An intelligent substation auxiliary system comprehensive monitoring platform

Abstract

The application discloses a kind of intelligent substation auxiliary system comprehensive monitoring platform, the application relates to the technical field of intelligent substation, the application includes the division module of perception domain, abnormal risk assessment module, abnormal processing module;The division module of perception domain is used to collect equipment historical failure data and divide equipment risk level, constructs the dynamic perception domain containing intersection area;Abnormal risk assessment module, with abnormal parameter calculation risk entropy value and diffusion index are obtained abnormal grade with equipment as node construction dependence atlas;Abnormal processing module, based on abnormal grade construction three-level resource pool adaptation processing demand.The application carries out to operation and maintenance resource configuration, improves abnormal response efficiency and reliability.

Description

Technical Field

[0001] This invention relates to the field of intelligent substation technology, specifically to an integrated monitoring platform for intelligent substation auxiliary systems. Background Technology

[0002] The integrated monitoring platform for intelligent substation auxiliary systems is a core support for ensuring the safe and stable operation of substations. However, existing technologies have the following shortcomings in practical applications:

[0003] First, existing solutions mostly rely on subjective human weighting or only preliminarily classify by equipment type, without constructing an objective classification model that takes into account the duration of the fault, economic loss, and the scope of its spread. This results in insufficient monitoring resources for high-risk equipment and waste of resources for low-risk equipment.

[0004] Secondly, existing sensing domains are mostly divided according to fixed ranges, without considering the correlation between devices, and no cross-level intersection area is set. When the sensing domain needs to be adjusted due to the upgrade of equipment risk, sensors need to be redeployed, which not only increases hardware costs, but also easily leads to data acquisition interruptions.

[0005] Third, existing assessments are mostly based on a single abnormal parameter and do not construct equipment dependency maps to quantify the correlation of fault spread, resulting in inaccurate classification of anomaly levels and difficulty in predicting the scope and severity of fault spread;

[0006] Therefore, there is an urgent need for a comprehensive monitoring platform for intelligent substation auxiliary systems. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a comprehensive monitoring platform for intelligent substation auxiliary systems, which solves the problems of single anomaly assessment, static sensing domain, and resource waste in existing monitoring platforms.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a comprehensive monitoring platform for intelligent substation auxiliary systems, comprising:

[0009] The sensing domain segmentation module collects historical fault data of all equipment in the substation, classifies the risk level of the equipment, and constructs dynamic sensing domains for equipment with different risk levels.

[0010] The abnormal risk assessment module uses substation equipment as nodes to construct an equipment dependency graph. For the collected abnormal parameters of the equipment, it calculates the corresponding abnormal risk entropy value. Combining the equipment dependency graph and the abnormal risk entropy value, it calculates the risk diffusion index and classifies the abnormal level of the equipment according to the index.

[0011] The anomaly handling module constructs a three-tiered processing resource pool based on the anomaly level of the equipment. Specifically, it includes: a first-tier resource pool serving first-tier anomalies, comprising the substation core processing node, backup power switching module, and emergency cut-off actuator, with 100% redundant resources reserved, and the CPU load of the core processing node ≤30%; a second-tier resource pool serving second-tier anomalies, comprising critical processing nodes and parameter adjustment module, with 50% redundant resources reserved, and the CPU load of the critical processing node ≤70%; and a third-tier resource pool serving third-tier anomalies, only calling edge basic nodes and log recording module, requiring no redundant resources, and the CPU load of the edge basic nodes ≤70%.

[0012] As a further aspect of the present invention, the specific steps for classifying equipment risk levels are as follows:

[0013] Four core indicators were collected from historical fault data of various equipment in the substation, including fault duration t, direct economic loss from the fault L, number of faults triggered by related equipment N, and maximum fluctuation of operating parameters △P in the 24 hours before the fault.

[0014] For each independent fault, construct a feature vector F=[t,L,N,△P]. If a certain equipment has ≤2 fault samples, then supplement the fault samples of the same model and the same operating conditions from the historical database.

[0015] Using the generated feature vector F as input, the DBSCAN algorithm is employed, and sample similarity is calculated using Euclidean distance to obtain preliminary clusters;

[0016] Retrieve the time series curves of the operating parameters of each sample in the cluster 24 hours before the failure, and calculate the dynamic time warping (DTW) distance between it and the typical time series sequence in the cluster. If the DTW distance of a sample is greater than the distance threshold, it is assigned to the transition cluster. The typical time series sequence in the cluster is the sample sequence that has the smallest sum of DTW distances with all other samples in the cluster.

[0017] Calculate the average economic loss of all clusters And calculate the cumulative distribution function (CDF), and automatically classify the cluster levels according to the CDF;

[0018] All fault samples are categorized according to the equipment to which they belong, and the total number of fault samples for each equipment and the number of samples for each risk level are counted. For each equipment, the percentage of each risk level is calculated according to the formula "Percentage of a certain level = Number of samples of that level / Total number of fault samples of the equipment × 100%". The risk level with the largest percentage is selected as the final risk level of the equipment. If multiple levels have the same percentage, the higher level is selected according to the principle of strict risk.

