Fault isolation and voltage recovery monitoring method for multi-terminal flexible DC power transmission system

Abstract

The invention relates to the technical field of operation and maintenance of an offshore wind power flexible direct current power transmission system, and discloses a multi-terminal flexible direct current power transmission system fault isolation and voltage recovery monitoring method, which comprises the following steps: acquiring degradation rate data to generate a degradation rate vector; identifying a degradation stage and screening degradation acceleration stage equipment; predicting a fault probability and screening high-risk equipment; sorting fault moments to generate a potential cascade fault time sequence; a cascade fault is rehearsed in the digital twinborn model, and an increment isolation scheme is solved; generating a pre-reconstruction topology time sequence scheme; performing a topology preparation action when the fault probability exceeds a preventive intervention threshold; when an actual fault occurs, the fault position is compared, and an isolation scheme is executed; and after isolation is completed, updating the prediction model and executing voltage recovery control. Linkage of degradation monitoring and topology reconstruction is established, and active prevention before a fault occurs is realized.

Description

Technical Field

[0001] This invention relates to the field of operation and maintenance technology for offshore wind power flexible DC transmission systems, and more specifically, to a method for fault isolation and voltage recovery monitoring of multi-terminal flexible DC transmission systems. Background Technology

[0002] In the field of operation and maintenance of offshore wind power flexible DC transmission systems, key equipment such as IGBT modules and DC capacitors may enter a deterioration acceleration stage after long-term operation. During the deterioration acceleration stage, the rate of change of equipment health status is significantly accelerated, and multiple devices may fail one after another in a short period of time, forming a cascading failure sequence.

[0003] Existing technologies employ an iterative topology reconstruction method, which uses a passive response mode of fault detection, solution, and execution to handle faults.

[0004] The existing technology has the following problems: equipment degradation monitoring and topology reconfiguration decision-making operate independently and do not form an effective linkage, which makes it impossible for the system to make proactive preparations using degradation prediction information; when waiting for the actual fault to occur before starting iterative reconfiguration, the electrical impact generated by the first fault may have already spread along the topology connection to adjacent areas, causing the fault impact range to expand, missing the best isolation opportunity, and affecting the safety and reliability of system operation. Summary of the Invention

[0005] This invention provides a method for fault isolation and voltage recovery monitoring of a multi-terminal flexible DC transmission system, addressing the technical problems in related technologies such as the lack of predictive topology adjustment capability, difficulty in formulating isolation plans in advance, and delayed fault response leading to increased losses when cascading faults occur in offshore wind power flexible DC transmission systems due to equipment deterioration.

[0006] This invention provides a method for fault isolation and voltage recovery monitoring in a multi-terminal flexible DC transmission system, comprising the following steps: Collect degradation rate index data of key equipment in offshore wind power flexible DC system and generate degradation rate vector of each equipment; The degradation rate vector is matched with the degradation stage feature template library to identify the degradation stage identifier of each device, and devices in the degradation acceleration stage are selected to generate a set of devices in the degradation acceleration stage. For each device in the set of devices in the accelerated degradation stage, an autoregressive model is used to predict the degradation trajectory, and a short-term failure probability time series distribution is calculated in combination with the failure threshold. A subset of high-risk devices is generated based on the early warning threshold. Based on the peak times of the predicted failure probability of each device in the high-risk device subset, a potential cascading failure time sequence is generated. In the digital twin model, faults are injected sequentially according to the potential cascading fault time sequence, the incremental isolation topology scheme is solved, and the pre-reconstruction topology time sequence scheme is generated. When the failure probability of the first high-risk device exceeds the preventive intervention threshold, the initial topology preparation action in the pre-reconstruction topology timing scheme is executed; If a fault actually occurs, the actual fault location will be compared with the potential cascading fault time sequence, and an isolation scheme will be executed according to the degree of matching. After isolation is completed, the parameters of the autoregressive model are updated, and voltage recovery control is executed.

[0007] Furthermore, the degradation rate index data includes temperature change rate, insulation resistance decrease rate, and partial discharge pulse frequency; the degradation rate vector is composed of the temperature change rate, insulation resistance decrease rate, and partial discharge pulse frequency; and each component of the degradation rate vector is normalized using Z-score.

[0008] Furthermore, matching the degradation rate vector with the degradation stage feature template library includes: The similarity between the degradation rate vector and the feature templates of each degradation stage is calculated using a dynamic time warping algorithm. The degradation stage with the highest similarity is used as the current degradation stage identifier.

[0009] Furthermore, the generation method of the degradation stage feature template library includes: Collect time series vectors of degradation rates of historically failed equipment throughout its entire lifespan; The degradation rate vector time series is divided into normal operation stage, early degradation stage, accelerated degradation stage and near-failure stage according to the time window before equipment failure. Statistical analysis was performed on the degradation rate vectors in each stage, and the mean, variance, and trend of each component were extracted as the degradation stage feature templates for that stage.

[0010] Furthermore, the calculation method for the short-term failure probability time-series distribution includes: The difference between the predicted degradation trajectory value and the failure threshold is denoted as the degradation margin. After normalizing the degradation margin, the failure rate function value is obtained by substituting it into the exponential decay function. The cumulative failure probability is obtained by integrating the failure rate function value over time and performing an exponential transformation.

[0011] Furthermore, the input of the autoregressive model is the recent degradation rate time series of the equipment, and the output is the degradation rate prediction value, the upper bound prediction value, and the lower bound prediction value of the degradation trajectory at each time in the future prediction time domain; the conservative failure probability is calculated based on the upper bound prediction value as the screening criterion; the degradation state prediction value is obtained by cumulatively integrating the degradation rate prediction value in the time dimension.

