A new energy station cluster active sensing method and system

Abstract

The present application belongs to the field of new energy power system fault perception technology, and provides a new energy station cluster active perception method and system, comprising: by comparing and analyzing the fault propagation time and perception delay time of the fault node in the historical cascading failure, identifying the perception lag fault and judging whether the cluster perception system exists the perception lag phenomenon, if exists, determining the reference propagation time by the fault propagation time stability analysis and generating the fault propagation library, then combining the physical parameters and the correction coefficient to calculate the fault node perception time, matching the target fault propagation chain according to the current fault characteristics, finally comparing the reference propagation time with the perception time, judging the lag risk and hierarchical regulation, solving the problems of new energy station cluster fault perception lag and intervention not timely, improving the fault perception accuracy and timeliness, and guaranteeing the stable operation of power grid.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of fault perception technology of new energy power system, and particularly relates to a new energy station cluster active perception method and system. BACKGROUND

[0002] With large-scale development and grid connection of new energy power generation (such as wind power, photovoltaic power, energy storage, etc.), new energy station clusters have become an important part of power systems. However, new energy stations have characteristics such as large output fluctuation, complex topology structure, and fast fault propagation speed. Traditional fault perception methods are mostly based on a single station or a local area, and have problems such as high perception delay, insufficient prediction ability for cascading faults, and easy to miss intervention opportunity.

[0003] In actual operation, when a fault occurs in a new energy station cluster, the fault will quickly propagate along the power grid topology. If the perception system cannot timely identify the upstream key fault node, the intervention measure will be delayed, and then serious accidents such as cascading tripping and voltage collapse will be caused. In the prior art, fault perception mostly relies on centralized data processing, and the data transmission and analysis delay is large. Moreover, the time sequence characteristics of fault propagation and the identification priority of key nodes are not fully considered, it is difficult to quantify the perception lag risk, and it is also impossible to develop accurate control strategies for different fault propagation paths, resulting in that the perception system and the fault propagation speed are not matched, the fault diffusion cannot be effectively blocked, and the safe and stable operation of the power grid is seriously affected.

[0004] Therefore, the application provides a new energy station cluster active perception method and system. SUMMARY

[0005] In order to make up for the deficiencies of the prior art and solve at least one technical problem proposed in the background.

[0006] The technical scheme adopted by the application to solve the technical problems is: a new energy station cluster active perception method, characterized by comprising:

[0007] Step 1: By comparing and analyzing the fault propagation time and the perception delay time of the fault nodes in the historical cascading faults, identify the perception lag faults in the historical cascading faults, and determine whether the cluster perception system has a fault perception lag phenomenon by frequency analysis;

[0008] Step 2: If there is, perform stability analysis on the fault propagation time of the adjacent fault nodes in the fault propagation chain, determine the reference fault propagation time of the adjacent fault nodes, integrate each fault propagation chain and the reference fault propagation time of the adjacent fault nodes, and generate a fault propagation library;

[0009] Step 3: Calculate the theoretical sensing time of the fault node based on the physical parameters of the fault node and sensing node in the fault propagation database, calculate the correction coefficient based on the historical sensing time of the fault node, and combine the theoretical sensing time and the correction coefficient to calculate the sensing time of each fault node in the fault propagation database.

[0010] Step 4: Identify the target fault propagation chain by matching the current fault characteristics with the fault propagation database;

[0011] Step 5: Compare and analyze the baseline fault propagation time and sensing time of the current fault node and the downstream node in the target fault propagation chain to determine whether there is a risk of sensing lag in the current fault. If so, determine the control priority of the target fault propagation chain based on the sensing lag index and carry out hierarchical control.

[0012] Furthermore, the method for identifying the sensing lag fault is as follows:

[0013] For each historical cascading failure event:

[0014] The first fault node is identified by sudden changes in electrical quantities, and the time of the fault occurrence is recorded.

[0015] Based on the power grid topology, the fault occurrence time of downstream nodes is found sequentially, and the faulty nodes are arranged according to the fault occurrence time to generate the corresponding fault propagation chain.

[0016] For each chain of failure propagation:

[0017] Calculate the fault propagation time between adjacent faulty nodes, that is, the time difference between the fault of the upstream node and the fault of the downstream node;

[0018] Calculate the perception delay for each fault node, which is the time difference between the time the fault occurs and the time the system recognizes the fault;

[0019] The propagation-sensing time difference is calculated by subtracting the sensing delay of the upstream node from the fault of the upstream node to the fault propagation time of the downstream node.

[0020] Each fault node in the fault propagation chain is weighted according to the fault propagation time sequence, and the weighted fault chain identification rate is calculated.

[0021] Based on the comparative analysis of the timing of upstream node failure and the timing of downstream node failure, the delayed intervention time and the actual time available for intervention are obtained, and the intervention window miss rate is calculated proportionally.

[0022] For any fault propagation chain, the weighted fault chain unidentification rate U ω For: U ω =1-C ω C ω The weighted fault chain identification rate;

[0023] The perception lag index is obtained by adding the weighted fault chain unrecognition rate and the intervention window missing rate, and the perception lag index is compared with a preset lag index, and if the perception lag index is greater than the preset lag index, the historical cascading failure corresponding to the fault propagation chain is a perception lag failure.

[0024] Further, the calculation manner of the weighted fault chain recognition rate is:

[0025] Each node in the fault propagation chain is numbered according to the fault propagation time sequence, the source node is level 1, the first node downstream is level 2, and the like, each node is weighted according to the node number, and the formula is: node weight = total node number - node number + 1;

[0026] The weighted fault chain recognition rate is calculated, and the formula is: weighted fault chain recognition rate = sum of weights of recognized nodes / sum of weights of all nodes in the fault propagation chain.

[0027] Further, the acquisition manner of the lag intervention time and the actual intervenable time is:

[0028] If the upstream node failure occurrence time is earlier than the downstream node failure trigger time, the lag intervention time is calculated , and the formula is: , wherein T down is the downstream node failure trigger time, T up is the upstream node failure occurrence time, and T min is the minimum execution time of intervention;

[0029] The determination manner of the minimum execution time of intervention is obtained according to the technical parameters of the regulation and control equipment and statistical historical data;

[0030] If , the lag intervention time is ;

[0031] If the upstream node failure occurrence time is later than the downstream node failure trigger time, the perception lag has caused the entire intervention window to be completely missed at this time, and the lag intervention time is the difference between the upstream failure occurrence time and the downstream failure trigger time;

[0032] The actual intervenable time is the time period between the downstream failure trigger time and the upstream failure occurrence time.

[0033] Further, the determination manner of whether the cluster perception system has a failure perception lag phenomenon is:

[0034] The proportion of the perception lag failure in the historical cascading failure is obtained, and the proportion of the perception lag failure is obtained.

[0035] The perception lag fault proportion is compared with a preset proportion, and if the perception lag fault proportion is greater than the preset proportion, the cluster perception system has a fault perception lag phenomenon.

[0036] Further, the determination process of the reference fault propagation time is:

[0037] The perception lag fault data is screened out from the historical cascading failure data, and a fault propagation chain corresponding to the perception lag fault is extracted;

[0038] Based on any one fault propagation chain:

[0039] For each pair of adjacent fault nodes:

[0040] The fault propagation time of the adjacent fault node is extracted from all fault propagation chains containing the adjacent fault node;

[0041] The coefficient of variation of the fault propagation time of the adjacent fault node in each fault propagation chain is calculated, and the coefficient of variation is compared with a preset coefficient of variation, if the coefficient of variation is less than the preset coefficient of variation, the reference propagation time is the mean value of the fault propagation time, otherwise, the minimum value is taken.

