundation for subsequent tie-and-transfer recovery.

[0007] 2. This invention further constructs a recovery action graph with stable fault core areas as constraints, and incorporates the scale of unrecovered load, boundary drift risk, and degree of violation of operational constraints into the comprehensive evaluation of candidate recovery sequences, and obtains the optimal recovery sequence through ant colony algorithm. After the recovery action is executed, a drift residual is generated based on the updated bidirectional drift intensity and fault core area score, and it is determined whether to re-search for the optimal recovery sequence based on the change status of stable fault core areas and drift residuals. Through the above method, adaptive optimization of the interconnection and transfer path can be achieved while ensuring the continuous isolation of the fault core area, reducing the risk of repeated transfer, false recovery, and secondary abnormal triggering during the recovery process, while enhancing the response capability of recovery actions to boundary changes, and improving the stability, accuracy, and overall recovery efficiency of rapid self-healing recovery of large-area power outages in the distribution network. Attached Figure Description

[0008] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 A flowchart illustrating a method for rapid self-healing recovery of large-area power outages in a distribution network based on multi-source collaboration, provided in an embodiment of the present invention. Figure 2 This is a functional block diagram of a system for rapid self-healing recovery of a large-area power outage in a power distribution network based on multi-source collaboration, provided as an embodiment of the present invention. Detailed Implementation

[0009] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0010] Example: This example provides a method for rapid self-healing and recovery of large-area power outages in distribution networks based on multi-source collaboration. See [link to example]. Figure 1 Specifically, including: S1. Collect equipment operating parameters at monitoring points along the candidate recovery path, determine reference status quantities, and divide the period before and after the connection transfer into two phases, with the closing time of the first connection switch as the boundary. S2. Based on equipment operating parameters, the period before and after the power supply transfer, extract the outward expansion, retraction, and continuous power loss that characterize the changes in the fault boundary, and generate the bidirectional drift intensity and fault core area score of the monitoring point based on the outward expansion, retraction, and continuous power loss. S3. Set the judgment threshold based on the fault core area score, filter the target monitoring points and map them to the physical section, and extract the stable fault core area and the first round of disconnecting switches; S4. Construct a recovery action graph with stable faulty core regions as constraints, determine the comprehensive cost of candidate recovery sequences by combining bidirectional drift intensity, and search for the optimal recovery sequence in the recovery action graph using the ant colony algorithm. S5. Sequentially execute the isolation actions corresponding to the first set of disconnect switches and the communication transfer actions included in the optimal recovery sequence. After each action, update the bidirectional drift intensity and fault core area score online to obtain the drift residual. Based on the changes in the stable fault core area and the drift residual, determine whether to regenerate the optimal recovery sequence.

[0011] In an embodiment of the present invention, equipment operating parameters are collected at monitoring points along the candidate recovery path to determine reference state quantities. The time period before and after the communication transfer is divided, with the closing time of the first communication switch as the boundary. This embodiment is applied to a 10kV underground cable double-ring network power supply system in the core urban area, targeting large-scale power outages caused by upstream power supply failure. First, candidate restoration paths are determined, including all sectional switches and tie switches of the lost feeder. All monitoring points on the candidate restoration paths are consecutively numbered according to the topological order from the side of the lost power station to the side of the tie power source. Monitoring points are set on the power supply side and load side of each sectional switch and tie switch, with each monitoring point corresponding to a line fault indicator installed at a specific location. When multiple tie power sources exist in the candidate restoration path, they are ordered sequentially according to their transfer capacity from largest to smallest, and the corresponding monitoring points are consecutively numbered following the monitoring point sequence on the side of the lost power station. Each monitoring point synchronously collects fault indication status, three-phase voltage RMS value, branch current, and switch status as equipment operating parameters through a distribution automation terminal at a sampling interval of 2 milliseconds. The fault indication status is taken from the action holding signal of the line fault indicator installed at the corresponding monitoring point; a value of 1 is output after fault action, and a value of 0 is output after no action or reset. The effective values ​​of the three-phase voltages are obtained from the measurement results of the voltage transformers installed at the corresponding monitoring points, and are calculated using the full-cycle Fourier algorithm, with a calculation window of one power frequency cycle. Branch currents are obtained from the measurement results of the current transformers installed at the corresponding monitoring points. Switch status is obtained from the remote signaling change signal of the corresponding switchgear; the output value is 1 when the switch is in the closed position and 0 when it is in the open position. After completing the acquisition of equipment operating parameters, a time interval of 20 power frequency cycles before the fault initiation time is selected as the pre-fault reference time window. The reason for selecting this length of the pre-fault reference time window is that this time length can completely cover the stable operating state of the system before the fault, effectively avoiding the influence of instantaneous voltage fluctuations and harmonic interference on the reference calculation results. The reference voltage is calculated based on the effective values ​​of the three-phase voltages within the pre-fault reference time window. The formula for calculating the reference voltage is as follows: ,in Let r be the reference voltage value corresponding to the monitoring point numbered r. This is the median calculation function. This serves as a reference time window before the failure. For monitoring point numbered r, The effective value of the three-phase voltage at time t. The reason for using the median to calculate the reference voltage is that the median has a strong resistance to abnormal sampling values ​​and impulse interference. Compared with the arithmetic mean, it can effectively avoid the reference reference shift caused by individual abnormal sampling points, ensuring the robustness and accuracy of the reference voltage.

[0012] The direction of the reference current is calculated based on the branch current within the reference time window before the fault. The formula for calculating the direction of the reference current is as follows: ,in The direction of the reference current corresponding to the monitoring point numbered r is shown. For monitoring point numbered r, The branch current at time t. The reason for using the median to calculate the reference current direction is that it can effectively eliminate the interference of instantaneous harmonic currents under normal operating conditions before the fault on the current direction judgment, ensuring that the reference current direction can accurately represent the normal power transmission direction of the system before the fault. After completing the calculation of the reference voltage and reference current direction, the voltage validity is calculated based on the reference voltage. The formula for calculating the voltage validity is: ,in Let r be the voltage validity at time t. Let r be the effective value of the three-phase voltage at time t. It is a minimal positive number with the same dimensions as voltage, and its value is taken as... The reason for setting a very small positive number is to avoid the calculation error of zero denominator when the reference voltage is zero. The reason for calculating the voltage validity is to eliminate the differences in transformer ratio and line rated voltage between different monitoring points through normalization, and to convert the energized state of different monitoring points to a unified quantization range, so as to accurately characterize the degree of real-time voltage loss at each monitoring point.

[0013] The reverse power transmission indication is determined based on the consistency between the branch current and the reference current direction. The formula for calculating the reverse power transmission indication is as follows: in Let r be the reverse power supply indication at time t. This is a two-parameter function for determining sign consistency. When and When the signs are the same, the function outputs a value of 1; when... and When the sign is opposite, the function outputs -1; when... When the value is 0, the function outputs a value of 1. When the value is 0, the judgment is based on the positive direction of power transmission in the system before the fault. The consistency of sign. The reason for calculating the reverse power transmission indicator is to convert the consistency of current direction into a quantitative index of 0 to 1. When reverse power transmission occurs at the monitoring point, the reverse power transmission indicator is 1, and when no reverse power transmission occurs, it is 0. This can accurately characterize the reverse power transmission status of the monitoring point after the interconnection transfer.

[0014] After calculating the voltage validity and reverse power supply indication, the fault initiation time, the closing time of the first tie switch, and the last state change time after the tie switch transfer are obtained sequentially. The fault initiation time is determined as follows: when more than half of the monitoring points belonging to the power-loss feeder in the candidate recovery path simultaneously experience a drop in the three-phase voltage validity value to below 10% of the reference voltage of the corresponding monitoring point, and this overall state persists for more than 3 power frequency cycles, the first time this voltage drop exceeds the limit is marked as the fault initiation time. The reason for using this method to determine the fault initiation time is that it can effectively avoid misjudgment of the fault initiation time caused by non-fault disturbances such as instantaneous voltage drops and lightning interference, ensuring that the fault initiation time accurately corresponds to the actual time of fault occurrence. The closing time of the first tie switch is taken from the remote signaling change data of the distribution automation system. The moment when the first tie switch on the candidate recovery path changes from the open state to the closed state is marked as the closing time of the first tie switch. The method for determining the last state change time after the interconnection switch is as follows: After the first interconnection switch is closed, when there is no change in the remote signaling of all monitoring points on the candidate recovery path, and the change in the effective value of the three-phase voltage of all monitoring points does not exceed 0.5% of the corresponding reference voltage and the change in the branch current does not exceed 1% of the corresponding rated current, and this stable state lasts for more than 10 power frequency cycles, the time when the last remote signaling change or effective change in remote measurement occurred before this stable state is marked as the last state change time after the interconnection switch. The reason for using this method to determine the time is that it ensures that the electrical transient process caused by the transfer of power supply is completely quelled, the fault boundary state enters a stable phase, and the time period division is not affected by state fluctuations during the transient process.

[0015] After determining the three key moments, the period from the fault initiation moment to the closing moment of the first tie switch is divided into the pre-tie power transfer period. The period from the moment the first contact switch was closed to the moment of the last state change after the contact switch was switched on is divided into the post-contact switch period. The period before the communication transfer is used to extract feature information of the fault boundary expanding outward to the healthy side, and the period after the communication transfer is used to extract feature information of the fault boundary shrinking back.