[0012] Furthermore, solving for the incremental isolation topology scheme includes: Based on the system topology connection matrix, a breadth-first traversal is performed starting from the fault point, and all nodes and connection lines with a traversal depth within a preset number of layers are included in the influence domain. A mixed-integer optimization model is used to solve for the break point selection scheme that minimizes isolation loss. The objective function of the mixed-integer optimization model includes a break point operation cost term and a load loss weighting term. The incremental isolation topology scheme is merged with the existing isolation scheme, and the isolation breakpoint set in the subsequent stage is a superset of the isolation breakpoint set in the previous stage.

[0013] Furthermore, the initial topology preparation action adopts a hierarchical strategy: When the failure probability is between the preventive intervention threshold and the medium risk threshold, perform a warm-up operation on the backup transmission path. When the probability of failure exceeds the medium risk threshold, a power margin pre-adjustment operation is performed simultaneously. The preheating operation includes sending a closing command to the relevant circuit breaker of the standby line and applying a test voltage lower than the rated value to the standby line.

[0014] Furthermore, implementing isolation protocols based on the degree of matching includes: The device matching degree and the time matching degree are calculated. The device matching degree is obtained by comparing the consistency between the actual fault device identifier and the predicted fault device identifier in the potential cascading fault time sequence. The time matching degree is obtained by normalizing the time difference between the actual fault time and the predicted fault time. The weighted average of the device matching degree and the time matching degree is used as the matching degree; When the matching degree is higher than the matching threshold, the isolation instruction of the corresponding stage in the pre-reconstructed topology timing scheme is directly invoked; When the matching degree is lower than the matching threshold, the process is executed after incremental correction based on the pre-reconstructed topology timing scheme.

[0015] Furthermore, the autoregressive model parameter update includes: using the degradation rate time series before the actual failure as the new training sample, using the actual failure time as the label value, adding the new training sample to the historical training set using a sliding window method, re-estimating the coefficient parameters of the autoregressive model using the least squares method, and using the updated autoregressive model for subsequent equipment failure prediction.

[0016] The beneficial effects of this invention are as follows: This invention identifies devices in the accelerated degradation stage by matching degradation rate vectors with a degradation stage feature template library, and uses an autoregressive model to predict the short-term failure probability time series distribution of each device, thereby generating a potential cascading failure time series. It establishes a linkage relationship between device degradation monitoring and cascading failure prediction, enabling the system to know possible cascading evolution paths before a failure occurs. By injecting faults sequentially into the digital twin model according to the potential cascading fault time sequence and solving the incremental isolation topology scheme, a pre-reconstructed topology time sequence scheme covering the entire cascading evolution process is generated in advance, transforming the passive response mode into an active prevention mode. By pre-opening backup transmission paths and pre-adjusting power margins when the probability of failure of the first high-risk device exceeds the preventive intervention threshold, the system has completed the initial topology preparation actions when the fault actually occurs. It can directly call the isolation instructions in the pre-reconfiguration topology timing scheme to perform rapid isolation, shortening the response time from the occurrence of the fault to the completion of isolation and reducing the risk of the electrical impact of the first fault spreading to adjacent areas along the topology connection. Attached Figure Description

[0017] Figure 1 This is a flowchart of a fault isolation and voltage recovery monitoring method for a multi-terminal flexible DC transmission system according to the present invention; Figure 2 This is a comparison chart of the original degradation rate indicators of the equipment in an example of the present invention; Figure 3 This is a vector heatmap of the standardized degradation rate of equipment in an example of the present invention; Figure 4 This is a similarity distribution diagram for identifying the equipment degradation stage in an example of the present invention; Figure 5 This is a high-risk equipment degradation trajectory prediction diagram in an example of the present invention; Figure 6 This is a graph showing the evolution trend of cumulative equipment failure probability in an example of the present invention; Figure 7 This is a power transfer configuration diagram of the pre-reconfiguration topology timing scheme in an example of the present invention; Figure 8 This is a voltage recovery timing curve of the converter station in an example of the present invention. Detailed Implementation

[0018] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, some features described in the examples may be combined in other examples.

[0019] At least one embodiment of the present invention discloses a method for fault isolation and voltage recovery monitoring of a multi-terminal flexible DC transmission system, such as... Figure 1 As shown, it includes the following steps: S100: Collect degradation rate data and generate degradation rate vectors: Collect degradation rate index data of key equipment in the offshore wind power flexible DC system in real time, perform vectorization processing on the collected degradation rate index data, and generate degradation rate vectors for each equipment.

[0020] It should be noted that the above degradation rate indicators include temperature change rate, insulation resistance decrease rate, and partial discharge pulse frequency. The temperature change rate reflects the rate of thermal stress accumulation in the equipment, the insulation resistance decrease rate reflects the aging process of the insulation material, and the partial discharge pulse frequency reflects the degree of dielectric breakdown risk. The key equipment mentioned above includes IGBT modules and DC capacitors. The expression for the above degradation rate vector is: ; in, Indicates the first Degradation rate vector of each device Indicates the first Temperature change rate of each device Indicates the first The rate of decrease in insulation resistance of each device Indicates the first The partial discharge pulse frequency of the device This indicates the device index.

[0021] Furthermore, since the temperature change rate, insulation resistance decrease rate, and partial discharge pulse frequency have different dimensions and numerical ranges, Z-score normalization is applied to each component of the degradation rate vector to eliminate the influence of dimensional differences on subsequent matching calculations.

