A power distribution automation fault operation and maintenance management method and system

Abstract

The application provides a power distribution automation fault operation and maintenance management method and system, relating to the technical field of power distribution. The method comprises the following steps: collecting electrical data of a power grid intelligent terminal; extracting real-time electrical characteristics of non-fault electrical disturbance; comparing the real-time electrical characteristics with an electrical characteristic disturbance fingerprint library, and inferring manual operation behavior according to the comparison result; updating the logical state of the corresponding equipment in the power grid digital model according to the inferred manual operation behavior; verifying the updated power grid digital model, and performing fault diagnosis and generating an isolation strategy based on the verified power grid digital model. The method aims to solve the problem of inability to identify manual operation behavior and model disconnection, can effectively identify manual operation behavior of equipment without communication function, ensure the consistency of the power grid digital model and the physical reality, thereby reducing fault diagnosis errors and power outage time.

Description

Technical Field

[0001] This invention relates to the field of power distribution technology, and more specifically, to a method and system for power distribution automation fault operation and maintenance management. Background Technology

[0002] In the field of power distribution automation, the core of fault operation and maintenance management lies in establishing a digital model that can accurately reflect the grid connection relationships in real time. Traditional systems rely on electrical data collected by smart grid terminals and manually entered topology information to construct this model, but this approach has several problems in actual operation. First, for equipment without communication capabilities, manual operation behavior cannot be directly fed back to the system, causing a disconnect between the digital model and physical reality. Second, during equipment maintenance, system upgrades, or network transformations, firmware compatibility issues, data entry errors, and other reasons may lead to structural errors in the grid digital model. Furthermore, when the system topology model contains errors, subsequent manual operation information may not be correctly identified and updated, further exacerbating the discrepancy between the model and reality.

[0003] The main technical problems faced by existing technologies include: the inability to effectively identify manual operations from devices lacking communication capabilities; difficulty in ensuring consistency between digital models and physical reality after system upgrades or network modifications; and the lack of effective verification and correction mechanisms when model errors occur. These problems lead to the system making incorrect diagnoses based on flawed models when real faults occur, potentially expanding the scope of power outages and prolonging fault recovery time. This risk from model inaccuracies is particularly pronounced in modern distribution networks with extensive distributed power generation.

[0004] There is currently no effective technical solution to the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for fault operation and maintenance management of power distribution automation, which aims to solve the problem of the inability to identify human operation behavior and the disconnect between the model and the system. It can effectively identify human operation behavior of equipment without communication functions, ensure the consistency between the digital model of the power grid and the physical reality, thereby reducing fault diagnosis errors and power outage time.

[0006] In a first aspect, the present invention provides a method for fault operation and maintenance management of power distribution automation, comprising the following steps: S1. After establishing an electrical characteristic disturbance fingerprint database corresponding to manual operation of equipment without communication function, collect electrical data from the smart grid terminal; S2. Real-time detection of non-faulty electrical disturbances in the electrical data, and extraction of the real-time electrical characteristics of the non-faulty electrical disturbances; S3. Compare the real-time electrical features with the electrical feature perturbation fingerprint database, and infer the manual operation behavior of devices without communication functions based on the comparison results; S4. Update the logical state of the corresponding equipment in the power grid digital model based on the inferred human operation behavior; S5. Verify the updated power grid digital model, and based on the verified power grid digital model, perform fault diagnosis and generate isolation strategies.

[0007] The power distribution automation fault operation and maintenance management method provided by this invention can effectively identify manual operation behaviors of equipment without communication functions, ensure the consistency between the power grid digital model and physical reality, thereby reducing fault diagnosis errors and power outage time.

[0008] Secondly, the present invention provides a power distribution automation fault operation and maintenance management system, comprising: The data acquisition module is used to acquire electrical data from smart grid terminals after establishing an electrical feature disturbance fingerprint database corresponding to manual operations of devices without communication capabilities. An extraction module is used to detect non-faulty electrical disturbances in the electrical data in real time and extract the real-time electrical characteristics of the non-faulty electrical disturbances; The comparison module is used to compare the real-time electrical features with the electrical feature disturbance fingerprint database, and infer the manual operation behavior of devices without communication functions based on the comparison results. The update module is used to update the logical state of the corresponding equipment in the power grid digital model based on the inferred human operation behavior. The control module is used to verify the updated power grid digital model and, based on the verified power grid digital model, perform fault diagnosis and generate isolation strategies.

[0009] As can be seen from the above, the power distribution automation fault operation and maintenance management method provided by the present invention solves the problem of being unable to identify manual operation behavior and model disconnection by collecting electrical data, detecting non-fault disturbances, extracting features, comparing fingerprint database to infer operation behavior, updating model and verifying it. It can effectively identify manual operation behavior of equipment without communication function, ensure the consistency between the power grid digital model and physical reality, thereby reducing fault diagnosis errors and power outage time.

[0010] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings. Attached Figure Description

[0011] Figure 1This is a flowchart of a power distribution automation fault operation and maintenance management method provided in an embodiment of the present invention.

[0012] Figure 2 This is a schematic diagram of a power distribution automation fault operation and maintenance management system provided in an embodiment of the present invention.

[0013] Label Explanation: 100. Acquisition Module; 200. Extraction Module; 300. Comparison Module; 400. Update Module; 500. Control Module. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0015] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0016] In distribution automation systems, real-time analysis models of power grid connections rely on node connection tables, which accurately record equipment connection status, operating conditions, and topological relationships. However, during the system's lifecycle, errors may be introduced during equipment maintenance, software upgrades, and data entry, leading to inconsistencies between the digital model and physical reality. Specifically, firmware updates to feeder terminal units can cause compatibility differences with the master station system when parsing non-standard topology data encoding. Furthermore, semantic ambiguity during geographic information system (GIS) data entry can result in incorrect labeling of equipment logical states. This inconsistency impairs the accuracy of fault location, isolation strategy formulation, and power restoration path planning, ultimately affecting the safe and stable operation of the power grid.

[0017] For example, during partial expansion and renovation of the distribution network, the initial logical state of newly added disconnecting switches was incorrectly marked as "normally closed" when entered into the geographic information system, while their actual physical state was "normally open." Furthermore, due to a discrepancy between the firmware update of the feeder terminal unit and the parsing logic of the master station system when processing non-standard bit field definitions, the aforementioned error was activated during the verification and integration of the entire network topology data during system software upgrades, resulting in structural errors in the node connection table. When scheduled power outages for maintenance, the signals from manually disconnecting critical disconnecting switches were not correctly reflected in the real-time topology diagram, further decoupling the digital model from physical reality. Consequently, in the event of a sudden line short-circuit fault, the system performed fault diagnosis based on the incorrect topology model, incorrectly locating the fault at a non-existent node and erroneously expanding the power outage area to the normally supplied area.

[0018] If the above problems are not addressed, fault diagnosis results will be unreliable, isolation strategies may incorrectly isolate actual live lines, leading to an expansion of the power outage area. The confidence level of fault information reported by the system will decrease, providing conflicting diagnostic results and making effective manual intervention difficult for maintenance personnel. Furthermore, incorrect isolation operations may trigger secondary faults, increasing the risk to power grid operation.

[0019] For reference, see the appendix. Figure 1 This invention provides a method for fault operation and maintenance management of power distribution automation, comprising the following steps: S1. After establishing an electrical characteristic disturbance fingerprint database corresponding to manual operation of equipment without communication function, collect electrical data from the smart grid terminal; S2. Real-time detection of non-faulty electrical disturbances in electrical data and extraction of real-time electrical characteristics of non-faulty electrical disturbances; S3. Compare the real-time electrical characteristics with the electrical characteristic disturbance fingerprint database, and infer the manual operation behavior of devices without communication functions based on the comparison results; S4. Update the logical state of the corresponding equipment in the power grid digital model based on the inferred human operation behavior; S5. Verify the updated power grid digital model, and based on the verified power grid digital model, perform fault diagnosis and generate isolation strategies.

[0020] For ease of understanding, the following explains some key terms in this embodiment: Electrical Feature Disturbance Fingerprint Database: This database stores electrical feature disturbance patterns corresponding to manual operations of devices without communication capabilities. When a device performs a specific operation (such as switching on or off), it generates unique electrical signal changes in the power grid. These changes, after processing and extraction, form the "electrical feature disturbance fingerprint." The establishment of this fingerprint database is the foundation for this method to identify operations of devices without communication capabilities.

[0021] Smart power grid terminals refer to devices deployed in the distribution network that can collect electrical data and have a certain degree of intelligent processing capabilities, such as feeder terminal units (FTUs) and distribution terminal units (DTUs). These terminals are an important source of real-time power grid operation data.

[0022] Non-faulty electrical disturbances refer to changes in electrical parameters in the power grid that are not caused by equipment failure. These disturbances can be caused by various factors such as normal human operation, load fluctuations, and capacitor switching. This method focuses on disturbances caused by human operation.

[0023] A digital power grid model refers to a model that digitally models the physical power grid and stores it in a computer system in the form of data structures (such as topology diagrams and node connection tables). This model reflects information such as the connection relationships and operating status of power grid equipment and serves as the basis for various analyses and decisions made by distribution automation systems.

[0024] Isolation strategy: refers to a scheme that, when a power grid fault occurs, disconnects the faulty area from the healthy power grid by operating switching equipment, in order to limit the scope of the fault and ensure power supply to the non-faulty areas.

[0025] This embodiment provides a method for fault operation and maintenance management of power distribution automation, which includes the following steps: After establishing an electrical feature disturbance fingerprint database corresponding to manual operations on devices lacking communication capabilities, electrical data from the power grid smart terminal is collected. This step is fundamental to the entire method. The establishment of the electrical feature disturbance fingerprint database aims to provide a basis for subsequent identification of manual operations on devices lacking communication capabilities. For example, in a controlled environment, a preset operation can be performed on a device lacking communication capabilities, and electrical data near the operation point and on related lines can be collected simultaneously. This allows the extraction of electrical features unique to the operation, which are then associated and stored with information such as device type, operation behavior, and operation location to form a fingerprint database. Electrical data collection from the power grid smart terminal can be conducted in various ways, such as periodically polling the smart terminal to obtain data, or using an event-triggered mechanism where the smart terminal actively reports data when a change in the power grid state is detected. This data may include real-time values, instantaneous values, or transient waveforms of parameters such as voltage, current, frequency, and power.

[0026] This process involves real-time detection of non-faulty electrical disturbances in electrical data and extraction of their real-time electrical characteristics. The aim is to identify non-faulty disturbances related to human operation from massive amounts of electrical data. For example, thresholds can be set to monitor sudden changes in parameters such as voltage and current; when the parameter change exceeds a preset threshold, an electrical disturbance is considered likely. To distinguish between faulty and non-faulty disturbances, a comprehensive judgment can be made by combining the disturbance's duration, waveform characteristics, and load changes in the disturbance's location. Specifically, the comprehensive judgment process incorporates the following key dimensions: First, consider the duration of the disturbance. Fault-related disturbances typically last until the protection device trips and clears the fault; their duration can be relatively long or exhibit specific temporal characteristics. Non-fault-related electrical disturbances, such as transient processes caused by switching operations, generally have shorter durations, typically ranging from tens to hundreds of milliseconds, and the power grid quickly returns to its normal steady state after the operation. By setting a reasonable duration threshold, non-fault-related disturbances with transient characteristics can be initially screened out.

[0027] Secondly, waveform characteristics are crucial. Different types of electrical events leave unique "fingerprints" on voltage and current waveforms. Fault-related disturbances, such as short circuits or ground faults, typically cause significant voltage drops and sharp current increases, and may be accompanied by severe waveform distortion or asymmetry. Non-fault-related electrical disturbances, such as the opening and closing operations of disconnecting switches, may produce specific high-frequency transient oscillations, voltage spikes, or current surges, but their amplitude and duration are usually within the system's tolerance range, and the waveform recovery speed is fast. By performing time-frequency analysis on the collected electrical data, the frequency components, transient energy distribution, amplitude, phase, and other characteristics of the disturbance can be extracted and compared with preset fault and non-fault characteristic patterns to identify the similarity of their waveform characteristics.

[0028] Furthermore, the load changes in the disturbance area are crucial. Non-faulty electrical disturbances are often directly related to load changes within the area. For example, the starting of a large motor can cause a sudden increase in load and a voltage drop in a localized area; the switching of capacitor banks can cause changes in reactive power and transient voltage fluctuations. Faulty disturbances, on the other hand, may lead to the disconnection of loads in the faulty area or the emergence of abnormal current paths, and their load change patterns differ significantly from those caused by normal operation. By monitoring real-time load data in the disturbance area and comparing it with historical load curves or typical load change patterns, it is possible to help determine whether the disturbance is caused by load changes.

