Abnormal result determination method and device of substation equipment and electronic equipment

Abstract

The application discloses a substation equipment abnormal result determination method and device and electronic equipment. The method comprises the following steps: acquiring sensing data corresponding to a plurality of detection positions of a substation equipment; determining type data weight values and dynamic alarm threshold values corresponding to the plurality of detection positions; determining state indexes corresponding to the plurality of detection positions according to the sensing data and the type data weight values corresponding to the plurality of detection positions; comparing the state indexes corresponding to the plurality of detection positions with the corresponding dynamic alarm threshold values to determine alarm event triggering results; inputting operation data corresponding to a plurality of alarm positions into corresponding position anomaly detection models to obtain abnormal results corresponding to the plurality of alarm positions; and determining a comprehensive abnormal result according to the abnormal results corresponding to the plurality of alarm positions. The application solves the technical problem of low detection result accuracy when determining whether the substation equipment is abnormal in the related art.

Description

Technical Field

[0001] This invention relates to the field of power equipment anomaly detection, and more specifically, to a method, apparatus, and electronic device for determining anomaly results of substation equipment. Background Technology

[0002] Currently, with the improvement of the intelligence level of substation equipment and the widespread application of online monitoring technology, the online monitoring technology for substations has various monitoring methods and uses a large number of sensors to collect operational data from multiple detection parts of the equipment. However, each sensing system is independent of each other, and alarm business management is difficult. Therefore, when determining whether substation equipment is abnormal, there is a technical problem of low accuracy of detection results.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This invention provides a method, apparatus, and electronic device for determining abnormal results of substation equipment, in order to at least solve the technical problem of low accuracy of detection results when determining whether substation equipment is abnormal in related technologies.

[0005] According to one aspect of the present invention, a method for determining abnormal results of substation equipment is provided, comprising: acquiring sensing data corresponding to multiple detection points of the substation equipment, wherein the sensing data includes raw sensor data, equipment runtime data, and real-time environmental data; determining type data weight values ​​and dynamic alarm thresholds corresponding to the multiple detection points, wherein the dynamic alarm thresholds are dynamically determined based on the equipment operation stage, real-time environmental data, and power supply requirements; and determining a state index corresponding to the multiple detection points based on the sensing data and type data weight values ​​corresponding to the multiple detection points, wherein the state indexes are expressed as... The method quantifies the degree of abnormal risk corresponding to the detection location; compares the state index corresponding to each of the multiple detection locations with the corresponding dynamic alarm threshold to determine the alarm event triggering result corresponding to each of the multiple detection locations; inputs the operating data corresponding to each of the multiple alarm locations into the corresponding location anomaly detection model to obtain the anomaly result corresponding to the multiple alarm locations, wherein the multiple alarm locations are the locations among the multiple detection locations whose corresponding alarm event triggering result is an alarm; based on the anomaly result corresponding to the multiple alarm locations, a comprehensive anomaly result is determined, wherein the comprehensive anomaly result includes anomaly prediction diffusion path and associated location risk probability.

[0006] Optionally, based on the anomaly results corresponding to the multiple alarm locations, a comprehensive anomaly result is determined, including: when the substation equipment includes multiple devices, determining the structural and functional dependencies between the multiple substation devices, and constructing a device association knowledge graph, wherein the device association knowledge graph includes target triples, the target triples include device elements, detection location elements, and dependency relationship elements, and the structural and functional dependencies include at least one of the following: electrical connection relationship, thermal conduction relationship, and functional coupling relationship; the comprehensive anomaly result is determined based on the anomaly results corresponding to the multiple alarm locations and the device association knowledge graph.

[0007] Optionally, determining a comprehensive anomaly result based on the anomaly results corresponding to the multiple alarm locations includes: determining location anomaly features corresponding to each of the multiple alarm locations based on the anomaly results corresponding to the multiple alarm locations; mapping the multiple location anomaly features to a unified feature space to obtain multiple unified anomaly features in the unified feature space; performing feature fusion on the multiple unified anomaly features to obtain anomaly fusion features; and determining the comprehensive anomaly result based on the anomaly fusion features.

[0008] Optionally, after determining the comprehensive anomaly result based on the anomaly results corresponding to the multiple alarm locations, the method further includes: if the comprehensive anomaly result includes associated risk locations, determining a first correlation between the sensing data corresponding to the multiple alarm locations and the anomaly result, determining a second correlation between the sensing data corresponding to non-alarm locations and non-anomaly results, and determining a third correlation between the sensing data corresponding to associated risk locations and the comprehensive anomaly result, wherein the associated risk locations are those among the multiple detection locations whose corresponding associated location risk probability is greater than a risk threshold; determining the weight ranking of the type data affecting the determination result based on the first correlation, the second correlation, and the third correlation; and updating the type data weight values ​​corresponding to the multiple detection locations based on the weight ranking.

[0009] Optionally, before inputting the operational data corresponding to multiple alarm locations into the corresponding location anomaly detection models to obtain the anomaly results corresponding to the multiple alarm locations, the method further includes: acquiring sample data, wherein the sample data includes historical anomaly results, corresponding processing feedback data, and subsequent equipment operation status data corresponding to multiple sample locations; and using a composite loss function to train the initial anomaly detection models for the corresponding locations based on the sample data to obtain the corresponding location sample detection models, wherein the composite loss function includes a perceptual similarity loss term and a weighted cross-entropy term, wherein the perceptual similarity loss term is used to determine the loss value between the predicted operational data and the actual operational data, and the weighted cross-entropy term is used to determine the loss value between the predicted anomaly probability and the actual anomaly label.

[0010] Optionally, based on the sensing data and type data weight values ​​corresponding to the multiple detection sites respectively, a state index corresponding to each of the multiple detection sites is determined, including: performing multi-dimensional preprocessing on the sensing data corresponding to the multiple detection sites respectively to obtain standardized data corresponding to each of the multiple detection sites, wherein the multi-dimensional preprocessing includes at least one of the following: removing abnormal data caused by sensor drift based on the standard deviation principle, unifying the data dimensions using the standard score standardization method, and aligning the time-series data collected across sensors; and determining the state index corresponding to each of the multiple detection sites based on the standardized data and type data weight values ​​corresponding to each of the multiple detection sites respectively.

[0011] Optionally, after comparing the state index corresponding to each of the multiple detection locations with the corresponding dynamic alarm threshold to determine the alarm event triggering result corresponding to each of the multiple detection locations, the method further includes: if the alarm event triggering result corresponding to the target detection location is an alarm, determining whether there is previous alarm data within a predetermined time period before the current alarm event; if there is no previous alarm data, generating a detection location abnormality entry corresponding to the target detection location in response to the current alarm event; and / or, if there is previous alarm data, updating the alarm count corresponding to the detection location abnormality entry corresponding to the target detection location.

[0012] According to one aspect of the present invention, an abnormal result determination device for substation equipment is provided, comprising: an acquisition module, configured to acquire sensing data corresponding to multiple detection points of the substation equipment, wherein the sensing data includes original sensor data, equipment runtime data, and real-time environmental data; a first determination module, configured to determine type data weight values ​​and dynamic alarm thresholds corresponding to the multiple detection points, wherein the dynamic alarm thresholds are dynamically determined based on the equipment operation stage, real-time environmental data, and power supply requirements; and a second determination module, configured to determine a state index corresponding to the multiple detection points based on the sensing data and type data weight values ​​corresponding to the multiple detection points, wherein the state index is expressed as a table. The system includes: a quantitative value representing the degree of abnormal risk at the corresponding detection site; a comparison module comparing the state index corresponding to each of the multiple detection sites with the corresponding dynamic alarm threshold to determine the alarm event triggering result corresponding to each of the multiple detection sites; a third determination module inputting the operating data corresponding to the multiple alarm sites into the corresponding site abnormality detection model to obtain the abnormal results corresponding to the multiple alarm sites, wherein the multiple alarm sites are the sites among the multiple detection sites whose corresponding alarm event triggering result is an alarm; and a fourth determination module determining a comprehensive abnormal result based on the abnormal results corresponding to the multiple alarm sites, wherein the comprehensive abnormal result includes an abnormality prediction diffusion path and the associated site risk probability.

[0013] According to one aspect of the present invention, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the substation equipment abnormality result determination method described in any of the preceding claims.

[0014] According to one aspect of the present invention, a computer-readable storage medium is provided, which, when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, enables the electronic device to perform the substation equipment abnormality determination method described in any of the preceding claims.