[0019] As a further aspect of the present invention, automatically classifying cluster levels according to CDF specifically includes: classifying all clusters... Sort by largest to smallest, and calculate each one in turn. The corresponding cumulative distribution function (CDF) value is used to classify the cluster as high-risk if CDF > CDFmax, medium-risk if CDFmin ≤ CDF ≤ CDFmax, and low-risk if CDF < CDFmin. CDFmin and CDFmax are the lower and upper limits of the threshold, respectively, and both need to be set according to the actual situation.

[0020] As a further aspect of the present invention, the specific steps for constructing a dynamic sensing domain for devices with different risk levels are as follows:

[0021] Centered on the physical body of the high-risk equipment, a core sensing domain with a radius of R1 is defined. R1 needs to cover the equipment body and all directly related key components, and also include the equipment connection link band with a width of W0.

[0022] In the edge region of the core sensing domain, an annular intersection zone is defined, and it is determined that this region must overlap with the sensing domain of the medium-risk equipment. The sensors within the intersection zone must support cross-domain data sharing. The edge region is located from R1-W1 to R1 of the high-risk equipment body, where W1 is the width of the intersection zone.

[0023] Centered on the medium-risk equipment, a basic sensing domain with a radius of R2 is defined. The boundary of the basic sensing domain near the high-risk equipment is extended outward by △R, so that the intersection reserved area of ​​the basic sensing domain and the core sensing domain completely overlaps. The overlapping area must include the connection link between the medium-risk equipment and the high-risk equipment. Here, △R is the extension width, and R2 < R1.

[0024] On the side of the basic sensing domain far from high-risk devices, an intersection area with a width of W2 is defined to overlap with the sensing domain of low-risk devices. This area needs to be deployed with shared sensors that support low-power data transmission, where W2 < W1.

[0025] Taking the functional areas of the substation as units, a rectangular gridded sensing domain with a size of A×B is defined to cover all low-risk equipment within the functional areas. The functional areas refer to the auxiliary areas within the substation that are not core power production but support operation and maintenance, safety, and environment. Here, A is the length and B is the width.

[0026] Extend the rectangular gridded sensing domain outward by △R1 from the boundary closest to the medium-risk equipment to ensure that the extended boundary can cover the W2 intersection area of ​​the medium-risk equipment. If the functional area is adjacent to a high-risk equipment, it is necessary to additionally ensure that the extended boundary covers the edge of the W1 intersection zone of the core sensing domain.

[0027] In the intersection area of ​​the rectangular gridded sensing domain and the core sensing domain, a standardized data interface is preset.

[0028] As a further aspect of the present invention, a fiber optic grating sensor, a partial discharge sensor, a current transformer, and a temperature and humidity sensor are deployed in the core sensing domain to collect vibration frequency, vibration acceleration, maximum amplitude of partial discharge, effective current value, temperature, and humidity in real time; an intelligent current transformer and a temperature and humidity sensor are deployed in the basic sensing domain to collect effective current value, temperature, and humidity in real time; and a temperature and humidity sensor is deployed in the rectangular gridded sensing domain to collect temperature and humidity in real time.

[0029] As a further aspect of the present invention, if a certain sensing domain experiences a similar anomaly exceeding a specified threshold within a cycle, the sensing domain is upgraded by one level. If the sensing domain is a core sensing domain, no adjustment is made. After adjusting the sensing domain level, the required corresponding data is dynamically obtained. The levels of the three sensing domains are: core sensing domain > basic sensing domain > rectangular gridded sensing domain.

[0030] As a further aspect of the present invention, the specific steps for dynamically obtaining the required corresponding data are as follows:

[0031] Based on the intersection of the sensing domains at each level, the core parameters that can be shared are identified: the core sensing domain opens up "vibration frequency, vibration acceleration, and maximum amplitude of partial discharge" to the basic sensing domain; the basic sensing domain opens up "effective value of current" to the rectangular gridded sensing domain, forming a two-way shared list.

[0032] Deploy data relay nodes in the intersection area, unify data format and transmission protocol, and compare the required parameters of the sensing domain after the adjustment level with the existing parameters of the current sensing domain;

[0033] The system automatically sends data request to the intersection area of ​​the core sensing domain and obtains missing parameters in real time, while pausing the acquisition of low-priority data.

[0034] The retrieved data from the intersection area are timestamped and spatially mapped to ensure that the shared data meets the monitoring requirements of the high-level sensing domain.

[0035] As a further aspect of the present invention, the specific steps for constructing a device dependency map are as follows:

[0036] The device dependencies are divided into three types of directed edges: power supply dependent edges, control dependent edges, and cooling dependent edges. The physical correspondence of each type of edge is clearly defined: power supply dependent edges correspond to power transmission paths, control dependent edges correspond to command transmission paths, and cooling dependent edges correspond to cooling medium / equipment supply paths. The physical location from the starting device to the ending device is marked for each type of edge.

