Power Internet of Things equipment access registration method and system in multi-protocol environment

Abstract

The invention discloses a method and a system for accessing and registering power Internet of Things equipment in a multi-protocol environment. The method comprises an original communication protocol identification step, a communication data semantic identification step, a target communication protocol conversion step and a device registration step. According to the invention, protocol identification, dynamic protocol conversion and equipment unique ID registration are carried out on equipment accessed to the electric power Internet of Things for the first time, intelligent identification and semantic level conversion of heterogeneous protocols are realized, incremental learning of unknown protocols is supported, real-time communication between equipment and a platform and between equipment is supported, and the real-time communication between equipment and platform is realized. And the compatibility and the adaptive capability of the electric power Internet of Things system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid equipment, and in particular to a method and device for accessing and registering power Internet of Things devices in a multi-protocol environment. Background Art

[0002] With the intelligent evolution of the power Internet of Things (IoT), the conflict between the automatic access requirements of terminal devices and the heterogeneous protocol environment is becoming increasingly prominent. Terminal devices from different manufacturers (such as smart meters, distribution terminals, and transmission sensors) generally face protocol heterogeneity barriers. For example, leading manufacturers tend to adopt international standard protocols (such as IEC 61850), while small and medium-sized manufacturers rely on proprietary protocols or outdated industry standards (such as DL / T 645-1997). Even devices with the same functional type may use completely incompatible communication stacks depending on the manufacturer, resulting in the inability of cross-vendor devices to communicate directly on the same network. The widespread phenomenon of non-standard communication protocols also directly affects the stable operation of devices and the accurate transmission of data.

[0003] Currently, terminal devices must be plug-and-play with a variety of access protocols and authentication mechanisms. However, the complexity of protocols (according to industry statistics, there are over 200 mainstream protocols) and the fragmentation of authentication standards have led to access detection requiring simultaneous handling of multi-dimensional technical challenges, including communication protocol compatibility verification, network load balancing, and device security compliance scanning. Traditional access detection solutions rely primarily on static protocol library matching and manual rule configuration, resulting in low adaptation efficiency, insufficient semantic parsing depth, and a disconnect between security detection and access processes. These solutions are no longer able to meet the access efficiency and reliability requirements of the digital transformation of the power grid. Summary of the Invention

[0004] Purpose of the invention: In response to the problems in the above-mentioned background technology, the present invention proposes a method and system for access and registration of power Internet of Things devices in a multi-protocol environment, which solves the barriers to device interconnection and interoperability in the power Internet of Things caused by heterogeneous protocols, and realizes automatic, secure and efficient access of terminal devices under diversified protocols and authentication mechanisms.

[0005] In order to achieve the above-mentioned object of the invention, the present invention adopts the following technical solutions:

[0006] A method for accessing and registering power Internet of Things devices in a multi-protocol environment, comprising:

[0007] Original communication protocol identification step: Extract the original protocol information of the current device based on the document information or transmission message actively sent by the power Internet of Things device; if the device does not actively send protocol information or the protocol information extraction fails, extract key features from the original communication data, match the key features with the protocol feature library, and identify the original protocol information of the current device. If the match fails, use the protocol classifier to classify the unmatched unconventional protocols. If the classification is unsuccessful, update the protocol feature library through online learning and execute the original communication protocol identification step again;

[0008] Communication data semantic recognition steps: Preprocess the original communication data to standardize data consistency; use natural language processing technology to understand the communication data protocol, identify professional entities in the power Internet of Things, combine domain knowledge graphs to establish contextual semantic associations, and convert the communication data into an intermediate standard representation;

[0009] Target communication protocol conversion step: Based on rule base matching, the intermediate standard representation is converted into data in the target communication protocol format. If the current rule base matching is unsuccessful, the intelligent mapping model is used to calculate the semantic similarity between the intermediate standard representation and the target protocol, and the cross-protocol mapping rules are generated by combining the knowledge graph constraints, and the rule base is automatically updated.

[0010] Device registration steps: Send the current device information and its original protocol information to the current regional IoT server. The server will generate a unique device ID based on the current device information and its original protocol information, perform device identity authentication and security testing, and officially connect the device to the power IoT system.

[0011] Preferably, in the original communication protocol identification step, the protocol classifier adopts a hybrid neural network model including a one-dimensional convolutional neural network and a bidirectional long short-term memory network, wherein the one-dimensional convolutional neural network extracts local spatial features, and the bidirectional long short-term memory network captures the order and grammatical relationship between these features; the input of the hybrid neural network model is a vectorized representation of the message byte sequence, and the output is a preset probability distribution of conventional protocol types, and whether it is an unconventional protocol is determined according to the probability distribution threshold; if the classifier successfully identifies, it enters the communication data semantic recognition and target communication protocol conversion step; if it is an unconventional protocol, it enters the online learning step.

[0012] Preferably, the online learning step includes:

[0013] The protocol feature analysis engine extracts session pattern features from unconventional protocols that exceed the protocol classifier threshold, including one or more of the following: message timing relationships and field position distribution entropy values. Based on the session pattern features, an incremental feature vector is generated through an external query interface and local cluster analysis to update the protocol feature library and classifier model parameters.

[0014] The protocol feature library stores static feature fingerprints of multiple communication protocols and their various versions and dynamic incremental features of unknown protocols, where the static feature fingerprints include one or more of the fixed port number, message header identifier and regular expression template of the protocol instruction set, and the dynamic incremental features continuously update the feature fingerprints of unknown protocols through an online learning mechanism; the protocol feature library supports regular expression matching, version update, and unknown protocol update.