[0015] In this embodiment of the invention, sensing data corresponding to multiple detection points of substation equipment is acquired, including raw sensor data, equipment runtime data, and real-time environmental data. Type data weight values ​​and dynamic alarm thresholds corresponding to the multiple detection points are determined, wherein the dynamic alarm thresholds are dynamically determined based on the equipment operation stage, real-time environmental data, and power supply requirements. Based on the sensing data and type data weight values ​​corresponding to the multiple detection points, a state index corresponding to each detection point is determined, where the state index represents a quantified value of the abnormal risk level of the corresponding detection point. The state indices corresponding to the multiple detection points are compared with the corresponding dynamic alarm thresholds to determine the alarm event triggering results corresponding to the multiple detection points. The operating data corresponding to the multiple alarm points are input into the corresponding part anomaly detection model to obtain the anomaly results corresponding to the multiple alarm points, where the multiple alarm points are the parts among the multiple detection points whose corresponding alarm event triggering results are alarms. Based on the anomaly results corresponding to the multiple alarm points, a comprehensive anomaly result is determined, where the comprehensive anomaly result includes anomaly prediction propagation paths and associated part risk probabilities. By employing dynamic weight allocation and dynamic alarm threshold adaptation, along with multi-part independent detection and correlation analysis, this method first calculates the status index of each detected part based on raw sensor data, equipment runtime data, and real-time environmental data, combined with dynamic type data weight values. Then, alarm parts are screened using alarm thresholds dynamically determined by equipment operation stage, real-time environmental data, and power supply requirements. After refining the anomaly results through a part anomaly detection model, correlation analysis is used to predict the anomaly propagation path and associated part risks. This achieves the goal of accurately locating anomalies in individual parts and comprehensively controlling the associated risks of multiple parts, thereby improving the technical effect of enhancing the accuracy of substation equipment anomaly detection and the comprehensiveness of risk prediction. This solves the technical problem of low accuracy in detection results when determining whether substation equipment is abnormal in related technologies. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0017] Figure 1 This is a flowchart of a method for determining abnormal results of substation equipment according to an embodiment of the present invention;

[0018] Figure 2 This is a structural block diagram of a substation equipment alarm system provided by an optional embodiment of the present invention;

[0019] Figure 3 This is a flowchart illustrating a substation equipment alarm method provided by an optional embodiment of the present invention;

[0020] Figure 4 This is a structural block diagram of a substation equipment anomaly result determination device according to an embodiment of the present invention. Detailed Implementation

[0021] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0022] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0023] Example 1

[0024] According to an embodiment of the present invention, an embodiment of a method for determining abnormal results of substation equipment is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0025] Figure 1 This is a flowchart of a method for determining abnormal results of substation equipment according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0026] Step S102: Obtain sensing data corresponding to multiple detection points of the substation equipment, wherein the sensing data includes raw sensor data, equipment runtime data and real-time environmental data.

[0027] Substation equipment refers to various equipment within a substation that performs functions such as power conversion, transmission, and control, including main transformers, gas-insulated switchgear (GIS), and switchgear, and is the core carrier of power supply in the power system.

[0028] Among them, the detection location refers to the specific location on the substation equipment that is monitored by sensors, such as the windings of the main transformer, the busbars of the GIS, and the contacts of the switchgear. These are key nodes for sensing the status of the equipment.

[0029] Among them, sensing data refers to various data sets used to reflect the operating status of the detected parts, and is the basic data source for subsequent status assessment.

[0030] Among them, raw sensor data refers to physical or chemical quantity data directly collected by the sensor, such as temperature values ​​collected by temperature sensors and hydrocarbon concentrations collected by oil chromatography sensors. This is first-hand data that has not been processed.

[0031] Among them, equipment runtime data refers to the cumulative operating time of substation equipment since its commissioning, which is an important basis for judging the degree of equipment aging.

[0032] Real-time environmental data refers to real-time environmental parameters at the substation site, such as ambient temperature, humidity, and weather conditions, which are external factors affecting the operating status of equipment.

[0033] In this step, multiple sensors are used to collect raw sensor data, equipment runtime data, and real-time environmental data from multiple detection points of the substation equipment in real time, forming a comprehensive set of sensing data.

[0034] This step enables the acquisition of multi-dimensional basic data reflecting the equipment status, avoiding the limitations of a single data dimension, providing a comprehensive and accurate data source for subsequent status index calculations, ensuring the accuracy of subsequent anomaly risk assessments, and reducing misjudgments caused by missing data from the source.

[0035] Step S104: Determine the type data weight values ​​and dynamic alarm thresholds corresponding to multiple detection locations, wherein the dynamic alarm thresholds are dynamically determined based on the equipment operation stage, real-time environmental data, and power supply requirements.

[0036] Among them, the type data weight value refers to the weight coefficient assigned to different types of sensing data based on factors such as the importance of sensing data and sensor accuracy, which is used to quantify the degree of influence of different data on state assessment.

[0037] Among them, the dynamic alarm threshold refers to the alarm judgment standard that is dynamically adjusted according to the actual operating scenario of the equipment, rather than a fixed value. It is the core basis for judging whether the detection part has an alarm.

[0038] The equipment operation phase refers to different stages in the entire life cycle of substation equipment, including the break-in period after commissioning, the stable operation period, and the aging period. The operating characteristics of the equipment are different in different stages.

[0039] Among them, power supply guarantee requirements refer to power supply guarantee requirements formulated for specific scenarios (such as extreme weather), which are divided into different levels such as Level 1, Level 2, and Level 3.

[0040] In this step, the weight value of the type data is determined by combining factors such as the importance of the detection part and the accuracy of the sensor. At the same time, the alarm threshold of each detection part is dynamically adjusted and determined based on the equipment operation stage, real-time environmental data and power supply requirements.

[0041] This step allows for the reasonable allocation of type data weights, highlighting the impact of key data and avoiding data interference with evaluation results. Dynamic alarm thresholds adapt to different operating scenarios, solving the problem of false alarms or missed alarms that are prone to occur at different stages and in different environments when using fixed thresholds, thus improving the adaptability and accuracy of alarm judgment.

[0042] Step S106: Based on the sensing data and type data weight values ​​corresponding to the multiple detection sites, determine the state index corresponding to each of the multiple detection sites, where the state index represents the quantitative value of the abnormal risk level of the corresponding detection site.

[0043] The status index, calculated using the weighted values ​​of perceived and type data, directly reflects the level of abnormal risk at the detection site. It is expressed through a quantified value of the degree of abnormal risk, using specific numbers (e.g., 0-100) to indicate the probability and severity of abnormalities at the detection site; a higher value indicates a higher risk.

[0044] In this step, the multi-dimensional sensing data of each detection site are weighted according to the corresponding data type weight value to obtain the state index of the detection site, thereby quantifying the abnormal risk.

[0045] This step transforms scattered, multi-dimensional sensing data into unified quantitative indicators, solving the problem of incomparability between different types of data. It makes the abnormal risks of the detected parts intuitive and quantifiable, providing a unified standard for subsequent comparison with alarm thresholds and improving the objectivity and accuracy of alarm judgment.

[0046] Step S108: Compare the state index corresponding to each of the multiple detection locations with the corresponding dynamic alarm threshold to determine the alarm event triggering result corresponding to each of the multiple detection locations.

[0047] Among them, the alarm event triggering result refers to the judgment result of whether the detected part needs to be alarmed after comparing the status index with the dynamic alarm threshold, including two categories: alarm triggered and alarm not triggered.

[0048] In this step, the status index of each detection location is compared with the corresponding dynamic alarm threshold one by one. If the status index is higher than the dynamic alarm threshold, it is determined that an alarm has been triggered; otherwise, it is determined that no alarm has been triggered. Finally, the alarm event triggering results of all detection locations are determined.

[0049] This step allows for the rapid identification of alarm locations with potential risks, focusing on subsequent handling and avoiding indiscriminate processing of all detected locations. This improves the targeted nature of alarm event handling and reduces ineffective maintenance costs.

[0050] Step S110: Input the running data corresponding to multiple alarm locations into the corresponding location anomaly detection model to obtain the anomaly results corresponding to the multiple alarm locations. Among them, the multiple alarm locations are the locations where the corresponding alarm event trigger result is an alarm among the multiple detection locations.

[0051] Among them, the alarm location refers to the detection location that triggers an alarm event, and it is the key object of subsequent anomaly analysis.

[0052] Among them, operational data refers to various parameters of the equipment corresponding to the alarm location during operation, including core operational parameters in the sensing data, as well as equipment load, electrical connection parameters, etc., which are key data for in-depth analysis of anomalies.

[0053] Among them, the part anomaly detection model refers to a specialized anomaly analysis model trained on the characteristics of a single detection part, such as a model based on the lightweight gradient boosting machine algorithm, which is used to refine the analysis of anomalies in a single part.

[0054] Among them, abnormal results refer to the specific abnormal information of the alarm location obtained through model analysis, including the type of abnormality, such as equipment defects, sensor failures, degree of abnormality, and probability of occurrence of abnormality.

[0055] In this step, the operational data of the selected alarm locations are input into the location anomaly detection model customized for each location, and the specific anomaly results of each alarm location are output through model analysis.

[0056] This step overcomes the limitation of only knowing the alarm but not the specific anomaly, refines key information such as the anomaly type and degree, avoids directionlessness in handling due to ambiguous anomaly information, provides accurate single-location anomaly basis for subsequent comprehensive anomaly analysis, and improves the pertinence and efficiency of anomaly handling.

[0057] Step S112: Based on the abnormal results corresponding to multiple alarm locations, determine the comprehensive abnormal result, which includes the predicted diffusion path of the abnormality and the risk probability of the associated location.

[0058] Among them, the comprehensive anomaly result refers to the overall anomaly assessment result obtained by combining the anomaly results of multiple alarm locations with the analysis of the correlation between devices. It is a global extension analysis of the anomaly of a single location.

[0059] Among them, the anomaly prediction diffusion path refers to the predicted route of an anomaly spreading from the current alarm location to other related locations, such as the diffusion path of an anomaly in the main transformer winding temperature to the GIS busbar.

[0060] Among them, the risk probability of related parts refers to the probability of abnormalities occurring in other parts that are related to the alarm part, such as electrical connections, heat conduction, etc., which quantifies the risk level of related parts.

[0061] In this step, the abnormal results of all alarm locations are integrated, and the structural and functional dependencies between substation equipment are combined to analyze the potential spread trend of the abnormality and the risks of related locations, ultimately forming a comprehensive abnormal result that includes the predicted spread path of the abnormality and the risk probability of related locations.