[0022] S200, Identify the degradation stage and filter devices in the accelerated degradation stage: Match the degradation rate vector with the pre-stored degradation stage feature template library, use the dynamic time warping algorithm to calculate the similarity between the degradation rate vector and each degradation stage feature template, identify the current degradation stage identifier of each device, filter devices in the accelerated degradation stage based on the degradation stage identifier, and generate a set of devices in the accelerated degradation stage.

[0023] It should be noted that the above-mentioned degradation stage feature template library contains degradation stage feature templates for multiple degradation stages. Each degradation stage includes the normal operation stage, the initial degradation stage, the accelerated degradation stage, and the near-failure stage. The degradation stage feature template corresponding to each degradation stage is the typical degradation rate vector distribution feature of that degradation stage.

[0024] Furthermore, the above-mentioned degradation stage feature template is generated as follows: collect the degradation rate vector time series of historically failed equipment throughout its entire life cycle, divide the degradation rate vector time series into four periods according to the time window before equipment failure: normal operation stage, initial degradation stage, accelerated degradation stage, and near-failure stage, perform statistical analysis on the degradation rate vector in each period, and extract the mean, variance, and trend of each component as the degradation stage feature template for that degradation stage.

[0025] Furthermore, the input of the aforementioned dynamic time warping algorithm is the current degradation rate vector sequence and the degradation stage feature template, and the output is the similarity value between the current degradation rate vector sequence and the degradation stage feature template. The dynamic time warping algorithm can handle the time scale difference between the degradation rate vector sequence and the degradation stage feature template, adapt to the speed difference of degradation process of different devices, and take the degradation stage with the highest similarity as the current degradation stage identifier.

[0026] S300, predicting failure probability and screening high-risk equipment: For each piece of equipment in the set of equipment in the accelerated degradation stage, extract the recent degradation rate time series of each piece of equipment, use an autoregressive model to predict the degradation rate time series, generate the degradation trajectory prediction results of each piece of equipment in the short term, combine the failure threshold of each piece of equipment to calculate the failure probability of the degradation trajectory prediction results, generate the short-term failure probability time series distribution of each piece of equipment, and filter the short-term failure probability time series distribution based on the warning threshold to generate a subset of high-risk equipment.

[0027] It should be noted that the failure thresholds mentioned above are the critical degradation values ​​at which the equipment experiences functional failure. The short-term failure probability time-series distribution described above describes the probability of failure for each device at each predicted time in the future. The formula for calculating the failure probability is as follows: ; in, Indicates the first Each device at time The cumulative failure probability, Indicates the first Each device at time The failure rate function, This indicates the predicted time length from the current moment. Represents the integral variable. Represents an exponential function. This indicates the device index.

[0028] Furthermore, the above failure rate function The calculation method is as follows: First, calculate the difference between the predicted degradation trajectory value and the failure threshold, denoted as the degradation margin. Then, normalize the degradation margin to the interval between zero and one. Finally, substitute the normalized degradation margin into the exponential decay function to obtain the failure rate function value. The specific expression of the failure rate function is as follows: ; in, Indicates the baseline failure rate. Indicates the degradation acceleration factor. Indicates the failure threshold. Indicates the first Each device at time The predicted value of the deterioration state, Represents an exponential function. This indicates the device index. The failure rate function value increases when the predicted degradation state value approaches the failure threshold.

[0029] Furthermore, the aforementioned warning threshold is a failure probability threshold used to screen high-risk equipment. When the failure probability of an equipment exceeds the warning threshold at any time in the prediction time domain, the equipment is included in the high-risk equipment subset.

[0030] Furthermore, the input to the aforementioned autoregressive model is the recent degradation rate time series of the equipment, and the output is the predicted degradation rate, upper bound prediction, and lower bound prediction of the degradation trajectory at each time point within the future prediction time domain. Historical degradation rate data is used during training. The output layer of the autoregressive model is a fully connected layer, using mean squared error as the loss function and least squares method for parameter estimation. A conservative failure probability is calculated based on the upper bound prediction, and an optimistic failure probability is calculated based on the lower bound prediction. The conservative failure probability is used as the selection criterion.

[0031] Furthermore, the above-mentioned predicted degradation status values The calculation method is as follows: the degradation rate prediction value output by the autoregressive model is cumulatively integrated over time to obtain the cumulative change in degradation state relative to the current time. This cumulative change is then added to the current degradation state measurement value to obtain the degradation state prediction value for future time. The specific expression for the degradation state prediction value is as follows: ; in, Indicates the first The measured value of the degradation status of each device at the current moment. Indicates the first Each device at time The predicted rate of degradation, For time integration variables, This indicates the device index.

[0032] S400, Sort the failure times and generate a potential cascading failure time series: Sort the predicted failure probability peak times of each device in the high-risk device subset to generate a potential cascading failure time series. The potential cascading failure time series includes the order in which each high-risk device may fail and the time interval between failures of adjacent devices.

[0033] It should be noted that the predicted peak time of failure probability mentioned above refers to the moment when the failure probability in the short-term failure probability time series distribution of each device reaches its maximum value, or the moment when the failure probability first exceeds the critical threshold. The expression for the above potential cascading failure time series is: ; in, This represents a potential cascading failure time sequence. This indicates the device identifier that is the first potential failure in a potential cascading failure time sequence. This indicates the device identifier that represents the second potential failure in a potential cascading failure time sequence. In the potential cascading failure time sequence, the first... A possible faulty device identifier. This indicates the predicted failure time of the first potentially faulty device. This indicates the predicted failure time of the second potentially faulty device. Indicates the first Predict the failure time of a possible faulty device, and satisfy the following conditions: , This represents the total number of devices in the high-risk device subset.