[0037] Retrieve historical fault records from the substation and calculate the strength W of each dependent edge using the formula. AB That is, W AB= The number of times in history that a failure of device A directly caused a failure of device B / the total number of failures of device B. If there is no failure record for device B, refer to the industry failure data of the same model of device.

[0038] The graph is constructed using a node-edge-strength structure. New fault data is automatically counted monthly, and the strength of related dependent edges is recalculated. When equipment is decommissioned or added, the corresponding nodes and dependent edges are automatically deleted or added.

[0039] As a further aspect of the present invention, the specific operation for calculating the abnormal risk entropy value corresponding to the device is as follows:

[0040] For the collected abnormal parameters, three core abnormal features are extracted, and K-means clustering is used to divide the feature intervals, calculating the probability density of each interval:

[0041] The three core abnormal characteristics include abnormal fluctuation amplitude, abnormal duration, and historical frequency of occurrence. By retrieving historical data of similar abnormalities and calculating the sample proportion of each interval, the probability density of that interval can be obtained.

[0042] Substituting the probability densities of the three core anomaly features into the entropy formula... ,in, This represents the probability density of the abnormal fluctuation range. This represents the probability density of the interval of abnormal duration. The probability density of the frequency interval of historical occurrences;

[0043] If at least one abnormal parameter is collected, the average entropy value corresponding to each abnormal parameter is calculated, and this average value is used as the abnormal risk entropy value of the device.

[0044] As a further aspect of the present invention, the specific steps for calculating the risk diffusion index are as follows:

[0045] Starting with the abnormal source device, traverse all dependency edges in the device dependency graph that start with the abnormal source device, filter out associated devices with a dependency strength greater than a specified threshold, and count the number of devices N1.

[0046] Calculate the risk diffusion index D=H×N1. If D is greater than the historical maximum diffusion index of the abnormal source device, then re-examine the screening results of highly dependent devices.

[0047] Based on the correspondence between the historical anomaly index of substations and economic losses, a two-level threshold is set: if D > Dmax, the equipment is classified as a Level 1 anomaly; if Dmin ≤ D ≤ Dmax, the equipment is classified as a Level 2 anomaly; if D < Dmin, the equipment is classified as a Level 3 anomaly. The priority of the three levels of anomalies is: Level 1 anomaly > Level 2 anomaly > Level 3 anomaly. Among them, Dmin and Dmax are the lower limit and upper limit of the risk diffusion index, respectively.

[0048] This invention provides a comprehensive monitoring platform for intelligent substation auxiliary systems, which has the following advantages compared with existing technologies:

[0049] (1) This invention collects historical data from multiple dimensions, such as fault duration, economic loss, and number of associated faults, and combines DBSCAN clustering, DTW time series verification and CDF quantile to classify equipment risk levels, thus avoiding the defects of existing technologies that rely on subjective human judgment or single-dimensional classification.

[0050] (2) This invention constructs a core, basic and rectangular gridded sensing domain with intersection area for equipment with different risk levels, which overcomes the problems of high hardware cost and data acquisition interruption when adjusting the existing static isolated sensing domain, and improves the system's adaptability to changes in equipment risk.

[0051] (3) The present invention constructs a three-level processing resource pool according to the anomaly level. The first-level resource pool has high redundancy to ensure rapid response to high-risk anomalies. The second-level resource pool has medium redundancy to prevent anomaly escalation. The third-level resource pool has lightweight configuration to avoid waste. It can reduce the processing cost of low-level anomalies while ensuring reliable processing of high-level anomalies. Attached Figure Description

[0052] Figure 1 This is the system principle block diagram of the present invention;

[0053] Figure 2 The flowchart illustrates the steps involved in dividing the dynamic sensing domain according to this invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] Example 1

[0056] like Figure 1 This invention provides a comprehensive monitoring platform for intelligent substation auxiliary systems, comprising:

[0057] The sensing domain segmentation module collects historical fault data of all equipment in the substation and classifies the risk level of the equipment.

[0058] Because the risks of equipment within a substation vary greatly: failures of equipment such as main transformers may cause a power outage for the entire station, while failures of equipment such as lighting have a limited impact on the entire station. If the equipment is not divided, it will lead to a misallocation of operation and maintenance resources. After division, resources can be allocated according to risk priority, and at the same time, it provides a basis for the configuration of the sensing domain and data collection below. More core parameters are collected in the area where high-risk equipment is located, while the collection is simplified in the area where low-risk equipment is located.

[0059] The specific operational steps for classifying the risk level of the equipment are as follows:

[0060] Four core indicators were collected from historical fault data of various equipment in the substation, including fault duration t, direct economic loss from the fault L, number of faults triggered by related equipment N, and maximum fluctuation of operating parameters △P in the 24 hours before the fault.

[0061] The duration t of the fault directly determines the duration of the fault's impact on the substation system. The longer the duration, the more severe the equipment damage, and the greater the impact on the reliability of the power grid.

