Low-voltage distribution network fault diagnosis method based on light storage device trusted access

Abstract

The application discloses a low-voltage area distribution network fault diagnosis method based on light storage equipment trusted access, and belongs to the technical field of power grid fault diagnosis. The steps comprise the following: in a trusted computing environment constructed by a security chip, a protocol deep analysis is performed on a data packet of a light storage equipment to obtain trusted standardized data; a minimum feature subset required for fault diagnosis is dynamically selected based on a maximum correlation minimum redundancy algorithm, and encrypted transmission is collected to obtain efficient sensing data; the efficient sensing data is fused with historical fault cases to construct an edge fault diagnosis knowledge unit; when a fault occurs, the edge fault diagnosis knowledge unit is used to perform parallel execution of fault type rapid identification and power supply recovery potential pre-evaluation. The application adopts protocol deep analysis and a trusted computing environment, can ensure the integrity and security of non-standard access data, and dynamically optimizes the dimension of measurement data through a maximum correlation minimum redundancy algorithm, thereby improving the real-time performance and efficiency of fault type rapid identification and power supply recovery potential pre-evaluation.

Description

Technical Field

[0001] This invention belongs to the field of power grid fault diagnosis technology, and in particular relates to a fault diagnosis method for low-voltage distribution networks based on reliable access of photovoltaic and energy storage equipment. Background Technology

[0002] With the transformation of the energy structure, the construction of new power systems is accelerating. Among these, the low-voltage distribution network, as a crucial link in supporting the integration of clean energy sources such as distributed photovoltaic (PV) and energy storage, is increasingly important for its intelligent management. At the end of low-voltage distribution areas, the large-scale integration of distributed photovoltaic (PV) and energy storage (ESS) devices is transforming the distribution network from a traditional unidirectional radial network to a complex network with multi-level interaction between source, grid, load, and storage. To effectively support, efficiently absorb, and safely and stably operate these dispersed and heterogeneous resources, unified and precise data acquisition, status monitoring, and flexible control of the end-point PV and energy storage devices are required.

[0003] Existing research, CN119231510A, discloses an intelligent management system for photovoltaic (PV) and energy storage (ESS) equipment, including PV equipment, PV inverters, energy storage systems, loads, distribution networks, transformers, and a PV-ESS management platform. The technical solution in this application primarily focuses on performance prediction and optimized scheduling of the upper-level management platform, relying on deep learning and improved particle swarm optimization algorithms to minimize economic costs. However, in practical low-voltage distribution area applications, the variety of end-point PV and ESS equipment, the complexity of communication protocols, and the existence of numerous non-standard protocols result in a lack of unified standards for underlying data acquisition and communication, making it difficult to achieve efficient and reliable data acquisition from heterogeneous resources.

[0004] Meanwhile, existing research has failed to fully consider the security and integrity of the data source, and cannot effectively verify the reliability of the collected electrical variable data. Once the underlying data is tampered with, it will directly lead to a reduction in the accuracy of upper-level optimization scheduling and fault diagnosis. Especially in complex scenarios of rapid recovery after a power outage due to a distribution network fault, the lack of reliable edge diagnostic capabilities for the status, fault type, and power supply potential of photovoltaic and energy storage equipment makes it difficult to make rapid and accurate response decisions to emergencies. This limits the improvement of the intelligent management level of the distribution network and the enhancement of user satisfaction with electricity use. Summary of the Invention

[0005] To address the aforementioned issues, the present invention aims to provide a fault diagnosis method for low-voltage distribution networks based on trusted access to photovoltaic and energy storage devices. This method employs deep protocol parsing and a trusted computing environment to ensure the integrity and security of non-standard access data. Furthermore, it dynamically optimizes the dimensions of measurement data through a maximum correlation minimum redundancy algorithm, thereby improving the real-time performance and efficiency of rapid fault type identification and power restoration potential pre-assessment.

[0006] To achieve the above objectives, this application adopts the following technical solution:

[0007] A fault diagnosis method for low-voltage distribution networks based on trusted access to photovoltaic and energy storage devices includes the following steps:

[0008] S1. In a trusted computing environment built with a security chip, the data packets obtained from the optical storage device are subjected to deep application layer protocol parsing to obtain the parsed data;

[0009] S2. Standardize the parsed data and add a data imprint containing timestamps and integrity check codes to it using a trusted computing environment to obtain trusted standardized data.

[0010] S3. Obtain the target variable used to distinguish different types of distribution network faults, and dynamically select features from the reliable standardized data based on the target variable using the maximum correlation minimum redundancy algorithm to obtain the minimum feature subset;

[0011] S4. Collect and encrypt the measurement data corresponding to the smallest feature subset to obtain efficient sensing data;

[0012] S5. Obtain historical fault case data and integrate the historical fault case data with efficient perception data to construct an edge fault diagnosis knowledge unit;

[0013] S6. When a fault occurs, the fault type is quickly identified and the power supply recovery potential is pre-assessed in parallel based on the edge fault diagnosis knowledge unit to obtain the diagnosis and assessment results, and the diagnosis and assessment results are reported to the upper-level master station.

[0014] Preferably, in S1, the parsed data includes:

[0015] Data packets from the optical storage device are acquired in parallel via power line communication channels and wireless communication channels.

[0016] Obtain a protocol feature library for identifying the communication mechanisms and data formats of application layer protocols;

[0017] In a trusted computing environment, parsed data is generated by matching the application layer content of data packets with the prefixes and suffixes of a protocol signature library.

[0018] Preferably, in S2, the obtained reliable standardized data includes:

[0019] Obtain a mapping rule library for unifying data fields, and perform format unification processing on the parsed data based on the mapping rule library to generate a standardized data stream;

[0020] In a trusted computing environment, a hardware cryptographic engine is used to compute integrity verification codes for standardized data streams.

[0021] The time stamp and integrity check code are encapsulated into a standardized data stream to generate trusted standardized data.

[0022] Preferably, in S3, the minimum feature subset includes:

[0023] Calculate the mutual information between each feature and the target variable in the credible standardized data to obtain the feature correlation index;

[0024] Calculate the mutual information between each feature to obtain the feature redundancy index;

[0025] Based on feature correlation and feature redundancy indices, features are iteratively selected to construct a minimum feature subset.

[0026] Preferably, in S3, after obtaining the minimum feature subset, the process further includes updating the minimum feature subset, specifically as follows:

[0027] Obtain operating parameters that characterize the current distribution network status;

[0028] Based on the operating parameters, the target variable is adjusted through decision logic to obtain a new target variable;

[0029] Based on the new target variable, the feature correlation index and feature redundancy index are recalculated, and the minimum feature subset is updated.