[0042] Further, the calculation method of the perception time of each fault node in the fault propagation library is:

[0043] Based on the physical law of the power system, the theoretical calculation of each component of the perception time is performed:

[0044] The data transmission delay is calculated according to the signal propagation speed and the transmission path parameter, and the formula is: T 传输 = line length / signal propagation speed + data frame size / link bandwidth;

[0045] The analysis and decision delay is calculated based on the algorithm processing capacity, and the formula is: T 分析 = 1 / sampling frequency x algorithm processing period number;

[0046] The theoretical perception time is T 基础 = T 传输 + T 分析 ;

[0047] The same fault as the current fault type and physical parameter is screened out from the perception lag fault data, and a historical sample set is obtained: {(T base,i , T act,i )|i=1,2,...,n}, wherein n is the number of faults in the historical sample set, T base,i is the theoretical perception time of the i-th perception lag fault, and T act,i is the actual perception time of the i-th perception lag fault;

[0048] For each perception lag fault in the historical sample set, calculate the deviation rate δ of the theoretical perception time and the actual perception time, and calculate the average deviation rate δ of the perception time of all perception lag faults in the historical sample set to obtain the average deviation rate δ 平 ;

[0049] The correction coefficient K is K = 1 + δ 平 ;

[0050] For each fault node in the fault propagation library, multiply the theoretical perception time by the correction coefficient to obtain the perception time.

[0051] Further, the matching manner of the target fault propagation chain is:

[0052] All fault propagation chains containing the current fault node are screened from the fault propagation library to form a candidate chain set;

[0053] For each fault propagation chain in the candidate chain set:

[0054] If the current fault node is the source node, each fault propagation chain is the target fault propagation chain;

[0055] Otherwise, determine the level of the current fault node in the candidate chain, and trace back to the upstream node, screen out all fault nodes from the source node to the current fault node in the candidate chain, and count the number of fault nodes, until the source node, and calculate the proportion of the number of fault nodes to the total number of nodes in the fault propagation chain to obtain the proportion of matching nodes;

[0056] Compare the proportion of matching nodes with the preset proportion of matching nodes, if the proportion of matching nodes is greater than the preset proportion of matching nodes, the fault propagation chain is the target fault propagation chain.

[0057] Further, the manner of judging whether the current fault exists a perception lag risk and regulating is:

[0058] For any target fault propagation chain:

[0059] Compare the perception time T 感知 of the current fault node with the reference fault propagation time T 传播 between the current fault node and the downstream node.

[0060] If T 感知 > T 传播 , the perception of the current fault node will lag behind the downstream node, and there is a perception lag risk.

[0061] Arrange the target fault propagation chain in descending order according to the perception lag index to obtain the regulation priority of the target fault propagation chain.

[0062] A new energy station cluster active perception system comprises the following modules:

[0063] Fault perception lag phenomenon judgment module: by comparing and analyzing the fault propagation time and the perception delay time of the fault nodes in the historical cascading faults, identify the perception lag faults in the historical cascading faults, and judge whether the cluster perception system has the fault perception lag phenomenon through frequency analysis;

[0064] Fault propagation library construction module: if there is, perform stability analysis on the fault propagation time of adjacent fault nodes in the fault propagation chain, determine the reference fault propagation time of the adjacent fault nodes, integrate each fault propagation chain and the reference fault propagation time of the adjacent fault nodes, and generate a fault propagation library;

[0065] Perception time calculation module: calculate the theoretical perception time of the fault nodes according to the physical parameters of the fault nodes and the perception nodes in the fault propagation library, calculate the correction coefficient according to the historical perception time of the fault nodes, and calculate the perception time of each fault node in the fault propagation library by combining the theoretical perception time and the correction coefficient;

[0066] Target fault propagation chain identification module: identify the target fault propagation chain by matching the current fault characteristics with the fault propagation library;

[0067] Perception lag risk judgment and control module: compare and analyze the reference fault propagation time and the perception time of the current fault nodes and the downstream nodes in the target fault propagation chain, judge whether the current fault has the perception lag risk, if there is, determine the control priority of the target fault propagation chain according to the perception lag index, and perform hierarchical control.

[0068] The beneficial effects of the present application are as follows: the multi-dimensional analysis accurately identifies the perception lag fault, judges whether the cluster perception system has a systematic lag problem, avoids misjudgment of occasional delays or ignores systematic risks, provides accurate basis for subsequent optimization, reduces unnecessary resource investment, at the same time guarantees the safety of power grid, constructs a fault propagation library, integrates the benchmark propagation time of adjacent fault nodes and line parameters, provides rich and reliable historical data support for subsequent analysis of current fault, makes the analysis of fault propagation law more targeted and accurate, lays a solid foundation for fault prediction, combines with the physical parameter calculation theory to perceive time, and introduces a correction coefficient to make up for the deviation between theory and practice, effectively improves the accuracy of fault point perception time prediction, avoids the idealized deviation of pure theoretical calculation, ensures that the perception time is highly consistent with the actual operation, quickly matches the target fault chain based on the current fault characteristics, can quickly lock the possible fault propagation path, especially when the current fault node is not the source node, the matching node proportion is calculated by tracing the upstream node, which improves the accuracy and efficiency of target fault chain matching, saves time for subsequent risk judgment, and adopts different risk judgment and control strategies according to the number of target fault chains, not only can accurately identify the perception lag risk and calculate the lag time, but also can improve the perception speed and intervention efficiency from different dimensions through hierarchical control (communication transmission layer, data processing layer, and perception parameter optimization layer), ensure that the perception speed can cover the fault propagation speed, effectively block the fault diffusion, and guarantee the stable operation of the new energy station cluster and the power grid. BRIEF DESCRIPTION OF DRAWINGS

[0069] The present application will be further described below with reference to the accompanying drawings.

[0070] Figure 1 is a step flow chart of a new energy station cluster active perception method according to embodiment 1 of the present application;

[0071] Figure 2 is a logic judgment diagram for judging whether the cluster perception system has a fault perception lag phenomenon according to embodiment 1 of the present application;

[0072] Figure 3 is a program block diagram of a new energy station cluster active perception system according to embodiment 2 of the present application. DETAILED DESCRIPTION

[0073] In order to make the technical means, creative features, purposes and effects realized by the present application easy to understand, the present application will be further described below with reference to the specific embodiments.

[0074] Embodiment 1: please refer to Figure 1 The new energy station cluster active perception method according to the embodiment of the present application comprises the following steps:

[0075] Step one: By comparing the fault propagation time and the perception delay time of the fault nodes in the historical cascading failure, identify the perception lag fault in the historical cascading failure, and determine whether the cluster perception system has fault perception lag phenomenon through frequency analysis;

[0076] Please refer to Figure 2 As shown in FIG. 1, in step one, the process of identifying the perception lag fault in the historical cascading failure includes:

[0077] Obtain the historical cascading failure data of the new energy station cluster perception system, including fault event records, perception system logs, and power grid topology parameters, wherein the historical cascading failure refers to a fault event in which more than one station in the new energy station cluster fails;

[0078] Among them, the fault event record includes fault type, fault node and fault start and end timestamp, the perception system log includes data collection time, data transmission completion time and fault identification time, and the power grid topology parameter includes station distance, line impedance, tie line switch state, etc.

