Digital twin fault diagnosis method and device for ac-dc power system

Abstract

This invention provides a digital twin fault diagnosis method and apparatus for AC / DC power systems. The method includes: responding to a received fault notification, triggering a fault inversion process and acquiring operating condition data of the power system; mapping the operating condition data to initial cross-sectional parameters of a digital twin simulation model; extracting fuzzy fault descriptions from the fault notification, performing semantic parsing, and generating simulation parameter configurations corresponding to at least one set of fault hypotheses; executing multiple time-domain simulation tasks corresponding to each simulation parameter configuration in parallel based on the initial cross-sectional parameters, generating simulation waveform data corresponding to each fault hypothesis; calculating the multi-dimensional feature similarity between the simulation waveform data corresponding to each fault hypothesis and the fault recording data, and sorting at least one set of fault hypotheses based on the multi-dimensional feature similarity, and outputting the fault diagnosis result. This invention significantly improves the intelligence level and processing efficiency of fault analysis.

Description

Technical Field

[0001] This invention relates to the field of power system automation and digital twin technology, and in particular to a digital twin fault diagnosis method and apparatus for AC / DC power systems. Background Technology

[0002] With the rapid development of ultra-high voltage AC / DC hybrid power grids, the scale and complexity of power systems are increasing, placing higher demands on the rapid and accurate location and analysis of system faults. Currently, power system fault analysis mainly relies on manual analysis of fault recording data and protection action reports, which has the following technical shortcomings: First, traditional fault analysis methods employ static expert rule matching, relying on a pre-set fault database for simple keyword matching, which cannot handle complex multiple faults and novel fault modes. When the fault description is ambiguous or incomplete, it is difficult to accurately locate the true cause of the fault, resulting in low efficiency and insufficient accuracy in fault analysis.

[0003] Secondly, existing digital twin technology in power systems is mainly limited to equipment condition monitoring and predictive maintenance, lacking deep integration with fault diagnosis systems. Existing digital twin models often employ offline modeling methods, failing to achieve real-time state synchronization at the time of a fault. This results in a state deviation between the virtual model and the physical system at the moment of the fault, affecting the accurate reproduction of the fault scenario.

[0004] Furthermore, traditional intelligent fault diagnosis methods are mainly based on machine learning and deep learning technologies, relying on a large number of historical fault samples for model training. This leads to problems such as imbalanced samples and insufficient ability to identify novel fault modes. Existing systems lack an organic combination of physical mechanisms and data-driven approaches, making it unable to effectively handle fault scenarios outside of the training samples.

[0005] Furthermore, existing simulation verification systems lack automatic operating condition synchronization mechanisms. The initial simulation cross-sections often use fixed parameters or are manually set, failing to reflect the actual operating conditions at the time of the fault. This results in significant discrepancies between simulation results and actual fault scenarios, affecting the reliability of fault inversion. Simultaneously, current fault diagnosis systems primarily analyze based on single fault assumptions, lacking parallel simulation verification capabilities and the ability to simultaneously verify and quantitatively compare multiple fault probabilities. Data silos are a serious problem in fault analysis; the fault diagnosis platform, operational data platform, and simulation system lack real-time data interaction mechanisms, requiring extensive manual data transfer and format conversion, severely restricting the timeliness and automation level of fault analysis.

[0006] Therefore, there is an urgent need for an intelligent fault diagnosis method that can realize the transformation from fuzzy fault description to accurate simulation verification, and has the ability to automatically synchronize operating conditions and objectively evaluate them, in order to solve the problems existing in the above-mentioned technologies. Summary of the Invention

[0007] This invention provides a digital twin fault diagnosis method and apparatus for AC / DC power systems, which significantly improves the intelligence level and processing efficiency of fault analysis.

[0008] In a first aspect, the present invention provides a digital twin fault diagnosis method for AC / DC power systems, the method comprising the following steps: In response to the received fault notification, the fault inversion process is triggered, and the operating condition data of the AC / DC power system within a preset time window before the fault occurred are obtained. The operating condition data is mapped to the initial section parameters of the digital twin simulation model to achieve state synchronization between the digital twin simulation model and the AC / DC power system. Extract fuzzy fault descriptions from the fault notifications, perform semantic parsing on the fuzzy fault descriptions, and generate at least one set of simulation parameter configurations corresponding to fault assumptions; Based on the initial cross-section parameters, multiple time-domain simulation tasks corresponding to each of the simulation parameter configurations are executed in parallel to generate simulation waveform data corresponding to each of the fault assumptions. Calculate the multidimensional feature similarity between the simulation waveform data and the fault recording data corresponding to each fault hypothesis, sort the at least one set of fault hypotheses based on the multidimensional feature similarity, and output the fault diagnosis result.

[0009] According to the present invention, a digital twin fault diagnosis method for AC / DC power systems includes mapping the operating condition data to the initial section parameters of the digital twin simulation model, comprising: The operating condition data is subjected to state standardization processing to obtain standardized operating condition data; the state standardization processing includes: converting Chinese state descriptions into digital quantities and unified physical units, and allocating and calculating power reference values ​​according to operating mode; The standardized working condition data is subjected to parameter rationality verification, and abnormal data is identified and corrected to obtain the initial cross-sectional parameters.

[0010] According to the present invention, a digital twin fault diagnosis method for AC / DC power systems includes extracting fuzzy fault descriptions from fault notifications, performing semantic parsing on the fuzzy fault descriptions, and generating at least one set of simulation parameter configurations corresponding to fault hypotheses, comprising: The protection action information and the fuzzy fault description are extracted from the fault notification using regular expressions or pattern matching techniques. The semantic parsing engine is invoked, and based on a predefined fuzzy semantic mapping rule library, the protection action information and the fuzzy fault description are converted into simulation parameter configurations corresponding to each fault hypothesis; the simulation parameter configurations include fault time, fault type, fault resistance, and fault parameters.

[0011] According to the present invention, a digital twin fault diagnosis method for AC / DC power systems includes, based on the initial cross-sectional parameters, executing multiple time-domain simulation tasks corresponding to each of the simulation parameter configurations in parallel to generate simulation waveform data corresponding to each of the fault assumptions, including: A multi-threaded parallel computing architecture is adopted. By calling the electromagnetic transient simulation application programming interface in the cloud, the multiple time-domain simulation tasks are executed in parallel based on the initial cross-sectional parameters, and the simulation waveform data corresponding to each fault assumption are obtained.

[0012] According to the present invention, a digital twin fault diagnosis method for AC / DC power systems includes calculating the multidimensional feature similarity between the simulated waveform data and the fault recording data corresponding to each fault hypothesis, comprising: For any analog signal corresponding to the fault hypothesis, calculate the Pearson correlation coefficient between the simulated waveform data and the fault recording data corresponding to the fault hypothesis; For any of the aforementioned fault assumptions, calculate the state consistency matching rate between the simulated waveform data and the fault recording data corresponding to the fault assumption. By integrating the Pearson correlation coefficients and state consistency matching rates of each channel through a weighted average method, the comprehensive similarity of each fault hypothesis is obtained, and the comprehensive similarity of each fault hypothesis is determined as the multidimensional feature similarity.

[0013] According to the present invention, a digital twin fault diagnosis method for AC / DC power systems includes both analog and digital signals in the simulated waveform data and the fault recording data; the analog signals include current waveforms and voltage waveforms, and the digital signals include protection trip signals and switch position status signals.

[0014] According to the present invention, a digital twin fault diagnosis method for AC / DC power systems includes sorting the at least one set of fault hypotheses based on the similarity of each of the multidimensional features and outputting fault diagnosis results, comprising: The fault hypotheses are sorted in descending order according to the multidimensional feature similarity to obtain the sorting result; The sorting result is output as the fault diagnosis result.

[0015] According to the present invention, a digital twin fault diagnosis method for AC / DC power systems includes, in the step of acquiring operating condition data of the AC / DC power system within a preset time window before the fault occurs, the method comprises: The power setting value and voltage reference value within the preset time window before the time of the fault occurrence are obtained through the analog quantity query interface; The switch status and protection activation / deactivation signals within the preset time window before the fault occurred are obtained through the digital status interface. The power setting value, the voltage reference value, the switch status, and the protection activation / deactivation signal are determined as the operating condition data.

[0016] According to the present invention, a digital twin fault diagnosis method for AC / DC power systems, before triggering the fault inversion process in response to a received fault notification and acquiring the operating condition data of the AC / DC power system within a preset time window before the fault occurred, further includes: The fault notification is received through the data interface layer, and the fault inversion process is initiated. The fault notification is validated by parameters; the parameter validation includes at least one of the following: time format validation, time reasonableness validation, and data integrity validation; If the fault notification passes the parameter verification, three parallel data acquisition channels are activated to acquire the fuzzy fault description, the operating condition data, and the waveform file in parallel.