[0030] Preferably, in S4, the efficient sensing data obtained includes:

[0031] Extract the corresponding voltage, current, and power measurement data from the smallest feature subset;

[0032] Obtain the symmetric encryption key used to encrypt measurement data;

[0033] The measurement data is encrypted using a symmetric encryption key to generate encrypted measurement data.

[0034] The encrypted measurement data is transmitted to the edge processing to obtain efficient sensing data.

[0035] Preferably, in S5, the edge fault diagnosis knowledge unit constructed includes:

[0036] Use real-time feature values ​​from the data to efficiently detect query indexes;

[0037] Pattern matching and correlation analysis are performed on historical fault case data to filter out historical fault cases that are similar to the current power grid status.

[0038] By integrating the selected historical fault cases with efficient perception data, edge fault diagnosis knowledge units are generated.

[0039] Preferably, in S6, the diagnostic and evaluation results include:

[0040] Use the highly efficient sensing data captured at the moment of the current fault as a real-time query mode;

[0041] In the edge fault diagnosis knowledge unit, the type of fault, such as short circuit, open circuit or ground fault, is identified by comparing the similarity between the real-time query mode and the historical fault mode.

[0042] Acquire real-time data on the capacity and distribution of optical storage devices, and pre-calculate reliability indicators for power restoration based on this data;

[0043] By integrating fault types and reliability metrics, diagnostic and assessment results are generated.

[0044] Preferably, in S6, after obtaining the diagnostic and evaluation results, the following is also included:

[0045] Obtain the distribution network topology information that depicts the interconnections within the transformer substation area;

[0046] Downstream tracing analysis is performed based on distribution network topology information and edge fault diagnosis knowledge units to assess the scope of fault impact.

[0047] Obtain critical load level information, and calculate the recovery scheduling priority based on critical load level information, fault impact range and real-time optical storage capacity data;

[0048] The scope of the fault impact and the priority of recovery scheduling are integrated into the diagnosis and assessment results.

[0049] The present invention has the following advantages:

[0050] This invention ensures that measurement data acquired from heterogeneous photovoltaic and energy storage devices is format-consistent, original, reliable, and tamper-proof by parsing the source data and adding data imprinting within a trusted computing environment built on a secure chip. This invention establishes a full-link trust mechanism from data generation to application, improving the data quality relied upon for subsequent fault diagnosis and decision analysis, and laying a solid data foundation for accurate power grid condition assessment.

[0051] This invention utilizes the Maximum Relevance Minimum Redundancy (mRMR) algorithm to dynamically optimize and select features from massive measurement data, constructing a minimal feature subset with high information density and small size. Based on this subset, targeted data acquisition and encrypted transmission are performed, reducing the consumption of communication bandwidth and edge computing resources, and achieving lightweight and high-efficiency sensing of the power grid status. This invention's intelligent dimensionality reduction method effectively solves the shortcomings of data redundancy and transmission congestion in traditional monitoring systems.

[0052] This invention deeply integrates edge computing and artificial intelligence technologies. By constructing an edge fault diagnosis knowledge unit at the end of the transformer substation that combines real-time sensing data with historical fault cases, it achieves localization and proactive diagnostic capabilities. When a fault occurs, this knowledge unit supports the parallel and rapid execution of fault type identification and power restoration potential assessment, transforming passive fault reporting into proactive situation analysis and shortening fault response time.

[0053] This invention further incorporates distribution network topology information and critical load levels into the edge-side analysis and decision-making process, enabling accurate assessment of the fault impact range and intelligent calculation of recovery scheduling priorities. This invention not only diagnoses faults but also provides specific and actionable recovery strategy suggestions, making fault handling more comprehensive and forward-looking, effectively ensuring the power supply reliability of critical users, and improving the overall resilience and self-healing level of the distribution network. Attached Figure Description

[0054] Figure 1 This is a flowchart illustrating the fault diagnosis method for low-voltage distribution networks based on trusted access of photovoltaic and energy storage devices according to the present invention.

[0055] Figure 2 This is a performance comparison chart of the minimum feature subset in Embodiment 1 of this application;

[0056] Figure 3 This is a visualization of the fault recovery scheduling priority in Embodiment 1 of this application;

[0057] Figure 4 This is a schematic diagram of the module of the low-voltage distribution network fault diagnosis system based on trusted access of photovoltaic and energy storage equipment according to the present invention. Detailed Implementation

[0058] Example 1: As Figure 1 As shown, the fault diagnosis method for low-voltage distribution networks based on trusted access of photovoltaic and energy storage devices includes the following steps:

[0059] S1. In a trusted computing environment built with a security chip, the data packets obtained from the optical storage device are subjected to deep application layer protocol parsing to obtain the parsed data;

[0060] S2. Standardize the parsed data and add a data imprint containing timestamps and integrity check codes to it using a trusted computing environment to obtain trusted standardized data.

[0061] S3. Obtain the target variable used to distinguish different types of distribution network faults, and dynamically select features from the reliable standardized data based on the target variable using the maximum correlation minimum redundancy algorithm to obtain the minimum feature subset;

[0062] S4. Collect and encrypt the measurement data corresponding to the smallest feature subset to obtain efficient sensing data;

[0063] S5. Obtain historical fault case data and integrate the historical fault case data with efficient perception data to construct an edge fault diagnosis knowledge unit;

[0064] S6. When a fault occurs, the fault type is quickly identified and the power supply recovery potential is pre-assessed in parallel based on the edge fault diagnosis knowledge unit to obtain the diagnosis and assessment results, and the diagnosis and assessment results are reported to the upper-level master station.

[0065] By performing deep protocol parsing in a hardware-level trusted computing environment built with secure chips, and adding data imprints containing timestamps and integrity check codes to the parsed data, the originality and immutability of the data accessed by optical storage devices are ensured from the source, thus building a unified, trusted, and standardized data foundation.

[0066] Building upon this foundation, a maximum relevance and minimum redundancy algorithm is used to dynamically optimize the features of massive amounts of data, selecting the smallest subset of features that are most strongly correlated with distribution network fault types and have the lowest information redundancy. This guides efficient data acquisition and encrypted transmission. Furthermore, at the edge, this efficiently perceived real-time data is integrated with accumulated historical fault cases to construct an edge fault diagnosis knowledge unit with context-aware capabilities. When a fault occurs, this knowledge unit supports the parallel and rapid execution of two tasks: fault type identification and power restoration potential assessment. Ultimately, it generates and reports comprehensive diagnostic and assessment results, achieving a closed loop from reliable data acquisition and efficient perception to edge intelligent decision-making.