[0062] The direct economic loss L from a fault is the most direct variable representing the risk consequences in substation operation and maintenance. It includes equipment maintenance costs, power outage losses, and user compensation costs. It directly reflects the actual degree of harm caused by the fault and is a common quantifiable indicator for all equipment faults.

[0063] Substation equipment exhibits strong correlation. The number N of faults triggered by related equipment directly reflects the scope of fault spread. The larger N is, the more the fault has spread from a single device to the related system, potentially causing a systemic failure.

[0064] The maximum fluctuation range of operating parameters △P in the 24 hours before the failure reflects the strength of the warning signal before the failure. Equipment failure does not happen suddenly, but there are usually abnormal fluctuations in parameters. The larger △P is, the more obvious the abnormal signs before the failure are, and the higher the suddenness and severity of the risk.

[0065] For each independent fault, a feature vector F=[t,L,N,△P] is constructed. Multiple fault data of the same equipment are not merged. If the fault samples of a certain equipment are ≤2, the fault samples of the same model and the same operating conditions are supplemented from the historical database.

[0066] Using the generated feature vector F as input, the DBSCAN algorithm is employed, and the sample similarity is calculated through Euclidean distance (the smaller the distance, the closer F is, and the more similar the risk features are) to obtain a preliminary cluster.

[0067] DBSCAN can automatically cluster data based on data density without pre-setting the number of clusters, and can naturally form clusters of different risk levels, which is more in line with the continuous characteristics of substation fault risk.

[0068] Retrieve the time series curves of the operating parameters of each sample in the cluster 24 hours before the failure, and calculate the dynamic time warping (DTW) distance between it and the typical time series sequence in the cluster (to measure the similarity of the time series curve shape). If the DTW distance of a sample is greater than the distance threshold, it is assigned to the transition cluster.

[0069] The typical time series sequence in the cluster is the sample sequence that has the smallest sum of DTW distances to all other samples in the cluster;

[0070] Calculate the average economic loss of all clusters It also calculates the cumulative distribution function (CDF) and automatically classifies clusters into different levels based on the CDF.

[0071] All clusters Sort by largest to smallest, and calculate each one in turn. The corresponding cumulative distribution function (CDF) value is used to classify the cluster into a high-risk level if CDF > CDFmax; a medium-risk level if CDFmin ≤ CDF ≤ CDFmax; and a low-risk level if CDF < CDFmin. CDFmin and CDFmax are the lower and upper limits of the threshold, respectively, and both need to be set according to the actual situation.

[0072] All fault samples are categorized according to the equipment to which they belong, and the total number of fault samples for each equipment and the number of samples for each risk level are counted. For each equipment, the proportion of each risk level is calculated according to the formula "Proportion of a certain level = Number of samples of that level / Total number of fault samples of the equipment × 100%", and the risk level with the largest proportion is selected as the final risk level of the equipment.

[0073] If multiple levels have the same percentage, the higher level shall be selected according to the principle of stricter risk.

[0074] The system divides the dynamic sensing domain based on devices with different risk levels, and after adjusting the sensing domain level, dynamically obtains the required data. The specific operation is as follows:

[0075] Based on the intersection of the sensing domains at each level, the core parameters that can be shared are identified: the core sensing domain opens up "vibration frequency, vibration acceleration, and maximum amplitude of partial discharge" to the basic sensing domain; the basic sensing domain opens up "effective value of current" to the rectangular gridded sensing domain, forming a two-way shared list.

[0076] Deploy data relay nodes in the intersection area to unify data formats and transmission protocols, ensuring that sensor data from different levels of domains can be directly accessed;

[0077] Comparing the required parameters of the sensing domain after the adjustment level with the existing parameters of the current sensing domain, for example, when the basic sensing domain is upgraded to the core sensing domain, it is necessary to supplement "vibration frequency, vibration acceleration, and maximum amplitude of partial discharge". By searching the data list of the intersection area, it was found that these parameters have been collected in the intersection area of ​​the core sensing domain.

[0078] The system automatically sends data request to the intersection area of ​​the core sensing domain and obtains missing parameters in real time. At the same time, it suspends the collection of low-priority data, such as the original humidity in the basic sensing domain, to avoid resource conflicts.

[0079] The retrieved data from the intersection area are calibrated with timestamps (e.g., if the data acquisition time of the core sensing domain is 50ms earlier than that of the basic sensing domain, the time difference can be made up by interpolation algorithm) and spatially mapped (e.g., if the sensor in the core sensing domain is 3 meters away from the medium-risk equipment, the temperature value can be corrected by distance attenuation model to eliminate spatial error) to ensure that the shared data meets the monitoring requirements of the high-level sensing domain.

[0080] The abnormal risk assessment module uses substation equipment as nodes to construct an equipment dependency graph. Based on the collected real-time abnormal parameters, it calculates the abnormal risk entropy value corresponding to the equipment. Combining the equipment dependency graph and the abnormal risk entropy value, it calculates the risk diffusion index and classifies the abnormal level of the equipment according to the index.