[0015] Preferably, the external query interface is used to initiate an external query request to an external data source when encountering an unknown protocol to obtain relevant information of the unknown protocol, wherein the external data source includes at least one of the following: a public protocol knowledge base, a third-party protocol parsing service API, and an expert collaboration platform;

[0016] The protocol feature analysis engine formats and parses the protocol feature data obtained from the external query interface, determines the difference between the unknown protocol features and the query results through comparative analysis, generates feature update proposals, and assists in generating protocol feature fingerprints;

[0017] The newly generated and verified feature fingerprint is automatically written into the protocol feature library, and the newly added or changed feature fingerprint is extracted based on the updated protocol feature library to generate an incremental training data set, and the online learning algorithm is used to update the weight of the classifier model.

[0018] Preferably, in the communication data semantic recognition step, natural language processing technology is used to understand the communication data protocol, identify professional entities in the power Internet of Things, establish contextual semantic associations in combination with the domain knowledge graph, and convert the communication data into an intermediate standard representation, specifically including:

[0019] Dynamically convert the binary message of the communication data into structured text according to the original protocol, and then identify the specific fields and professional entities in the structured text through a migration model based on a general NER model or a BERT model, and obtain the temporal dependency relationship of the entities in the message;

[0020] Construct a knowledge graph in the field of power Internet of Things, extract triples in the form of "entity-relationship-entity" based on the knowledge graph, and supplement them with implicit semantic associations;

[0021] By establishing a unified data model across protocols and adopting regularized field definitions, the original communication data is converted into an intermediate standard representation with a unified format.

[0022] Preferably, in the target communication protocol conversion step, the intelligent mapping model calculates the semantic similarity between the intermediate standard representation and the target protocol, generates cross-protocol mapping rules in combination with knowledge graph constraints, and automatically updates the rule base, including:

[0023] The semantic similarity between the intermediate standard representation and the target protocol format is calculated using a bidirectional long short-term memory network and a Transformer model, a cross-protocol mapping suggestion is generated, logical constraints are supplemented in combination with a domain knowledge graph, and candidate conversion rules are generated;

[0024] Double verification is performed on the generated candidate conversion rules, including automatic verification and rule credibility evaluation, and it is automatically selected whether to join manual review, and the rule library is automatically updated after verification; if the newly generated rule fails, re-mapping is triggered, if re-mapping fails more than a specified number of times, the generated data is discarded or transferred to manual, and an error log is generated.

[0025] Preferably, the method further comprises a security detection step, which is performed before the protocol identification step and continues throughout the entire method execution period, and if an abnormality is detected, the device connection is blocked and an alarm is given, otherwise the protocol identification step is entered to analyze the device communication protocol; the security detection step includes one or more of traffic baseline detection and blacklist filtering, wherein:

[0026] Traffic baseline detection: a traffic feature baseline is established by pre-training and analyzing the device traffic characteristics in the power Internet of Things, and if the traffic data of the access device exceeds the preset threshold range, it is determined to be abnormal;

[0027] Blacklist filtering: according to the blacklist database provided and maintained by the power Internet of Things server, including one or more of malicious IP addresses, MAC addresses, and device identifiers, the known attack source device is directly blocked; the blacklist database supports user-defined blacklist rules and supports automatic updating of the blacklist from external security intelligence sources.

[0028] Preferably, the method further comprises an online user feedback step, which automatically collects and processes online user feedback, uses an incentive mechanism to improve the enthusiasm and quality of user feedback, and guides the optimization of the domain knowledge graph and the rule library according to the feedback processing results.

[0029] In a second aspect, a power Internet of Things device access registration system in a multi-protocol environment includes:

[0030] An original communication protocol identification module: according to the document information or transmission message actively sent by the power Internet of Things device, the original protocol information of the current device is extracted; if the device does not actively send protocol information or the protocol information extraction fails, the key features are extracted from the original communication data, the key features are matched with the protocol feature library, the original protocol information of the current device is identified, if the matching fails, the protocol classifier is used to classify the irregular protocol that is not matched, if the successful classification cannot be performed, the protocol feature library is updated in the online learning module, and the above operation is performed again;

[0031] Communication data semantic recognition module: pre-processes raw communication data to standardize data consistency; uses natural language processing technology to understand communication data protocols, identify professional entities in the power Internet of Things, establishes contextual semantic associations based on domain knowledge graphs, and converts communication data into an intermediate standard representation;

[0032] Target communication protocol conversion module: This module converts the intermediate standard representation into data in the target communication protocol format based on rule base matching. If the current rule base matching fails, the module uses the intelligent mapping model to calculate the semantic similarity between the intermediate standard representation and the target protocol, combines the knowledge graph constraints to generate cross-protocol mapping rules, and automatically updates the rule base.

[0033] Device registration module: Sends the current device information and its original protocol information to the current regional IoT server. The server generates a unique device ID based on the current device information and its original protocol information, performs device identity authentication and security testing, and officially connects the device to the power IoT system.

[0034] In a third aspect, an electronic device comprises: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the method for accessing and registering power Internet of Things devices in a multi-protocol environment as described in the first aspect of the present invention is implemented.

[0035] In a fourth aspect, a computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the method for accessing and registering power Internet of Things devices in a multi-protocol environment as described in the first aspect of the present invention.

[0036] Beneficial effects:

[0037] (1) The present invention realizes intelligent identification and semantic-level conversion of heterogeneous protocols by performing protocol identification, dynamic protocol conversion and device unique ID registration for devices that are first connected to the power Internet of Things, supports incremental learning of unknown protocols, supports real-time communication between devices and platforms, and between devices, and improves the compatibility and adaptability of the power Internet of Things system.