[0067] In S1, the parsed data includes:

[0068] S11. Acquire data packets from the optical storage device in parallel through power line communication channel and wireless communication channel;

[0069] Operational data packets are collected from the optical storage equipment at the end of the low-voltage distribution area using two independent physical links: power line communication channels and wireless communication channels. Power line communication channels utilize existing power cables for data transmission, offering advantages such as low deployment costs and wide coverage; wireless communication channels, such as LoRa or 4G / 5G, provide high bandwidth and flexible access capabilities. Parallel acquisition improves the reliability and real-time performance of data collection, ensuring uninterrupted data flow even when a single channel is interfered with or interrupted.

[0070] In a low-voltage distribution area, the energy converter simultaneously acquires data via a narrowband power line carrier (NB-PLC) deployed on the transformer side and a built-in 4G channel. When a household's photovoltaic inverter communicates upstream via the low-speed, high-interference-resistance NB-PLC, the converter captures its periodic output data packets. Simultaneously, the household's energy storage device (ESS) transmits real-time state of charge (SOC) and alarm data packets via the 4G wireless network. The energy converter receives data streams from these two independent channels in parallel. Even if the NB-PLC channel experiences a rate reduction due to increased line noise, the 4G channel ensures the real-time acquisition of critical data packets, maintaining data continuity.

[0071] S12. Obtain a protocol feature library for identifying the communication mechanisms and data formats of application layer protocols;

[0072] After acquiring data packets, a pre-built protocol feature library is loaded. This library is the core for achieving compatibility analysis of protocols for optical storage devices from different manufacturers. The protocol feature library stores the communication mechanisms and data format characteristics of various application layer protocols. These characteristics exist in the form of explicit prefix and suffix code streams, key field identifiers, or specific data frame structures, forming fingerprints that identify different protocols.

[0073] Before initiating protocol parsing, the energy converter loads a pre-built protocol signature library. This library stores the communication fingerprints of mainstream photovoltaic and energy storage devices. For example, a manufacturer A's energy storage BMS system uses the CAN protocol, and its signature is stored as follows: the start flag of the application layer data packet is a specific CAN ID, such as 0x18FEE000, and the 5th byte of the data body is fixed as the ASCII code of the battery temperature, i.e., the prefix signature; while another manufacturer B's photovoltaic inverter uses a proprietary Modbus variant protocol, and its signature is stored as follows: the end of the data frame is fixed to include a CRC checksum followed by a specific frame tail marker 0xFE, i.e., the suffix signature.

[0074] S13. In a trusted computing environment, parsed data is generated by matching the application layer content of data packets with the protocol feature library by prefix and suffix.

[0075] The entire protocol parsing process is executed in a trusted computing environment isolated by a secure chip hardware. This environment ensures the confidentiality and integrity of the protocol parsing logic and the data to be processed. Within this secure environment, the application layer content of the captured data packets, i.e., the business data payload, is deeply matched against features in the protocol feature library. This matching process uses a prefix and suffix pattern recognition algorithm to scan the application layer content of the data packets, searching for start and end markers that match any protocol feature stored in the library. Once a match is successful, the specific application layer protocol followed by the data packet can be determined, and based on the data format definition of that protocol, the effective payloads such as voltage, current, power, and device status are accurately extracted, thereby generating structured parsed data suitable for subsequent processing.

[0076] Within a trusted computing environment isolated by a security chip, the energy converter receives data packets from a photovoltaic inverter. Protocol deep parsing extracts the application-layer payload from the data packets and compares it against a protocol signature library. First, scanning the payload's start end reveals no match for any known standard Modbus start flag; however, subsequent suffix matching identifies a specific frame tail flag, 0xFE, at the end of the payload, representing a proprietary Modbus variant protocol from vendor B's inverter.

[0077] After a successful match, it is confirmed that the data packet conforms to the protocol of vendor B. According to the definition of the protocol, the 16-bit unsigned integer code located at a specific offset in the data body, such as offset12, is accurately extracted and determined to be the real-time active power value of the inverter output, which is 15.3kW, thereby generating structured parsed data.

[0078] In S2, the obtained reliable standardized data includes:

[0079] S21. Obtain the mapping rule base for unifying data fields, and perform format unification processing on the parsed data based on the mapping rule base to generate a standardized data stream;

[0080] A mapping rule library for unifying data fields is loaded. This library predefines a set of standard data naming conventions, units, and formats. For example, it maps heterogeneous fields such as "Voltage A," "Ua," and "Phase_A_Voltage" reported by different manufacturers' devices to the standard "Phase A Voltage," specifying that the unit is volts and the data type is floating-point. Based on this mapping rule library, each piece of parsed data is traversed and matched, and the various data fields are renamed, converted in unit, and formatted according to the rules. After this format unification process, a standardized data stream with consistent content and a well-organized structure is generated.

[0081] The energy converter receives the parsed data PV_Volt_1 field (unit: mV) reported by photovoltaic inverter A and the parsed data ESS_Ua field (unit: V) reported by energy storage device B. It loads a mapping rule library, defining that both PV_Volt_1 and ESS_Ua must be mapped to the standard field "A-phase voltage," with the unit uniformly in V and the data type floating-point. The value of PV_Volt_1 is converted by a 1000-fold division, and both its value and the field name of ESS_Ua are replaced with "A-phase voltage," ultimately generating a standardized data stream with unified structure and units, such as in JSON format.

[0082] .

[0083] S22. Calculate integrity verification codes for standardized data streams using a hardware cryptographic engine in a trusted computing environment;

[0084] The standardized data stream is fed into a hardware cryptographic engine for processing within a trusted computing environment built on a secure chip. The hardware cryptographic engine utilizes efficient and secure hardware circuitry to execute a specified cryptographic hash algorithm, such as SHA-256, on the entire standardized data stream, thereby calculating a fixed-length digest value, which is the integrity check code. This process can be represented as:

[0085] ;

[0086] Where S represents the standardized data stream after format unification, HASH represents the hash function executed by the hardware cryptographic engine, and C1 is the final generated integrity check code.

[0087] Inside the energy converter's security chip, a standardized data stream is fed into a dedicated hardware SHA-256 engine. This engine performs a hash operation on the entire input standardized data stream, including all field names and values, with extremely high speed and security, calculating a unique 256-bit digest value H. This digest value H serves as the integrity check code for the data stream, ensuring that the hash value calculation process is completely isolated from the external environment during this critical data generation stage, preventing software-level tampering and malicious injection.