[0017] Secondly, the present invention also provides a digital twin fault diagnosis device for AC / DC power systems, the device comprising the following modules: The fault notification and data acquisition module is used to respond to the received fault notification, trigger the fault inversion process, and acquire the operating condition data of the AC / DC power system within a preset time window before the fault occurs. The operating condition synchronization and parameter mapping module is used to map the operating condition data to the initial section parameters of the digital twin simulation model, so as to realize the state synchronization between the digital twin simulation model and the AC / DC power system. The pattern mapping and hypothesis generation module is used to extract fuzzy fault descriptions from the fault notifications, perform semantic parsing on the fuzzy fault descriptions, and generate at least one set of simulation parameter configurations corresponding to fault hypotheses. The parallel simulation execution module is used to execute multiple time-domain simulation tasks corresponding to each of the simulation parameter configurations in parallel based on the initial cross-section parameters, and generate simulation waveform data corresponding to each of the fault assumptions. The similarity analysis and ranking module is used to calculate the multidimensional feature similarity between the simulation waveform data and the fault recording data corresponding to each fault hypothesis, and to rank the at least one set of fault hypotheses based on the multidimensional feature similarity, and output the fault diagnosis result.

[0018] According to the present invention, a digital twin fault diagnosis device for AC / DC power systems is provided, wherein the operating condition synchronization and parameter mapping module is specifically used for: The operating condition data is subjected to state standardization processing to obtain standardized operating condition data; the state standardization processing includes: converting Chinese state descriptions into digital quantities and unified physical units, and allocating and calculating power reference values ​​according to operating mode; The standardized working condition data is subjected to parameter rationality verification, and abnormal data is identified and corrected to obtain the initial cross-sectional parameters.

[0019] According to the present invention, a digital twin fault diagnosis device for AC / DC power systems is provided, wherein the pattern mapping and hypothesis generation module is specifically used for: The protection action information and the fuzzy fault description are extracted from the fault notification using regular expressions or pattern matching techniques. The semantic parsing engine is invoked, and based on a predefined fuzzy semantic mapping rule library, the protection action information and the fuzzy fault description are converted into simulation parameter configurations corresponding to each fault hypothesis; the simulation parameter configurations include fault time, fault type, fault resistance, and fault parameters.

[0020] Thirdly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the digital twin fault diagnosis method for AC / DC power systems as described above.

[0021] Fourthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the digital twin fault diagnosis method for AC / DC power systems as described above.

[0022] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the digital twin fault diagnosis method for AC / DC power systems as described above.

[0023] This invention provides a digital twin fault diagnosis method and apparatus for AC / DC power systems. First, in response to a received fault notification, a fault inversion process is triggered, and operating condition data of the AC / DC power system within a preset time window before the fault occurs are acquired. Then, the operating condition data is mapped to initial cross-sectional parameters of a digital twin simulation model to achieve state synchronization between the digital twin simulation model and the AC / DC power system. Further, a fuzzy fault description is extracted from the fault notification, and semantic parsing is performed on the fuzzy fault description to generate at least one set of simulation parameter configurations corresponding to fault hypotheses. Based on the initial cross-sectional parameters, multiple time-domain simulation tasks corresponding to each of the simulation parameter configurations are executed in parallel to generate simulation waveform data corresponding to each of the fault hypotheses. Then, the multidimensional feature similarity between the simulation waveform data and fault recording data corresponding to each fault hypothese is calculated, and the at least one set of fault hypotheses is sorted based on the multidimensional feature similarity to output the fault diagnosis result.

[0024] This invention constructs a complete digital twin fault inversion closed loop, realizing a fully automated process from event triggering, operating condition synchronization, fuzzy semantic conversion, parallel simulation to objective evaluation. It solves the problems of reliance on manual labor, inconsistent operating conditions, and inability to handle fuzzy descriptions in existing technologies, significantly improving the intelligence level and processing efficiency of fault analysis. It achieves intelligent fault diagnosis, from fuzzy fault description to accurate simulation verification, with automatic operating condition synchronization and objective evaluation capabilities. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0026] Figure 1 This is one of the flowcharts of the digital twin fault diagnosis method for AC / DC power systems provided by the present invention.

[0027] Figure 2 This is a schematic diagram illustrating the principle of the digital twin fault diagnosis method for AC / DC power systems provided by the present invention.

[0028] Figure 3 This is the second flowchart of the digital twin fault diagnosis method for AC / DC power systems provided by this invention.

[0029] Figure 4 This is a schematic diagram of the method for intelligent working condition synchronization and standardized processing provided by the present invention.

[0030] Figure 5 This is a flowchart illustrating the fault semantic mapping and hypothesis generation method provided by the present invention.

[0031] Figure 6 This is a schematic diagram showing the comparison and verification results between the simulated waveform of the fault assumption provided by this invention and the actual fault recording.

[0032] Figure 7 This is a schematic diagram of the state changes of digital signals provided by the present invention in fault simulation and waveform comparison.

[0033] Figure 8 This is a schematic diagram of the method flow for automated processing of the interface service layer provided by the present invention.

[0034] Figure 9 This is the third flowchart of the digital twin fault diagnosis method for AC / DC power systems provided by this invention.

[0035] Figure 10 This is a schematic diagram of the structure of the digital twin fault diagnosis device for AC / DC power systems provided by the present invention.

[0036] Figure 11 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0038] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same class, not limited in number; for example, a first node can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0039] This invention aims to solve key technical problems in power system fault analysis, such as the difficulty in converting fuzzy semantics into accurate simulation, simulation distortion caused by inconsistent operating conditions, and the lack of objective fault matching assessment.

[0040] (1) To resolve the contradiction between the ambiguity of fault description and the accuracy of simulation, an automatic mapping mechanism from qualitative fault description to quantitative simulation parameters is established to realize the parallel verification and screening of multiple sets of fault hypotheses, thereby improving the accuracy and reliability of fault inversion.

[0041] (2) To solve the problem of inconsistency between simulation conditions and actual operating conditions, the simulation model's operating state is completely consistent with the physical system state at the time of the fault by automatically acquiring real-time operating condition data and mapping the initial section, thereby improving the realism of fault reproduction.

[0042] (3) To solve the problem of lack of objective quantitative standards for fault type determination, a multi-dimensional similarity evaluation system based on feature correlation analysis is established to automatically compare and analyze simulated waveforms and actual recorded waveforms, so as to realize the objective identification and ranking of fault types.

[0043] The following is combined Figures 1 to 11 The present invention describes a digital twin fault diagnosis method and apparatus for AC / DC power systems.

[0044] Figure 1 This is one of the flowcharts illustrating the digital twin fault diagnosis method for AC / DC power systems provided by this invention, such as... Figure 1 As shown, the method includes the following: Step 101: In response to the received fault notification, trigger the fault inversion process and obtain the operating condition data of the AC / DC power system within a preset time window before the fault occurred.

[0045] The subject of the digital twin fault diagnosis method for AC / DC power systems provided by this invention can be an electronic device, or any other subject that can implement the digital twin fault diagnosis method for AC / DC power systems, such as other digital twin fault diagnosis systems for AC / DC power systems.

[0046] The core of this invention lies in constructing a fully automated closed-loop fault diagnosis system, encompassing event triggering, operational condition synchronization, semantic parsing, parallel simulation, and quantitative evaluation. The system employs an event-driven architecture and monitors fault notifications in real time.

[0047] The fault diagnosis service constructed in this invention adopts a distributed microservice architecture, including an interface service layer, a computation execution layer, and a data storage layer. The system supports 24 / 7 uninterrupted operation, and when a fault occurs in the power system, the fault diagnosis process is automatically triggered through an event-driven mechanism.

[0048] Figure 2 This is a schematic diagram illustrating the principle of the digital twin fault diagnosis method for AC / DC power systems provided by this invention, as shown below. Figure 2As shown, this schematic diagram illustrates the complete data flow and processing logic from the physical system to the fault diagnosis result. Specifically, the left side represents the physical system, whose output real waveforms (i.e., fault waveform data) are used for subsequent verification and comparison, and also synchronized to the simulation side via a working condition synchronization mechanism, transmitting the operating state parameters before the fault occurrence. The right side represents the simulation scenario, which is a co-simulation of primary and secondary components. It includes a primary model (simulating the main circuit of the power system, such as generators, transformers, transmission lines, converters, etc.) and a secondary model (simulating the control and protection system, such as protection devices, controllers, measurement units, etc.). By automatically setting the fault scenario and waveform channels, the fault diagnosis result is output. Data interaction between the physical system and the simulation scenario is achieved through parameter synchronization, ensuring that the initial profile of the simulation model is consistent with the state of the actual physical system at the moment of the fault. Based on this, the simulation scenario generates simulation waveforms reflecting the fault process through co-simulation of primary and secondary components. Finally, the real waveforms from the physical system and the simulation waveforms from the simulation scenario enter the verification and comparison stage. Through multi-dimensional feature similarity calculation and comparison, the system outputs the final fault diagnosis result. The schematic diagram clearly illustrates the core technical route of this invention: synchronous operation condition simulation, primary and secondary joint simulation, waveform comparison and verification, and result output.

[0049] The system achieves concurrent processing capabilities through multi-threading technology, supporting the simultaneous execution of multiple parallel simulation tasks. The system deployment architecture adopts a monolithic application model, with main components including a fault notification interface in the interface service layer (for receiving fault notifications), a historical data query interface in the data storage layer (for querying actual waveform recording data), and a computational execution layer consisting of a working condition synchronization engine, a simulation scheduling engine, and a similarity analysis engine. Data storage supports structured storage of fault reports, waveform recording files, working condition parameters, and analysis results.