[0038] (2) The present invention realizes accurate identification of device communication protocols, including known protocols, unconventional protocols and unknown protocols. The device protocol is matched using the protocol feature library to realize rapid identification of known protocols and basic conversion with intermediate standard representations. For unconventional protocols that cannot be initially identified, a protocol classifier is introduced to classify unconventional protocols that fail to match the protocol feature library. Online protocol learning is started to learn unknown protocols through external queries and local feature analysis, and the protocol feature library and classifier model are updated. NLP technology is used to perform semantic analysis on communication data, generate intermediate standard representations, realize semantic-level protocol conversion, and convert the intermediate standard representations into the target protocol format to realize cross-protocol communication between devices. The above functions are continuously improved through the online learning mechanism, and the protocol feature library and rule library are continuously updated to improve the system's adaptability to new protocols.

[0039] (3) The present invention ensures the security of the device access process through security detection and prevents malicious devices from accessing. It implements device registration and identity authentication to ensure that the device is legally connected to the power Internet of Things. In combination with the user online feedback mechanism, it optimizes and improves the performance of the protocol conversion process, improves the adaptability and accuracy of the protocol conversion, and continuously optimizes the overall performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Schematic diagram of the overall framework of the power Internet of Things device access registration system.

[0041] Figure 2 Flowchart of the registration method for power Internet of Things devices.

[0042] Figure 3 This is a schematic diagram of the framework of the original communication protocol identification module.

[0043] Figure 4 Schematic diagram of the protocol classifier identification model framework. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings.

[0045] Under the conditions and background of the present invention, the power Internet of Things system needs to access various types of devices from different manufacturers. These devices use different protocols such as Modbus, MQTT, HTTP, XMPP, IEC 60870-5-101 / 104, DL / T698, IEC61850, etc. to communicate. The embodiment of the present invention aims to describe the operation process and details of the power Internet of Things device access registration system in a multi-protocol environment. The goal of the system is to achieve automatic access, protocol conversion and unified management of these devices.

[0046] Reference Figure 1The power Internet of Things device access registration system of the embodiment of the present invention realizes real-time protocol parsing and cross-protocol communication between power Internet of Things devices in a multi-protocol environment, thereby achieving unified management of power Internet of Things devices. The system specifically includes the following functional modules:

[0047] Original communication protocol identification module: performs protocol identification on devices waiting to be connected to the power Internet of Things, extracts protocol features based on the original communication data transmitted from the power Internet of Things device to the system, and uses the protocol feature library to perform protocol matching; if failure occurs, the protocol classifier is used to classify unconventional protocols to obtain the original protocol information of the device; if the current protocol classifier cannot accurately classify, the protocol online learning module is entered, including requesting external queries and analyzing the extracted features of the message, updating the original communication protocol identification module, and executing the protocol identification step again;

[0048] Communication protocol online learning module: used to build a dynamic protocol signature library, mainly used to store the characteristic fingerprints of various protocols, support regular expression matching and protocol signature library updates; as well as external query and protocol signature analysis for unknown protocols, support updating the protocol signature library, and perform online learning on the protocol classifier;

[0049] Communication data semantic analysis module: pre-processes the original communication data to standardize data consistency; uses natural language processing (NLP) technology to understand the communication data protocol, establishes semantic associations based on domain knowledge graphs, and converts the communication data into an intermediate standard representation;

[0050] Target communication protocol conversion module: Builds a dynamic rule generation unit to automatically create protocol conversion logic based on NLP semantic analysis results and NLP intelligent mapping models, generates new rules, and automatically updates the rule base; converts the intermediate standard representation into data in the target communication protocol format based on rule base matching;

[0051] Device registration module: After the power IoT device passes security testing and completes protocol conversion, the system sends the newly added device information, including its original communication protocol version, to the IoT server in the current region. The server generates a unique device ID based on the newly added device information and its original communication protocol version, and stores the newly added device information and its unique ID in the server's device database, completing identity authentication at the same time.

[0052] User online feedback mechanism: The target device / service platform receives encapsulated target protocol communication data and voluntarily gives feedback to support online optimization of system functions, domain knowledge graphs, and rule bases.

[0053] As a preferred implementation method, the system not only implements core module functions such as protocol identification, protocol conversion, and device registration, but also includes security support from a security detection module; and the security detection module is executed earlier than the protocol identification module, that is, basic security detection is first performed when the device is first connected, and security tracking is performed, which is implemented throughout the entire execution period of the system.

[0054] The security detection module's basic security checks include traffic baseline detection and blacklist filtering. If any detection anomalies are detected, the device connection is blocked and an alarm is issued. Otherwise, the original communication protocol identification module is used to analyze the device communication protocol. Basic security checks include at least one of the following:

[0055] Traffic baseline detection: Through pre-training and analysis of the traffic characteristics of a wide range of devices in the power Internet of Things, a traffic feature baseline is established. If the traffic data of the access device exceeds the preset threshold range, it is judged as abnormal.

[0056] Blacklist filtering: Based on the blacklist database provided and maintained by the Power IoT server, including malicious IP addresses, MAC addresses, and device identifiers, known attack source devices are directly blocked. The server-provided blacklist database should support user-defined blacklist rules and automatic blacklist updates from external security intelligence sources. The dynamic update mechanism for the blacklist database includes: receiving threat intelligence push from the Power IoT management platform, automatically adding new attack source characteristics to the blacklist; and generating temporary blacklist entries and setting expiration times based on abnormal protocol behavior records reported by the original communication protocol identification module.

[0057] Abnormal handling method: immediately block the abnormal device connection and trigger the alarm mechanism, send the alarm information through the alarm interface, and send a security incident report to the power Internet of Things management platform. The report content shall at least include the abnormal feature summary, device identification and information stamp.

[0058] According to an embodiment of the present invention, the original communication protocol identification module includes:

[0059] Protocol feature extraction unit: When a power IoT device is first connected, it parses the document information or transmission message actively sent by the device and extracts the device identifier, protocol type, and version number. If the device does not actively send protocol information or the protocol information extraction fails, the key protocol features of the device are obtained from the original communication data and the dynamic protocol matching mechanism is activated.

