A cloud-edge collaborative driven physical information fusion power distribution network fault location system

Abstract

The application relates to the technical field of power distribution network fault positioning, and particularly discloses a cloud-edge collaborative driving physical information fusion power distribution network fault positioning system. In view of the demand for rapid and accurate positioning of power distribution network faults, a transient waveform acquisition module is adopted to capture transient data with a sampling frequency at the MHz level, a physical feature extraction module is used to generate a lightweight physical feature vector to reduce data transmission volume; an edge evidence anchoring module is used to ensure the authenticity and integrity of data sources, a cloud evidence checking module is used for data checking, a space-time physical consistency verification module is used for cross verification of multi-source data, a fault trust topology network is constructed, and finally a fault trust topology positioning module is used to output a high-accuracy fault section positioning result. The application effectively solves the problems of large bandwidth pressure and easy data tampering in centralized processing, improves the timeliness and reliability of fault positioning, and guarantees the stable operation of the power distribution network.

Description

Technical Field

[0001] This invention relates to the field of power distribution network fault location technology, and more specifically, to a cloud-edge collaborative physical information fusion power distribution network fault location system. Background Technology

[0002] Currently, the stable operation of the power distribution network is crucial for socio-economic activities, and quickly and accurately locating and eliminating line faults is one of the core challenges in ensuring power supply reliability. With the expansion of the power grid and the increasing complexity of its topology, especially with the integration of new energy power generation and distributed power sources, fault types are becoming more diversified and transient processes more complex. This places higher demands on traditional fault location technologies, requiring the rapid inference of the specific section where the fault occurred based on data from multiple monitoring points in the network within a very short time after the fault occurs.

[0003] To address this challenge, the industry has developed various fault location technologies. One type is the impedance method based on steady-state power frequency quantities, which calculates the electrical distance from the fault point to the measurement point by measuring changes in voltage and current before and after the fault. Another type is based on transient fault quantities, particularly the traveling wave method, which captures the transient traveling wave signal generated instantaneously by the fault and propagating along the line, and uses the time difference of its arrival at different monitoring points to calculate the fault location. The traveling wave method theoretically has high location accuracy because it is unaffected by fault resistance, changes in system operating mode, and transition resistance. Currently, solutions based on the traveling wave method typically require deploying data acquisition devices at multiple nodes and transmitting the acquired waveform data to a central station for centralized analysis.

[0004] However, existing technical solutions have some shortcomings in practical applications. For centralized traveling wave positioning systems, it is necessary to upload raw waveform data with sampling frequencies at the MHz level from multiple acquisition points across the entire network to the main station in real time during fault conditions. This generates a huge amount of communication data instantaneously, posing a severe challenge to the bandwidth and stability of the distribution network communication system. In environments with limited or unstable communication resources, data transmission delays or losses may occur, affecting the timeliness and accuracy of positioning. In addition, the distributed data acquisition environment makes it difficult to effectively guarantee the authenticity and integrity of each data source. During the transmission of data from edge acquisition points to the cloud central platform, there is a risk of unauthorized tampering or damage due to transmission errors. Traditional systems lack an effective mechanism to verify the credibility and integrity of the reported data source, which introduces uncertainty into subsequent analysis and decision-making. Summary of the Invention

[0005] In view of this, in order to solve the problems mentioned in the background technology, a cloud-edge collaborative physical information fusion power distribution network fault location system is proposed.

[0006] The objective of this invention can be achieved through the following technical solution: This invention provides a cloud-edge collaborative physical information fusion distribution network fault location system, comprising: a transient waveform acquisition module, which monitors the changes in electrical quantities at the edge nodes of the distribution network, acquires transient data with a sampling frequency at the MHz level at the trigger time, and uses a satellite-timed clock source for synchronization marking to generate a time-stamped original transient waveform sequence.

[0007] The physical feature extraction module extracts singularity and energy features from the time-stamped original transient waveform sequence, generating a physical feature vector containing waveform fingerprints.

[0008] The edge evidence anchoring module uses the device's private key to digitally sign and encapsulate the physical feature vector and the device's unique identification ID, generating an edge verifiable evidence chain data packet and uploading it.

[0009] The cloud-based evidence verification module receives edge-verifiable evidence chain data packets from the cloud management platform, uses the corresponding device public key to verify the authenticity and integrity of the source, and generates a valid evidence set.

[0010] The spatiotemporal physical consistency verification module, based on the global topology model of the distribution network, performs spatiotemporal physical consistency verification on the data of adjacent nodes within the valid evidence set, and generates a trust link marker characterizing the health of the line segment.

[0011] The fault trust topology localization module collects trust link markers to construct a fault trust topology network, identifies logical breakpoints or convergence points, and outputs fault segment localization results.

[0012] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) By performing on-site feature extraction of transient waveform sequences at the edge nodes of the power distribution network, the present invention can transform the original waveform data with a sampling frequency of MHz containing a large number of sampling points into a lightweight physical feature vector composed of a few key parameters such as the arrival time, polarity, steepness, energy integral value, and energy distribution entropy through wavelet transform and other algorithms. This processing method of information condensation at the source of data generation replaces the mode of uploading all the original waveform data to the cloud, and the amount of data reported is only a very small part of the original data. This mechanism reduces the instantaneous impact on the bandwidth of the cloud-edge communication link when a fault occurs, so that the system can operate stably under the existing power distribution communication network conditions, and ensure that key fault information can be effectively reported.

[0013] (2) This invention utilizes asymmetric encryption technology within the edge computing terminal to anchor the physical feature vector, generating an immutable cryptographic credential for each reported fault evidence data. Inside the edge terminal, its unique device private key, embedded in secure hardware, digitally signs the hash value containing the device identity and feature vector. Upon receiving the data, the cloud platform verifies the signature using a pre-stored device public key to confirm the authenticity of the data source and the integrity of the content. This mechanism establishes a trusted end-to-end link from data generation to data analysis, effectively preventing malicious tampering or accidental damage to data during transmission and storage, thereby eliminating invalid or forged fault information and providing a cryptographically verified reliable data foundation for subsequent fault analysis.