[0079] For each historical cascading failure event:

[0080] Determine the first fault node by electrical quantity mutation (such as voltage step-down starting point, power mutation to 0), and record the fault occurrence time T0;

[0081] According to the power grid topology, find the fault occurrence time of the downstream node in turn, for example: the first fault node is A station, the power grid topology is A->B->C line connection, and the abnormal time of downstream nodes B and C is: the time when the voltage of B station starts to deviate from the threshold T1 (such as the starting point of voltage from 500kV to 470kV), and the time when the inverter protection of C station acts (off-grid) T2;

[0082] Arrange each fault node according to the fault occurrence time to generate the corresponding fault propagation chain;

[0083] For each fault propagation chain:

[0084] Calculate the fault propagation time, which is the time difference between the upstream node fault and the downstream node fault ;

[0085] Calculate the perception delay, which is the total lag from fault occurrence to system identification, that is, the time difference between the fault occurrence time and the time when the system identifies the fault ;

[0086] The perception delay includes a data transmission delay and an analysis decision delay, wherein the data transmission delay is a time difference between a time when data arrives at the system and a time when the data is collected, and the analysis decision delay is a time difference between a time when the system identifies the fault and the time when the data arrives at the system;

[0087] The propagation-perception time difference is calculated, that is, a total perception delay of the upstream node The difference between the fault of the upstream node and the fault propagation time of the downstream node , ;

[0088] If , it indicates that the upstream fault identification lags behind the downstream fault propagation;

[0089] It should be noted that the diffusion logic of the fault propagation chain is: upstream node fault -> downstream node spread, early identification of the upstream node (especially the source node) is the core premise of blocking the propagation, and if the source node is missed, the downstream node cannot trace the root cause even if it identifies again, and the intervention measure will lag behind;

[0090] Each node in the fault propagation chain is numbered according to the fault propagation time sequence, the source node is level 1, the first node downstream is level 2, and so on, each node is assigned a weight according to the node number, and the formula is: node weight = total node number - node level + 1;

[0091] The weighted fault chain identification rate is calculated, and the formula is: weighted fault chain identification rate = sum of weights of identified nodes / sum of weights of all nodes in the fault chain;

[0092] It can be understood that the physical meaning of the weighted fault chain identification rate is: the identification value of the key node (upstream / source node) is highlighted through the weight, and whether the perception system has grasped the node that needs to be identified most in the fault chain is quantified;

[0093] The intervention window missing rate is calculated, that is, the ratio of the lagging intervention time to the actual intervention time;

[0094] The actual intervention time is the time period from the occurrence of the upstream fault to the time when the downstream node is affected by the fault (loses the intervention significance), that is, the time period between the downstream fault triggering time and the upstream fault occurrence time;

[0095] The lagging intervention time is the time that the subsequent intervention action cannot be completed before the downstream fault occurs due to the late identification of the upstream fault by the perception system, and finally misses the time to block the fault propagation, and the calculation method is:

[0096] If the upstream node fault occurrence time is earlier than the downstream fault triggering time, according to the downstream node fault triggering time T down , the upstream node fault occurrence time T up ​and the minimum intervention execution time T min , the remaining intervention time after identification is calculated , the formula is: ;

[0097] If , it means that the intervention can be completed before the downstream node failure trigger time, otherwise, the missed intervention time due to the perception lag is ;

[0098] If the upstream node failure occurs at a time later than the downstream node failure trigger time, the lag intervention time is the difference between the upstream failure occurrence time and the downstream failure trigger time at this time, the perception lag has caused the entire intervention window to be completely missed;

[0099] The upstream node failure occurrence time is the time when the new energy cluster perception system identifies the upstream node failure, the downstream node failure trigger time is the time when the downstream node actually fails, and the failure occurrence time in the "arrangement of each failure node in the failure chain according to the failure occurrence time" is the time when each failure node actually fails. When the time when the new energy cluster perception system identifies the upstream node failure is later than the time when the downstream node actually fails, it means that the perception lag has caused the entire intervention window to be completely missed, for example: the time when the new energy cluster perception system identifies the upstream node failure is 8:00, and the time when the downstream node actually fails is 7:00. When the new energy cluster perception system identifies the upstream node failure, the failure of the upstream node has propagated to the downstream node, causing the downstream node to fail, and the entire intervention window is completely missed;

[0100] Among them, the determination method of the minimum intervention execution time includes but is not limited to:

[0101] Regulation device technical parameters: such as the "response time" of inverter active power adjustment (usually marked by the manufacturer ≤50ms), the "trip time" of field station load shedding device (≤100ms);

[0102] Statistical historical data: from the cluster regulation system log, extract the time difference of "instruction issuing time → action completion time" of nearly 100 times of the same intervention (such as cutting A field station load), take the 95% quantile value (avoid extreme value, more conservative);

[0103] It can be understood that the physical meaning of the intervention window loss rate is: the proportion of the effective intervention opportunity wasted by the perception system due to lag, which directly reflects the matching degree of the perception speed and the failure propagation speed;

[0104] Since the weighted failure chain identification rate C ωIn contrast to the response of the intervention window miss rate to perceived lag risk, the weighted fault chain identification rate needs to be converted in the same direction. The core idea is to convert the identification completeness into the unidentified risk level so that the higher the value, the greater the perceived lag risk, which is consistent with the direction of the intervention window miss rate.

[0105] For any fault propagation chain, the weighted fault chain unidentification rate U ω For: U ω =1-C ω C ω The weighted fault chain identification rate;

[0106] The weighted failure chain unidentification rate and the intervention window miss rate are added together to obtain the perception lag index. The perception lag index is then compared with the preset lag index. If the perception lag index is greater than the preset lag index, the historical chain failure corresponding to the failure propagation chain is a perception lag failure.

[0107] The preset lag index is a risk threshold. When the actual perceived lag index exceeds the preset value, it indicates that the combined risk of unidentified fault propagation and missed intervention has exceeded the system's acceptable range. This must be classified as a perceived lag fault, triggering more urgent optimization. The preset lag index is set based on industry compliance and benchmarks, for example:

[0108] Power System (GB / T38953-2020 "Guidelines for the Safety and Stability of Power Systems"): requires that the perception lag of critical faults be ≤500ms, and the corresponding perception lag index (which needs to be converted in conjunction with specific algorithms) is usually no more than 0.1;

[0109] It is understandable that the physical meaning of the perception lag index is: to integrate the lag risk of identifying quality defects and wasting intervention opportunities into a directly comparable quantitative value, which comprehensively reflects the overall performance of the perception system in identifying key faults and supporting timely intervention. The larger the value, the higher the degree of perception lag.

[0110] In step one, the process of determining whether the cluster sensing system has a fault detection lag includes:

[0111] The proportion of perceived lag faults is obtained by statistically analyzing the proportion of perceived lag faults in historical cascading failures.

[0112] The percentage of perceived lag faults is compared with the preset percentage. If the percentage of perceived lag faults is greater than the preset percentage, then the cluster perception system has a fault perception lag phenomenon.