[0029] Finally, a key criterion is the operation of the protection devices. Faulty disturbances typically trigger the relay protection devices, causing the circuit breaker to trip and clear the fault. Non-faulty electrical disturbances, such as manual operation or normal load switching, usually should not be accompanied by protection action. Therefore, the system checks whether any relevant protection devices issue trip commands or action signals when an electrical disturbance occurs. If an electrical disturbance is detected but no protection action is taken, the likelihood of it being a non-faulty disturbance increases significantly.

[0030] During comprehensive judgment, the system logically combines and evaluates all the above information. For example, a rule engine or machine learning model can be established. The rule engine can be defined as follows: if an electrical disturbance meets the following conditions: "duration less than X milliseconds" AND "waveform characteristics conform to the transient pattern of switching operation" AND "load change in the disturbance area conforms to the normal switching pattern" AND "no accompanying protection action", then it is comprehensively judged as a non-faulty electrical disturbance and further inferred to be a specific human operation. Conversely, if the disturbance has a long duration, severe waveform distortion, and is accompanied by protection action, it may be judged as a faulty disturbance. This fusion and cross-validation of multi-dimensional information can effectively improve the system's accuracy in identifying power grid events, reduce misjudgments, and thus ensure the accuracy of subsequent power grid digital model updates and fault diagnosis.

[0031] In addition, when extracting real-time electrical features, signal processing can be performed on the detected disturbance data, such as Fourier transform and wavelet transform, to obtain features such as the frequency components, transient energy distribution, amplitude, phase and duration of the disturbance.

[0032] The process involves comparing real-time electrical features with an electrical feature perturbation fingerprint database, and inferring the human operation behavior of devices without communication capabilities based on the comparison results. This step is crucial for identifying operations performed by devices without communication capabilities. The comparison process can employ techniques such as pattern recognition and machine learning. For example, the extracted real-time electrical features can be used as input, and similarity calculations can be performed with various operation fingerprints stored in the fingerprint database. The device type, operation behavior, and location corresponding to the fingerprint with the highest similarity are inferred as the current human operation behavior. The comparison results can include a confidence score, indicating the reliability of the inference. For example, if the real-time electrical features have a similarity of 0.9 with the fingerprint of "disconnect switch opening" in the fingerprint database, it can be inferred that the disconnect switch has been opened.

[0033] Based on the inferred human intervention behavior, the logical state of the corresponding equipment in the power grid digital model is updated. This step aims to synchronously reflect human operations in the physical power grid into the digital model, eliminating inconsistencies between the model and reality. For example, if it is inferred that a disconnector switch without communication capabilities has been opened, the system will immediately update the state of that disconnector switch from "closed" to "open" in the power grid digital model. This update is not limited to the operated equipment itself, but may also involve the logical state and connection relationships of its associated equipment. For example, the opening of a disconnector switch may cause its connected downstream lines to become de-energized in the digital model, or change the topology of a ring network.

[0034] The updated power grid digital model is validated, and based on this validated model, fault diagnosis and isolation strategies are generated. This step is crucial for ensuring the accuracy and reliability of the digital model. The validation process can include multiple checks, such as verifying whether the updated model conforms to power grid operation procedures, whether it is consistent with the real-time status of nearby communicable devices, and whether there are any electrical logic inconsistencies. For example, if, after the model update, a line appears to be de-energized in the digital model, but its downstream smart terminal still reports current, it indicates that the model may be incorrect and requires further verification. Only when the digital model passes validation and is confirmed to be consistent with physical reality can fault diagnosis and isolation strategies be generated based on the model. For example, in the event of a fault, the system can use an accurate digital model to quickly locate the fault point and generate an optimal isolation strategy to minimize the power outage area.

[0035] The core technical concept of this solution lies in the detailed analysis of weak electrical signal disturbances in the power grid, comparing them with the "electrical fingerprints" generated by specific manual operations in a pre-established manner. This allows for the inverse inference of on-site operations performed by maintenance personnel on equipment lacking communication capabilities (such as old disconnect switches) without relying on communication functions. This "indirect sensing" mechanism enables the distribution automation system to correct the logical state of equipment in its internal digital model in real time and non-intrusively. This ensures that in the event of a sudden fault, the system can make fault diagnosis and isolation decisions based on a model highly consistent with the physical power grid, effectively avoiding misjudgments and the expansion of power outage areas due to information gaps.

[0036] The following example will provide a more detailed explanation of the above technical solution: Traditional distribution automation systems rely on a model capable of real-time analysis of grid connectivity for fault management, with the node connection table forming the foundation for all intelligent decisions. However, in practice, structural errors can occur in the node connection table of the digital grid model due to firmware compatibility issues with early-deployed feeder terminal units (FTUs), semantic ambiguity and distraction during data entry by maintenance personnel, and deficiencies in data verification and integration during master station software upgrades. For example, a feeder that should be radial might incorrectly form a closed loop in the model, or a real connection point might be incorrectly marked as disconnected. Given this significant disconnect between the digital model and physical reality, if a planned localized power outage is implemented in a certain area, the maintenance team will manually disconnect multiple isolating switches in the relevant area according to operating procedures. These manual operations are reported to the master station system via the field FTUs. However, due to structural errors in the node connection table of the master station, when the system receives a manual disconnection signal from a critical disconnecting switch, it cannot find the correct location of the device in its erroneous topology model. Even if it does find it, it cannot update its status with the existing erroneous records in the model. This results in the legitimate and important manual operation not being correctly reflected in the real-time topology diagram, causing a further disconnect between the digital model and physical reality. The system still "thinks" that some switches that have actually been disconnected are in the closed state, or that some lines are still energized, when in fact they have been safely isolated. In this situation, if a real line short-circuit fault occurs in the aforementioned area, the distribution automation system will quickly initiate the fault diagnosis process. However, because the node connection table in its core real-time power grid connection analysis model has been severely distorted during the upgrade process and has failed to synchronize subsequent manual operations, the system will not be able to accurately identify the exact location of the fault. Based on its erroneous topology perception, the system may issue unnecessary isolation commands, leading to an expansion of the power outage area far beyond the actual fault area. At the same time, due to the confusion in the system's internal logic, the confidence level of its reported fault location results is extremely low, and it may provide contradictory diagnostic information. This creates a fog of information and decision-making dilemmas for on-site maintenance personnel when attempting manual intervention and correction, as they cannot trust the information provided by the system and find it difficult to manually verify the true status of all equipment in a short period of time. In this situation, the time window for fault isolation and power restoration is significantly prolonged, and secondary faults may even occur due to maintenance personnel performing incorrect isolation operations under insufficient information or misguidance, further exacerbating the risks to power grid operation.

[0037] In response, this application proposes a fault operation and maintenance management method for power distribution automation, which aims to solve the problems of inaccurate fault diagnosis and incorrect isolation strategies caused by the inconsistency between digital models and physical reality in power distribution automation systems.

[0038] For example, suppose there is a disconnecting switch in a power distribution network that lacks communication capabilities, and maintenance personnel manually open it. Because the disconnecting switch lacks communication capabilities, its operational status cannot be directly reported to the distribution automation master station system. As a result, the status of the disconnecting switch in the power grid digital model of the master station system is still "closed," causing an inconsistency between the digital model and physical reality.

[0039] To address this issue, this method first establishes an electrical characteristic disturbance fingerprint database corresponding to manual operations on devices lacking communication capabilities, and then collects electrical data from smart grid terminals. Specifically, when maintenance personnel manually open the disconnect switch, their operation causes instantaneous changes in parameters such as voltage and current in the power grid. Smart grid terminals deployed near the disconnect switch and on related lines collect this electrical data in real time.

[0040] Subsequently, the system detects non-faulty electrical disturbances in these electrical data in real time and extracts their real-time electrical characteristics. For example, the system identifies current surges and voltage fluctuations caused by the opening of disconnecting switches and performs time-frequency analysis to extract their unique electrical characteristics, such as frequency components, transient energy distribution, amplitude, phase, and duration. These characteristics differ significantly from common fault characteristics in the power grid (such as short-circuit current and grounding current), thus enabling the differentiation of non-faulty disturbances.

[0041] Next, the system compares the extracted real-time electrical features with a pre-established electrical feature disturbance fingerprint database. This database stores electrical feature fingerprints generated by various devices without communication capabilities during different operations. Through comparison, the system can infer that the current non-faulty electrical disturbance was caused by the manual opening of the disconnector switch based on the fingerprint with the highest similarity. For example, if the similarity between the real-time electrical feature and the fingerprint for "disconnector switch opening" in the database reaches 0.9, the system infers that the disconnector switch has been opened.

[0042] Based on the inferred human operation behavior, the system immediately updates the logical state of the corresponding device in the power grid digital model. Specifically, the system updates the logical state of the disconnector switch in the power grid digital model from "closed" to "open" and adjusts the connectivity between the two nodes it connects accordingly. For example, if the disconnector switch connects to two lines, its opening operation will cause these two lines to become disconnected in the digital model.

[0043] Finally, the system verifies the updated power grid digital model. The verification process checks whether the operation conforms to the current power grid operating procedures, such as whether it is within the permitted operating time window and whether it matches known planned maintenance tasks. Simultaneously, the system checks whether the operation is consistent with the real-time status of other nearby communicable devices. For example, if a disconnector is inferred to be open, the system checks whether its downstream FTU reports zero current and whether its upstream voltage remains normal. If the verification passes, the status update is confirmed as valid, and the device's status in the digital model is permanently updated, thus correcting the discrepancy between the digital model and physical reality. If the verification fails, the system issues an alarm, prompting maintenance personnel to conduct manual verification and temporarily maintaining the original model status or marking it as an uncertain state. Based on the verified power grid digital model, the system can perform accurate fault diagnosis and generate isolation strategies. For example, if a line short-circuit fault occurs in the area, the system can quickly locate the fault point based on the accurate digital model and generate the optimal isolation strategy, such as disconnecting only the faulty line without misjudging other healthy areas, thereby minimizing the power outage area and improving power supply reliability.

[0044] This embodiment addresses the problem of traditional power distribution automation systems being unable to detect manual operations on equipment lacking communication capabilities by establishing an electrical feature disturbance fingerprint database. Compared to traditional solutions that rely on manual reporting or periodic inspections to update equipment status, this method can automatically identify these operations in real time, significantly improving the consistency between the digital model and physical reality. By detecting non-faulty electrical disturbances in real time and extracting their features, this method can effectively distinguish between normal operations and actual faults, avoiding misdiagnosis caused by misjudgment in traditional solutions. For example, in traditional solutions, the manual tripping of a disconnector switch might be mistaken for a line fault, triggering unnecessary protection actions. This method, through fingerprint comparison, can accurately infer manual operation behavior, solving the problem of untimely model updates due to missing information in traditional solutions. For example, in traditional solutions, the tripping operation of a disconnector switch without communication capabilities might take hours or even days to be manually updated into the digital model. This method can update the logical state of the corresponding equipment in the power grid digital model in real time based on the inferred manual operation behavior, ensuring the real-time performance and accuracy of the digital model. This contrasts sharply with the disconnect between the digital model and physical reality in traditional solutions. By validating the updated power grid digital model, this method further ensures the model's reliability and avoids fault diagnosis and strategy generation based on an incorrect model. For example, in traditional methods, an incorrect digital model may lead to inaccurate fault location or even generate incorrect isolation strategies, expanding the outage area. This method, based on the validated digital model, performs fault diagnosis and generates isolation strategies, significantly improving the accuracy and efficiency of fault handling and reducing power grid operation risks.

[0045] In some embodiments, the step of establishing an electrical feature perturbation fingerprint database corresponding to manual operations on devices without communication capabilities includes: A1. Under controlled conditions, perform preset operations on devices that do not have communication functions, and collect electrical data near the device operation point and on related lines while performing preset operations; A2. Process the collected electrical data and extract the electrical characteristics generated by the preset operations; A3. Associate the extracted electrical features with the corresponding device type, operation behavior, and operation location, and store them to form an electrical feature perturbation fingerprint database.

