Internet of Things-Based Intelligent Gateway Fault Detection System for Power Equipment

Abstract

The present invention discloses an intelligent gateway fault detection system for power equipment based on the Internet of Things, which relates to the technical field of intelligent gateways for power equipment and is used to solve the problem of inaccurate fault detection in existing power equipment monitoring systems. It includes: a power equipment access module, a data preprocessing and edge computing module, a data storage and analysis module, and a fault diagnosis and alarm module, with signal connections between the modules. The power equipment access module is used to collect the operation status data of power equipment; the data preprocessing and edge computing module cleans, converts the format of, and performs preliminary calculations on the collected data, reduces data redundancy, and improves processing efficiency; the data storage and analysis module stores the processed data and extracts feature information using big data analysis methods; the fault diagnosis and alarm module performs fault identification based on the analysis results and triggers an alarm signal to achieve accurate detection and efficient maintenance of the intelligent gateway of power equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent gateways for power equipment, and more specifically, to a fault detection system for intelligent gateways of power equipment based on the Internet of Things. Background Art

[0002] With the development of the power system towards intelligence and automation, a large number of power equipment requires efficient and stable remote monitoring and management. Traditional monitoring systems rely on fixed protocols and centralized data processing methods, suffering from problems such as poor device compatibility, incomplete data collection, and lagging fault detection, making it difficult to meet the requirements of modern power systems for efficient operation and maintenance.

[0003] Existing intelligent gateway products still have deficiencies in aspects such as multi-protocol compatibility, data processing capabilities, and fault diagnosis accuracy. Some devices only support a single communication protocol, resulting in difficulties in the interconnection and interoperability of devices from different manufacturers; on the other hand, traditional data analysis methods rely on cloud processing, making it difficult to identify and respond to on-site faults in a timely manner, increasing maintenance costs and operation and maintenance difficulties.

[0004] In view of the above problems, the present invention proposes a solution. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a fault detection system for intelligent gateways of power equipment based on the Internet of Things to solve the problems raised in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] In a preferred embodiment, it includes: a power equipment access module, a data preprocessing and edge computing module, a data storage and analysis module, and a fault diagnosis and alarm module, with signal connections between the modules;

[0008] The power equipment access module ensures the secure access of power equipment through protocol identification and encryption authentication, collects the operation data of power equipment, and adopts a hierarchical configuration strategy to achieve the automatic access and configuration of power equipment through the priority of power equipment with different protocols;

[0009] The data preprocessing and edge computing module cleans the operation data of power equipment through caching, parsing, interpolation, and anomaly detection, uses a decision tree classification model to judge the state of power equipment, analyzes the operation mode through K-Means clustering, and detects fault symptoms using the IsolationForest algorithm;

[0010] The data storage and analysis module adopts a hierarchical storage architecture to optimize the management of operation data of power equipment, analyzes the operation trend of power equipment using batch processing and stream processing, and identifies fault patterns using Fourier transform, STL decomposition, and wavelet transform;

[0011] The fault diagnosis and alarm module determines the cause of the fault through the fault knowledge base and reasoning methods, combines random forest and LSTM to identify the fault type and abnormal trend, and adopts a hierarchical alarm mechanism; it also supports real-time monitoring of the power equipment status through the SNMP protocol, remote management of power equipment through TR-069, and automatically generates an operation and maintenance report.

[0012] In a preferred embodiment, when the power equipment goes online, the power equipment access module automatically scans the communication ports of the power equipment, listens for and captures the operation data packets of the power equipment, and uses a protocol recognition engine to analyze the operation data packets of the power equipment using deep packet detection technology to identify the communication protocol used. For the supported protocols, the power equipment access module directly calls the corresponding parsing module for data decoding. For unknown protocols, the power equipment access module extracts protocol features and attempts to match the existing protocol library. When it cannot be recognized, the power equipment is marked as "pending manual configuration";

[0013] When the power equipment is first accessed, the power equipment access module uses TLS / SSL encryption authentication and OAuth2.0 power equipment authentication. After successful authentication, the power equipment access module assigns a unique session token to the power equipment;

[0014] The power equipment access module uses a remote terminal unit to poll the operation data of low-frequency power equipment and uses an event-triggered mechanism for the operation data of high-frequency abnormal power equipment, and parses the collected original operation data of the power equipment into a structured format through the protocol parsing layer for preliminary integrity check of the operation data of the power equipment;

[0015] The power equipment access module adopts a double-layer cache mechanism, maintains a short-term cache at the power equipment end based on Redis, starts a persistent cache based on SQLite when the power equipment network is abnormal, and automatically synchronizes and caches the operation data of the power equipment and marks the successfully uploaded operation data of the power equipment when the power equipment network is normal;

[0016] The power equipment access module uses a data conversion mapping table to convert the data structures of each protocol into a unified JSON format, and encapsulates the structured operation data of the power equipment in the standard format of timestamp + power equipment ID + measured value;

[0017] The power equipment access module adopts an adaptive data compression and hierarchical transmission strategy. In a low-speed network environment, lightweight compression algorithms such as LZ4 and Snappy are used. In a high-speed network environment, batch transmission Batching is used. In edge computing power equipment, the MQTT protocol is used. In the LPWAN Internet of Things scenario, the CBOR format is used;

[0018] The power equipment access module uses an LSTM-based sequence classification model to automatically identify the protocol type of power equipment. According to the identified protocol type, it automatically loads the corresponding parsing driver, and uses pattern matching and neural network structure inference to analyze the data frame structure of the power equipment operation data. Through Few-Shot Learning, it automatically maps the data fields.