[0113] The core role of the preset proportion is to distinguish between accidental perception lag and systematic perception lag. It needs to pass through the quantitative threshold to determine whether the lag failure of the cluster perception system is an individual accident or a widespread systemic defect. The preset proportion is set based on industry compliance and benchmarks, for example:

[0114] Electric power industry (such as “GB / T 36548-2018 New Energy Station Grid Connection Operation Technical Requirements”): It is clear that the failure lag rate of the cluster perception system (i.e., the proportion of perception lag failure) should not exceed 15%;

[0115] New energy grid connection dispatching specification: Some regional dispatching centers require that the perception lag failure proportion of core stations be ≤8%, so the preset proportion of the corresponding cluster needs to be ≤8%;

[0116] Industry best practices: The public operation and maintenance reports of leading new energy operators (such as State Energy Group and China Huaneng) show that the perception lag failure proportion of mature clusters is usually controlled at 8%~12%, which can be referenced for setting;

[0117] The role of determining whether the cluster perception system has a failure perception lag phenomenon is:

[0118] From the historical data dimension, the perception lag rule is mined, and the perception lag failure is accurately identified through multi-index quantitative analysis. At the same time, the lag failure proportion is calculated to determine whether the cluster perception system has a systematic lag problem. This avoids misjudging accidental delays as system defects, leading to resource waste, and prevents ignoring systematic lag, which may cause safety risks. It provides accurate problem positioning basis for subsequent targeted optimization;

[0119] Step two: if there is, perform stability analysis on the fault propagation time of adjacent fault nodes in the fault propagation chain to determine the baseline fault propagation time of adjacent fault nodes. Integrate each fault propagation chain and the baseline fault propagation time of adjacent fault nodes to generate a fault propagation library;

[0120] In step two, the process of performing stability analysis on the fault propagation time of adjacent fault nodes in the fault propagation chain includes:

[0121] In the historical fault record, filter out the perception lag failure and extract the corresponding fault propagation chain;

[0122] Based on any one fault propagation chain:

[0123] For each pair of adjacent fault nodes:

[0124] Extract the fault propagation time from all fault propagation chains containing the adjacent fault nodes, i.e., the time difference between the upstream node failure and the downstream node failure;

[0125] The coefficient of variation of the fault propagation time is calculated, and the coefficient of variation is compared with a preset coefficient of variation. If the coefficient of variation is less than the preset coefficient of variation, the reference propagation time is the mean value of the fault propagation time, otherwise, the minimum value is taken.

[0126] The preset coefficient of variation is set as a threshold for distinguishing whether the fault propagation rule is predictable or the propagation rule fluctuates randomly. The preset coefficient of variation is set based on historical fault record data, and specifically:

[0127] The fault propagation time data set of all adjacent fault node pairs in the historical data is extracted.

[0128] The coefficient of variation of the fault propagation time of each fault node pair is calculated.

[0129] The distribution of all coefficients of variation (such as drawing a histogram) is counted, and the natural breaking point of the data is found, that is, the coefficients of variation of most stable node pairs are concentrated in a certain interval, and the CV suddenly increases (corresponding to unstable node pairs) beyond the interval.

[0130] The preset coefficient of variation can be set as an approximate value of the natural breaking point (for example, if 80% of stable node pairs CV≤0.25, the breaking point is near 0.25, and the preset value can be set to 0.25).

[0131] The reference propagation time is a standardized time value set for a specific pair of adjacent fault nodes (such as node X as upstream and node Y as downstream), and the core is to represent the typical / credible time length of fault propagation from upstream to downstream between a pair of adjacent fault nodes.

[0132] All fault propagation chains are integrated, and the reference fault propagation time and connection line parameters of each pair of adjacent fault nodes are marked to generate a fault propagation library.

[0133] For example, the structure of the fault propagation library is as follows:

[0134] | Fault propagation chain ID | Propagation path | Adjacent fault nodes | Connection line parameters | Reference fault propagation time | Fault type label |;

[0135] | FC001 | A wind field -> B photovoltaic -> C energy storage | A-B | 110kV, X=5Ω | 30.3s | Voltage sag chain trip |;

[0136] | FC001 | A wind field -> B photovoltaic -> C energy storage | B-C | 35kV, X=2Ω | 20s | Voltage sag chain trip |;

[0137] | FC002 | D wind farm -> E photovoltaic | D-E | 110 kV, X = 6Ω | 18s | fan overload chain off-network |

[0138] The role of constructing the fault propagation library is:

[0139] Based on the historical propagation data of the perception lag fault, a standardized and structured fault propagation library is established, the benchmark propagation time and line parameters of adjacent fault nodes are integrated, the fault propagation law is converted into a reusable standardized data resource, a reliable historical reference is provided for the subsequent matching of the propagation path of the current fault and the comparison of the perception time, and the problem of lack of unified propagation data support in traditional analysis is solved;

[0140] Step three: calculate the theoretical perception time of the fault node according to the physical parameters of the fault node and the perception node in the fault propagation library, calculate the correction coefficient according to the historical perception time of the fault node, and calculate the perception time of each fault node in the fault propagation library by combining the theoretical perception time and the correction coefficient;

[0141] In step three, the calculation process of the perception time of each fault node in the fault propagation library includes:

[0142] The physical parameters of each fault node and the perception node in the fault propagation library include transmission parameters and algorithm parameters, wherein the transmission parameters include line length, signal propagation speed, data frame size and link bandwidth, and the algorithm parameters include sampling frequency and algorithm processing period number;

[0143] Based on the physical law of the power system, the theoretical calculation of each component of the perception time is carried out:

[0144] According to the signal propagation speed and the transmission path parameters, the data transmission delay is calculated, and the formula is: T 传输 = line length / signal propagation speed + data frame size / link bandwidth;

[0145] Based on the algorithm processing capacity, the analysis and decision delay is calculated, and the formula is: T 分析 = 1 / sampling frequency * algorithm processing period number;

[0146] The theoretical perception time is T 基础 = T 传输 + T 分析 ;

[0147] In the perception lag fault data, the same fault as the current fault type and physical parameters is selected to obtain the historical sample set: {(T base,i , T act,i )|i = 1, 2,..., n}, wherein n is the number of faults in the historical sample set, T base,i is the theoretical perception time of the i-th perception lag fault, and T act,ithe actual perception time of the ith perception lag fault;

[0148] For each perception lag fault in the historical sample set, the deviation rate δ of the theoretical perception time and the actual perception time is calculated, and the average deviation rate δ of the perception time deviation rate of all perception lag faults in the historical sample set is calculated to obtain the average deviation rate δ 平 ;

[0149] The correction coefficient K is K = 1 + δ 平 ;

[0150] For each fault node in the fault propagation library, the theoretical perception time is multiplied by the correction coefficient to obtain the perception time;

[0151] It can be understood that the physical meaning of the correction coefficient is that the essence of the correction coefficient is the fusion coefficient of the deviation law of the ideal physical model and the actual system, and the core function is to make up for the difference between the theoretical calculation and the actual operation. The correction coefficient neither deviates from the physical law, nor reflects the long-term operation characteristics of the system, avoiding the idealized deviation of pure theoretical prediction;

[0152] The function of calculating the perception time of each fault node in each fault propagation library is:

[0153] Function 1: Combining physical laws and historical operation deviations, precise prediction of fault node perception time is realized;

[0154] Function 2: The scientificity of the prediction is ensured by calculating the theoretical perception time through physical parameters, the deviation between the theoretical model and the actual system is compensated by introducing the correction coefficient, and the idealized error of pure theoretical calculation is avoided, providing accurate time quantization basis for subsequent judgment of the matching degree of the perception speed and the fault propagation speed;

[0155] Step four: for the current fault, according to the fault characteristics, the fault propagation chain matching is carried out in the fault propagation library to obtain the target fault propagation chain;

[0156] In step four, the acquisition process of the target fault propagation chain includes:

[0157] All fault propagation chains containing the current fault node are screened from the fault propagation library to form a candidate chain set;

[0158] For each fault propagation chain in the candidate chain set:

[0159] If the current fault node is the source node, each fault propagation chain is the target fault propagation chain;

[0160] Conversely, the level of the current fault node in the candidate chain is determined, and the upstream nodes are traced back to determine whether each node is faulty until the source node. All faulty nodes between the source node and the current fault node in the candidate chain are screened out, and the proportion of matching nodes is calculated by the total number of nodes in the fault propagation chain to obtain the proportion of matching nodes;