[0060] Protocol feature matching unit: used to instantly match the key features of the original communication data with the protocol features maintained by the system protocol feature library to determine the protocol type; if the protocol match fails, it is handed over to the protocol classifier for processing;

[0061] Protocol Classifier Unit: When the protocol signature library fails to match protocol features, a protocol classifier is used for further identification. This protocol classifier employs a hybrid neural network model combining a one-dimensional convolutional neural network (1D-CNN) and a bidirectional long short-term memory (BiLSTM) network. The 1D-CNN module extracts local spatial features, while the BiLSTM module captures long-range and bidirectional dependencies within the sequence. After the CNN extracts local features, the BiLSTM is responsible for understanding the order and grammatical relationships between these features. The model input is a vectorized representation of the message byte sequence, and the output is a preset probability distribution of conventional protocol types. The classifier determines whether the protocol is unconventional based on the probability distribution threshold. If the classifier successfully identifies the protocol, it proceeds to the communication data semantic recognition module and the target communication protocol conversion module. If the protocol is unconventional, the protocol continues to be processed in the protocol online learning unit.

[0062] According to an embodiment of the present invention, the communication protocol online learning module includes the following submodules:

[0063] Dynamic protocol signature library: used to store static signature fingerprints of multiple communication protocols and their versions, as well as dynamic incremental signatures of unknown protocols. Static signature fingerprints include fixed port numbers, message header identifiers, and regular expression templates for protocol instruction sets. Dynamic incremental signatures continuously update signature fingerprints of unknown protocols through an online learning mechanism. The protocol signature library supports regular expression matching, version updates, and updates for unknown protocols.

[0064] Protocol online learning unit: includes an external query interface and a protocol feature analysis engine. For unconventional protocols that exceed the protocol classifier threshold, it extracts session pattern features such as message timing relationships and field position distribution entropy values. Through external query interfaces and local feature clustering analysis, it generates incremental feature vectors and updates the protocol feature library and classifier model parameters.

[0065] The protocol online learning unit of the present invention includes an external query interface, a protocol feature analysis engine and an automatic update mechanism, wherein:

[0066] The external query interface is used to initiate an external query request to an external data source when encountering an unknown protocol to obtain relevant information about the unknown protocol. The external data source includes at least one of the following:

[0067] Public protocol knowledge bases, such as the RFC document library and the IANA protocol database;

[0068] Third-party protocol parsing service API, such as the Wireshark protocol parsing interface;

[0069] The expert collaboration platform is used to initiate consultation requests to the expert system. It is mainly used to fill the gaps in private protocols and supports manual submission of protocol documents or feature descriptions.

[0070] The protocol feature analysis engine is used to format and parse the protocol feature data obtained from the external query interface. It uses comparative analysis methods such as cosine similarity to determine the differences between unknown protocol features and query results, generate feature update proposals, and assist in generating protocol feature fingerprints.

[0071] The automatic update mechanism is used to automatically write newly generated and verified feature fingerprints into the protocol feature library, extract new or changed feature fingerprints based on the updated protocol feature library, generate incremental training data sets, and use online learning algorithms such as SGD and Adam to update the weights of the classifier model. The update frequency is configurable.

[0072] According to an embodiment of the present invention, the processing flow of the communication data semantic parsing module includes:

[0073] Data preprocessing, which is used to preprocess the original communication data, including noise elimination, coding format correction and data consistency standardization;

[0074] NLP technology implements protocol understanding by first dynamically converting the binary message of communication data into structured text based on the original protocol. It then uses a migration model based on a general NER model or BERT model to identify specific fields and professional entities in the structured text and obtain the temporal dependencies of the entities in the message.

[0075] Establish semantic association units, construct a knowledge graph in the field of power Internet of Things, extract triples (entity-relationship-entity) based on the knowledge graph, and supplement hidden semantic associations;

[0076] Generate an intermediate standard representation: By establishing a unified data model across protocols and adopting regularized field definitions, the original communication data is converted into an intermediate standard representation with a unified format.

[0077] According to an embodiment of the present invention, the target communication protocol conversion module includes:

[0078] The rule matching and conversion unit converts the intermediate standard representation into data in the target communication protocol format according to the rule base matching. If the match is successful, it converts the data into the target protocol format and encapsulates the data. If the match fails, it generates a log and triggers the NLP intelligent mapping unit to dynamically generate rules.

[0079] The NLP intelligent mapping unit is used to dynamically generate rules. It can use NLP technologies such as the BiLSTM+Transformer model to calculate the semantic similarity between the intermediate standard representation and the target protocol format, generate cross-protocol mapping suggestions, and combine domain knowledge graphs to supplement logical constraints and generate candidate conversion rules.

[0080] The rule base update mechanism performs double verification on the newly generated rules of the NLP intelligent mapping unit, including automated verification and rule credibility assessment. It allows you to choose whether to add a manual review module. After verification, the rule base is automatically updated. If the newly generated rule fails, remapping is triggered and a remapping threshold is set. If the threshold is exceeded and the result is not successful, the generated data is discarded or transferred to manual review, and an error log is generated.

[0081] According to an embodiment of the present invention, the online user feedback module supports user feedback for system optimization, ensuring automated feedback processes and effective processing of feedback data. The module designs incentive mechanisms and sets reward rules under expert guidance to ensure the enthusiasm and quality of user feedback. It also designs automated processing procedures to ensure timely submission of feedback information to the server for effective processing, and automatically optimizes domain knowledge graphs and rule bases based on prompt guidance.