[0046] Specifically, "under controlled conditions" refers to a state in an experimental or testing environment where external interference factors (such as load fluctuations, other equipment operation, environmental noise, etc.) are minimized or completely eliminated through human intervention or system settings. This ensures that the preset operations performed produce clear and repeatable electrical characteristics, avoiding data confusion or feature ambiguity caused by environmental complexity. For example, operations can be performed during off-peak hours at night or at a dedicated test site to ensure the stability and predictability of the power grid's operating status. Another controlled condition could be that, before performing the operation, the monitoring system confirms that the power grid is in a stable operating state with no other abnormal events, and records the baseline electrical data at this time for subsequent differential analysis. "Equipment without communication capabilities" refers to those devices in the distribution network that cannot directly report their status or operational information to the master station system via remote communication protocols (such as IEC 61850, Modbus, DNP3, etc.). These devices are usually older equipment, cost-sensitive equipment, or certain mechanical equipment, such as some manual disconnect switches, drop-out fuses, and load switches without communication modules. "Preset operations" refer to specific actions performed on equipment lacking communication capabilities according to predetermined operating procedures or test plans. These operations aim to simulate manual operations that may occur during actual operation and maintenance, such as opening / closing disconnect switches, engaging / disengaging fuses, and closing / opening grounding switches. By performing these known operations, the correspondence between operational behaviors and the resulting electrical characteristics can be clarified. "Collecting electrical data near the operating point of the equipment and on associated lines" refers to acquiring power grid parameters directly related to the operational behavior using high-precision sensors and data acquisition units when performing preset operations. These data acquisition points are typically located directly above, below, to the sides of the operated equipment, or at key locations such as feeders or branches connected to it. The collected electrical data may include, but is not limited to, voltage, current, active power, reactive power, frequency, harmonic content, and transient voltage / current waveforms. For example, a transient waveform recorder can be used to record high-frequency transient signals near the operating point, or a smart meter can be used to collect steady-state electrical quantities over a period of time before and after the operation.

[0047] It should be noted that the term "associated lines" typically refers to lines that are electrically directly connected to or physically close enough to the operating point of equipment that lacks communication capabilities, such that the operating behavior of that equipment (e.g., switching on and off) generates electrical disturbances on these lines that can be detected by the smart grid terminal. This association ensures that the collected electrical data effectively reflects the operating characteristics of the equipment. Specifically, this may include: 1. Directly connected lines: For example, when a disconnecting switch trips, both its directly connected upstream and downstream lines will be affected.

[0048] 2. Adjacent lines: In some cases, even without a direct electrical connection, detectable electrical disturbances may occur in other lines near the operating point due to physical effects such as electromagnetic induction.

[0049] 3. Lines within the same feeder or area: Typically, electrical disturbances caused by the operation of a device will propagate along its feeder or local area. Therefore, other lines within the same area may also be considered as related lines in order to comprehensively capture disturbance information.

[0050] "Processing the acquired electrical data" refers to performing a series of preprocessing and analyses on the raw acquired electrical data to eliminate noise, calibrate the data, and prepare for feature extraction. This may include data filtering (such as low-pass filtering and band-pass filtering to remove high-frequency noise or specific frequency interference), data synchronization (ensuring that data timestamps from different acquisition points are aligned), data normalization (unifying data of different dimensions into a comparable range), and baseline drift correction. For example, wavelet denoising algorithms can be used to process transient waveforms to highlight characteristic signals caused by the operation. "Extracting the electrical features generated by the preset operation" refers to identifying and quantifying physical quantities or patterns from the processed electrical data that can uniquely characterize a specific preset operation. These features should be discriminative, able to distinguish different operational behaviors. For example, voltage drop / rise amplitude, current surge rate, transient oscillation frequency, duration, energy spectral density, and harmonic distortion rate changes caused by the operation can be extracted. These features can be time-domain features (such as peak value, root mean square value, number of zero crossings), frequency-domain features (such as energy of specific frequency components, harmonic content), or time-frequency-domain features (such as wavelet coefficients, S-transform results).

[0051] "Associating extracted electrical features with corresponding equipment types, operational behaviors, and operational locations" refers to establishing a mapping relationship between electrical features and equipment, operations, and locations in the actual physical world. This association is the core of building the fingerprint database, enabling the system to infer specific operational events from electrical features. For example, a specific voltage sag waveform may be associated with "the opening operation of a certain type of disconnecting switch on a certain feeder in a certain substation." "Equipment type" refers to the specific model, specifications, or functional classification of equipment that does not have communication capabilities, such as "drop-out fuse," "manual disconnecting switch," and "load switch." "Operational behavior" refers to the specific actions performed on the equipment, such as "opening," "closing," "connecting," and "disconnecting." "Operation location" refers to the specific geographical location or topological location of the equipment in the power grid, such as "the disconnecting switch on the XX feeder of the XX substation." "Storing to form the electrical feature disturbance fingerprint database" refers to storing the above-mentioned association information in a structured manner, forming a database that can be queried and compared. The fingerprint database can be stored using a relational database (such as MySQL or PostgreSQL), which includes table structure definitions such as device ID, device type, operation type, operation location, feature vector (or feature parameter set), and feature generation time. Alternatively, it can be stored using a non-relational database (such as MongoDB) in document format, offering greater flexibility to adapt to different feature structures. Each entry in the fingerprint database represents a known, verified operation event and its unique electrical characteristic, a "fingerprint."

[0052] This solution ensures the accuracy and reliability of the fingerprint database through a systematic fingerprint database construction process. After electrical data is collected from the smart grid terminal, this data needs to be compared to infer human operation behavior. To ensure accurate comparison results, the underlying electrical feature perturbation fingerprint database must be of high quality. This solution first emphasizes data collection under controlled conditions. This means that during the execution of preset operations, external environmental factors are controlled to minimize the impact of noise and interference on electrical data, thereby ensuring the purity and consistency of the raw data. This high-quality raw data is a prerequisite for accurate extraction of electrical features. Next, the collected data undergoes refined processing to extract electrical features that uniquely characterize specific operational behaviors. This step ensures that the extracted features have sufficient discriminative power and representativeness, accurately reflecting the essence of the operational behavior and avoiding identification ambiguity caused by improper feature extraction. Finally, these extracted electrical features are explicitly associated with specific equipment types, operational behaviors, and the location of the operation and stored, constructing a structured and queryable fingerprint database. This associative storage method not only gives the fingerprint database a clear logical structure, facilitating rapid retrieval and comparison, but also greatly improves its practicality and comparison accuracy by binding features with multi-dimensional information. Through the above steps, this solution provides a solid and accurate foundation for subsequent real-time detection of non-faulty electrical disturbances and inference of human operation behavior, effectively solving the accuracy problems that may occur during the fingerprint database construction process, thereby improving the reliability of the entire power distribution automation fault operation and maintenance management method.

[0053] As a specific implementation method, establishing an electrical characteristic disturbance fingerprint database can be carried out with reference to the following example. First, for example, during the period of lowest grid load and most stable operation late at night, or in a dedicated test station, preset operations are performed on equipment in the distribution network that does not have communication capabilities, such as a certain type of pole-mounted disconnector. The preset operations can include two typical actions: "opening" and "closing". When performing these operations, using a high-precision transient waveform recorder and smart meters, instantaneous waveform data of voltage and current (sampling rate can reach several kHz to MHz) and steady-state voltage, current, and power data are simultaneously collected upstream, downstream, and on adjacent branches of the disconnector operation point. Second, the collected electrical data is processed to extract the electrical characteristics generated by the preset operations. Specifically, for instantaneous waveform data, wavelet transform or S-transform can be used for time-frequency analysis to identify and separate the transient impact components and high-frequency oscillation components generated at the moment of operation. For steady-state data, the changes in the root mean square values ​​of voltage and current, phase angle, and harmonic distortion rate before and after the operation can be calculated. For example, the characteristics of a disconnector switch opening operation might include: at the instant of opening, the downstream current at the operating point rapidly drops to zero, accompanied by a voltage dip or rise lasting several milliseconds, and possibly damped oscillations at a specific frequency. The amplitude, duration, and frequency components of these characteristics are quantified and extracted. Finally, the extracted electrical characteristics are associated with the corresponding equipment type, operating behavior, and location of the operation, forming an electrical feature disturbance fingerprint database. For example, the extracted voltage dip amplitude, current change rate, transient oscillation frequency, and other characteristic parameters are bound to information such as "a certain type of pole-mounted disconnector switch," "opening operation," and "XX substation XX feeder XX tower," and stored in a relational database. Each record in the database contains a unique fingerprint ID, equipment ID, equipment type, operation type, operation location coordinates, and a feature vector containing all quantified characteristic parameters. Thus, when similar electrical disturbances are subsequently collected in real time, the characteristics in the fingerprint database can be compared to quickly and accurately identify which equipment performed which manual operation at which location.

[0054] Through the above technical solution, this application effectively solves the problem of decreased accuracy of the electrical feature disturbance fingerprint database due to environmental interference, improper feature extraction, and insufficient correlation during the establishment process. By collecting data under controlled conditions, the purity and consistency of the original electrical data are ensured, providing a high-quality foundation for subsequent feature extraction. Refined processing of the collected data and extraction of representative electrical features ensure that the constructed fingerprint database accurately reflects the essence of different operational behaviors. Furthermore, the extracted electrical features are associated and stored in a multi-dimensional manner with equipment type, operational behavior, and location of operation, giving the fingerprint database a clear logical structure and high-precision matching capability. The establishment of this high-quality, high-accuracy electrical feature disturbance fingerprint database provides a reliable basis for real-time detection of non-faulty electrical disturbances and inference of manual operation behaviors of equipment without communication functions in subsequent power distribution automation fault operation and maintenance management methods. This significantly improves the accuracy of the system's perception of the power grid status and the level of intelligent operation and maintenance management, effectively avoiding fault diagnosis errors and isolation strategy failures caused by the disconnect between digital models and physical reality, thus ensuring the safe and stable operation of the power grid.

[0055] In some embodiments, the specific steps in step A2 include: A21. Perform time-frequency analysis on the collected electrical data to obtain the time-frequency analysis results; A22. Based on the time-frequency analysis results, identify and separate the frequency components and transient energy distributions related to the preset operation; A23. Based on the frequency components and transient energy distribution, extract the amplitude, phase, and duration characteristics generated by the preset operation, and use them as the electrical characteristics generated by the preset operation.

[0056] Time-frequency analysis (TFA) is a signal processing technique that aims to simultaneously reveal the time and frequency variations of a signal. Unlike traditional Fourier transform, which only provides the overall frequency components of a signal, TFA captures the dynamic evolution of frequency components at different points in time. This is crucial for analyzing non-stationary signals, especially electrical disturbances in power systems caused by transient events such as switching operations. TFA can be implemented in various ways, including but not limited to: Short-Time Fourier Transform (STFT), which involves performing a Fourier transform on a framed, windowed signal to generate a time-frequency graph; or using wavelet transform, which decomposes the signal using wavelet basis functions at different scales to analyze local features at different frequency resolutions. After obtaining the TFA results, it is necessary to accurately locate and extract electrical features closely related to a specific pre-defined operation. This is because the power grid may contain various background noises and non-operational disturbances. The purpose of identification and separation is to focus on the unique signal features generated by the operation itself to improve the accuracy of subsequent feature extraction. Implementation methods can include: setting thresholds or masks on the time-frequency graph based on a preset frequency range and time window to filter out irrelevant frequency components and background noise; or, using pattern recognition algorithms to train a model to identify the unique patterns formed by specific operations on the time-frequency graph and separate them from other signals. Amplitude, phase, and duration are key parameters describing electrical disturbance events, comprehensively characterizing the features of a transient event. Amplitude reflects the intensity of the disturbance, phase reveals the start or change relationship of the disturbance relative to a reference point, and duration quantifies the duration of the disturbance event. The combination of these features forms a unique "fingerprint" for distinguishing different operational behaviors. Extraction of these features can include: for the separated frequency components, obtaining amplitude features by calculating their instantaneous amplitude envelope, obtaining instantaneous phase information through methods such as Hilbert transform, and determining the duration by detecting the start and end points of the signal; or, integrating or peak detecting the transient energy distribution to obtain its energy intensity and duration, and combining this with the phase information of the original signal for extraction.

[0057] It should be noted that the "frequency components and transient energy distributions related to the preset operation" refer to electrical signal characteristics that are unique to or primarily generated by that specific preset operation (e.g., the opening and closing of a disconnector switch). These characteristics can distinguish the preset operation from other background noise, other non-faulty disturbances (such as load switching), or fault events. This correlation emphasizes the uniqueness and representativeness of the characteristics. Specifically, it includes: 1. Frequency component: refers to the specific frequency components exhibited in the frequency domain by electrical disturbances caused by preset operations. For example, switching operations may generate specific high-frequency or ultra-high-frequency components due to the arcing effect.