[0062] This step enables extended analysis from single-location anomalies to global anomaly risks, allowing for early prediction of the direction of anomaly propagation and associated risks. It helps maintenance personnel develop global prevention and control strategies, avoids dealing with only a single alarm location while ignoring potential cascading faults, reduces the risk of fault escalation, and improves the stability of substation equipment operation.

[0063] Through steps S102-S112 above, sensing data corresponding to multiple detection points of the substation equipment are acquired, including raw sensor data, equipment runtime data, and real-time environmental data. Type data weight values ​​and dynamic alarm thresholds corresponding to each of the multiple detection points are determined, wherein the dynamic alarm thresholds are dynamically determined based on the equipment operating stage, real-time environmental data, and power supply requirements. Based on the sensing data and type data weight values ​​corresponding to the multiple detection points, a state index corresponding to each of the multiple detection points is determined, wherein the state index represents the anomaly of the corresponding detection point. The risk level is quantified; the state index corresponding to each of the multiple detection sites is compared with the corresponding dynamic alarm threshold to determine the alarm event triggering result corresponding to each of the multiple detection sites; the operating data corresponding to each of the multiple alarm sites is input into the corresponding site anomaly detection model to obtain the anomaly result corresponding to the multiple alarm sites, wherein the multiple alarm sites are the sites among the multiple detection sites whose corresponding alarm event triggering result is an alarm; based on the anomaly result corresponding to the multiple alarm sites, a comprehensive anomaly result is determined, wherein the comprehensive anomaly result includes anomaly prediction diffusion path and associated site risk probability.

[0064] As an optional embodiment, determining a comprehensive anomaly result based on the anomaly results corresponding to the multiple alarm locations includes: when the substation equipment includes multiple devices, determining the structural and functional dependencies between the multiple substation devices, and constructing a device association knowledge graph, wherein the device association knowledge graph includes target triples, the target triples include device elements, detection location elements, and dependency relationship elements, and the structural and functional dependencies include at least one of the following: electrical connection relationship, thermal conduction relationship, and functional coupling relationship; determining the comprehensive anomaly result based on the anomaly results corresponding to the multiple alarm locations and the device association knowledge graph.

[0065] Among them, the structural and functional dependency relationship refers to the relationship between multiple substation devices in terms of physical structural connection and functional collaborative operation. It is the core link for mutual influence and interaction between devices.

[0066] Among them, the equipment association knowledge graph refers to a semantic network that stores multiple substation equipment, detection points and dependencies in a structured form, and is used to intuitively present the association logic between equipment.

[0067] Among them, the target triple refers to the basic data unit that constitutes the device association knowledge graph, which fully describes a set of association relationships through three association elements.

[0068] Among them, the equipment element refers to the core element representing the substation equipment itself in the target triplet, and is one of the participating subjects in the association relationship.

[0069] Among them, the detection location element refers to the element in the target triplet that represents the specific location being monitored on the substation equipment, and is the specific functional node of the correlation relationship.

[0070] Among them, the dependency element refers to the element in the target triple that represents the association type between the device element and the detection part element, or between device elements, and is the core of defining the association property.

[0071] Electrical connection relationship refers to a type of structural and functional dependency relationship, which refers to the physical connection between equipment to transmit electrical energy through conductors. For example, the high-voltage side of the main transformer is connected to the GIS busbar through copper busbars, and the conductor connection between the outgoing end of the switch cabinet and the cable joint.

[0072] Among them, thermal conduction relationship refers to a type of structural and functional dependency relationship, which refers to the connection formed between equipment or different parts of equipment through heat transfer. For example, when the main transformer winding overheats, the heat is transferred to the tank shell, and the heat from the GIS busbar is conducted to the adjacent cabinet metal frame. This heat transfer may cause abnormal temperature or abnormal equipment operation.

[0073] Among them, functional coupling relationship refers to a type of structural and functional dependency relationship, which refers to the collaborative operation association formed between equipment to realize specific power business functions. For example, the main transformer, GIS and switch cabinet work together to realize the function of power voltage reduction and distribution, and the circuit breaker and relay protection device work together to realize the function of rapid fault tripping.

[0074] In this embodiment, when multiple substation devices are involved, the physical structural connections and functional coordination relationships between the devices are first clarified, that is, the structural and functional dependencies, specifically including one or more of electrical connection relationships, thermal conduction relationships, and functional coupling relationships. Then, a structured device association knowledge graph is constructed using the target triplet of device element, detection part element, and dependency relationship element as the basic unit. Finally, the abnormal results of multiple alarm parts and the device association knowledge graph are combined to obtain a comprehensive abnormal result that includes the abnormal prediction and diffusion path and the risk probability of the associated parts.

[0075] This approach allows the equipment association knowledge graph to systematically organize the complex relationships between multiple substation devices, avoiding the one-sidedness of anomaly analysis caused by traditional methods that ignore the relationships between devices. By combining anomaly results with the comprehensive analysis of the knowledge graph, the propagation path of anomalies among multiple devices and the risk level of associated parts can be accurately predicted. This extends anomaly detection from a single device or single part to the entire link of multiple devices, effectively reducing the risk of fault expansion caused by unidentified anomalies in the relationships between devices, and improving the comprehensiveness and accuracy of anomaly detection in substation equipment.

[0076] As an optional embodiment, a comprehensive anomaly result is determined based on the anomaly results corresponding to multiple alarm locations, including: determining the location anomaly features corresponding to each of the multiple alarm locations based on the anomaly results corresponding to the multiple alarm locations; mapping the multiple location anomaly features to a unified feature space to obtain multiple unified anomaly features in the unified feature space; performing feature fusion on the multiple unified anomaly features to obtain anomaly fusion features; and determining the comprehensive anomaly result based on the anomaly fusion features.

[0077] Among them, the location anomaly features refer to the key features extracted from the anomaly results of a single alarm location that can characterize the nature of the anomaly in that location, such as the SF6 gas concentration exceeding the standard, the temperature rise suddenly, and the abnormal fluctuation of partial discharge. These features can accurately capture the core attributes of the anomaly and eliminate irrelevant interference information.

[0078] Among them, the unified feature space refers to the standardized feature carrier constructed to eliminate the dimensional and dimensional differences of abnormal features in different parts. For example, different types of abnormal features such as temperature, concentration, and current are mapped to the feature space in the 0-1 range, which can realize the unified comparison of different types of abnormal features and break down feature barriers.

[0079] Among them, unified anomaly features refer to the standardized features formed in a unified feature space after the location anomaly features are mapped. For example, mapping "SF6 gas concentration 1500μL / L" to 0.85 in the unified feature space can make anomaly features from different sources comparable, providing a basis for subsequent fusion.

[0080] Feature fusion refers to the process of integrating multiple unified abnormal features into a comprehensive feature through specific algorithms, such as weighted summation and neural network fusion. This can integrate abnormal information from multiple parts and avoid the limitations of a single feature.

[0081] Among them, the abnormal fusion feature refers to the core feature obtained after feature fusion that can comprehensively reflect the abnormal correlation of all alarm locations. For example, the fusion feature value of 0.78 for integrating the temperature and concentration anomalies of multiple alarm locations can centrally reflect the collaborative correlation of anomalies in multiple locations and highlight the global abnormal pattern.

[0082] In this embodiment, firstly, location anomaly features that reflect the anomaly essence are extracted from the abnormal results of each alarm location; then, these location anomaly features from different alarm locations and of different types are mapped to a pre-constructed unified feature space and transformed into unified anomaly features that can be directly compared; subsequently, multiple unified anomaly features are integrated through a feature fusion algorithm to obtain anomaly fusion features that can comprehensively reflect the anomaly correlation of all alarm locations; finally, based on the anomaly fusion features, a comprehensive anomaly result including anomaly prediction diffusion path and risk probability of associated locations is analyzed and obtained.

[0083] This approach unifies the feature space, resolving the issue of incompatibility caused by differing dimensions and units of abnormal features in different alarm locations, such as temperature, concentration, and discharge quantity, ensuring the rationality and accuracy of feature comparison. Feature fusion integrates abnormal information from multiple alarm locations, avoiding the one-sidedness of single-location anomaly analysis, and enabling fused anomaly features to fully reflect the inherent correlation between anomalies in multiple locations. The comprehensive anomaly results determined based on fused anomaly features can more comprehensively and accurately present global anomaly risks, providing maintenance personnel with more reliable decision-making basis, effectively reducing the risk of fault expansion due to the omission of multiple related anomalies, and further improving the comprehensiveness and accuracy of substation equipment anomaly detection.

[0084] As an optional embodiment, after determining the comprehensive anomaly result based on the anomaly results corresponding to multiple alarm locations, the method further includes: if the comprehensive anomaly result includes associated risk locations, determining a first correlation between the sensing data corresponding to the multiple alarm locations and the anomaly result; determining a second correlation between the sensing data corresponding to non-alarm locations and the non-anomaly result; and determining a third correlation between the sensing data corresponding to associated risk locations and the comprehensive anomaly result, wherein the associated risk location is the location among the multiple detection locations whose corresponding associated location risk probability is greater than a risk threshold; determining the weight ranking of the type data affecting the determination result based on the first correlation, second correlation, and third correlation; and updating the weight values ​​of the type data corresponding to the multiple detection locations based on the weight ranking.

[0085] Among them, associated risk locations refer to locations with a risk probability greater than the risk threshold among multiple detection locations. For example, switch cabinet contacts that are electrically connected to alarm locations can identify potential risk locations that need to be focused on, and prevent cascading failures in advance.