[0082] Reference Figure 2 The method for accessing and registering a power Internet of Things device provided by an embodiment of the present invention is completed based on the above-mentioned access and registration system. The specific process steps are as follows:

[0083] S1: When a new device attempts to access the system, the security detection module is first triggered;

[0084] Step 1.1: The security detection module performs a flow baseline test to check whether the flow data of the device exceeds the preset threshold range; for example: assuming that the access device is a smart meter device, taking model: DTSD341-MC3, protocol: DL / T645-2020 as an example, the system pre-trained smart meter flow baseline is 2-5KB / s. If the current initial flow is 3KB / s, it meets the threshold range. If the current flow is 6KB / s, the data flow is too high and is judged to be abnormal; assuming that the access device is a temperature and humidity sensor, taking a custom industrial protocol as an example, the system pre-trained area device flow baseline is 5-15KB / s. If the current sensor initial flow is 8KB / s, it meets the threshold range, otherwise it is judged to be abnormal;

[0085] Step 1.2: The security detection module also performs blacklist filtering, checking whether the device's IP address, MAC address, and device identifier are in the blacklist database. If the device's MAC address (e.g., 00:1A:2B:3C:4D) and IP address (e.g., 192.168.1.100) are not in the blacklist, and the digital certificate is verified by the power industry's dedicated CA center, the device is considered normal.

[0086] Step 1.3: If no abnormalities are found in the baseline detection and blacklist filtering of the access device traffic, the device passes the basic security detection and enters the protocol identification process; if an abnormality is found in any detection, the system immediately blocks the connection of the device and triggers the alarm mechanism. The alarm information is sent through the alarm interface, and a security incident report is sent to the power Internet of Things management platform. The report content includes at least a summary of the abnormal characteristics, device identification and timestamp.

[0087] S2: After the device passes the security test, it enters the original communication protocol identification module.

[0088] Reference Figure 3 ,The processing flow of the original communication protocol identification module includes the following steps:

[0089] Step 2.1: Enable the protocol feature extraction unit. If the device actively sends data containing protocol information, extract its key feature fields. If the device does not actively send protocol documents or version information (such as company-specific protocols or custom protocols), the protocol feature extraction unit extracts key features from the original message. For example, if the key features of the original communication message of the above-mentioned temperature and humidity sensor are extracted, "fixed port number: 502, byte sequence features: 0x02 0x01 0x00 0x0A 0x12 0x34 0x03" are obtained, where the first byte 0x02 is the custom device ID, 0x01 is the operation code (read), 0x00 0x0A is the register address (decimal 10), 0x12 0x34 is 16-bit data (decimal 4660), and the last byte 0x03 is the checksum.

[0090] Step 2.2: Match the extracted protocol features with the data in the protocol feature library. The protocol feature library already stores static feature fingerprints for some common protocols (such as Modbus, MQTT, and DL / T 645-2020), and matching is performed accordingly. For example, if the protocol of the smart meter device is matched, the feature fingerprint of DL / T 645-2020 in the protocol feature library is compared. If a match is successful, it is directly identified as a known protocol without triggering the classifier. If the protocol of the temperature and humidity sensor is matched, the Modbus RTU feature in the protocol feature library is compared based on similarity (the message header should be 0x01, and the function code range is 0x01-0x0F). If a mismatch is found, the protocol classifier is triggered.

[0091] Step 2.3: The protocol classifier classifies unconventional protocols and unknown protocols, inputs the message byte sequence into a vectorized representation into the 1D-CNN+BiLSTM model, and outputs the protocol type probability distribution; for example, taking the partial message byte sequence "0x02 0x01 0x00 0x0A0x12 0x34 0x03" of the above-mentioned temperature and humidity sensor as an example, it is vectorized and encoded into a fixed-length vector representation with a dimension of [seq_len, embedding_dim], which serves as the initial input of the ID-CNN module; the model structure is as follows Figure 4 As shown, the 1D-CNN module extracts local byte pattern features through sliding convolution kernels over multiple layers. The convolved local feature sequence undergoes activation and pooling before being fed into a bidirectional long short-term memory (BiLSTM) network. The BiLSTM models the feature sequence in both the forward and backward directions. The forward LSTM unit captures the positive dependencies between features, while the backward LSTM unit captures the negative dependencies between features. This contextual relationship helps to more accurately understand the semantic structure of the protocol. The BiLSTM outputs a bidirectional feature sequence, which is the concatenation of the forward and backward LSTM units. This is processed by an attention mechanism module and finally passes through a fully connected network, coupled with a softmax function, to output a probability distribution vector corresponding to multiple pre-set common protocol types. For example, the protocol type distribution probability output by the temperature and humidity sensor is "Modbus RTU: 35%, BACnet: 15%, IEC104: 12%, Other: 38%." This is below the preset probability threshold, making it an unknown protocol, and the protocol online learning module is initiated.

[0092] S3: Start the communication protocol online learning module for the unknown protocol to learn the characteristics of the custom protocol. The specific steps are as follows:

[0093] Step 3.1: Use the local protocol feature analysis engine to extract session pattern features from the temperature and humidity sensor device messages, such as message timing relationships and field position distribution entropy, and perform external queries and local clustering analysis. External queries are used to obtain data related to the requested message session pattern features, and local comparative analysis is performed, using methods such as similarity calculation or clustering to determine the best results for devices with unconventional protocols. Similarity, which considers the sample and the protocol information provided by each external query, calculates similarities and compares the values, is a basic classification method. Clustering, which considers both intra-class similarity and inter-class distance, using various calculation methods to bring similar classes closer together and different classes farther apart, is a more preferred feature classification method.