[0058] 2. Transient energy distribution: refers to the distribution pattern of energy in the time-frequency domain of electrical disturbances caused by a preset operation, at different points in time and within a specific frequency range. For example, a switching operation may produce a phenomenon where energy is concentrated within a specific frequency range for a very short time (transient).

[0059] This application's solution, through refined time-frequency analysis of the acquired electrical data, can simultaneously capture the dynamic changes of electrical disturbances in both time and frequency dimensions, thus overcoming the limitations of traditional single time-domain or frequency-domain analysis in handling transient and non-stationary signals. Specifically, firstly, time-frequency analysis is performed on the acquired electrical data to generate a three-dimensional time-frequency spectrum containing time, frequency, and energy information, laying the foundation for subsequent feature identification. Secondly, based on the results of this time-frequency analysis, the system can accurately identify and separate the frequency components and transient energy distributions associated with specific preset operations. This process, by focusing on the unique signal characteristics generated by the operation itself, effectively eliminates interference from background noise and irrelevant disturbances, ensuring the purity and relevance of the features of interest. Finally, from these identified and separated frequency components and transient energy distributions, the system further extracts key electrical features such as amplitude, phase, and duration. These features can comprehensively and accurately characterize the electrical disturbances caused by the preset operation, forming a highly discriminative "electrical fingerprint." By organically combining the above steps, the solution in this application ensures that the extracted electrical features are not only accurate but also have sufficient discriminative power, thus providing a solid data foundation for establishing a highly accurate and reliable electrical feature perturbation fingerprint database. This significantly improves the accuracy of inferring the manual operation behavior of devices without communication capabilities when comparing real-time electrical features with the fingerprint database, thereby ensuring the reliability of power grid digital model updates.

[0060] The following is a specific example. As a concrete implementation, after performing a preset operation on a device without communication capabilities and collecting electrical data, continuous wavelet transform (CWT) can be used to perform time-frequency analysis on this electrical data, generating a wavelet coefficient matrix. This matrix can visually display the energy distribution of the signal at different time and frequency scales, thus obtaining the time-frequency analysis results. For example, when a disconnector switch is manually disconnected, a transient disturbance with a very short duration and a wide frequency range may occur. Based on this time-frequency analysis result, an energy threshold and a specific frequency range (e.g., 5kHz to 50kHz) can be set to identify and separate the frequency components and transient energy distribution associated with the switch operation that appear within a specific time window. For example, an energy density-based clustering algorithm can be used to identify regions in the time-frequency spectrum where energy is concentrated and conforms to the preset operation characteristics. Subsequently, based on these separated frequency components and transient energy distributions, their features can be further extracted. For example, the peak amplitude of the transient disturbance, its initial phase angle in the power frequency cycle, and the duration from the start of the disturbance to decay to the background noise level can be calculated. These extracted amplitude, phase, and duration features serve as the electrical characteristics of the disconnecting switch's disconnection operation, and are used to construct an electrical feature perturbation fingerprint database.

[0061] Through the above technical solution, this application effectively solves the problem of inaccurate features or inability to effectively distinguish different operations caused by the lack of specific feature extraction methods in traditional methods. By introducing time-frequency analysis, the dynamic characteristics of electrical disturbances in time and frequency can be comprehensively captured, providing a rich and detailed data view for subsequent feature identification. Furthermore, by identifying and separating the frequency components and transient energy distribution related to the preset operation, the interference of background noise and irrelevant disturbances is effectively eliminated, ensuring the purity and specificity of the extracted features. Finally, by extracting key electrical features such as amplitude, phase, and duration, a highly discriminative "electrical fingerprint" for the preset operation can be formed. These accurate and discriminative electrical features significantly improve the accuracy and reliability of the electrical feature disturbance fingerprint database, thus laying a solid foundation for accurately inferring the manual operation behavior of equipment without communication functions, thereby ensuring the accuracy of power grid digital model updates and the reliability of fault diagnosis.

[0062] In some embodiments, the specific steps in step S3 include: S31. Based on the preset confidence threshold in the electrical feature perturbation fingerprint database, compare the real-time electrical features with the electrical feature perturbation fingerprint database to obtain preliminary comparison results; S32. Obtain non-electrical accompanying information related to the area where the non-faulty electrical disturbance occurs; the non-electrical accompanying information includes operation and maintenance work plan information, equipment geographical location information, and area load status information; S33. Based on non-electrical accompanying information, generate an operational probability prediction; S34. Based on the operational probability prediction, the preliminary comparison results are revised to obtain the revised comparison results; S35. Based on the revised comparison results, infer the manual operation behavior of devices that do not have communication functions.

[0063] In step S31, real-time electrical features refer to the features extracted from non-faulty electrical disturbances detected in electrical data collected from the smart grid terminal, reflecting the dynamic changes of the power grid at a specific moment. The electrical feature disturbance fingerprint database is a pre-established database that stores electrical features with unique patterns corresponding to manual operations of devices without communication capabilities. The comparison process can employ various pattern recognition algorithms, such as those based on Euclidean distance, cosine similarity, or dynamic time warping (DTW), to quantify the similarity between real-time electrical features and known operational features in the fingerprint database, thereby obtaining preliminary comparison results. These results are typically expressed as confidence levels (the electrical feature disturbance fingerprint database has preset confidence thresholds, which are generally set through expert experience) or similarity scores for different operational behaviors.

[0064] In step S32, the area where non-faulty electrical disturbances occur refers to the geographical range within the power grid where electrical disturbances occur. This range can be determined by analyzing the source, propagation path, or impact range of real-time electrical data. Non-electrical accompanying information is auxiliary data that exists independently of electrical data and provides operational context information. This can include maintenance and repair work plan information, equipment geographic location information, and regional load status information. Maintenance and repair work plan information refers to pre-planned inspection, maintenance, or operation tasks, typically including the work time, location, and type of equipment involved. Equipment geographic location information refers to the actual physical coordinates or topological location of equipment without communication capabilities within the power grid. Regional load status information refers to the changing trend or current level of power load within a specific area, such as a sudden increase or decrease in load. This non-electrical information provides important background and constraints for judging human intervention behavior.

[0065] It should be noted that the specific basis for defining and analyzing the areas where non-faulty electrical disturbances occur is, for example: Firstly, the core of analyzing the sources of real-time electrical data lies in identifying which power grid smart terminals (or those that detected the electrical disturbance first, or detected the disturbance signal with the strongest and most obvious characteristics). Power grid smart terminals, such as feeder terminal units (FTUs), are typically deployed near critical nodes or equipment in the distribution network. When equipment without communication capabilities is manually operated, the resulting electrical disturbances are first captured by smart terminals near the point of operation. By comparing the timestamps, signal amplitudes, or clarity of characteristics detected by different smart terminals, the initial location or main center of influence of the disturbance can be preliminarily determined. For example, if multiple FTUs detect the same disturbance, but one FTU's signal amplitude is significantly higher than the others, or its detection time is the earliest, then the equipment monitored by that FTU or its surrounding area is likely the direct source of the disturbance.

[0066] Secondly, regarding the analysis of disturbance propagation paths, electrical disturbances do not exist in isolation within the power grid but propagate along the physical connection paths of the grid. By analyzing the order in which different smart terminals detect disturbance signals and their signal characteristics (such as attenuation and distortion), the propagation direction and path of the disturbance in the power grid can be inferred. For example, if the disturbance signal is detected sequentially by FTU1, FTU2, and FTU3, and FTU1 is upstream of FTU2, and FTU2 is upstream of FTU3, this indicates that the disturbance may propagate from FTU1 to FTU3. Combining this with the power grid topology information, the propagation path of the disturbance can be plotted, thereby determining the sequence of lines and equipment affected by the disturbance. This analysis helps to understand how disturbances spread within the power grid and provides a basis for defining their scope of impact.

[0067] Finally, the analysis of the impact range of the disturbance aims to determine all devices and lines in the power grid actually affected by the electrical disturbance. This includes not only the disturbance source itself, but also all devices and lines monitored by smart terminals that detected the disturbance signal and whose signal strength reached a certain threshold. By mapping all smart terminals that detected the disturbance and the devices they monitored onto the power grid topology, a continuous, affected area of ​​the power grid can be identified. The boundary of this area is usually determined by the device or line monitored by the smart terminal that detected the disturbance at the furthest point, or by the switches or nodes that logically separate the affected area from the unaffected area. For example, if a disturbance caused by the operation of a disconnecting switch is detected by all FTUs on the same feeder, but not by FTUs on adjacent feeders, then that feeder is the main area affected by the disturbance.

[0068] In summary, the delineation and definition of non-faulty electrical disturbance areas are achieved by dynamically identifying disturbance sources, tracing propagation paths, and determining the geographical or logical scope encompassing all affected equipment and lines through real-time collected electrical data combined with the static topology model of the power grid. This multi-dimensional and dynamic analysis method ensures accurate delineation of disturbance areas, providing precise spatial and topological context for subsequent inferences about human intervention behaviors, updates to the power grid digital model, and fault diagnosis and isolation strategy generation.

[0069] Furthermore, the "non-electrical accompanying information related to the area where the non-faulty electrical disturbance occurs" refers to non-electrical information that can provide contextual support or evidence for determining whether human intervention occurred within the area where the non-faulty electrical disturbance occurred. This relevance is reflected in the temporal, spatial, and logical matching of information. Specific meanings include: 1. Relevance of Operation and Maintenance Work Plan Information: If the time of a non-faulty electrical disturbance overlaps with the operation time window of a certain device in the operation and maintenance work plan, and the location of the disturbance is consistent with the location of the planned operation device, then the plan information is considered to be related to the disturbance.

[0070] 2. Relevance of equipment geographical location information: If the geographical location of the area where the non-faulty electrical disturbance occurs is close to the geographical location of equipment that does not have communication function (such as disconnecting switches), then the geographical location information of the equipment is considered to be related to the disturbance.

[0071] 3. Correlation of regional load status information: If the load change trend in a region is consistent with the typical load change pattern caused by a specific manual operation (such as the opening and closing of a switch) when a non-faulty electrical disturbance occurs, then the load status information is considered to be related to the disturbance.

[0072] In step S33, the probability prediction of an operation is an assessment of the probability of a specific manual operation occurring in the current context. This prediction can be based on a comprehensive analysis of maintenance work plan information, equipment geographical location information, and regional load status information. For example, if a disconnector switch operation is scheduled in a certain area's maintenance plan, and a real-time electrical disturbance occurs near that disconnector switch, the probability prediction of this operation will be significantly increased.

[0073] In step S34, the correction process aims to fuse the comparison results based purely on electrical features with the predictions based on non-electrical context information to improve the accuracy of the inference. This correction can employ weighted fusion, Bayesian inference, or other multi-source information fusion algorithms. By introducing non-electrical accompanying information, misjudgments caused by the superposition of electrical disturbances or similarity can be effectively reduced, and the ability to identify genuine human operation behaviors can be improved.

[0074] In step S35, the system will select the operation with the highest confidence or the highest probability as the final inference result based on the confidence or probability of each operation in the corrected comparison results.

[0075] This application's solution combines purely electrical feature-based comparison with contextual analysis of non-electrical accompanying information to form a multi-dimensional and more robust inference mechanism. After electrical data is collected by the smart grid terminal, the system first identifies and extracts real-time electrical features of non-faulty electrical disturbances. These features are then initially compared with a pre-established electrical feature disturbance fingerprint database to identify potential human intervention behaviors. However, due to the complexity of grid operation, a single electrical feature comparison may be affected by the superposition of multiple non-faulty events, leading to uncertainty in the initial comparison results. To overcome this limitation, this application introduces non-electrical accompanying information, such as operation and maintenance work plans, equipment geographical location, and regional load status. This information provides additional context related to the area where the electrical disturbance occurs. The system uses this non-electrical information to generate an operation probability prediction, assessing the probability of a specific operation behavior occurring in the current context. Subsequently, this operation probability prediction is used to correct the initial comparison results. By fusing the similarity of electrical feature comparisons with the contextual support of non-electrical information, a more accurate and reliable corrected comparison result is obtained. Ultimately, based on this revised comparison result, the system can more accurately infer the human operation behavior of devices without communication capabilities. This method effectively solves the problem that relying solely on electrical characteristics is insufficient to accurately identify human operation behavior in complex power grid environments, thus providing a more solid and accurate foundation for subsequent updates to the power grid digital model, fault diagnosis, and isolation strategy generation.