[0050] Among them, fault notification refers to a digital or text message issued by the protection device or monitoring system when a fault occurs in the power system, containing information such as the station name, fault time, and preliminary fault identification information. Operating condition data refers to the operating status of the power system at a certain moment, such as the power, voltage, current, switch status, and control parameters of each device.

[0051] In practical applications, when a fault occurs (such as "grounding fault of high-end converter 1 at station S1"), the interface service layer receives a fault notification containing the station name "S1" and the fault time "2023-XX-XX 10:00:00.123". In response to this notification, the system immediately triggers the fault inversion process. Subsequently, the interface service layer automatically retrieves the operating status data of the converter station from the operating data platform within 5 minutes (a preset time window) prior to the fault occurrence, including but not limited to key parameters such as the operating power of stations S1 and S2, bipolar operation mode, and the engagement status of each valve group.

[0052] For example, Figure 3 This is the second flowchart of the digital twin fault diagnosis method for AC / DC power systems provided by this invention, as shown below. Figure 3 As shown, when a fault occurs, fault notifications and operating condition data are obtained from the operating data platform (fault diagnosis platform and next-generation platform) through the data interface, thereby triggering the fault inversion process.

[0053] Step 102: Map the operating condition data to the initial cross-sectional parameters of the digital twin simulation model to achieve state synchronization between the digital twin simulation model and the AC / DC power system.

[0054] Digital twins fully utilize data from physical models, sensor updates, and operational history to integrate multi-disciplinary, multi-physical, multi-scale, and multi-probabilistic simulation processes, completing mapping in virtual space to reflect the entire lifecycle of the corresponding physical equipment. A digital twin simulation model refers to a digitally mapped model constructed in virtual space that reflects the real-time state, behavior, and operational patterns of a physical power system. Initial cross-sectional parameters refer to a complete set of parameters used to start the digital twin simulation model, describing the system's entire operational state at the initial moment (i.e., the instant before a fault occurs).

[0055] In practical applications, the acquired operating condition data is processed and mapped to the initial cross-sectional parameters of the digital twin simulation model. These initial cross-sectional parameters accurately reproduce detailed information such as generator output, load level, transformer tap position, converter firing angle, and switch status in the AC / DC power system 0.1 seconds before the fault occurs. This ensures that the state of the digital twin model remains highly consistent with the actual physical system at the moment of the fault.

[0056] Step 103: Extract fuzzy fault descriptions from fault notifications, perform semantic parsing on the fuzzy fault descriptions, and generate at least one set of simulation parameter configurations corresponding to fault assumptions; Among them, fuzzy fault description refers to unstructured, qualitative text descriptions included in fault notifications or fault reports, such as "grounding fault of high-end converter 1 at station S1". Simulation parameter configuration refers to a set of quantitative parameters used to define the fault scenario in the time-domain simulation task, such as fault location, fault type, fault impedance, and start time.

[0057] A fault hypothesis refers to one or more possible, specific, and quantifiable fault scenarios anticipated through semantic parsing and rule mapping, based on a fuzzy fault description extracted from a fault notification. When a power system fault occurs, the original fault description is often qualitative and fuzzy natural language text, such as "S1 station pole 1 high-end converter grounding fault." Such a description is insufficient to directly drive accurate electromagnetic transient simulation. Therefore, this invention uses natural language processing and a predefined mapping rule base to parse and transform this fuzzy description into a set of quantitative, simulable fault scenarios. Each such scenario is a "fault hypothesis."

[0058] Each fault hypothesis must include at least the following precise simulation parameter configurations: Fault location identifier: such as "S1 station pole 1 valve group 1 converter high end"; Fault type coding: such as "grounding fault"; Fault parameters: such as fault resistance (e.g., 5Ω, 10Ω), fault start time, etc.

[0059] In practical applications, the system extracts a fuzzy fault description from the fault notification, such as "S1 station pole 1 high-end converter grounding fault". Through an internal semantic parsing engine, this fuzzy fault description is automatically parsed and at least one set of simulation parameter configurations corresponding to fault hypotheses is generated. For example, two sets of fault hypotheses are generated: "Fault Hypothesis A: S1 station pole 1 valve group 1 converter high-end grounding fault" and "Fault Hypothesis B: S1 station pole 1 valve group 1 converter low-end grounding fault".

[0060] Step 104: Based on the initial cross-sectional parameters, execute multiple time-domain simulation tasks corresponding to each simulation parameter configuration in parallel to generate simulation waveform data corresponding to each fault assumption.

[0061] Among them, the simulated waveform data refers to the data sequence obtained through time-domain simulation calculations, reflecting the changes of electrical quantities (voltage, current) and status quantities (switches, protection signals) over time after a fault occurs. In practical applications, based on the generated initial cross-sectional parameters, the system creates independent time-domain simulation tasks for each set of fault assumptions and submits them to the computing cluster for parallel execution. Multiple simulation tasks run synchronously, calculating simulation waveform data under their respective fault assumptions, including high-end converter valve-side current and DC bus voltage.

[0062] Step 105: Calculate the multidimensional feature similarity between the simulation waveform data and the fault recording data corresponding to each fault hypothesis, and sort at least one set of fault hypotheses based on the multidimensional feature similarity, and output the fault diagnosis results.

[0063] Fault waveform data refers to the sequence of actual electrical and state quantities recorded by fault waveform recording devices in a real power system over a period of time before and after a fault occurs, typically following the COMTRADE standard. The COMTRADE standard (IEEE C37.111) defines an internationally accepted format for exchanging fault waveform data, specifying the data structure, file format, and naming rules for waveform files. COMTRADE files usually contain a configuration file (.cfg), a data file (.dat), and a header file (.hdr).

[0064] Multidimensional feature similarity refers to a quantitative indicator that comprehensively evaluates the similarity between simulation data and actual waveform recording data from multiple dimensions (such as analog waveform morphology and digital state timing).

[0065] In practical applications, actual fault waveform data is obtained from AC / DC power systems. Subsequently, the multidimensional similarity between the simulated waveforms and actual waveforms for each set of fault hypotheses is calculated, and the results show the overall similarity of each set of fault hypotheses. Based on this, the system ranks the fault hypotheses and outputs fault diagnosis results, such as "The most likely fault is a high-side grounding fault in converter 1 of valve group 1 at station S1, with a confidence level of 94.2%." The ranking results provide fault handlers with a clear assessment of fault probability, significantly improving the efficiency and accuracy of fault analysis.

[0066] The method provided in this embodiment of the invention first responds to a received fault notification by triggering a fault inversion process and acquiring the operating condition data of the AC / DC power system within a preset time window before the fault occurs. Then, the operating condition data is mapped to the initial section parameters of a digital twin simulation model to achieve state synchronization between the digital twin simulation model and the AC / DC power system. Further, a fuzzy fault description is extracted from the fault notification, and semantic parsing is performed on the fuzzy fault description to generate at least one set of simulation parameter configurations corresponding to fault hypotheses. Based on the initial section parameters, multiple time-domain simulation tasks corresponding to each simulation parameter configuration are executed in parallel to generate simulation waveform data corresponding to each fault hypothesis. Then, the multidimensional feature similarity between the simulation waveform data and fault recording data corresponding to each fault hypothesis is calculated, and at least one set of fault hypotheses is sorted based on the multidimensional feature similarity to output the fault diagnosis result.

[0067] This invention constructs a complete digital twin fault inversion closed loop, realizing a fully automated process from event triggering, operating condition synchronization, fuzzy semantic conversion, parallel simulation to objective evaluation. It solves the problems of reliance on manual labor, inconsistent operating conditions, and inability to handle fuzzy descriptions in existing technologies, significantly improving the intelligence level and processing efficiency of fault analysis. It achieves intelligent fault diagnosis, from fuzzy fault description to accurate simulation verification, with automatic operating condition synchronization and objective evaluation capabilities.

[0068] It should be noted that each embodiment of the present application can be freely combined, reordered, or executed independently, without relying on or depending on a fixed execution sequence.

[0069] According to a digital twin fault diagnosis method for AC-DC power systems provided by the present invention, operating condition data is mapped to initial section parameters of a digital twin simulation model, comprising: performing state standardization processing on the operating condition data to obtain standardized condition data; the state standardization processing includes: converting Chinese state descriptions into digital quantities and unifying physical units, as well as distributing and calculating power reference values according to operation modes; performing parameter rationality verification on the standardized condition data, identifying and correcting abnormal data therein, to obtain the initial section parameters.

[0070] In some embodiments, the specific implementation of mapping the operating condition data to the initial section parameters of the digital twin simulation model includes the following steps: The state standardization processing refers to the process of converting condition data from different data sources and in various formats (such as text, numbers, different units) into data with a unified format and dimension that can be directly recognized by the simulation model. For example, it includes: converting Chinese state descriptions into digital states and unifying physical units, as well as distributing and calculating power reference values according to operation modes. The parameter rationality verification refers to the process of checking the standardized condition data, and identifying and correcting abnormal data points that are out of the normal range, sudden, missing, etc.