[0081] The specific steps for constructing a device dependency graph are as follows:

[0082] Device dependencies are categorized into three types of directed edges: power supply dependencies, control dependencies, and cooling dependencies. The physical correspondence of each type of edge is then clearly defined.

[0083] The power supply depends on the power transmission path, such as "110kV line → 110kV switchgear → low-voltage side of main transformer". A line fault will cause the switchgear to lose power, which will then affect the main transformer.

[0084] Control-dependent edges correspond to instruction transmission paths, for example, "PLC controller ⇒ main transformer tap changer". A PLC failure will cause the tap changer to be unable to adjust.

[0085] Cooling-dependent edge corresponds to the supply path of cooling medium / equipment, for example, "main transformer cooler ⇨ main transformer body". Cooler failure will cause the main transformer oil temperature to rise.

[0086] Each type of edge is labeled with the physical location from the starting device to the ending device to avoid ambiguous logical relationships;

[0087] Retrieve historical fault records from the substation and calculate the strength W of each dependent edge using the formula. AB That is, W AB= The number of times in history that a failure of device A directly caused a failure of device B / the total number of failures of device B. If device B has no failure records (such as new equipment), then refer to the industry failure data of the same model of equipment.

[0088] The graph is constructed using a node-edge-strength structure (devices are nodes, category-dependent edges are connections, and strength is an edge attribute). New fault data is automatically counted monthly, and the strength of related dependent edges is recalculated. When a device is decommissioned or added, the corresponding node and dependent edge are automatically deleted or added.

[0089] The specific steps to calculate the abnormal risk entropy value corresponding to this device are as follows:

[0090] For the collected abnormal parameters (such as excessive real-time current), three core abnormal features are extracted, and K-means clustering is used to divide the feature intervals, calculating the probability density of each interval:

[0091] The three core abnormal characteristics include abnormal fluctuation amplitude, abnormal duration, and historical frequency of occurrence. By retrieving historical data of similar abnormalities (such as current exceeding the standard) and calculating the sample proportion of each interval, the probability density of that interval is obtained.

[0092] Substituting the probability densities of the three core anomaly features into the entropy formula... ,in, This represents the probability density of the abnormal fluctuation range. This represents the probability density of the interval of abnormal duration. The probability density of the frequency interval of historical occurrences;

[0093] If there is more than one abnormal parameter collected, the average entropy value corresponding to each abnormal parameter is calculated, and the average value is used as the abnormal risk entropy value of the device.

[0094] By combining the equipment dependency map and the anomaly risk entropy value, the risk diffusion index is calculated. The specific steps are as follows:

[0095] Starting with the abnormal source device, traverse all dependency edges in the device dependency graph that start with the abnormal source device, filter out associated devices with a dependency strength greater than a specified threshold, and count the number of devices N1.

[0096] Calculate the risk diffusion index D=H×N1. If D is greater than the historical maximum diffusion index of the abnormal source device, re-examine the screening results of highly dependent devices to ensure that the index does not fluctuate abnormally.

[0097] Based on the correlation between historical anomaly indices of substations and economic losses, a secondary threshold is set:

[0098] If D > Dmax, the device is classified as a Level 1 anomaly; if Dmin ≤ D ≤ Dmax, the device is classified as a Level 2 anomaly; if D < Dmin, the device is classified as a Level 3 anomaly. The priority of the three levels of anomalies is: Level 1 anomaly > Level 2 anomaly > Level 3 anomaly. Dmin and Dmax are the lower and upper limits of the risk diffusion index, respectively.

[0099] The exception handling module constructs a three-tiered processing resource pool based on different exception levels, specifically including:

[0100] The primary resource pool serves primary anomalies and includes substation core processing nodes, backup power switching modules, and emergency disconnect actuators, reserving 100% redundant resources. The CPU load of the core processing nodes is ≤30%.

[0101] Level 1 anomaly handling requires real-time calculation of isolation paths and issuance of cut-off commands. High load (such as CPU ≥ 70%) will cause command delays. At the same time, this level of anomaly is accompanied by "fault source power supply link failure". Backup power switching can ensure that the core control equipment is powered continuously, while the emergency cut-off actuator can directly physically isolate the fault source and avoid relying on manual operation.

[0102] Level 1 anomaly handling cannot have any single point of failure. If only the primary cluster is used, once the primary node goes down, the interruption of the processing flow will cause the fault to spread. 100% redundancy can ensure that "when the primary node fails, the backup node takes over within a specified time period", ensuring that the processing is not interrupted.

[0103] The secondary resource pool serves secondary anomalies and includes critical processing nodes and parameter adjustment modules, reserving 50% redundant resources. The CPU load of the critical processing nodes is ≤70%.

[0104] Level 2 anomaly handling mainly involves parameter adjustment. Unlike Level 1, it does not require high-frequency calculation of isolation paths. 50% load is sufficient to meet the requirements. The root cause of this type of anomaly is often that the operating parameters exceed the threshold, rather than a physical fault. Adjusting the parameters directly can restore normal operation without shutting down the equipment.