[0014] (3) This invention calls the global topology model of the distribution network on the cloud platform to perform spatiotemporal physical consistency verification on a set of valid evidence from multiple nodes that has passed cryptographic verification. This allows isolated multi-point data to be jointly verified under global physical constraints. The mechanism verifies whether the traveling wave has completely traversed the line segment by calculating the theoretical propagation delay of the traveling wave of adjacent nodes and comparing it with the actual measured time difference. If the time difference matches and the polarity conforms to the traversal characteristics, it indicates that the segment is a healthy traveling wave transmission channel, thus generating a trust link. This cross-verification can effectively eliminate isolated pseudo-feature data caused by misjudgment of a single measurement point, electromagnetic interference, or complex waveform reflection and refraction phenomena. By constructing a fault trust topology network composed of healthy segments, the area where the trust link is interrupted is inferred to be the fault segment, realizing in-depth screening and confirmation of fault information, thereby improving the accuracy and reliability of the final fault segment location result. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the system module structure connection of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 The present invention provides a cloud-edge collaborative physical information fusion power distribution network fault location system, including: a transient waveform acquisition module, a physical feature extraction module, an edge evidence anchoring module, a cloud evidence verification module, a spatiotemporal physical consistency verification module, and a fault trust topology location module.

[0019] The transient waveform acquisition module is connected to the physical feature extraction module, the physical feature extraction module is connected to the edge evidence anchoring module, the edge evidence anchoring module is connected to the cloud evidence verification module, the cloud evidence verification module is connected to the spatiotemporal physical consistency verification module, and the spatiotemporal physical consistency verification module is connected to the fault trust topology location module.

[0020] The transient waveform acquisition module monitors changes in electrical quantities at the edge nodes of the distribution network. At the trigger moment, it acquires transient data with a sampling frequency in the MHz range and uses a satellite-timed clock source for synchronization marking to generate a time-stamped original transient waveform sequence.

[0021] In a specific embodiment of the present invention, the specific process of generating a time-stamped original transient waveform sequence includes: continuously monitoring the instantaneous rate of change of voltage and current signals.

[0022] When the instantaneous rate of change exceeds the preset trigger threshold, the preset time window centered on the trigger moment is locked.

[0023] Collect continuous transient voltage and transient current data with a sampling frequency in the MHz range within a preset time window.

[0024] Using a satellite-synchronized clock source to generate nanosecond-level timestamps, the collected transient voltage and transient current data are marked point by point to generate a time-stamped original transient waveform sequence.

[0025] Specifically, the execution entity is an edge computing terminal deployed at the feeder switches or transformer nodes of the distribution network. This edge computing terminal integrates a digital signal processor (DSP) synchronized with a satellite-timed clock source and a high-speed analog-to-digital converter (ADC). Under normal operating conditions, the DSP controls the high-speed ADC to continuously sample the analog voltage and current signals of the power grid lines at a preset sampling frequency, storing the digitized sampled values ​​in a first-in-first-out (FIFO) circular data buffer. Simultaneously, the DSP calculates the rate of change of voltage or current amplitude between two adjacent sampling points in real time, i.e., the instantaneous rate of change. The system continuously compares the absolute value of this calculated instantaneous rate of change with a preset trigger threshold. When the instantaneous rate of change of voltage or current in any phase exceeds the preset trigger threshold, the system determines that a fault transient event has occurred and defines this moment as the trigger moment. Once triggered, the DSP immediately locks onto a preset time window centered on the trigger moment. This preset time window includes historical data before the trigger moment and data to be collected after the trigger moment. The processor extracts historical waveform data from the pre-trigger window from the circular data buffer and continues to collect data until the post-trigger window is full, thus obtaining a continuous segment of transient voltage and current data with sampling frequencies at the MHz level. Subsequently, the system invokes a satellite-synchronized clock source to precisely label each sampling point in the MHz-level transient voltage and current data with a global synchronization timestamp with nanosecond resolution. Finally, the edge computing terminal integrates the voltage and current waveform data with synchronization timestamps into a structured dataset, generating a time-stamped raw transient waveform sequence for subsequent processing steps.

[0026]

[0027]

[0028] In the above formula, and They represent the first time in the second month. The and the first Voltage sample values ​​acquired at each sampling time. and They represent the first time in the second month. The and the first The current sample value obtained at each sampling time. The sampling period is denoted by , and its reciprocal is the sampling frequency. and They represent the first time. The instantaneous rate of change of voltage and current calculated at each sampling time. The trigger condition is... or ,in and These are the preset voltage and current trigger thresholds, respectively. The above formulas are checked for dimensions; taking the formula for the instantaneous rate of change of voltage as an example... Its dimension is volt (V). The dimension of is seconds (s), therefore The unit of measurement is volts per second (V / s). The corresponding trigger threshold... The dimension of is also volts per second (V / s), and the two have the same dimension, which is consistent with physical logic.

[0029] The edge computing terminal is an embedded device integrating data acquisition, processing, and communication functions. It possesses at least one processor with a clock speed of at least 1 GHz, at least 512 MB of RAM, and an interface supporting time synchronization via BeiDou BDS or GPS. The satellite-synchronized clock source refers to a local clock that receives satellite second pulse signals through the aforementioned interface and is aligned with Coordinated Universal Time (UTC), with a synchronization accuracy better than 100 ns. The sampling frequency for transient voltage and current data is set between 1 MHz and 10 MHz to ensure the capture of the steep rising edge of the fault traveling wavefront. The trigger threshold is set based on statistical analysis of historical monitoring data from at least 30 consecutive days under normal operating conditions for a specific distribution network. It is typically set to 3 to 5 times the standard deviation of the instantaneous rate of change of voltage or current under normal operating conditions. This aims to ensure sensitivity to real faults while avoiding false triggers caused by routine operations or load fluctuations. The total duration of the preset time window is typically set to 20 ms to 50 ms. For example, it can be set to include data from 5 ms to 10 ms before the trigger and from 15 ms to 40 ms after the trigger, to completely record the entire process from the pre-fault steady state to the fault transient state and then to the post-fault initial steady state. The time-stamped raw transient waveform sequence is a data structure in the format of a time series array containing multiple elements. Each element is a data tuple, which contains at least a nanosecond-level timestamp, a voltage sample value, and a current sample value.