[0019] In a preferred embodiment, when the power equipment is accessed, the power equipment access module automatically scans the TCP / UDP ports or serial ports, listens for the power equipment operation data packets sent by the power equipment, and adapts to TCP devices and serial port devices: MQTT corresponds to 1883, IEC61850 MMS corresponds to 102, ModbusTCP corresponds to 502, ModbusRTU and DL / T645 are read through UART / RS485. Then, the power equipment access module listens to the data stream and extracts the protocol feature fields: for the MQTT protocol, it listens and extracts the CONNECT, SUBSCRIBE, and PUBLISH fixed message headers; for the Modbus protocol, it listens and extracts the message start identifier and function code; for the DL / T645 protocol, it listens and extracts the 68H frame header, power equipment address field, and checksum; for the IEC61850 protocol, it listens and extracts the unique ASN.1 encoding of MMS / GOOSE / SV. Finally, the power equipment access module matches the extracted feature fields with the protocol library to determine the protocol type.

[0020] The power equipment access module loads the corresponding parsing driver according to the protocol type: for power equipment with the MQTT protocol, it parses the JSON payload data; for power equipment with the Modbus protocol, it reads the register data; for power equipment with the DL / T645 protocol, it decodes the electricity meter data frame and verifies the CRC checksum; for power equipment with the IEC61850 protocol, it parses the MMS / GOOSE / SV messages and establishes a logical node mapping. Finally, it initializes the communication for power equipment with the four protocols of MQTT, Modbus, DL / T645, and IEC61850.

[0021] The power equipment access module converts the data structures of different protocols into a computable data format: for MQTT data, it directly parses the JSON format; for Modbus protocol data, it reads the data from the register address and converts the fields according to the mapping table; for DL / T645 protocol data, it decodes the data frame and parses the fields according to the protocol specification; for IEC61850 protocol data, it extracts the logical node data in the MMS / GOOSE message.

[0022] The power equipment access module determines the configuration priorities of each protocol based on the real-time performance T of the power equipment operation data, the importance I of the power equipment operation data, the access complexity C of the power equipment, the communication reliability R of the power equipment, and the power consumption constraint E of the power equipment. The specific steps are as follows:

[0023] Step A1: Measure the value of the real-time performance T of the power equipment operation data through the requirements of the power equipment for the update frequency of the power equipment operation data, measure the value of the importance I of the power equipment operation data through the category of the power equipment, determine the value of the access complexity C of the power equipment according to the complexity of the protocol, measure the value of the communication reliability R of the power equipment by the network quality and the support situation of the power equipment protocol, and measure the value of the power consumption constraint E of the power equipment through the power consumption mode of the power equipment;

[0024] Step A2: Determine the comprehensive priority Ptotal of the protocol power equipment according to the quantified values of the real-time performance T of the power equipment operation data, the importance I of the power equipment operation data, the access complexity C of the power equipment, the communication reliability R of the power equipment, and the power consumption constraint E of the power equipment. The specific basis is the formula: Ptotal = Qt×T + Qi×I + Qc×C + Qr×R + Qe×E, where Qt represents the weight of the real-time performance T of the power equipment operation data, Qi represents the weight of the importance I of the power equipment operation data, Qc represents the weight of the access complexity C of the power equipment, Qr represents the weight of the communication reliability R of the power equipment, and Qe represents the weight of the power consumption constraint E of the power equipment, indicating the importance of different factors;

[0025] The power equipment access module sorts all newly accessed power equipment according to the comprehensive priority Ptotal, and starts from the newly accessed power equipment with the highest priority, and gradually configures it to the newly accessed power equipment with a lower score.

[0026] In a preferred embodiment, the data preprocessing and edge computing module receives the power equipment operation data transmitted by the power equipment access module. The edge computing power equipment temporarily stores the power equipment operation data through the data buffer queues Kafka and Redis. The data preprocessing and edge computing module uses a protocol parser to parse the original power equipment operation data and standardizes the power equipment operation data of different protocols into a unified format;

[0027] The data preprocessing and edge computing module processes the lost power equipment operation data by using moving average filling or interpolation method;

[0028] The data preprocessing and edge computing module calculates the mean and standard deviation of the operation data of power equipment using Z-Score applicable to normally distributed data, calculates the IQR value of the operation data of power equipment using IQR applicable to non-normally distributed data, and uses DBSCAN applicable to anomaly detection in complex environments to identify isolated points by calculating the density of neighboring points; determines whether the operation data of power equipment is abnormal, and finally standardizes the format of the operation data of power equipment.