[0016] It should be noted that the pre-interconnection and power transfer period refers to the time interval from the moment the fault begins to the moment the first interconnection switch closes after a large-scale power outage occurs in the distribution network. During this period, the power-loss section has not yet received power back through adjacent power supply paths, and the network is mainly in the stage of the fault propagation impact becoming apparent, the initial undervoltage state being maintained, and the initial exposure of boundary anomaly information. During this period, the fault indication status, voltage drop status, current interruption status, and local boundary expansion phenomena collected by each monitoring point mainly reflect the initial response formed by the combined effects of the original fault path, cable energy storage release, and the interruption of the original power supply relationship after the fault occurs. Therefore, this period is used to extract the cumulative change information of the fault boundary expanding to the healthy side and serves as the initial observation interval for identifying the actual fault impact range.

[0017] It should be noted that the period after the tie-and-switch power transfer refers to the time interval from the closing of the first tie switch to the end of the last state change after the tie-and-switch power transfer. During this period, some healthy areas in the power-loss section have been connected to adjacent power supply paths through the tie channel, and the network enters the process of power restoration, power flow redistribution, and boundary reorganization. During this period, the voltage recovery status, current direction change status, fault indication fading status, and local boundary shrinkage phenomenon collected by each monitoring point mainly reflect the subsequent response formed by reverse power supply, topology reconstruction, and local boundary correction after the tie-and-switch power transfer action is implemented. Therefore, this period is used to extract the change information of the fault boundary from the apparent expansion state to the convergence of the stable core region, and serves as the subsequent observation interval for judging the degree of boundary shrinkage and the effectiveness of the recovery action.

[0018] In embodiments of the present invention, based on equipment operating parameters, the period before and after the power supply transfer, the expansion, contraction, and continuous power loss characteristics of the fault boundary change are extracted. Based on the expansion, contraction, and continuous power loss, a bidirectional drift intensity and fault core area score for the monitoring point are generated, including: After completing the calculation of voltage validity, reverse power transmission indication, and the division of the pre- and post-interconnection power transfer periods, the degree of voltage loss is determined based on the voltage validity. The formula for calculating the degree of voltage loss is as follows: ,in Let r represent the voltage loss level at time t for monitoring point r. Let r be the voltage validity at time t. The formula is a maximum value function. The reason for using this formula to calculate the voltage loss degree is that voltage validity accurately represents the ratio of the real-time voltage at the monitoring point to the stable reference voltage before the fault. Subtracting the voltage validity from 1 directly yields the voltage loss ratio, converting the positive representation of voltage validity into the negative representation of voltage loss. Simultaneously, by using the maximum value function, the minimum value of the voltage loss degree is limited to 0, avoiding negative values ​​when the voltage validity exceeds 1 due to voltage rise after power transfer. This ensures that the physical meaning of the voltage loss degree is always non-negative; a larger value indicates a more severe voltage drop at the monitoring point and a clearer power outage state, which can be directly used for subsequent fault boundary feature extraction. In this embodiment, all remote signaling and telemetry quantities are sampled synchronously with a uniform sampling interval of 2 milliseconds. This ensures that fault indication status, voltage validity, and reverse power supply indication quantities at the same time come from the same sampling moment, avoiding calculation distortion caused by timing misalignment.

[0019] After calculating the voltage loss level, based on the fault indication status and voltage loss level during the period before the interconnection power transfer, the outward expansion quantity is extracted to characterize the cumulative extent of the fault boundary expanding towards the healthy side before the interconnection power transfer. In this embodiment, the line fault indicator used is a live-reset type fault indicator. When a fault occurs, the fault indication status changes from 0 to 1 and remains there. When the monitoring point line is restored to power and remains so for more than 5 power frequency cycles, the fault indication status automatically resets to 0. The formula for calculating the continuous time of the outward expansion quantity is as follows: ,in The expansion amount of monitoring point numbered r, For the period before the connection and transfer, the interval is defined as a left-closed, right-open interval. , This represents the fault indication status of monitoring point r at time t. Let r represent the voltage loss level at time t for monitoring point r. Let be the integral time element. The reason for using this formula to calculate the outward expansion is that the outward expansion is used to characterize the cumulative extent of the fault boundary's expansion towards the healthy side before the interconnection and power transfer. Only when the monitoring point simultaneously meets two conditions—the fault indication state is in an active state and there is a significant voltage deficiency—does it meet the characteristic of the fault boundary expanding to that monitoring point. By integrating the characteristic quantities that meet the conditions during the period before the interconnection and power transfer, the characteristic intensity of the entire outward expansion process can be fully accumulated, avoiding misjudgments caused by instantaneous interference at a single moment. Simultaneously, it can quantify the degree to which different monitoring points are affected by the boundary expansion. In actual engineering implementation, based on the equally spaced discrete sampling data of the distribution automation terminal, the continuous integration is converted into discrete summation calculation. The discrete calculation formula is: ,in The sampling interval should be consistent with the sampling interval of the equipment operating parameters. To determine the total number of valid sampling points within the period preceding the contact transfer, For the time corresponding to the k-th sampling point within the time period before the connection transfer, all sampling times are less than the closing time of the first connection switch. When there are no valid sampling points in the period before the power transfer, the external expansion value is 0. The reason for using the discrete summation method is that the equipment operating parameters collected by the distribution automation system are all discrete time series. This conversion method can ensure that technicians in the relevant technical field can directly complete the calculation of the external expansion based on the sampling data of the existing distribution terminal without additional hardware modification.

[0020] After extracting the expansion quantity, the contraction quantity is extracted based on the fault indication fading status, voltage validity, and reverse power supply indication during the period following the interconnection and power transfer. This extraction characterizes the cumulative degree of fault boundary contraction after the interconnection and power transfer. The fault indication fading status is defined as 1 minus the fault indication status at the corresponding time. When the fault indication status is 0, the fault indication fading status is 1; when the fault indication status is 1, the fault indication fading status is 0. The formula for calculating the continuous time of the contraction quantity is as follows: in Let r be the amount of retraction at monitoring point numbered r. For the period following the transfer of power, the interval is defined as a left-closed, right-open interval. , This represents the fault indication fading status of monitoring point r at time t. Let r be the voltage validity at time t. Let r be the reverse power supply indication at time t. This is an indicator function that outputs 1 when the condition within the parentheses is met and 0 when the condition is not met. The integral time element is used. The reason for setting a voltage validity criterion of not less than 0.9 is to clarify that a monitoring point voltage recovery to more than 90% of its pre-fault value constitutes effective voltage recovery, avoiding misjudgments caused by minor voltage disturbances. The reason for setting a reverse power supply indication of not less than 0.95 is to clarify that only monitoring points in a stable reverse power supply state are included in the receding feature accumulation, avoiding misjudgments of reverse power supply caused by sampling noise. The reason for using this formula to calculate the receding amount is that the receding amount characterizes the degree of accumulated receding of the fault boundary after the connection and power transfer. The core characteristic of fault boundary receding is that after the connection and power transfer, the monitoring point that originally experienced a fault action due to boundary expansion simultaneously exhibits three synchronous characteristics: fault indication fading, effective voltage recovery, and stable reverse power supply. By limiting the effective feature boundary through the indication function, and multiplying the three feature quantities, it can be ensured that only states that simultaneously meet all three conditions are included in the receding feature accumulation, effectively eliminating interference from non-receding scenarios. By integrating the feature quantities within the time period after the connection and power transfer, the feature intensity of the entire boundary receding process can be fully accumulated, quantifying the degree of boundary receding at different monitoring points.

[0021] After extracting the recovery amount, based on the fault indication status and voltage loss degree during the periods before and after the power supply transfer, the continuous power loss is extracted to characterize the degree of continuous voltage loss at the monitoring point throughout the entire recovery event. The formula for calculating the continuous time of the continuous power loss is as follows: ,in The continuous power loss of monitoring point r is... This is a combined time interval between the period before and after the communication and supply transfer, covering the complete event chain from the occurrence of the fault to the stabilization of the boundary state, with no overlap or gaps between the intervals. This represents the fault indication status of monitoring point r at time t. Let r represent the voltage loss level at time t for monitoring point r. Let be the integral time infinitesimal element. The reason for using this formula to calculate the continuous power loss is that the continuous power loss is used to characterize the degree of continuous voltage loss of the monitoring point throughout the entire recovery event. The monitoring point in the true stable fault core area will continuously maintain the fault action state and voltage loss state throughout the entire event chain, while the drift boundary point will only meet this condition for part of the time period. By integrating the feature quantity in the complete event chain over time, the total duration and intensity of the monitoring point being in the fault power loss state can be quantified, effectively distinguishing the true fault core area from the drift boundary point, and providing a core basis for subsequent fault core area identification.

[0022] It should be noted that the outward expansion refers to the cumulative amount used to characterize the apparent expansion of the fault boundary towards the healthy side during the period before the connection is transferred. It reflects the strength of the monitoring point's sustained under-pressure state while the fault indication remains valid. The larger the outward expansion, the longer the monitoring point exhibits a response state consistent with the outward movement of the fault boundary during the initial stage after the fault occurs. That is, the more likely the monitoring point is to be affected by the initial impact of the fault, the release of cable energy, or local boundary disturbances, thus presenting an apparent abnormal location further outward than the actual fault core area. Therefore, the outward expansion is used to describe the cumulative trend of the fault boundary expanding outward relative to the actual core area during the pre-recovery stage.