[0030] For example, suppose an edge computing terminal is deployed on a 10 kV distribution network feeder, and its sampling frequency is set to 2 MHz, i.e., the sampling period is... The trigger threshold for the instantaneous rate of change of voltage is 0.5 μs. Based on historical data analysis, the setting was set to 1.5 kV / μs. During normal operation, the system continuously samples, and if the collected voltage sample... The voltage was 10.05 kV. At the next sampling time, due to a ground fault at the remote end, the voltage sample collected... The voltage suddenly drops to 9.20 kV. At this point, the system calculates the instantaneous rate of change. =1.7 kV / μs. Since the calculated 1.7 kV / μs is greater than the preset trigger threshold of 1.5 kV / μs, the system immediately triggers fault recording. Assuming the preset time window is from 5 ms before triggering to 20 ms after triggering, the system will lock the data within this time range. Finally, the system generates a time-stamped raw transient waveform sequence containing data tuples, where each tuple is in the format {timestamp, voltage value, current value}.

[0031] The physical feature extraction module extracts singularity and energy features from the time-stamped original transient waveform sequence, generating a physical feature vector containing waveform fingerprints.

[0032] In a specific embodiment of the present invention, the process of extracting singularity and energy features from the time-stamped original transient waveform sequence to generate a physical feature vector containing waveform fingerprints includes: using a wavelet transform algorithm to perform multi-scale decomposition on the time-stamped original transient waveform sequence.

[0033] Search for the modulus maxima in the decomposed sequence of detail coefficients.

[0034] Identify the arrival time, polarity, and steepness characteristics of transient traveling wavefronts.

[0035] In a specific embodiment of the present invention, generating a physical feature vector containing waveform fingerprints further includes the following steps: extracting a waveform segment of a preset length after the fault triggering time.

[0036] The transient energy integral value within the waveform segment is calculated, and the energy distribution entropy is calculated based on the energy ratio of the sub-time window, thus forming the physical fingerprint characteristics of the fault waveform.

[0037] Arrival time, wavefront polarity, wavefront steepness characteristics, transient energy integral value, and energy distribution entropy are concatenated in a preset order to generate a physical feature vector.

[0038] Specifically, the execution entity remains the digital signal processor (DSP) within the edge computing terminal. This DSP first invokes a pre-defined wavelet transform algorithm module to perform multi-scale decomposition on the voltage or current data in the time-stamped original transient waveform sequence. This decomposition process projects the original transient waveform sequence onto a series of wavelet basis functions at different scales, generating multiple levels of detail coefficient sequences and approximate coefficient sequences. The processor focuses on analyzing the first or second level detail coefficient sequences corresponding to the high-frequency band. By searching for modulus maxima points on these detail coefficient sequences, singularity abrupt changes in the signal can be identified; these abrupt changes correspond to the wavefront of the transient traveling wave. The system records the timestamp corresponding to the first modulus maxima point exceeding a preset singularity detection threshold as the wavefront arrival time. Simultaneously, the system reads the sign of the sampled values ​​of the original transient waveform sequence at the wavefront arrival time, with positive signs indicating positive polarity and negative signs indicating negative polarity, thereby determining the wavefront polarity. The amplitude of the modulus maxima point or the maximum value within a nearby small window is used as the wavefront steepness feature, characterizing the steepness of the wavefront. Next, the processor extracts a waveform segment of a preset length from the time-stamped original transient waveform sequence, starting from the fault trigger moment. The processor calculates the sum of the squares of the amplitudes of all sampled values ​​within this segment to obtain the transient energy integral value. To further extract waveform morphology information, the processor evenly divides this waveform segment into multiple sub-time windows, calculates the energy within each sub-time window, and constructs a probability distribution based on the proportion of energy in each sub-time window to the total energy. Then, it calculates the Shannon entropy of this probability distribution to obtain the energy distribution entropy. This energy distribution entropy, together with the transient energy integral value, constitutes the physical fingerprint feature of the fault waveform. Finally, the processor combines these five feature values ​​into a one-dimensional array according to a preset order, such as arrival time, wavefront polarity, wavefront steepness, transient energy integral value, and energy distribution entropy, to generate the physical feature vector of the edge nodes.

[0039]

[0040]

[0041] In the above formula, The first segment of a waveform of a preset length Each sample value, This represents the total number of sampling points within the segment. The sampling period. This is the transient energy integral value. It is used in calculating the energy distribution entropy. At that time, the waveform segment was divided into A time window, It is the first Energy integral value of each time window It is the first The proportion of energy in each time window to the total energy. Perform a dimensional check on the above formula, taking a voltage signal as an example. The dimension of is V. If the dimension of is s, then The dimension of this is V²·s, which conforms to the dimension of energy per unit resistance in physics. For , It is the ratio of energy to energy, a dimensionless quantity, therefore its logarithm is also dimensionless, and the final result is entropy. It is a dimensionless quantity, which is consistent with the definition of information entropy.

[0042] Among these, the wavelet transform algorithm typically uses the Daubechies wavelet, such as the db4 wavelet, which has good performance in detecting singularities in transient signals. The number of multi-scale decomposition levels is set according to the sampling frequency and the range of fault frequencies of interest, generally ranging from 4 to 8 levels. The singularity detection threshold is set based on 1.5 to 2.0 times the maximum detail coefficient value after performing the same wavelet transform analysis on the background noise data collected under normal operating conditions, to ensure the reliability of the detection. Wavefront polarity is represented by a numerical value, for example, +1 represents positive polarity and -1 represents negative polarity. The preset waveform segment duration is set to 1 ms to 5 ms, which is sufficient to cover the initial high-frequency oscillation process after the fault. The number of sub-time windows is also considered in the energy distribution entropy calculation. The value is typically set to 10 to 20 to strike a balance between computational complexity and feature representation capability. A physical feature vector is a structured data set containing five numerical elements, and its data type can be uniformly set to 64-bit floating-point numbers to ensure precision.