[0076] The following is a concrete example to illustrate this. Suppose that in a certain power distribution area, the system detects a non-faulty electrical disturbance in real time and extracts its real-time electrical characteristics. The system compares these real-time electrical characteristics with an electrical disturbance fingerprint database. The initial comparison results show that the similarity of "disconnector A is open" is 0.6, while the similarity of "load switching" is 0.55. Both are below the preset confidence threshold of 0.8, indicating that it is difficult to determine the specific operation based solely on electrical characteristics. At this point, the system will obtain non-electrical accompanying information related to the area where the electrical disturbance occurred. For example, the system may find maintenance work plan information showing that a planned maintenance operation was scheduled at disconnector A in the area that afternoon; equipment geographical location information confirms that disconnector A is located within the area where the disturbance occurred; and area load status information shows that the load in the area did not show significant abnormal fluctuations during the period when the disturbance occurred.

[0077] The system integrates this non-electrical accompanying information to calculate a comprehensive probability score for each possible event. For example, for the event "disconnector A is open," the system can use a weighted summation model for calculation: Score = w1 * Time_Match_1 + w2 * Geo_Proximity_1 + w3 * Env_Exclusion_1 + w4 * Load_Trend_Match_1. Here, Time_Match_1 represents the degree of matching between the maintenance work plan information and the time of the disturbance; Geo_Proximity_1 represents the degree of matching (proximity) between the equipment's geographical location information and the location of the disturbance; Env_Exclusion_1 represents the exclusion factor for environmental factors (such as weather); and Load_Trend_Match_1 represents the degree of matching between the regional load status information and the preset typical load characteristics. By assigning preset weights w1, w2, w3, and w4 (which can be preset through expert experience) and normalizing each matching degree or exclusion factor, the calculated operational probability prediction score for "disconnector A is open" is as high as 0.9.

[0078] Specifically, the calculation process for Time_Match_1 can be as follows: 1. The system first identifies all planned operations related to the "disturbance occurrence area" from the "Operation and Maintenance Work Plan Information" that are within the current time window (e.g., within a few hours or a day before or after).

[0079] 2. For each identified planned operation, determine whether the "disturbance occurrence time" falls within its "planned operation time window".

[0080] 3. If the “disturbance occurrence time” falls precisely within the planned time window, the matching degree can be set to a high value (e.g., 1.0).

[0081] 4. If the "disturbance occurrence time" falls outside the planned time window but is still within the acceptable "tolerable time range" (e.g., 30 minutes before or after the planned window), the degree of matching can be calculated using a decay function based on the magnitude of the time deviation. For example, an inverse proportional function or an exponential decay function can be used; the larger the time deviation, the lower the degree of matching.

[0082] 5. If the “Disturbance Occurrence Time” is completely outside the planned time window and tolerance time range, the matching degree is set to a low value (e.g., 0.0).

[0083] 6. The final Time_Match_1 value is usually the highest match calculated among all relevant planned operations, and then normalized so that its value range is between [0,1].

[0084] The calculation process for Geo_Proximity_1 can be as follows: 1. The system first determines the center point or boundary of the "disturbance occurrence area".

[0085] 2. For all equipment within or near the "disturbance zone" that may be manually operated (e.g., disconnect switches without communication capabilities), calculate the spatial distance between their geographical location and the center point of the "disturbance zone". The spatial distance can be calculated using Euclidean distance or a more precise spherical distance (such as the Haversine formula).

[0086] 3. Convert the calculated spatial distance into a proximity score. This is usually done using an inverse distance weighting function (e.g., 1 / (1+distance)) or a Gaussian decay function, where the smaller the distance, the higher the proximity score.

[0087] 4. A "maximum association distance threshold" (e.g., 500 meters) can be set. If the device distance exceeds this threshold, the proximity score can be set to a very low value (e.g., 0.0).

[0088] 5. The final Geo_Proximity_1 value is usually the highest proximity score calculated among all relevant devices, and then normalized so that its value range is between [0,1].

[0089] The calculation process for Env_Exclusion_1 can be as follows: 1. The system monitors environmental data in real time to identify whether there are severe weather or special environmental conditions that may hinder or prohibit manual on-site operations.

[0090] 2. Pre-define a mapping rule between environmental conditions and exclusion factors. For example: If there are extreme weather conditions such as "lightning warning", strong winds (exceeding the preset threshold) or heavy rain, these conditions will usually make on-site manual operation unsafe or prohibited. In this case, Env_Exclusion_1 can be set to a low value (e.g., 0.0 or a very small value close to 0.0) to indicate that the operation is almost impossible to occur under the current environment.

[0091] If environmental conditions are normal and no factors are found to hinder manual operation, Env_Exclusion_1 can be set to a high value (e.g., 1.0), indicating that environmental factors do not constitute exclusion.

[0092] 3. Env_Exclusion_1 is a multiplier factor, and its value range is usually between [0,1], where 0 represents complete exclusion and 1 represents no exclusion.

[0093] The calculation process for Load_Trend_Match_1 can be as follows: 1. The system analyzes the real-time load data of the "disturbance occurrence area" before and after the "disturbance occurrence time" and extracts the characteristics of its load change trend, such as the amplitude, duration, rate of change, and step characteristics of the load change.

[0094] 2. Compare the extracted real-time load change trend features or load curve segments with the load patterns of various known non-fault events stored in the "Typical Load Change Trend Library".

[0095] 3. The comparison can employ various pattern recognition technologies: Correlation analysis: Calculate the correlation coefficient between real-time load trends and typical load trends.

[0096] Dynamic Time Warping (DTW): Measures the similarity between two load curves that may exhibit scaling or contraction along the time axis.

[0097] Feature matching: Compare the extracted load change features (such as step amplitude and rise time) with the feature thresholds or ranges of typical events.

[0098] 4. The higher the matching degree, the more similar the real-time load change trend is to the load pattern of a typical non-fault event.

[0099] 5. The final Load_Trend_Match_1 value is usually the highest matching degree calculated among all typical non-faulty events, and then normalized so that its value range is between [0,1].

[0100] Subsequently, the system corrects the preliminary comparison results based on the predicted probability of the operation. The system obtains the confidence level of each operation behavior in the preliminary comparison results (e.g., "disconnector A is open" is 0.6, "load switching" is 0.55), and obtains the predicted confidence level of each operation behavior in the predicted probability of the operation (e.g., "disconnector A is open" is 0.9). The system determines the fusion weights based on these predicted confidence levels (determining fusion weights can be based on information entropy theory, which measures the uncertainty or randomness of information. The system can calculate the entropy value of the confidence distribution of each operation behavior in the preliminary comparison results, and the entropy value of the confidence distribution of each operation behavior prediction in the operation probability prediction. The lower the entropy value, the less uncertain the information is, meaning the information source is more reliable or has higher indicative power, and therefore can be assigned a higher fusion weight. For example, if the confidence distribution of the preliminary comparison results is relatively dispersed with high entropy values, while the confidence distribution of the operation probability prediction is relatively concentrated with low entropy values, the system will assign a higher weight to the operation probability prediction to allow it to play a greater role in the fusion calculation), and performs a weighted fusion calculation on them. For example, through fusion calculation, the final confidence level of "disconnect switch A is open" is increased to 0.88. The system selects the event with the highest final confidence level as the inference result. Since 0.88 exceeds the preset confidence level threshold of 0.85, the system ultimately infers that the manual operation behavior of the device without communication function is "disconnect switch A is open".

[0101] Through the above technical solution, this application effectively solves the technical problem that in complex power grid environments, the superposition of multiple non-faulty electrical events makes it difficult to accurately identify manual operation behaviors of equipment without communication functions based solely on electrical characteristics. By introducing non-electrical accompanying information such as operation and maintenance work plan information, equipment geographical location information, and regional load status information, and fusing and correcting them with the comparison results of electrical characteristics, this application can significantly improve the accuracy and reliability of inferring manual operation behaviors. This enables more accurate inputs in subsequent steps such as power grid digital model updates, fault diagnosis, and isolation strategy generation, thereby avoiding the disconnect between the digital model and physical reality caused by misjudgment or omission of manual operations, and thus improving the overall efficiency and security of fault operation and maintenance management of distribution automation systems. Especially in situations with complex power grid topology, increased distributed power source access, and variable power flow direction, this solution can more accurately identify the impact of manual operations on the power grid state, providing a strong guarantee for the stable operation of the power grid.

[0102] In some embodiments, the specific steps in step S33 include: S331. Assign weights to the operation and maintenance work plan information, equipment geographical location information, and regional load status information in the non-electrical accompanying information; S332. Based on the assigned weights, perform weighted calculations on the degree of matching between the operation and maintenance work plan information and the disturbance occurrence time, the degree of matching between the equipment geographical location information and the disturbance occurrence location, and the degree of matching between the regional load status information and the preset typical load characteristics, and generate an operational probability prediction based on the weighted calculation results.

[0103] Specifically, regarding the weighting of maintenance operation plan information, equipment geographical location information, and regional load status information within the non-electrical accompanying information, weighting refers to assigning different numerical coefficients to different non-electrical accompanying information based on their importance, reliability, or relevance in inferring human operational behavior. These weights reflect the system's level of trust in each information source or its degree of influence on the outcome. One implementation method is the expert experience method, where the weights of various information items are manually set or adjusted based on the experience of power system maintenance experts and historical data analysis. For example, maintenance operation plan information typically directly indicates the likelihood of an operation, thus it can be assigned a higher weight; equipment geographical location information provides spatial constraints for operations, so its weight is moderate; regional load status information reflects real-time operating conditions, and its weight can be adjusted based on its relevance to specific operational behaviors. Another implementation method is to utilize machine learning methods to train a model using historical data, automatically learning and optimizing the weights of various information items. For example, known operational behaviors can be used as labels, and various non-electrical accompanying information items can be used as features. A supervised learning algorithm (such as regression analysis, decision trees, or neural networks) can be used to train the model to obtain the optimal weight combination, thereby maximizing the accuracy of operational behavior inference.

[0104] Regarding the weighted calculation of the matching degree of operation and maintenance work plan information with the disturbance occurrence time, the matching degree of equipment geographical location information with the disturbance occurrence location, and the matching degree of regional load status information with preset typical load characteristics, the weighted calculation refers to multiplying the matching degree of each non-electrical accompanying information by its pre-assigned weight, and then summing the products to obtain a comprehensive evaluation value. For the matching degree of each piece of information, for example: The degree of match between maintenance work plan information and the timing of disturbances: For example, if an electrical disturbance is detected at a specific time, the system will query the maintenance work plan to determine if there are any plans to operate the disconnect switches in that area during that time period. The higher the degree of time matching, the more likely the electrical disturbance is related to planned manual operations.

[0105] The degree of matching between the device's geographical location information and the location of the disturbance: For example, if an electrical disturbance is located near a pole, the system checks whether there are any devices on or near that pole that lack communication capabilities and assesses their geographical proximity. The higher the degree of location matching, the more likely the electrical disturbance is related to human operation of that device.

[0106] The system compares the area's load status information to preset typical load characteristics. For example, it analyzes the area's load data before and after an electrical disturbance, observing its trend (e.g., whether the load drops suddenly or remains stable), and compares this trend with typical load change patterns stored in a fingerprint database caused by specific manual operations (such as switch opening and closing). It also compares the data with the area's normal load fluctuation trends to rule out other non-operational load changes. For instance, if opening a disconnector switch typically causes a momentary drop in downstream load, the system checks if real-time load data shows a similar downward trend. A higher degree of matching indicates that the electrical disturbance is more likely related to that manual operation.

[0107] These matching degrees are then normalized, multiplied by their corresponding weights, and summed. Alternatively, fuzzy logic reasoning can be employed. The matching degree of each piece of information is defined as a fuzzy variable, and a fuzzy rule base is established. Based on the assigned weights, weighted fuzzy reasoning is performed on the fuzzy variables to obtain a fuzzy prediction of operational probability, which is then defuzzified to obtain a precise prediction value.

