A power grid industry data processing method and related device

Abstract

This application discloses a data processing method and related equipment for the power grid industry, applicable to the fields of power grid information security and data processing technology. This application uses a preset intelligent data recognition model to perform semantic parsing and power grid topology correlation sensitivity analysis on multi-source heterogeneous power grid data to obtain the real-time data security level. The real-time data security level is then matched with the corresponding target encryption strength in a preset quantum encryption strategy library. A dynamic quantum key is jointly generated based on the real-time data security level and quantum true random data. The multi-source heterogeneous power grid data is then encrypted using the dynamic quantum key and the target encryption strength to generate ciphertext data, which is transmitted via a quantum channel at the transmitting end. When the receiving end decrypts the ciphertext data, it performs online incremental updates to the preset intelligent data recognition model based on the decryption result. This effectively improves the accuracy of the preset intelligent data recognition model in classifying and classifying data and enhances the security of data transmission.

Description

Technical Field

[0001] This application relates to the field of power grid information security and data processing technology, and in particular to a data processing method and related equipment for the power grid industry. Background Technology

[0002] In related technologies, with the rapid advancement of smart grid construction, the amount of data generated by the power grid industry is growing exponentially. Among them, core power grid data is directly related to the safe and stable operation of power infrastructure and the security of user privacy, and is the core object of power network security protection. Currently, existing power grid data security processing methods suffer from static and fixed data classification and grading with low accuracy. The separation between data classification and encryption mechanisms makes it impossible to simultaneously ensure transmission security and real-time performance. Furthermore, the plug-in cryptography system is vulnerable to quantum cracking, resulting in low data transmission security. At the same time, the security protection system lacks adaptive iteration capabilities, leading to a continuous decline in protection capabilities.

[0003] In summary, the technical problems existing in the relevant technologies need to be improved. Summary of the Invention

[0004] The main objective of this application is to propose a data processing method and related equipment for the power grid industry, which can effectively improve the accuracy of data classification and the security of data transmission.

[0005] To achieve the above objectives, one aspect of this application proposes a data processing method for the power grid industry, the method comprising the following steps: Acquire multi-source heterogeneous power grid data within the target area, wherein the multi-source heterogeneous power grid data includes structured data, semi-structured data, and unstructured data; By performing semantic parsing and power grid topology correlation sensitivity analysis on the multi-source heterogeneous power grid data through a preset intelligent data recognition model, the real-time security level of the data is obtained. The real-time security level of the data is matched with the corresponding encryption strength in the preset quantum encryption strategy library as the target encryption strength, and a dynamic quantum key is jointly generated based on the real-time security level of the data and the quantum true random data. After encrypting the multi-source heterogeneous power grid data according to the dynamic quantum key and the target encryption strength to generate ciphertext data, the ciphertext data is transmitted through a quantum channel via the sending end. After the receiving end decrypts the encrypted data, it associates the decrypted data with the real-time security level of the data to form a decryption log. The decryption log includes a data identifier, a timestamp, the real-time security level of the data, a decryption success indicator or a decryption failure indicator, decryption time, and a quantum bit error rate. A driving signal is generated based on the decryption log to drive the preset intelligent data recognition model to perform online incremental updates.

[0006] In some embodiments, the preset intelligent data recognition model includes an input layer, a semantic embedding layer, a power grid topology layer, and a sensitivity scoring layer; The input layer is used to perform unified vector encoding on the multi-source heterogeneous power grid data to obtain initial data features; The semantic embedding layer is used to perform semantic feature mining on the initial data features using a pre-trained language model, and to capture long-distance semantic dependencies in the initial data features to obtain semantically enhanced features; The power grid topology layer is used to aggregate the association features of adjacent nodes based on the topological relationships of power grid equipment within the target area to obtain topology enhancement features; The sensitivity scoring layer is used to fuse the semantic enhancement features and the topology enhancement features, and then output the probability distribution of each security level through a fully connected classification network. Based on the probability distribution, the real-time security level of the multi-source heterogeneous power grid data is determined.

[0007] In some embodiments, the power grid topology layer adopts a three-layer graph attention network overlay structure; the aggregation of adjacent node association features based on the topological relationships of power grid equipment within the target area to obtain topology enhancement features includes: Each layer of the graph attention network is controlled to calculate the attention coefficients between adjacent nodes in the power grid equipment topology through a multi-head attention mechanism; The attention coefficients of all the adjacent nodes are normalized to obtain normalized attention weights; The topology enhancement feature is obtained by weighting and aggregating the initial data features of all nodes based on the normalized attention weights.

[0008] In some embodiments, the training process of the preset intelligent data recognition model includes the following steps: A labeled dataset is constructed as a training sample set, which includes power grid topology, historical power grid operation data, and the actual safety level labels corresponding to the historical operation data. The training sample set is input into the preset intelligent data recognition model to obtain the training prediction security level probability of each historical power grid operation data and the training topology enhancement features of two adjacent nodes in the power grid topology. The sensitivity score loss is calculated based on the actual security level label and the trained predicted security level probability, and the topology consistency loss is calculated based on the trained topology enhancement features. Calculate the composite total loss based on the sensitivity scoring loss and the topology consistency loss; The model parameters of the preset intelligent data recognition model are iteratively updated based on the backpropagation of the composite total loss.

[0009] In some embodiments, the step of matching the real-time data security level with the corresponding encryption strength in a preset quantum encryption strategy library as the target encryption strength, and jointly generating a dynamic quantum key based on the real-time data security level and quantum true random data, includes: The real-time security level of the data is mapped to parameters of the quantum key distribution protocol. The quantum key distribution protocol parameters are matched with the corresponding encryption strength in a preset quantum encryption strategy library as the target encryption strength; Dynamic quantum keys are jointly generated based on the parameters of the quantum key distribution protocol and quantum true random data. The quantum true random data includes true random binary sequences randomly extracted based on vacuum fluctuations or photon arrival times in quantum optics processes.

[0010] In some embodiments, the update frequency of the dynamic quantum key is adaptively controlled by the parameters of the quantum key distribution protocol; When the parameters of the quantum key distribution protocol are greater than or equal to a preset high-sensitivity threshold, the dynamic quantum key is updated by generating a new quantum key independently in a single data transmission. When the parameters of the quantum key distribution protocol are less than or equal to a preset low-sensitivity threshold, the same quantum key is reused within a preset time window; When the quantum key distribution protocol parameters are greater than the preset low sensitivity threshold and less than the preset high sensitivity threshold, the quantum key is periodically updated based on the number of transmission batches or timestamps.

[0011] In some embodiments, generating a drive signal based on the decryption log to drive the preset intelligent data recognition model to perform online incremental updates includes: The decryption log is used to calculate the decryption failure rate of consecutive batches in real time. When the failure rate of the continuous batch decryption exceeds the preset failure rate threshold, it is determined that the data distribution corresponding to the decryption log has drifted or suffered a network attack. Obtain the incremental decrypted text; Freeze the parameters of the semantic embedding layer and the power grid topology layer in the preset intelligent data recognition model, and fine-tune the number of layers of the sensitivity scoring layer according to the incremental decrypted text.