[0161] The proportion of matching nodes is compared with the preset proportion of matching nodes. If the proportion of matching nodes is greater than the preset proportion, the candidate chain is the target fault propagation chain;

[0162] The core role of the preset matching node is to balance the accuracy and coverage of the fault propagation chain matching, which needs to be determined comprehensively in combination with three core dimensions of business scenarios, system characteristics and historical data quality. The essence is to find the balance point of not missing key chains and not introducing invalid chains. The setting process includes:

[0163] Extract key indicators from the fault propagation library:

[0164] The average length of the propagation chain of the historical perceived lag fault (if the chain length is 10 nodes, 90% needs to match 9 nodes, if the chain length is 3 nodes, 90% needs to match 3 nodes, which needs to be adjusted in combination with the chain length);

[0165] The path stability of the historical propagation chain (with the aid of the coefficient of variation: low coefficient of variation → stable path → high proportion);

[0166] The completeness rate of historical data (such as 80% of the chain has complete path records → the proportion should not exceed 80%, to avoid no matching chain);

[0167] Find the balance point with the recall rate-precision rate curve (P-R curve):

[0168] Recall rate: the proportion of real target chains selected out (the higher the better to avoid missing selection);

[0169] Precision: the proportion of real target chains in the selected chains (the higher the better to be accurate);

[0170] Test the P-R value corresponding to different preset proportions (such as 60%, 70%, 80%, 90%), and select the proportion with a recall rate not lower than an acceptable threshold (such as 80%) and the highest precision as the final value;

[0171] The role of matching the target fault propagation chain is:

[0172] Quickly lock the possible propagation path of the current fault, determine the target fault propagation chain by screening the candidate chain containing the current fault node and judging the node matching degree, solve the problem of fuzzy prediction and long time consumption in traditional fault analysis, gain key time for subsequent lag risk judgment and control, and ensure the timeliness of intervention decision;

[0173] Step five: Compare the benchmark fault propagation time and the perception time of the current fault node and the downstream node in the target fault propagation chain, judge whether the current fault exists the risk of perception lag, if so, determine the regulation priority of the target fault propagation chain according to the perception lag index, and carry out hierarchical regulation;

[0174] In step five, the process of judging whether the current fault exists the risk of perception lag includes:

[0175] For any target fault propagation chain:

[0176] Compare the perception time T 感知 of the current fault node with the fault benchmark propagation time T 传播 between the current fault node and the downstream node;

[0177] If T 感知 ≤ T 传播 , the perception speed can cover the propagation speed, no compensation is needed, and the data is processed according to the normal process;

[0178] Otherwise, the perception of the current fault node will lag behind the downstream node, and there is a risk of perception lag;

[0179] If there is a risk of perception lag, the predicted perception time T 感知 of the current fault node is subtracted from the fault benchmark propagation time T 传播 between the upstream node and the downstream node to obtain the lag time;

[0180] Arrange the target fault propagation chain in descending order of the perception lag index to obtain the regulation priority of the target fault propagation chain;

[0181] Regulate the target fault propagation chain according to the regulation priority, including:

[0182] Communication transmission layer regulation:

[0183] Dynamically improve the data flow priority of the current fault node (such as including the collected data of the current fault node in the highest priority queue of QoS to avoid being squeezed out of bandwidth by other low priority data);

[0184] Enable the backup communication link of the current fault node (such as switching from wireless transmission to optical fiber link, and the transmission delay can be reduced from 50 ms to within 10 ms);

[0185] Data processing layer regulation:

[0186] Allocate exclusive edge computing resources for the current fault node (avoid high load of centralized server leading to analysis delay, such as scheduling the fault analysis task of X to the adjacent edge node);

[0187] Lightweight fault identification algorithm for the current fault node (such as using voltage sag threshold to directly determine instead of complex machine learning model, analysis delay from 30ms to 5ms);

[0188] The perception parameter optimization layer regulates:

[0189] Temporarily increase the electrical quantity collection frequency of the current fault node (such as from 50Hz collection to 200Hz, capture voltage / power mutation signal earlier);

[0190] Fine-tune the fault identification threshold of the current fault node (such as voltage sag identification threshold from "5% drop" to "3% drop", trigger identification in advance, combined with multi-parameter verification (voltage + power) to avoid false positives);

[0191] The role of judging whether there is a lag risk and regulating is:

[0192] According to the number and parameters of the target fault propagation chain, accurately judge the perception lag risk of the current fault, develop hierarchical regulation strategies (communication transmission layer, data processing layer, perception parameter optimization layer) according to the risk level, improve the perception speed from three core dimensions of data transmission, processing, and perception parameters, ensure that the perception speed covers the fault propagation speed, effectively block the fault diffusion, and solve the problem of single strategy and insufficient pertinence in traditional intervention.

[0193] The technical scheme and advantages of the embodiments of the present application are as follows: by comparing and analyzing the fault propagation time and the sensing delay time of the fault nodes in the historical cascading faults, the sensing lag faults are identified in the historical cascading faults, and whether the fault sensing lag phenomenon exists in the cluster sensing system is determined through frequency analysis; if the fault sensing lag phenomenon exists, the stability of the fault propagation time of the adjacent fault nodes in the fault propagation chain is analyzed, the reference fault propagation time of the adjacent fault nodes is determined, the fault propagation library is integrated by integrating each fault propagation chain and the reference fault propagation time of the adjacent fault nodes, the theoretical sensing time of the fault nodes is calculated according to the physical parameters of the fault nodes and the sensing nodes in the fault propagation library, the correction coefficient is calculated according to the historical sensing time of the fault nodes, the sensing time of each fault node in the fault propagation library is calculated by combining the theoretical sensing time and the correction coefficient, the target fault propagation chain is identified by matching the current fault characteristics with the fault propagation library, the reference fault propagation time and the sensing time of the current fault node and the downstream node in the target fault propagation chain are compared and analyzed, whether the current fault exists the sensing lag risk is determined, if the sensing lag risk exists, the control priority of the target fault propagation chain is determined according to the sensing lag index, and hierarchical control is performed. By comparing and analyzing the fault propagation time and the sensing delay time of the fault nodes, the sensing lag faults are identified and the fault sensing lag phenomenon is determined; if the fault sensing lag phenomenon exists, the reference fault propagation time is determined according to the stability analysis of the fault propagation time of the adjacent fault nodes in the fault propagation chain, the fault propagation library is constructed, the sensing time of each fault node is calculated by combining the physical parameters of the fault nodes and the sensing nodes in the fault propagation library and the historical sensing time of the fault nodes, the target fault propagation chain is identified by matching the current fault characteristics with the fault propagation library, the reference fault propagation time and the sensing time of the current fault node and the downstream node in the target fault propagation chain are compared and analyzed, the sensing lag risk of the current fault is determined, if the sensing lag risk exists, the control priority of the target fault propagation chain is determined according to the sensing lag index, and hierarchical control is performed, the problems of the cluster fault sensing lag and the intervention not being timely in the new energy station are solved, the fault sensing accuracy and timeliness are improved, and the stable operation of the power grid is ensured.

[0194] Embodiment 2: please refer to Figure 3 The new energy station cluster active sensing system provided by the embodiments of the present application comprises the following modules:

[0195] The fault sensing lag phenomenon judgment module: by comparing and analyzing the fault propagation time and the sensing delay time of the fault nodes in the historical cascading faults, the sensing lag faults are identified in the historical cascading faults, and whether the fault sensing lag phenomenon exists in the cluster sensing system is determined through frequency analysis;

[0196] The process of identifying the sensing lag faults in the historical cascading faults comprises:

[0197] The historical cascading failure data of the new energy station cluster perception system is acquired, including failure event records, perception system logs, and power grid topology parameters, wherein the historical cascading failure refers to a failure event in which more than one station in the new energy station cluster fails;

[0198] The failure event records include failure types, failure nodes, and failure start and end time stamps, the perception system logs include data collection times, data transmission completion times, and failure identification times, and the power grid topology parameters include station-to-station distances, line impedances, and tie-line switch states.