[0105] While 100% redundancy is not required for Level 2 anomalies, it is necessary to prevent processing interruptions that could lead to upgrades. If the secondary cluster has no backup, node failures can cause delays in parameter adjustments. 50% redundancy can ensure that "backup nodes take over within a specified time," thus avoiding upgrade risks.

[0106] The three-tier resource pool serves three-tier anomalies, only calling edge basic nodes and the logging module, requiring no redundant resources, and the CPU load of the edge basic nodes is ≤70%;

[0107] Level 3 anomaly handling only requires logging and lightweight alerts, without the need for real-time calculations or adjustments. The computing power of edge nodes is sufficient to support this. If core or critical nodes are used, they will consume computing resources for high-priority anomalies, leading to resource mismatch.

[0108] Level 3 anomaly handling failure has no impact. Even if the edge node fails and the logs are not recorded, the operation and maintenance personnel can still find it during inspection. It will not lead to the spread of the fault or escalation, while configuring redundant resources will cause the resources to be idle.

[0109] Example 2

[0110] This embodiment further discloses a method for dividing the dynamic sensing domain based on devices with different risk levels, building upon Embodiment 1. Figure 2 As shown, the specific content includes:

[0111] Centered on the physical body of the high-risk equipment, a core sensing domain with a radius of R1 is defined. R1 must cover the equipment body and all directly related key components, such as the cooling system and high-voltage side switch, and also include the equipment connection link band with a width of W0 (covering the connection path between the equipment and the protection device and the power supply).

[0112] In the edge region of the core sensing domain, an annular intersection zone is defined, and it is determined that this region must overlap with the sensing domain of the medium-risk equipment. The sensors within the intersection zone must support cross-domain data sharing. The edge region is located from R1-W1 to R1 of the high-risk equipment body, where W1 is the width of the intersection zone.

[0113] The ring intersection zone W1 is a dynamically adjusted resource buffer layer. When the sensing domain where the medium-risk equipment is located is upgraded in the future, the sensor resources in this area can be directly reused without the need to redeploy the hardware.

[0114] Centered on the medium-risk equipment, a basic sensing domain with a radius of R2 is defined. The boundary of the basic sensing domain near the high-risk equipment is extended outward by △R, so that the intersection reserved area (R1-W1 to R1) of the basic sensing domain and the core sensing domain completely overlaps. The overlapping area must include the connection link between the medium-risk equipment and the high-risk equipment. Here, △R is the extension width, and R2 < R1.

[0115] The extended design of △R forces the intersection of the basic sensing domain and the core sensing domain: if the boundary is defined only by R2, the intersection may be insufficient due to the device spacing problem. After the extension, it can ensure that the overlapping area is sufficient to cover the cross-device association link, providing a data connection basis for subsequent domain upgrades.

[0116] On the side of the basic sensing domain far from high-risk devices, an intersection area with a width of W2 is defined to overlap with the sensing domain of low-risk devices. This area needs to be deployed with shared sensors that support low-power data transmission, where W2 < W1.

[0117] Using the functional areas of a substation as units, a rectangular gridded sensing domain of size A×B is defined to cover all low-risk equipment within the functional area. Here, A is the length and B is the width, which are adapted according to the actual size of the functional area. The functional area refers to the auxiliary area within the substation that is not core power production but supports operation and maintenance, safety, and environment. The equipment in these areas has unified functions (such as communication, lighting, and fire protection), is concentrated in location (such as dedicated rooms), and has low risk. This zoning avoids dividing each low-risk equipment into domains one by one, which simplifies the configuration of the sensing domain and facilitates the intersection with the core sensing domain.

[0118] Extend the rectangular gridded sensing domain outward by △R1 from the boundary closest to the medium-risk equipment to ensure that the extended boundary can cover the W2 intersection area of ​​the medium-risk equipment. If the functional area is adjacent to a high-risk equipment, it is necessary to additionally ensure that the extended boundary covers the edge of the W1 intersection zone of the core sensing domain.

[0119] In the intersection area of ​​the rectangular gridded sensing domain and the core sensing domain, a standardized data interface is preset, and historical data and real-time monitoring data of the core sensing domain can be directly retrieved through the interface during subsequent adjustments.

[0120] Fiber optic grating sensors, partial discharge sensors, current transformers, and temperature and humidity sensors are deployed in the core sensing domain to collect vibration frequency, vibration acceleration, maximum amplitude of partial discharge, effective current value, temperature, and humidity in real time; intelligent current transformers and temperature and humidity sensors are deployed in the basic sensing domain to collect effective current value, temperature, and humidity in real time; and temperature and humidity sensors are deployed in the rectangular gridded sensing domain to collect temperature and humidity in real time.