[0071] In practical applications, in step 102, the condition synchronization engine first performs state standardization processing. For example, the original value of "switch state" obtained from the data platform is the Chinese character "closed", and the system converts it into the digital quantity "1"; the original value of "power reference value" is "500MW (megawatt)", and the system uniformly converts it into a per-unit value or retains the basic unit, and performs distribution calculation according to the current bipolar operation mode (for example, 250MW per pole).

[0072] Next, the system performs parameter rationality verification on the standardized condition data. For example, if a standardized voltage value on the valve side of a converter transformer is displayed as "0.2 p.u.", while the normal operation range should be between "0.95-1.05 p.u.", the system determines that this is abnormal data. According to preset rules, the system may mark the value as invalid, and use the historical data of the previous moment or the average value of adjacent nodes for correction, or directly issue an alarm to ensure that the initial section parameters entering the simulation model are credible.

[0073] For example, for a schematic diagram of the intelligent condition synchronization process, refer to Figure 4 as shown in Figure 4This is a schematic diagram of the intelligent operating condition synchronization and standardized processing method provided by the present invention. It should be noted that this diagram actually shows the data capture and processing process related to operating condition synchronization. The method includes: This flowchart uses a timeline as its main thread to illustrate the complete processing from the moment a fault occurs to the standardized output of operating condition data.

[0074] The top of the screen uses time as a scale, indicating the specific time of the fault occurrence as 2025-05-30 14:18:53. Below the timeline, a time window for capturing operating condition data is defined, for example: steady-state operating conditions n minutes before the fault (e.g., the first 5 minutes) (14:13:53-14:18:53). This window is used to ensure that the acquired operating condition data represents the steady-state operating state before the fault occurred, avoiding interference from transient processes.

[0075] The captured operating condition data is divided into two main categories: analog parameters and digital status (including discrete signals such as switch states). This raw data enters the operating condition data standardization processing stage, the core of which is the standardization and verification of the data before the mapping result. This stage specifically includes three operations: 1. Chinese status value conversion: For example, converting the switch status "close / open" into the digital value "1 / 0"; 2. Power allocation calculation: The total power reference value is allocated and calculated according to the current operating mode; 3. Constraint check: Perform parameter rationality checks on the standardized data and automatically identify abnormal data.

[0076] After the above standardization and verification processes, the final output is standard chemical condition data that can be used for initial cross-section mapping of the simulation model.

[0077] The flowchart fully demonstrates the entire process of this invention, from determining the time window and parallel acquisition of analog and digital data to performing multidimensional standardization and verification, ensuring a high degree of consistency between the initial state of the digital twin simulation model and the actual physical system. The technical advantages are: (1) Precise mapping of digital twin: The real-time synchronization technology of physical-virtual state of power system is established, and the accuracy of the simulation initial conditions and the actual system is higher than the industry standard. (2) Intelligent selection of time window: Automatically calculates the steady state period before the fault, avoids transient interference, and ensures the representativeness and reliability of the operating condition data. (3) Multidimensional parameter coordination: Simultaneously processes multiple key operating parameters, including power, voltage, and switch status, to achieve comprehensive operating condition reproduction. (4) Adaptive anomaly handling: Built-in parameter rationality verification algorithm automatically identifies and corrects abnormal data, improving the credibility of simulation results.

[0078] The method provided in this invention solves the problem that multi-source heterogeneous data is difficult to be directly identified by the simulation model, and improves the robustness and reliability of the simulation input through automated abnormal data identification and correction, thus ensuring the accuracy of the initial state of the digital twin model.

[0079] According to the present invention, a digital twin fault diagnosis method for AC / DC power systems extracts fuzzy fault descriptions from fault notifications, performs semantic parsing on the fuzzy fault descriptions, and generates at least one set of simulation parameter configurations corresponding to fault hypotheses, including: Regular expressions or pattern matching techniques are used to extract protection action information and fuzzy fault descriptions from fault notifications. The semantic parsing engine is invoked, and based on a predefined fuzzy semantic mapping rule library, the protection action information and fuzzy fault description are converted into simulation parameter configurations corresponding to each fault hypothesis. The simulation parameter configurations include fault time, fault type, fault resistance, and fault parameters.

[0080] Regular expression / pattern matching is a technique used to find, extract, and replace strings from text based on specific patterns. The semantic parsing engine is a module capable of understanding the meaning of natural language or semi-structured text and transforming it into machine-understandable, structured data. The fuzzy semantic mapping rule base is a pre-built knowledge base that stores the correspondence between fuzzy fault phenomenon descriptions (such as "converter grounding," "protection action A tripped") and precise simulation parameters (such as fault location ID, fault type code, and fault resistance range).

[0081] In practical applications, in step 103, the system uses regular expression technology to extract key information from the complete fault report. For example, using regular expression for converter (high-end|low-end) grounding, the high-end is successfully matched and extracted from the text "After analysis, the fault cause is high-end grounding of converter 1 at station S1".

[0082] Subsequently, the system invokes the semantic parsing engine to input the extracted information into a predefined fuzzy semantic mapping rule base. This rule base contains hundreds of precise fault scenarios. Based on keywords such as "S1 station," "pole 1," "converter," "high-end," and "grounding," the engine matches the corresponding rule entries and automatically generates a set of precise simulation parameter configurations, including fault time, fault type, and fault resistance: {fault type code: GND_FLT, fault resistance: [5,10,20]Ω, fault start time: fault notification time - 0.1s}.

[0083] For example, Figure 5 This is a flowchart illustrating the fault semantic mapping and hypothesis generation method provided by the present invention, as shown below. Figure 5As shown, this process reveals an automatic conversion mechanism from fuzzy fault descriptions to multiple sets of precise fault hypotheses. The method includes: Fuzzy fault description: Its specific input is "S1 station pole 1 high-end converter grounding fault", which is derived from the automatic parsing of fault reports using Extensible Markup Language (XML).

[0084] The fuzzy fault description was then fed into the semantic parsing engine. This engine performed keyword extraction and pattern matching to parse the unstructured natural language text.

[0085] After the analysis is completed, the process enters the fault hypothesis generation stage, where the system automatically outputs nine sets of fault hypotheses. Each set of fault hypotheses includes specific simulation parameter configurations.

[0086] The diagram details the nine sets of fault hypotheses generated, categorized into three main types: The converter system faults include: high-end grounding fault of converter 1 in valve group 1 of station S1, low-end grounding fault of converter 1 in valve group 1 of station S1, and mid-point grounding fault of converter 1 in valve group 1 of station S1. The faults on the YY transformer side include: single-phase grounding of phase A of YY transformer 1Y ... The faults on the YD transformer side include: single-phase grounding of phase A of YD transformer 1 in valve group 1 of station S1, single-phase grounding of phase B of YD transformer 1 in valve group 1 of station S1, and single-phase grounding of phase C of YD transformer 1 in valve group 1 of station S1.

[0087] The diagram illustrates that the simulation parameter configuration for each automatically generated fault hypothesis includes at least the fault time, fault resistance, and fault type.

[0088] The flowchart fully demonstrates how this invention utilizes a semantic parsing engine to expand a fuzzy fault description into multiple quantifiable simulation hypotheses covering different fault locations and types, providing complete parameterized input for subsequent parallel simulation verification. The technical advantages are: (1) Intelligent semantic understanding: Breaking through the limitations of traditional keyword matching, deep semantic parsing of fault descriptions is achieved through regular expressions and pattern recognition technology. (2) Scalable knowledge base: Hundreds of precise fault scenarios cover more than 95% of common fault types, significantly exceeding the coverage of traditional systems. (3) High mapping accuracy: The conversion accuracy from fuzzy description to precise parameters reaches over 90%, eliminating the subjective differences of manual interpretation. (4) Dynamic expansion capability: Adopting a configurable mapping rule base, it supports the rapid addition of new fault types, adapting to the needs of power system technology development.

[0089] The method provided in this invention realizes the automatic and intelligent conversion from vague and qualitative fault descriptions such as "grounding of high-end converter of station pole 1" to precise simulation parameters such as "fault location ID=101", thus solving the semantic gap problem in fault inversion.

[0090] According to the present invention, a digital twin fault diagnosis method for AC / DC power systems is provided, which, based on initial cross-sectional parameters, executes multiple time-domain simulation tasks corresponding to each simulation parameter configuration in parallel to generate simulation waveform data corresponding to each fault hypothesis, including: A multi-threaded parallel computing architecture is adopted. By calling the cloud-based electromagnetic transient simulation application programming interface, multiple time-domain simulation tasks are executed in parallel based on the initial cross-sectional parameters, and the simulation waveform data corresponding to each fault assumption are generated.

[0091] Specifically, in some embodiments, step 104 is implemented through the following steps: The multi-threaded parallel computing architecture allows multiple threads within a single process to execute different computational tasks simultaneously, thereby leveraging the capabilities of multi-core processors to improve computational efficiency. The cloud-based electromagnetic transient simulation application programming interface (API) is deployed on a cloud computing platform and is specifically designed for performing electromagnetic transient simulation calculations. Users can submit simulation tasks and obtain simulation results by calling the API, without needing to concern themselves with the configuration of the underlying hardware and simulation software.