[0029] The data preprocessing and edge computing module trains a decision tree-based power equipment status classification model to classify the status data of power equipment and determine whether the power equipment is in one of the three states: normal, warning, and fault; and uses K-Means clustering to perform clustering analysis on the operation data of power equipment with long-term operation to identify normal operation modes and detect abnormal modes; uses the IsolationForest algorithm to detect early fault signs by analyzing the characteristic values of the operation data of power equipment.

[0030] In a preferred embodiment, the data storage and intelligent analysis module uses a column storage format for time series data, a relational database applicable to storing information such as power equipment archives, operation status, and maintenance logs for structured data, and an object storage method for unstructured data. Then, it adopts a time slicing strategy to store the operation data of power equipment according to time windows. The data storage and intelligent analysis module also establishes a multi-level index and adopts a caching mechanism.

[0031] The data storage and intelligent analysis module uses an outlier detection algorithm to filter invalid operation data of power equipment and scales the operation data of power equipment to a unified range through standardization.

[0032] The data storage and intelligent analysis module performs batch analysis on historical operation data of power equipment through batch processing applicable to offline big data analysis and stream processing applicable to real-time data analysis for anomaly detection. It also uses Fourier transform to analyze the frequency characteristics of signals, uses STL to separate long-term trends, periodic changes, and random fluctuations, and detects signal mutation points through wavelet transform technology to identify abnormal modes.

[0033] In a preferred embodiment, the fault diagnosis and alarm module establishes a power equipment fault knowledge base to store known fault modes, possible causes, and corresponding treatment plans, and uses rule-based reasoning (RBR) with IF-THEN rules to judge known faults and case-based reasoning (CBR) to infer possible fault causes through similarity matching of historical fault cases. The fault diagnosis and alarm module uses a multi-classification model, random forest, to identify fault categories and uses LSTM to analyze time series data to detect abnormal operation trends; and sets an alarm mechanism according to the fault impact range and urgency.

[0034] The fault diagnosis and alarm module uses the SNMP protocol to obtain the status of power equipment in real time, remotely upgrades the firmware through TR-069 remote management that supports remote modification of power equipment parameters to repair software problems, and also automatically generates an operation and maintenance report containing major fault statistics, the operation trend of power equipment, and maintenance suggestions based on NLG.

[0035] The present invention discloses an intelligent gateway fault detection system for power equipment based on the Internet of Things, which relates to the technical field of intelligent gateways for power equipment and is used to solve the problem of inaccurate fault detection in existing power equipment monitoring systems. It includes: a power equipment access module, a data preprocessing and edge computing module, a data storage and analysis module, and a fault diagnosis and alarm module, with signal connections between the modules. The power equipment access module is used to collect the operation status data of power equipment; the data preprocessing and edge computing module cleans, converts the format of, and performs preliminary calculations on the collected data to reduce data redundancy and improve processing efficiency; the data storage and analysis module stores the processed data and extracts feature information using big data analysis methods; the fault diagnosis and alarm module performs fault identification based on the analysis results and triggers an alarm signal to achieve accurate detection and efficient maintenance of the intelligent gateway of power equipment. Description of the Drawings

[0036] Figure 1 It is a schematic structural diagram of the intelligent gateway fault detection system for power equipment based on the Internet of Things of the present invention. Detailed Embodiments

[0037] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0038] Embodiment

[0039] The present invention discloses an intelligent gateway fault detection system for power equipment based on the Internet of Things, including: a power equipment access module, a data preprocessing and edge computing module, a data storage and analysis module, and a fault diagnosis and alarm module, with signal connections between the modules.

[0040] The power equipment access module is mainly responsible for connecting to the intelligent gateway of power equipment, collecting the operation data of power equipment, including: current, voltage, temperature, communication status, and transmitting the standardized data to the data preprocessing and edge computing module.

[0041] Specifically, the power equipment access module adapts to the data access of different types of power equipment by being compatible with protocols such as MQTT, Modbus, DL / T645, and IEC61850.

[0042] First, when the power equipment goes online, the power equipment access module automatically scans the communication ports of the power equipment, listens for and captures data packets, and uses a protocol recognition engine to analyze the power equipment data packets using Deep Packet Inspection (DPI) technology to identify the communication protocols in use, such as MQTT, Modbus, DL / T645, and IEC61850. If it is a supported protocol, the power equipment access module directly calls the corresponding parsing module to decode the data. If it is an unknown protocol, the power equipment access module extracts the protocol features and attempts to match the existing protocol library. If it still cannot be identified, the power equipment is marked as "pending manual configuration".