[0121] If a certain sensing domain experiences a similar anomaly exceeding a specified threshold within a cycle, the sensing domain is upgraded to the next level. If the sensing domain is a core sensing domain, no adjustment is made. The order of the three sensing domains is: core sensing domain > basic sensing domain > rectangular gridded sensing domain.

[0122] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0123] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. An intelligent substation auxiliary system integrated monitoring platform, characterized in that, Comprise: A perception domain division module collects historical fault data of all devices in the substation, divides the risk level of the devices, and constructs a dynamic perception domain for devices with different risk levels; An abnormal risk assessment module takes substation devices as nodes to construct a device dependency graph, calculates the abnormal risk entropy value of the collected device abnormal parameters, combines the device dependency graph and the abnormal risk entropy value to calculate the risk diffusion index, and divides the device abnormal level according to the index. The specific operation is: collect four types of core indicators in the historical fault data of each device in the substation, including fault duration t, fault direct economic loss L, number of associated device triggered faults N, and maximum operating parameter fluctuation amplitude △P within 24 hours before the fault; construct a feature vector F = [t, L, N, △P] for each independent fault. If the number of fault samples of a device is ≤2, supplement the fault samples of the same type and working condition device from the historical database; With the generated feature vector F as input, DBSCAN algorithm is adopted, and the similarity of samples is calculated by Euclidean distance to obtain preliminary clusters; the time series curve of each sample in the cluster within 24 hours before failure is called, and the dynamic time warping DTW distance between it and the typical time series sequence in the cluster is calculated. If the DTW distance of a sample is greater than the distance threshold, the sample is divided into a transition cluster. The typical time series sequence in the cluster is the sample sequence corresponding to the minimum sum of DTW distances with all other samples in the cluster The average economic loss of all clusters is calculated, and the cumulative distribution function CDF is calculated to automatically divide the cluster level; all fault samples are classified according to the equipment they belong to, and the total number of fault samples and the number of samples of each risk level of each equipment are counted. For each equipment, the proportion of each risk level is calculated according to the formula "proportion of a certain level = number of samples of the level / total number of fault samples of the equipment × 100%", and the risk level with the largest proportion is selected as the final risk level of the equipment. If there are multiple levels with the same proportion, the higher level is selected according to the principle of strict risk. An abnormal processing module constructs a three-level processing resource pool based on the abnormal level of the device, including: a first-level resource pool serving a first-level abnormality, containing a substation core processing node, a backup power supply switching module, and an emergency shutdown executor, reserving 100% redundant resources, and the CPU load of the core processing node is ≤30%; a second-level resource pool serving a second-level abnormality, containing a key processing node and a parameter adjustment module, reserving 50% redundant resources, and the CPU load of the key processing node is ≤70%; a third-level resource pool serving a third-level abnormality, only calling an edge basic node and a log recording module, without redundant resources, and the CPU load of the edge basic node is ≤70%. 2.The integrated monitoring platform of an intelligent substation auxiliary system according to claim 1, characterized in that, The automatic division of the cluster levels according to the CDF specifically includes: calculating the average economic loss of all clusters According to the descending order, the average economic loss of each cluster is calculated one by one The corresponding cumulative distribution function CDF value, if CDF>CDFmax, the cluster is divided into a high-risk level; if CDFmin≤CDF≤CDFmax, the cluster is divided into a medium-risk level; if CDF CDFmin and CDFmax are the lower limit and upper limit of the threshold, which need to be set according to the actual situation. 3.The integrated monitoring platform of an intelligent substation auxiliary system according to claim 1, characterized in that, The specific steps for constructing a dynamic perception domain for devices with different risk levels are: Take the physical ontology of a high-risk device as the center, and divide a core perception domain with a radius of R1. R1 needs to cover the device ontology and all directly associated key components, and also includes a device connection link belt with a width of W0; In the edge area of the core perception domain, an annular intersection belt is divided, which needs to be forced to overlap with the perception domain of a medium-risk device. The sensors in the intersection belt need to support cross-domain data sharing. The edge area is from R1-W1 to R1 of the high-risk device ontology, where W1 is the width of the intersection belt; Take the medium-risk device as the center, and divide a basic perception domain with a radius of R2. Expand the boundary of the basic perception domain close to the high-risk device side by △R, so that the intersection reserved area of the basic perception domain and the core perception domain completely overlaps. The overlapping area needs to include the connection link of the medium-risk device and the high-risk device, where △R is the expansion width, and R2 < R1; In the side of the basic perception domain away from the high-risk device, an intersection area with a width of W2 is divided, which is used to overlap with the perception domain of a low-risk device. The area needs to deploy shared sensors that support low-power data transmission, where W2 < W1; Divide a rectangular grid perception domain with size A x B in the functional area of the substation to cover all low-risk devices in the functional area. The functional area represents auxiliary areas in the substation that are not core power production but support operation and maintenance, safety, and environment. Where A is the length and B is the width. The boundary of the rectangular grid perception domain near the side of the medium-risk equipment is extended outward by △R1, ensuring that the extended boundary covers the W2 intersection area of the medium-risk equipment. If the functional area is adjacent to a high-risk equipment, the extended boundary must additionally cover the edge of the W1 intersection zone of the core perception domain; In the intersection area of the rectangular grid perception domain and the core perception domain, a standardized data interface is preset.