[0023] It should be noted that the shrinkage amount refers to the cumulative amount used to characterize the degree of convergence of the fault boundary from an outward expansion state to an inward state during the period after the interconnection and power transfer. It reflects the strength of the boundary retreat exhibited by the monitoring point when the fault indication gradually fades, the local voltage recovers, and the current direction changes to reverse power supply. The larger the shrinkage amount, the more obviously the monitoring point shows a trend of returning from the abnormal boundary point to the healthy power supply boundary after the interconnection and power transfer action is implemented. That is, the monitoring point is more likely to be a drift point that was mistakenly brought into the fault boundary in the previous stage, rather than a point that exists stably in the real fault core area. Therefore, the shrinkage amount is used to describe the cumulative trend of the fault boundary converging towards the real fault core area after the interconnection and power transfer.

[0024] It should be noted that sustained power loss refers to the cumulative amount of power loss during the entire process before and after the power reconnection, characterizing the period when a monitoring point remains in a state of effective fault indication and power outage. It reflects the stability of the monitoring point's power outage response throughout the entire recovery event chain. A larger sustained power loss indicates that the monitoring point maintains a power outage state under the influence of the fault for a longer period and is less likely to recover to normal due to partial power restoration after the power reconnection. Therefore, this monitoring point is more likely to be located within the actual fault core area or within the stability-affected area adjacent to the actual fault core area. Sustained power loss is used to characterize whether a monitoring point possesses sustained and stable fault attributes throughout the entire event process, thus providing a basis for distinguishing between the actual fault core area and the short-term drift boundary.

[0025] After extracting the expansion, contraction, and continuous power loss from all monitoring points, robust normalization based on the median and absolute median difference is performed on the expansion, contraction, and continuous power loss to obtain normalized expansion, contraction, and continuous power loss. The reason for using robust normalization based on the median and absolute median difference is that characteristic quantities in distribution network fault scenarios are easily affected by transient disturbances, sampling anomalies, and other factors, resulting in extreme outliers. Traditional maximum / minimum value normalization or mean / variance normalization is highly sensitive to outliers, easily leading to severe distortion of the normalization results. The median and absolute median difference, however, have strong resistance to outlier interference, ensuring the robustness and comparability of the normalization results even in scenarios with extreme values. Furthermore, no manually preset thresholds are required, making it adaptable to distribution network topologies of different sizes and fault scenarios of varying severity. The basic calculation rules for the absolute median difference are determined, and the formula for calculating the absolute median difference is as follows: ,in The absolute median difference corresponding to the feature sequence X. Let i be the i-th element in the feature sequence X. The absolute median difference is the absolute deviation of each element in the sequence from the median of the sequence. The reason for using this formula to calculate the absolute median difference is that the absolute median difference can accurately characterize the dispersion of the feature sequence, and is not affected by extreme outliers in the sequence. It can be used in conjunction with the median to achieve fully robust normalization.

[0026] The external expansion is robustly normalized, and the formula for calculating the normalized external expansion is as follows: in This is the normalized extrapolation value corresponding to the monitoring point numbered r. This refers to the expansion quantity corresponding to the monitoring point numbered r. The median of the extrapolation sequence for all monitoring points. The absolute median difference of the extrapolation sequences for all monitoring points. It is a very small positive number, and its value is [value]. Set the smallest positive number. The reason for this is to avoid the calculation error of a zero denominator when the sum of the median and the absolute median difference is zero. When the sum of the median and the absolute median difference of the expansion of all monitoring points is zero, the normalized expansion of all monitoring points is uniformly set to 0. The reason for using this formula to normalize the expansion is that by using the sum of the median and the absolute median difference as the normalization benchmark, the amplitude difference of the expansion under different fault scenarios can be eliminated, and the expansion of all monitoring points can be transformed into a unified dimensionless interval, achieving horizontal comparability of the expansion characteristics of different monitoring points.

[0027] The retraction amount is robustly normalized, and the formula for calculating the normalized retraction amount is as follows: in Let r be the normalized shrinkage amount corresponding to the monitoring point numbered r. Let r be the amount of retraction corresponding to the monitoring point numbered r. This represents the median of the shrinkage sequence for all monitoring points. The absolute median difference of the shrinkage sequence for all monitoring points. It is a very small positive number, and its value is [value]. When the sum of the median and absolute median differences of the retraction amounts at all monitoring points is 0, the normalized retraction amount for all monitoring points is uniformly set to 0. The reason for using this formula to normalize the retraction amount is that by adopting a robust normalization rule consistent with the expansion amount, the amplitude differences of the retraction amount under different fault scenarios can be eliminated, ensuring that the normalized expansion amount and the retraction amount are at the same quantization level, thus providing a unified numerical benchmark for the subsequent calculation of bidirectional drift intensity.

[0028] The continuous power loss is robustly normalized. The formula for calculating the normalized continuous power loss is as follows: ,in Let r be the normalized continuous power loss corresponding to the monitoring point numbered r. Let r be the continuous power loss corresponding to the monitoring point numbered r. The median of the continuous power loss sequence at all monitoring points. The absolute median difference of the continuous power loss sequence at all monitoring points. The value is a very small positive number. When the sum of the median and absolute median difference of the continuous power loss at all monitoring points is 0, the normalized continuous power loss at all monitoring points is uniformly set to 0. The reason for using this formula to normalize the continuous power loss is that by adopting a consistent and robust normalization rule, the amplitude difference of the continuous power loss under different fault scenarios can be eliminated, ensuring that the normalized continuous power loss can be reasonably matched with the bidirectional drift intensity, thus providing a reliable basis for the subsequent calculation of the fault core area score.

[0029] After robust normalization of the three characteristic quantities, a bidirectional drift intensity is generated based on the normalized expansion and normalized contraction quantities to characterize the degree of contraction after the fault boundary of the monitoring point expands. The formula for calculating the bidirectional drift intensity is as follows: ,in The bidirectional drift intensity corresponds to the monitoring point numbered r. This is the normalized extrapolation value corresponding to the monitoring point numbered r. Let r be the normalized shrinkage amount corresponding to the monitoring point numbered r. The reason for using this formula to calculate the bidirectional drift intensity is that the core feature of bidirectional drift at the fault boundary is that the monitoring point simultaneously exhibits obvious outward expansion and shrinkage characteristics. Multiplying the normalized outward expansion amount by the normalized shrinkage amount can simultaneously amplify the coupling effect of the two characteristics. When the monitoring point only has outward expansion characteristics or only has shrinkage characteristics, the bidirectional drift intensity value is low. When the monitoring point has significant outward expansion and shrinkage characteristics at the same time, the bidirectional drift intensity value increases significantly. This can accurately quantify the probability that the monitoring point belongs to the drift boundary point after outward expansion and shrinkage, perfectly matching the physical characteristics of bidirectional drift at the fault boundary.

[0030] After calculating the bidirectional drift intensity, a fault core region score is generated based on the normalized continuous power loss and the bidirectional drift intensity to characterize the probability that the monitoring point belongs to a true stable fault core region. The formula for calculating the fault core region score is as follows: ,in The fault core area corresponding to monitoring point number r is scored. Let r be the normalized continuous power loss corresponding to the monitoring point numbered r. The bidirectional drift intensity corresponds to the monitoring point numbered r. The value is a very small positive number. The reason for using this formula to calculate the fault core area score is that monitoring points within a true stable fault core area have the core characteristics of significant continuous power loss and weak bidirectional drift. This formula uses the normalized continuous power loss as the numerator and the sum of the normalized continuous power loss and bidirectional drift intensity as the denominator, which can constrain the fault core area score within the range of 0 to 1. The more significant the continuous power loss characteristic and the weaker the bidirectional drift characteristic of the monitoring point, the closer the fault core area score is to 1, indicating a higher probability that the monitoring point belongs to a true stable fault core area. Conversely, the more significant the bidirectional drift characteristic and the weaker the continuous power loss characteristic of the monitoring point, the closer the fault core area score is to 0, indicating a higher probability that the monitoring point belongs to a drift boundary point. This can achieve accurate differentiation between true fault core areas and drift boundary points, providing clear quantitative criteria for the subsequent identification of stable fault core areas.

[0031] It should be noted that bidirectional drift intensity refers to the combined strength of a monitoring point exhibiting both outward expansion and backward contraction of the fault boundary during the self-healing process of a large-scale power outage. It reflects the degree to which the monitoring point shows an abnormal response of extending its boundary towards the healthy side before the power supply is restored, and the degree of regression of the boundary towards the actual fault area after the power supply is restored. A larger bidirectional drift intensity indicates that the monitoring point more clearly experienced a process of first being apparently included in the fault boundary and then withdrawing from it. This means the monitoring point is more likely to be a drift point formed by the combined effects of transient disturbances, power restoration, and boundary resetting, rather than a point that remains consistently stable within the core of the actual fault. Therefore, bidirectional drift intensity is used to characterize the two-stage oscillation characteristics of the fault boundary during the recovery process and to identify boundary drift positions that easily interfere with the judgment of the fault core area.

[0032] It should be noted that the fault core area score is an evaluation metric used to characterize the probability that a monitoring point belongs to a true stable fault core area. Essentially, it reflects the relative relationship between the stability of the fault attributes maintained by the monitoring point throughout the recovery event and the significance of its bidirectional drift phenomenon. A larger fault core area score indicates that the monitoring point exhibits more prominent continuous power outages and continuous fault responses, while being less affected by the shrinkage after boundary expansion. Therefore, it is more likely to be located within or adjacent to the stable impact range of a true fault core area. Conversely, a smaller fault core area score indicates that although the monitoring point may exhibit anomalies at a certain stage, these anomalies are more likely to originate from boundary drift rather than the sustained effects of a true fault. Therefore, the fault core area score is used to screen stable fault core areas from multi-source time-series responses and provides direct evidence for subsequent isolation location determination and communication-based power transfer recovery path searching.