[0034] Furthermore, the aforementioned critical threshold is the fault probability boundary value for determining whether the equipment enters a high-risk state. The critical threshold is higher than the warning threshold and is used to identify the moment when the equipment is most likely to fail in the short-term fault probability time series distribution.

[0035] S500, simulates cascading faults and solves incremental isolation schemes: For potential cascading fault time sequences, faults are injected sequentially in the digital twin model according to the potential cascading fault time sequence. The influence domain of each fault point is calculated based on the system topology connection matrix. For each newly added fault point, the incremental isolation topology scheme is solved using a mixed integer optimization model. The incremental isolation topology scheme is then merged with the existing isolation scheme. The isolation breakpoint set and power transfer path of each stage are recorded.

[0036] It should be noted that the aforementioned digital twin model is a digital mapping model of the offshore wind power flexible DC system. The digital twin model includes the system's electrical topology, equipment parameters, and operating status. The aforementioned system topology connection matrix describes the electrical connection relationships between nodes in the system. The aforementioned influence domain refers to the range of equipment and lines that may be affected by a fault point through electrical connections. The influence domain is specifically calculated as follows: starting from the fault point, a breadth-first traversal is performed along the connection relationships of the system topology connection matrix, including all nodes and connecting lines within a preset traversal depth within the influence domain. The aforementioned topology merging process refers to integrating newly added incremental isolation topology schemes with existing isolation schemes at the topology level to ensure the compatibility of incremental isolation topology schemes at each stage.

[0037] Furthermore, the aforementioned preset number of layers is a traversal depth limit for the influence domain. The preset number of layers is determined based on the system topology and fault propagation characteristics, with typical values ​​ranging from two to four layers, ensuring that the influence domain covers the directly adjacent and secondary adjacent areas of the fault point.

[0038] Furthermore, the input to the aforementioned mixed-integer optimization model includes the current system topology state, the location of the fault point, and the influence domain range. The output is the breakpoint selection scheme that minimizes the isolation loss. The objective function of the mixed-integer optimization model is: ; in, Represents the set of candidate isolation breaks. Indicates the break The operational cost, Representing the break point as a binary decision variable Whether or not to be selected, This represents a set of areas that may experience load loss. Indicates the area Load weight, Representing a region for continuous variables Loss of load, Represents the set of candidate isolation breaks The breakpoint index in the middle, This represents the set of regions that may experience load loss. The region index is solved using the branch and bound method for optimization.

[0039] Furthermore, due to operational costs and load weight Since they have different dimensions and numerical ranges, in order to ensure the comparability of the two items in the objective function, the operation cost and load weight are normalized by mean normalization based on the range, so that the numerical ranges of the two items are unified to the interval between zero and one.

[0040] Furthermore, the above mixed-integer optimization model includes the following constraints: the fault point must be completely isolated from the healthy area, that is, at least one break point must be selected to disconnect the fault point from the rest of the system; the on / off state of each break point must meet the topological connectivity requirements to ensure that the load in the non-faulty area can still be powered through at least one path; the power transfer of each line must not exceed the rated capacity of the line; the total amount of load loss must not exceed the maximum allowable load loss limit of the system.

[0041] Furthermore, the above topology merging process follows the following constraints: the set of isolation breaks in subsequent stages is a superset of the set of isolation breaks in previous stages, that is, the isolation actions that have been performed cannot be undone; the adjustment of power transfer paths adopts an incremental approach, adding or adjusting some power transfer paths while maintaining the established power transfer paths.

[0042] Furthermore, the costs of the above operations The method of determination is: based on the fracture surface The type, current power, and operating frequency of the circuit breaker are comprehensively evaluated. Higher operating value is assigned to high-voltage circuit breaker types, circuit breakers with higher current power carrying capacity, and circuit breakers with higher recent operating frequencies. The weighted sum of these three evaluation results yields the circuit breaker value. The operational costs.

[0043] S600 integrates the pre-simulation results and generates a pre-reconstruction topology timing scheme: Based on the results of the digital twin pre-simulation, a pre-reconstruction topology timing scheme is generated. The pre-reconstruction topology timing scheme includes the isolation break sequence, backup transmission path configuration and voltage reference value adjustment curve corresponding to each predicted time.

[0044] It should be noted that the above isolation break sequence is a list of break identifiers that need to be activated at each stage, arranged in chronological order. The above backup transmission path configuration refers to the backup lines used for power transfer at each stage and their transmission capacity allocation. The above voltage reference value adjustment curve is the time-series change trajectory of the DC voltage setpoint of the converter station at each stage.

[0045] Furthermore, the output of the aforementioned pre-reconfigured topology timing scheme is a structured control instruction set, which includes disconnection action instructions, line switching instructions, and voltage setpoint instructions. Each instruction contains the execution time, target device identifier, and operating parameters. The control instruction set can be directly sent to the field control system for execution.

[0046] S700, determine intervention conditions and execute preventive topology preparation actions: when the failure probability of the first high-risk device exceeds the preventive intervention threshold, execute the initial topology preparation actions in the pre-reconfiguration topology timing scheme in advance. The initial topology preparation actions include pre-opening backup transmission paths and pre-adjusting the power margin of adjacent converter stations.

[0047] It should be noted that the aforementioned preventative intervention threshold is a fault probability threshold that triggers preventative actions. This threshold is lower than the actual fault determination threshold and is used to allow preparation time before a fault actually occurs. The aforementioned pre-opening of backup transmission paths refers to placing the relevant switches of the backup transmission paths in a standby state or a low-power operating state. The aforementioned pre-adjustment of the power margin of adjacent converter stations refers to reducing the current power output of adjacent converter stations to reserve capacity for power transfer after a fault.