[0088] Assume that the standardized data stream L is a binary serialized content:

[0089] ;

[0090] in, This represents the voltage of phase A. Indicates active power;

[0091] The hardware SHA-256 engine performs a hash operation on the data L, obtaining a 256-bit digest value H. If any bit in the data stream L changes, for example... become Calculated This will be completely different, thus verifying the integrity of the data.

[0092] S23. Encapsulate the time stamp and integrity check code into a standardized data stream to generate trusted standardized data.

[0093] A high-precision, tamper-proof time stamp is obtained from a secure time source and encapsulated along with a calculated integrity check code into a standardized data stream's data structure, typically as a metadata field in the header or footer. This encapsulation process adds a data imprint, ultimately outputting trusted, standardized data with a trusted identity and integrity verification.

[0094] The system obtains the current high-precision time stamp from a secure time synchronization source, such as an internal TCXO crystal oscillator synchronizing with an external GPS / BeiDou time synchronization signal, for example, 2025-11-18T09:00:00.123Z. Subsequently, this timestamp, along with a 256-bit integrity check code H, is appended and encapsulated as a "data imprint" metadata field to the end of the standardized data stream's data frame. The final output data frame not only contains the operational information of the optical storage device but also possesses a reliable timestamp and complete tamper-proof verification, becoming trusted standardized data that can be consumed by upper-layer secure applications.

[0095] In S3, the minimum feature subset includes:

[0096] S31. Calculate the mutual information between each feature and the target variable in the credible standardized data to obtain the feature correlation index;

[0097] The target variable is a discrete label used to distinguish different types of distribution network faults. For example, 0 represents normal operation, 1 represents a single-phase short circuit, 2 represents a two-phase short circuit, and 3 represents a line break fault. This target variable usually corresponds one-to-one with reliable standardized data in a historical fault case library, together forming a labeled training dataset.

[0098] The core stage of feature selection involves two key computational steps. The first step is calculating the feature correlation index, which involves calculating the mutual information between each feature in the reliable standardized data, such as phase A voltage, phase B current, and active power, and the target variable. Mutual information is an indicator that measures the degree of statistical dependence between two variables, and its calculation formula is:

[0099] ;

[0100] Where F represents a feature variable to be evaluated, and T represents the target variable. It is the joint probability distribution of feature F taking the value f and target variable T taking the value t, while and These are marginal probability distributions. These probability values ​​can be obtained through statistical analysis of labeled, historically reliable, and standardized data. A feature's relevance index. The larger the value, the stronger the correlation between the feature and the fault type.

[0101] Assume the target variable y represents "fault type", with 0 being normal, 1 being short circuit, and 2 being open circuit. The characteristic variables to be evaluated are... Including phase A current And active power P. Statistical analysis of historical reliable standardized data revealed that when a short-circuit fault occurs (y=1), Values ​​in Joint probability of intervals It is very high, while the value of P is in joint probability Very low. By substituting into the mutual information formula... The current in phase A can be calculated. Mutual information value with fault type y The mutual information value of active power P is 0.85. It is 0.42, thus determining A higher characteristic correlation index indicates a stronger association with distinguishing fault types. Assuming that the A-phase current is obtained by statistically analyzing 1000 historical fault cases... The joint probability distribution of discrete values ​​and fault type y As shown in Table 1:

[0102] Table 1. Joint probability distribution of discrete values ​​of phase A current and fault type

[0103]

[0104] Mutual Information It is obtained by summing the four terms:

[0105] Normal value retrieval , normal :

[0106] ;

[0107] Normal value retrieval , Short circuit :

[0108] ;

[0109] Abnormal value , normal :

[0110] ;

[0111] Abnormal value , Short circuit :

[0112] ;

[0113] Total mutual information value After detailed calculations, the mutual information value between phase A current and fault type was determined. Approximately This value is used to compare with other characteristics, such as the mutual information value of 0.42 for active power P, to determine the characteristic correlation index of the A-phase current.

[0114] S32. Calculate the mutual information between each feature to obtain the feature redundancy index;

[0115] The second step is to calculate the feature redundancy index, which is to calculate the mutual information between any two features. The higher the value of this indicator, the greater the overlap in information provided by these two features, indicating redundancy.

[0116] After evaluating the correlation between all features and the target variable, the redundancy between features is calculated. Assuming that, except for phase A current... And the voltage of phase A The volatility and harmonic distortion rate of phase A current As candidate features. Calculate. and Mutual information between It was found that since the larger the current, the higher the harmonic distortion rate is usually, the two are statistically highly dependent. The calculated mutual information value of 0.78 is very high, indicating that... and There is information redundancy between them. In contrast, and A low mutual information value for volatility indicates that the two provide little overlap in the information they offer.

[0117] S33. Based on feature correlation index and feature redundancy index, iteratively select features to construct the minimum feature subset.

[0118] After calculating these two metrics, an iterative approach is used to construct a minimum feature subset. The feature with the highest mutual information with the target variable, i.e., the highest feature relevance index, is selected as the first member of the initial subset. In each iteration, the algorithm selects a feature from the unselected features that maximizes the following criterion for adding to the subset, which combines relevance and redundancy:

[0119] ;

[0120] in, E represents the candidate features, and E is the currently selected subset of features. It is the size of the subset. It is the correlation between candidate features and the target variable. It is the candidate feature and all selected features in the subset. The sum of redundancies. Based on this criterion, the algorithm selects, at each step, the feature most relevant to the target variable and least redundant with the selected features. This iterative process continues until the feature subset reaches a preset size or the gain is no longer significant; the final feature set is the minimum feature subset.

[0121] In the initial stage, the algorithm selects the A-phase current with the highest feature correlation index. Add subset E. In subsequent iterations, the algorithm evaluates all unselected features, such as phase A voltage. volatility and .Although Correlation with the target variable It's also very high, but compared to the selected features... Redundancy Extremely high. And... Although volatility is correlated Slightly lower, but it is comparable to The redundancy is extremely low. According to the mRMR criterion... After calculation, it was found that Volatility has the highest gain value, so it is selected to be added to the subset, thereby ensuring that the features in the subset can effectively distinguish the fault types while minimizing the duplication of information between them.

[0122] Assuming the algorithm selects the A-phase current with the highest feature correlation index. As Add subset E, at this time In subsequent iterations, the algorithm evaluates two candidate features: candidate features Harmonic distortion rate: Highly correlated with short-circuit faults, but not with the selected phase A current. High redundancy. Correlation. Redundancy Candidate features Volatility / Voltage Volatility): Moderately correlated with short-circuit faults, but significantly correlated with the selected phase A current. Almost no redundancy.