[0086] The first correlation refers to the correspondence between the sensing data corresponding to multiple alarm locations and their own abnormal results. For example, the SF6 gas concentration of 1500 μL / L indicates a moderate SF6 gas leakage anomaly, which can reveal the mapping logic between effective abnormal data and abnormal results.

[0087] Among them, non-alarm locations refer to locations among multiple detection locations where alarm events do not trigger alarms, such as humidity sensor installation locations that do not trigger alarms, which can provide a comparative reference for normal operating status.

[0088] Among them, non-abnormal results refer to the judgment results corresponding to non-alarm parts that are not abnormal. For example, if the humidity sensor detection part is operating normally, it can be compared with abnormal situations to help verify the rationality of the correlation.

[0089] The second correlation refers to the correspondence between the sensing data corresponding to the non-alarm location and its own non-abnormal results. For example, normal data with an ambient humidity of 45% has no abnormal results, which can clearly show the matching pattern between normal data and non-abnormal results.

[0090] The third correlation refers to the corresponding relationship between the perceived data corresponding to the associated risk location and the comprehensive anomaly result. For example, the GIS bus temperature of 75℃ leads to a certain degree of main transformer anomaly spread risk, which can uncover the implicit correlation between potential risk location data and global anomalies.

[0091] Among them, the risk probability of associated parts refers to the quantitative value of the probability that parts that are related to the alarm parts will have an anomaly, such as 78%, which can intuitively reflect the risk level of associated parts.

[0092] Among them, the risk threshold refers to the critical probability value for determining whether the detection site is a related risk site, such as 60%, which can clearly define the screening criteria for related risk sites and avoid indiscriminate attention.

[0093] Among them, categorized data refers to different types of perceived data, such as raw sensor data, device runtime data, and real-time environmental data, which can distinguish the impact of different dimensions of data on anomaly detection results.

[0094] Among them, weighted ranking refers to the ranking of the degree of influence of various types of data on the anomaly detection results based on the analysis of three types of correlations. For example, the original sensor data is greater than the real-time environmental data, which is greater than the equipment runtime data. This can clearly identify the key data types that affect the data, as well as the ranking of the degree of influence of more subdivided data within each type of data.

[0095] Among them, the type data weight value refers to the weight coefficient assigned to different types of sensing data. For example, the weight of the original sensor data is 40%, which can quantify the degree of influence of different types of data and optimize the calculation of the state index.

[0096] In this embodiment, when the overall anomaly result includes associated risk locations, the first correlation between the sensing data of the alarm location and its own anomaly result is determined, the second correlation between the sensing data of the non-alarm location and its own non-anomaly result is determined, and the third correlation between the sensing data of the associated risk location and the overall anomaly result is determined. Then, based on these three types of correlations, the influence of each type of data on the anomaly detection result is analyzed to form a weighted ranking. Finally, according to the weighted ranking, the weight values ​​of the type data corresponding to each of the multiple detection locations are updated.

[0097] This approach comprehensively covers the correlation logic between data and results from three dimensions: abnormal, normal, and potential risk, avoiding the bias in weight allocation caused by single-dimensional analysis. Weight ranking can accurately identify the types of data that play a key role in anomaly detection, and the updated weight values ​​of the types of data can make the subsequent state index calculation more in line with actual operating rules, improving the accuracy of the state index. At the same time, it enables the method to have self-optimization capabilities, continuously adjusting the weights as maintenance data accumulates, gradually reducing false alarms and missed alarms caused by unreasonable data weights, and further improving the accuracy and adaptability of substation equipment anomaly detection.

[0098] As an optional embodiment, before inputting the operational data corresponding to multiple alarm locations into the corresponding location anomaly detection model to obtain the anomaly results corresponding to the multiple alarm locations, the method further includes: acquiring sample data, wherein the sample data includes historical anomaly results, corresponding processing feedback data, and subsequent equipment operation status data corresponding to multiple sample locations; and using a composite loss function to train the initial anomaly detection model for the corresponding location based on the sample data to obtain the corresponding location sample detection model, wherein the composite loss function includes a perceptual similarity loss term and a weighted cross-entropy term, wherein the perceptual similarity loss term is used to determine the loss value between the predicted operational data and the actual operational data, and the weighted cross-entropy term is used to determine the loss value between the predicted anomaly probability and the actual anomaly label.

[0099] Among them, sample data refers to historical datasets used to train the anomaly detection model, such as historical SF6 gas leakage results of a main transformer winding, corresponding maintenance records, and subsequent 3 months of operating status data. This data provides a real and effective foundation for model training and ensures the reliability of model learning.

[0100] Among them, the sample location refers to the substation equipment detection location used as training samples, such as GIS busbar, switchgear contacts, and main transformer oil temperature detection point. This allows the model to learn the abnormal patterns of different locations in a targeted manner, thereby improving the model's adaptability.

[0101] Among them, historical anomaly results refer to the judgment information when anomalies occurred in the sample part in the past. They can provide the model with learning basis such as anomaly type and degree, and help the model identify similar anomalies.

[0102] Among them, processing feedback data refers to relevant information on the handling after an anomaly occurs in the sample area, such as defect category, such as equipment defect, handling measures, such as replacing seals, and retest results, such as eliminating leakage. This allows the model to learn the correlation between anomalies and handling effects and optimize the anomaly judgment logic.

[0103] Among them, the subsequent operating status data of the equipment refers to the continuous monitoring data after the abnormal treatment of the sample part or during normal operation, such as the gas pressure data one month after the SF6 leak treatment, and the oil temperature time series data during the normal operation of the main transformer. It can provide the model with a reference for abnormal recovery or continuous status, and improve the accuracy of abnormality determination.

[0104] Among them, the composite loss function refers to the model training loss calculation function composed of multiple loss terms, such as the combination of perceptual similarity loss term and weighted cross-entropy term, which can take into account different training objectives and comprehensively optimize model performance.

[0105] The initial anomaly detection model refers to the basic model framework that has not been trained or initially constructed, such as the initial model structure built on a lightweight gradient booster, which can provide the basic architecture for subsequent training and reduce the cost of building the model from scratch.

[0106] Among them, the part sample detection model refers to a special anomaly detection model that is trained with sample data and adapted to a specific detection part. For example, an anomaly detection model optimized for the main transformer winding can accurately match the characteristics of the part and improve the accuracy of single part anomaly identification.

[0107] Among them, the perceptual similarity loss term refers to the module in the composite loss function used to calculate the deviation between the predicted operating data and the actual operating data. For example, it calculates the similarity loss between the SF6 gas pressure predicted by the model and the actual monitored pressure. This can ensure the accuracy of the model's prediction of the core operating parameters of the equipment and reduce misjudgments caused by parameter deviations.

[0108] The weighted cross-entropy term refers to the module in the composite loss function used to calculate the deviation between the predicted anomaly probability and the true anomaly label. For example, assigning more weight to the anomaly label and less weight to the normal label in calculating the loss can increase the model's attention to scarce anomaly samples and improve the anomaly recall rate.

[0109] Predictive operational data refers to the model's output of predicted equipment operating parameters, such as the model's predicted main transformer winding temperature data for the next two hours. This provides a trend reference for determining the severity of anomalies and assists in accurately identifying gradual anomalies. It's important to note that this prediction refers to the future operating status of the equipment or related equipment after an anomaly occurs. For example, after a main transformer temperature anomaly, the prediction might include changes in SF6 gas pressure in the GIS or load fluctuations in the switchgear within the following hour. This overcomes the limitation of only identifying current anomalies and supports subsequent analysis of related anomalies and prediction of risk propagation. In other words, equipment anomalies are often gradual, such as slow parameter drift due to insulation aging, or cascading, such as an anomaly in part A spreading to part B. Predictive operational data can predict the possible operating status of equipment in the future, enabling early warning rather than responding only after the anomaly becomes severe.

[0110] Among them, real-world operational data refers to the operational parameter data actually monitored at the sample site, such as the SF6 gas concentration and temperature values ​​collected by the sensor in real time. This data can serve as a benchmark for model training, ensuring that the model prediction results match the actual operational scenario.

[0111] Among them, the loss value refers to the quantified value of the deviation between the model's prediction result and the actual situation, calculated by the loss function. For example, the perceptual similarity loss value is 0.1 and the weighted cross-entropy loss value is 0.05. It can intuitively reflect the model training effect and provide a basis for optimizing model parameters.

[0112] Among them, the predicted anomaly probability refers to the quantitative value of the possibility that an anomaly exists in a sample part output by the model. For example, if the model predicts that the anomaly probability of a certain part is 92%, it can provide a quantitative reference for anomaly judgment and avoid the limitations of absolute judgment.

[0113] Among them, the real anomaly label refers to the identifier of whether there is an anomaly in the sample part, such as abnormal (1) normal (0), which can provide clear supervision signals for model training and ensure that the model learns the correct anomaly recognition logic.

[0114] In this embodiment, before inputting the operational data of multiple alarm locations into the corresponding location anomaly detection model, historical anomaly results, processing feedback data, and subsequent operational status data of multiple sample locations are collected to form sample data. Then, a composite loss function containing a perceptual similarity loss term and a weighted cross-entropy term is used to train each initial anomaly detection model to obtain a location detection model suitable for the corresponding location.