[0048] Furthermore, the execution of the above-mentioned initial topology preparation actions adopts a hierarchical strategy: when the failure probability is between the preventive intervention threshold and the medium risk threshold, only the warm-up operation of the backup transmission path is performed; when the failure probability exceeds the medium risk threshold, the power margin pre-adjustment operation is performed simultaneously.

[0049] Furthermore, the aforementioned medium-risk threshold is the intermediate threshold in the tiered strategy. The medium-risk threshold lies between the preventive intervention threshold and the actual fault determination threshold, and is used to distinguish the intensity of preparatory actions corresponding to different risk levels.

[0050] Furthermore, the preheating operation of the aforementioned backup transmission path specifically involves: sending a closing command to the relevant circuit breaker of the backup line to put the circuit breaker into a standby state; simultaneously applying a test voltage lower than the rated value to the backup line to check the insulation status and connection integrity of the line, ensuring that the backup line can respond quickly when it receives the formal commissioning instruction.

[0051] S800, compare fault location and execute isolation scheme: If a fault actually occurs, compare the actual fault location with the potential cascading fault timing sequence, calculate the matching degree, and when the matching degree is higher than the matching threshold, directly call the isolation instruction of the corresponding stage in the pre-reconstructed topology timing scheme to execute fast isolation; when the matching degree is lower than the matching threshold, perform incremental correction processing based on the pre-reconstructed topology timing scheme, and merge the corrected isolation scheme with the executed isolation scheme in the topology before execution.

[0052] It should be noted that the above-mentioned matching degree calculation includes two dimensions: device matching degree and time matching degree. The device matching degree is calculated by comparing the consistency between the actual faulty device identifier and the predicted faulty device identifier in the potential cascading fault time series. The device matching degree is one when the actual faulty device is in the predicted sequence, and zero otherwise. The time matching degree is obtained by calculating the time difference between the actual fault time and the predicted fault time, and normalizing the time difference to the interval between zero and one. The smaller the time difference, the higher the time matching degree. The final matching degree is the weighted average of the device matching degree and the time matching degree. The above-mentioned incremental correction processing refers to using a mixed integer optimization model to solve for the supplementary isolation gap and adjust the power transfer path based on the pre-reconstructed topology time series scheme and the deviation between the actual fault location and the predicted location.

[0053] Furthermore, the aforementioned matching threshold is a threshold value for determining the degree of matching of a pre-reconstructed topology time series scheme that can be directly used. When the degree of matching is higher than the matching threshold, it is considered that the actual fault and the predicted sequence are highly consistent, and the pre-reconstructed topology time series scheme can be directly called. The typical value is 0.7 to 0.9.

[0054] S900, update the prediction model and execute voltage recovery control: After isolation is completed, update the parameters of the autoregressive model based on the actual fault data, solve the voltage recovery timing optimization model based on the final stable topology, generate the voltage recovery curve of each converter station and send it out for execution, and at the same time calculate the hit rate of the pre-reconfiguration topology timing scheme for subsequent strategy optimization.

[0055] It should be noted that the above hit rate is a statistical value of the matching degree between the actual fault sequence and the potential cascading fault time sequence, reflecting the accuracy of prediction and the effectiveness of the pre-reconstruction topology time sequence scheme.

[0056] Furthermore, the specific method for updating the parameters of the aforementioned autoregressive model is as follows: the degradation rate time series before the actual failure occurs is used as the new training sample, the actual failure time is used as the label value, the new training sample is added to the historical training set using a sliding window method, the coefficient parameters of the autoregressive model are re-estimated using the least squares method, and the updated autoregressive model is used for subsequent equipment failure prediction.

[0057] Furthermore, the input of the voltage recovery timing optimization model is the isolated system topology state and the current voltage value of each station, and the output is the timing trajectory of the voltage of each station recovering from the current value to the rated value. The timing trajectory satisfies the voltage change rate constraint and the inter-station voltage difference constraint. The output layer of the voltage recovery timing optimization model is a fully connected layer, and the sequential quadratic programming algorithm is used for optimization during the solution process.

[0058] Furthermore, the aforementioned voltage change rate constraint is an upper limit on the voltage adjustment speed of the converter station, used to avoid the impact of rapid voltage changes on system stability. The aforementioned inter-station voltage difference constraint limits the allowable range of voltage difference between adjacent converter stations, used to prevent excessive voltage differences from causing circulating currents or equipment overload.

[0059] Furthermore, the output form of the voltage recovery curve is a sequence of voltage setpoints for each converter station at each time. The voltage setpoint sequence can be directly converted into a reference input signal for the converter station voltage controller and sent to each converter station for execution through the field communication network.

[0060] Based on the steps in the above-mentioned fault isolation and voltage recovery monitoring method for a multi-terminal flexible DC transmission system, this method is applied to the following example, with the following background: A certain offshore wind power flexible DC transmission system is located in a wind farm cluster in the eastern sea area. The system includes three offshore converter stations ( , , ) and an onshore converter station ( The system, connected via ±500kV DC transmission lines, has a total installed capacity of 1200MW. Having been in operation for five years, it has entered a critical period for equipment degradation monitoring. On March 15, 20XX, the system's operation and maintenance center's equipment health management platform detected abnormal fluctuations in the degradation rate indicators of several key pieces of equipment, necessitating the initiation of a predictive topology pre-reconfiguration process.