[0012] To achieve the above objectives, another aspect of this application provides a data processing apparatus for the power grid industry, the apparatus comprising: The first module is used to acquire multi-source heterogeneous power grid data within the target area, wherein the multi-source heterogeneous power grid data includes structured data, semi-structured data, and unstructured data. The second module is used to perform semantic parsing and power grid topology correlation sensitivity analysis on the multi-source heterogeneous power grid data through a preset intelligent data recognition model to obtain the real-time data security level. The third module is used to match the real-time security level of the data with the corresponding encryption strength in the preset quantum encryption strategy library as the target encryption strength, and to jointly generate a dynamic quantum key based on the real-time security level of the data and the quantum true random data. The fourth module is used to encrypt the multi-source heterogeneous power grid data according to the dynamic quantum key and the target encryption strength to generate ciphertext data, and then transmit the ciphertext data through a quantum channel via the sending end; The fifth module is used to associate the decrypted data with the real-time security level of the data after the receiving end decrypts the encrypted data to form a decryption log. The decryption log includes a data identifier, a timestamp, the real-time security level of the data, a decryption success indicator or a decryption failure indicator, decryption time, and a quantum bit error rate. The sixth module is used to generate a driving signal based on the decryption log to drive the preset intelligent data recognition model to perform online incremental updates.

[0013] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described above.

[0014] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods described above.

[0015] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer program product, including a computer program that, when executed by a processor, implements the aforementioned method.

[0016] The embodiments of this application include at least the following beneficial effects: This application provides a data processing method and related equipment for the power grid industry. This method acquires multi-source heterogeneous power grid data, including structured data, semi-structured data, and unstructured data, within a target area. Then, it uses a preset intelligent data recognition model to perform semantic parsing and power grid topology correlation sensitivity analysis on the multi-source heterogeneous power grid data to obtain the real-time data security level. Next, it matches the real-time data security level with the corresponding encryption strength in a preset quantum encryption strategy library as the target encryption strength. Finally, it generates a dynamic quantum key based on the real-time data security level and quantum true random data, and then uses the dynamic quantum key and the target encryption strength to process the multi-source heterogeneous power grid data. After heterogeneous power grid data is encrypted to generate ciphertext data, it is transmitted through a quantum channel at the transmitting end. When the receiving end decrypts the ciphertext data, it associates the decrypted data with the real-time security level of the data to form a decryption log, which includes a data identifier, timestamp, the real-time security level of the data, a decryption success or failure indicator, decryption time, and quantum bit error rate. Then, a driving signal is generated based on the decryption log to drive the preset intelligent data recognition model to perform online incremental updates. This can effectively improve the accuracy of the preset intelligent data recognition model in classifying and classifying data. Combined with the generation of dynamic quantum keys, the security of data transmission is further improved. Attached Figure Description

[0017] Figure 1 This is a flowchart of the power grid industry data processing method provided in the embodiments of this application; Figure 2 This is a complete implementation flowchart of the power grid industry data processing method provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of the power grid industry data processing device provided in the embodiments of this application; Figure 4 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0019] It is understood that the terms “first,” “second,” etc., used in this application may be used herein to describe various concepts, but unless otherwise stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the words “if,” “when,” or “in response to a determination” as used herein may be interpreted as “when…” or “when…” or “in response to a determination.”

[0020] As used in this application, the terms "at least one", "multiple", "each", "any", etc., "at least one" includes one, two or more, "multiple" includes two or more, "each" refers to each of the corresponding multiples, and "any" refers to any one of the multiples.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0022] In related technologies, existing methods for secure power grid data processing have the following problems: First, the static and fixed data classification and grading system suffers from low accuracy and poor adaptability: Current power grid data classification primarily relies on the Cybersecurity Classified Protection 2.0 standard, employing fixed static rules based on data source and data type—a coarse-grained, passive classification model. Power grid data exhibits highly dynamic sensitivity; the sensitivity of the same type of data varies significantly across different power grid topologies and equipment operating conditions. Traditional static classification mechanisms fail to recognize the relationship between data semantic changes and equipment topology, easily leading to over-protection of low-sensitivity data and lack of protection for high-sensitivity data. This wastes significant computing and communication resources and poses a major risk of data leakage. Second, data classification and encryption mechanisms are disconnected and lack synergy: In the existing technology system, the data classification module and the data encryption module are deployed independently and run serially, forming technology silos. After the system completes static classification, a unified and fixed encryption strategy is generally used to encrypt all data, which cannot dynamically adjust the protection strength according to the real-time sensitivity of the data. Security resources cannot be accurately allocated as needed, and the security and real-time performance of data transmission cannot be balanced. Third, traditional cryptographic systems are vulnerable to quantum-based cracking, making quantum encryption difficult to implement in engineering: Traditional power grid data encryption relies on classical cryptographic algorithms such as AES, SM series national cryptographic standards, RSA, and ECC. Their security boundaries are based on traditional mathematical problems and cannot withstand the rapid cracking attacks of Shor's algorithm in quantum computing, posing a long-term security risk of "storing now, cracking later." Quantum key distribution and quantum secure direct communication technologies possess theoretically unconditional security characteristics and are core technologies for resisting quantum attacks. However, quantum encryption devices are expensive, have limited key generation rates, and high communication latency. Existing technologies lack differentiated quantum protection strategies; full-scale high-strength encryption leads to system overload and excessive latency, while lightweight encryption cannot guarantee the security of core data, making large-scale engineering applications difficult. Fourth, the security protection system lacks adaptive iteration capabilities, and its protection capabilities continue to decline: the operating conditions of the power grid, the characteristics of data distribution, and network attack methods are constantly changing dynamically. The traditional fixed model and fixed encryption strategy protection system cannot be optimized and upgraded autonomously. After long-term operation, the accuracy of classification decreases, the encryption strategy does not match the actual scenario, and the effectiveness of protection continues to decrease.

[0023] In view of this, this application provides a data processing method and related equipment for the power grid industry, which can effectively improve the accuracy of data classification and the security of data transmission.

[0024] The power grid industry data processing method provided in this application relates to the field of power grid information security and data processing technology. This power grid industry data processing method can be applied to a terminal, a server, or software running on a terminal or server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or vehicle terminal, but is not limited to these. The server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network. The software can be an application implementing the power grid industry data processing method, but is not limited to the above forms.

[0025] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0026] It should be noted that in all specific embodiments of this application, when processing data related to user identity or characteristics, such as user information, user behavior data, user historical data, and user location information, user permission or consent is obtained first. Furthermore, the collection, use, and processing of this data comply with relevant laws, regulations, and standards. In addition, when embodiments of this application require access to sensitive personal information of users, separate permission or consent from the user is obtained through pop-ups or redirection to confirmation pages. Only after obtaining the user's separate permission or consent is the necessary user-related data required for the proper functioning of these embodiments acquired.