[0123] Correlation Redundancy Calculate the mRMR gain value: , Gain: , Gain in volatility: After calculation using the mRMR criterion, it was found that... The feature with the highest gain (volatility) is selected for inclusion in the subset by the algorithm. This ensures that the features in the subset can effectively distinguish fault types while minimizing information duplication, thus achieving efficient perception.

[0124] In S3, after obtaining the minimum feature subset, the process also includes updating the minimum feature subset, specifically:

[0125] S34. Obtain the operating parameters that characterize the current distribution network status;

[0126] A series of operating parameters characterizing the current state of the distribution network are acquired in real time. These operating parameters are not high-frequency electrical quantity measurements, but rather quasi-static information describing the macroscopic operating background of the power grid, such as whether the current load level is peak or off-peak, whether it is summer or winter, the penetration rate and real-time power generation of distributed photovoltaic systems, and whether there are line maintenance or topology changes. These operating parameters together depict the current operating condition of the power grid.

[0127] Real-time acquisition of macroscopic operating parameters of the current distribution network, including: time stamped as 12:00 noon in summer, peak load and peak photovoltaic generation period, overall photovoltaic penetration rate of the distribution network of 40%, real-time load rate of distribution transformers of 85%, and monitoring of a 10kV line undergoing planned maintenance resulting in temporary topology changes. These parameters together constitute the background operating conditions of the current power grid facing high-voltage operation and topological constraints.

[0128] S35. Based on the operating parameters, adjust the target variable through decision logic to obtain a new target variable;

[0129] After acquiring the operating parameters, a preset decision logic is activated. This logic is a set of expert rules or a small decision model. Its core function is to dynamically adjust the target variable used for fault diagnosis based on the input operating parameters. For example, during the summer peak load and solar power generation period, the fault patterns in the distribution network may be more inclined towards high-impedance grounding or voltage anomalies caused by distributed power source disturbances. In this case, the decision logic will adjust the original general target variable to a new target variable that focuses more on identifying this specific type of fault. This adjustment does not change the physical nature of the fault, but rather changes the priority and granularity of the diagnostic task.

[0130] The decision logic receives operating parameters based on peak summer load and a 40% photovoltaic penetration rate. Expert rules determine that under these conditions, the fault is more likely to be a high-voltage transient fault caused by distributed generation (DG) or a high-impedance grounding fault under heavy load. Therefore, the decision logic will use the original general objective variable... Adjust to the new target variable ,in Two subcategories, “3: High-voltage transient faults caused by DG” and “4: High-impedance grounding”, have been added, and the diagnostic weight priority of the original short-circuit fault y=1 has been reduced.

[0131] S36. Based on the new target variable, recalculate the feature correlation index and feature redundancy index, and update the minimum feature subset.

[0132] Once the new target variable is obtained, the feature selection process will be re-executed based on it. The updated labeled historical data will be used to recalculate the mutual information between each feature in the trusted standardized data and the new target variable, resulting in a completely new set of feature correlation indices. Due to the change in the target variable, features that were previously not strongly correlated with common fault types may now show a high correlation with new fault types under specific operating conditions.

[0133] Subsequently, based on this new set of feature correlation indices and the previously calculated feature redundancy indices, the iterative selection process of the maximum correlation minimum redundancy algorithm is run again to construct a new minimum feature subset that adapts to the current distribution network state. For example... Figure 2 As shown, Figure 2 The graphs compare the fault diagnosis accuracy of the mRMR criterion and the maximum correlation (MR) criterion alone under different numbers of features. The graphs show that the mRMR criterion can achieve the same diagnostic accuracy with fewer features, effectively eliminating redundant information between features and realizing efficient perception of the power grid status.

[0134] Use the new target variable Based on high-voltage transient fault categories, the characteristics were recalculated. The mutual information. The results show that, Previously unimportant features, such as the abrupt change rate of the positive-sequence voltage component and the rate of change of harmonic content, showed significantly enhanced correlation with the "DG-induced high-voltage transient fault" category, resulting in improved feature correlation indices. Subsequently, using these updated feature correlation indices, the mRMR iterative selection process was run again to select the minimum feature subset. The high redundancy current peak feature is replaced with the highly correlated voltage positive sequence component mutation rate, thereby constructing a minimal feature subset that is suitable for the current high photovoltaic penetration conditions.

[0135] In S4, the efficient sensing data obtained includes:

[0136] S41. Extract the corresponding voltage, current and power measurement data from the smallest feature subset;

[0137] A minimal subset of features is used as the precise instruction for data acquisition. This subset contains a series of feature names, such as the RMS value of phase A voltage, the instantaneous value of phase B current, or the total three-phase active power. Based on these names, a targeted request is sent to the data source, i.e., the photovoltaic storage device or its interface unit, to collect only the real-time measurement data corresponding to these specific features, thereby extracting a subset of measurement data containing key information such as voltage, current, and power.

[0138] Assuming that the minimum feature subset is determined through mRMR algorithm filtering as {instantaneous value of phase A current, effective value of phase B voltage, and total three-phase active power}, this subset is sent as a command to the photovoltaic-storage equipment interface unit. The interface unit accurately extracts the three data values ​​corresponding to the current moment from the real-time data cache of the inverter and energy storage BMS: instantaneous value of phase A current 80.5A, effective value of phase B voltage 219.8V, and total three-phase active power 5.5kW. This avoids the collection and transmission of more than 30 unnecessary non-critical features, such as temperature, frequency, and reactive power.

[0139] S42. Obtain the symmetric encryption key used to encrypt the measurement data;

[0140] After data acquisition is complete, a symmetric encryption key is needed to ensure data security during transmission. This key is typically generated and distributed by a secure key management center and pre-distributed to edge acquisition devices and processing units via a secure channel, ensuring that both communicating parties possess the same key.

[0141] The energy converter retrieves the symmetric encryption key used for this communication from the security key storage area inside the security chip. The key It is a 256-bit random string that has passed security authentication and key negotiation mechanisms during the initialization phase. It is issued from the upper-level key management center and pre-placed in the energy converter and designated edge processing unit to ensure the confidentiality of the key during transmission and storage.

[0142] S43. Use a symmetric encryption key to encrypt the measurement data and generate encrypted measurement data;

[0143] After obtaining the key, the subset of collected measurement data will be encrypted. This process utilizes standard symmetric encryption algorithms, such as the Advanced Encryption Standard (AES), to transform the data. The process is represented as follows:

[0144] ;

[0145] in, This represents the raw measurement data extracted from the smallest feature subset. Se represents the symmetric encryption key, and Se represents the symmetric encryption function. It is the generated encrypted measurement data.