[0043] Furthermore, when the power equipment is first accessed, the power equipment access module uses TLS / SSL encryption authentication and OAuth2.0 power equipment authentication to prevent illegal power equipment from accessing. First, the power equipment access module sends an authentication request to the power equipment, and the power equipment returns identity information, such as the power equipment ID, MAC address, firmware version, etc. At the same time, the service uses public-private key encryption (RSA or ECC) to verify the identity information to ensure the trustworthiness of the power equipment. After successful authentication, the power equipment access module assigns a unique session token (Token) to the power equipment and establishes a secure communication channel.

[0044] Furthermore, the power equipment access module uses a remote terminal unit (RTU). For low-frequency data (such as temperature and voltage), it uses a polling method to request power equipment data at fixed time intervals. For high-frequency abnormal data (such as instantaneous current mutation), it uses an event-triggered mechanism. When the power equipment detects that the threshold is exceeded, it immediately pushes the data to the collection end. And the raw data collected is parsed into a structured format through the protocol parsing layer, and a preliminary data integrity check (such as missing value and duplicate value detection) is performed.

[0045] Furthermore, the power equipment access module adopts a double-layer cache mechanism (memory + local storage), maintaining a short-term cache (based on Redis or EdgeDB) at the power equipment end to store data in the recent few seconds. If the power equipment network is abnormal, a persistent cache (based on SQLite / LevelDB) is started to store data for a longer time. After the network is restored, the power equipment access module automatically synchronizes the cached data to the server and marks the data that has been successfully uploaded to avoid duplicate transmission.

[0046] Furthermore, the power equipment access module uses a data conversion mapping table (DataMapping) to convert the data structures of each protocol into a unified JSON / Protobuf format, and encapsulates the structured data in the standard format of timestamp + power equipment ID + measurement value.

[0047] Furthermore, the power equipment access module adopts adaptive data compression (AdaptiveDataCompression) and hierarchical transmission strategy (HierarchicalTransmission) to reduce data transmission overhead and improve transmission reliability. Specifically, in a low-speed network environment, lightweight compression algorithms (such as LZ4, Snappy) are used to reduce the size of data packets. In a high-speed network environment, batching is adopted to merge multiple data packets before sending, reducing the request frequency. For edge computing power equipment, the MQTT protocol (QoS level 1 or 2) is used to ensure reliable data transmission. In 5G or LPWAN IoT scenarios, the CBOR (Compact Binary Object Representation) format, which is smaller than JSON, is adopted to improve bandwidth utilization.

[0048] Finally, the power equipment access module uses an LSTM-based sequence classification model to automatically identify the protocol type of power equipment, and according to the identified protocol type, automatically loads the corresponding parsing driver, and adopts pattern matching and neural network structure reasoning to analyze the data frame structure of power equipment. Through Few-Shot Learning, it automatically maps data fields such as current, voltage, temperature, etc., instead of relying on fixed field mapping rules, to achieve automatic configuration of newly added power equipment.

[0049] It should be noted that MQTT adopts the publish-subscribe mode. Power equipment actively pushes data, and the server only needs to subscribe to the corresponding topic to obtain the data without making an active request. At the same time, MQTT power equipment generally uses JSON or Protobuf format, and the data fields are dynamic. The data structures of different power equipment may be different, and dynamic parsing needs to be performed according to the specific definition of the power equipment. Modbus adopts the master-slave query mode. The server must actively send a request to the power equipment before the power equipment will return data. Therefore, a polling mechanism needs to be maintained on the server side. At the same time, Modbus adopts a fixed-length binary format, and the data is stored in fixed register addresses. The meaning corresponding to each register needs to be parsed by looking up the mapping table. DL / T645 is also a request-response mechanism, but it also involves the identity authentication of power equipment. Only after the authentication is passed can the server read the data. And DL / T645 uses a strict fixed frame format, and the byte position of each field is fixed. It needs to be disassembled according to the byte bits during parsing, otherwise data parsing errors will occur. IEC61850 is more complex. It not only supports the server polling mode (MMS), but also supports power equipment to actively broadcast data (GOOSE / SV). The diversity of this communication mechanism makes it impossible to use a unified method to automatically configure all power equipment. IEC61850 also adopts a logical node (LN) structure, and the data is stored in a hierarchical manner. When parsing, multiple levels of data objects need to be read and logical mapping needs to be performed. As a result, the data formats and parsing methods of different protocols are different, and it is impossible to adapt all power equipment through a general automatic configuration method.

[0050] Therefore, the power equipment access module adopts a hierarchical configuration strategy to achieve the automatic access and configuration of power equipment through steps such as protocol identification, driver loading, and data standardization.