[0027] The embodiments of this application will be described in detail below with reference to the accompanying drawings: Figure 1 This is an optional flowchart of the power grid industry data processing method provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S110 to S160: Step S110: Obtain multi-source heterogeneous power grid data within the target area, wherein the multi-source heterogeneous power grid data includes structured data, semi-structured data, and unstructured data; Step S120: Perform semantic parsing and power grid topology correlation sensitivity analysis on multi-source heterogeneous power grid data through a preset intelligent data recognition model to obtain the real-time security level of the data; Step S130: Match the real-time security level of the data with the corresponding encryption strength in the preset quantum encryption strategy library as the target encryption strength, and jointly generate a dynamic quantum key based on the real-time security level of the data and the quantum true random data; Step S140: After encrypting the multi-source heterogeneous power grid data according to the dynamic quantum key and the target encryption strength to generate ciphertext data, the ciphertext data is transmitted through the quantum channel by the sending end; Step S150: After the receiving end decrypts the encrypted data, it associates the decrypted data with the real-time security level of the data to form a decryption log. The decryption log includes a data identifier, a timestamp, the real-time security level of the data, a decryption success identifier or a decryption failure identifier, decryption time, and a quantum bit error rate. Step S160: Generate a driving signal based on the decryption log to drive the preset intelligent data recognition model to perform online incremental updates.

[0028] It is understood that the structured data in this embodiment includes telemetry and telecontrol data (such as voltage, current, power, switch status, etc.) and power quality monitoring data (harmonics, frequency, flicker, etc.) collected by SCADA (Supervisory Control and Data Acquisition) systems. This type of data is usually stored in relational databases or time-series databases, with regular data formats and clearly defined fields. The semi-structured data in this embodiment includes IEC61850 / CIM model configuration files (SCL, RDF formats), fault waveform files in XML format, and transient data in COMTRADE format. This type of data has self-describing characteristics but its structure is flexible and variable. The unstructured data in this embodiment includes dispatch voice command text, equipment inspection images (infrared thermal images, visible light photographs), operation and maintenance record documents (Word / PDF format), dispatch log text, etc. This type of data has no fixed format constraints.

[0029] In this embodiment, after obtaining multi-source heterogeneous power grid data, the data will be cleaned for quality. The quality cleaning process includes, but is not limited to, missing value handling (removing measurement points with a missing rate exceeding 30% in telemetry data), outlier detection (identifying and marking outliers using the 3σ principle or interquartile range method), and timestamp alignment (normalizing timestamps from different data sources to millisecond-level precision, using the system clock of the data acquisition and monitoring control system as the primary reference).

[0030] It is understood that the preset intelligent data recognition model in this embodiment is a deep learning model that integrates Transformer semantic embedding and Graph Attention Network (GAT). Specifically, the preset intelligent data recognition model in this embodiment includes an input layer, a semantic embedding layer, a power grid topology layer, and a sensitivity scoring layer. The input layer is used to perform unified vector encoding on the multi-source heterogeneous power grid data to obtain initial data features; the semantic embedding layer is used to perform semantic feature mining on the initial data features using a pre-trained language model and capture long-distance semantic dependencies in the initial data features to obtain semantically enhanced features; the power grid topology layer is used to aggregate the association features of adjacent nodes based on the topological relationships of power grid equipment within the target area to obtain topology enhanced features; the sensitivity scoring layer is used to fuse the semantic and topology enhanced features, and then output the probability distribution of each security level through a fully connected classification network, determining the real-time security level of the multi-source heterogeneous power grid data based on the probability distribution.

[0031] Specifically, in this embodiment, when performing unified vector encoding on multi-source heterogeneous power grid data, the input layer uses a combination of numerical normalization (Min-Max normalization to the [0,1] interval) and discrete feature embedding (EntityEmbedding, which maps high-dimensional discrete variables to low-dimensional dense vectors with an embedding dimension of 128) for structured data. For semi-structured data, a parse tree structure traversal encoding method is used to encode path-value pairs in the XML / JSON tree structure into sequence features. Each node label is mapped to a 256-dimensional vector through word embedding, and the overall representation is output after aggregation by a bidirectional long short-term memory network (BiLSTM). For unstructured data, word segmentation (based on a word segmentation tool for the power industry dictionary, with a dictionary size of approximately 120,000 power industry professional terms) and WordPiece word embedding encoding method are used, with the maximum sequence length truncated to 512 tokens.

[0032] The semantic embedding layer employs a Transformer-based pre-trained language model to extract deep semantic features from the initial data, aiming to mine semantic and numerical statistical features related to the power industry. After capturing long-distance semantic dependencies in the data through a multi-head self-attention mechanism (with 12 heads), it outputs a semantically enhanced feature vector with a dimension of 768. Specifically, the pre-trained language model in this embodiment can be a BERT-based model that has undergone domain-adaptive fine-tuning on a power industry corpus (containing approximately 500 million tokens of professional texts such as power regulations, dispatching procedures, equipment manuals, and fault case libraries).

[0033] The power grid topology association layer adopts a three-layer graph attention network superposition structure. Each layer of the graph attention network calculates the attention coefficients between adjacent nodes in the power grid equipment topology relationship through a multi-head attention mechanism (e.g., each layer can set up 4 attention heads). The attention coefficients... The calculation formula is: ; In the formula, The attention mechanism can be used to learn parameter vectors; The weight matrix is ​​a learnable weight matrix; These are the feature vectors of node u and node v in the current network layer, respectively. This is a vector concatenation operation; The activation function is a linear unit with leakage correction, and the slope in the negative interval is set to 0.2.

[0034] Attention coefficients for all adjacent nodes Perform Softmax normalization to obtain normalized attention weights. Based on the initial data features of all nodes in the neighborhood, weighted and aggregated using normalized weights, the topology enhancement features of node u in the next layer network are obtained. The calculation formula is as follows: ; In the formula, This represents the node attention weights after Softmax normalization; Represents the set of all adjacent nodes of the central node u; This represents the weight matrix specific to the graph attention network at layer l. Represents a non-linear activation function; Indicates the node u-th The output topology-enhanced feature vector.

[0035] The sensitivity scoring layer is used to fuse semantic enhancement features and topology enhancement features. It outputs the probability distribution of each security level through a fully connected classification network, and selects the security level corresponding to the highest probability from the probability distribution as the real-time security level of the multi-source heterogeneous power grid data. The security levels are divided into M levels according to the power data security protection specifications. In this embodiment, M=5, and the five levels from low to high are: Level 1 (public information), Level 2 (internal information), Level 3 (sensitive information), Level 4 (important secrets), and Level 5 (core secrets). Specifically, the fusion strategy of the sensitivity scoring layer in this embodiment is as follows: the semantic enhancement features (768 dimensions) and the topology enhancement features (768 dimensions) are concatenated to obtain a 1536-dimensional fused feature vector, which is then processed by two fully connected network layers. The two fully connected network layers are as follows: the first layer outputs a dimension of 256, uses the ReLU activation function and applies Dropout regularization (dropout rate 0.3); the second layer outputs a dimension of M=5, followed by a Softmax activation function to output the probability distribution.