[0115] This approach ensures that the sample data covers the history of anomalies, the handling process, and the subsequent status, providing a comprehensive and multi-dimensional learning basis for model training and avoiding the one-sidedness of model learning caused by single data. The perceptual similarity loss term of the composite loss function guarantees the model's prediction accuracy for the core operating parameters of the equipment, while the weighted cross-entropy term solves the problem of the model's insensitivity to anomaly identification caused by the scarcity of anomaly samples. The two work together to enable the trained part sample detection model to accurately output the anomaly results of alarm parts. Ultimately, this improves the accuracy and reliability of single-part anomaly identification, provides high-quality basic data for the determination of subsequent comprehensive anomaly results, effectively reduces the global anomaly analysis error caused by the deviation of single-part anomaly identification, and further reduces the false alarm rate and false negative rate of substation equipment anomaly detection.

[0116] As an optional embodiment, a state index corresponding to each of the multiple detection sites is determined based on the sensing data and type data weight values ​​corresponding to each of the multiple detection sites. This includes: performing multi-dimensional preprocessing on the sensing data corresponding to each of the multiple detection sites to obtain standardized data corresponding to each of the multiple detection sites. The multi-dimensional preprocessing includes at least one of the following: removing abnormal data caused by sensor drift based on the standard deviation principle, unifying the data dimensions using the standard score standardization method, and aligning the time-series data collected across sensors on the time axis. The state index corresponding to each of the multiple detection sites is determined based on the standardized data and type data weight values ​​corresponding to each of the multiple detection sites.

[0117] Multi-dimensional preprocessing refers to the multi-step standardization process performed on the perceived data, such as a combination of processes including anomaly removal, unified units, and time-series alignment, which can comprehensively optimize data quality.

[0118] Standardized data refers to regularized data obtained after multi-dimensional preprocessing. For example, standardized temperature data is 0.75 and concentration data is 0.62, which can eliminate data differences and facilitate unified calculation.

[0119] Among them, the standard deviation principle refers to the rule of filtering outliers based on the degree of data dispersion. For example, a range of 3 times the standard deviation is used as the boundary of normal data, which can accurately identify abnormal data caused by sensor drift, etc.

[0120] Sensor drift refers to the phenomenon where sensor-collected data deviates from the true value. For example, after long-term use, the temperature sensor may collect values ​​that are 2°C higher than the actual value, which can clearly identify the source of abnormal data that needs to be removed.

[0121] Abnormal data refers to data that deviates from the normal operating pattern of the equipment, such as the instantaneous change in SF6 gas concentration. This can prevent abnormal data from interfering with the status assessment results.

[0122] The standard score standardization method refers to the method of transforming data of different dimensions into standard normal distribution data. For example, temperature (°C) and concentration (μL / L) can be transformed into data with a mean of 0 and a variance of 1, which can break down the dimensional barriers between different types of data.

[0123] Among them, data units refer to the units of measurement of data, such as temperature in °C or electric current in A. These units can clearly define the measurement attributes of data and solve the problem that data with different units cannot be directly compared.

[0124] Cross-sensor acquisition refers to the method of simultaneously acquiring data from multiple different types of sensors. For example, a temperature sensor and an SF6 sensor can simultaneously acquire data from the same device, enabling multi-dimensional data complementarity.

[0125] Time-series data refers to continuous data collected in chronological order, such as a main transformer temperature data sequence collected every 5 minutes, which can reflect the time-varying trend of equipment status.

[0126] Time axis alignment refers to adjusting the time-series data collected by different sensors to the same time point. For example, aligning temperature data collected once every 1 minute with oil chromatography data collected once every 5 minutes can ensure the time synchronization of the data.

[0127] In this embodiment, multi-dimensional preprocessing is first performed on the sensing data of multiple detection sites. Specifically, one or more of the following processes can be selected: removing abnormal data caused by sensor drift based on the standard deviation principle, unifying the data dimensions using the standard score standardization method, and aligning the time-series data collected across sensors to obtain standardized data. Then, combined with the type data weight value corresponding to each detection site, the state index of each detection site is determined by weighted calculation to quantify the abnormal risk.

[0128] This approach utilizes the standard deviation principle in multi-dimensional preprocessing to eliminate invalid data such as sensor drift, preventing outliers from interfering with state assessment. The standard score standardization method unifies the measurement standards of data with different dimensions, making previously incomparable data such as temperature and concentration comparable. Time axis alignment ensures the synchronization of cross-sensor time-series data, avoiding calculation deviations caused by differences in acquisition time. The standardized data after preprocessing is more reliable and consistent. The state index calculated by combining type data weights accurately reflects the true anomaly risk at the detection site, providing an accurate basis for subsequent comparison with dynamic alarm thresholds. This improves the accuracy and reliability of the entire anomaly detection process, reducing false alarms or missed alarms caused by data quality issues.

[0129] As an optional embodiment, after comparing the state index corresponding to each of the multiple detection locations with the corresponding dynamic alarm threshold to determine the alarm event triggering result corresponding to each of the multiple detection locations, the method further includes: if the alarm event triggering result corresponding to the target detection location is an alarm, determining whether there is previous alarm data within a predetermined time period before the current alarm event; if there is no previous alarm data, generating an abnormal entry for the target detection location corresponding to the current alarm event in response to the current alarm event; and / or, if there is previous alarm data, updating the alarm count corresponding to the abnormal entry for the target detection location.

[0130] Among them, the target detection location refers to the specific detection location that is currently identified as an alarm, such as GIS busbar or switch cabinet contacts, which can clearly identify the specific object of alarm analysis.

[0131] The alarm event refers to the alarm behavior currently triggered by the target detection part, such as the main transformer winding temperature exceeding the standard alarm at 14:30 on October 15, 2024, which can accurately locate the time and object of a single alarm.

[0132] The predetermined time period refers to the set time range used to determine whether there are repeated alarms, such as 24 hours or 12 hours, which can clearly define the judgment period for repeated alarms.

[0133] Among them, the previous alarm data refers to the alarm records that have been triggered by the target detection part within the predetermined time period before this alarm event, such as the temperature alarm data of the same main transformer winding at 16:00 on October 14, 2024, which can distinguish between new alarms and repeated alarms.

[0134] Among them, the abnormal entry for the detection location refers to the structured data that records the alarm information of the target detection location, such as entries containing the location name, the time of the first alarm, and the alarm type, which can systematically manage alarm information.

[0135] Among them, the alarm count refers to the cumulative number of alarms corresponding to the abnormal items in the detected area. For example, if the same area accumulates 3 alarms within 24 hours, it can intuitively reflect the frequency of the abnormality.

[0136] In this embodiment, when the alarm event of the target detection part triggers an alarm, it is first determined whether there is previous alarm data for the part within a predetermined time period before this alarm; if not, an abnormal entry for the target detection part corresponding to this alarm is generated; if it exists, the alarm count corresponding to the abnormal entry for the detection part is updated (accumulated by 1).

[0137] This method, by combining the predetermined time period with previous alarm data, can effectively distinguish between new and duplicate alarms, avoiding repeated responses and recordings of the same anomaly, and reducing ineffective maintenance workload. The generation of anomaly entries for detected locations and the updating of alarm counts can systematically organize alarm information, allowing maintenance personnel to clearly understand the frequency and duration of anomalies in each location, facilitating accurate judgment of the severity of anomalies (e.g., high-frequency duplicate alarms may indicate that the fault has not been resolved). At the same time, the structured anomaly entries and cumulative alarm counts provide data support for subsequent fault tracing and equipment status analysis, further improving the standardization of alarm management and the scientific nature of maintenance decisions, and reducing the difficulty of alarm business management.

[0138] Based on the above embodiments and optional embodiments, an optional implementation method is provided, which is described in detail below.

[0139] In related technologies, on the one hand, with the improvement of the intelligence level of substation equipment and the widespread application of online monitoring technology, various sensors are deployed in large numbers to collect operational data from multiple detection points of the equipment. However, in existing monitoring technologies, the various sensing systems are independent of each other, and data processing has obvious limitations: Firstly, the use of fixed weight allocation to integrate multi-dimensional data does not consider the differences in sensor accuracy and data importance, resulting in the weakening of the role of key data; secondly, alarm thresholds are mostly fixed values, which cannot adapt to the different operating characteristics of equipment during the break-in period, stable period, and aging period, nor can they respond to real-time environmental changes and differences in power supply requirements, thus frequently causing duplicate alarms, false alarms, and missed alarms. On the other hand, for online monitoring technology of substations, due to the diversity of monitoring methods and the large number and types of sensors used, there are problems such as the independence of various sensing systems and the difficulty of alarm business management. Specifically, the sensors used to monitor the detection points of various substation equipment often have problems such as duplicate alarms and false alarms, which require repeated responses and processing of alarm events of substation equipment, affecting the overall processing efficiency of alarm events and increasing the difficulty of alarm business management.