[0043] For example, the processor in the edge computing terminal receives a time-stamped raw transient waveform sequence. First, the processor uses the db4 wavelet to perform a 5-level decomposition of the voltage data of the raw transient waveform sequence. During the analysis of the level 1 detail coefficients, a modulus maxima far exceeding the singularity detection threshold is found at the sampling point corresponding to the trigger time. Therefore, the system determines the wavefront arrival time as this timestamp, which can be represented as a nanosecond value of 15123450500 for ease of calculation. Since the voltage value drops from 10.05 kV to 9.20 kV at this time, a negative change, the wavefront polarity is recorded as -1. The amplitude of this modulus maxima point is 2.1, and this value is recorded as the wavefront steepness feature. Next, the system extracts a 2 ms waveform segment from the trigger time, containing 4000 sampling points. By summing the squares of these 4000 voltage sampling values ​​and multiplying by the sampling period of 0.5 μs, the transient energy integral value of 0.021 V²·s is calculated. Subsequently, the 2 ms segment was divided into 10 sub-windows of 0.2 ms each, and the energy percentage of each sub-window was calculated. The probabilities are {0.45, 0.25, 0.12, 0.08, 0.04, 0.02, 0.01, 0.01, 0.01, 0.01}. Based on this probability distribution, the energy distribution entropy is calculated. Finally, the system concatenates these five feature values ​​in sequence to generate the physical feature vector of the edge node as [15123450500,-1,2.1,0.021,2.08].

[0044] The edge evidence anchoring module uses the device's private key to digitally sign and encapsulate the physical feature vector and the device's unique identification ID, generating an edge verifiable evidence chain data packet and uploading it.

[0045] In a specific embodiment of the present invention, the specific process of generating an edge verifiable evidence chain data packet includes: obtaining the device unique identifier ID preset by the edge computing terminal and combining it with the physical feature vector.

[0046] The integrity hash value of the combined data is calculated using a one-way hash function.

[0047] The device's private key stored in the edge computing terminal is used to perform a digital signature operation on the integrity hash value, generating a digital signature string.

[0048] The device's unique identifier (ID), physical feature vector, and digital signature string are encapsulated into an edge-verifiable evidence chain data package.

[0049] Specifically, the execution is performed by the main processor within the edge computing terminal or its integrated security coprocessor. First, the processor obtains the device's unique identifier (ID) from a region embedded in read-only memory or a secure element during device manufacturing or deployment. Next, the processor combines the device's unique identifier ID with the physical feature vector of the edge node generated in the previous step. This combination process is accomplished through a deterministic serialization operation, such as converting the string representation of the device's unique identifier ID and each floating-point number in the physical feature vector into a byte stream in a preset order and concatenating them to form the original data byte string. Subsequently, the processor invokes an onboard cryptographic function library, taking the original data byte string as input, and performs calculations using a one-way hash function, specifically the secure hash algorithm SHA-256, to generate a fixed-length 256-bit integrity hash value. After this, the processor issues an instruction to its internally integrated Trusted Platform Module (TPM) or Hardware Security Module (HSM) to request a signature operation on the integrity hash value. The hardware security module uses its internally stored, externally unreadable device private key to perform a digital signature operation on the integrity hash value, such as using the Elliptic Curve Digital Signature Algorithm (ECDSA), to generate an unforgeable digital signature string. Finally, the processor encapsulates the device's unique identifier ID, the unmodified physical feature vector of the edge node, and the digital signature string into a standardized data structure, such as a JSON object, forming a verifiable chain of evidence data packet at the edge. This data packet is then uploaded to a pre-defined cloud management platform address via a secure communication protocol such as HTTPS or MQTTS.

[0050]

[0051]

[0052] The formula above describes the core process of evidence anchoring. In the first row of the formula, Represents the unique identifier ID of the device. The physical feature vector representing the edge node. This represents a serialization function that converts input data into a deterministic byte stream. This indicates a byte stream concatenation operation. This indicates the 256-bit version of the Secure Hash Algorithm, whose output is the integrity hash value. In the second line of the formula, Represents the device's private key. This represents a function that uses the private key to perform digital signature operations, and its input is the integrity hash value. The output is the digital signature string. .

[0053] The device's unique identifier (ID) is a globally unique identifier, such as a 128-bit Universally Unique Identifier (UUID), which is programmed into a one-time programmable memory at the factory to ensure its immutability. The one-way hash function SHA-256 is chosen based on its current security standard and widespread application in cryptography, guaranteeing data integrity and collision resistance. The device's private key is part of the asymmetric key pair, generated according to industry-recognized elliptic curve standards such as secp256k1 or P-256. The private key's lifecycle is entirely within the hardware security module; no external software can directly access it, ensuring absolute signature security. The digital signature string is typically encoded in DER format or further Base64 encoded for transmission in text formats such as JSON.

[0054] For example, the processor first obtains the device's unique identifier ID, assuming it's "a1b2c3d4-e5f6-4789-a0b1-c2d3e4f5g6h7". Next, it serializes and concatenates this ID string with the generated physical feature vector of the edge node [15123450500,-1,2.1,0.021,2.08], forming a byte stream to be processed. Then, the system performs a SHA-256 hash operation on this byte stream, obtaining a 32-byte integrity hash value, whose hexadecimal representation might be "1a8efc9c5d6f45b563a4d3e72d6f8b3d4f1a2b3c4d5e6f7a8b9c0d1e2f3a4b5c". Subsequently, the processor sends this hash value to its hardware security module, which uses the device's private key to sign it using the ECDSA algorithm. After the signing operation is complete, the hardware security module returns a digital signature string, which, after Base64 encoding, may be "MEQCIDn8GZ / Lw8xJ6n2Y…4W7FqP5Z6A==". Finally, the processor assembles these three pieces of information into a JSON-formatted edge-verifiable chain of evidence data packet. This JSON data packet is then sent to the cloud management platform via an encrypted HTTPS connection.

[0055] The cloud-based evidence verification module receives edge-verifiable evidence chain data packets from the cloud management platform, uses the corresponding device public key to verify the authenticity and integrity of the source, and generates a valid evidence set.