[0036] It is understood that the training process of the preset intelligent data recognition model in this embodiment includes the following steps: A labeled dataset is constructed as the training sample set, which includes the power grid topology, historical power grid operation data, and the actual safety level labels corresponding to the historical operation data. The training sample set is input into the preset intelligent data recognition model to obtain the training prediction security level probability of each power grid historical operation data and the training topology enhancement features of two adjacent nodes in the power grid topology structure. The sensitivity score loss is calculated based on the real security level label and the predicted security level probability trained, and the topology consistency loss is calculated based on the topology enhancement features trained. Calculate the composite total loss based on the sensitivity score loss and the topology consistency loss; The model parameters of the preset intelligent data recognition model are iteratively updated based on backpropagation of the composite total loss.

[0037] Specifically, the training sample set in this embodiment is formed by collecting historical power grid operation data and manually labeling it with security level tags according to power data security protection specifications. ; In the formula, This is the i-th training sample (containing complete data content and metadata information). Let i be the true security level label corresponding to the i-th training sample. ∈{1,2,3,4,5}, the total number of safety level categories is M, each training sample is a complete power grid data record and its corresponding metadata information, including the data source device ID, acquisition timestamp (accurate to milliseconds), data type identifier (telemetry / remote signaling / message / image / text, etc.), and the voltage level to which it belongs.

[0038] To enhance the generalization ability of the preset intelligent data recognition model, this embodiment also employs an EDA (Easy Data Augmentation) strategy for unstructured text samples, including four operations: synonym replacement (based on an electricity thesaurus), random insertion, random swapping, and random deletion, with an augmentation factor of 3 times; and noise injection (adding Gaussian noise) is used for structured numerical samples. ) and scaling transformation (random scaling factor 0.9) 1.1) Strategy.

[0039] In this embodiment, after obtaining the training sample set, the training sample set is input into a preset intelligent data recognition model. After being correlated through the input layer, semantic embedding layer, and power grid topology layer in the preset intelligent data recognition model, the sensitivity scoring layer outputs a training prediction security level probability vector. ; In the formula, Predict the probability vector for the security level of the i-th sample; The model predicts the probability that the i-th sample belongs to the c-th security level, with a value in the range (0,1), satisfying the normalization constraint. ;T is the vector transpose operator.

[0040] During training, this embodiment constructs a weighted composite total loss function that includes sensitivity scoring loss and topology consistency loss, and iteratively updates the model parameters through backpropagation to achieve model convergence training. Specifically, the Adam optimizer is used during training, and the initial learning rate is set to... (The Transformer backbone adopts a hierarchical learning rate decay strategy with a decay coefficient of 0.95). The batch size is set to 32 or 64 according to the GPU memory capacity. The total number of training rounds is set to 50. An early stopping strategy is adopted (the program terminates when the validation set loss does not decrease for 5 consecutive rounds).

[0041] Specifically, the composite total loss function in this embodiment Loss due to sensitivity score With topology consistency loss The weighted fusion is obtained. Among them, the sensitivity score loss... The multi-class cross-entropy loss is used, and the calculation formula is as follows: ; In the formula, N is the total number of training samples in a single batch; M is the total number of security level categories. This is the true security level label indicator value for the i-th sample belonging to the c-th level. It is 1 when sample i belongs to category c, and 0 otherwise. The training prediction security level probability of the preset intelligent data recognition model to predict that the i-th sample belongs to the c-th security level is taken in the range of (0,1).

[0042] Topology consistency loss The calculation formula is: ; In the formula, Adjacent device nodes in the power grid topology ,node Training topology-enhanced feature vectors; is the Sigmoid activation function; T is the vector transpose operator.

[0043] The physical meaning of the topology consistency loss in this embodiment lies in the fact that data generated by adjacent devices in the power grid topology should have inherent consistency in terms of security sensitivity (for example, monitoring data of feeders connected to the same bus should not show extreme differences in their security levels). By maximizing the inner product likelihood probability of the features of adjacent nodes, the model is forced to learn the sensitivity representation of topology awareness, so that adjacent device nodes on the topology have similar representations in the feature space. This constraint is equivalent to introducing prior knowledge of the power grid topology into the preset intelligent data recognition model, which can effectively reduce the risk of overfitting in data-sparse scenarios and improve the stability of the preset intelligent data recognition model in classifying topology edge nodes.

[0044] Composite total loss function The calculation formula is: ; In the formula, The weighting coefficient for the sensitivity score loss. Let the topology consistency loss weight coefficients satisfy the following conditions: + =1. In a preferred embodiment, α=0.7 and β=0.3. This ratio is determined by grid search and can achieve an optimal balance between classification accuracy (measured by validation set accuracy) and topological consistency (measured by cosine similarity of features of neighboring nodes).

[0045] Specifically, the training process of the preset intelligent data recognition model in this embodiment can be based on approximately 1.2 million labeled samples (approximately 3.6 million samples after data augmentation), and the training time on an NVIDIA A100 (80GB) GPU is approximately 18 hours. The training curve shows that the preset intelligent data recognition model reaches its minimum loss on the validation set in the 32nd round. At this point, the classification accuracy on the test set reaches 96.2%, the macro average F1-score is 0.952, and the precision and recall rates for each level are all higher than 94%. The inference stage uses FP16 half-precision acceleration, with an average classification time of approximately 85ms per data point and a batch throughput of approximately 320 data points / second, meeting the real-time data processing requirements of the scheduling center.

[0046] It is understood that, after obtaining the real-time data security level, this embodiment matches the real-time data security level with the corresponding encryption strength in the preset quantum encryption strategy library as the target encryption strength, and jointly generates a dynamic quantum key based on the real-time data security level and quantum true random data. Specifically, the preset quantum encryption strategy library in this embodiment contains M encryption strength levels (M=5), and each encryption strength level corresponds to a set of quantum key distribution (QKD) protocol parameters. These quantum key distribution protocol parameters include: Quantum bit encoding basis selection strategy: such as the mixing ratio of Z basis (right-angle basis) and X basis (diagonal basis). For high security level, asymmetric basis selection (70% Z basis and 30% X basis) is used to improve key generation efficiency, while symmetric basis selection (50% each) is used for low security level to simplify implementation. Decoy photon pulse intensity: This includes the specific numerical settings for signal state intensity μ, decoy state intensity ν, and vacuum state intensity 00. For high security levels, a multi-intensity decoy state scheme is adopted (intensity number ≥ 3, intensity interval ≥ 0.3) to more accurately estimate the amount of eavesdropped information. Error correction protocol type and parameters: including the selection of low-density parity-check code (LDPC) or polar code and its code rate setting, number of error correction iterations, damping factor of belief propagation algorithm, etc.