[0140] In view of this, the optional embodiments of the present invention provide a method for determining abnormal results of substation equipment and a method for alarming substation equipment, which solves the technical problem of low accuracy of detection results when determining whether substation equipment is abnormal in related technologies, as well as the problem that various sensors used to monitor substation equipment often generate repeated alarms, requiring repeated responses and processing of substation equipment alarm events, which affects the overall processing efficiency of alarm events and increases the difficulty of alarm service management. Figure 2 This is a structural block diagram of a substation equipment alarm system provided by an optional embodiment of the present invention. Figure 3 This is a flowchart illustrating an optional embodiment of a substation equipment alarm method provided by the present invention, which is described below:

[0141] An optional embodiment of the present invention provides a method for determining abnormal results of substation equipment. The method involves acquiring sensing data corresponding to multiple detection points of the substation equipment, including raw sensor data, equipment runtime data, and real-time environmental data; determining type data weight values ​​and dynamic alarm thresholds corresponding to the multiple detection points, wherein the dynamic alarm thresholds are dynamically determined based on the equipment operation stage, real-time environmental data, and power supply requirements; determining state indices corresponding to the multiple detection points based on the sensing data and type data weight values, wherein the state index represents a quantitative value of the abnormal risk level of the corresponding detection point; comparing the state indices corresponding to the multiple detection points with the corresponding dynamic alarm thresholds to determine the alarm event triggering results corresponding to the multiple detection points; inputting the operating data corresponding to the multiple alarm points into the corresponding part anomaly detection model to obtain the abnormal results corresponding to the multiple alarm points, wherein the multiple alarm points are the parts among the multiple detection points whose corresponding alarm event triggering results are alarms; and determining a comprehensive abnormal result based on the abnormal results corresponding to the multiple alarm points, wherein the comprehensive abnormal result includes anomaly prediction propagation paths and associated part risk probabilities. By employing dynamic weight allocation and dynamic alarm threshold adaptation, along with multi-part independent detection and correlation analysis, this method first calculates the status index of each detected part based on raw sensor data, equipment runtime data, and real-time environmental data, combined with dynamic type data weight values. Then, alarm parts are screened using alarm thresholds dynamically determined by equipment operation stage, real-time environmental data, and power supply requirements. After refining the anomaly results through a part anomaly detection model, correlation analysis is used to predict the anomaly propagation path and associated part risks. This achieves the goal of accurately locating anomalies in individual parts and comprehensively controlling the associated risks of multiple parts, thereby improving the technical effect of enhancing the accuracy of substation equipment anomaly detection and the comprehensiveness of risk prediction. This solves the technical problem of low accuracy in detection results when determining whether substation equipment is abnormal in related technologies.

[0142] An optional embodiment of the present invention also provides a substation equipment alarm method, executed on a status monitoring platform of a status monitoring server. The status monitoring platform is communicatively connected to multiple sensors. The method includes: acquiring real-time sensing data of various detection parts of various substation equipment collected by multiple sensors; processing and analyzing the first sensing data of any detection part of any substation equipment collected by any sensor to determine whether the first state of any detection part of any substation equipment triggers an alarm event; if the alarm event is triggered, determining whether there is previous alarm data corresponding to any sensor, any substation equipment, or any detection part within a predetermined time period before the alarm event is triggered; if previous alarm data exists, updating the alarm count corresponding to any sensor, any substation equipment, or any detection part; if no previous alarm data exists, generating alarm data for the current alarm event in response to the current alarm event; and determining the alarm type and alarm level of the current alarm event based on the alarm data for processing the current alarm event.

[0143] Optionally, if the first state of any detection part of any substation equipment does not trigger this alarm event, then it is determined whether the first state of any detection part of any substation equipment has not triggered an alarm event for a predetermined period of time; if the first state of any detection part of any substation equipment has not triggered an alarm event for a predetermined period of time, then the alarm state corresponding to the first state of any detection part of any substation equipment is updated to the reset state.

[0144] Optionally, after generating the alarm data for this alarm event, the method further includes: recording the alarm data in the alarm record list, wherein the alarm data includes the device identifier of any sensor, the device name of any substation equipment, the location name of any detection part, and the trigger time of this alarm event.

[0145] Optionally, determining whether the first state of any detection part of any substation equipment has not triggered an alarm event for a continuous predetermined time includes: querying the corresponding alarm data from the alarm record list based on the device identifier of any sensor to obtain a first query result; and determining whether the first state of any detection part of any substation equipment has not triggered an alarm event for a continuous predetermined time based on the first query result.

[0146] Optionally, if the first state of any detection part of any substation equipment triggers this alarm event, the alarm state corresponding to the first state of any detection part of any substation equipment is updated to an alarm status and displayed in the real-time alarm list.

[0147] Optionally, processing and analyzing the first sensing data of any detection part of any substation equipment collected by any sensor to determine whether the first state of the detection part of the substation equipment triggers the current alarm event includes: processing and analyzing the first sensing data of any detection part of any substation equipment collected by any sensor to determine the first state value of any detection part of any substation equipment; comparing the first state value with the first alarm threshold corresponding to any sensor to determine whether the first state of the detection part of the substation equipment triggers the current alarm event.

[0148] Optionally, in response to the processing operation of this alarm event, the processing information of this alarm event is obtained, including the defect category, defect level, processing measures, and retesting status of any detection part of any substation equipment; the alarm status corresponding to the first state of any detection part of any substation equipment is updated to the processed status.

[0149] Optionally, determining whether there is any previous alarm data corresponding to any sensor, any substation equipment, or any detection point within a predetermined time period prior to triggering this alarm event includes: querying the corresponding alarm data from the alarm record list based on the device identifier of any sensor to obtain a second query result; and determining whether there is any previous alarm data corresponding to any sensor, any substation equipment, or any detection point within a predetermined time period prior to triggering this alarm event based on the second query result.

[0150] Optionally, a variety of sensors are included, such as oil chromatography sensors, leakage current sensors, water level sensors, water immersion sensors, temperature sensors, humidity sensors, SF6 gas monitoring sensors, infrared sensors, ultra-high frequency partial discharge sensors, ultrasonic partial discharge sensors, and transient ground voltage partial discharge sensors.

[0151] Among them, the oil chromatography sensor, leakage current sensor, water level sensor, water immersion sensor, temperature sensor, humidity sensor, and SF6 gas monitoring sensor can respectively collect oil chromatography data, leakage current data, water level data, water inflow data, temperature data, humidity data, and SF6 gas leakage data from the detection points of substation equipment. The infrared sensor, UHF partial discharge sensor, ultrasonic partial discharge sensor, and transient ground voltage partial discharge sensor can respectively collect infrared thermal imaging data, UHF partial discharge data, ultrasonic data, and transient ground voltage data from the detection points of substation equipment.

[0152] An optional embodiment of the present invention also relates to a substation equipment alarm system. The substation equipment alarm system includes a status monitoring server and multiple sensors, each of which may include one or more sensors of the same type. The multiple sensors are used to collect sensing data from various detection points of various substation equipment in the substation, in order to monitor the status of each detection point of each substation equipment. The status monitoring server includes a status monitoring platform, which can communicate and connect with the multiple sensors, thereby acquiring the sensing data from each detection point of each substation equipment collected in real time by the multiple sensors.

[0153] It should be noted that each type of sensor can collect a specific type of sensing data, and different types of sensors collect different sensing data. For example, an oil chromatography sensor can collect oil chromatography data of the detection area of ​​substation equipment, while an infrared sensor can collect infrared thermal imaging data of the detection area of ​​substation equipment.

[0154] In an embodiment of the present invention, the status monitoring platform of the status monitoring server can process and analyze the first sensing data of any detection part of any substation equipment collected by any sensor to determine whether the first state of any detection part of any substation equipment triggers the current alarm event. If the current alarm event is triggered, it is determined whether there is previous alarm data corresponding to any sensor, any substation equipment, or any detection part within a predetermined time period before the current alarm event is triggered. If previous alarm data exists, the alarm count corresponding to any sensor, any substation equipment, or any detection part is updated. If no previous alarm data exists, in response to the current alarm event, current alarm data is generated for the current alarm event. Then, based on the current alarm data, the alarm type and alarm level of the current alarm event are determined in order to handle the current alarm event.

[0155] According to embodiments of the present invention, the equipment types of each substation device include, but are not limited to, the substation main transformer, gas-insulated substation (GIS), and substation switchgear. In other words, the substation equipment alarm system according to the present invention can centrally monitor and manage the status of equipment such as the substation main transformer, gas-insulated substation (GIS), and substation switchgear in each substation.

[0156] The condition monitoring platform can communicate and connect with various sensors. These sensors are used to collect sensing data from various detection points of various substation equipment in the substation, so as to monitor the status of each detection point of each substation equipment. In some embodiments, the various sensors may include, for example, oil chromatography sensors, leakage current sensors, water level sensors, water immersion sensors, temperature sensors, humidity sensors, SF6 gas monitoring sensors, infrared sensors, ultra-high frequency partial discharge sensors, ultrasonic partial discharge sensors, and transient ground voltage partial discharge sensors.

[0157] Step 210: The status monitoring platform can acquire real-time sensing data from various sensors at various detection points of equipment in various substations.

[0158] Here, it should be understood that the installation location of each sensor corresponds to a detection part of a substation device.

[0159] Step 220: The status monitoring platform can process and analyze the first sensing data of any detection part of any substation equipment collected by any sensor to determine whether the first state of any detection part of any substation equipment triggers this alarm event.

[0160] It should be noted that the first sensing data is the sensing data corresponding to any given sensor. The first state is the state corresponding to any given sensor and the first sensing data. For example, if any given sensor is an infrared sensor, then the first sensing data is infrared thermal imaging data, and the first state is temperature.

[0161] In some embodiments, in step 220, the specific method by which the status monitoring platform determines whether the first state of any detection part of any substation equipment triggers the current alarm event is as follows: The first sensing data (e.g., infrared thermal imaging data) of any detection part of any substation equipment collected by any sensor is processed and analyzed to determine the first state value (e.g., temperature value) of any detection part of the substation equipment. Then, the first state value can be compared with the first alarm threshold corresponding to any sensor to determine whether the first state of the detection part of the substation equipment triggers the current alarm event.

[0162] It should be noted that before implementing the substation equipment alarm method of the present invention, the corresponding alarm threshold can be reasonably configured for each sensor based on the actual situation of the substation site, the power supply situation, and the weather conditions.

[0163] If the first state of any detection part of any substation equipment triggers this alarm event, then the following steps 230 can be executed.