[0094] Step 3.2: Initiate a query request through an external query interface to external query services supported by the system, such as public protocol knowledge bases (such as RFC document libraries) and third-party protocol parsing service APIs (such as Wireshark interfaces). Query the message characteristics of the communication data of the above-mentioned devices. For example, query the IANA database for the registered protocol for port 502, which returns Modbus-related information but does not match the custom field. Call the Wireshark API to parse the data stream, and the prompt "Unknown protocol, manual analysis recommended" will be displayed. Further request is made to the expert platform to upload the original message header and protocol characteristics to obtain suggestions.

[0095] Step 3.3: Perform local feature analysis based on the external query results, generate new feature vectors based on the original features, and automatically update the protocol feature library;

[0096] Step 3.4: The automatic update mechanism writes the newly generated feature fingerprint into the protocol feature library and generates an incremental training data set based on the updated feature library. It uses an online learning algorithm (such as the Adam algorithm) to update the weights of the protocol classifier model. The update frequency can be automatically adjusted according to the system configuration. After the update is completed, the system executes the protocol identification step again and successfully identifies the custom protocol of the device.

[0097] S4: After successfully identifying the access device protocol type, perform communication data semantic analysis.

[0098] Step 4.1: Data preprocessing: Preprocess the raw communication data sent by the device, including noise elimination, encoding format correction, and data consistency standardization. For example, for the communication data of the smart meter device mentioned above, filter out abnormal bytes generated by power line interference in carrier communication, convert the GB / T 1988 character set data sent by the meter into UTF-8, standardize the electrical energy unit to "kWh", and convert "0.001kWh" in the raw data into a standard numerical format.

[0099] Step 4.2: Use NLP technology to convert the binary message into structured text; for example, the partial message of the smart meter device is 68 01 02 03 68 11 04 33 12 34CRC, which is converted into the structured text "device_id = 010203, operation = read, unit = 1104, value = 3312.34kWh, check = CRC". In this example, according to the clear field definition in the DL / T645 protocol document, a specific byte (68) is used as the frame start identifier and frame start delimiter (twice 68), followed by the device identification segment (01 02 03), function / operation segment (11 04), data segment (33 12 34), and check segment (CRC). The message example here is a hexadecimal byte representation of a binary stream. The original binary bit stream is a string of 01s, which is not convenient to represent. The machine reads information in bytes, and hexadecimal bytes are consistent with the message characteristics of most industrial equipment. The message shown is a partial structure written in a format similar to the DL / T645 protocol. It is only for exemplary explanation. The bytes between the frame header (68) and the checksum (CRC) have fixed field positions except for the last data segment (n).

[0100] Step 4.3: Use a NER or BERT model to identify specific fields and professional entities, such as "device_id," "unit," and "value." Existing mature pre-trained NER models can perform deep semantic encoding on the entire structured text. Combined with a sequence annotation mechanism, each word / character is labeled with an entity category. After the model outputs a label sequence, entity concatenation and key-value mapping are performed, thereby achieving automatic extraction and semantically structured representation of protocol semantic information. For example, professional entities such as "device number (device_id)," "operation type (operation)," "unit of measurement (unit)," and "data value (value)" in the message can be identified as described above. Furthermore, semantic associations can be established by combining knowledge graphs in the power Internet of Things domain. For example, "device ID" can be associated with "smart meter," and "data type: electricity" can be associated with "metering data."

[0101] Step 4.4: Generate an intermediate standard representation of a regularized and standardized template based on the above-mentioned understanding of protocol semantics. The intermediate standard representation is a universal structured template whose purpose is to map the irregular message semantics of various protocols into a universal format within the system, supporting semantic alignment and conversion between different protocols. It can be designed according to enterprise needs and industry standards. The following intermediate standard representation template written by the present embodiment is only used as an example and is not intended to be limiting: { "device_info":{"device_id":"CN-WS-DTSD341-20250603", "device_type":"Smart Meter","manufacturer":"X Group"}, "data_info":{"timestamp":1685355600,"operation":"read","entity":{"name":"active_energy_total","value":1234.56,"unit":"kWh"},"protocol_metadata":{"protocol_type":"DL / T 645-2020","version":

[0102] "V1.0"}}}

[0103] S5: Target communication protocol conversion module, which receives the intermediate standard representation data and converts it into the target protocol format.

[0104] Step 5.1: The rule matching and conversion unit searches the rule base for a rule that matches the target protocol. The target rule can be the unified communication protocol specified by the system server or the communication protocol of the target IoT system / terminal. The rule is used to convert the common fields and semantic information in the intermediate standard representation into a data structure or byte message in the specific target protocol format. The system maintains a protocol mapping rule base that includes field naming specifications, byte formats, encoding methods, and other content for different protocols. The field-level mapping is as described above: "ISR.device_id = 010203" → "target protocol.dev_id = 0x01 0x02 0x03" for the smart meter device. The target protocol format combination rules are followed to splice the fields according to the target protocol message format. If a rule for converting the intermediate standard representation (ISR) to the target protocol format already exists in the rule base, the data is directly converted to the target protocol format according to the rule. If there is no direct matching rule in the rule base, NLP intelligent mapping is triggered.

[0105] Step 5.2: NLP intelligent mapping can use the BiLSTM+Transformer model to calculate the semantic similarity between the ISR and the target protocol, and generate cross-protocol mapping rules in combination with knowledge graph constraints (such as unit constraints); its processing flow is semantic matching and structural mapping. First, each field in the ISR (such as "device_id, unit, value, timestamp") and the fields in the target protocol description are embedded and encoded respectively. The encoding uses BiLSTM to extract the contextual semantic representation of each field. The field vector output by BiLSTM will be sent to the Transformer model. The global dependency relationship between the ISR and target protocol fields is modeled through the multi-head attention mechanism, and deeper semantic alignment features are learned to identify the most likely semantic mapping relationship; the semantic similarity score is calculated for each ISR field and the target protocol field to form a candidate "field pair mapping table", and the matching field pairs are automatically generated into mapping rule entries under the attribute semantics and type constraints provided by the domain knowledge graph.