[0033] In an embodiment of the present invention, a judgment threshold is set based on the fault core area score, target monitoring points are screened and mapped to physical sections, and stable fault core areas and the first-round disconnect switch set are extracted, including: After calculating the fault core area scores for all monitoring points along the candidate recovery path, a decision threshold is determined based on the fault core area scores of all monitoring points. The formula for calculating the decision threshold is as follows: ,in The threshold for determining the faulty core area. The median of the fault zone score sequence for all monitoring points. The absolute median difference is the sum of the scores of all monitoring points for the fault core area. The calculation rule for the absolute median difference is consistent with that in the previous embodiment. The reason for using this formula to calculate the judgment threshold is that it is adaptively generated based on robust statistics of the fault core area score sequence. This allows it to adapt to different scales of distribution network topologies, fault scenarios of varying severity, and different numbers of monitoring points. Furthermore, the statistics based on the median and absolute median difference possess strong resistance to outlier interference, preventing threshold distortion caused by abnormal scores at individual monitoring points and ensuring the stability and reliability of the judgment results. When the fault core area scores of all monitoring points are exactly the same, the judgment threshold is taken as this unified score value.

[0034] After calculating the judgment threshold, monitoring points with fault core area scores not less than the judgment threshold are selected as target monitoring points from the topologically ordered monitoring point sequence. The topologically ordered monitoring point sequence is the sequence of monitoring points numbered consecutively according to the topological order from the power substation side to the tie power source side, as described in the previous embodiment. This ensures that the selected target monitoring points completely correspond to the actual physical topology of the distribution network, avoiding deviations in subsequent segment mapping due to disordered monitoring point sorting. During the selection process, the fault core area score of each monitoring point is compared with the judgment threshold one by one. When the fault core area score of a monitoring point is greater than or equal to the judgment threshold, that monitoring point is marked as a target monitoring point. When the fault core area scores of all monitoring points are less than the judgment threshold, the monitoring point that is the first to experience a drop in the effective value of the three-phase voltage to below 10% of the corresponding reference voltage at the fault initiation time, and that this drop continues for more than 3 power frequency cycles, is marked as a target monitoring point. This covers extreme weak fault characteristic scenarios and ensures the completeness of fault segment identification.

[0035] After selecting target monitoring points, these points are mapped to physical segments between adjacent sectionalizing switches. All sectionalizing switches on the candidate recovery path are consecutively numbered according to a topological order identical to the monitoring point sequence. The underground cable segment between two consecutively numbered sectionalizing switches is defined as an independent physical segment. Each physical segment uniquely corresponds to a set of upstream and downstream monitoring points. The upstream monitoring point is the load-side monitoring point of the upstream sectionalizing switch in that physical segment, and the downstream monitoring point is the power-side monitoring point of the downstream sectionalizing switch. Each monitoring point uniquely belongs to only one physical segment. The load-side monitoring point of the upstream sectionalizing switch uniquely belongs to its downstream physical segment, and the power-side monitoring point of the downstream sectionalizing switch uniquely belongs to its upstream physical segment. When at least one target monitoring point exists in the unique set of upstream and downstream monitoring points corresponding to a physical segment, that physical segment is marked as a fault-associated segment. The reason for adopting this mapping rule is that it can accurately map the discrete monitoring point scoring results to the actual physical cable sections of the distribution network, realize a one-to-one correspondence between electrical characteristic quantities and physical fault locations, and avoid the problems of misalignment and duplicate marking of monitoring points and fault sections.

[0036] After completing the physical segment mapping, the smallest continuous segment covering all fault-related segments is extracted from all mapped fault-related segments as the stable fault core area. During the extraction process, all fault-related segments are first sorted according to a topological order consistent with the monitoring point sequence. The first sorted fault-related segment is determined as the starting segment, and the last sorted fault-related segment is determined as the ending segment. All continuous physical segments between the starting and ending segments are integrated into the smallest continuous segment covering all fault-related segments. Discretely distributed fault-related segments are defined as those where at least one non-fault-related segment separates two fault-related segments. Even with discretely distributed fault-related segments, the upstream fault-related segment is still considered the starting segment, and the downstream fault-related segment is considered the ending segment; all intermediate continuous physical segments are included in the smallest continuous segment. The reason for adopting this extraction rule is that the impact range of underground cable faults has the characteristic of continuous propagation. There is a high probability that there are hidden fault features that have not been captured by the monitoring points in the sections between discrete fault-related sections. Including the intermediate section can ensure the physical continuity of the fault core area and avoid incomplete fault isolation due to section omission. At the same time, using the upstream and downstream fault-related sections as the boundary can ensure the minimization of the isolation range and avoid unnecessary load loss.

[0037] After extracting the stable fault core area, the sectionalizing switches surrounding both sides of the stable fault core area are determined as the first set of disconnecting switches. The sectionalizing switch corresponding to the upstream boundary of the stable fault core area closest to the side of the lost electronic station is the upstream disconnecting switch, which is the upstream sectionalizing switch for the starting section of the stable fault core area. The sectionalizing switch corresponding to the downstream boundary of the stable fault core area closest to the tie power supply side is the downstream disconnecting switch, which is the downstream sectionalizing switch for the ending section of the stable fault core area. The upstream and downstream disconnecting switches together form the first set of disconnecting switches. During the selection process, priority is given to sectionalizing switches that are topologically directly adjacent to the boundary of the stable fault core area, whose rated breaking current matches the maximum short-circuit current of the line, whose remote signaling status has not changed for three consecutive sampling periods, and whose telemetry error does not exceed 0.5% of the rated value. Topologically directly adjacent is defined as having no other sectionalizing switch intervals between the sectionalizing switch and the boundary section of the stable fault core area. The reason for adopting this selection rule is that it can achieve complete electrical isolation between the stable fault core area and the upstream and downstream healthy sections with minimal switching operations. This provides a fault-free safety boundary for subsequent interconnection and power restoration, avoiding unnecessary load losses due to excessive isolation range, and preventing secondary tripping caused by incomplete fault isolation due to insufficient isolation range. When the starting section of the stable fault core area is the first physical section downstream of the incoming switch of the power-loss feeder, it is determined that the upstream boundary extends to the busbar side of the power-loss substation, and the incoming switch of the power-loss feeder is included in the first round of disconnecting switches. When the ending section of the stable fault core area is the last physical section upstream of the interconnecting switch, it is determined that the downstream boundary extends to the interconnecting switch side, and the corresponding interconnecting switch is included in the first round of disconnecting switches. This covers extreme scenarios of boundary extension.

[0038] It should be noted that a stable fault core area refers to the smallest continuous physical segment extracted from all abnormal response segments during the self-healing recovery process of a large-scale power outage. This segment is based on the fault core area scores of each monitoring point, topological location relationships, and boundary continuity constraints, and can stably represent the continuous impact range of the actual fault. The stable fault core area does not reflect a temporary, apparent abnormal boundary at a particular moment, but rather the core affected area that maintains strong fault attributes throughout the entire event chain before and after power transfer and does not significantly fluctuate with the outward expansion and subsequent contraction of the boundary. The more clearly defined the stable fault core area, the more thoroughly the drift components in the fault boundary are stripped away, and the more concentrated the continuous impact range of the actual fault. Therefore, the stable fault core area serves as a direct constraint object for subsequent isolation operations and recovery path construction to avoid mistaking short-term drift points as actual fault segments, thus expanding the power outage area or triggering repeated power transfers.

[0039] It should be noted that the initial set of disconnecting switches refers to a group of sectional switches defined around the boundaries of the stable fault core area, which are prioritized for disconnection at the beginning of the recovery process. The purpose of this switch set is to first electrically separate the stable fault core area from the healthy sections outside it, allowing subsequent communication and power transfer actions to be established on the network foundation that has already achieved fault core isolation. The initial set of disconnecting switches does not correspond to arbitrary switch combinations, but rather to the boundary switch set that directly corresponds to the physical location of the stable fault core area and can surround and disconnect it. Therefore, it essentially reflects the initial physical isolation of the actual fault's continued impact range. The initial set of disconnecting switches is used to quickly establish a clear boundary between the fault area and the recoverable area in the initial recovery phase, thereby providing the preconditions for subsequent communication and power transfer and self-healing recovery of the healthy sections.

[0040] In embodiments of the present invention, a recovery action graph is constructed with stable faulty core regions as constraints, the comprehensive cost of candidate recovery sequences is determined by combining bidirectional drift intensity, and the optimal recovery sequence is obtained by searching the recovery action graph using an ant colony algorithm, including: A recovery action graph is constructed using feasible network states after isolating the stable fault core area as nodes, where edges between nodes represent legal switching operations. The initial node of the recovery action graph is defined as the network state where, after executing the first round of disconnecting switch operations, the stable fault core area is completely electrically isolated from upstream and downstream healthy sections, all non-faulty sections are de-energized, and all tie switches and sectionalizing switches of non-faulty sections are in the open state. The uniqueness rule for feasible network states is that two network states are considered the same node only if their topological connectivity, the open / closed status of all switches, the energized status of all sections, and the load status of all tie power sources are completely identical. The terminating node of the recovery action graph is defined as the network state where all non-faulty sections form a connected path with the tie power source, or where there are no remaining legal switching operations to execute. Each of the remaining nodes corresponds to the network topology state, switch open / closed state, section energized state, and load power supply state after a valid switch operation is completed. The edges between nodes only connect two network state nodes that are associated with a single switch closing operation. The direction of the edges is from the node before the operation to the node after the operation, thus constructing a complete directed acyclic recovery action graph.