[0047] Specifically, the higher the real-time data security level, the higher the matching encryption strength level, and the greater the intensity modulation amplitude of the decoy state pulses during quantum key distribution (i.e., increasing the intensity distinction between the decoy state and the signal state to enhance the detection sensitivity against photon number separation attacks), and the more iterations the error correction protocol undergoes (to reduce the impact of the qubit error rate on the final key consistency). Specifically, at the lowest security level, a simplified BB84 protocol is used, with only one strength level set for the decoy state; at the highest security level, an improved BB84 protocol based on the decoy state or a measurement-device-independent QKD protocol is used, with three or more strength levels set for the decoy state, and the number of error correction iterations increases to three times the baseline level.

[0048] In some embodiments, the configuration process for each level of policy is shown in Table 1 below: Table 1

[0049] The policy matching logic in this embodiment is as follows: the real-time security level c∈{1,2,3,4,5} of the data is directly mapped to the policy library index, and the corresponding QKD parameter set is queried and loaded. When the security level is the lowest level (level 1), a simplified BB84 protocol is used, the decoy state is set with only one strength level (equivalent to the scheme without a decoy state), and the error correction iteration is 10 times; when the security level is the highest level (level 5), a dual-field QKD protocol is used, the decoy state is set with 4 strength levels, the error correction adopts a concatenated scheme of polar code and BCH code, and the number of error correction iterations is increased to 5 times that of the baseline level (10 times), that is, 50 times.

[0050] In this embodiment, the dynamic quantum key is generated jointly by quantum true random numbers and security level encoding. Specifically, the real-time security level of the data output by the intelligent data recognition model is mapped to a level quantization value g (i.e., g=c, taking the value of a positive integer from 1 to 5). Based on the physical parameters of the quantum key distribution device, the modulation parameters for quantum state preparation (including the voltage bias of the phase modulator and the extinction ratio setting of the amplitude modulator, specifically the phase modulation voltage) are dynamically adjusted. The bias adjustment range is ±0.2. (Amplitude modulation extinction ratio set to 20~30dB) to generate dynamic quantum keys corresponding to the security level. .

[0051] Among them, dynamic quantum key The generation logic is as follows: ; In the formula, R is a dynamic quantum key (with a fixed output length of 256 bits, compatible with the AES-256 encryption algorithm); R is a truly random binary sequence output by a quantum true random number generator (with dynamically adjusted length, not less than 256 bits). This sequence is extracted based on the randomness of vacuum fluctuations or photon arrival time in quantum optics processes, and has information-theoretic security. Its randomness has passed all 15 tests of the NISTSP800-22 test suite. This is a real-time security level quantification value, which is a positive integer (1≤g≤5). ∥ is the fixed-length binary code corresponding to the security level quantization value g; ∥ is the binary string concatenation operation; Hash is a quantum collision resistant hash function, preferably SHA-384 or SHA-3 series hash functions, with an output length of 384 bits, and the first 256 bits are truncated as the final key.

[0052] It is understood that the update frequency of the dynamic quantum key in this embodiment is adaptively controlled by the quantum key distribution protocol parameter g. Specifically, a high-sensitivity threshold is set according to the requirements of power data security classification management. Low sensitivity threshold Then, perform the following operations: When the quantum key distribution protocol parameter g is greater than or equal to a preset high-sensitivity threshold The current multi-source heterogeneous power grid data is determined to be in a highly sensitive state. Therefore, a one-time pad mechanism is adopted, where a completely new quantum key is generated independently for each data transmission to update the dynamic quantum key, ensuring absolute key security. The key length is no less than the data length. For long data, a mode combining AES-256-GCM with one-time pad quantum keys is used, ensuring both the one-time use characteristic of the key and encryption efficiency. In this mode, the key is immediately destroyed from the cache after use, leaving no persistent trace.

[0053] When the quantum key distribution protocol parameter g is less than or equal to a preset low-sensitivity threshold If the data is determined to be in a low-sensitivity state, the same quantum key is reused within a preset time window. Specifically, the preset time window can be set to 5 minutes. Multiple data transmissions within the preset time window share the same quantum key to reduce the consumption of quantum true random numbers and key negotiation overhead. Experimental statistics show that this strategy can reduce the consumption of quantum true random numbers by approximately 45% to 60%.

[0054] When the quantum key distribution protocol parameter g is greater than the preset low sensitivity threshold And less than the preset high sensitivity threshold The data is determined to be in a moderately sensitive state, and the quantum key is updated periodically based on the number of transmission batches or timestamps. Specifically, the quantum key is updated every 100 batches of data transmitted or every 60 seconds, whichever comes first. This strategy balances security and efficiency: the 60-second update cycle ensures the limited use of the key while avoiding the overhead of renegotiating for each transmission.

[0055] This embodiment performs adaptive quantum fusion encryption on multi-source heterogeneous power grid data after obtaining the dynamic quantum key. Based on the dynamic quantum key and matched encryption strength parameters, this embodiment can perform quantum key distribution encryption or quantum-secure direct communication encryption on multi-source heterogeneous power grid data, generating ciphertext data and completing quantum channel transmission. Specifically, this embodiment includes the following steps: Step S210: Encryption mode selection strategy; This embodiment automatically selects the encryption mode based on the real-time security level g, the data volume, and the real-time requirements: When using quantum key distribution encryption mode: the sender and receiver first transmit and measure quantum states through a quantum channel, negotiate to generate a shared quantum key, and then use this shared key in conjunction with a classical symmetric encryption algorithm (preferably AES-256-GCM, which supports authentication encryption and can simultaneously guarantee confidentiality and integrity) to encrypt the power grid data. The encrypted data is then transmitted through a classical channel. This mode is suitable for scenarios with large data volumes (>1KB) and relatively relaxed real-time requirements, such as batch telemetry data transmission and fault waveform file transmission.

[0056] When using quantum-secure direct communication encryption mode: the transmitter directly encodes multi-source heterogeneous power grid data into quantum states (using phase encoding or polarization encoding schemes, with each quantum bit encoding 2 bits of information), and transmits it to the receiver through a quantum channel. The receiver can directly measure and acquire the data without prior key negotiation. This mode is suitable for control command data transmission scenarios with extremely high real-time requirements (end-to-end latency <10ms) and small data volume (<1KB), such as circuit breaker opening and closing commands, AVC regulation commands, etc.

[0057] Step S220: Encryption execution process; QKD encryption mode execution flow: The transmitting end prepares decoy state and signal state optical pulse sequences of corresponding strengths according to the matched QKD (quantum key distribution) protocol parameters; The light pulse is transmitted to the receiver via a quantum channel, and the receiver randomly selects a measurement basis for measurement. Both parties publicly disclose the measurement basis information through a classical channel, and select and retain measurement results with consistent basis vectors as the original key; Perform error estimation, error correction (according to the matching error correction protocol and number of iterations), and privacy amplification (the privacy amplification factor is dynamically adjusted according to the estimated amount of eavesdropped information, ranging from 1.2 to 2.0), and generate the final shared key; The data is encrypted using a shared key with AES-256-GCM to generate ciphertext and an authentication tag. The ciphertext, initialization vector (IV, 12 bytes), authentication tag (16 bytes), and data header information (including security level identifier and timestamp) are encapsulated into a data packet and transmitted to the receiving end via a classic channel.