[0056] In a specific embodiment of the present invention, the specific process of using the corresponding device public key to verify the authenticity and integrity of the source and generate a valid evidence set includes: parsing the received edge verifiable evidence chain data packet and extracting the device's unique identity ID.

[0057] Retrieve the device public key corresponding to the device's unique identifier ID from the key management database.

[0058] The device's public key is used to decrypt and verify the digital signature string in the edge verifiable evidence chain data packet. The extracted device unique identifier ID is then combined with the physical feature vector in the edge verifiable evidence chain data packet. The hash value of the combined data is then recalculated and compared with the decryption result.

[0059] Data packets that fail verification are discarded, and data packets that pass verification are retained to form a valid set of evidence.

[0060] Specifically, the execution entity is a cloud management platform deployed on a cloud server cluster. This platform launches one or more public-facing listening services, such as a highly available API gateway or an MQTT broker cluster, to receive edge-verifiable evidence chain data packets uploaded via secure channels from various edge computing terminals in the power distribution network. To handle potential concurrent data surges during failures, the platform establishes a receive buffer behind the listening service. This receive buffer is implemented using a distributed message queue system, where all incoming data packets are published as messages to designated topics or queues. Next, the platform's background processing service launches a consumer group consisting of multiple parallel worker threads to asynchronously pull and consume data packets from the receive buffer. For each acquired data packet, the worker thread first parses it, extracting the device's unique identifier ID, the edge node's physical feature vector, and the digital signature string. Subsequently, the worker thread uses the device's unique identifier ID as the key to query a centralized key management database and retrieve the pre-registered device public key corresponding to that device. After obtaining the public key, the worker thread locally re-executes the serialization and hash calculation process, which is completely identical to that of the edge evidence anchoring module. This involves concatenating the device's unique identifier (ID) and physical feature vector in the data packet and calculating its SHA-256 hash value. Finally, the worker thread calls a cryptography library to decrypt and verify the digital signature string in the data packet using the obtained device public key, and rigorously compares the verification result with the locally recalculated hash value. If the verification is successful, indicating that the data packet's origin is authentic and its content has not been tampered with, the worker thread stores the core feature data from this data packet into a temporary in-memory database or table, forming a valid evidence set. Conversely, if the public key is missing, decryption verification fails, or the hash value mismatch occurs, the worker thread discards the data packet and records a security alert log, indicating the suspicious device ID and event time. Ultimately, only the valid evidence set that has passed all verifications is retained for further analysis.

[0061]

[0062] The above formula expresses the core logic of source authenticity verification. Among them, , and These represent the device ID, physical feature vector, and digital signature string, respectively, parsed from the received edge-verifiable evidence chain data packet. Representative according to The device public key retrieved from the key management database. and Functions and The operator definitions are completely consistent with those in the edge evidence anchoring module. It is a boolean function that uses the public key. This function verifies whether the signature is valid for a given data digest. The function returns true only if it does. Only when this condition is met is the data packet considered valid.

[0063] The cloud management platform is a distributed application built on a microservices architecture, deployed in an elastic and scalable cloud environment to ensure high availability and high throughput. The receive buffer preferably uses industrial-grade message queue middleware such as Apache Kafka, which can smooth out peak and off-peak traffic, decouple data reception and processing, and provide data persistence to prevent data loss. The key management database is a high-security database instance, such as AWS KMS or a self-built HashiCorp Vault, which stores the mapping between the unique device identifiers (IDs) of all legitimate edge devices and their public keys, and performs strict authentication and access control. Public key registration is typically completed during the device authentication and authorization phase when the edge device first connects to the network. The valid evidence set is a temporary storage space with a time-to-live (TTL), such as a Redis hash table or a partition in an in-memory data grid, used to aggregate all valid evidence belonging to the same failure event (typically within a second or millisecond time window), providing a data foundation for subsequent spatiotemporal correlation analysis.

[0064] For example, the MQTT broker of the cloud management platform receives a message published to the topic "fault / evidence," with a JSON data packet generated by the edge evidence anchoring module as its payload. A background worker thread consumes and parses this message, obtaining the device ID "a1b2c3d4-e5f6-4789-a0b1-c2d3e4f5g6h7." The worker thread uses this ID to query the key management database and successfully retrieves the public key of the device bound to it. Next, the worker thread serializes and concatenates the ID string and feature vector [15123450500,-1,2.1,0.021,2.08] according to the exact same rules as the edge evidence anchoring module, and then calculates its SHA-256 hash value, resulting in "1a8efc9c5d6f45b563a4d3e72d6f8b3d4f1a2b3c4d5e6f7a8b9c0d1e2f3a4b5c". Finally, the worker thread calls the ECDSA verification function, using the retrieved public key to verify the signature string "MEQCIDn8GZ / Lw8xJ6n2Y…4W7FqP5Z6A==", checking if it matches the hash value just calculated. The verification shows a perfect match, and the function returns true. Therefore, the data packet is confirmed as authentic and valid. The system stores this evidence, including its device ID and physical feature vector, into a valid evidence set named "event_20231026083015_cluster," waiting for evidence from other relevant nodes to arrive for the next step of correlation verification.

[0065] The spatiotemporal physical consistency verification module, based on the global topology model of the distribution network, performs spatiotemporal physical consistency verification on the data of adjacent nodes within the valid evidence set, and generates a trust link marker characterizing the health of the line segment.

[0066] In a specific embodiment of the present invention, the specific process of performing spatiotemporal physical consistency verification on adjacent node data within the valid evidence set to generate a trust link marker characterizing the health of a line segment includes: based on the global topology model of the distribution network, identifying one or more pairs of adjacent edge nodes in the valid evidence set in terms of electrical connection relationship.

[0067] Extract the arrival times of the traveling waves recorded by adjacent edge nodes and calculate the actual transmission time difference.

[0068] The theoretical transmission delay is calculated based on the line length and preset traveling wave velocity in the global topology model of the distribution network.

[0069] In a specific embodiment of the present invention, generating a trusted link marker characterizing the health of a line segment further includes the following steps: comparing the actual transmission time difference with the theoretical transmission delay to calculate the delay deviation.