[0041] After completing the basic framework of the recovery action diagram, a switching operation that simultaneously satisfies the following conditions—not crossing the stable fault core area, meeting the power supply requirements of the current topology, and satisfying the power flow feasibility under the current state—is determined to be a legal switching operation. Legal switching operations are only applicable to sectionalizing switches and tie switches that are in the open state and located in non-faulty sections, and the operation method is a single-switch closing operation. The rule for determining not crossing the stable fault core area is that the topology connection path corresponding to the switching operation must not pass through any physical section within the stable fault core area, and the network topology after the operation must not directly electrically connect the stable fault core area to any tie power source or lost substation bus, thereby ensuring that the fault core area remains completely isolated and preventing the fault range from expanding during the recovery process. The rule for satisfying the power supply requirements of the current topology is that in the network topology after the switching operation, the load section to be restored corresponding to the switching operation must have at least one fault-free continuous path connected to a tie power source, and the available reserve capacity of the tie power source corresponding to this path must not be less than the sum of the total active power of all loads to be restored on this path. The available reserve capacity of the interconnecting power supply is calculated by subtracting the total active power currently continuously supplied by the interconnecting power supply after the fault from the rated maximum power supply capacity of the upstream substation of the interconnecting power supply. In scenarios where multiple interconnecting power supplies are used together, the available reserve capacity of each interconnecting power supply is calculated separately, and the power supply capacity is aggregated for statistical purposes. This ensures that the interconnecting power supply has sufficient power supply capacity to support the load to be restored after the restoration operation, avoiding power overload. The criteria for determining the feasibility of power flow under the current state are as follows: after the switching operation, the network topology is calculated using the distribution network forward-backward substitution method. In the power flow calculation, the load adopts a constant power model, and the load power factor is kept consistent with the reference time window before the fault. The interconnecting power supply is set as the balancing node, the voltage reference is the system rated voltage, the iteration convergence accuracy is set to 0.0001, the maximum number of iterations is set to 50, and the load current of all lines in the calculation results does not exceed the rated current carrying capacity of the corresponding line, and the effective voltage deviation of all load nodes does not exceed ±7% of the system rated voltage. This ensures that the network state after the restoration operation meets the national standard requirements for safe operation of the distribution network, avoiding equipment overload or voltage exceeding the limit. After the switch closing operation is completed, the rule for determining the energized state of a section is that if the section forms a fault-free continuous connection path with at least one interconnecting power source, it is determined to be energized; otherwise, it is determined to be de-energized.

[0042] After determining the rules for valid switching operations and constructing the recovery action diagram, for any candidate recovery sequence, the total load power that remains unrecovered after executing the candidate recovery sequence is determined. The total bidirectional drift intensity along the recovery path corresponding to the candidate recovery sequence is determined as the drift risk intensity, and the combined constraint violation amount consisting of current exceeding limits and voltage instability is determined. A candidate recovery sequence is defined as an operation sequence that starts from the initial node of the recovery action diagram and sequentially executes a single switch closing operation along the connecting edge direction until the termination node is reached. Each element in the sequence corresponds to one valid switch closing operation. The formula for calculating the total load power that remains unrecovered is... ,in To execute candidate recovery sequences The total load power that has not yet been restored To execute candidate recovery sequences The set of load nodes that are still in a state of power loss Let be the rated total active power of the nth load node. The reason for using this formula to calculate the total load power that has not yet been restored is that it can directly quantify the load restoration effect of the candidate restoration sequence. The smaller the value, the larger the scale of load that can be restored by the sequence, and the better the restoration effect.

[0043] The formula for calculating the intensity of drift risk is as follows: ,in Candidate recovery sequence The corresponding drift risk intensity, Candidate recovery sequence After execution, the set of monitoring points corresponding to all energized sections will be generated. The bidirectional drift intensity corresponding to the monitoring point numbered r is consistent with the bidirectional drift intensity calculated in the previous embodiment. The reason for using this formula to calculate the drift risk intensity is that the total bidirectional drift intensity of the energized section after recovery can directly quantify the risk of reactivating the bidirectional drift of the fault boundary during the execution of the recovery sequence. The larger the value, the higher the risk of the sequence triggering boundary re-drift and causing secondary malfunction.

[0044] The formula for calculating the total amount of violations of the comprehensive constraint is as follows: in Candidate recovery sequence The corresponding comprehensive constraint violation volume, To execute candidate recovery sequences Then, all energized lines and load nodes are collected. Calculate the effective value of the power flow current for the m-th line within the set. Let m be the rated current carrying capacity of the m-th line. Let be the effective voltage value of the m-th load node within the set. This is the lowest permissible effective voltage value for the load node, taken as 93% of the system's rated voltage. This is the highest permissible effective voltage value for the load node, taken as 107% of the system's rated voltage. This is a function for maximizing the value. The reason for using this formula to calculate the comprehensive constraint violation amount is that it only accumulates the current and voltage exceeding the safe operating range in the energized area after restoration. The contribution value of the part that does not exceed the limit is 0. This can accurately quantify the total degree of violation of the distribution network safety operation constraints after the candidate restoration sequence is executed. The larger the value, the higher the operational safety risk of the sequence.

[0045] After calculating the three core indicators, an initial candidate recovery sequence set is generated. This initial candidate recovery sequence set consists of 100 non-repeating candidate recovery sequences generated using a random walk based on the recovery action graph. The random walk constraint rules are: starting from the initial node, proceeding along the connecting edge direction, not repeatedly passing through the same node, until reaching the termination node. The criterion for non-repeating sequences is that the switching operation sequence and operation object of two sets of sequences are not completely consistent, and the number of operation steps of each set of sequences does not exceed the total number of segment switches and tie switches in the non-faulty section of the candidate recovery path. Based on the statistical results of the total unrecovered load power, drift risk intensity, and comprehensive constraint violation amount corresponding to the initial candidate recovery sequence set, the total unrecovered load power, drift risk intensity, and comprehensive constraint violation amount of the current candidate recovery sequence are normalized, and the comprehensive cost of the candidate recovery sequence is generated based on the normalization result. The normalization process adopts the robust normalization rules consistent with the aforementioned embodiments to eliminate the differences in dimensions and amplitudes between different indicators. The normalization benchmark is uniformly adopted using the statistical results of the initial candidate recovery sequence set, and all subsequent candidate recovery sequences are normalized using this benchmark. The normalized calculation formula for the total unrestored load power is as follows: ,in Candidate recovery sequence The corresponding normalized total unrestored load power, The sequence is composed of the total unrecovered load power of all sequences in the initial candidate recovery sequence set. This is the median calculation function. The absolute median difference calculation function follows the same calculation rules as in the aforementioned embodiments. It is a very small positive number. When the sum of the difference between the median and the absolute median of the total unrecovered load power of all sequences in the initial candidate recovery sequence set is 0, the normalized total unrecovered load power of all sequences is uniformly set to 0.

[0046] The normalized formula for calculating drift risk intensity is as follows: ,in, Candidate recovery sequence The corresponding normalized drift risk intensity, This is a sequence composed of the drift risk intensities of all sequences in the initial candidate recovery sequence set. It is a very small positive number. When the sum of the median and absolute median differences of the drift risk intensity of all sequences in the initial candidate recovery sequence set is 0, the normalized drift risk intensity of all sequences is uniformly set to 0. The normalized calculation formula for the comprehensive constraint violation is as follows: ,in Candidate recovery sequence The corresponding normalized comprehensive constraint violation amount, The sequence is composed of the combined constraint violations of all sequences in the initial candidate recovery sequence set. The value is a very small positive number. When the sum of the median and absolute median differences of the comprehensive constraint violations of all sequences in the initial candidate recovery sequence set is 0, the normalized comprehensive constraint violation value of all sequences is uniformly set to 0. The reason for adopting this robust normalization rule is that, based on the statistical characteristics of the initial candidate recovery sequence set, an adaptive normalization benchmark is generated, which can eliminate the dimensional differences between different indicators, ensure that the three indicators are at the same quantitative level, and at the same time have a strong ability to resist outlier interference, avoiding distortion of the normalization results caused by individual extreme sequences.

[0047] After normalizing the three indicators, the comprehensive cost of the candidate recovery sequence is generated. The formula for calculating the comprehensive cost is as follows: ,in Candidate recovery sequence The corresponding comprehensive cost. The reason for using this formula to calculate the comprehensive cost is that the three normalized indicators correspond to the three core optimization objectives of self-healing recovery of large-scale power outages in distribution networks: load restoration effect, boundary drift risk, and power grid operation safety. Linear addition can achieve balanced optimization of multiple objectives without the need for manual weight setting, avoiding the influence of subjective bias on the optimization results. The smaller the comprehensive cost value, the better the comprehensive performance of the candidate recovery sequence, providing a clear and reproducible objective function for subsequent optimization algorithms.