[0146] A subset of the collected raw measurement data, such as the A-phase current of 80.5A, is encrypted using the obtained symmetric encryption key in the hardware encryption of the security chip. The data is encrypted using the AES-256 algorithm. The original data block is transformed into an unreadable ciphertext block. For example, the original 64-byte data is converted into 64-byte ciphertext, thus generating encrypted measurement data, effectively preventing eavesdropping or cracking during data transmission over the public network. Assuming a subset of the original measurement data... The corresponding 16-byte 128-bit block after formatting is:

[0147] ;

[0148] Use AES-256 symmetric encryption function and 256-bit key Encryption is performed, and the original data block is converted into an unreadable ciphertext block. :

[0149] ;

[0150] The ciphertext block is represented in 16-byte hexadecimal format. For example, the original 64-byte data is converted into 64-byte ciphertext, thus generating encrypted measurement data, effectively preventing data from being eavesdropped on or cracked during transmission over the public network.

[0151] S44. Transmit the encrypted measurement data to the edge processing to obtain efficient sensing data.

[0152] This encrypted measurement data is transmitted to a designated edge processing unit via power line communication or a wireless channel. This filtered and encrypted data, which ultimately arrives at the edge processing unit and is ready for fault diagnosis, is defined as high-efficiency sensing data.

[0153] The energy converter encapsulates the encrypted measurement data into TCP / IP packets and transmits them rapidly to the edge processing unit deployed in the regional substation via internal fiber optic or 4G channels. Because this data consists of a minimal feature set filtered by mRMR and is highly encrypted, it consumes few resources and has high security along the transmission path. When this data successfully arrives at the edge processing unit and awaits decryption and application, it forms efficient sensing data.

[0154] In S5, the edge fault diagnosis knowledge units constructed include:

[0155] S51. Use real-time feature values ​​in the efficiently perceived data as the query index;

[0156] The high-efficiency sensing data acquired in real time serves as the entry point for dynamic queries. When the edge processing unit receives a new set of high-efficiency sensing data, it extracts the real-time feature values ​​contained within, such as phase A voltage being 0V and phase B current being 300A, and uses this vector composed of multiple feature values ​​as the query index.

[0157] The edge processing unit receives new high-efficiency sensing data, which has been filtered by mRMR and includes the following real-time features: instantaneous A-phase current 300A, effective C-phase voltage 5V, and total three-phase active power 0kW. These values ​​are then processed... Extracted to form a three-dimensional real-time feature vector. This serves as the sole query index for subsequent retrievals in the historical case database.

[0158] S52. Perform pattern matching and correlation analysis on historical fault case data to filter out historical fault cases that are similar to the current power grid status.

[0159] This query index is used to perform a search on locally stored historical fault case data. This historical fault case data is a structured database containing a large number of confirmed distribution network fault events. Each case includes detailed information such as the feature value vector at the time of the fault, the ultimately diagnosed fault type, fault location, and environmental parameters. The core of the search is pattern matching and association analysis. The similarity between the current real-time feature value vector and the feature value vector of each historical fault case in the database is calculated. This similarity can be measured by calculating the distance between them, for example, using Euclidean distance.

[0160] ;

[0161] Where D represents the Euclidean distance between the two vectors. It is the real-time value of the z-th feature in the current high-efficiency sensing data. This is the z-th feature value corresponding to a certain historical fault case. All historical fault cases with a distance D less than a preset threshold are selected, meaning these cases are considered highly similar to the current power grid state in terms of electrical characteristics.

[0162] use As a query index, it iterates through 500 historical failure cases stored locally. For each case in the database... Calculate its relationship with The Euclidean distance D. For example, in a certain historical case. According to calculations, the distance Only 7.8; while another case distance Far exceeding 300. Filter out all distances. Cases that are highly similar to the current state in terms of electrical characteristics and whose historical diagnostic results are mostly "C-phase metallic grounding fault" are selected.

[0163] S53. Integrate the selected historical fault cases with efficient perception data to generate edge fault diagnosis knowledge units.

[0164] These selected similar historical fault cases, along with their accompanying complete information such as fault type and handling measures, are integrated with the current high-efficiency perception data. This integration is not a simple patchwork, but rather the construction of a completely new, content-rich composite data structure, which is the edge fault diagnosis knowledge unit. It includes "what is the current situation" and "what similar situations were like in the past."

[0165] The edge processing unit will select five similar historical fault cases, including their diagnosed fault type, location, duration, and real-time, efficient sensing data at the current moment. This is integrated into a composite data structure. This composite structure, namely the edge fault diagnosis knowledge unit, not only contains the currently observed electrical characteristics, but also the empirical information that "in 90% of similar historical cases, the fault was diagnosed as metallic grounding and located in the C branch of transformer area 8," providing direct knowledge support for subsequent rapid identification and assessment.

[0166] In S6, the diagnostic and assessment results include:

[0167] S61. Use the highly efficient sensing data captured at the moment of the current fault as a real-time query mode;

[0168] S6. When a fault occurs, the edge fault diagnosis knowledge unit performs parallel operations for rapid fault type identification and power restoration potential assessment. This process starts the instant the fault signal is triggered. The highly efficient sensing data captured at this instant, i.e., a vector containing key electrical characteristic values, is used as a real-time query mode.

[0169] When the transformer area protection device activates and triggers a fault signal, the power converter immediately locks the high-efficiency sensing data captured within 50 milliseconds prior to the protection activation. This data is a vector containing three feature values. For example, if the A-phase current sudden change value is +350A, the zero-sequence current value is 5A, and the B-phase voltage drop depth is 95%, this vector is set to real-time query mode.

[0170] S62. In the edge fault diagnosis knowledge unit, the type of short circuit, open circuit or ground fault is identified by comparing the similarity between the real-time query mode and the historical fault mode.

[0171] Two core tasks are carried out in parallel. The first task is rapid fault type identification. Within the edge fault diagnosis knowledge unit, the real-time query pattern is compared with multiple historical fault patterns extracted from historical fault cases stored in the knowledge unit. By calculating the vector distance or correlation coefficient between the real-time pattern and each historical pattern, one or more historical cases with the highest similarity can be quickly identified. Based on the confirmed fault labels recorded in these highly matched historical cases, the most likely type of the current fault is quickly inferred through methods such as weighted voting or confidence assessment, thereby identifying whether it is a short circuit, open circuit, or ground fault.