[0108] Regarding the generation of the operation probability prediction based on the weighted calculation results, generating the operation probability prediction refers to quantitatively assessing the likelihood of a specific device performing a certain manual operation at a specific time point based on the weighted calculation results. This prediction result is typically a probability value or confidence score, used to subsequently correct the electrical feature comparison results. As one implementation method, a threshold judgment can be used. The weighted calculation result is compared with a preset threshold; if it exceeds the threshold, the operation probability is considered high; if it is below the threshold, the operation probability is considered low. The prediction result can be a discrete level such as "high," "medium," or "low," or a normalized score can be directly output. As another implementation method, the weighted calculation result can be input into a probability model (such as a sigmoid function) to convert it into a probability value between 0 and 1, directly representing the probability of the operation occurring.

[0109] This application's solution introduces a weighting mechanism, enabling the system to identify and quantify the relative importance of different non-electrical accompanying information in inferring human operational behavior. Specifically, when assigning weights to maintenance work plan information, equipment geographical location information, and regional load status information, the system can assign different influence factors based on the strength of the correlation, reliability, or timeliness of these information with human operational behavior. For example, a maintenance plan that explicitly indicates that a certain piece of equipment will be operated at a specific time should have a higher weight than geographical location information that only indicates the area where the equipment is located. Based on this, the system performs a weighted calculation of the matching degree of each piece of information according to these assigned weights. This means that information with higher weights will play a greater role in the comprehensive evaluation, and its matching degree will have a more significant impact on the final operational probability prediction result. For example, when the maintenance work plan information is highly matched with the real-time operation time, even if the support provided by the equipment geographical location information or regional load status information is low, the high-weighted maintenance plan information can effectively improve the confidence of the operational probability prediction. Ultimately, through this weighted calculation, the system can generate a more accurate and reliable operational probability prediction. This approach, combined with the aforementioned technique of generating operational probability predictions based on non-electrical accompanying information to correct preliminary comparison results, significantly improves the accuracy and robustness of the overall method. When preliminary comparison results may be uncertain due to the complexity or ambiguity of electrical characteristics, a refined weighted operational probability prediction can more accurately reflect the probability of actual operations. For example, when the electrical characteristic comparison results have low confidence in a certain operational behavior, if the weighted operational probability prediction shows that the operational behavior is highly likely (e.g., because there is a clear maintenance plan), the system can more confidently correct the preliminary comparison results, thereby avoiding misjudgments or omissions. This mechanism ensures that when inferring the manual operation behavior of devices without communication capabilities, the system not only considers electrical disturbance characteristics but also fully utilizes the complementarity and importance differences of multi-source non-electrical accompanying information, making the final inference results closer to physical reality and effectively solving the problem of inaccurate predictions caused by uneven information importance.

[0110] The following is a concrete example. Suppose the system needs to determine whether a disconnector switch S101 has been manually disconnected. When assigning weights to the maintenance work plan information, equipment geographical location information, and regional load status information in the non-electrical accompanying information, the system can, based on historical data and expert experience, set the weight q1 of the maintenance work plan information to 0.5, the weight q2 of the equipment geographical location information to 0.3, and the weight q3 of the regional load status information to 0.2. Based on these assigned weights, the system performs a weighted calculation of the matching degree of each piece of information and generates an operational probability prediction. Specifically, the system can calculate a comprehensive probability score, for example: Operational Probability Prediction Score = q1 * Time_Match_2 + q2 * Geo_Proximity_2 + q3 * Load_Trend_Match_2. Wherein, Time_Match_2 represents the degree of matching between the disturbance occurrence time and the S101 operation time in the maintenance work plan; Geo_Proximity_2 represents the degree of matching (proximity) between the disturbance occurrence location and the geographical location of the S101 equipment; Load_Trend_Match_2 represents the degree of matching between the load change trend of the area where S101 is located and the typical load characteristics of the disconnecting switch operation. For example, if the disturbance occurs within the planned maintenance time window of S101 (Time_Match_2=1.0), the real-time electrical disturbance source is located near the physical location of S101 (Geo_Proximity_2=0.9), and the load in the area decreases at the time of the disturbance, consistent with the disconnecting switch operation (Load_Trend_Match_2=0.8), then the operation probability prediction score = 0.5*1.0+0.3*0.9+0.2*0.8=0.5+0.27+0.16=0.93. This 0.93 represents the predicted probability of S101 being manually disconnected. As another specific implementation, if the system detects an exclusion factor strongly correlated with a specific operational behavior in the non-electrical accompanying information—for example, if the maintenance work plan explicitly states that the equipment is not allowed to operate during the current time period—an exclusion factor Env_Exclusion_2 can be introduced. In this case, the predicted probability score can be adjusted as: Predicted Probability Score = (w1*Time_Match_2 + w2*Geo_Proximity_2 + w3*Load_Trend_Match_2)*Env_Exclusion_2. If Env_Exclusion_2 is 0 (indicating explicit exclusion), then regardless of how high the matching degree of other factors is, the final prediction score will be 0, thus effectively avoiding misjudgment. It should be noted that the value of Env_Exclusion_2 is either 0 or 1; a value of 1 indicates no exclusion (i.e., the equipment is allowed to operate during the current time period).

[0111] Through the above technical solution, this application effectively solves the problem of inaccurate prediction of operational probability caused by the uneven importance of non-electrical accompanying information in traditional methods. By assigning weights to maintenance operation plan information, equipment geographical location information, and regional load status information, and performing weighted calculations based on these weights, the system can more precisely evaluate the impact of each piece of information on the inference of human operation behavior. This allows the system to prioritize non-electrical accompanying information that is more indicative, reliable, or relevant when generating operational probability predictions, thereby significantly improving the accuracy and reliability of the predictions. Specifically, when combined with the above-mentioned technical solution that compares real-time electrical features with an electrical feature perturbation fingerprint database and infers the human operation behavior of equipment without communication functions based on the comparison results, this weighted and optimized operational probability prediction can more effectively correct the initial comparison results. For example, even if the electrical feature comparison results have some ambiguity, if the high-weighted maintenance operation plan information clearly indicates a certain operation, the system can more accurately identify the operation, avoiding misjudgments caused by insufficient information from a single source. This not only improves the accuracy of inferring human operational behavior, but also enhances the robustness of the entire power distribution automation fault operation and maintenance management method, enabling the power grid digital model to reflect physical reality more timely and accurately, providing a solid foundation for subsequent fault diagnosis and isolation strategy generation, and ultimately ensuring the safe and stable operation of the power grid.

[0112] In some embodiments, the specific steps in step S34 include: S341. Obtain the confidence level of each operation in the preliminary comparison results; S342. Obtain the confidence level of each operational behavior in the operational probability prediction; S343. Determine the fusion weights based on the confidence levels of each operational behavior in the preliminary comparison results and the prediction confidence levels of each operational behavior in the operational probability prediction. S344. Based on the fusion weight, the confidence of each operation in the preliminary comparison results and the prediction confidence of each operation in the operation probability prediction are weighted and fused to obtain the corrected comparison results.

[0113] Specifically, when obtaining the confidence level of each operation in the initial comparison results, this step aims to quantify the reliability of each possible operation obtained based on electrical feature comparison. This ensures that subsequent correction processes prioritize operations supported by high confidence from electrical data. For example, this can be achieved by calculating the similarity between real-time electrical features and fingerprints in an electrical feature perturbation fingerprint database, and normalizing the similarity values ​​to a confidence score between 0 and 1. Alternatively, a pre-trained machine learning model, such as a support vector machine or neural network, can be used to classify the electrical feature comparison results and output the posterior probability of each operation as its confidence level.

[0114] When obtaining the predicted confidence level for each operational behavior in the operational probability prediction, this step aims to quantify the degree to which non-electrical accompanying information supports the probability of a specific operational behavior occurring. This allows the correction process to incorporate external contextual factors, such as maintenance plans or regional load status. For example, based on a pre-defined rule engine, the corresponding predicted confidence level can be calculated according to the degree of matching between maintenance work plan information, equipment geographical location information, and regional load status information and specific operational behaviors. Alternatively, a Bayesian network or decision tree model can be used, taking this non-electrical accompanying information as input and outputting the probability of occurrence of each operational behavior as the predicted confidence level.

[0115] When determining the fusion weights based on the confidence levels of each operational behavior in the preliminary comparison results and the predicted confidence levels of each operational behavior in the predicted operational probability, this step aims to dynamically adjust the influence of the two information sources in the final corrected result to address differences in information reliability under different scenarios. For example, based on information entropy theory, the entropy values ​​of the confidence distribution of the preliminary comparison results and the confidence distribution of the predicted operational probability can be calculated, and the side with the lower entropy value (less information uncertainty) can be assigned a higher weight. Alternatively, fuzzy logic reasoning can be used to dynamically adjust the weight allocation based on the relative levels of the confidence levels of the preliminary comparison results and the predicted confidence levels, as well as the degree of difference between them, through a pre-set fuzzy rule base. In addition, a machine learning model can be trained using historical data. This model takes the preliminary comparison confidence level, the predicted confidence level, and other relevant contextual information as input and outputs the optimal weight combination for weighted fusion.

[0116] It should be noted that in step S31, the system compares the real-time electrical characteristics with the electrical characteristic disturbance fingerprint database. This comparison process does not produce a single result, but rather calculates a preliminary confidence level for each preset operation behavior with a high degree of matching in the fingerprint database, for all or some of them. Therefore, the "confidence level of each operation behavior in the preliminary comparison result" is actually a list containing multiple candidate operation behaviors and their respective confidence levels. For example, the comparison result may show that the confidence level of "disconnecting switch A is open" is 0.6, the confidence level of "load switching" is 0.55, and so on.

[0117] Similarly, in step S33, the system generates an operational probability prediction based on non-electrical accompanying information. This prediction process also assesses the probability of a series of potential operational behaviors occurring in the current context. For example, by combining maintenance work plan information, equipment geographical location information, and regional load status information, the system calculates a prediction confidence level for operational behaviors such as "disconnecting switch A is open," "load switching," and "capacitor bank switching." Therefore, "the prediction confidence level of each operational behavior in the operational probability prediction" is also a list containing multiple candidate operational behaviors and their respective prediction confidence levels.

[0118] When determining the fusion weights and performing weighted fusion calculations, the system processes these common candidate operations. This means that the system maintains a unified list of candidate operations internally. For each operation in the list, the system obtains its confidence level in the initial comparison results and its predicted confidence level in the operation probability prediction. If an operation appears only in one list but has no explicit confidence level or predicted confidence level in another list, the system will assign it a very low value (e.g., 0 or close to 0) by default to ensure that all candidate operations can participate in the fusion calculation, thereby achieving a logical one-to-one correspondence.

[0119] Therefore, the design of the steps in this application ensures a clear correspondence between the confidence levels of each operational behavior in the preliminary comparison results and the predicted confidence levels of each operational behavior in the operational probability prediction. This correspondence is based on the system's unified management and evaluation of all potential operational behaviors. In this way, the system can independently fuse the confidence levels of each candidate operational behavior to obtain a corrected comparison result, and then infer the most reliable human operational behavior. This multi-source information fusion mechanism can improve the accuracy and reliability of the final inference even when there is ambiguity or uncertainty in a single information source (such as electrical feature comparison), by introducing complementary non-electrical accompanying information for correction.

[0120] When calculating the weighted fusion of the confidence levels of each operational behavior in the preliminary comparison results and the predicted confidence levels of each operational behavior in the operational probability predictions, based on fusion weights, to obtain the corrected comparison results, this step aims to integrate the two confidence levels to generate a more accurate and robust final inference result, thereby improving the accuracy and reliability of inferences about human operational behaviors. For example, a linear weighted summation method can be used, where the corrected confidence level equals the weighted average of the confidence levels of the preliminary comparison results and the operational probability predictions, with the sum of the weights being 1. Alternatively, a product fusion method can be used, where the corrected confidence level equals the product of the confidence levels of the preliminary comparison results and the operational probability predictions, with exponential adjustments based on the weights. Furthermore, the Dempster-Shafer evidence theory (DS evidence theory) can be applied, treating the preliminary comparison results and operational probability predictions as different sources of evidence, and fusing them through evidence combination rules to handle uncertainties and conflicting information.