[0106] Taking the self-written ISR in this embodiment as an example for intelligent mapping and conversion to the MQTT v3.1.1 target protocol, the specific mapping rules are independently written and set by experts or engineers based on specific scenarios and are not based on the standards in this embodiment. In addition, all conversion rules in this embodiment are briefly described for understanding only and are not intended to be limiting. For example, device_info.device_id is mapped to the {device_id} parameter in the MQTT topic, active_energy_total→

[0107] The data_type field of the MQTT payload, value→the value field of the payload, the data type is Float32 (retain two decimal places), the unit→the unit field of the payload, the value is kWh, the operation type operation in the read→the operation field of the payload, the value is get (in line with the MQTT subscription / publish semantics), etc. The detailed rule generation content is not repeated here.

[0108] Step 5.3: The rule base is automatically updated, and the newly generated rules are automatically stored after passing the dual verification of automated verification (verification of successful protocol conversion is a verification mechanism automatically completed by the program, which is mainly reflected in whether the protocol conversion is successful, including semantic consistency verification before and after protocol conversion, protocol structure compatibility verification, backtesting verification (running verification on real historical instances)) and credibility assessment (the output probability of the industry rule credibility model is >90%); if the newly generated rules fail the dual verification, they will be re-intelligently mapped with the custom number of repetitions as the threshold, and an error report will be generated.

[0109] Step 5.4: Encapsulate the original communication data and the original communication protocol type into the message format of the target protocol and send it to the target server platform.

[0110] S6: Protocol registration module. Once device protocol conversion is complete, the system sends the newly added device information and its original communication protocol version to the regional IoT server. The server generates a unique ID based on the device information and stores it in the device database. This completes device identity authentication and security testing, officially connecting the device to the power IoT system. If the device is already registered, communication can proceed directly.

[0111] S7: The user feedback mechanism is used to optimize the system. For example, when operation and maintenance personnel view data through the management platform, they provide feedback on the data accuracy of the device and the system performance of the encapsulated target protocol communication data received by the target server platform. If the data is accurate and the system performs well, it is manually marked as "conversion accurate";

[0112] The system's feedback mechanism can also be used to generate feedback data as supplementary data for optimizing the system model. The supplementary data is used to optimize the domain knowledge graph, rule base and NLP model parameters, and supports real-time online updates.

[0113] This embodiment demonstrates that the power Internet of Things device access and registration system in a multi-protocol environment described in the present invention can realize automatic access, protocol conversion and unified management functions of devices, aiming to solve the device interconnection barriers caused by the diversity of access device protocols in the power Internet of Things, as well as the problems caused by protocol complexity and non-standardization.

[0114] Another embodiment of the present invention provides an electronic device, comprising: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the steps of the method for accessing and registering power Internet of Things devices in a multi-protocol environment as described above are implemented.

[0115] Another embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method for accessing and registering a power Internet of Things device in a multi-protocol environment as described above are implemented.

[0116] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, devices (systems), electronic devices, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0117] The present invention is described with reference to flowcharts of methods according to embodiments of the present invention. It should be understood that each process in the flowcharts and combinations of processes in the flowcharts can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts. Figure 1 A device that specifies functions in a process or multiple processes.

[0118] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A function specified in a process or multiple processes.

[0119] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 The steps of a specified function in a process or multiple processes.

Claims

1. A method for accessing and registering power Internet of Things devices in a multi-protocol environment, characterized in that: The method comprises: Original communication protocol identification step: Extract the original protocol information of the current device based on the document information or transmission message actively sent by the power Internet of Things device; if the device does not actively send protocol information or the protocol information extraction fails, extract key features from the original communication data, match the key features with the protocol feature library, and identify the original protocol information of the current device. If the match fails, use the protocol classifier to classify the unmatched unconventional protocols. If the classification is unsuccessful, update the protocol feature library through online learning and execute the original communication protocol identification step again; Communication data semantic recognition steps: Preprocess the original communication data to standardize data consistency; use natural language processing technology to understand the communication data protocol, identify professional entities in the power Internet of Things, combine domain knowledge graphs to establish contextual semantic associations, and convert the communication data into an intermediate standard representation; Target communication protocol conversion step: Based on rule base matching, the intermediate standard representation is converted into data in the target communication protocol format. If the current rule base matching is unsuccessful, the intelligent mapping model is used to calculate the semantic similarity between the intermediate standard representation and the target protocol, and the cross-protocol mapping rules are generated by combining the knowledge graph constraints, and the rule base is automatically updated. Device registration steps: Send the current device information and its original protocol information to the current regional IoT server. The server will generate a unique device ID based on the current device information and its original protocol information, perform device identity authentication and security testing, and officially connect the device to the power IoT system.

2. The method according to claim 1, characterized in that In the original communication protocol identification step, the protocol classifier adopts a hybrid neural network model including a one-dimensional convolutional neural network and a bidirectional long short-term memory network, wherein the one-dimensional convolutional neural network extracts local spatial features, and the bidirectional long short-term memory network captures the order and grammatical relationship between these features; the input of the hybrid neural network model is a vectorized representation of the message byte sequence, and the output is a preset conventional protocol type probability distribution, and whether it is an unconventional protocol is determined according to the probability distribution threshold; if the classifier successfully identifies, it enters the communication data semantic recognition and target communication protocol conversion step; if it is an unconventional protocol, it enters the online learning step.