[0048] After constructing the recovery action graph and determining the comprehensive cost calculation rules for candidate recovery sequences, the ant colony algorithm is used to search for the optimal recovery sequence in the recovery action graph. First, the initial settings for the ant colony algorithm are completed, with the ant population size set to 50, the maximum number of iterations set to 200, and the initial pheromone value set to... Set as The initial pheromone concentration of all legal connections in the restored motion graph is uniformly set to... The reason for setting this initial value and uniform assignment rule is to ensure that all legal edges in the initial iteration phase have the same pheromone base, avoiding inherent bias in the search due to differences in initial pheromone, and ensuring the algorithm's global exploration capability. At the same time, the iteration stagnation threshold is set to 30 iterations, and the criterion for determining iteration stagnation is that the decrease in the global optimal comprehensive cost after 30 consecutive iterations is less than... An iteration is considered stalled when this condition is met, and the iteration is terminated early to balance the algorithm's optimization accuracy and computational efficiency. A pheromone importance factor is set. A value of 1 indicates the importance factor of heuristic information. The value is set to 2. This parameter is set to balance the algorithm's global exploration capability and local optimization capability, preventing the algorithm from getting stuck in local optima or having excessively slow convergence. This sets the minimum pheromone value. for maximum pheromone content for The reason for setting these upper and lower limits is to avoid premature convergence of the algorithm due to excessive accumulation of pheromones, and to avoid loss of effective path information due to excessive decay of pheromones.

[0049] After completing the initialization settings, the local comprehensive cost of a single-step action is determined as the basis for generating heuristic information. First, the local comprehensive cost of a single-step action is defined. A single-step action corresponds to the edge connecting the current state node x to the next state node y in the restored action graph. The local comprehensive cost of a single-step action... The calculation formula is ,in Let $x$ be the local comprehensive cost of a single-step action from node $x$ to node $y$. This is the normalized value corresponding to the change in the total unrestored load power before and after executing this single-step action. This is the normalized value corresponding to the change in the total bidirectional drift intensity of the newly added charged segment before and after performing this single-step action. To determine the normalized value corresponding to the change in the overall constraint violation amount of the energized region before and after executing the single-step action, all normalization calculations use the unified normalization benchmark determined by the initial candidate recovery sequence set in the aforementioned embodiments. The change amount is the value after executing the single-step action minus the value before execution. The reason for using this formula to calculate the local comprehensive cost of the single-step action is that the calculation rules for the local cost of the single step and the comprehensive cost of the entire sequence are completely consistent. By using the change amount, the change in comprehensive cost brought about by a single switching operation can be accurately quantified, ensuring the consistency between heuristic information and global optimization objectives, so that the ant's single-step selection always serves the global optimal objective.

[0050] After completing the local comprehensive cost calculation for a single-step action, heuristic information for the corresponding edges is generated. The formula for calculating the heuristic information is as follows: ,in This provides heuristic information for the edges connecting node x to node y. It is a very small positive number, and its value is [value]. The reason for using this formula to calculate heuristic information is that heuristic information is inversely proportional to the local comprehensive cost of a single-step action. The smaller the local comprehensive cost, the greater the heuristic information corresponding to the edge, which can guide the ant to prioritize the single-step action with better comprehensive performance, thus accelerating the convergence speed of the algorithm. After completing the heuristic information calculation, the transition probability between state nodes is determined based on the heuristic information and pheromone to generate multiple candidate recovery sequences. Each ant starts from the initial node of the recovery action graph to perform path search. Each ant maintains a tabu list, which records the nodes that the ant has already visited in this search. Each time an ant completes a node transition, it immediately adds the newly arrived node to the tabu list. The ant cannot choose the edge corresponding to a node that already exists in the tabu list. Before each iteration, all ants' tabu lists are completely cleared and reset to ensure that there are no duplicate switching operations in the candidate recovery sequences and no backtracking of the network state. The formula for calculating the transition probability from the current state node x to the candidate next state node y of the k-th ant in the q-th iteration is as follows: ,in Let be the probability that the k-th ant moves from node x to node y. Let x be the current pheromone concentration corresponding to the edge connecting node x to node y. This provides heuristic information for the edge connecting node x to node y. Let x be the set of all legal edges pointing to the next state nodes that are not included in the tabu list. For pheromone importance factors, This is the importance factor for heuristic information. The reason for using this formula to calculate the transition probability is that it can combine the historical path experience represented by pheromones with the single-step action benefits represented by heuristic information, achieving a balance between global exploration and local optimization, so that ants can choose the path with better overall performance with a higher probability.

[0051] After an ant completes a node transfer, it checks if it has reached the termination node of the recovery action graph. If it reaches the termination node, the current path search is terminated, and a valid candidate recovery sequence is generated. If an ant has no legal edges to traverse at the current node and has not reached the termination node, the current path search is terminated, the generated sequence is deemed invalid and discarded, and is not included in subsequent comprehensive cost calculations and pheromone updates. After completing the path search for all ants in this iteration, all valid candidate recovery sequences for this iteration are obtained. Based on the comprehensive cost calculation formula in the aforementioned embodiment, the comprehensive cost of each valid candidate recovery sequence is calculated. Simultaneously, after each iteration, the comprehensive cost of the generated valid sequence is compared with the globally saved optimal comprehensive cost. If a smaller comprehensive cost exists, the globally optimal comprehensive cost and the corresponding optimal candidate recovery sequence are updated to ensure the complete preservation of the globally optimal sequence. Subsequently, the adaptive evaporation rate is determined based on the comprehensive cost of the generated valid candidate recovery sequences. The adaptive evaporation rate corresponding to the q-th iteration is... The calculation formula is ,in The pheromone adaptive evaporation rate for the q-th iteration. Let be the median of the combined cost of all valid candidate recovery sequences in the q-th iteration. The maximum value of the combined cost of all valid candidate recovery sequences in the q-th iteration is given by . It is a very small positive number, and its value is [value]. .pass and The function constrains the evaporation rate within the range of 0.1 to 0.9. The reason for using this formula to calculate the adaptive evaporation rate is that it allows for dynamic adjustment of the evaporation rate based on the overall cost distribution of candidate recovery sequences in the current iteration. Simultaneously, upper and lower bound constraints prevent algorithm failure caused by an evaporation rate of 0 or 1 in extreme scenarios. When the overall cost difference between sequences is large in the current iteration, a smaller evaporation rate preserves the pheromones of high-quality paths, accelerating algorithm convergence. Conversely, when the overall cost difference between sequences is small in the current iteration, a larger evaporation rate weakens the pheromone influence of historical paths, enhancing the algorithm's global exploration capability and avoiding getting trapped in local optima.

[0052] After completing the adaptive volatility calculation, the pheromone is updated based on the comprehensive cost of each effective candidate recovery sequence. The pheromone update formula is as follows: in Let be the pheromone concentration corresponding to the edge connecting node x to node y in the (q+1)th iteration. The pheromone concentration corresponding to this edge in the q-th iteration. Let q be the adaptive evaporation rate for the qth iteration. The indicator function is defined as follows: when the edge from node x to node y belongs to the k-th valid candidate recovery sequence. When the condition is met, the indicator function outputs 1; otherwise, it outputs 0. When the same edge is contained by multiple valid sequences, the pheromone increments are directly added together. For the kth valid candidate recovery sequence The corresponding overall cost, It is a very small positive number. The reason for using this formula to update pheromones is that the historical information of inferior paths can be weakened through the evaporation mechanism, while the pheromone of superior paths can be strengthened through the incremental mechanism. The superior path with a smaller overall cost corresponds to a larger pheromone increment, which can guide the ants in subsequent iterations to prioritize the path and gradually converge towards the global optimum.

[0053] After the pheromone update is completed, the updated pheromone concentration is limited to between the minimum and maximum pheromone values. When the updated pheromone concentration is less than the minimum pheromone value, it is set as the minimum; when the updated pheromone concentration is greater than the maximum pheromone value, it is set as the maximum. This prevents pheromone saturation or disappearance, completing one full iteration. Then, it is determined whether the termination conditions are met. There are two termination conditions: the first is that the number of iterations reaches the preset maximum number of iterations; the second is that the number of consecutive iterations reaches the preset iteration stagnation threshold and meets the iteration stagnation judgment condition. The algorithm stops iterating if either of the two termination conditions is met. After the iteration stops, the candidate recovery sequence corresponding to the globally optimal comprehensive cost, which is saved throughout the process, is output as the optimal recovery sequence. When there are multiple candidate recovery sequences with the same comprehensive cost and all being the minimum, the sequence with the fewest switching operation steps is selected as the optimal recovery sequence. When the number of operation steps is also the same, the sequence with the highest load recovery rate after execution is selected as the optimal recovery sequence. This ensures that the optimal recovery sequence has the best comprehensive performance, the fastest recovery execution speed, and the best recovery effect.