[0051] First, the power equipment access module determines the communication protocol used by the newly accessed power equipment to ensure correct subsequent data parsing. Specifically, when the power equipment is accessed, the power equipment access module automatically scans TCP / UDP ports or serial ports, listens for data packets sent by the power equipment, and adapts to TCP power equipment and serial port power equipment, such as: MQTT (default 1883), IEC61850 MMS (102), Modbus TCP (502), Modbus RTU, DL / T645 is read through UART / RS485. Then the power equipment access module listens to the data stream and extracts protocol characteristic fields. For example, for the MQTT protocol, it listens and extracts: fixed message headers such as CONNECT, SUBSCRIBE, PUBLISH. For the Modbus protocol, it listens and extracts: message start identifier, function code (such as 0x03 for reading registers). For the DL / T645 protocol, it listens and extracts: 68H frame header, power equipment address field, checksum. For the IEC61850 protocol, it listens and extracts: specific ASN.1 encoding of MMS / GOOSE / SV. Finally, the power equipment access module matches the extracted characteristic fields with the protocol library to determine the protocol type. When the matching fails, an LSTM sequence classification model is used for protocol inference. If it still cannot be recognized, it is marked as "pending manual configuration" and stored in the exception log.

[0052] After identifying the protocol, the power equipment access module loads the corresponding parsing driver according to the protocol type. For example: for power equipment with the MQTT protocol, it parses JSON / Protobuf payload data and registers subscribed topics. For power equipment with the Modbus protocol, it reads register data and maps it to a standard data structure. For power equipment with the DL / T645 protocol, it decodes the electric meter data frame and verifies the CRC checksum. For power equipment with the IEC61850 protocol, it parses MMS / GOOSE / SV messages and establishes a logical node mapping. Finally, communication initialization is performed on power equipment with the four protocols of MQTT, Modbus, DL / T645, and IEC61850 to ensure reliable data transmission.

[0053] Furthermore, the power equipment access module converts data structures of different protocols into computable data formats. Specifically, for MQTT data, it directly parses the JSON / Protobuf format and extracts the power equipment data fields. For Modbus protocol data, it reads data from register addresses and converts fields according to the mapping table. For DL / T645 protocol data, it decodes the data frame and parses fields according to the protocol specification. For IEC61850 protocol data, it extracts the logical node (LN) data in the MMS / GOOSE message. Then, it uniformly converts all data into the format of timestamp + power equipment ID + data type + data value. And it uses a data mapping table (DataMapping) to convert protocol-specific fields into standard fields, and uploads the converted data to the data storage and analysis module.

[0054] Furthermore, in the hierarchical data transmission strategy, the power equipment access module adopts QoS = 1 or QoS = 2 for power equipment using the MQTT protocol. For power equipment using the Modbus protocol, the data is preferentially stored locally and uploaded in batches after the network is restored, while ensuring reliable data transmission.

[0055] When multiple power equipment of different protocols are connected simultaneously, resources (such as computing power, network bandwidth, data storage) are limited. Configuring all power equipment at the same time may lead to computing resource conflicts, communication bottlenecks, and data integrity problems. Therefore, in this embodiment, the power equipment access module comprehensively considers the data real-time performance (T), the importance of power equipment data (I), the access complexity of power equipment (C), the communication reliability of power equipment (R), and the power consumption constraint of power equipment (E) to determine the configuration priority of each protocol, and completes the power equipment configuration in the optimal order in turn to ensure that critical data is processed first, communication is stable, and resources are reasonably utilized.

[0056] First, the power equipment access module defines and quantifies the evaluation criteria for each factor. Specifically, the value of data real-time performance (T) is measured by the requirement of the power equipment for data update frequency. For example, for high real-time performance requirements (such as power equipment that needs to collect data every second), T = 3; for medium real-time performance requirements (such as power equipment that needs to collect data every minute), T = 2; for low real-time performance requirements (such as power equipment that needs to collect data every hour), T = 1. The value of the importance of power equipment data (I) is measured by the category of the power equipment. For example, for high-importance power equipment (such as core power equipment of the power grid), I = 3; for medium-importance power equipment (such as auxiliary power equipment, general sensors), I = 2; for low-importance power equipment (such as non-critical detection power equipment), I = 1. The value of the access complexity of the power equipment (C) is determined according to the complexity of the protocol. For example, for high-complexity protocols (such as IEC61850), C = 3; for medium-complexity protocols (such as DL / T645), C = 2; for low-complexity protocols (such as Modbus, MQTT), C = 1. The value of the communication reliability of the power equipment (R) is measured by the network quality and the support situation of the power equipment protocol. For example, for high-reliability power equipment (such as power equipment that uses the TCP / IP protocol and has a multi-path redundant network), R = 3; for medium-reliability power equipment (such as Wi-Fi network transmission), R = 2; for low-reliability power equipment (such as power equipment that uses a relatively unstable wireless transmission technology), R = 1. The value of the power consumption constraint of the power equipment (E) is measured by the power consumption mode of the power equipment. For example, for high power consumption constraints (such as battery-powered, low-power power equipment), E = 3; for medium power consumption constraints (such as power equipment with an external power supply), E = 2; for low power consumption constraints (such as power equipment that is always connected to the power grid), E = 1.