[0199] For each historical cascading failure event:

[0200] The first failure node is determined through an electrical quantity mutation (such as a voltage sudden drop start point or a power mutation of 0), and the failure occurrence time T0 is recorded;

[0201] According to the power grid topology, the failure occurrence times of downstream nodes are sequentially found, for example: the first failure node is station A, the power grid topology is a line connection of A->B->C, and the abnormal times of the downstream nodes B and C are: the time T1 at which the voltage of station B starts to deviate from the threshold value (such as the start point of the voltage decreasing from 500 kV to 470 kV), and the time T2 at which the inverter protection of station C acts (is disconnected from the grid);

[0202] The failure nodes are arranged according to the failure occurrence times to generate corresponding failure propagation chains;

[0203] For each failure propagation chain:

[0204] The failure propagation time, that is, the time difference between the failure of the upstream node and the failure of the downstream node ;

[0205] The perception delay is calculated, which is the total lag from the failure occurrence to the system identification, that is, the time difference between the failure occurrence time and the time when the system identifies the failure ;

[0206] The perception delay includes a data transmission delay and an analysis and decision delay, wherein the data transmission delay is the time difference between the time when the data arrives at the system and the time when the data is collected, and the analysis and decision delay is the time difference between the time when the system identifies the failure and the time when the data arrives at the system;

[0207] The propagation-perception time difference is calculated, that is, the difference between the total perception delay of the upstream node and the failure propagation time of the upstream node to the downstream node ; , ;

[0208] If , it is indicated that the upstream failure identification lags behind the downstream failure propagation.

[0209] It should be noted that the spread logic of the fault propagation chain is: upstream node failure → downstream node spread, early identification of the upstream node (especially the source node) is the core premise of blocking propagation, if the source node is missed, the downstream node cannot trace the root cause, and the intervention measure will be delayed;

[0210] Number each node in the fault propagation chain according to the fault propagation time sequence, the source node is level 1, the first node downstream is level 2, and so on, assign weights to each node according to the node number, the formula is: node weight = total node number - node level + 1;

[0211] Calculate the weighted fault chain identification rate, the formula is: weighted fault chain identification rate = sum of weights of identified nodes / sum of weights of all nodes in the fault chain;

[0212] It can be understood that the physical meaning of the weighted fault chain identification rate is: through the weight, highlight the identification value of the key node (upstream / source node), and quantify whether the perception system has grasped the node that needs to be identified first in the fault chain;

[0213] Calculate the missed intervention window rate, which is the ratio of the lagging intervention time to the actual intervenable time;

[0214] Among them, the actual intervenable time is the time period from the occurrence of upstream failure to the time when the downstream node is affected by the failure (loses the meaning of intervention), that is, the time period between the downstream failure trigger time and the upstream failure occurrence time;

[0215] The lagging intervention time is the time that the subsequent intervention action cannot be completed before the downstream failure occurs due to the late identification of the upstream failure by the perception system, and finally misses the time to block the fault propagation, which is calculated as follows:

[0216] If the upstream node failure occurrence time is earlier than the downstream node failure trigger time, according to the downstream node failure trigger time T down , the upstream node failure occurrence time T up and the minimum execution time of intervention T min , the remaining intervenable time after identification is calculated as , the formula is: ;

[0217] If , it means that the intervention can be completed before the downstream node failure trigger time, otherwise, the missed intervention time due to perception lag is ;

[0218] If the upstream node failure occurs at a time later than the downstream node failure trigger time, the perception lag has caused the entire intervention window to be completely missed at this time, and the lag intervention time is the difference between the upstream failure occurrence time and the downstream failure trigger time;

[0219] The determination manner of the minimum intervention execution time includes but is not limited to:

[0220] The technical parameters of the regulation equipment, such as the "response time" of the active power adjustment of the inverter (usually marked by the manufacturer as less than or equal to 50 ms), and the "trip time" of the load shedding device of the field station (less than or equal to 100 ms);

[0221] Statistical historical data: Extract the time difference between the "instruction issuing time" and the "action completion time" of nearly 100 times of the same intervention (such as cutting the load of field A) from the cluster regulation system log, and take the 95% quantile value (to avoid extreme values and be more conservative);

[0222] It can be understood that the physical meaning of the intervention window loss rate is the proportion of the effective intervention opportunity wasted by the perception system due to lag, which directly reflects the matching degree of the perception speed and the fault propagation speed;

[0223] Since the weighted fault chain identification rate C ω The weighted fault chain identification rate needs to be converted in the same direction first, and the core idea is to convert the identification integrity into an un-identified risk degree, so that the higher the value is, the greater the perception lag risk is, which is consistent with the direction of the intervention window loss;

[0224] For any fault propagation chain, the weighted fault chain un-identified rate U ω is: ω U ω = 1-C ω , wherein C

[0225] The weighted fault chain un-identified rate and the intervention window loss rate are added to obtain the perception lag index, and the perception lag index is compared with the preset lag index. If the perception lag index is greater than the preset lag index, the historical cascading failure corresponding to the fault propagation chain is a perception lag fault;

[0226] It can be understood that the physical meaning of the perception lag index is to integrate the lag risks in the two dimensions of identification quality defects and intervention opportunity waste to form a quantitative value that can be directly compared, and comprehensively reflect the overall performance of the perception system in identifying key faults and supporting timely intervention. The greater the value is, the higher the perception lag degree is;

[0227] The process of judging whether the cluster perception system has a fault perception lag phenomenon includes:

[0228] In the historical cascading failure, the proportion of the statistical perception lag failure is obtained, and the proportion of the perception lag failure is obtained.

[0229] The proportion of the perception lag failure is compared with the preset proportion, and if the proportion of the perception lag failure is greater than the preset proportion, the cluster perception system has a failure perception lag phenomenon.

[0230] The failure propagation library construction module: if there is, the stability of the failure propagation time of the adjacent failure nodes in the failure propagation chain is analyzed, the reference failure propagation time of the adjacent failure nodes is determined, the reference failure propagation time of each failure propagation chain and the adjacent failure nodes is integrated, and the failure propagation library is generated.

[0231] The process of analyzing the stability of the failure propagation time of the adjacent failure nodes in the failure propagation chain includes:

[0232] In the historical failure record, the perception lag failure is screened out, and the corresponding failure propagation chain is extracted.

[0233] Based on any one failure propagation chain:

[0234] For each pair of adjacent failure nodes:

[0235] The failure propagation time is extracted from all failure propagation chains containing the adjacent failure nodes, that is, the time difference between the upstream node failure and the downstream node failure;

[0236] The coefficient of variation of the failure propagation time is calculated, and the coefficient of variation is compared with the preset coefficient of variation, if the coefficient of variation is less than the preset coefficient of variation, the reference propagation time is taken as the mean value of the failure propagation time, otherwise, the minimum value is taken.

[0237] All failure propagation chains are integrated, and the reference failure propagation time of each pair of adjacent failure nodes and the connection line parameters are marked to generate the failure propagation library.

[0238] The perception time calculation module: according to the physical parameters of the failure nodes and the perception nodes in the failure propagation library, the theoretical perception time of the failure nodes is calculated, and according to the historical perception time of the failure nodes, the correction coefficient is calculated, and the perception time of each failure node in the failure propagation library is calculated by combining the theoretical perception time and the correction coefficient.