[0054] In embodiments of the present invention, the isolation actions corresponding to the first set of disconnecting switches and the communication transfer actions included in the optimal recovery sequence are executed sequentially. After each action, the bidirectional drift intensity and fault core area score are updated online to obtain the drift residual. Based on the changes in the stable fault core area and the drift residual, it is determined whether to regenerate the optimal recovery sequence, including: After determining the initial set of disconnecting switches and solving for the optimal recovery sequence, the isolation actions corresponding to the initial set of disconnecting switches are executed to isolate the stable fault core area. The execution order of the isolation actions is to first open the disconnecting switches on the upstream side of the stable fault core area, and then open the disconnecting switches on the downstream side of the stable fault core area. This execution order is adopted to avoid equipment damage caused by reverse fault current during the isolation process, ensuring the safety of the isolation operation. Before each switch operation, the five-prevention interlocking conditions of the distribution automation system are verified. The operation command is only issued after confirming that the interlocking conditions are met and there is no risk of opening under load. After each disconnecting switch opening operation is completed, the remote signaling status of the corresponding switch is collected through the distribution automation terminal. After confirming that the switch is in a fully open state and the open position signal has remained unchanged for three consecutive power frequency cycles, the isolation action is considered to be completed. This ensures the reliable execution of the isolation action and avoids incomplete fault isolation caused by false opening of the switch. If the disconnector switch is opened and the remote signaling status does not show that the switch is open, or the electrical detection shows that there is still electrical connection between the stable fault zone and the healthy zone, the disconnection process should be stopped immediately and the opening operation should be retried once. If the retried operation is still unsuccessful, an alarm should be issued and the process should be switched to manual emergency handling. No subsequent recovery actions should be performed.

[0055] After all the first-round disconnector switches have completed their tripping actions, the voltage and current signals on both sides of the stable fault core area are collected through the distribution automation terminal. Once it is confirmed that there is no electrical connection between the stable fault core area and the upstream and downstream healthy sections, the isolation operation of the stable fault core area is considered complete. Subsequently, the initial online drift residual is calculated. The calculation process uses the time range from the fault initiation time to the isolation completion time as the time frame, and calculates the bidirectional drift intensity and fault core area score for each monitoring point according to the same rules as after subsequent actions. The initial online drift residual is then obtained using the online drift residual calculation formula. This serves as a baseline for residual comparison after the initial recovery action, ensuring consistency in the calculation rules before and after. Subsequently, the interconnection and transfer actions are executed sequentially according to the optimal recovery sequence. Each action element in the optimal recovery sequence corresponds to the closing operation of a sectionalizing switch or interconnecting switch. Only the closing operation of a single switch is executed at a time. After the previous closing operation is completed and the online drift verification is completed, the next closing operation is executed. This ensures the controllability of the recovery process and avoids the expansion of the fault range caused by the simultaneous execution of multiple operations.

[0056] Before each switch closing operation, the five-prevention interlocking conditions and power flow feasibility are verified to ensure there is no short-circuit risk or power overload risk before issuing the operation command. If the remote signaling status does not indicate closing after the switch closing operation is completed, the closing operation is immediately retried once. If the retried operation is still unsuccessful, the switch action is skipped, and the next legal action in the optimal recovery sequence is executed. At the same time, the legality judgment rules for subsequent actions are updated. After each recovery action is completed, five power frequency cycles are waited. If the fluctuation range of the three-phase voltage RMS value of all monitoring points does not exceed ±0.5% of the corresponding reference voltage and the fluctuation range of the branch current does not exceed ±1% of the corresponding rated current within three consecutive sampling cycles, the electrical transient process is determined to have completely subsided. Then, the equipment operating parameters of all monitoring points on the candidate recovery path are re-collected through the distribution automation terminal at a sampling interval of 2 milliseconds. The equipment operating parameters include fault indication status, three-phase voltage RMS value, branch current, and switch status. The collection rules are completely consistent with those in the aforementioned embodiment. If a monitoring point experiences an abnormal situation such as communication interruption or sampling value overflow, the monitoring point will be removed from the current calculation sequence and will not be included in subsequent calculations. At the same time, an abnormal alarm will be issued.

[0057] After data collection, the time period following the current recovery action is updated, using the completion time of this recovery action as the new time node. The time period before the recovery action remains a fixed interval from the fault initiation time to the closing time of the first communication switch. The time period after the recovery action is updated to the closing time of the first communication switch to the completion time of this recovery action. The reference voltage and reference current directions still use the fixed benchmarks calculated from the reference time window before the fault in the aforementioned embodiments, and do not change with the execution of the recovery action, thus ensuring the consistency of the characteristic quantity calculation benchmarks. Then, based on the updated dual time periods, the expansion, contraction, and continuous power loss of each monitoring point are recalculated, with the calculation rules remaining completely consistent with the aforementioned embodiments. Robust normalization is then performed based on the updated characteristic quantities, and the bidirectional drift intensity and fault core area score after the current action are updated accordingly. The calculation rules for the updated bidirectional drift intensity and fault core area score remain completely consistent with the aforementioned embodiments.

[0058] After updating the bidirectional drift intensity and fault core region score, an online drift residual is generated based on the updated bidirectional drift intensity and the updated fault core region score. The formula for calculating the online drift residual is as follows: ,in The online drift residual after the h-th recovery action is completed. This represents the updated bidirectional drift intensity corresponding to the monitoring point numbered r after the h-th recovery action. This is the updated fault core region score corresponding to the monitoring point numbered r after the h-th recovery action. The reason for using this formula to calculate the online drift residual is that the monitoring point with the greater bidirectional drift intensity and the lower the fault core region score contributes more to the drift residual. This can accurately quantify the residual risk of bidirectional drift at the fault boundary under the current network state. The larger the value, the higher the risk of further drift at the boundary and the worse the stability of the fault core region.

[0059] After completing the online drift residual calculation, based on the updated fault core area score, and according to the judgment threshold calculation rules and stable fault core area extraction rules in the aforementioned embodiments, the stable fault core areas in the current network state are re-identified. A state determination is performed. If the upstream start segment and downstream end segment of the re-identified stable fault core area are completely consistent with the corresponding segments of the original stable fault core area, and this state continues for 3 consecutive sampling periods, it is determined that the stable fault core area has not changed. Simultaneously, the online drift residual after this recovery action is smaller than the online drift residual after the previous recovery action, and the subsequent actions of the optimal recovery sequence continue to be executed. If the upstream start segment or downstream end segment of the re-identified stable fault core area differs from the corresponding segments of the original stable fault core area, and this difference continues for 3 consecutive sampling periods, it is determined that the stable fault core area has changed. In this case, the execution of the current optimal recovery sequence is immediately stopped, the actual open / closed state of all switches and the network energized state are locked, and the maximum number of recalculations for the optimal recovery sequence is set to 3. If the number of recalculations does not exceed the maximum number of recalculations, the newly identified stable fault core area is used as a constraint, and the network state after the current actions is taken is used as the initial node. The recovery action graph is reconstructed according to the rules in the aforementioned embodiments, the comprehensive cost of candidate recovery sequences is calculated, and the updated optimal recovery sequence is obtained again using the ant colony algorithm. During the recalculation process, the loads that have been restored remain energized and no tripping operation is performed to prevent the restored loads from losing power again. Subsequent interconnection and power transfer actions are then performed according to the updated optimal recovery sequence. If the number of recalculations exceeds the maximum number of recalculations, the automatic recovery process is immediately stopped, an alarm is issued, and the process is transferred to manual emergency response. Using this stoppage and recalculation rule, the system can respond promptly to changes in the fault boundary during the recovery process, avoiding secondary false isolation, repeated power transfer, or secondary tripping caused by recovery actions based on the original fault core area. At the same time, the upper limit on the number of recalculations prevents the process from falling into an infinite loop, ensuring the stability and reliability of the recovery process.

[0060] After all the interconnection and power transfer actions in the optimal recovery sequence are completed, the energized status and power data of all load nodes are collected again. The total load power of the restored load set and the total load power of the entire de-energized load set are calculated, and the recovery rate is determined based on the ratio of the two. The formula for calculating the recovery rate is as follows: ,in The load recovery rate of this large-scale power outage self-healing recovery The set of restored loads refers to the set of all normal load nodes within the unstable fault zone that are energized after all restoration actions have been completed. This refers to the set of all power-loss loads, specifically the set of all normal load nodes within the unstable fault core area that initially loses power after the fault occurs. Let be the rated total active power of the nth load node. The reason for using this formula to calculate the recovery rate is that it accurately quantifies the actual effect of this self-healing recovery by using the ratio of active power, eliminating the interference of fault loads in the fault core area on the recovery rate calculation. The closer the value is to 1, the better the recovery effect.

[0061] like Figure 2 The diagram shown is a functional block diagram of a system for rapid self-healing recovery of large-area power outages in a power distribution network based on multi-source collaboration, provided by an embodiment of the present invention.

[0062] In this embodiment, the functions of each module / unit are as follows: The time period segmentation module is used to collect equipment operating parameters at the monitoring points of the candidate recovery path, determine the reference status quantity, and divide the time period before and after the connection transfer with the closing time of the first connection switch as the boundary. The drift quantization module is used to extract the outward expansion, retraction, and continuous power loss that characterize the changes in the fault boundary based on equipment operating parameters, the period before and after the power transfer, and to generate the bidirectional drift intensity and fault core area score of the monitoring point based on the outward expansion, retraction, and continuous power loss. The core area identification module is used to set a judgment threshold based on the fault core area score, filter target monitoring points and map them to physical sections, and extract stable fault core areas and the first round of disconnect switches. The path optimization module is used to construct a recovery action graph with stable fault core regions as constraints, determine the comprehensive cost of candidate recovery sequences by combining bidirectional drift intensity, and search for the optimal recovery sequence in the recovery action graph using the ant colony algorithm. The execution review module is used to sequentially execute the isolation actions corresponding to the first set of disconnect switches and the communication transfer actions included in the optimal recovery sequence. After each action, it updates the bidirectional drift intensity and fault core area score online to obtain the drift residual. Based on the changes in the stable fault core area and the drift residual, it determines whether to regenerate the optimal recovery sequence.