[0057] Furthermore, a comprehensive score is calculated based on the five evaluation factors of data real-time performance (T), the importance of power equipment data (I), the access complexity of the power equipment (C), the communication reliability of the power equipment (R), and the power consumption constraint of the power equipment (E) to determine the configuration priority of the power equipment. Specifically, according to the formula: Ptotal = Qt×T + Qi×I + Qc×C + Qr×R + Qe×E, where Qt represents the weight of data real-time performance (T), Qi represents the weight of the importance of power equipment data (I), Qc represents the weight of the access complexity of the power equipment (C), Qr represents the weight of the communication reliability of the power equipment (R), Qe represents the weight of the power consumption constraint of the power equipment (E), indicating the importance of different factors. Ptotal represents the comprehensive priority of the power equipment of this protocol. The higher the priority, the higher the priority of the power equipment of this protocol.

[0058] The power equipment access module sorts all newly accessed power equipment according to the comprehensive priority Ptotal. The access order will start from the newly accessed power equipment with the highest priority and gradually configure to the newly accessed power equipment with lower scores, ensuring that important data that needs to be processed in real time (such as relay status, transformer monitoring data, etc.) is configured first to ensure the efficient operation of power equipment.

[0059] The data preprocessing and edge computing module mainly cleans and optimizes the operation data of power equipment transmitted by the power equipment access module, and uses edge computing power equipment for preliminary analysis to improve the quality of power equipment operation data and reduce network transmission pressure.

[0060] First, the data preprocessing and edge computing module receives the operation data of power equipment transmitted by the power equipment access module. The edge computing device temporarily stores the data through the data buffer queues Kafka and Redis to prevent loss when the data flow surges. At the same time, the data preprocessing and edge computing module uses a protocol parser (ProtocolParser) to parse the original data and standardize the data in different protocols into a unified format, such as JSON or CSV.

[0061] Furthermore, for small-scale missing data (such as the missing of a single data point) in the data preprocessing and edge computing module, it uses moving average filling or interpolation methods (linear interpolation, Lagrange interpolation), and judges the data missing ratio and deletes abnormal records. For example, when the data missing ratio exceeds 30%, the data is determined to be invalid and deleted.

[0062] At the same time, the data preprocessing and edge computing module uses three anomaly detection algorithms to ensure effective anomaly detection under different data distributions. Specifically, it uses Z-Score (standard score detection) applicable to normally distributed data (such as voltage and current) to calculate the mean and standard deviation of the data. If a data point deviates more than 3 times the standard deviation from the mean, it is marked as abnormal; it uses IQR (interquartile range method) applicable to non-normally distributed data (such as equipment operation duration) to calculate the IQR value (Q3 - Q1) of the data. If the data exceeds [Q1 - 1.5×IQR, Q3 + 1.5×IQR], it is marked as abnormal; it uses DBSCAN (density-based clustering anomaly detection) applicable to anomaly detection in complex environments (such as load fluctuation data) to identify isolated points by calculating the density of neighboring points and judge whether it is abnormal. Finally, the data format is standardized. For example, the time format is standardized (such as unified to UTC timestamp), unit conversion (such as kV converted to V, MW converted to kW), and field names are unified (to avoid inconsistent field naming for different devices).

[0063] Furthermore, the data preprocessing and edge computing module trains a power equipment status classification model based on DecisionTree to classify the collected data such as voltage, current, and temperature, and determine whether the equipment is in one of the three states: normal, warning, or fault. K-Means clustering is used to perform clustering analysis on the long-term operation data of the equipment to identify the normal operation mode and detect abnormal modes. The IsolationForest algorithm is used to analyze the data eigenvalue in advance to discover possible fault signs. For example, a sudden increase in current may mean motor overload, and a continuous rise in equipment temperature may mean a cooling system failure.

[0064] The data storage and intelligent analysis module is mainly used to store and manage data, and perform in-depth analysis to provide a decision-making basis for fault detection.

[0065] First, the data storage and intelligent analysis module adopts different storage architectures according to the time-series data, structured data, and unstructured data included in the power equipment operation data. Specifically, for time-series data (such as current, voltage, power, etc.), a column storage format is adopted to optimize the writing and query of high-frequency data; for structured data (such as equipment status, operation and maintenance records, etc.), a relational database (PostgreSQL / MySQL) suitable for storing information such as equipment files, operation status, and maintenance logs is used; for unstructured data (such as fault logs, pictures, videos, etc.), an object storage method is adopted to support the fast access of large files and a distributed storage (HDFS / MinIO) suitable for storing large-scale log files and video surveillance data. Then, a time slicing (TimePartitioning) strategy is adopted to store the data according to time windows to improve the query efficiency, and a multi-level index (time index + equipment ID index) is established to optimize the data retrieval performance. Finally, a caching mechanism (Redis / Memcached) is adopted to reduce the database query load and improve the response speed.