[0058] Execution flow of quantum-safe direct communication mode: The transmitting end groups the data to be transmitted into bytes, and each 2 bits is mapped to one of the four orthogonal phases of the quantum state (0, π / 2, π, 3π / 2). The phase modulator is controlled to load the corresponding phase onto the optical pulse to prepare the encoded quantum state sequence, and the average photon number is controlled between 0.1 and 0.3 to ensure security; The encoded quantum state sequence is transmitted to the receiving end through an optical fiber quantum channel with a pulse interval of 100ns to 200ns. The receiver uses the basis vector information pre-shared with the transmitter to perform zero-difference measurement, recovers the phase value from the measurement result, and demaps to restore the binary data; Perform integrity verification on the demodulated data; if the verification fails, request a retransmission.

[0059] Step S230: Data packet encapsulation and transmission; After encryption, the ciphertext data, along with necessary quantum state measurement basis information and error correction verification information, is encapsulated into a data packet and transmitted to the receiving end via a quantum channel or a classical-quantum hybrid channel. During transmission, the quantum channel is used for quantum state transmission (using dense wavelength division multiplexing technology, with quantum optical signal wavelengths of 1550nm and classical optical signal wavelengths of 1310nm / 1490nm, transmitted via a shared fiber), while the classical channel is used to transmit basis alignment information, error correction information, and ciphertext data.

[0060] The data packet header format is designed as shown in Table 2 below (total header overhead of 20 bytes): Table 2

[0061] Understandably, after receiving the encrypted data, the receiving end performs quantum decryption based on the corresponding authorization information and generates a feedback signal based on the decryption result. This signal drives the preset intelligent data recognition model to perform online incremental updates, achieving adaptive optimization of the security level and encryption strategy. Specifically, after receiving the encrypted data, the receiving end queries the corresponding authorized decryption strategy based on the real-time security level g carried in the data packet header. When using QKD encryption mode: The receiving end uses a local quantum key distribution device to match the same QKD protocol parameters (including basis selection ratio, error correction protocol type, and iteration number) based on the header security level identifier. It then performs quantum state measurements and basis vector comparisons with the sending end to generate a shared key. After that, it decrypts the ciphertext data using AES-256-GCM and verifies the integrity of the authentication tag. If tag verification fails, the data packet is discarded and an exception is recorded.

[0062] When using quantum-secure direct communication encryption mode: the receiver directly performs zero-difference measurement on the quantum state, demodulates the phase value based on the pre-shared basis vector information, and recovers the binary data sequence. During the measurement process, the signal-to-noise ratio and measurement result confidence of each pulse are recorded synchronously. If the confidence is lower than a preset threshold (0.85), the bit is marked as a low-confidence bit, triggering the error correction and retransmission mechanism.

[0063] It is understandable that this embodiment performs closed-loop optimization on the preset intelligent data recognition model after completing the decryption of the encrypted data. Specifically, as shown in the example... Figure 2 As shown, the closed-loop optimization process in this embodiment includes the following steps: Step S310: After the receiving end completes decryption, it associates the content digest (SHA-256 hash value, 32 bytes) of the decrypted data with the security level and records it to form a structured decryption log. The decryption log includes the following fields: data identifier (unique ID of the data packet, UUID format); timestamp (accurate to milliseconds); security level g (1~5); decryption success / failure indicator (Boolean value); decryption time (in milliseconds); quantum bit error rate (QBER) (percentage, recorded only in QKD mode); measurement result confidence level (recorded only in quantum secure direct communication mode); data source device ID; and data length (bytes). Furthermore, the decryption log is stored in JSONLines format on a security log server and retained for 90 days, supporting subsequent audit tracing and anomaly analysis.

[0064] Step S320: Calculate the continuous batch decryption failure rate in real time based on the decryption log; when the continuous batch decryption failure rate exceeds the preset failure rate threshold, determine whether the data distribution corresponding to the decryption log has drifted or suffered a network attack.

[0065] Specifically, this embodiment calculates the decryption failure rate of consecutive batches (100 data entries per batch) in real time. When the failure rate of consecutive batches exceeds a preset threshold (preferably set to 5%), it determines that the data distribution has drifted or has been subjected to a network attack. At this time, a security alarm is triggered (reported to the operation and maintenance management center via SNMPTrap, and a text message is sent to the on-duty personnel), and the batch of data is marked as an abnormal sample and temporarily stored in an isolation area for subsequent model updates and security analysis.

[0066] QBER Anomaly Monitoring: In addition to the decryption failure rate, this embodiment also monitors the sliding window average of QBER (window size 1000 pulses). When the window average exceeds 80% of the theoretical tolerance (e.g., the tolerance of the dual-field QKD protocol is 11%, and the warning threshold is set to 8.8%), an warning is triggered but the service is not interrupted. Maintenance personnel can use this information to perform channel maintenance in advance.

[0067] Step S330: Obtain the incremental decryption text, freeze the parameters of the semantic embedding layer and the power grid topology layer in the preset intelligent data recognition model, and fine-tune the number of layers of the sensitivity scoring layer according to the incremental decryption text.

[0068] Specifically, incremental decryption samples are retrieved, and all learnable parameters of the model's semantic embedding layer (Transformer part) and power grid topology association layer (three-layer GAT) are frozen (requires_grad=False). Only the parameters of the sensitivity scoring layer (fully connected classification layer) are fine-tuned to complete online incremental learning.

[0069] Specifically, successfully decrypted sample data and their corresponding security levels (recorded by the system before encryption and transmitted along with the ciphertext, verified by digital signature as weakly supervised labels, with a label confidence weight set to 0.9) are used as new training samples. The fully connected classification parameters of the sensitivity scoring layer are updated using mini-batch stochastic gradient descent. Incremental learning uses the same optimizer Adam as the initial training, with a learning rate set to 1 / 10 of the initial training learning rate. The batch size is set to 8 or 16, the number of incremental training rounds is set to 3 to 5, and the performance change is evaluated on the cached validation set after each round.

[0070] During online incremental learning, this embodiment also introduces an Elastic Weight Consolidation (EWC) strategy, adding a parameter importance constraint term to the loss function to suppress catastrophic forgetting. Incremental Loss Function for: ; In the formula, This is the EWC regularization coefficient, preferably 100; Model parameters after initial training convergence The corresponding diagonal elements of the Fisher information matrix; For the sensitivity scoring layer One parameter; These are the baseline parameter values ​​frozen after the initial training is completed.

[0071] Specifically, the incremental learning triggering conditions in this embodiment can be set with two triggering mechanisms: Scheduled trigger: Incremental learning is automatically triggered once every 24 hours of operation; Event trigger: Triggered when the decryption failure rate exceeds 3% for three consecutive batches (below the alarm threshold of 5% but remains high), to quickly adapt to data distribution drift.