[0092] In practical applications, in step 104, a multi-threaded thread pool is started through the simulation scheduling engine. For each set of generated fault hypotheses, the engine allocates an independent simulation thread for each set. Each thread encapsulates complete simulation parameters (initial cross-sectional parameters, fault parameter configuration). Then, the Representational State Transfer API (RESTful API) of the cloud-deployed electromagnetic transient simulation platform is called via Hypertext Transfer Protocol (HTTP). After receiving the request, the API allocates computing resources in the cloud to execute two time-domain simulation tasks in parallel. After the simulation is completed, each thread receives the simulation waveform data in JavaScript Object Notation (JSON) format returned by the cloud and summarizes it.

[0093] The method provided in this invention utilizes cloud computing resources to achieve large-scale parallel processing capabilities, which can simultaneously verify multiple fault hypotheses, reducing the computation time of traditional serial simulation, which requires several hours, to minutes, and significantly improving the real-time performance and computational efficiency of fault analysis.

[0094] According to the present invention, a digital twin fault diagnosis method for AC / DC power systems calculates the multidimensional feature similarity between the simulated waveform data and the fault recording data corresponding to each fault hypothesis, including: For any analog signal corresponding to a fault hypothesis, calculate the Pearson correlation coefficient between the simulated waveform data and the fault recording data corresponding to the fault hypothesis; For any fault hypothesis, calculate the state consistency matching rate between the simulated waveform data and the fault recording data corresponding to the fault hypothesis. By integrating the Pearson correlation coefficients and state consistency matching rates from multiple channels using a weighted average method, the comprehensive similarity of each fault hypothesis is obtained, and the comprehensive similarity of each fault hypothesis is determined as the multidimensional feature similarity.

[0095] In some embodiments, the calculation of the multidimensional feature similarity between the simulation waveform data and the fault recording data corresponding to each fault hypothesis in step 105 is achieved through the following steps: The Pearson correlation coefficient is a statistic that measures the degree of linear correlation between two continuous variables, ranging from -1 to 1. A value closer to 1 indicates a stronger positive correlation, while a value closer to -1 indicates a stronger negative correlation. The state consistency matching rate is a proportional indicator that measures whether the states of two binary signal sequences (such as protection trip signals) are consistent at each time point, ranging from 0 to 1. The overall similarity is calculated by using a weighted average method to combine the Pearson correlation coefficients and state consistency matching rates from multiple channels (measurement points) into a single value representing the overall matching degree.

[0096] In practical applications, for any hypothetical fault corresponding to an analog signal (e.g., "phase A current on the valve side of pole 1 high-end converter"), the analog signal in its simulated waveform is calculated; simultaneously, the actual waveform data of the corresponding channel is extracted from the fault waveform data. Then, the Pearson correlation coefficient (r) between these two time series is calculated. The calculation formula is as follows: in, The Pearson correlation coefficient is the relationship between the time series corresponding to the analog signal of the simulated waveform and the time series corresponding to the analog signal of the actual waveform. For the sampling points of the simulated waveform, For actual waveform recording sampling points, These are the arithmetic means of the corresponding sequences, where n is the total number of sampling points. The Pearson correlation coefficient r ranges from [-1, 1]. When r is close to 1, it indicates a high positive correlation; when it is close to -1, it indicates a high negative correlation; and when it is close to 0, it indicates no linear correlation.

[0097] For example, Figure 6 This is a schematic diagram illustrating the comparison and verification results between the simulated waveform of the fault hypothesis provided by this invention and the actual fault recording. Specifically, it shows the waveform comparison results for a certain fault hypothesis. The fault hypothesis involves the IDL measurement point of grounding S1-2 on the P1 DC line, and the specific channel is identified as "P1_DC_LINE_S1-2_GND_D_IDL (NRAIn-PPR_0)(S1P1_C_P)". Figure 6 As shown in the figure, the horizontal axis represents time (t / s), and the vertical axis represents amplitude, reflecting the magnitude of the electrical quantity. The red solid line represents the actual fault waveform obtained from the physical system, i.e., the IDL fault waveform, and the blue dashed line represents the simulated waveform generated after digital twin simulation based on this fault assumption, the IDL (NRAIn-PPR_0) (S1P1_C_P) simulated waveform. The Pearson correlation coefficient for this comparison is 93.9%. In this example, 93.9% indicates that the simulated waveform and the actual waveform have a very high positive correlation, verifying the accuracy of the fault assumption.

[0098] This figure illustrates the practical application effect of the multidimensional feature similarity evaluation system of the present invention. By combining waveform visualization with quantitative correlation coefficients, it provides intuitive and objective evidence support for fault diagnosis results.

[0099] Secondly, for any fault assumption, the system extracts the binarized sequence of the digital signal (e.g., protection trip signal, switch position, etc.) from the simulated waveform and the actual recorded waveform (1 indicates action, 0 indicates no action). The system calculates the state consistency matching rate (M) using the following formula: Where M represents the state consistency matching rate between the digital state of the simulated waveform and the digital state of the actual recorded waveform. The digital state of the simulated waveform. The digital state of the actual recorded waveform is represented by I(·), which is an indicator function. The value is 1 when the states are equal and 0 otherwise. n is the total number of sampling points. The state consistency matching rate M ranges from [0, 1]. The closer M is to 1, the higher the signal consistency.

[0100] For example, Figure 7 This is a schematic diagram illustrating the state changes of the digital signal provided by this invention in fault simulation and waveform comparison, as shown below. Figure 7As shown, this figure illustrates the comparison of digital signals under a specific fault assumption, specifically "P1_DC_LINE_S1-2_GND_D_LPTW_TRIP (1-41-NRUnpack-77) (S2P1_C_P) Digital Comparison". The horizontal axis represents time (t / s), and the vertical axis represents the digital status value (0 / 1). "1" indicates action (e.g., protection issues a trip command), and "0" indicates no action (e.g., protection does not trip). The legend includes two curves: LPTW_TRIP Fault Recording (0 / 1): Represents the recorded waveform data of this digital signal in the actual physical system; LPTW_TRIP(1-41-NRUnpack-77) (S2P1_C_P) Simulation waveform (0 / 1): Represents the simulation waveform of this signal generated based on the digital twin simulation model.

[0101] As can be seen from the data in the graph, the state values ​​of both curves remained at 0.00 throughout the entire time window from 0.0 seconds to 1.6 seconds, without any jump to 1. This indicates that the protection trip signal did not activate under this fault scenario, and the simulated waveform is completely consistent with the actual recorded waveform.

[0102] This diagram can be used to calculate the state consistency matching rate of digital signals. According to the method of this invention, the system compares the digital signals in the simulated waveform with the corresponding signals in the actual recorded waveform point by point, and calculates the degree of matching using the state consistency matching rate formula. This example demonstrates a typical protection trip signal timing, which is an important basis for evaluating the similarity of digital signals in fault diagnosis.

[0103] Finally, the similarity results from multiple channels are integrated using a weighted average method to automatically generate a ranked list of fault types and output the most probable fault scenarios and their confidence levels. For example, the correlation coefficient r and matching rate M calculated for all analog channels (such as three-phase current and voltage) and digital channels (such as multiple protection trip signals) of any fault hypothesis are integrated using a weighted average method (e.g., analog weight 0.6, digital weight 0.4), resulting in a comprehensive similarity of 0.942 for fault hypothesis A. Similarly, the comprehensive similarity of 0.879 for fault hypothesis B is calculated.

[0104] The method provided in this invention uses a dual algorithm of Pearson correlation coefficient and state consistency to eliminate the subjectivity of manual judgment, and the evaluation results are repeatable and verifiable. It analyzes both analog and digital signals simultaneously, which effectively improves the accuracy of fault identification compared to single-dimensional analysis. The optimized algorithm implementation enables the similarity calculation of large-scale waveform data to be completed in seconds, supporting the needs of real-time fault analysis.

[0105] According to the present invention, a digital twin fault diagnosis method for AC / DC power systems includes both analog and digital signals in the simulated waveform data and the fault recording data. The analog signals include current waveforms and voltage waveforms, and the digital signals include protection trip signals and switch position status signals.

[0106] Specifically, analog signals refer to electrical quantities that change continuously over time, such as current and voltage. Digital signals refer to signals that are discrete in both time and value, and typically represent the state of equipment, such as the opening and closing of switches or the activation or deactivation of protection mechanisms.

[0107] The following is a detailed implementation example: The simulated waveform data generated by the simulation scheduling engine, and the fault recording data obtained from the actual system, both contain at least two types of signals. The first type is analog signals, specifically including but not limited to: the A, B, and C phase current waveforms near the fault point, the bus three-phase voltage waveforms, and the converter DC-side voltage waveforms. The second type is digital signals, specifically including but not limited to: the trip output signals of the line protection device, the circuit breaker position status signals, and the reclosing start and output signals. These signals together provide a comprehensive data foundation for subsequent multi-dimensional feature similarity calculations.

[0108] The method provided in this embodiment of the invention clarifies the specific connotation of the data object and ensures the feasibility of the technical solution of the invention.