[0066] Furthermore, the data storage and intelligent analysis module adopts an outlier detection algorithm to filter invalid data, improve data accuracy, and scale the data to a unified range through standardization (Min-MaxScaling / Z-scoreNormalization) to prevent the influence of different dimensions on the analysis results.

[0067] Finally, the data storage and intelligent analysis module performs batch analysis on historical data through batch processing (Hadoop + HDFS) suitable for offline big data analysis and stream processing (Apache Spark Streaming / Flink) suitable for real-time data analysis of anomaly detection, to mine the operation trends of the devices. At the same time, the data storage and intelligent analysis module also uses time series analysis methods to mine the operation rules of the devices. For example, it uses Fourier transform (FFT) to analyze the frequency characteristics of signals and identify periodic fault patterns. It uses STL (Seasonal-Trend Decomposition) to separate long-term trends, periodic changes, and random fluctuations to improve prediction accuracy, and detects signal mutation points through wavelet transform (Wavelet Transform) technology to identify abnormal patterns. Finally, the data storage and intelligent analysis module pushes the fault analysis results to the fault diagnosis and alarm module to achieve quick response.

[0068] The fault diagnosis and alarm module is mainly used to identify power equipment faults, trigger alarms, and provide remote maintenance support based on the fault analysis results transmitted by the data storage and intelligent analysis module.

[0069] First of all, the fault diagnosis and alarm module establishes a power equipment fault knowledge base, stores known fault patterns, possible causes, and corresponding treatment plans, and uses rule-based reasoning (RBR) to use IF-THEN rules to judge known faults (for example: "If the voltage fluctuation exceeds 10%, it may be a line fault") and case-based reasoning (CBR) to infer the cause of the fault by matching the similarity of historical fault cases (for example: "The current fault characteristics are similar to the historical records of a certain equipment fault, and it may be transformer aging"). Then, the fault diagnosis and alarm module uses the multi-classification model Random Forest to identify fault categories (such as short circuit, overload, line aging), and uses LSTM (Long Short-Term Memory Network) to analyze time series data to detect abnormal operation trends and give early warnings of potential faults.

[0070] Furthermore, the fault diagnosis and alarm module sets an alarm mechanism according to the scope and urgency of the fault, and notifies the staff to handle it. For example, according to the scope and urgency of the fault, a three-level alarm mechanism is set: minor anomalies, faults that do not affect the operation of the equipment are low-level alarms, and such alarms are only stored in the log and reported to the operation and maintenance personnel regularly; the performance of the equipment decreases, faults that may affect the operation are medium-level alarms, and such alarms are pushed to the operation and maintenance personnel through text messages, emails, and APPs, and the alarm log is recorded and marked as "to be processed"; serious faults, faults that affect the safety of the equipment or the entire power grid are high-level alarms, and such alarms are immediately connected to the SCADA system, and other devices are linked to execute emergency plans (such as automatically cutting off the circuit), and the operation and maintenance person in charge is directly warned by voice calls, WeChat / DingTalk.

[0071] Furthermore, the fault diagnosis and alarm module uses the SNMP protocol to obtain the device status (temperature, current, voltage) in real time, and remotely upgrades the firmware through TR-069 remote management that supports remote modification of device parameters to fix software problems.

[0072] Finally, the fault diagnosis and alarm module automatically generates an operation and maintenance report based on NLG, including main fault statistics, device operation trends, and maintenance suggestions, for decision-making reference.

[0073] The above formulas are all dimensionless and take their numerical values for calculation. The formula is obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0074] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product.