[0164] Step 230: The status monitoring platform can determine whether there is any previous alarm data corresponding to any of the above sensors, substation equipment, or detection points within a predetermined time period before the triggering of this alarm event.

[0165] In some embodiments, the predetermined time period can be, for example, 24 hours. That is, in step 230, the status monitoring platform can determine whether there is any previous alarm data corresponding to any of the aforementioned sensors, substation equipment, or detection points within the 24 hours prior to triggering this alarm event. However, the present invention is not limited thereto. The predetermined time period can be set by those skilled in the art according to actual circumstances.

[0166] If, within the predetermined time period prior to triggering this alarm event, there is previous alarm data corresponding to any of the aforementioned sensors, substation equipment, or detection points, then there is no need to respond to this alarm event, and the following steps 240 can be performed.

[0167] Step 240: The status monitoring platform updates the alarm count corresponding to any of the above sensors, any substation equipment, or any detection location (alarm count incremented by 1).

[0168] It should be noted that, according to an embodiment of the present invention, if previous alarm data exists within a predetermined time period before the current alarm event is triggered, the problem of duplicate alarms can be avoided by counting the number of alarms.

[0169] In addition, if there is no previous alarm data corresponding to any of the above sensors, substation equipment, or detection points within the predetermined time period prior to triggering this alarm event, then the following steps 250-260 can be executed.

[0170] Step 250: The status monitoring platform can respond to this alarm event and generate alarm data for this alarm event.

[0171] Step 260: The status monitoring platform can determine the alarm type and alarm level of this alarm event based on the alarm data, so as to handle the alarm event based on the alarm type and alarm level.

[0172] In some embodiments, the status monitoring platform can determine the alarm type, alarm level, and contact information of the current alarm event based on the current alarm data, so as to process the current alarm event based on the alarm type, alarm level, and contact information.

[0173] In some embodiments, if the first state of any detection part of any substation equipment triggers this alarm event, the status monitoring platform can also update the alarm status corresponding to the first state of any detection part of any substation equipment to the alarm status, and can display the alarm status (alarm status) corresponding to the first state of any detection part of any substation equipment in the real-time alarm list.

[0174] In some embodiments, after determining the alarm type and alarm level of the current alarm event, the condition monitoring platform can also obtain processing information for the current alarm event in response to a processing operation on the current alarm event (e.g., a user clicking a processing button on a page and entering processing information). Here, the processing information may include, for example, information such as the defect category, defect level, processing measures, and retesting status of any detection part of any substation equipment. The defect category may include, for example, one or more of the following: substation equipment body defects, sensor defects, and communication equipment defects. Furthermore, the condition monitoring platform can update the alarm status (from the alarm in progress state) corresponding to the first state of any detection part of any substation equipment to the processed state.

[0175] In addition, the status monitoring platform can also suppress alarm events by ceasing to respond to alarm events indicating the first state of any detection point on any substation equipment, and by no longer displaying the alarm status corresponding to the first state of any detection point on any substation equipment in the real-time alarm list. It should be noted that by performing alarm suppression, duplicate responses to known alarm events can be avoided.

[0176] In some embodiments, the substation equipment alarm method may further include steps 270-280.

[0177] In some embodiments, if the first state of any detection part of any substation equipment triggers this alarm event, the following steps 270 can be performed.

[0178] Step 270: The condition monitoring platform can determine whether the first state of any detection part of any substation equipment has not triggered an alarm event for a predetermined duration. In other words, the condition monitoring platform can determine whether the continuous duration for which the first state of any detection part of any substation equipment has not triggered an alarm event has reached a predetermined duration.

[0179] In some embodiments, the predetermined duration is, for example, 3 days. That is, in step 270, the status monitoring platform can determine whether the first status of any detection part of any substation equipment has not triggered an alarm event for 3 consecutive days.

[0180] If no alarm event is triggered for a predetermined duration (e.g., 3 days), the following steps 280 can be performed.

[0181] Step 280: Update the alarm status (from "in alarm" to "reset") corresponding to the first state of any detection part of any substation equipment to "reset". At this time, the alarm status corresponding to the first state of any detection part of any substation equipment will no longer be displayed in the real-time alarm list.

[0182] It should be noted that if no alarm event is triggered for a predetermined period of time when the status of the substation equipment detection part is continuously monitored, the corresponding alarm status can be changed to the reset status. This can avoid the problem of false alarms caused by extreme abnormal conditions.

[0183] In some embodiments, after the status monitoring platform generates alarm data for this alarm event in step 250, it can record this alarm data in the alarm record list. This alarm data may include the device identifier of any of the aforementioned sensors, the device name of any substation equipment, the location name of any detection point, and the trigger time of this alarm event.

[0184] It should be noted that after generating alarm data for each alarm event, the status monitoring platform can record the generated alarm data in the alarm record list. Each alarm data entry in the alarm record list can include the device identifier of the sensor corresponding to the alarm data, the device name of the substation equipment corresponding to the alarm data, the location name of the corresponding detection part, and the trigger time of the alarm event corresponding to the alarm data.

[0185] Based on this, in step 270, when the status monitoring platform determines whether the first state of any detection part of any substation equipment has not triggered an alarm event for a predetermined period of time, it can specifically query the corresponding alarm data from the alarm record list based on the device identifier of any of the aforementioned sensors to obtain a first query result. Furthermore, based on the first query result, it can determine whether the first state of any detection part of any substation equipment has not triggered an alarm event for a predetermined period of time. For example, if no alarm data is found in the alarm record list based on the device identifier of any of the aforementioned sensors, it can be determined based on the first query result that the first state of any detection part of any substation equipment has not triggered an alarm event for a predetermined period of time. As another example, if one or more alarm data are found in the alarm record list based on the device identifier of any of the aforementioned sensors, the trigger time of the alarm event corresponding to one or more alarm data can be determined based on the first query result. Then, based on the trigger time of the alarm event corresponding to one or more alarm data, it can be determined whether the first state of any detection part of any substation equipment has not triggered an alarm event for a predetermined period of time.

[0186] In step 230, when the status monitoring platform determines whether there is any previous alarm data corresponding to any of the aforementioned sensors, substation equipment, or detection points within a predetermined time period prior to triggering the current alarm event, it can specifically query the corresponding alarm data from the alarm record list based on the device identifier of any of the aforementioned sensors to obtain a second query result. Furthermore, based on the second query result, it can determine whether there is any previous alarm data corresponding to any of the aforementioned sensors, substation equipment, or detection points within the predetermined time period prior to triggering the current alarm event. For example, if no previous alarm data is found in the alarm record list based on the device identifier of any of the aforementioned sensors, it can be determined based on the second query result that there is no previous alarm data corresponding to any of the aforementioned sensors, substation equipment, or detection points within the predetermined time period prior to triggering the current alarm event. For example, if one or more previous alarm data are found in the alarm record list based on the device identifier of any of the above sensors, the trigger time of the alarm event corresponding to one or more previous alarm data can be determined based on the second query result. Then, it can be determined whether the trigger time of the alarm event corresponding to each previous alarm data is within a predetermined time period before the current alarm event is triggered, so as to determine whether there is previous alarm data corresponding to any of the above sensors, any substation equipment, or any detection part within the predetermined time period before the current alarm event is triggered.

[0187] Therefore, according to the substation equipment alarm method in this embodiment of the invention, the status monitoring platform can acquire real-time sensing data from various detection parts of various substation equipment collected by multiple sensors, and process and analyze the first sensing data of any detection part of any substation equipment collected by any sensor to determine whether the first state of any detection part of any substation equipment triggers the current alarm event. If the current alarm event is triggered, it is determined whether there is previous alarm data corresponding to any sensor, any substation equipment, or any detection part within a predetermined time period before the current alarm event is triggered. If previous alarm data exists, the alarm count corresponding to any sensor, any substation equipment, or any detection part is updated. If there is no previous alarm data, in response to the current alarm event, current alarm data is generated for the current alarm event, and then the alarm type and alarm level of the current alarm event are determined based on the current alarm data in order to process the current alarm event. Based on this, the present invention can avoid the problem of repeated alarms by counting alarm counts, thereby avoiding repeated responses and processing of alarm events, improving the overall processing efficiency of alarm events, and reducing the difficulty of alarm business management.

[0188] Furthermore, according to the technical solution of the present invention, if the status of the detection part of the substation equipment does not trigger an alarm event for a predetermined period of time, the corresponding alarm status can be changed to a reset status, thus avoiding the problem of false alarms caused by extreme abnormal conditions.

[0189] Through the above optional implementation methods, at least the following beneficial effects can be achieved: The status monitoring platform can acquire real-time sensing data from various detection points of various substation equipment collected by multiple sensors, and process and analyze the first sensing data of any detection point of any substation equipment collected by any sensor to determine whether the first state of any detection point of any substation equipment triggers the current alarm event. If the current alarm event is triggered, it is determined whether there is previous alarm data corresponding to any sensor, any substation equipment, or any detection point within a predetermined time period before the current alarm event is triggered. If previous alarm data exists, the alarm count corresponding to any sensor, any substation equipment, or any detection point is updated. If no previous alarm data exists, in response to the current alarm event, current alarm data is generated for the current alarm event, and then the alarm type and alarm level of the current alarm event are determined based on the current alarm data in order to handle the current alarm event. Based on this, the present invention, by performing alarm counting, can avoid the problem of repeated alarms, thereby avoiding repeated responses and processing of alarm events, improving the overall processing efficiency of alarm events, and reducing the difficulty of alarm business management. If no alarm event is triggered for a predetermined period of time in the status of the detection part of the substation equipment, the corresponding alarm status can be changed to the reset status. This can avoid the problem of false alarms caused by extreme abnormal conditions.