[0070] Extract the wavefront polarity of adjacent edge nodes for their respective records.

[0071] If the time delay deviation is within the preset allowable range, it indicates that the traveling wave has completely traversed the line segment between the adjacent edge node pairs, and the wavefront polarity of the adjacent edge node pairs satisfies the same or opposite direction characteristics of the traversing wave. Then, the line segment is determined to be a healthy segment, and a trust link marker representing the health of the line segment is generated.

[0072] Specifically, the execution entity is the spatiotemporal-physical consistency verification engine within the cloud management platform. This engine first retrieves the set of valid evidence from the platform's in-memory database and simultaneously loads the global distribution network topology model. The global distribution network topology model stores the node information of all edge computing terminals and their electrical connections in the form of a graph data structure. Then, the verification engine traverses each piece of evidence in the set of valid evidence. For any piece of evidence, the engine searches for the node in the global distribution network topology model based on its unique device identifier (ID) and identifies all directly electrically connected adjacent nodes. Next, the engine checks whether the device IDs of these adjacent nodes exist in the current set of valid evidence. If they do, an adjacent edge node pair is formed. For each formed adjacent edge node pair, the engine extracts the arrival time of the traveling waves from their respective physical feature vectors. The engine calculates the absolute difference between the arrival times of these two traveling waves to obtain the actual transmission time difference. Simultaneously, the engine queries the actual length of the line between these nodes from the edge attributes of the global distribution network topology model and queries the traveling wave speed of the corresponding line type from a physical parameter library, calculating the theoretical transmission delay by dividing the two. The engine compares the actual transmission time difference with the theoretical transmission delay to determine if the delay deviation is within a preset allowable range. Furthermore, the engine extracts the recorded wavefront polarity from the physical feature vectors of the pair of nodes and verifies whether the wavefront polarity satisfies a logical interlocking relationship based on preset fault direction judgment logic. Only when both conditions—the transmission time difference deviation being within the allowable range and the wavefront polarity satisfying the logical interlocking relationship—are met simultaneously, does the engine determine that the spatiotemporal physical relationship between the pair of adjacent edge nodes is valid, and generates a trusted link tag containing the IDs of both nodes, storing it in a new result set.

[0073]

[0074]

[0075] The formula in the first row above is used to calculate the time delay deviation. .in, and These are the arrival times of the traveling waves of adjacent nodes 1 and 2 extracted from the set of valid evidence. It is the line length between node 1 and node 2 obtained from the global topology model of the distribution network. The traveling wave velocity is retrieved from the physical parameter database. The second line of the formula contains the logical conditions for generating a trusted link. . This is the upper limit of the preset allowable range. The physical meaning of this logical judgment condition is to verify whether the traveling wave has completely traversed the line between node 1 and node 2. If so, it means that the propagation time of the traveling wave between the two nodes is equal to the theoretical transmission time, that is, no fault has occurred in this section, and the traveling wave is unobstructed. and These are the wavefront polarities recorded by node 1 and node 2, respectively, represented by +1 or -1. This represents the polarity characteristic relationship of the crossing wave, which usually depends on the polarity installation direction of the current transformer. For example, when the polarity definitions of the current transformers at both ends are consistent, the polarity response of the crossing wave at the inflow and outflow ends should be consistent (or opposite, depending on the equipment definition). This is used here to assist in verifying the crossing properties of the traveling wave. Only when... Only when the result is true is the line segment determined to be a healthy and trustworthy segment, and a trust link tag containing the IDs of the two nodes is generated and stored in a new result set.

[0076] The global topology model of the distribution network is a digital network diagram. Its nodes represent locations where edge computing terminals are installed, such as substations, feeder switches, and distribution transformers. Edges represent overhead lines or cables connecting these nodes, and it stores static parameters such as line type, length, and impedance. (Traveling wave velocity) It is not a fixed value; it depends on the medium of the transmission line. For overhead lines, its value is close to the speed of light c, approximately 2.99 × 10⁻⁶. 8 m / s; for underground cables, due to their high dielectric constant, the value is typically 0.5 to 0.7 times the speed of light. The system will automatically select the appropriate wave velocity value based on the line type defined in the topology model. Preset allowable range. The settings comprehensively consider the minor errors in GPS / BDS clock synchronization, the sampling clock jitter of the analog-to-digital converter, and the calculation errors of the traveling wave identification algorithm, and are typically set between 50 ns and 200 ns. The logical interlocking relationship is based on physical principles: when a fault occurs outside the line segment (i.e., the traveling wave crosses the segment), the time difference between the two monitoring points should strictly conform to the transmission delay, and the waveform polarity should conform to the characteristics of the same traveling wave sequence. However, when a fault occurs inside the line segment, both ends detect the initial traveling wave from the fault point, and their time difference is usually much smaller than the theoretical transmission delay, failing to meet the aforementioned delay deviation condition, thus preventing the generation of a trusted link.

[0077] For example, the spatiotemporal physical consistency verification engine retrieves evidence uploaded by device "a1b2c3d4-e5f6-4789-a0b1-c2d3e4f5g6h7" from the valid evidence set "event_20231026083015_cluster". Assume that this set also contains another valid piece of evidence from device "b2c3d4e5-f6g7-5890-b1c2-d3e4f5g6h7i8", whose physical feature vector is [15123454550,1,1.8,0.018,2.25]. The engine loads the global topology model and finds that the two devices are directly adjacent in the topology, separated by a 1.2 km overhead line. The engine then begins verifying these two node pairs. First, it extracts the arrival times of the traveling waves at the two nodes, which are 15123450500 ns and 15123454550 ns respectively, and calculates the actual transmission time difference to be 4050 ns. Next, the engine obtains the traveling wave velocity of the overhead line from the physical parameter database as 2.99 × 10⁻⁶. 8 m / s, the calculated theoretical transmission delay is 4013.4 ns. Delay deviation It is 36.6 ns. Assuming a preset allowable range. The time consistency verification is passed because 36.6 ns ≤ 200 ns. Subsequently, the engine extracts the wavefront polarities of the two nodes, which are -1 and +1 respectively. Their product is -1, satisfying the logical interlocking relationship of less than 0, and the spatial consistency verification also passes. Since both conditions are met, the engine ultimately determines that the spatiotemporal physical relationship between the two nodes is valid and generates a trusted link marker, whose content can be represented as {"link":["a1b2c3d4-e5f6-4789-a0b1-c2d3e4f5g6h7","b2c3d4e5-f6g7-5890-b1c2-d3e4f5g6h7i8"],"status":"trusted"}, and stores it in the result set.