4. The integrated monitoring platform for an intelligent substation auxiliary system according to claim 3, characterized in that, In the core perception domain, fiber grating sensors, partial discharge sensors, current transformers, temperature and humidity sensors are deployed to collect vibration frequency, vibration acceleration, maximum partial discharge amplitude, current effective value, temperature and humidity in real time. In the basic perception domain, intelligent current transformers and temperature and humidity sensors are deployed to collect current effective value, temperature and humidity in real time. In the rectangular grid perception domain, temperature and humidity sensors are deployed to collect temperature and humidity in real time.

5. The integrated monitoring platform for intelligent substation auxiliary system according to claim 4, characterized in that, If a perception domain has more than a specified threshold of the same type of abnormality in a period, the perception domain is upgraded by one level. If the perception domain is a core perception domain, no adjustment is made. After adjusting the levels of the perception domains, the required corresponding data is dynamically obtained. The levels of the three perception domains are as follows: core perception domain > basic perception domain > rectangular grid perception domain. 6.The integrated monitoring platform of an intelligent substation auxiliary system according to claim 5, characterized in that, The specific steps for dynamically obtaining the required corresponding data are as follows: Based on the intersection area of each level of perception domain, the core parameters that can be shared are sorted out: the core perception domain opens "vibration frequency, vibration acceleration, maximum partial discharge amplitude" to the basic perception domain; the basic perception domain opens "current effective value" to the rectangular grid perception domain, forming a bidirectional sharing list; Data relay nodes are deployed in the intersection area to unify data formats and transmission protocols, and to compare and adjust the parameter requirements of the perception domains after adjustment with the existing parameters of the perception domains; The system automatically sends a data call request to the intersection area of the core perception domain and obtains the missing parameters in real time, while suspending the collection of low-priority data; The intersection area data is timestamped and spatially mapped to ensure that the shared data meet the monitoring requirements of high-level perception domains. 7.The integrated monitoring platform of an intelligent substation auxiliary system according to claim 1, characterized in that, The specific operation steps for constructing the device dependency graph are as follows: The device dependency is divided into three types of directed edges, including power supply dependency edges, control dependency edges, and cooling dependency edges. The physical correspondence of each type of edge is determined: the power supply dependency edge corresponds to the power transmission path, the control dependency edge corresponds to the command transmission path, and the cooling dependency edge corresponds to the cooling medium / device supply path. Each type of edge is labeled with the physical location from the starting device to the terminating device. Call the historical fault records of the substation, calculate the strength W of each dependent edge according to the formula AB , that is AB W = the number of times that the failure of device A directly leads to the failure of device B in history / the total number of failures of device B, if there is no failure record of device B, refer to the industry failure data of the same type of device; The graph is constructed using the node-edge-intensity structure. New fault data is automatically counted every month, and the intensity of the relevant dependency edges is recalculated. When a device is retired or added, the corresponding nodes and dependency edges are automatically deleted or added. 8.The integrated monitoring platform of an intelligent substation auxiliary system according to claim 1, characterized in that, The specific operation for calculating the abnormal risk entropy value corresponding to the device is as follows: For the collected abnormal parameters, three core abnormal features are extracted, and the feature intervals are divided using K-means clustering. The probability density of each interval is calculated: The three core abnormal features include abnormal fluctuation amplitude, abnormal duration, and historical occurrence frequency. Historical similar abnormal data is called to calculate the sample proportion of each interval, which is the probability density of the interval. The probability density of the three core abnormal features is substituted into the entropy value formula, i.e. wherein, is the probability density of the abnormal fluctuation amplitude interval, is the probability density of the abnormal duration interval, is the probability density of the historical occurrence frequency interval; If the collected abnormal parameters are not less than one, the average value of the entropy value corresponding to each abnormal parameter is calculated as the abnormal risk entropy value of the device. 9.The integrated monitoring platform of an intelligent substation auxiliary system according to claim 1, characterized in that, The specific steps for calculating the risk diffusion index are as follows: Starting from the abnormal source device, traverse all dependency edges in the device dependency graph that start from the abnormal source device, filter out associated devices with dependency strength greater than a specified threshold, and count the number of devices N1. Calculate the risk diffusion index D=H×N1, if D is greater than the historical maximum diffusion index of the abnormal source device, recheck the high dependency device filtering result; Based on the corresponding relationship between the historical abnormal index of the substation and the economic loss, set the secondary threshold: if D>Dmax, the device is classified as a first-level abnormality; if Dmin≤D≤Dmax, the device is classified as a second-level abnormality; if D<Dmin, the device is classified as a third-level abnormality. The priority of the three levels of abnormality is: first-level abnormality>second-level abnormality>third-level abnormality, wherein Dmin and Dmax are the lower limit and upper limit of the risk diffusion index.

Images