[0109] According to the present invention, a digital twin fault diagnosis method for AC / DC power systems sorts at least one set of fault hypotheses based on the similarity of various multidimensional features and outputs fault diagnosis results, including: The fault hypotheses are sorted in descending order according to the multidimensional feature similarity to obtain the sorting results; The sorting results will be output as the fault diagnosis results.

[0110] Specifically, in some embodiments, after calculating the overall similarity of each set of fault hypotheses, the system automatically generates a fault type and sorts all fault hypotheses in descending order of their overall similarity. The sorting results are as follows: 1. Short circuit fault of the No. 1 thyristor valve bridge arm of the 1YD converter of the 1st pole 1 valve group of S1 station - similarity: 0.938; 2. Short circuit fault at the neutral point of the high-end converter of the S1 station pole 1 valve group 1 - grounding fault - similarity: 0.879; 3. Short circuit fault of the bridge arm of valve group 1YD converter No. 1 thyristor and valve 6 at station 1 - similarity: 0.635.

[0111] The system outputs the sorting results as the final fault diagnosis results, for example, by sending them to the maintenance personnel's terminal in JSON format, or by displaying them on the system interface.

[0112] The method provided in this invention provides maintenance personnel with a clear ranking of fault probabilities based on a comprehensive similarity intelligent ranking mechanism, which greatly improves decision-making efficiency.

[0113] According to the present invention, a digital twin fault diagnosis method for AC / DC power systems acquires operating condition data of the AC / DC power system within a preset time window before the fault occurs, including: The power setting value and voltage reference value within a preset time window before the fault occurred can be obtained through the analog query interface. The switch status and protection activation / deactivation signals within a preset time window before the fault occurs are obtained through the digital status interface. The power setting value, voltage reference value, switch status, and protection activation / deactivation signals are defined as operating condition data.

[0114] Specifically, in some embodiments, acquiring the operating condition data of the AC / DC power system within a preset time window before the fault occurs is achieved through the following steps: The analog query interface is an API used to query continuously changing quantities (such as power and voltage) from a data platform (such as a SCADA system). The digital status interface is an API used to query discrete status quantities (such as switch positions and protection activation / deactivation status) from a data platform.

[0115] In practical applications, the system does not passively receive operating condition data, but actively acquires it through specific interfaces.

[0116] First, the system sends a request to the operation data platform via the analog query interface to obtain the active power setpoint (e.g., 500MW), reactive power reference value, and AC bus voltage reference value (e.g., 525kV) of stations S1 and S2 within 5 minutes before the fault occurred. Second, the system obtains the AC side circuit breaker position (closed / open), DC side disconnect switch status, and related protection function activation / deactivation signals (e.g., "converter differential protection activated" is "1") of the high-end converter of pole 1 of station S1 within the same time period via the digital status interface.

[0117] The method provided in this invention clarifies the technical path and data type for data acquisition, providing a specific and operable technical solution for subsequent operating condition synchronization, and ensuring that the system can accurately capture the full picture of steady-state operation before the fault.

[0118] According to the digital twin fault diagnosis method for AC / DC power systems provided by the present invention, before triggering the fault inversion process in response to a received fault notification and acquiring the operating condition data of the AC / DC power system within a preset time window before the fault occurrence time, the method further includes: Receive fault notifications through the data interface layer and initiate the fault inversion process; Perform parameter verification on fault notifications; parameter verification includes at least one of the following: time format verification, time reasonableness verification, and data integrity verification; If the fault notification passes parameter verification, activate three parallel data acquisition channels to acquire fuzzy fault descriptions, operating condition data, and waveform files in parallel.

[0119] Specifically, the data interface layer is the outermost layer of the system architecture, responsible for data interaction with external systems, receiving input and returning output. Parameter verification checks the validity of received fault notification data to prevent invalid or erroneous data from triggering subsequent processing flows. Parallel data acquisition channels are independent processing flows that are opened simultaneously to acquire different types of data (fault description, operating condition, waveform recording).

[0120] In practical applications, in some embodiments, the method further includes a preprocessing step before the fault inversion process is initiated, including: First, the system's data interface layer exposes a fault notification receiving endpoint in the form of a RESTful API. When a fault notification in JSON format containing {"station":"S1 station", "fault_time":"2023-XX-XX 10:00:00.123", "report_id":"RPT001"} arrives, the interface layer initiates the processing flow.

[0121] Secondly, the interface layer performs rigorous parameter validation on received fault notifications to ensure the integrity and correctness of input data. The system standardizes and validates the time format, preventing future timestamps or obviously incorrect timestamps from causing subsequent processing anomalies. For example, the system verifies whether the time format is "YYYY-MM-DD HH:MM:SS.SSS" and checks whether the fault time is a reasonable past time (not a future time). Simultaneously, the data integrity check module verifies the validity of key parameters such as station name and file name, intercepting anomalies such as empty values ​​and special characters. If validation fails, the system immediately returns detailed error information to prevent invalid data from entering subsequent processing stages. This pre-validation mechanism significantly improves the system's robustness and reliability. For example, the data integrity check module verifies whether the "station name" and "report ID" fields are empty and whether they contain special characters.

[0122] Finally, after all verifications passed, the system immediately activated three parallel data acquisition channels. The first channel is responsible for acquiring fault description information. It connects to the fault information source through an authentication mechanism, automatically downloads the specified fault report file, and stores it locally to provide data support for subsequent fault information analysis. The second channel is dedicated to acquiring operating condition data. It obtains system operating parameters before and after the fault occurrence from the data platform through analog query interfaces and digital status interfaces, including key operating information such as power setpoints, voltage reference values, switch status, and protection activation / deactivation signals. The third channel performs batch download tasks of waveform recording files. Based on the file list parsed from the fault description information, it automatically acquires the corresponding waveform recording files and performs preliminary format verification and data integrity checks. For example, channel 1 downloads the fault report file from the file server based on the "report ID" (acquiring an ambiguous fault description); channel 2 acquires operating condition data through analog and digital interfaces based on the fault time; and channel 3 downloads the corresponding COMTRADE format waveform recording file based on the list of files that may be attached to the report.

[0123] For example, Figure 8 This is a schematic diagram of the method flow for automated processing of the interface service layer provided by the present invention, as shown below. Figure 8 As shown, this process details the complete automated processing from fault occurrence to standardized data output. The method includes: The process begins with the occurrence of a fault, which then triggers the fault notification receiving step, where the interface automatically receives the fault parameters. Next, the system proceeds to the parameter verification and processing step, where it performs time format validation and data integrity checks on the fault parameters.

[0124] If the verification fails, the process will branch to return an error message and terminate the current processing to prevent invalid data from entering subsequent stages.

[0125] If the verification is successful, the system will start three data acquisition channels in parallel: XML report acquisition, obtaining authentication token, and downloading fault report files; operating condition data acquisition, performing analog parameter query and digital status acquisition in parallel; and waveform file download: batch downloading waveform files and parsing COMTRADE format.

[0126] The data acquired from the above three channels are converged to the data standardization processing layer. This processing layer integrates a variety of standardization processing functions, including: XML parsing to extract protection actions, operating condition parameter format conversion, waveform data time synchronization, and abnormal data processing.

[0127] After standardization, the standardized data is output and finally passed to the working condition synchronization engine for further processing.

[0128] The flowchart clearly demonstrates the fully automated mechanism of the interface service layer of this invention, from pre-verification and parallel data acquisition to standardized output, reflecting efficient and robust multi-source heterogeneous data processing capabilities. The technical advantages are: (1) Intelligent parameter verification: It has a built-in comprehensive parameter verification mechanism, including time format verification and data integrity check, which effectively prevents analysis errors caused by dirty data; (2) Fault-tolerant design: It has an exception handling and automatic retry mechanism to ensure the reliability of data transmission during network fluctuations or system load peaks; (3) Real-time communication guarantee: It realizes the real-time push of simulation progress, and the user experience is significantly better than the traditional polling method.

[0129] The method provided in this invention prevents dirty data from entering the core analysis process through a pre-verification mechanism, thereby improving the robustness of the system; and significantly improves the acquisition efficiency of multi-source heterogeneous data through parallel acquisition channels, thus ensuring the timeliness of the entire fault diagnosis process.

[0130] Compared with the prior art, the advantages of the present invention include: (1) A complete digital twin fault inversion closed-loop system was constructed, realizing a fully automated process from fuzzy fault description to accurate simulation verification and then to objective similarity assessment, which significantly improved the intelligence level and processing efficiency of fault analysis.

[0131] (2) The present invention adopts event-driven real-time operating condition synchronization technology, which can automatically capture the system operating status at the time of the fault occurrence, and realize the dynamic initialization of the simulation model through parameter mapping and standardization processing, thus ensuring the high consistency between the virtual simulation environment and the physical system.

[0132] (3) The present invention establishes a multi-hypothesis verification mechanism based on parallel time-domain simulation, which can simultaneously perform simulation calculations of multiple fault scenarios and achieve large-scale parallel processing through cloud computing resources, thereby greatly improving the computational efficiency and coverage of fault analysis.

[0133] (4) This invention innovatively uses a multidimensional feature correlation analysis algorithm, combined with the calculation of the Pearson correlation coefficient of analog waveforms and the state consistency assessment of digital signals, to establish an objective fault matching measurement system, eliminating the subjective judgment factors in traditional methods.