[0239] The calculation process of the perception time of each failure node in the failure propagation library includes:

[0240] The physical parameters of the failure nodes and the perception nodes in the failure propagation library include transmission parameters and algorithm parameters, wherein the transmission parameters include line length, signal propagation speed and data frame size, and the algorithm parameters include sampling frequency and algorithm processing period number.

[0241] Based on the physical law of power system, the theoretical calculation of each component of the perception time is carried out:

[0242] According to the signal propagation speed and the transmission path parameter calculation data transmission delay, the formula is: 传输 = line length / signal propagation speed + data frame size / link bandwidth;

[0243] Based on the algorithm processing capacity calculation analysis decision delay, the formula is: 分析 = 1 / sampling frequency x algorithm processing period number;

[0244] The theoretical perception time is T 基础 =T 传输 +T 分析 ;

[0245] In the perception lag fault data, the same fault as the current fault type and physical parameter is screened to obtain the historical sample set: {(T base,i , T act,i )|i=1, 2,..., n}, wherein, n is the number of faults in the historical sample set, T base,i is the theoretical perception time of the i-th perception lag fault, and T act,i is the actual perception time of the i-th perception lag fault;

[0246] For each perception lag fault in the historical sample set, the deviation rate δ of the theoretical perception time and the actual perception time is calculated, and the average deviation rate K=1+δ 平 of all perception lag fault perception time deviation rates in the historical sample set is calculated;

[0247] The correction coefficient K is K=1+δ 平 ;

[0248] For each fault node in the fault propagation library, the theoretical perception time is multiplied by the correction coefficient to obtain the perception time;

[0249] It can be understood that the physical meaning of the correction coefficient is that the essence of the correction coefficient is the fusion coefficient of the deviation law of the ideal physical model and the actual system, and the core function is to make up for the difference between the theoretical calculation and the actual operation. The correction coefficient neither deviates from the physical law, nor reflects the long-term operation characteristics of the system, avoiding the idealized deviation of pure theoretical prediction;

[0250] The target fault propagation chain identification module: for the current fault, the fault propagation chain matching is carried out in the fault propagation library according to the fault characteristics, and the target fault propagation chain is obtained;

[0251] The acquisition process of the target fault propagation chain includes:

[0252] Screening all fault propagation chains containing the current fault node from the fault propagation library to form a candidate chain set;

[0253] For each fault propagation chain in the candidate chain set:

[0254] If the current fault node is the source node, each fault propagation chain is the target fault propagation chain;

[0255] Conversely, determine the level of the current fault node in the candidate chain, and trace upstream nodes to determine whether each node is faulty until the source node, screen out all fault nodes between the source node and the current fault node in the candidate chain, and calculate the proportion of matching nodes based on the total number of nodes in the fault propagation chain to obtain the proportion of matching nodes;

[0256] Compare the proportion of matching nodes with the preset proportion of matching nodes, if the proportion of matching nodes is greater than the preset proportion of matching nodes, the candidate chain is the target fault propagation chain;

[0257] The perception lag risk judgment and regulation module: comparing and analyzing the benchmark fault propagation time and the perception time of the current fault node and the downstream node in the target fault propagation chain, judging whether the current fault exists a perception lag risk, if so, determining the regulation priority of the target fault propagation chain according to the perception lag index, and performing hierarchical regulation;

[0258] The process of judging whether the current fault exists a perception lag risk includes:

[0259] For any target fault propagation chain:

[0260] Compare the perception time T 感知 of the current fault node with the fault benchmark propagation time T 传播 between the current fault node and the downstream node;

[0261] If T 感知 ≤ T 传播 , the perception speed can cover the propagation speed, no compensation is needed, and the data is processed according to the normal process;

[0262] Conversely, the perception of the current fault node will lag behind the downstream node, and there is a perception lag risk;

[0263] If there is a perception lag risk, the predicted perception time T 感知 of the current fault node is subtracted from the fault benchmark propagation time T 传播 between the upstream node and the downstream node to obtain the lag time;

[0264] Arrange the target fault propagation chain in descending order of the perception lag index to obtain the regulation priority of the target fault propagation chain;

[0265] The target fault propagation chain is regulated in order of regulation priority, including:

[0266] Communication transmission layer regulation:

[0267] Dynamically improving the data flow priority of the current fault node (such as including the collected data of the current fault node in the highest priority queue of QoS to avoid being squeezed out of bandwidth by other low priority data);

[0268] Enabling the backup communication link of the current fault node (such as switching from wireless transmission to optical fiber link, and the transmission delay can be reduced from 50ms to within 10ms);

[0269] Data processing layer regulation:

[0270] Allocating exclusive edge computing resources for the current fault node (to avoid high load of centralized servers causing analysis delay, such as scheduling the fault analysis task of X to a nearby edge node);

[0271] Enabling the lightweight fault identification algorithm of the current fault node (such as using voltage sag threshold to directly judge instead of complex machine learning model, and the analysis delay is reduced from 30ms to 5ms);

[0272] Sensing parameter optimization layer regulation:

[0273] Temporarily increasing the electrical quantity collection frequency of the current fault node (such as increasing from 50Hz to 200Hz to capture voltage / power mutation signals earlier);

[0274] Fine-tuning the fault identification threshold of the current fault node (such as adjusting the voltage sag identification threshold from "5%" to "3%", triggering identification earlier, and combining multiple parameter verification (voltage + power) to avoid false positives).

[0275] The above shows and describes the basic principles, main features and advantages of the present application. It should be understood by those skilled in the art that the present application is not limited to the above embodiments, and the above embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for active sensing in a cluster of new energy power stations, characterized in that: include: Step 1: By comparing and analyzing the fault propagation time and perception delay time of fault nodes in historical cascading failures, identify perception lag faults in historical cascading failures, and determine whether there is a fault perception lag phenomenon in the cluster perception system through frequency analysis. Step 2: If it exists, perform stability analysis on the fault propagation time of adjacent fault nodes in the fault propagation chain, determine the baseline fault propagation time of adjacent fault nodes, integrate each fault propagation chain and the baseline fault propagation time of adjacent fault nodes to generate a fault propagation library. Step 3: Calculate the theoretical sensing time of the fault node based on the physical parameters of the fault node and sensing node in the fault propagation database, calculate the correction coefficient based on the historical sensing time of the fault node, and combine the theoretical sensing time and the correction coefficient to calculate the sensing time of each fault node in the fault propagation database. Step 4: Identify the target fault propagation chain by matching the current fault characteristics with the fault propagation database; Step 5: Compare and analyze the baseline fault propagation time and sensing time of the current fault node and the downstream node in the target fault propagation chain to determine whether there is a risk of sensing lag in the current fault. If so, determine the control priority of the target fault propagation chain based on the sensing lag index and carry out hierarchical control.