[0075] Those of ordinary skill in the art can realize that the modules and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application of the technical solution and the inventive constraints. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0076] In addition, the functional modules in each embodiment of this application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0077] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0078] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. Intelligent gateway fault detection system for power equipment based on the Internet of Things, It is characterized by comprising: power equipment access module, data preprocessing and edge computing module, data storage and analysis module, fault diagnosis and alarm module, and signal connection between modules; The power equipment access module is configured as follows: When the power equipment is online, it automatically scans the TCP / UDP port or serial port, monitors and captures the uplink data packets; uses deep packet inspection to extract the traffic feature sequence, and uses the long short-term memory network to classify the communication protocol of the features; directly loads the corresponding parsing driver for the supported protocols, and uses few-sample learning to map the fields for unknown protocols. If it still cannot be identified, it is marked as "pending manual configuration"; TLS / SSL and OAuth2.0 dual identity authentication are performed during the first access, and the identity of the power equipment is verified through asymmetric encryption. After successful authentication, a unique session token is allocated and a secure channel is established; Adopt polling-event hybrid collection: low-frequency data is obtained by the remote terminal unit according to the set periodic polling, and instantaneous abnormal data is pushed immediately by the intelligent collection port when the threshold is triggered; Use Redis to build a short-term cache in seconds, and switch to a persistent cache based on SQLite when the network is abnormal; automatically synchronize and upload cached data with deduplication after the network is restored; The multi-protocol original frame is converted into a unified JSON structure through the data mapping table and encapsulated in the format of "timestamp + power equipment ID + measurement value"; Adaptively select LZ4 or Snappy compression, batch transmission, MQTT or CBOR packet encoding according to the link status to achieve layered transmission; For each gateway to be connected, the real-time performance T of the power equipment operation data, the importance I of the data, the complexity C of the access, the reliability R of the communication and the power consumption constraint E are quantified respectively, and the comprehensive priority is calculated according to the formula P_total = Q_t·T + Q_i·I + Q_c·C + Q_r·R+ Q_e·E. The connection is established, the bandwidth is allocated and the processing order is carried out from high to low according to P_total; The data preprocessing and edge computing module cleans the power equipment operation data through temporary storage, parsing, interpolation and anomaly detection, applies the decision tree classification model to judge the power equipment status, analyzes the operation mode through K-Means clustering, and uses the IsolationForest algorithm to detect fault signs; The data storage and analysis module uses a hierarchical storage architecture to optimize the management of power equipment operation data, and uses batch processing and stream processing to analyze the operation trend of power equipment, and uses Fourier transform, STL decomposition and wavelet transform to identify fault modes; The fault diagnosis and alarm module determines the cause of the fault through the fault knowledge base and reasoning method, combines random forest and LSTM to identify the fault type and abnormal trend, and adopts a hierarchical alarm mechanism; it also supports the SNMP protocol to monitor the status of power equipment in real time, TR-069 to remotely manage power equipment, and automatically generates operation and maintenance reports.

2. According to the Internet of Things-based power equipment intelligent gateway fault detection system according to claim 1, it is characterized in that; The data preprocessing and edge computing module receives the power equipment operation data transmitted by the power equipment access module. The edge computing power equipment temporarily stores the power equipment operation data through the data buffer queues Kafka and Redis. The data preprocessing and edge computing module uses a protocol parser to parse the original power equipment operation data and standardize the power equipment operation data of different protocols into a unified format. The data preprocessing and edge computing module uses moving average filling or interpolation method to process the missing power equipment operation data; The data preprocessing and edge computing module uses Z-Score, which is applicable to normally distributed data, to calculate the mean and standard deviation of the power equipment operation data, and uses IQR, which is applicable to non-normally distributed data, to calculate the IQR value of the power equipment operation data. It uses DBSCAN, which is applicable to anomaly detection in complex environments, to identify isolated points through neighboring point density calculation; it determines whether the power equipment operation data is abnormal, and finally standardizes the power equipment operation data format; The data preprocessing and edge computing module trains a power equipment status classification model based on a decision tree, classifies the power equipment status data, and determines whether the power equipment is in normal, warning, or fault status; And K-Means clustering is used to perform cluster analysis on the long-term operation data of power equipment to identify normal operation modes and detect abnormal modes; The IsolationForest algorithm is used to analyze the characteristic values ​​of power equipment operation data to detect fault signs in advance.

3. The power equipment intelligent gateway fault detection system based on the Internet of Things according to claim 2 is characterized in that: The data storage and intelligent analysis module uses column storage format for time series data, and a relational database suitable for storing power equipment archives, operating status, maintenance logs and other information for structured data. Object storage is used for unstructured data, and then a time slicing strategy is adopted to store power equipment operating data according to time windows. The data storage and intelligent analysis module also establishes multi-level indexes and adopts a cache mechanism; The data storage and intelligent analysis module uses an outlier detection algorithm to filter invalid power equipment operation data and scales the power equipment operation data to a uniform range through standardization; The data storage and intelligent analysis module performs batch analysis on historical power equipment operation data through batch processing suitable for offline big data analysis and stream processing suitable for real-time data analysis for anomaly detection. It also uses Fourier transform to analyze the frequency characteristics of signals, uses STL to separate long-term trends, periodic changes and random fluctuations, and uses wavelet transform technology to detect signal mutation points and identify abnormal patterns.

4. The power equipment intelligent gateway fault detection system based on the Internet of Things according to claim 3 is characterized in that: The fault diagnosis and alarm module establishes a knowledge base of power equipment faults, stores known fault modes, possible causes and corresponding treatment plans, and uses reasoning RBR to use IF-THEN rules to judge and reason about known faults. CBR infers the cause of faults through similarity matching of historical fault cases. The fault diagnosis and alarm module uses a multi-classification model random forest to identify fault categories, and uses LSTM to analyze time series data and detect abnormal operation trends; and sets an alarm mechanism based on the scope and urgency of the fault impact; The fault diagnosis and alarm module uses the SNMP protocol to obtain the status of power equipment in real time, remotely upgrades the firmware and fixes software problems through TR-069 remote management that supports remote modification of power equipment parameters, and automatically generates operation and maintenance reports based on NLG that include major fault statistics, power equipment operation trends, and maintenance recommendations.

Images