[0190] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0191] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0192] Example 2

[0193] According to embodiments of the present invention, an apparatus for implementing the above-described method for determining abnormal results of substation equipment is also provided. Figure 4 This is a structural block diagram of a substation equipment anomaly result determination device according to an embodiment of the present invention, such as... Figure 4 As shown, the device includes: an acquisition module 402, a first determination module 404, a second determination module 406, a comparison module 408, a third determination module 410, and a fourth determination module 412. The device will be described in detail below.

[0194] The acquisition module 402 is used to acquire sensing data corresponding to multiple detection points of the substation equipment, wherein the sensing data includes raw sensor data, equipment runtime data, and real-time environmental data; the first determination module 404, connected to the acquisition module 402, is used to determine the type data weight value and dynamic alarm threshold corresponding to the multiple detection points, wherein the dynamic alarm threshold is dynamically determined based on the equipment operation stage, real-time environmental data, and power supply requirements; the second determination module 406, connected to the first determination module 404, is used to determine the state index corresponding to the multiple detection points based on the sensing data and type data weight value corresponding to the multiple detection points, wherein the state index represents the quantitative value of the abnormal risk level of the corresponding detection point; the comparison module 4 08, connected to the second determining module 406, is used to compare the state index corresponding to multiple detection locations with the corresponding dynamic alarm threshold to determine the alarm event triggering result corresponding to the multiple detection locations; the third determining module 410, connected to the comparison module 408, is used to input the running data corresponding to multiple alarm locations into the corresponding location anomaly detection model to obtain the anomaly result corresponding to the multiple alarm locations, wherein the multiple alarm locations are the locations among the multiple detection locations whose corresponding alarm event triggering result is alarmed; the fourth determining module 412, connected to the third determining module 410, is used to determine the comprehensive anomaly result based on the anomaly result corresponding to the multiple alarm locations, wherein the comprehensive anomaly result includes the anomaly prediction diffusion path and the risk probability of the associated location.

[0195] It should be noted that the above-mentioned acquisition module 402, first determination module 404, second determination module 406, comparison module 408, third determination module 410 and fourth determination module 412 correspond to steps S102 to S112 in the method for determining abnormal results of substation equipment. The instances and application scenarios implemented by multiple modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiment 1.

[0196] Example 3

[0197] According to another aspect of the present invention, an electronic device is also provided, comprising: a processor; and a memory for storing processor-executable instructions, wherein the processor is configured to execute instructions to implement the substation equipment abnormality determination method of any of the above embodiments.

[0198] Example 4

[0199] According to another aspect of the present invention, a computer-readable storage medium is also provided, which, when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, enables the electronic device to perform the substation equipment abnormality determination method described above.

[0200] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0201] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0202] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0203] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0204] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0205] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0206] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for determining abnormal results of substation equipment, characterized in that, include: Acquire sensing data corresponding to multiple detection points of substation equipment, wherein the sensing data includes raw sensor data, equipment runtime data, and real-time environmental data; Determine the type data weight values ​​and dynamic alarm thresholds corresponding to multiple detection locations, wherein the dynamic alarm thresholds are dynamically determined based on the equipment operation stage, real-time environmental data, and power supply requirements; Based on the sensing data and type data weight values ​​corresponding to the multiple detection sites, a state index corresponding to each of the multiple detection sites is determined, wherein the state index represents a quantitative value of the abnormal risk level of the corresponding detection site. The state index corresponding to each of the multiple detection locations is compared with the corresponding dynamic alarm threshold to determine the alarm event triggering result corresponding to each of the multiple detection locations. The running data corresponding to multiple alarm locations are respectively input into the corresponding location anomaly detection model to obtain the anomaly results corresponding to the multiple alarm locations. The multiple alarm locations are the locations among the multiple detection locations where the corresponding alarm event trigger result is an alarm. Based on the abnormal results corresponding to the multiple alarm locations, a comprehensive abnormal result is determined, wherein the comprehensive abnormal result includes the predicted diffusion path of the abnormality and the risk probability of the associated location.

2. The method according to claim 1, characterized in that, Based on the anomaly results corresponding to the multiple alarm locations, a comprehensive anomaly result is determined, including: When the substation equipment includes multiple devices, the structural and functional dependencies between the multiple substation devices are determined, and a device association knowledge graph is constructed. The device association knowledge graph includes target triples, which include device elements, detection part elements, and dependency elements. The structural and functional dependencies include at least one of the following: electrical connection relationship, thermal conduction relationship, and functional coupling relationship. The comprehensive anomaly result is determined based on the anomaly results corresponding to the multiple alarm locations and the device association knowledge graph.

3. The method according to claim 1, characterized in that, Based on the anomaly results corresponding to the multiple alarm locations, a comprehensive anomaly result is determined, including: Based on the abnormal results corresponding to the multiple alarm locations, determine the location abnormality characteristics corresponding to each of the multiple alarm locations; Multiple abnormal features are mapped to a unified feature space to obtain multiple unified abnormal features in the unified feature space. The multiple unified anomaly features are fused to obtain anomaly fusion features; The comprehensive anomaly result is determined based on the aforementioned anomaly fusion characteristics.

4. The method according to claim 1, characterized in that, After determining the comprehensive anomaly result based on the anomaly results corresponding to the multiple alarm locations, the process also includes: When the comprehensive anomaly result includes associated risk locations, a first correlation is determined between the sensing data corresponding to the multiple alarm locations and the anomaly result, a second correlation is determined between the sensing data corresponding to the non-alarm locations and the non-anomaly result, and a third correlation is determined between the sensing data corresponding to the associated risk locations and the comprehensive anomaly result. The associated risk locations are those locations among the multiple detection locations where the risk probability of the associated location is greater than the risk threshold. Based on the first association relationship, the second association relationship, and the third association relationship, determine the weight ranking of the types of data that affect the determination result; Based on the weighted sorting, update the type data weight values ​​corresponding to the multiple detection sites respectively.

5. The method according to claim 1, characterized in that, Before inputting the operational data corresponding to multiple alarm locations into the corresponding location anomaly detection model to obtain the anomaly results corresponding to the multiple alarm locations, the process also includes: Acquire sample data, wherein the sample data includes historical abnormal results corresponding to multiple sample parts, corresponding processing feedback data, and subsequent operating status data of the equipment; A composite loss function is used to train the initial anomaly detection model for the corresponding part based on the sample data, thereby obtaining the corresponding part sample detection model. The composite loss function includes a perceptual similarity loss term and a weighted cross-entropy term. The perceptual similarity loss term is used to determine the loss value between the predicted running data and the actual running data, and the weighted cross-entropy term is used to determine the loss value between the predicted anomaly probability and the actual anomaly label.

6. The method according to claim 1, characterized in that, Based on the sensing data and type data weight values ​​corresponding to the multiple detection sites, a state index corresponding to each of the multiple detection sites is determined, including: The sensing data corresponding to the multiple detection sites are preprocessed in multiple dimensions to obtain standardized data corresponding to the multiple detection sites. The multi-dimensional preprocessing includes at least one of the following: removing abnormal data caused by sensor drift based on the standard deviation principle, unifying the data dimensions using the standard score standardization method, and aligning the time-series data collected across sensors on the time axis. Based on the standardized data and type data weight values ​​corresponding to the multiple detection sites, the state index corresponding to each of the multiple detection sites is determined.

7. The method according to any one of claims 1 to 6, characterized in that, After comparing the state indices corresponding to the multiple detection locations with the corresponding dynamic alarm thresholds to determine the alarm event triggering results for each of the multiple detection locations, the method further includes: If the alarm event corresponding to the target detection location results in an alarm, determine whether there was any previous alarm data within a predetermined time period before this alarm event; In the absence of the previous alarm data, in response to the current alarm event, an abnormal entry for the detection part corresponding to the target detection part is generated for the current alarm event; and / or, in the presence of the previous alarm data, the alarm count corresponding to the abnormal entry for the detection part corresponding to the target detection part is updated.

8. A device for determining abnormal results of substation equipment, characterized in that, include: The acquisition module is used to acquire sensing data corresponding to multiple detection parts of the substation equipment, wherein the sensing data includes raw sensor data, equipment runtime data, and real-time environmental data. The first determining module is used to determine the type data weight value and dynamic alarm threshold corresponding to multiple detection parts, wherein the dynamic alarm threshold is dynamically determined based on the equipment operation stage, real-time environmental data and power supply requirements. The second determining module is used to determine the state index corresponding to each of the multiple detection sites based on the sensing data and type data weight values ​​corresponding to the multiple detection sites respectively, wherein the state index represents the quantitative value of the abnormal risk level of the corresponding detection site. The comparison module is used to compare the state index corresponding to the multiple detection locations with the corresponding dynamic alarm threshold to determine the alarm event triggering result corresponding to the multiple detection locations. The third determining module is used to input the running data corresponding to multiple alarm locations into the corresponding location anomaly detection model to obtain the anomaly results corresponding to the multiple alarm locations, wherein the multiple alarm locations are the locations among the multiple detection locations whose corresponding alarm event trigger results are alarms; The fourth determining module is used to determine a comprehensive anomaly result based on the anomaly results corresponding to the multiple alarm locations, wherein the comprehensive anomaly result includes anomaly prediction propagation path and risk probability of associated locations.

9. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the method for determining abnormal results of substation equipment as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, When the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device is able to perform the substation equipment abnormality determination method as described in any one of claims 1 to 7.

Images