[0063] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for rapid self-healing and recovery of large-area power outages in distribution networks based on multi-source collaboration, characterized in that, include: S1. Collect equipment operating parameters at monitoring points along the candidate recovery path, determine reference status quantities, and divide the period before and after the connection transfer into two phases, with the closing time of the first connection switch as the boundary. S2. Based on equipment operating parameters, the period before and after the power supply transfer, extract the outward expansion, retraction, and continuous power loss that characterize the changes in the fault boundary, and generate the bidirectional drift intensity and fault core area score of the monitoring point based on the outward expansion, retraction, and continuous power loss. S3. Set the judgment threshold based on the fault core area score, filter the target monitoring points and map them to the physical section, and extract the stable fault core area and the first round of disconnecting switches; S4. Construct a recovery action graph with stable faulty core regions as constraints, determine the comprehensive cost of candidate recovery sequences by combining bidirectional drift intensity, and search for the optimal recovery sequence in the recovery action graph using the ant colony algorithm. S5. Sequentially execute the isolation actions corresponding to the first set of disconnect switches and the communication transfer actions included in the optimal recovery sequence. After each action, update the bidirectional drift intensity and fault core area score online to obtain the drift residual. Based on the changes in the stable fault core area and the drift residual, determine whether to regenerate the optimal recovery sequence.

2. The method for rapid self-healing and recovery of large-area power outages in a distribution network based on multi-source collaboration as described in claim 1, characterized in that, Equipment operating parameters are collected at monitoring points along the candidate recovery path to determine reference state quantities. The timeframe before and after the first communication switch closure is used as the boundary, including: The fault indication status, three-phase voltage RMS value, branch current and switch status are collected at the monitoring points of the candidate recovery path as the equipment operating parameters; Select a reference time window before the fault, determine the reference voltage based on the effective value of the three-phase voltage within the reference time window before the fault, and determine the reference current direction based on the branch current within the reference time window before the fault. The voltage validity is determined based on the reference voltage, and the reverse power supply indication is determined based on the consistency between the branch current and the reference current direction. The fault initiation time, the first contact switch closing time, and the last state change time after contact transfer are obtained. The time from the fault initiation time to the first contact switch closing time is divided into the period before contact transfer, and the time from the first contact switch closing time to the last state change time after contact transfer is divided into the period after contact transfer.

3. The method for rapid self-healing and recovery of large-area power outages in a distribution network based on multi-source collaboration as described in claim 2, characterized in that, Extract the expansion, contraction, and sustained power loss that characterize the changes in the fault boundary, including: The degree of voltage loss is determined based on the voltage effectiveness. Based on the fault indication status and voltage loss level during the period before the interconnection transfer, the outward expansion amount is extracted to characterize the cumulative extent of the fault boundary expanding to the healthy side before the interconnection transfer. Based on the fault indication fading status, voltage validity, and reverse power supply indication during the period after the interconnection and power transfer, the shrinkage amount is extracted to characterize the cumulative degree of fault boundary shrinkage after the interconnection and power transfer. Based on the fault indication status and voltage loss level during the period before and after the power supply transfer, the continuous power loss is extracted to characterize the continuous voltage loss level of the monitoring point throughout the entire recovery event.

4. The method for rapid self-healing and recovery of large-area power outages in a distribution network based on multi-source collaboration as described in claim 3, characterized in that, Generate bidirectional drift intensity and fault core area scores for monitoring points, including: The expansion amount, the retraction amount, and the continuous power loss are respectively subjected to robust normalization processing to obtain normalized expansion amount, normalized retraction amount, and normalized continuous power loss amount; The bidirectional drift intensity is generated based on the normalized expansion and normalized contraction to characterize the degree of contraction after the fault boundary of the monitoring point expands. A fault core region score is generated based on normalized continuous power loss and bidirectional drift intensity to characterize the probability that the monitoring point belongs to a true stable fault core region.

5. The method for rapid self-healing and recovery of large-area power outages in a distribution network based on multi-source collaboration as described in claim 4, characterized in that, Based on the fault zone score, a judgment threshold is set, target monitoring points are screened and mapped to physical sections, and stable fault zones and the first set of disconnect switches are extracted, including: The judgment threshold is determined based on the median and absolute median difference of the fault core area scores of all monitoring points; In a topologically ordered sequence of monitoring points, monitoring points with a fault core area score not less than the judgment threshold are selected as target monitoring points. The target monitoring point is mapped to the physical segment between adjacent sectional switches, and the smallest continuous segment covering the physical segment is extracted from the mapped physical segment as the stable fault core area; The sectional switches surrounding the stable fault core area on both sides are identified as the first set of isolating switches.

6. The method for rapid self-healing and recovery of large-area power outages in a distribution network based on multi-source collaboration as described in claim 5, characterized in that, A recovery action graph is constructed using stable faulty core regions as constraints. The comprehensive cost of candidate recovery sequences is determined by combining bidirectional drift intensity, including: The recovery action graph is constructed using the feasible network state after isolating the stable faulty core region as nodes, wherein the edges between nodes represent legal switching operations; Switching operations that satisfy the conditions of not crossing the stable fault core region, meeting the power supply requirements of the current topology, and satisfying the power flow feasibility under the current state are determined to be legal switching operations. For any candidate recovery sequence, determine the total load power that has not been recovered after executing the candidate recovery sequence, determine the total bidirectional drift intensity on the recovery path corresponding to the candidate recovery sequence as the drift risk intensity, and determine the comprehensive constraint violation amount composed of current over-limit and voltage instability. Based on the statistical results of the total unrecovered load power, drift risk intensity, and comprehensive constraint violation amount corresponding to the initial candidate recovery sequence set, the total unrecovered load power, drift risk intensity, and comprehensive constraint violation amount of the current candidate recovery sequence are normalized, and the comprehensive cost of the candidate recovery sequence is generated based on the normalization result.

7. The method for rapid self-healing and recovery of large-area power outages in a distribution network based on multi-source collaboration as described in claim 6, characterized in that, The optimal recovery sequence is obtained by searching the recovery action graph using the ant colony algorithm, including: The ant colony algorithm is used to search in the recovered action graph, and the local comprehensive cost of a single action is determined as the basis for generating heuristic information; The transition probabilities between state nodes are determined based on heuristic information and pheromones to generate multiple candidate recovery sequences; The adaptive volatility rate is determined based on the comprehensive cost of the generated candidate recovery sequences, and the pheromone is updated in combination with the comprehensive cost of each candidate recovery sequence. When the termination condition is met, the candidate recovery sequence with the minimum overall cost is output as the optimal recovery sequence.

8. The method for rapid self-healing and recovery of large-area power outages in a distribution network based on multi-source collaboration as described in claim 7, characterized in that, After each action, the bidirectional drift intensity and fault core region score are updated online to obtain the drift residual. Based on the changes in the stable fault core region and the drift residual, it is determined whether to regenerate the optimal recovery sequence, including: Execute the isolation action corresponding to the first set of disconnect switches to complete the isolation of the stable fault core area; Execute the communication and transfer actions sequentially according to the optimal recovery sequence; After each recovery action is completed, the operating parameters of the device are re-acquired, and the bidirectional drift intensity and fault core area score after the current action are updated accordingly. Drift residuals are generated based on the updated bidirectional drift intensity and the updated fault core region score; If the re-identified stable faulty core region has not changed, and the drift residual is smaller than the drift residual after the previous recovery action, then the subsequent actions of the optimal recovery sequence will continue to be executed. If the re-identified stable faulty core region changes, the current recovery sequence is terminated, and the newly identified stable faulty core region is used as a constraint to search for the optimal recovery sequence again.

9. A method for rapid self-healing and recovery of large-area power outages in a distribution network based on multi-source collaboration, as described in claim 8, is characterized in that... After all the interconnection and transfer operations are completed, the total load power of the restored load set and the total load power of the entire power-loss load set are calculated, and the recovery rate is determined based on the ratio of the two.

10. A system for rapid self-healing recovery of large-area power outages in distribution networks based on multi-source collaboration, applied in the method for rapid self-healing recovery of large-area power outages in distribution networks based on multi-source collaboration as described in any one of claims 1-9, characterized in that, The system includes: The time period segmentation module is used to collect equipment operating parameters at the monitoring points of the candidate recovery path, determine the reference status quantity, and divide the time period before and after the connection transfer with the closing time of the first connection switch as the boundary. The drift quantization module is used to extract the outward expansion, retraction, and continuous power loss that characterize the changes in the fault boundary based on equipment operating parameters, the period before and after the power transfer, and to generate the bidirectional drift intensity and fault core area score of the monitoring point based on the outward expansion, retraction, and continuous power loss. The core area identification module is used to set a judgment threshold based on the fault core area score, filter target monitoring points and map them to physical sections, and extract stable fault core areas and the first round of disconnect switches. The path optimization module is used to construct a recovery action graph with stable fault core regions as constraints, determine the comprehensive cost of candidate recovery sequences by combining bidirectional drift intensity, and search for the optimal recovery sequence in the recovery action graph using the ant colony algorithm. The execution review module is used to sequentially execute the isolation actions corresponding to the first set of disconnect switches and the communication transfer actions included in the optimal recovery sequence. After each action, it updates the bidirectional drift intensity and fault core area score online to obtain the drift residual. Based on the changes in the stable fault core area and the drift residual, it determines whether to regenerate the optimal recovery sequence.

Images