[0061] The current system operating power is 950MW, offshore converter station Configured IGBT module and DC capacitor The degradation rate index has been rising recently, and the converter station IGBT module It also shows an accelerating trend of deterioration, and there is a risk of cascading failures in the short term.

[0062] like Figures 2-8 As shown, an example of the implementation process is as follows: An implementation example of S100 is as follows: The system monitored the converter station via a distributed sensor network from 08:00:00 to 08:30:00 on March 15, 20XX. , and Data collection on degradation rate indicators for key equipment, specifically IGBT modules. The temperature monitoring sensor records the junction temperature change rate every 5 minutes, the insulation resistance tester measures the gate-emitter insulation resistance decrease rate every 10 minutes, and the partial discharge detection system statistically analyzes the discharge pulse frequency within a 30-minute window in real time. After data acquisition, the raw data is vectorized and Z-score normalized.

[0063] Table 1. Input data for S100 (original degradation rate index) in, Indicates the original rate of temperature change. Indicates the rate of decrease in original insulation resistance. This indicates the original partial discharge pulse frequency.

[0064] Table 2. Output data of S100 (standardized degradation rate vector) in, Indicates the first Temperature change rate of each device after standardization Indicates the first The rate of decrease in insulation resistance after standardization of each piece of equipment Indicates the first The standardized partial discharge pulse frequency of each device Indicates the first Degradation rate vector of each device This indicates the device index. With equipment Taking the rate of temperature change as an example, the original value ℃ / h, global average ℃ / h, global standard deviation ℃ / h, standardized calculation is as follows: ; The implementation example of S200 is as follows: The degradation rate vector generated in step 1 is matched with a pre-stored degradation stage feature template library. The template library contains feature templates for four degradation stages, each template representing a typical degradation rate vector distribution feature for that stage. The dynamic time warping algorithm is used to calculate the similarity between the degradation rate vector sequence of each device over the past 7 days and each template. The similarity value ranges from 0 to 1, with a higher value indicating a higher degree of matching.

[0065] Table 3 Input data for S200 (deterioration stage feature template library) Table 4. Output data of S200 (deterioration stage identification results) equipment The highest similarity with the feature template of the accelerated degradation stage was 0.88, exceeding the identification threshold of 0.70, and therefore it was identified as belonging to the accelerated degradation stage. The set of devices in the accelerated degradation stage is as follows: .

[0066] The implementation example of S300 is as follows: For three devices in the accelerated degradation phase, the degradation rate time series over the past 168 hours (7 days) was extracted, and an autoregressive model was used to predict the degradation trajectory for the next 72 hours. The model output includes predicted degradation rate and predicted degradation state. Combined with a device failure threshold, the cumulative failure probability of each device at any future time was calculated. A warning threshold of 0.35 was set, and devices whose failure probability exceeds this threshold at any time within a short period were selected.

[0067] Table 5 Intermediate data for S300 (deterioration trajectory prediction results) in, Indicates the first The measured value of the degradation status of each device at the current moment. Indicates the failure threshold. Indicates the predicted time (number of hours from the current time). Indicates the first Each device at time The predicted rate of degradation, Indicates the first Each device at time The predicted value of the deterioration state, Indicates the first Each device at time The failure rate function value, Indicates the first Each device at time The cumulative failure probability, This indicates the device index.

[0068] With equipment Taking the calculation at a prediction time of 48 hours as an example, the baseline failure rate Degradation acceleration coefficient The predicted value of the deterioration state is calculated by integration: ; The failure rate function value is: ; The cumulative failure probability is: ; Table 6. S300 Output Data (High-Risk Equipment Screening Results) The high-risk equipment subset is ,equipment Although it is in the accelerated deterioration stage, the short-term failure probability has not reached the warning threshold, so it is not included for the time being.

[0069] The implementation example of S400 is as follows: For two devices in the high-risk equipment subset, extract their predicted failure probability peak times and sort them chronologically to generate a potential cascading failure time series. and The peak time is 48 hours after the current time, that is, 08:30:00 on March 17, 20XX, but The peak probability is higher (0.43 > 0.38), making its failure risk more urgent and ranking it higher.

[0070] Table 7. S400 Output Data (Potential Cascade Fault Timing Sequence) in, This represents the position index in the time sequence of potential cascading failures. Indicates the first A possible faulty device identifier. Indicates the first Predicted failure time for potentially faulty equipment.

[0071] The potential cascading failure time sequence is expressed as follows: ; in At 08:30:00 on March 17, XX year, both devices are predicted to enter a high-risk state at the same time with a time interval of 0, which belongs to a near-synchronous cascading risk scenario.

[0072] The implementation examples of S500 and S600 are as follows: In the digital twin model, faults are injected sequentially according to the potential cascading fault time sequence. First, the faults are injected into the equipment. Fault, calculate its impact area covering the converter station busbar segment and connecting lines The incremental isolation scheme is solved using a mixed-integer optimization model, and the break point is selected. and Isolation is implemented, and power is transmitted through a backup line. Transfer. Then inject into the device. The fault affected a section of the busbar. Solve for the newly added fracture. Power is further directed to the converter station The route Transfer. Integrate the results of the two stages to generate a pre-reconstructed topology timing scheme, including the isolation break sequence, backup transmission path configuration, and voltage reference value adjustment curve, forming a structured control instruction set to be executed.

[0073] Table 8 Input data for step 5 (first-stage fault injection information) Table 9 Intermediate data for step 5 (parameters of the mixed-integer optimization model in the first stage) in, Represents the set of candidate isolation breaks The breakpoint index in the middle, Indicates the break The operation cost (after normalization), Representing the break point as a binary decision variable Whether it has been selected (1 indicates selection, 0 indicates no selection), This represents the set of regions that may experience load loss. The region index in Indicates the area The load weight (after normalization), Representing a region for continuous variables The amount of load loss.