[0134] For example, a DC transmission system is used as a typical case study to demonstrate the complete implementation process and effects of the present invention. This case is based on actual operating data, and the fault type is "grounding fault of high-end converter 1 at station S1".

[0135] Data source for the example: The DC transmission project adopts the bipolar polarity return line operation mode. At the time of the fault, the system operating parameters are as follows: the operating power of S1 station, the operating power of S2 station, the bipolar operation mode, the engagement status of each valve group, and other key parameters are managed through a parameter array.

[0136] Implementation steps and code execution: Step 1: The data interface layer receives the fault notification, and after parameter verification, triggers the functions to obtain the token and file from the fault report file to automatically download the fault report.

[0137] Step 2: The fault report parsing module analyzes the fault content, identifies protection actions, and generates fault hypotheses. It then generates various precise fault scenario configurations through mapping rules.

[0138] Step 3: The fault simulation method performs operating condition synchronization and obtains real-time operating parameters to generate standardized operating condition parameter configurations.

[0139] Step 4: Execute multiple simulation tasks in parallel. Each task calls the CloudPSS API to perform time-domain simulation and generate waveform data files in JSON format.

[0140] Step 5: Call the analog and digital waveform algorithms to perform similarity analysis, calculate the comprehensive matching degree, and finally output the matching degree ranking results.

[0141] Actual running results: The system output sorting results are: [(“S1 station pole 1 valve group 1 converter high-end grounding fault”, 0.942), (“S1 station pole 1 valve group 1 converter low-end grounding fault”, 0.879), ...], the first fault type perfectly matches the actual fault, verifying the technical effectiveness of the invention. The entire processing flow is completed within 40 minutes, significantly improving fault analysis efficiency.

[0142] For example, Figure 9 This is the third flowchart of the digital twin fault diagnosis method for AC / DC power systems provided by this invention. Figure 9 As shown, the method includes: The fault occurred, and the fault name was "S1 station pole 1 high-end converter fault occurred"; Phase 1: Fault Notification and Data Acquisition This includes the fault diagnosis platform automatically sending fault notifications and automatically obtaining fault reports and actual waveform recordings.

[0143] Phase Two: Fuzzy Mapping and Hypothesis Generation This includes parsing and identifying protection action information from fault reports, generating accurate fault hypotheses through a mapping rule base, and each fault hypothesis containing simulation parameter configurations.

[0144] Phase 3: Operating Condition Synchronization and Parameter Mapping This includes standardized processing of operating condition data, initial cross-section setting of simulation models, and constraint condition checking and parameter verification.

[0145] Phase 4: Parallel Simulation Execution This includes multi-threaded parallel scheduling of multiple simulation tasks, cloud-based time-domain simulation calculations on the simulation platform, automatic export of waveform data for key protection channels, and standardized storage of result data.

[0146] Phase 5: Similarity Analysis and Ranking This includes calculating the Pearson correlation coefficient for analog quantities, matching the consistency of digital quantities in terms of state, weighted averaging of multi-channel similarity, ranking fault types by confidence, determining the optimal matching result, and pushing the real-time results to the front-end interface.

[0147] Finally, the output is: High-side grounding fault of converter 1 at station S1 (highest matching degree).

[0148] The following describes the digital twin fault diagnosis device for AC / DC power systems provided by the present invention. The digital twin fault diagnosis device for AC / DC power systems described below can be referred to in correspondence with the digital twin fault diagnosis method for AC / DC power systems described above.

[0149] Figure 10 This is a schematic diagram of the structure of the digital twin fault diagnosis device for AC / DC power systems provided by the present invention, as shown below. Figure 10 As shown, the digital twin fault diagnosis device 1000 for AC / DC power systems includes the following modules: The fault notification and data acquisition module 1010 is used to respond to the received fault notification, trigger the fault inversion process, and acquire the operating condition data of the AC / DC power system within a preset time window before the fault occurs. The operating condition synchronization and parameter mapping module 1020 is used to map the operating condition data to the initial section parameters of the digital twin simulation model, so as to realize the state synchronization between the digital twin simulation model and the AC / DC power system. The pattern mapping and hypothesis generation module 1030 is used to extract fuzzy fault descriptions from the fault notification, perform semantic parsing on the fuzzy fault descriptions, and generate at least one set of simulation parameter configurations corresponding to fault hypotheses. The parallel simulation execution module 1040 is used to execute multiple time-domain simulation tasks corresponding to each of the simulation parameter configurations in parallel based on the initial cross-section parameters, and generate simulation waveform data corresponding to each of the fault assumptions. The similarity analysis and ranking module 1050 is used to calculate the multidimensional feature similarity between the simulation waveform data and the fault recording data corresponding to each fault hypothesis, and to rank the at least one set of fault hypotheses based on the multidimensional feature similarity, and output the fault diagnosis result.

[0150] According to the present invention, a digital twin fault diagnosis device 1000 for AC / DC power systems is provided, wherein the operating condition synchronization and parameter mapping module 1020 is specifically used for: The operating condition data is subjected to state standardization processing to obtain standardized operating condition data; the state standardization processing includes: converting Chinese state descriptions into digital quantities and unified physical units, and allocating and calculating power reference values ​​according to operating mode; The standardized working condition data is subjected to parameter rationality verification, and abnormal data is identified and corrected to obtain the initial cross-sectional parameters.

[0151] According to the present invention, a digital twin fault diagnosis device 1000 for AC / DC power systems is provided, wherein the mode mapping and hypothesis generation module 1030 is specifically used for: The protection action information and the fuzzy fault description are extracted from the fault notification using regular expressions or pattern matching techniques. The semantic parsing engine is invoked, and based on a predefined fuzzy semantic mapping rule library, the protection action information and the fuzzy fault description are converted into simulation parameter configurations corresponding to each fault hypothesis; the simulation parameter configurations include fault time, fault type, fault resistance, and fault parameters.

[0152] According to the present invention, a digital twin fault diagnosis device 1000 for AC / DC power systems is provided, wherein the parallel simulation execution module 1040 is specifically used for: A multi-threaded parallel computing architecture is adopted. By calling the electromagnetic transient simulation application programming interface in the cloud, the multiple time-domain simulation tasks are executed in parallel based on the initial cross-sectional parameters, and the simulation waveform data corresponding to each fault assumption are obtained.

[0153] According to the present invention, a digital twin fault diagnosis device 1000 for AC / DC power systems is provided, wherein the similarity analysis and ranking module 1050 is specifically used for: For any analog signal corresponding to the fault hypothesis, calculate the Pearson correlation coefficient between the simulated waveform data and the fault recording data corresponding to the fault hypothesis; For any of the aforementioned fault assumptions, calculate the state consistency matching rate between the simulated waveform data and the fault recording data corresponding to the fault assumption. By integrating the Pearson correlation coefficients and state consistency matching rates of each channel through a weighted average method, the comprehensive similarity of each fault hypothesis is obtained, and the comprehensive similarity of each fault hypothesis is determined as the multidimensional feature similarity.

[0154] According to the present invention, a digital twin fault diagnosis device 1000 for AC / DC power systems includes both analog and digital signals in the simulated waveform data and the fault recording data; the analog signals include current waveforms and voltage waveforms, and the digital signals include protection trip signals and switch position status signals.

[0155] According to the present invention, a digital twin fault diagnosis device 1000 for AC / DC power systems is provided, wherein the similarity analysis and ranking module 1050 is further used for: The fault hypotheses are sorted in descending order according to the multidimensional feature similarity to obtain the sorting result; The sorting result is output as the fault diagnosis result.

[0156] According to the present invention, a digital twin fault diagnosis device 1000 for AC / DC power systems is provided, wherein the fault notification and data acquisition module 1010 is specifically used for: The power setting value and voltage reference value within the preset time window before the time of the fault occurrence are obtained through the analog quantity query interface; The switch status and protection activation / deactivation signals within the preset time window before the fault occurred are obtained through the digital status interface. The power setting value, the voltage reference value, the switch status, and the protection activation / deactivation signal are determined as the operating condition data.

[0157] According to the present invention, a digital twin fault diagnosis device 1000 for AC / DC power systems is provided, wherein the fault notification and data acquisition module 1010 is further configured to: The fault notification is received through the data interface layer, and the fault inversion process is initiated. The fault notification is validated by parameters; the parameter validation includes at least one of the following: time format validation, time reasonableness validation, and data integrity validation; If the fault notification passes the parameter verification, three parallel data acquisition channels are activated to acquire the fuzzy fault description, the operating condition data, and the waveform file in parallel.