[0078] The fault trust topology localization module collects trust link markers to construct a fault trust topology network, identifies logical breakpoints or convergence points, and outputs fault segment localization results.

[0079] In a specific embodiment of the present invention, the specific process of collecting trust link tags to construct a fault trust topology network includes: collecting all generated trust link tags.

[0080] Map all trusted link tags onto the global topology model of the distribution network.

[0081] Preserve the nodes and edges covered by the trusted links to generate a fault trust topology subgraph composed of verified nodes.

[0082] In a specific embodiment of the present invention, the specific process of identifying logical breakpoints or convergence points and outputting fault segment location results includes: searching for logical breakpoints or convergence points in the fault trust topology subgraph.

[0083] A logical breakpoint refers to a node on an established trust link where the downstream trust link is interrupted.

[0084] A convergence point is a point where a trust link fails to be established between two adjacent nodes, and the fault directions recorded by each node are opposite to each other.

[0085] Identify the boundary nodes corresponding to the above logical breakpoints or convergence points.

[0086] The line segment between the boundary nodes is identified as the final fault location, and the fault segment location result is output.

[0087] Specifically, the execution entity is the fault location and decision-making module on the cloud management platform. This module first aggregates all trusted link markers generated in the previous step, which together form an edge set consisting of healthy line segments. Then, the module maps this edge set onto the original global distribution network topology model. By retaining all nodes and edges included in the trusted links and removing all unverified nodes and edges, a fault-trusted topology subgraph reflecting healthy areas in the network is constructed. Next, the location module executes a search algorithm on this subgraph to identify the boundaries of faulty segments, i.e., to find the points of interruption in the trusted link network. The core of this search algorithm is to find logical breakpoints or convergence points in the fault-trusted topology subgraph. A logical breakpoint refers to a node with an established trusted link where all its downstream links are interrupted (meaning the downstream segment failed the traversal verification and may be faulty), while links pointing in the opposite direction exist. A convergence point refers to two adjacent nodes that do not form a trust link between them (i.e., the traveling wave does not cross the segment between them, implying a fault within it), but each has formed a trust link with other parts of the network, and their recorded fault direction indicators are opposite (both pointing towards each other, further confirming the fault is in the middle). Once such a pair of boundary nodes is identified, the system locks the physical line segment between these two boundary nodes as the final fault location. Finally, the system encapsulates the location information of this fault segment, such as the starting and ending tower numbers or GPS coordinate range, along with a confidence assessment calculated based on the amount of evidence involved in the location and the magnitude of the delay deviation, into a standardized location report, and outputs it through API, push notifications, or a visual interface for use by operations and maintenance personnel.

[0088] The fault trust topology subgraph is a dynamic subset of the global distribution network topology model. It visually demonstrates the propagation path of transient traveling wave information in the network after a fault occurs, ensuring consistent verification. The search algorithm can be a graph traversal-based algorithm, such as Depth-First Search (DFS) or Breadth-First Search (BFS), combined with pruning based on the fault direction information of each node. The boundary node identification logic is as follows: if node A and node B are adjacent, and A determines the fault is downstream while B determines the fault is upstream, but a trust link cannot be formed between A and B (e.g., due to time difference mismatch), then A and B constitute a pair of boundary nodes, and the fault point is locked between them. Confidence assessment can be calculated using an empirical formula. For example, setting the base confidence to 95%, the confidence increases by 0.5% for each node found that participates in forming the fault trust topology subgraph. Simultaneously, the confidence is inversely proportional to the average time delay deviation of the final boundary node pair; the smaller the deviation, the higher the confidence. The location report is a structured XML or JSON file that includes a unique identifier for the faulty section, a description of the geographical location, the time of the fault, a list of evidence used for location, and a confidence score, facilitating integration with work order management systems or geographic information systems (GIS).

[0089] For example, the fault location and decision-making module receives a set of trust link tags containing {"link":["a1b2c3d4-e5f6-4789-a0b1-c2d3e4f5g6h7","b2c3d4e5-f6g7-5890-b1c2-d3e4f5g6h7i8"],"status":"trusted"}. Suppose that during the event, another trust link tag {"link":["c3d4e5f6-a1b2-…","a1b2c3d4-e5f6-…"],"status":"trusted"} was also generated, but a trust link was not formed between node "b2c3d4e5-f6g7-…" and its downstream node "d4e5f6g7-b2c3-…". The module plots these trust links on the global topology graph, forming a linear fault trust topology subgraph from an upstream node "c3…" to node "a1…" and then to node "b2…". At node "b2c3d4e5-f6g7-5890-b1c2-d3e4f5g6h7i8", the link further downstream is broken. Inspection reveals that node "b2…" records the fault direction upstream (wavefront polarity +1), while its downstream node "d4…" also records the fault direction upstream. This does not conform to normal propagation patterns, therefore a trust link could not be formed between them, and nodes "b2…" and "d4…" constitute a pair of boundary nodes. The system therefore identifies the line segment between these two boundary nodes as the final fault location. Finally, the system generated a location report, which stated: "Fault location result: The fault is located on the 10kV-F9 feeder, between tower 032 (node ​​b2...) and tower 033 (node ​​d4...). Confidence level: 98.5%." This report was immediately pushed to the monitoring screen of the distribution network emergency repair dispatch center.