[0074] In solving the first stage of the mixed-integer optimization model, the objective function is: ; Solution results selection of fracture surface and (Right now Loss of load , The objective function value is 0.60.

[0075] Table 10 Output data of steps 5 and 6 (pre-reconstruction topology timing scheme) The following is an example of the implementation of S700: System continuous monitoring equipment The failure probability was detected to have risen to 0.38 at 14:00:00 on March 16, 20XX, exceeding the preventive intervention threshold of 0.36. The system automatically triggered initial topology preparation. Since 0.38 falls between the preventive intervention threshold of 0.36 and the medium-risk threshold of 0.42, the system only performed a backup transmission path warm-up operation: [the system then] sent [the data] to the backup line. Related circuit breakers and Send a closing pre-command to put the circuit breaker into a standby state, and simultaneously... Apply a test voltage of 70% of the rated value (350kV) to check the insulation condition and connection integrity of the line, and confirm that the standby line can be put into operation quickly.

[0076] Table 11 S700 Execution Record (Initial Topology Preparation Actions) Implementation Examples of S800 March 17, 20XX, 09:15:00, equipment An actual fault occurred. The system immediately compared the actual fault location with the timing sequence of potential cascading faults: the actual faulty device was... The first device of the predicted sequence is The equipment matching degree is 1; the actual fault time is 09:15:00, the predicted time is 08:30:00, the time difference is 45 minutes, and the time matching degree is calculated to be 0.85 (based on time difference normalization); the overall matching degree is... If the value is higher than the matching threshold of 0.80, the system directly invokes the isolation instruction of the first stage of the pre-reconstruction topology timing scheme to send a signal to the breakpoint. and Issue a trip command and simultaneously send a message to the backup line. The commissioning order was issued, and rapid isolation was completed.

[0077] Table 12 S800 Execution Record (Fault Isolation Execution) The following is an example of the implementation of S900: After isolation is completed, the system updates the autoregressive model parameters based on actual fault data. The device... The degradation rate time series of the 168 hours prior to the failure was used as new training samples, and the actual failure time was used as the label value. These samples were added to the historical training set using a sliding window approach, and the autoregressive model coefficients were re-estimated using the least squares method. Simultaneously, based on the finally stable topology (which has been isolated)... backup line (During commissioning) Solve the voltage recovery timing optimization model to generate voltage recovery curves for each converter station: Converter station The voltage linearly recovered from the current value of 470kV to 495kV within 30 minutes at the converter station. The voltage linearly recovers from 515kV to 500kV within 30 minutes, with a voltage change rate constraint of 0.8kV / min. The hit rate of the pre-reconstruction topology timing scheme is calculated: the actual faulty device perfectly matches the first device in the predicted sequence, resulting in a 100% hit rate.

[0078] Table 13 S900 Output Data (Voltage Recovery Curve and Hit Rate Statistics) It should be noted that the data flow is explained as follows: The data flow throughout the entire implementation process reflects a complete closed loop from equipment degradation monitoring to preventative topology reconfiguration: First, the original degradation rate index data (Table 1) were processed by Z-score standardization to generate a degradation rate vector (Table 2), which eliminated the difference in dimensions and provided standard input for subsequent matching; Secondly, the degradation rate vector is dynamically time-warped and matched with the degradation stage feature template library (Table 3) to identify the set of devices in the degradation acceleration stage (Table 4), realizing the transformation from raw monitoring data to degradation stage discrimination; Third, the time series data of the equipment in the accelerated degradation stage are input into the autoregressive model to predict and generate degradation trajectories and calculate the failure probability in combination with the failure threshold (Table 5), and a subset of high-risk equipment is screened out (Table 6), thus completing the mapping from degradation state to failure risk quantification. Fourth, the peak failure probability moments of high-risk equipment are sorted to form a potential cascading failure time sequence (Table 7), providing a failure injection order for digital twin simulation; Fifth, inject faults sequentially into the digital twin model (Table 8), solve the incremental isolation scheme using a mixed integer optimization model (Table 9), integrate and generate a pre-reconstruction topology timing scheme (Table 10), and form an executable control instruction set; Sixth, when the detected fault probability exceeds the preventive intervention threshold, the initial topology preparation action (Table 11) is performed in advance to complete the preheating of the backup line; Seventh, when an actual fault occurs, the contingency plan can be directly invoked through comparison and matching to perform rapid isolation (Table 12), which significantly shortens the response time; Finally, the prediction model is updated based on actual fault data, the voltage recovery curve is solved and the hit rate is statistically analyzed (Table 13), providing feedback for subsequent strategy optimization.

[0079] Throughout the data flow process, the output of each step serves as the input for the next step, ensuring logical continuity and data consistency from equipment degradation monitoring data to topology reconfiguration control commands, and achieving effective linkage between degradation prediction information and topology reconfiguration decisions.