[0172] Rapid comparison of edge diagnostic knowledge units Similarity to 10 historical fault modes in the knowledge unit. The correlation coefficient with Mode 3, representing "single-phase high-impedance grounding," is 0.92, while the correlation coefficient with Mode 1, representing "phase-to-phase metallic short circuit," is only 0.35. Based on the high similarity of 0.92, it can be quickly deduced that the most likely type of the current fault is a high-impedance grounding fault.

[0173] S63. Obtain real-time data on the capacity and distribution of optical storage devices, and pre-calculate the reliability indicators for power restoration based on this data;

[0174] The second task is a pre-assessment of power restoration potential, which is independent of fault identification and conducted simultaneously. This involves real-time querying and obtaining the latest status of all photovoltaic and energy storage devices within the distribution area, specifically including the rated capacity, current available capacity or state of charge (SOC) of each device, and its geographical or electrical topology location within the distribution network. Based on this data, a reliability index for power restoration is pre-calculated. This index aims to quantify the ability to utilize existing photovoltaic and energy storage resources to form isolated power systems and supply power to critical loads. For example, this reliability index is defined as:

[0175] ;

[0176] Where R is the pre-evaluation reliability index, which is a dimensionless value; The total electrical energy that can be provided by all available photovoltaic and energy storage devices in the region is obtained by summing up the real-time available capacity of each device. It is a pre-set critical load power value that needs to be prioritized for restoration; This refers to the target duration of power restoration. This indicator directly reflects the margin of existing energy storage resources in meeting the continuous power supply needs of critical loads.

[0177] The SOC status of five energy storage devices within the fault area can be queried in real time, and the total available electrical energy can be obtained by summarizing them. The value is 45 kWh. Assume the critical load power that needs to be restored first is... 10kW, target duration The estimated time is 4 hours. The reliability index for power restoration is pre-calculated. for The index of 1.125 indicates that the existing photovoltaic and energy storage resources can meet the continuous power supply needs of critical loads for 4 hours, with a margin of 12.5%.

[0178] S64. Integrate fault types and reliability indicators to generate diagnostic and evaluation results.

[0179] The fault types identified by the first task and the reliability indicators calculated by the second task are integrated, packaged into a structured diagnosis and evaluation result, and immediately reported to the upper-level master station.

[0180] The identified fault type "high impedance grounding fault" and the calculated reliability index " The data, along with the timestamp of the fault and key characteristic values, is encapsulated into a structured JSON message packet. This message packet is immediately encrypted and used to generate diagnostic and assessment results, ready to be reported to the distribution network master station via a secure channel.

[0181] S6, after obtaining the diagnostic and assessment results, also includes:

[0182] S65. Obtain the distribution network topology information that depicts the connection relationships within the transformer area;

[0183] Before performing diagnostic tasks, it is necessary to preload or acquire in real time the distribution network topology information depicting the interconnections within the transformer substation. This information acts like a "digital map" of the power grid, precisely defining the electrical connections and hierarchical structure between transformers, feeders, switches, photovoltaic and energy storage devices, and various user loads.

[0184] The energy converter synchronously obtains the latest topology information of the No. 10 transformer area from the geographic information system (GIS) of the distribution network master station. This information is stored in the form of a graph database, which accurately depicts the physical and electrical connections from the low-voltage side of the transformer to each user's electricity meter. This includes the connection of switch K1 to 5 residential buildings and 1 energy storage unit, and the connection of branch line No. 5 to Building A hospital and 2 photovoltaic households.

[0185] S66. Based on the power distribution network topology information and edge fault diagnosis knowledge unit, perform downstream tracing analysis to assess the scope of fault impact.

[0186] Once a fault occurs, downstream tracing analysis is performed based on the initially identified fault characteristics and the locations of associated equipment, combined with the distribution network topology information. This analysis algorithm starts from the fault point or the upstream equipment node closest to the fault point and traverses all downstream branches along the electrical path in the topology graph, marking all load nodes along this path as affected. Through this tracing, a precise assessment and output of the fault's impact range can be achieved, resulting in a list of all users or equipment that lost power due to the fault.

[0187] Assume the fault point diagnosed by the knowledge unit is located downstream of switch K1. Using topology information, a depth-first search (DFS) traversal is performed starting from node K1, tracing all downstream connections. The analysis results quickly locate and mark all load nodes that lost power due to the tripping of K1. The assessed impact range of the fault is: 5 residential buildings, Hospital Building A, and 1 energy storage unit, and a list is generated. .

[0188] S67. Obtain critical load level information, and calculate the recovery scheduling priority based on critical load level information, fault impact range and real-time optical storage capacity data;

[0189] Obtain a pre-configured list of critical load levels, categorizing loads within the distribution area according to their importance (e.g., hospitals, schools, traffic lights are high-level). Associate this critical load level information with the newly assessed fault impact area and, combined with real-time photovoltaic (PV) and energy storage capacity data, calculate the restoration scheduling priority. Select all high-level critical loads within the fault impact area and, based on the distribution and capacity of available PV and energy storage resources, generate an optimal, step-by-step power restoration sequence. This sequence, the restoration scheduling priority, clarifies which critical loads should be restored power first when resources are limited. Figure 3 As shown, the bubble chart illustrates the recovery scheduling priorities between loads of different importance levels and available energy storage units within the fault's impact area. The size of the circles is proportional to the calculated recovery scheduling priorities, guiding the emergency recovery decisions of the upper-level master station.

[0190] Obtaining critical load level information reveals that Hospital Building A is at the highest level, Level I, while the residential building is at Level III. Based on the fault impact range list, ESS-1, with a real-time available capacity of 20 kWh, is located on an adjacent node to the hospital. Calculating the recovery priority: Due to limited resources, ESS-1 should be isolated and switched to islanded mode first, prioritizing power supply to the Level I load, Hospital Building A. Therefore, the recovery scheduling priority sequence is calculated as follows: .

[0191] S68. Integrate the scope of the fault impact and the priority of recovery scheduling into the diagnosis and assessment results.

[0192] The assessed scope of the fault impact and the calculated recovery scheduling priority will be integrated with the fault type and reliability indicators into the final diagnosis and assessment results, forming a more comprehensive situation report to be submitted to the upper-level master station.

[0193] List of the scope of impact of the assessed failure and the calculated recovery scheduling priority sequence This information is integrated with fault type and reliability metrics. The final diagnosis and assessment result is a comprehensive situation report, including the nature of the fault, its scope of impact, and actionable emergency recovery steps, which is immediately reported to the upper-level master station for dispatchers to make decisions as soon as possible.