[0158] Figure 11 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 11As shown, the electronic device may include: a processor 1110, a communications interface 1120, a memory 1130, and a communication bus 1140, wherein the processor 1110, the communications interface 1120, and the memory 1130 communicate with each other via the communication bus 1140. The processor 1110 can call logic instructions in the memory 1130 to execute a digital twin fault diagnosis method for AC / DC power systems, the method including: In response to the received fault notification, the fault inversion process is triggered, and the operating condition data of the AC / DC power system within a preset time window before the fault occurred are obtained. The operating condition data is mapped to the initial section parameters of the digital twin simulation model to achieve state synchronization between the digital twin simulation model and the AC / DC power system. Extract fuzzy fault descriptions from the fault notifications, perform semantic parsing on the fuzzy fault descriptions, and generate at least one set of simulation parameter configurations corresponding to fault assumptions; Based on the initial cross-section parameters, multiple time-domain simulation tasks corresponding to each of the simulation parameter configurations are executed in parallel to generate simulation waveform data corresponding to each of the fault assumptions. Calculate the multidimensional feature similarity between the simulation waveform data and the fault recording data corresponding to each fault hypothesis, sort the at least one set of fault hypotheses based on the multidimensional feature similarity, and output the fault diagnosis result.

[0159] Furthermore, the logical instructions in the aforementioned memory 1130 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0160] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program that can be stored on a non-transitory computer-readable storage medium, wherein when the computer program is executed by a processor, the computer is able to execute the digital twin fault diagnosis method for AC / DC power systems provided by the above methods, the method comprising: In response to the received fault notification, the fault inversion process is triggered, and the operating condition data of the AC / DC power system within a preset time window before the fault occurred are obtained. The operating condition data is mapped to the initial section parameters of the digital twin simulation model to achieve state synchronization between the digital twin simulation model and the AC / DC power system. Extract fuzzy fault descriptions from the fault notifications, perform semantic parsing on the fuzzy fault descriptions, and generate at least one set of simulation parameter configurations corresponding to fault assumptions; Based on the initial cross-section parameters, multiple time-domain simulation tasks corresponding to each of the simulation parameter configurations are executed in parallel to generate simulation waveform data corresponding to each of the fault assumptions. Calculate the multidimensional feature similarity between the simulation waveform data and the fault recording data corresponding to each fault hypothesis, sort the at least one set of fault hypotheses based on the multidimensional feature similarity, and output the fault diagnosis result.

[0161] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the digital twin fault diagnosis method for AC / DC power systems provided by the methods described above, the method comprising: In response to the received fault notification, the fault inversion process is triggered, and the operating condition data of the AC / DC power system within a preset time window before the fault occurred are obtained. The operating condition data is mapped to the initial section parameters of the digital twin simulation model to achieve state synchronization between the digital twin simulation model and the AC / DC power system. Extract fuzzy fault descriptions from the fault notifications, perform semantic parsing on the fuzzy fault descriptions, and generate at least one set of simulation parameter configurations corresponding to fault assumptions; Based on the initial cross-section parameters, multiple time-domain simulation tasks corresponding to each of the simulation parameter configurations are executed in parallel to generate simulation waveform data corresponding to each of the fault assumptions. Calculate the multidimensional feature similarity between the simulation waveform data and the fault recording data corresponding to each fault hypothesis, sort the at least one set of fault hypotheses based on the multidimensional feature similarity, and output the fault diagnosis result.

[0162] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0163] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0164] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A digital twin fault diagnosis method for AC / DC power systems, characterized in that, include: In response to the received fault notification, the fault inversion process is triggered, and the operating condition data of the AC / DC power system within a preset time window before the fault occurred are obtained. The operating condition data is mapped to the initial section parameters of the digital twin simulation model to achieve state synchronization between the digital twin simulation model and the AC / DC power system. Extract fuzzy fault descriptions from the fault notifications, perform semantic parsing on the fuzzy fault descriptions, and generate at least one set of simulation parameter configurations corresponding to fault assumptions; Based on the initial cross-section parameters, multiple time-domain simulation tasks corresponding to each of the simulation parameter configurations are executed in parallel to generate simulation waveform data corresponding to each of the fault assumptions. Calculate the multidimensional feature similarity between the simulation waveform data and the fault recording data corresponding to each fault hypothesis, sort the at least one set of fault hypotheses based on the multidimensional feature similarity, and output the fault diagnosis result.

2. The digital twin fault diagnosis method for AC / DC power systems according to claim 1, characterized in that, The process of mapping the operating condition data to the initial cross-sectional parameters of the digital twin simulation model includes: The operating condition data is subjected to state standardization processing to obtain standardized operating condition data; the state standardization processing includes: converting Chinese state descriptions into digital quantities and unified physical units, and allocating and calculating power reference values ​​according to operating mode; The standardized working condition data is subjected to parameter rationality verification, and abnormal data is identified and corrected to obtain the initial cross-sectional parameters.

3. The digital twin fault diagnosis method for AC / DC power systems according to claim 1, characterized in that, The step of extracting a fuzzy fault description from the fault notification, performing semantic parsing on the fuzzy fault description, and generating at least one set of simulation parameter configurations corresponding to fault hypotheses includes: The protection action information and the fuzzy fault description are extracted from the fault notification using regular expressions or pattern matching techniques. The semantic parsing engine is invoked, and based on a predefined fuzzy semantic mapping rule library, the protection action information and the fuzzy fault description are converted into simulation parameter configurations corresponding to each fault hypothesis; the simulation parameter configurations include fault time, fault type, fault resistance, and fault parameters.

4. The digital twin fault diagnosis method for AC / DC power systems according to claim 1, characterized in that, Based on the initial cross-sectional parameters, multiple time-domain simulation tasks corresponding to each of the simulation parameter configurations are executed in parallel to generate simulation waveform data corresponding to each of the fault assumptions, including: A multi-threaded parallel computing architecture is adopted. By calling the electromagnetic transient simulation application programming interface in the cloud, the multiple time-domain simulation tasks are executed in parallel based on the initial cross-sectional parameters, and the simulation waveform data corresponding to each fault assumption are obtained.

5. The digital twin fault diagnosis method for AC / DC power systems according to any one of claims 1-4, characterized in that, The calculation of the multidimensional feature similarity between the simulated waveform data and the fault recording data corresponding to each fault hypothesis includes: For any analog signal corresponding to the fault hypothesis, calculate the Pearson correlation coefficient between the simulated waveform data and the fault recording data corresponding to the fault hypothesis; For any of the aforementioned fault assumptions, calculate the state consistency matching rate between the simulated waveform data and the fault recording data corresponding to the fault assumption. By integrating the Pearson correlation coefficients and state consistency matching rates of each channel through a weighted average method, the comprehensive similarity of each fault hypothesis is obtained, and the comprehensive similarity of each fault hypothesis is determined as the multidimensional feature similarity.

6. The digital twin fault diagnosis method for AC / DC power systems according to claim 5, characterized in that, The simulated waveform data and the fault recording data both include analog signals and digital signals; the analog signals include current waveforms and voltage waveforms, and the digital signals include protection trip signals and switch position status signals.

7. The digital twin fault diagnosis method for AC / DC power systems according to any one of claims 1-4, characterized in that, The process of ranking the at least one set of fault hypotheses based on the similarity of each of the multidimensional features and outputting fault diagnosis results includes: The fault hypotheses are sorted in descending order according to the multidimensional feature similarity to obtain the sorting result; The sorting result is output as the fault diagnosis result.

8. The digital twin fault diagnosis method for AC / DC power systems according to any one of claims 1-4, characterized in that, The acquisition of operating condition data of the AC / DC power system within a preset time window before the fault occurs includes: The power setting value and voltage reference value within the preset time window before the time of the fault occurrence are obtained through the analog quantity query interface; The switch status and protection activation / deactivation signals within the preset time window before the fault occurred are obtained through the digital status interface. The power setting value, the voltage reference value, the switch status, and the protection activation / deactivation signal are determined as the operating condition data.

9. The digital twin fault diagnosis method for AC / DC power systems according to any one of claims 1-4, characterized in that, Before responding to the received fault notification, triggering the fault inversion process, and obtaining the operating condition data of the AC / DC power system within a preset time window before the fault occurred, the process further includes: The fault notification is received through the data interface layer, and the fault inversion process is initiated. The fault notification is validated by parameters; the parameter validation includes at least one of the following: time format validation, time reasonableness validation, and data integrity validation; If the fault notification passes the parameter verification, three parallel data acquisition channels are activated to acquire the fuzzy fault description, the operating condition data, and the waveform file in parallel.

10. A digital twin fault diagnosis device for AC / DC power systems, characterized in that, include: The fault notification and data acquisition module is used to respond to the received fault notification, trigger the fault inversion process, and acquire the operating condition data of the AC / DC power system within a preset time window before the fault occurs. The operating condition synchronization and parameter mapping module is used to map the operating condition data to the initial section parameters of the digital twin simulation model, so as to realize the state synchronization between the digital twin simulation model and the AC / DC power system. The pattern mapping and hypothesis generation module is used to extract fuzzy fault descriptions from the fault notifications, perform semantic parsing on the fuzzy fault descriptions, and generate simulation parameter configurations corresponding to at least one set of fault hypotheses. The parallel simulation execution module is used to execute multiple time-domain simulation tasks corresponding to each of the simulation parameter configurations in parallel based on the initial cross-section parameters, and generate simulation waveform data corresponding to each of the fault assumptions. The similarity analysis and ranking module is used to calculate the multidimensional feature similarity between the simulation waveform data and the fault recording data corresponding to each fault hypothesis, and to rank the at least one set of fault hypotheses based on the multidimensional feature similarity, and output the fault diagnosis result.

Images