[0121] This application optimizes the correction process of comparison results by introducing a confidence level acquisition and dynamic weighted fusion mechanism, thereby more reliably inferring human operation behavior and reducing errors caused by information conflicts. Specifically, acquiring the confidence level of each operation behavior in the preliminary comparison results provides a reliability quantification index based on electrical feature comparison, ensuring that the correction process prioritizes high-confidence electrical data; acquiring the prediction confidence level of each operation behavior in the operation probability prediction provides a probability assessment based on non-electrical accompanying information, allowing the correction to incorporate external factors such as operation and maintenance plans or load status. Based on this, the fusion weight is determined according to the confidence level of the preliminary comparison results and the operation probability prediction confidence level, allowing dynamic adjustment of the weight ratio to avoid fixed weights amplifying errors when information is inconsistent. Finally, a weighted fusion calculation is performed based on the fusion weight, integrating the advantages of electrical and non-electrical information to generate a more balanced correction result and improve the robustness of the overall inference. This scheme, by comparing real-time electrical characteristics with an electrical feature disturbance fingerprint database to obtain preliminary comparison results and acquiring non-electrical accompanying information to generate operational probability predictions, effectively solves the limitations or uncertainties that may exist with a single information source by intelligently fusing these two information sources. When the preliminary comparison results show multiple low-confidence matches due to composite signals, or when the confidence of a single match is not high, the scheme of this application can introduce an operational probability prediction score to correct the matching results. This provides a more comprehensive and reliable decision-making basis when there are conflicts or uncertainties in the information, significantly improving the accuracy and reliability of inferring the manual operation behavior of equipment without communication functions, and thus laying a solid foundation for subsequent updates to the power grid digital model, fault diagnosis, and the generation of isolation strategies.

[0122] The following is a specific example to illustrate this. Suppose that after the system detects a non-faulty electrical disturbance in real time, it compares the real-time electrical features with the electrical feature disturbance fingerprint database and initially infers that the confidence level of the operation "disconnector A is open" is 0.6, which is lower than the preset confirmation threshold of 0.8. This means that the system cannot be completely certain that the operation has occurred based solely on the electrical features. Simultaneously, the system obtains non-electrical accompanying information related to the area where the non-faulty electrical disturbance occurred. For example, maintenance work plan information shows that there is a planned maintenance task for "disconnector A" on that day; equipment geographical location information confirms that "disconnector A" is located within the maintenance area; and area load status information shows that the area's load has decreased, consistent with a planned power outage operation. Based on this non-electrical accompanying information, the system generates a predicted confidence level of 0.9 for the operation "disconnector A is open." At this point, the system needs to fuse these two confidence levels. The system will determine the fusion weight based on the initial comparison result confidence level (0.6) and the predicted operation probability confidence level (0.9). For example, considering the low confidence level of electrical feature comparison and the high support of external context information, the system can dynamically assign a higher weight to the predicted operational probability. For instance, the weight of the initial comparison result confidence level is 0.4, and the weight of the predicted operational probability confidence level is 0.6. Then, the system performs a weighted fusion calculation: the corrected confidence level = 0.4 * 0.6 + 0.6 * 0.9 = 0.24 + 0.54 = 0.78. If the system's preset final confirmation threshold is 0.85, and 0.78 is still below this threshold, the system will issue an alarm, prompting maintenance personnel to conduct manual verification. However, if the fused confidence level exceeds 0.85, for example, if the confidence level of electrical feature comparison is 0.7, the predicted operational probability confidence level is 0.95, and the fusion weights are 0.5 and 0.5 respectively, then the corrected confidence level = 0.5 * 0.7 + 0.5 * 0.95 = 0.35 + 0.475 = 0.825. At this point, if the system sets the final confirmation threshold to 0.8, then 0.825, being higher than this threshold, will be confirmed as a manual operation. In this way, even if the confidence level of a single information source is low, by dynamically weighting and fusing information from multiple sources, the system can still improve the final confidence level of its inference about manual operation behavior, thus making a more accurate judgment.

[0123] Through the above technical solution, this application effectively solves the problem of deviation in correction results caused by information conflicts or inconsistencies when fusing preliminary comparison results and confidence information of operational probability prediction. By dynamically acquiring and fusing the two confidence levels and determining the fusion weight based on their reliability, this application can more accurately and reliably infer the manual operation behavior of devices without communication capabilities. This significantly improves the consistency between the power grid digital model and physical reality, thus providing a more accurate and reliable foundation for subsequent fault diagnosis and isolation strategy generation, avoiding misjudgments and unnecessary expansion of power outage areas due to model inaccuracies, and improving the operation and maintenance efficiency of the distribution automation system and the safety of power grid operation.

[0124] In some embodiments, the specific steps in step S4 include: S41. Based on the preset power grid topology rules and equipment linkage logic, and according to the inferred human operation behavior, deduce the impact of the operation behavior on the logical state of the corresponding equipment in the power grid digital model, as well as the logical state and connection relationship of the affected equipment. S42. Based on the derived impact, update the logical state of the corresponding equipment in the power grid digital model, as well as the logical state and connection relationship of the affected equipment.

[0125] Among them, the predefined power grid topology rules refer to predefined structural constraints describing the physical connections and logical relationships between devices in the power grid. These rules ensure the rationality and consistency of the power grid model. For example, they could be a series of graph theory-based rules stipulating that there cannot be more than one direct connection path between any two switches, or that a bus node must be connected to at least one switch or transformer. Alternatively, they could be rules based on power system engineering specifications, such as specifying the connection methods between devices of different voltage levels, or restricting the specific structure of ring or radial networks. Device linkage logic refers to predefined rules that change the state or connection relationship of other related devices when the state of one device in the power grid changes. For example, it could be a set of logic based on relay protection principles, where when a circuit breaker trips, its upstream or downstream disconnecting switches may need to perform corresponding operations to isolate the fault. It could also be logic based on operating procedures, such as when a bypass switch closes, the main switch connected in parallel with it must disconnect to avoid short circuits or unsafe operating modes. The system derives the impact of operational actions on the logical states of corresponding devices in the power grid digital model, as well as the logical states and connections of affected devices. This aims to comprehensively assess the impact of a manual operation on the entire power grid digital model, not just the operating device itself, ensuring the integrity and accuracy of model updates. The system can utilize preset power grid topology rules and device linkage logic to traverse relevant nodes and edges in the power grid digital model, simulating state changes after an operational action occurs, thereby identifying all affected devices and their new logical states and connections. Alternatively, a rule-based reasoning engine can be used, encoding the power grid topology rules and device linkage logic into a series of production rules. When a manual operation is inferred, the rule engine is triggered to automatically derive the states and connections of all affected devices. Updating the logical states of corresponding devices in the power grid digital model, as well as the logical states and connections of affected devices, involves actually modifying the data in the power grid digital model based on the impact derived in the previous step, ensuring consistency with physical reality. The system can directly modify the attribute fields of corresponding and affected devices in the digital model, such as updating the switch state from "closed" to "open" and adjusting the connectivity flags of relevant connection points. Alternatively, a transactional update mechanism can be used to package all the states and connections that need to be updated into a single transaction, ensuring that all updates either succeed or are rolled back, in order to maintain the data consistency of the digital model.

[0126] This application's solution adjusts the logical state of the corresponding device in the internally maintained power grid digital model immediately after inferring the manual operation behavior of a device lacking communication capabilities. This digital model is typically stored in memory as a graph structure, where each node represents a device or connection point, and each edge represents a line or switch. Further, based on preset power grid topology rules and device linkage logic, the method derives the impact of this operation behavior on the logical state of the corresponding device in the power grid digital model, as well as the logical state and connection relationships of the affected devices. These rules and logic serve as a knowledge base for power grid structure and device interaction, enabling the system to simulate the consequences of the operation behavior in the digital model, thereby identifying all affected devices and their new logical states and connection relationships. Subsequently, the system updates the logical state of the corresponding device in the power grid digital model, as well as the logical states and connection relationships of all affected devices, based on these derived impacts. This systematic approach ensures that the digital model accurately reflects the physical reality of the power grid after manual operation, thus avoiding inconsistencies that could lead to incorrect fault diagnosis or isolation strategies. This goes beyond simple point-to-point state updates when updating the power grid digital model, ensuring the integrity of the digital model across the entire affected network segment, which is crucial for subsequent fault diagnosis and isolation strategy generation. Without this comprehensive update, even if manual operations are correctly identified, the digital model will still be inaccurate, thereby compromising the reliability of the entire fault management process.

[0127] In one specific implementation, after inferring the manual operation of disconnecting switch S1 being opened, the system does not simply update the state of S1 from "closed" to "open". Instead, it first deduces the full impact of this operation based on preset grid topology rules and equipment linkage logic. For example, preset grid topology rules may stipulate that after a disconnecting switch is opened, its connected downstream lines will no longer be energized unless another power source is connected. Equipment linkage logic may further stipulate that if S1 is the main disconnecting switch on a substation outgoing circuit, its opening may mean that all downstream load switches and user-side equipment on that circuit will lose power. Specifically, the system first locates disconnecting switch S1 in the grid digital model and marks its logical state from "closed" to "open". Subsequently, the system traverses the downstream topology path of S1. If S1 is connected to a feeder and there is no other power source on that feeder, the system deduces that all load switches, branch switches, and distribution transformers and user equipment connected to these switches on that feeder will be de-energized. Based on these derivation results, the system updates the logical states of the affected devices in the power grid digital model. For example, it updates the state of the load switch to "power outage" and the state of the user equipment to "no power supply," and modifies the relevant connection relationships. For example, it marks the connection between S1 and the downstream feeder as "disconnected" and the connection between the feeder and the downstream equipment as "no power connection." This process ensures that the power grid digital model can accurately reflect the electrical connectivity and equipment status of the entire affected area after the disconnecting switch S1 is opened.

[0128] The proposed solution, when updating the power grid digital model, comprehensively deduces the impact of manual operation on the logical state and connection relationships of corresponding and affected equipment based on preset power grid topology rules and equipment linkage logic, and updates accordingly. This allows the power grid digital model to more accurately and comprehensively reflect the actual operating state of the power grid. In this way, the disconnect between the digital model and physical reality is effectively resolved, significantly improving the accuracy and reliability of the power grid digital model. In subsequent fault diagnosis and isolation strategy generation processes, based on the highly accurate digital model, the system can more precisely locate fault points and formulate more reasonable and efficient isolation strategies, thereby reducing the scope and duration of power outages and improving the operation and maintenance efficiency and power supply reliability of the distribution automation system.

[0129] In some embodiments, step S5, the step of verifying the updated power grid digital model, includes: S51. Obtain multi-source verification information related to the updated state of the power grid digital model; the multi-source verification information includes real-time electrical data, power grid topology information, operating procedures, and planned maintenance task information; S52. Perform comprehensive analysis on the multi-source verification information to obtain the analysis results; S53. Based on the analysis results, assess the consistency between the updated digital power grid model state and physical reality.

[0130] Specifically, the multi-source verification information refers to a collection of data from various sources and of different types used to verify the accuracy of the digital power grid model's status. This can be obtained, for example, from real-time electrical data acquired through a data interface from a SCADA / DMS system, power grid topology information from a GIS system, and operating procedures and planned maintenance task information from an operation and maintenance management system. Alternatively, it can utilize local environmental data collected by an IoT sensor network, historical operating data analysis results, and manual inspection reports as supplementary information sources. The real-time electrical data refers to the electrical quantity measurements and status information of each device in the power grid at the current moment. This can be obtained, for example, through telemetry and telesignaling data periodically reported by smart terminals (such as feeder terminal units, distribution terminal units, and transformer terminal units), or through voltage, current, power, and frequency data directly collected by smart meters, sensors, and other devices. The power grid topology information refers to structured data describing the physical and logical connections between devices in the power grid. This can be obtained, for example, from line maps, equipment coordinates, and connection point information acquired from a Geographic Information System (GIS) or Distribution Management System (DMS), or through node-edge graph structure data automatically generated based on equipment connection relationships using topology analysis algorithms. The operating procedures refer to the standard procedures, rules, and safety requirements that guide the operation and maintenance of power grid equipment. These can be, for example, various operation tickets, safety regulations, and dispatch instructions stored in electronic document form, or the key logic and conditions in the procedures encoded into a set of rules executable by a rule engine. The planned maintenance task information refers to the details of pre-arranged activities for inspecting, maintaining, repairing, or replacing power grid equipment. This can be, for example, information obtained from a production management system or maintenance planning system, such as maintenance work orders, outage scope, operation time, and equipment involved, or a maintenance task list stored in a structured database, including fields such as task ID, equipment ID, planned start / end time, and maintenance type.