[0072] After optimizing the incremental learning process, the effectiveness of the incremental learning was verified: after 30 days of continuous system operation, approximately 15,000 valid incremental samples were collected, and incremental learning was triggered 12 times (10 timed triggers and 2 event triggers). Specifically, in this embodiment, after incremental learning, the accuracy of the pre-defined intelligent data recognition model on recent data improved from the initial 96.2% to 97.1%, an increase of 0.9 percentage points. Simultaneously, due to the protection of EWC regularization, the accuracy on the initial test set (old data distribution) only decreased from 96.2% to 95.9%, a decrease of only 0.3 percentage points, indicating that catastrophic forgetting was effectively suppressed. This result verifies that the incremental learning strategy proposed in this embodiment, which involves freezing the underlying network (preserving general feature extraction capabilities) combined with EWC regularization to fine-tune the top-level classifier (adapting to the new distribution), has good stability and effectiveness in continuous learning scenarios.

[0073] Step S340: The encryption strategy is adjusted adaptively.

[0074] In this embodiment, after incremental learning is completed, the classification decision boundary of the sensitivity scoring layer is updated. The system recalculates the security level distribution ratio of the entire dataset and synchronously pushes the updated level distribution data to the quantum key management module. The quantum key management module automatically adjusts the capacity of the quantum key pre-generation buffer pool corresponding to each security level. If the proportion of high-level (g=4, 5) data increases, the number of keys pre-generated corresponding to high security levels will be increased to ensure that keys respond in real time when highly sensitive data arrives, without negotiation waiting delay; if the proportion of low-level (g=1, 2) data increases, the capacity of the pre-generated buffer pool corresponding to low security levels will be reduced to free up quantum true random number resources for the generation of keys of higher priority levels.

[0075] Among them, the capacity of the buffer pool The following formula: ; In the formula, The percentage of data at level g is estimated, and T is the average throughput of key negotiation (transactions / second). The average time (in seconds) for key negotiation at level g. Set to the minimum buffer size (set to 5).

[0076] The dynamic adjustment mechanism in this embodiment, while ensuring the immediate availability of high-level data keys (response latency <20ms), can reduce the overall consumption of quantum true random numbers by approximately 12% compared to the fixed buffer strategy.

[0077] As can be seen from the above, the method of this application embodiment has the following beneficial effects: First, this embodiment constructs a deep learning model that integrates Transformer semantic embedding and graph attention networks. Through the semantic embedding layer, it fully mines the deep semantic information of power data, and through the power grid topology association layer, it introduces equipment topology relationship constraints. This ensures that the determination of data security level is not only based on the data content itself, but also incorporates the data's location sensitivity within the power grid topology and the influence of associated equipment. Compared to traditional rule-based or shallow machine learning-based classification methods, this embodiment can achieve unified modeling and accurate classification of multi-source heterogeneous data, significantly reducing the false positive rate.

[0078] Secondly, this embodiment quantifies the security level into level parameters and automatically matches them with a preset quantum encryption strategy library, achieving refined management of "encryption on demand". High-security-level data is matched with high-strength quantum encryption protocols, while low-security-level data is matched with lightweight quantum encryption protocols, avoiding over-encryption or insufficient protection of all data in a "one-size-fits-all" manner.

[0079] Third, this embodiment introduces quantum true random numbers as the key entropy source and combines them with security level coding to generate dynamic quantum keys through a quantum collision-resistant hash function, enabling the key to possess both information-theoretic randomness and level correlation. The key update frequency adaptively adjusts with the security level, and the one-time pad mechanism in highly sensitive scenarios ensures forward security, thereby effectively resisting store-and-replay attacks and key exhaustion attacks in quantum computing environments, providing future-proof quantum-resistant security protection for power grid data.

[0080] Fourth, this embodiment establishes a complete closed loop of "grading → encryption → decryption → feedback → optimization" through batch statistical analysis of decryption results and online incremental learning. When an abnormally high decryption failure rate or a persistently high QBER is detected, it can automatically identify data distribution drift or potential attacks and trigger rapid fine-tuning of model parameters. Incremental learning adopts a strategy of freezing the underlying general feature extraction network, fine-tuning only the top-level classifier, and supplementing it with EWC regularization constraints. This achieves an effective balance between the absorption of new knowledge and the retention of old knowledge, and can continuously evolve with the accumulation of power grid operation data. Under long-term operation, the generalization ability and robustness of the grading model are continuously enhanced.

[0081] Fifth, this embodiment establishes a direct feedback loop from incremental learning results to key management strategies. Based on the updated level distribution statistics of the sensitivity scoring layer, it dynamically adjusts the capacity of the quantum key pre-generation buffer pool corresponding to each security level. This allows for an additional reduction in the overall consumption of quantum true random numbers compared to a fixed buffer strategy, while ensuring the immediate availability of high-security-level data keys. This further improves the utilization efficiency of quantum encryption resources and reduces operating costs.

[0082] like Figure 3As shown in the illustration, this application also provides a data processing device for the power grid industry, which includes: The first module is used to acquire multi-source heterogeneous power grid data within the target area, including structured data, semi-structured data, and unstructured data. The second module is used to perform semantic parsing and power grid topology correlation sensitivity analysis on multi-source heterogeneous power grid data through a preset intelligent data recognition model to obtain the real-time data security level. The third module is used to match the real-time security level of the data with the corresponding encryption strength in the preset quantum encryption strategy library as the target encryption strength, and to jointly generate a dynamic quantum key based on the real-time security level of the data and the quantum true random data. The fourth module is used to encrypt the multi-source heterogeneous power grid data according to the dynamic quantum key and the target encryption strength to generate ciphertext data, and then transmit the ciphertext data through the quantum channel via the sending end. The fifth module is used to associate the decrypted data with the real-time security level of the data after the receiving end decrypts the encrypted data to form a decryption log. The decryption log includes a data identifier, a timestamp, the real-time security level of the data, a decryption success indicator or a decryption failure indicator, decryption time, and a quantum bit error rate. The sixth module is used to generate drive signals based on the decryption logs to drive the preset intelligent data recognition model to perform online incremental updates.

[0083] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0084] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0085] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0086] Please see Figure 4 , Figure 4 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 410 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 420 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 420 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 420 and is called and executed by the processor 410 using the methods described above in the embodiments of this application. Input / output interface 430 is used to realize information input and output; The communication interface 440 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 450 transmits information between various components of the device (e.g., processor 410, memory 420, input / output interface 430, and communication interface 440); The processor 410, memory 420, input / output interface 430 and communication interface 440 are connected to each other within the device via bus 450.