3. The method according to claim 2, characterized in that The online learning steps include: The protocol feature analysis engine extracts session pattern features from unconventional protocols that exceed the protocol classifier threshold, including one or more of the following: message timing relationships and field position distribution entropy values. Based on the session pattern features, an incremental feature vector is generated through an external query interface and local cluster analysis to update the protocol feature library and classifier model parameters. The protocol feature library stores static feature fingerprints of multiple communication protocols and their various versions and dynamic incremental features of unknown protocols, where the static feature fingerprints include one or more of the fixed port number, message header identifier and regular expression template of the protocol instruction set, and the dynamic incremental features continuously update the feature fingerprints of unknown protocols through an online learning mechanism; the protocol feature library supports regular expression matching, version update, and unknown protocol update.

4. The method according to claim 3, characterized in that The external query interface is used to initiate an external query request to an external data source when encountering an unknown protocol to obtain relevant information of the unknown protocol, wherein the external data source includes at least one of the following: a public protocol knowledge base, a third-party protocol analysis service API, and an expert collaboration platform; The protocol feature analysis engine formats and parses the protocol feature data obtained from the external query interface, determines the difference between the unknown protocol features and the query results through comparative analysis, generates feature update proposals, and assists in generating protocol feature fingerprints; The newly generated and verified feature fingerprint is automatically written into the protocol feature library, and the newly added or changed feature fingerprint is extracted based on the updated protocol feature library to generate an incremental training data set, and the online learning algorithm is used to update the weight of the classifier model.

5. The method according to claim 1, wherein In the communication data semantic recognition step, natural language processing technology is used to understand the communication data protocol, identify professional entities in the power Internet of Things, establish contextual semantic associations based on the domain knowledge graph, and convert the communication data into an intermediate standard representation. Specifically, the following steps are involved: Dynamically convert the binary message of the communication data into structured text according to the original protocol, and then identify the specific fields and professional entities in the structured text through a migration model based on a general NER model or a BERT model, and obtain the temporal dependency relationship of the entities in the message; Construct a knowledge graph for the power Internet of Things (IoT) domain, extract triples in the form of "entity-relationship-entity" based on the knowledge graph, and supplement them with implicit semantic associations; By establishing a unified data model across protocols and adopting regularized field definitions, the original communication data is converted into an intermediate standard representation with a unified format.

6. The method according to claim 1, characterized in that In the target communication protocol conversion step, the intelligent mapping model calculates the semantic similarity between the intermediate standard representation and the target protocol, generates cross-protocol mapping rules based on knowledge graph constraints, and automatically updates the rule base, including: Using a bidirectional long short-term memory network and a Transformer model, the semantic similarity between the intermediate standard representation and the target protocol format is calculated to generate cross-protocol mapping suggestions. This is supplemented with logical constraints using domain knowledge graphs to generate candidate conversion rules. Perform double verification on the generated candidate conversion rules, including automated verification and rule credibility assessment. You can choose whether to add manual review. After verification, the rule base is automatically updated. If the newly generated rule fails, remapping is triggered. If remapping fails after exceeding the specified number of times, the generated data is discarded or transferred to manual review, and an error log is generated.

7. The method according to claim 1, characterized in that The method further includes a security detection step, which is executed before the protocol identification step and is continuously implemented throughout the execution of the method. If a detection anomaly occurs, the device connection is blocked and an alarm is issued. Otherwise, the protocol identification step is entered to parse the device communication protocol. The security detection step includes one or more of traffic baseline detection and blacklist filtering, wherein: Traffic baseline detection: Through pre-training and analysis of the traffic characteristics of devices in the power Internet of Things, a traffic feature baseline is established. If the traffic data of the connected device exceeds the preset threshold range, it is determined to be abnormal; Blacklist filtering: Based on the blacklist database provided and maintained by the power IoT server, including one or more of malicious IP addresses, MAC addresses, and device identifiers, devices with known attack sources are directly blocked. The blacklist database supports user-defined blacklist rules and supports automatic blacklist updates from external security intelligence sources.

8. The method according to claim 1, characterized in that The method also includes an online user feedback step, which automatically collects and processes user online feedback, uses an incentive mechanism to improve the enthusiasm and quality of user feedback, and guides the optimization of the domain knowledge graph and rule base based on the feedback processing results.

9. A power Internet of Things device access registration system in a multi-protocol environment, characterized by: The system comprises: Original communication protocol identification module: Extracts the original protocol information of the current device based on the document information or transmission message actively sent by the power Internet of Things device. If the device does not actively send protocol information or the protocol information extraction fails, it extracts key features from the original communication data, matches the key features with the protocol feature library, and identifies the original protocol information of the current device. If the match fails, the protocol classifier is used to classify the unmatched unconventional protocols. If the classification is unsuccessful, the online learning module is entered to update the protocol feature library and perform the above operations again. Communication data semantic recognition module: pre-processes raw communication data to standardize data consistency; uses natural language processing technology to understand communication data protocols, identify professional entities in the power Internet of Things, establishes contextual semantic associations based on domain knowledge graphs, and converts communication data into an intermediate standard representation; Target communication protocol conversion module: This module converts the intermediate standard representation into data in the target communication protocol format based on rule base matching. If the current rule base matching fails, the module uses the intelligent mapping model to calculate the semantic similarity between the intermediate standard representation and the target protocol, combines the knowledge graph constraints to generate cross-protocol mapping rules, and automatically updates the rule base. Device registration module: Sends the current device information and its original protocol information to the current regional IoT server. The server generates a unique device ID based on the current device information and its original protocol information, performs device identity authentication and security testing, and officially connects the device to the power IoT system.

10. An electronic device, characterized in that: include: one or more processors; Memory; And one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and when the programs are executed by the processors, the power Internet of Things device access registration method in a multi-protocol environment according to any one of claims 1-8 is implemented.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for accessing and registering a power Internet of Things device in a multi-protocol environment according to any one of claims 1 to 8 is implemented.

Images