[0131] It should be noted that the "multi-source verification information related to the updated state of the power grid digital model" refers to information that can be used to verify whether the updated device logic states and connection relationships in the power grid digital model are consistent with physical reality. This relevance is reflected in the fact that the information can directly or indirectly reflect the correctness of the model update. Specifically, this includes: 1. Correlation of real-time electrical data: If the state of a switch in the digital model is updated to "open", the real-time electrical data associated with that switch (such as downstream current) should be zero, and the upstream voltage should remain normal.

[0132] 2. Relevance of power grid topology information: If a connection in the digital model changes, the change should be consistent with the actual physical topology of the power grid. For example, the disconnection of a line should not lead to unreasonable topological islands.

[0133] 3. Relevance of operating procedures: The equipment operation reflected in the digital model should comply with the power grid's operating procedures. For example, the operation time and operation type should be within the range allowed by the procedures.

[0134] 4. Relevance of planned maintenance task information: The equipment operations reflected in the digital model should be consistent with the known planned maintenance tasks. For example, the disconnection operation of a disconnecting switch should correspond to the planned power outage maintenance task of the equipment.

[0135] Comprehensive analysis of the multi-source verification information refers to cross-referencing, logical reasoning, and data fusion of information from different sources and of different types to discover correlations, conflicts, or consistency among the information. This can be achieved, for example, by using a rule engine for logical judgment, checking whether the inferred operations conform to the operating procedures' time windows, safety distances, and other regulations; and by comparing real-time electrical data with topology information to verify the electrical logic consistency of equipment status. Alternatively, data mining or machine learning algorithms can be used to learn patterns from multi-source data and identify anomalies or inconsistencies, such as using cluster analysis to identify electrical data patterns that do not match planned maintenance tasks. Based on the analysis results, evaluating the consistency between the updated power grid digital model status and physical reality refers to determining whether the equipment status and connection relationships in the power grid digital model accurately reflect the actual physical state of the power grid, based on the comprehensive analysis results. This can be achieved, for example, by comparing the analysis results with preset consistency evaluation indicators (such as consistency scores and deviation thresholds); if the indicators are met, consistency is considered achieved; otherwise, inconsistency is considered. Alternatively, an expert system or decision support system can be used to generate an evaluation report based on the analysis results and provide suggestions on the reliability of the model status.

[0136] This application's solution, after inferring the manual operation behavior of devices lacking communication capabilities and updating the logical state of the corresponding devices in the power grid digital model, marks this update as a "state to be verified" and initiates a lightweight verification process. This process first acquires multi-source verification information related to the updated power grid digital model state. This information covers real-time electrical data, power grid topology information, operating procedures, and planned maintenance tasks, thus providing a basis for a comprehensive verification of the model state. Subsequently, the system comprehensively analyzes this multi-source verification information. This includes verifying operational compliance, i.e., comparing whether the inferred operation conforms to the current power grid operating procedures and planned maintenance tasks; and checking electrical logic consistency, i.e., verifying whether the inferred operation matches the real-time state of other nearby communicable devices. For example, if a disconnector is inferred to be open, the system checks whether its downstream feeder terminal unit reports zero current and whether its upstream voltage remains normal. Finally, based on the results of the comprehensive analysis, the system evaluates the consistency between the updated power grid digital model state and physical reality. If the verification passes, the status update is confirmed as valid, and the status of the device in the digital model is permanently updated, thus correcting the disconnect between the digital model and physical reality. If the verification fails, the system will issue an alarm, prompting maintenance personnel to conduct manual verification, and temporarily maintain the original model status or mark it as an uncertain state. In this way, the solution proposed in this application provides a multi-dimensional and multi-level verification mechanism for the accuracy of the power grid digital model, effectively avoiding the disconnect between the model and physical reality caused by insufficient information from a single source or information conflicts, thereby providing a reliable foundation for subsequent fault diagnosis and isolation strategy generation.

[0137] The following is a concrete example to illustrate this. Suppose the system infers the manual operation of "disconnecting switch S1 being disconnected" by comparing it with an electrical feature disturbance fingerprint database, and updates the logical state of disconnecting switch S1 in the power grid digital model accordingly. To verify the accuracy of this update, the system initiates a verification process. First, the system acquires multi-source verification information, including: real-time current data from the downstream feeder terminal unit and real-time voltage data from the upstream feeder terminal unit of disconnecting switch S1 from the SCADA system; the geographical location of disconnecting switch S1 and its connected line topology information from the GIS system; and the operating procedures and planned maintenance tasks for disconnecting switch S1 or its area for the current period from the operation and maintenance management system. Next, the system performs a comprehensive analysis of this information. On one hand, the system verifies whether the disconnection operation of disconnecting switch S1 conforms to the current operating procedures, such as whether it is within the permitted operation time window and whether it matches the known planned maintenance tasks. On the other hand, the system checks the electrical logic consistency; for example, if disconnecting switch S1 is inferred to be disconnected, the system checks whether the downstream feeder terminal unit of S1 reports zero current and whether the upstream voltage of S1 remains normal. If all checks pass—for example, the operation complies with procedures, the downstream current of S1 is indeed zero, and the upstream voltage is normal—the evaluation result is "consistent," the status update of disconnector S1 is confirmed as valid, and the status of S1 in the digital model is permanently updated. Conversely, if any check fails—for example, the feeder terminal unit downstream of S1 still reports current—the evaluation result is "inconsistent," the system will issue an alarm, prompting maintenance personnel to conduct manual verification, and temporarily maintain the original model status or mark it as an uncertain state.

[0138] Through the above technical solution, this application significantly enhances the accuracy and reliability of the updated power grid digital model. By introducing multi-source verification information such as real-time electrical data, power grid topology information, operating procedures, and planned maintenance task information, and conducting comprehensive analysis, the consistency between the model state and physical reality can be thoroughly verified. This effectively solves the problem that traditional verification methods may rely on a single or insufficient information source, avoiding the disconnect between the digital model and physical reality caused by data entry errors, system compatibility issues, or deviations in operating procedures. Therefore, it ensures the accuracy of the power grid digital model in fault diagnosis and isolation strategy generation, reduces the risk of misjudgment and erroneous operation, and improves the overall operation and maintenance efficiency and safety of the distribution automation system.

[0139] Reference Appendix Figure 2 This invention provides a power distribution automation fault operation and maintenance management system (this power distribution automation fault operation and maintenance management system adopts the power distribution automation fault operation and maintenance management method of the above embodiments, and the specific process is referred to the corresponding steps above), including: The data acquisition module 100 is used to acquire electrical data from the smart grid terminal after establishing an electrical feature disturbance fingerprint database corresponding to manual operations of devices without communication functions. Extraction module 200 is used to detect non-faulty electrical disturbances in electrical data in real time and extract the real-time electrical characteristics of the non-faulty electrical disturbances. The comparison module 300 is used to compare real-time electrical features with the electrical feature disturbance fingerprint database, and infer the manual operation behavior of devices without communication functions based on the comparison results. The update module 400 is used to update the logical state of the corresponding equipment in the power grid digital model based on the inferred human operation behavior. The control module 500 is used to verify the updated power grid digital model and, based on the verified power grid digital model, perform fault diagnosis and generate isolation strategies.

[0140] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0141] The above description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for fault operation and maintenance management of power distribution automation, characterized in that, Includes the following steps: S1. After establishing an electrical characteristic disturbance fingerprint database corresponding to manual operation of equipment without communication function, collect electrical data from the smart grid terminal; S2. Real-time detection of non-faulty electrical disturbances in the electrical data, and extraction of the real-time electrical characteristics of the non-faulty electrical disturbances; S3. Compare the real-time electrical features with the electrical feature perturbation fingerprint database, and infer the manual operation behavior of devices without communication functions based on the comparison results; S4. Update the logical state of the corresponding equipment in the power grid digital model based on the inferred human operation behavior; S5. Verify the updated power grid digital model, and based on the verified power grid digital model, perform fault diagnosis and generate isolation strategies.

2. The power distribution automation fault operation and maintenance management method according to claim 1, characterized in that, The steps for establishing an electrical feature perturbation fingerprint database corresponding to manual operations on devices without communication capabilities include: A1. Under controlled conditions, perform preset operations on devices that do not have communication functions, and collect electrical data near the device's operation point and on associated lines while performing the preset operations; A2. Process the collected electrical data and extract the electrical characteristics generated by the preset operation; A3. The extracted electrical features are associated with the corresponding device type, operation behavior, and operation location, and stored to form the electrical feature perturbation fingerprint database.

3. The power distribution automation fault operation and maintenance management method according to claim 2, characterized in that, The specific steps in step A2 include: A21. Perform time-frequency analysis on the collected electrical data to obtain the time-frequency analysis results; A22. Based on the time-frequency analysis results, identify and separate the frequency components and transient energy distribution related to the preset operation; A23. Based on the frequency components and the transient energy distribution, extract the amplitude, phase, and duration characteristics generated by the preset operation, and use them as the electrical characteristics generated by the preset operation.

4. The power distribution automation fault operation and maintenance management method according to claim 1, characterized in that, The specific steps in step S3 include: S31. Based on the preset confidence threshold in the electrical feature perturbation fingerprint database, the real-time electrical feature is compared with the electrical feature perturbation fingerprint database to obtain a preliminary comparison result; S32. Obtain non-electrical accompanying information related to the area where the non-faulty electrical disturbance occurs; S33. Based on the aforementioned non-electrical accompanying information, generate an operational probability prediction; S34. Based on the predicted operational probability, the preliminary comparison result is corrected to obtain the corrected comparison result; S35. Based on the corrected comparison results, infer the manual operation behavior of devices that do not have communication functions.

5. The power distribution automation fault operation and maintenance management method according to claim 4, characterized in that, The non-electrical accompanying information includes operation and maintenance work plan information, equipment geographical location information, and regional load status information.

6. The power distribution automation fault operation and maintenance management method according to claim 5, characterized in that, The specific steps in step S33 include: S331. Assign weights to the operation and maintenance work plan information, equipment geographical location information, and regional load status information in the non-electrical accompanying information; S332. Based on the assigned weights, perform a weighted calculation on the degree of matching between the operation and maintenance work plan information and the disturbance occurrence time, the degree of matching between the equipment geographical location information and the disturbance occurrence location, and the degree of matching between the regional load status information and the preset typical load characteristics, and generate the operation probability prediction based on the weighted calculation results.

7. The power distribution automation fault operation and maintenance management method according to claim 5, characterized in that, The specific steps in step S34 include: S341. Obtain the confidence level of each operation behavior in the preliminary comparison results; S342. Obtain the prediction confidence level of each operation behavior in the operation probability prediction; S343. Determine the fusion weight based on the confidence level of each operation behavior in the preliminary comparison results and the prediction confidence level of each operation behavior in the operation probability prediction. S344. Based on the fusion weight, the confidence of each operation in the preliminary comparison result and the prediction confidence of each operation in the operation probability prediction are weighted and fused to obtain the corrected comparison result.

8. The power distribution automation fault operation and maintenance management method according to claim 1, characterized in that, The specific steps in step S4 include: S41. Based on preset power grid topology rules and equipment linkage logic, and according to the inferred manual operation behavior, deduce the impact of the operation behavior on the logical state of the corresponding equipment in the power grid digital model, as well as the logical state and connection relationship of the affected equipment. S42. Based on the derived impact, update the logical state of the corresponding device in the power grid digital model, as well as the logical state and connection relationship of the affected device.

9. The power distribution automation fault operation and maintenance management method according to claim 1, characterized in that, Step S5, the steps for verifying the updated power grid digital model, include: S51. Obtain multi-source verification information related to the updated state of the power grid digital model; the multi-source verification information includes real-time electrical data, power grid topology information, operating procedures, and planned maintenance task information; S52. Perform comprehensive analysis on the multi-source verification information to obtain the analysis results; S53. Based on the analysis results, evaluate the consistency between the updated digital power grid model state and physical reality.

10. A power distribution automation fault operation and maintenance management system, characterized in that, include: The data acquisition module is used to acquire electrical data from smart grid terminals after establishing an electrical feature disturbance fingerprint database corresponding to manual operations of devices without communication capabilities. An extraction module is used to detect non-faulty electrical disturbances in the electrical data in real time and extract the real-time electrical characteristics of the non-faulty electrical disturbances; The comparison module is used to compare the real-time electrical features with the electrical feature disturbance fingerprint database, and infer the manual operation behavior of devices without communication functions based on the comparison results. The update module is used to update the logical state of the corresponding equipment in the power grid digital model based on the inferred human operation behavior. The control module is used to verify the updated power grid digital model and, based on the verified power grid digital model, perform fault diagnosis and generate isolation strategies.

Images