[0080] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A method for fault isolation and voltage recovery monitoring in a multi-terminal flexible DC transmission system, characterized in that, Includes the following steps: Collect degradation rate index data of key equipment in offshore wind power flexible DC system and generate degradation rate vector of each equipment; The degradation rate vector is matched with the degradation stage feature template library to identify the degradation stage identifier of each device, and devices in the degradation acceleration stage are selected to generate a set of devices in the degradation acceleration stage. For each device in the set of devices in the accelerated degradation stage, an autoregressive model is used to predict the degradation trajectory, and a short-term failure probability time series distribution is calculated in combination with the failure threshold. A subset of high-risk devices is generated based on the early warning threshold. Based on the peak times of the predicted failure probability of each device in the high-risk device subset, a potential cascading failure time sequence is generated. In the digital twin model, faults are injected sequentially according to the potential cascading fault time sequence, the incremental isolation topology scheme is solved, and the pre-reconstruction topology time sequence scheme is generated. When the failure probability of the first high-risk device exceeds the preventive intervention threshold, the initial topology preparation action in the pre-reconstruction topology timing scheme is executed; If a fault actually occurs, the actual fault location will be compared with the potential cascading fault time sequence, and an isolation scheme will be executed according to the degree of matching. After isolation is completed, the parameters of the autoregressive model are updated, and voltage recovery control is executed.

2. The method for fault isolation and voltage recovery monitoring of a multi-terminal flexible DC transmission system according to claim 1, characterized in that, The degradation rate index data includes temperature change rate, insulation resistance decrease rate, and partial discharge pulse frequency; the degradation rate vector is composed of the temperature change rate, insulation resistance decrease rate, and partial discharge pulse frequency; each component of the degradation rate vector is normalized using Z-score.

3. The method for fault isolation and voltage recovery monitoring of a multi-terminal flexible DC transmission system according to claim 1, characterized in that, Matching the degradation rate vector with the degradation stage feature template library includes: The similarity between the degradation rate vector and the feature templates of each degradation stage is calculated using a dynamic time warping algorithm. The degradation stage with the highest similarity is used as the current degradation stage identifier.

4. The method for fault isolation and voltage recovery monitoring of a multi-terminal flexible DC transmission system according to claim 1, characterized in that, The generation methods of the degradation stage feature template library include: Collect time series vectors of degradation rates of historically failed equipment throughout its entire lifespan; The degradation rate vector time series is divided into normal operation stage, early degradation stage, accelerated degradation stage and near-failure stage according to the time window before equipment failure. Statistical analysis was performed on the degradation rate vectors in each stage, and the mean, variance, and trend of each component were extracted as the degradation stage feature templates for that stage.

5. The method for fault isolation and voltage recovery monitoring of a multi-terminal flexible DC transmission system according to claim 1, characterized in that, The calculation method for the short-term failure probability time-series distribution includes: The difference between the predicted degradation trajectory value and the failure threshold is denoted as the degradation margin. After normalizing the degradation margin, the failure rate function value is obtained by substituting it into the exponential decay function. The cumulative failure probability is obtained by integrating the failure rate function value over time and performing an exponential transformation.

6. The method for fault isolation and voltage recovery monitoring of a multi-terminal flexible DC transmission system according to claim 1, characterized in that, The input to the autoregressive model is the recent degradation rate time series of the equipment, and the output is the degradation rate prediction value, the upper bound prediction value, and the lower bound prediction value of the degradation trajectory at each time in the future prediction time domain. The conservative failure probability is calculated based on the upper bound prediction value as a screening criterion. The degradation state prediction value is obtained by cumulatively integrating the degradation rate prediction value in the time dimension.

7. The method for fault isolation and voltage recovery monitoring of a multi-terminal flexible DC transmission system according to claim 1, characterized in that, Solving for incremental isolation topology schemes includes: Based on the system topology connection matrix, a breadth-first traversal is performed starting from the fault point, and all nodes and connection lines with a traversal depth within a preset number of layers are included in the influence domain. A mixed-integer optimization model is used to solve for the break point selection scheme that minimizes isolation loss. The objective function of the mixed-integer optimization model includes a break point operation cost term and a load loss weighting term. The incremental isolation topology scheme is merged with the existing isolation scheme, and the isolation breakpoint set in the subsequent stage is a superset of the isolation breakpoint set in the previous stage.

8. The method for fault isolation and voltage recovery monitoring of a multi-terminal flexible DC transmission system according to claim 1, characterized in that, The initial topology preparation process adopts a hierarchical strategy: When the failure probability is between the preventive intervention threshold and the medium risk threshold, perform a warm-up operation on the backup transmission path. When the probability of failure exceeds the medium risk threshold, a power margin pre-adjustment operation is performed simultaneously. The preheating operation includes sending a closing command to the relevant circuit breaker of the standby line and applying a test voltage lower than the rated value to the standby line.

9. The method for fault isolation and voltage recovery monitoring of a multi-terminal flexible DC transmission system according to claim 1, characterized in that, Implementing isolation schemes based on matching degree includes: The device matching degree and the time matching degree are calculated. The device matching degree is obtained by comparing the consistency between the actual fault device identifier and the predicted fault device identifier in the potential cascading fault time sequence. The time matching degree is obtained by normalizing the time difference between the actual fault time and the predicted fault time. The weighted average of the device matching degree and the time matching degree is used as the matching degree; When the matching degree is higher than the matching threshold, the isolation instruction of the corresponding stage in the pre-reconstructed topology timing scheme is directly invoked; When the matching degree is lower than the matching threshold, the process is executed after incremental correction based on the pre-reconstructed topology timing scheme.

10. The method for fault isolation and voltage recovery monitoring of a multi-terminal flexible DC transmission system according to claim 1, characterized in that, The autoregressive model parameter update includes: using the degradation rate time series before the actual failure as the new training sample, using the actual failure time as the label value, adding the new training sample to the historical training set using a sliding window method, re-estimating the coefficient parameters of the autoregressive model using the least squares method, and using the updated autoregressive model for subsequent equipment failure prediction.

Images