2. The active sensing method for a new energy power station cluster according to claim 1, characterized in that: The method for identifying the sensing lag fault is as follows: For each historical cascading failure event: The first fault node is identified by sudden changes in electrical quantities, and the time of the fault occurrence is recorded. Based on the power grid topology, the fault occurrence time of downstream nodes is found sequentially, and the faulty nodes are arranged according to the fault occurrence time to generate the corresponding fault propagation chain. For each chain of failure propagation: Calculate the fault propagation time between adjacent faulty nodes, where the fault propagation time is the time difference between the fault of the upstream node and the fault of the downstream node; Calculate the perception delay for each fault node, where the perception delay is the time difference between the time the fault occurs and the time the system recognizes the fault; The propagation-sensing time difference is calculated by subtracting the sensing delay of the upstream node from the fault of the upstream node to the fault propagation time of the downstream node. Each fault node in the fault propagation chain is weighted according to the fault propagation time sequence, and the weighted fault chain identification rate is calculated. Based on the comparative analysis of the timing of upstream node failure and the timing of downstream node failure, the delayed intervention time and the actual time available for intervention are obtained, and the intervention window miss rate is calculated proportionally. For any fault propagation chain, the weighted fault chain unidentification rate U ω For: U ω =1-C ω C ω The weighted fault chain identification rate; The weighted failure chain unidentification rate and the intervention window miss rate are added together to obtain the perception lag index. The perception lag index is then compared with the preset lag index. If the perception lag index is greater than the preset lag index, the historical chain failure corresponding to the failure propagation chain is a perception lag failure.

3. The active sensing method for a new energy power station cluster according to claim 2, characterized in that: The weighted fault chain identification rate is calculated as follows: Number each node in the fault propagation chain according to the fault propagation sequence, with the source node being level 1, the first downstream node being level 2, and so on. Assign weight to each node according to its number, with the formula: Node weight = Total number of nodes - Node number + 1. The weighted fault chain identification rate is calculated using the formula: Weighted fault chain identification rate = Sum of weights of identified nodes / Sum of weights of all nodes in the fault propagation chain.

4. The active sensing method for a new energy power station cluster according to claim 2, characterized in that: The method for obtaining the delayed intervention time and the actual interventionable time is as follows: If the upstream node failure occurs earlier than the downstream node failure trigger time, calculate the lag intervention time. The formula is: , among which, T down T is the time when the downstream node failure is triggered. up T represents the time when the upstream node failure occurs. min Minimum execution time for intervention; The method for determining the minimum intervention execution time is based on the technical parameters of the control equipment and statistical historical data; like Then the delayed intervention time for ; If the upstream node failure occurs later than the downstream node failure trigger, the perception lag will cause the entire intervention window to be missed. In this case, the delayed intervention time is the difference between the upstream failure occurrence time and the downstream failure trigger time. The actual intervention time is the period between the downstream fault triggering time and the upstream fault occurrence time.

5. The active sensing method for a new energy power station cluster according to claim 1, characterized in that: The method for determining whether the cluster sensing system has a fault detection lag is as follows: The proportion of perceived lag faults is obtained by statistically analyzing the proportion of perceived lag faults in historical cascading failures. The percentage of perceived lag faults is compared with the preset percentage. If the percentage of perceived lag faults is greater than the preset percentage, then the cluster perception system has a fault perception lag phenomenon.

6. The active sensing method for a new energy power station cluster according to claim 1, characterized in that: The process for determining the baseline fault propagation time is as follows: Filter out the data of perceived lag faults from the historical cascading fault data, and extract the fault propagation chain corresponding to the perceived lag faults. Based on any fault propagation chain: For each pair of adjacent faulty nodes: Extract the fault propagation time of the adjacent fault node from all fault propagation chains that contain the adjacent fault node; Calculate the coefficient of variation of the fault propagation time of the adjacent fault node in each fault propagation chain, and compare the coefficient of variation with the preset coefficient of variation. If the coefficient of variation is less than the preset coefficient of variation, the baseline propagation time is the average fault propagation time; otherwise, the minimum value is taken.

7. The active sensing method for a new energy power station cluster according to claim 1, characterized in that: The calculation method for the perception time of each fault node in the fault propagation database is as follows: Based on the physical laws of power systems, theoretical calculations are performed on the components of sensing time: The data transmission delay is calculated based on the signal propagation speed and transmission path parameters, using the formula: T 传输 = Line length / Signal propagation speed + Data frame size / Link bandwidth; The decision delay is calculated and analyzed based on the algorithm's processing capabilities, and the formula is: T 分析 =1 / sampling frequency × number of algorithm processing cycles; The theoretical perception time is T 基础 =T 传输 +T 分析 ; By filtering out faults with the same fault type and physical parameters as the current fault from the perceived lag fault data, a historical sample set is obtained: {(T base,i T act,i Let T be the number of faults in the historical sample set, where T = 1, 2, ..., n. base,i Let T be the theoretical sensing time for the i-th sensing lag fault. act,i The actual sensing time for the i-th sensing lag fault; For each sensing lag fault in the historical sample set, calculate the deviation rate δ between the theoretical sensing time and the actual sensing time, and calculate the mean of the sensing time deviation rates for all sensing lag faults in the historical sample set to obtain the average deviation rate δ. 平 ; The correction factor K is K=1+δ 平 ; For each fault node in the fault propagation database, the sensing time is obtained by multiplying the theoretical sensing time by the correction coefficient.

8. The active sensing method for a new energy power station cluster according to claim 1, characterized in that: The matching method for the target fault propagation chain is as follows: Filter all fault propagation chains that contain the current fault node from the fault propagation library to form a candidate chain set; For each fault propagation chain in the candidate chain set: If the current faulty node is the source node, then each fault propagation chain is a target fault propagation chain; Conversely, determine the level of the current faulty node in the candidate chain, trace back to the upstream nodes, filter out all faulty nodes in the candidate chain from the source node to the current faulty node, and count the number of faulty nodes until the source node. Calculate the ratio of the number of faulty nodes to the total number of nodes in the fault propagation chain to obtain the proportion of matching nodes. The proportion of matched nodes is compared with the preset proportion of matched nodes. If the proportion of matched nodes is greater than the preset proportion of matched nodes, then the fault propagation chain is the target fault propagation chain.

9. The active sensing method for a cluster of new energy power stations according to claim 1, characterized in that: The method for determining whether there is a risk of perception lag in the current fault and implementing corresponding adjustments is as follows: For any target fault propagation chain: Compare the perception time T of the current faulty node 感知 And the baseline fault propagation time T between the current faulty node and downstream nodes 传播 ; If T 感知 >T 传播 If the current faulty node is detected and identified later than the downstream node, there is a risk of detection lag. The target fault propagation chain is arranged in descending order of perception lag index to obtain the control priority of the target fault propagation chain.

10. A new energy power station cluster active sensing system, the system being used to execute a new energy power station cluster active sensing method as described in any one of claims 1-9, characterized in that: Includes the following modules: Fault perception lag judgment module: By comparing and analyzing the fault propagation time and perception delay time of fault nodes in historical cascading faults, it identifies faults with perception lag in historical cascading faults, and judges whether there is a fault perception lag phenomenon in the cluster perception system through frequency analysis. Fault propagation library construction module: If it exists, perform stability analysis on the fault propagation time of adjacent fault nodes in the fault propagation chain, determine the baseline fault propagation time of adjacent fault nodes, integrate each fault propagation chain and the baseline fault propagation time of adjacent fault nodes, and generate a fault propagation library. Perception Time Calculation Module: Calculates the theoretical perception time of the fault node based on the physical parameters of the fault node and the perception node in the fault propagation database, calculates the correction coefficient based on the historical perception time of the fault node, and calculates the perception time of each fault node in the fault propagation database by combining the theoretical perception time and the correction coefficient. Target Fault Propagation Chain Identification Module: Identifies the target fault propagation chain by matching the current fault characteristics with the fault propagation database; Perception Lag Risk Judgment and Control Module: Compares and analyzes the baseline fault propagation time and perception time of the current fault node and downstream nodes in the target fault propagation chain to determine whether there is a perception lag risk in the current fault. If so, it determines the control priority of the target fault propagation chain based on the perception lag index and performs hierarchical control.

Images