[0090] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A cloud-edge collaborative physical information fusion power distribution network fault location system, characterized in that, include: The transient waveform acquisition module monitors changes in electrical quantities at the edge nodes of the distribution network. At the trigger moment, it acquires transient data with a sampling frequency in the MHz range and uses a satellite-timed clock source for synchronization marking to generate a time-stamped original transient waveform sequence. The physical feature extraction module extracts singularity and energy features from the time-stamped original transient waveform sequence, generating a physical feature vector containing waveform fingerprints. The edge evidence anchoring module uses the device's private key to digitally sign and encapsulate the physical feature vector and the device's unique identification ID, generating an edge verifiable evidence chain data package and uploading it. The cloud-based evidence verification module receives edge-verifiable evidence chain data packets from the cloud management platform, uses the corresponding device public key to verify the authenticity and integrity of the source, and generates a valid evidence set. The spatiotemporal physical consistency verification module, based on the global topology model of the distribution network, performs spatiotemporal physical consistency verification on the data of adjacent nodes within the valid evidence set, and generates a trust link marker characterizing the health of the line segment. The fault trust topology localization module collects trust link markers to construct a fault trust topology network, identifies logical breakpoints or convergence points, and outputs fault segment localization results.

2. The cloud-edge collaborative physical information fusion power distribution network fault location system according to claim 1, characterized in that, The specific process of generating the time-stamped original transient waveform sequence includes: Continuously monitor the instantaneous rate of change of voltage and current signals; When the instantaneous rate of change exceeds the preset trigger threshold, the preset time window centered on the trigger moment is locked. Collect continuous transient voltage and transient current data with a sampling frequency in the MHz range within a preset time window; Using a satellite-synchronized clock source to generate nanosecond-level timestamps, the collected transient voltage and transient current data are marked point by point to generate a time-stamped original transient waveform sequence.

3. The cloud-edge collaborative physical information fusion power distribution network fault location system according to claim 1, characterized in that, The specific process of extracting singularity and energy features from the time-stamped original transient waveform sequence to generate a physical feature vector containing waveform fingerprints includes: The wavelet transform algorithm is used to perform multi-scale decomposition on the original time-scaled transient waveform sequence; Search for the modulus maxima in the decomposed sequence of detail coefficients; Identify the arrival time, polarity, and steepness characteristics of transient traveling wavefronts.

4. The cloud-edge collaborative physical information fusion power distribution network fault location system according to claim 3, characterized in that, The generation of the physical feature vector containing the waveform fingerprint also includes the following steps: Extract a waveform segment of a preset length after the fault triggering time; Calculate the transient energy integral value within the waveform segment, and calculate the energy distribution entropy based on the energy proportion of the sub-time window to form the physical fingerprint characteristics of the fault waveform; Arrival time, wavefront polarity, wavefront steepness characteristics, transient energy integral value, and energy distribution entropy are concatenated in a preset order to generate a physical feature vector.

5. The cloud-edge collaborative physical information fusion power distribution network fault location system according to claim 1, characterized in that, The specific process for generating the edge-verifiable evidence chain data packet includes: Obtain the pre-set unique device identifier ID of the edge computing terminal and combine it with the physical feature vector; The integrity hash value of the combined data is calculated using a one-way hash function; The device's private key stored in the edge computing terminal is used to perform a digital signature operation on the integrity hash value, generating a digital signature string. The device's unique identifier (ID), physical feature vector, and digital signature string are encapsulated into an edge-verifiable chain of evidence data package.

6. The cloud-edge collaborative physical information fusion power distribution network fault location system according to claim 1, characterized in that, The specific process of using the corresponding device public key to verify the authenticity and integrity of the source and generate a valid evidence set includes: Parse the received edge verifiable evidence chain data packets to extract the device's unique identification ID; Retrieve the device public key corresponding to the device's unique identifier ID from the key management database; The device's public key is used to decrypt and verify the digital signature string in the edge verifiable evidence chain data packet. The extracted device unique identifier ID is then combined with the physical feature vector in the edge verifiable evidence chain data packet. The hash value of the combined data is then recalculated and compared with the decryption result. Data packets that fail verification are discarded, and data packets that pass verification are retained to form a valid set of evidence.

7. A cloud-edge collaborative physical information fusion power distribution network fault location system according to claim 1, characterized in that, The specific process of performing spatiotemporal-physical consistency verification on adjacent node data within the valid evidence set to generate a trusted link marker characterizing the health of the line segment includes: Based on the global topology model of the distribution network, identify one or more pairs of adjacent edge nodes in terms of electrical connection from the set of valid evidence; Extract the arrival times of the traveling waves recorded by adjacent edge nodes and calculate the actual transmission time difference; The theoretical transmission delay is calculated based on the line length and preset traveling wave velocity in the global topology model of the distribution network.

8. A cloud-edge collaborative physical information fusion power distribution network fault location system according to claim 7, characterized in that, The process of generating a trusted link tag that characterizes the health of a line segment further includes the following steps: The actual transmission time difference is compared with the theoretical transmission delay to calculate the delay deviation; Extract the wavefront polarity of adjacent edge nodes for their respective records; If the time delay deviation is within the preset allowable range, it indicates that the traveling wave has completely traversed the line segment between the adjacent edge node pairs, and the wavefront polarity of the adjacent edge node pairs satisfies the same or opposite direction characteristics of the traversing wave. Then, the line segment is determined to be a healthy segment, and a trust link marker representing the health of the line segment is generated.

9. A cloud-edge collaborative physical information fusion power distribution network fault location system according to claim 1, characterized in that, The specific process of constructing a fault trust topology network by aggregating trust link tags includes: Aggregate all generated trust link tokens; Map all trusted link tags to the global topology model of the distribution network; Preserve the nodes and edges covered by the trusted links to generate a fault trust topology subgraph composed of verified nodes.

10. A cloud-edge collaborative physical information fusion power distribution network fault location system according to claim 9, characterized in that, The specific process of identifying logical breakpoints or convergence points and outputting fault segment location results includes: Search for logical breakpoints or convergence points in the fault trust topology subgraph; A logical breakpoint refers to a node on an established trust link where the downstream trust link is interrupted. A convergence point is a point where a trust link fails to be established between two adjacent nodes, and the fault directions recorded by each node are opposite to each other. Identify the boundary nodes corresponding to the above logical breakpoints or convergence points; The line segment between the boundary nodes is identified as the final fault location, and the fault segment location result is output.

Images