[0194] Example 2: As Figure 4 As shown, the low-voltage distribution network fault diagnosis system based on trusted access of photovoltaic and energy storage devices is used to implement the method in Example 1, including:

[0195] The trusted protocol parsing module is used to perform deep application layer protocol parsing on data packets obtained from optical storage devices in a trusted computing environment built by a security chip, and obtain parsed data.

[0196] The Trusted Data Standardization Module is used to standardize and transform the parsed data, and use a trusted computing environment to add a data imprint containing timestamps and integrity check codes to it, thus obtaining trusted standardized data.

[0197] The perception dimension optimization module is used to obtain target variables for distinguishing different types of distribution network faults, and dynamically selects features from trusted standardized data based on the target variables using the maximum correlation minimum redundancy algorithm to obtain the minimum feature subset;

[0198] The measurement data acquisition and transmission module is used to acquire and encrypt the measurement data corresponding to the smallest feature subset to obtain efficient sensing data;

[0199] The edge knowledge unit construction module is used to acquire historical fault case data and integrate the historical fault case data with efficient perception data to construct edge fault diagnosis knowledge units;

[0200] The edge fault diagnosis and assessment module is used to perform rapid fault type identification and power restoration potential pre-assessment in parallel based on the edge fault diagnosis knowledge unit when a fault occurs, obtain diagnosis and assessment results, and report the diagnosis and assessment results to the upper-level master station.

[0201] The implementation details of each module are the same as in Example 1.

Claims

1. A method for fault diagnosis of low-voltage distribution networks based on trusted access of photovoltaic and energy storage equipment, characterized by the following steps: include: S1. In a trusted computing environment built with a security chip, the data packets obtained from the optical storage device are subjected to deep application layer protocol parsing to obtain the parsed data; S2. Standardize the parsed data and add a data imprint containing timestamps and integrity check codes to it using a trusted computing environment to obtain trusted standardized data. S3. Obtain the target variable used to distinguish different types of distribution network faults, and dynamically select features from the reliable standardized data based on the target variable using the maximum correlation minimum redundancy algorithm to obtain the minimum feature subset; S4. Collect and encrypt the measurement data corresponding to the smallest feature subset to obtain efficient sensing data; S5. Acquire historical fault case data and integrate it with high-efficiency sensing data to construct edge fault diagnosis knowledge units, including: Use real-time feature values ​​from the data to efficiently detect query indexes; Pattern matching and correlation analysis are performed on historical fault case data to filter out historical fault cases that are similar to the current power grid status. The selected historical fault cases are integrated with efficient perception data into a composite data structure to generate edge fault diagnosis knowledge units. S6. When a fault occurs, the fault type is quickly identified and the power supply recovery potential is pre-assessed in parallel based on the edge fault diagnosis knowledge unit to obtain the diagnosis and assessment results, and the diagnosis and assessment results are reported to the upper-level master station.

2. The method for fault diagnosis of low-voltage distribution network based on trusted access of photovoltaic storage equipment according to claim 1, characterized in that, In S1, the parsed data includes: Data packets from the optical storage device are acquired in parallel via power line communication channels and wireless communication channels. Obtain a protocol feature library for identifying the communication mechanisms and data formats of application layer protocols; In a trusted computing environment, parsed data is generated by matching the application layer content of data packets with the prefixes and suffixes of a protocol signature library.

3. The method for fault diagnosis of low-voltage distribution network based on trusted access of photovoltaic and energy storage equipment according to claim 1, characterized in that, In S2, the obtained reliable standardized data includes: Obtain a mapping rule library for unifying data fields, and perform format unification processing on the parsed data based on the mapping rule library to generate a standardized data stream; In a trusted computing environment, a hardware cryptographic engine is used to compute integrity verification codes for standardized data streams. The time stamp and integrity check code are encapsulated into a standardized data stream to generate trusted standardized data.

4. The method for fault diagnosis of low-voltage distribution network based on trusted access of photovoltaic and energy storage equipment according to claim 1, characterized in that, In S3, the minimum feature subset includes: Calculate the mutual information between each feature and the target variable in the credible standardized data to obtain the feature correlation index; Calculate the mutual information between each feature to obtain the feature redundancy index; Based on feature correlation and feature redundancy indices, features are iteratively selected to construct a minimum feature subset.

5. The method for fault diagnosis of low-voltage distribution network based on trusted access of photovoltaic storage equipment according to claim 1, characterized in that, In S3, after obtaining the minimum feature subset, the process also includes updating the minimum feature subset, specifically: Obtain operating parameters that characterize the current distribution network status; Based on the operating parameters, the target variable is adjusted through decision logic to obtain a new target variable; Based on the new target variable, the feature correlation index and feature redundancy index are recalculated, and the minimum feature subset is updated.

6. The method for fault diagnosis of low-voltage distribution network based on trusted access of photovoltaic storage equipment according to claim 1, characterized in that, In S4, the efficient sensing data obtained includes: Extract the corresponding voltage, current, and power measurement data from the smallest feature subset; Obtain the symmetric encryption key used to encrypt measurement data; The measurement data is encrypted using a symmetric encryption key to generate encrypted measurement data. The encrypted measurement data is transmitted to the edge processing to obtain efficient sensing data.

7. The method for fault diagnosis of low-voltage distribution network based on trusted access of photovoltaic storage equipment according to claim 1, characterized in that, In S6, the diagnostic and assessment results include: Use the highly efficient sensing data captured at the moment of the current fault as a real-time query mode; In the edge fault diagnosis knowledge unit, the type of fault, such as short circuit, open circuit or ground fault, is identified by comparing the similarity between the real-time query mode and the historical fault mode. Acquire real-time data on the capacity and distribution of optical storage devices, and pre-calculate reliability indicators for power restoration based on this data; By integrating fault types and reliability metrics, diagnostic and assessment results are generated.

8. The method for fault diagnosis of low-voltage distribution network based on trusted access of photovoltaic storage equipment according to claim 1, characterized in that, S6, after obtaining the diagnostic and assessment results, also includes: Obtain the distribution network topology information that depicts the interconnections within the transformer substation area; Downstream tracing analysis is performed based on distribution network topology information and edge fault diagnosis knowledge units to assess the scope of fault impact. Obtain critical load level information, and calculate the recovery scheduling priority based on critical load level information, fault impact range and real-time optical storage capacity data; The scope of the fault impact and the priority of recovery scheduling are integrated into the diagnosis and assessment results.

Images