[0087] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0088] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0089] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0090] It is understood that the content of the above method embodiments is applicable to the embodiments of this program product. The specific functions implemented by the embodiments of this program product are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0091] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0092] This application provides a data processing method and related equipment for the power grid industry. After acquiring multi-source heterogeneous power grid data (including structured, semi-structured, and unstructured data) within a target area, a preset intelligent data recognition model is used to perform semantic parsing and power grid topology correlation sensitivity analysis on the multi-source heterogeneous power grid data to obtain the real-time data security level. Then, the real-time data security level is matched with the corresponding encryption strength in a preset quantum encryption strategy library as the target encryption strength. A dynamic quantum key is jointly generated based on the real-time data security level and quantum true random data. Finally, the multi-source heterogeneous power grid data is processed based on the dynamic quantum key and the target encryption strength. After encrypting and generating ciphertext data, the ciphertext data is transmitted through a quantum channel at the sending end. When the receiving end decrypts the ciphertext data, it associates the decrypted data with the real-time security level of the data to form a decryption log, which includes a data identifier, timestamp, the real-time security level of the data, a successful decryption flag or a failed decryption flag, decryption time, and quantum bit error rate. Then, a driving signal is generated based on the decryption log to drive the preset intelligent data recognition model to perform online incremental updates. This can effectively improve the accuracy of the preset intelligent data recognition model in classifying and classifying data. Combined with the generation of dynamic quantum keys, the security of data transmission is further improved.

[0093] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0094] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0095] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0096] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0097] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0098] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0099] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0100] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0101] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0102] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0103] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A data processing method for the power grid industry, characterized in that, The method includes the following steps: Acquire multi-source heterogeneous power grid data within the target area, wherein the multi-source heterogeneous power grid data includes structured data, semi-structured data, and unstructured data; By performing semantic parsing and power grid topology correlation sensitivity analysis on the multi-source heterogeneous power grid data through a preset intelligent data recognition model, the real-time security level of the data is obtained. The real-time security level of the data is matched with the corresponding encryption strength in the preset quantum encryption strategy library as the target encryption strength, and a dynamic quantum key is jointly generated based on the real-time security level of the data and the quantum true random data. After encrypting the multi-source heterogeneous power grid data according to the dynamic quantum key and the target encryption strength to generate ciphertext data, the ciphertext data is transmitted through a quantum channel via the sending end. After the receiving end decrypts the encrypted data, it associates the decrypted data with the real-time security level of the data to form a decryption log. The decryption log includes a data identifier, a timestamp, the real-time security level of the data, a decryption success indicator or a decryption failure indicator, decryption time, and a quantum bit error rate. A driving signal is generated based on the decryption log to drive the preset intelligent data recognition model to perform online incremental updates.

2. The method according to claim 1, characterized in that, The preset intelligent data recognition model includes an input layer, a semantic embedding layer, a power grid topology layer, and a sensitivity scoring layer; The input layer is used to perform unified vector encoding on the multi-source heterogeneous power grid data to obtain initial data features; The semantic embedding layer is used to perform semantic feature mining on the initial data features using a pre-trained language model, and to capture long-distance semantic dependencies in the initial data features to obtain semantically enhanced features; The power grid topology layer is used to aggregate the association features of adjacent nodes based on the topological relationships of power grid equipment within the target area to obtain topology enhancement features; The sensitivity scoring layer is used to fuse the semantic enhancement features and the topology enhancement features, and then output the probability distribution of each security level through a fully connected classification network. Based on the probability distribution, the real-time security level of the multi-source heterogeneous power grid data is determined.

3. The method according to claim 2, characterized in that, The power grid topology layer adopts a three-layer graph attention network superposition structure; the topology enhancement features obtained by aggregating the association features of adjacent nodes based on the topology relationships of power grid equipment within the target area include: Each layer of the graph attention network is controlled to calculate the attention coefficients between adjacent nodes in the power grid equipment topology through a multi-head attention mechanism; The attention coefficients of all the adjacent nodes are normalized to obtain normalized attention weights; The topology enhancement feature is obtained by weighting and aggregating the initial data features of all nodes based on the normalized attention weights.

4. The method according to claim 1, characterized in that, The training process of the preset intelligent data recognition model includes the following steps: A labeled dataset is constructed as a training sample set, which includes power grid topology, historical power grid operation data, and the actual safety level labels corresponding to the historical operation data. The training sample set is input into the preset intelligent data recognition model to obtain the training prediction security level probability of each historical power grid operation data and the training topology enhancement features of two adjacent nodes in the power grid topology. The sensitivity score loss is calculated based on the actual security level label and the trained predicted security level probability, and the topology consistency loss is calculated based on the trained topology enhancement features. Calculate the composite total loss based on the sensitivity scoring loss and the topology consistency loss; The model parameters of the preset intelligent data recognition model are iteratively updated based on the backpropagation of the composite total loss.

5. The method according to claim 1, characterized in that, The step of matching the real-time security level of the data with the corresponding encryption strength in a preset quantum encryption strategy library as the target encryption strength, and jointly generating a dynamic quantum key based on the real-time security level of the data and quantum true random data, includes: The real-time security level of the data is mapped to parameters of the quantum key distribution protocol. The quantum key distribution protocol parameters are matched with the corresponding encryption strength in a preset quantum encryption strategy library as the target encryption strength; Dynamic quantum keys are jointly generated based on the parameters of the quantum key distribution protocol and quantum true random data. The quantum true random data includes true random binary sequences randomly extracted based on vacuum fluctuations or photon arrival times in quantum optics processes.

6. The method according to claim 5, characterized in that, The update frequency of the dynamic quantum key is adaptively adjusted by the parameters of the quantum key distribution protocol; When the parameters of the quantum key distribution protocol are greater than or equal to a preset high-sensitivity threshold, the dynamic quantum key is updated by generating a new quantum key independently in a single data transmission. When the parameters of the quantum key distribution protocol are less than or equal to a preset low-sensitivity threshold, the same quantum key is reused within a preset time window; When the quantum key distribution protocol parameters are greater than the preset low sensitivity threshold and less than the preset high sensitivity threshold, the quantum key is periodically updated based on the number of transmission batches or timestamps.

7. The method according to claim 1, characterized in that, The step of generating a driving signal based on the decryption log to drive the preset intelligent data recognition model to perform online incremental updates includes: The decryption log is used to calculate the decryption failure rate of consecutive batches in real time. When the failure rate of the continuous batch decryption exceeds the preset failure rate threshold, it is determined that the data distribution corresponding to the decryption log has drifted or suffered a network attack. Obtain the incremental decrypted text; Freeze the parameters of the semantic embedding layer and the power grid topology layer in the preset intelligent data recognition model, and fine-tune the number of layers of the sensitivity scoring layer according to the incremental decrypted text.

8. A data processing device for the power grid industry, characterized in that, The device includes: The first module is used to acquire multi-source heterogeneous power grid data within the target area, wherein the multi-source heterogeneous power grid data includes structured data, semi-structured data, and unstructured data. The second module is used to perform semantic parsing and power grid topology correlation sensitivity analysis on the multi-source heterogeneous power grid data through a preset intelligent data recognition model to obtain the real-time data security level. The third module is used to match the real-time security level of the data with the corresponding encryption strength in the preset quantum encryption strategy library as the target encryption strength, and to jointly generate a dynamic quantum key based on the real-time security level of the data and the quantum true random data. The fourth module is used to encrypt the multi-source heterogeneous power grid data according to the dynamic quantum key and the target encryption strength to generate ciphertext data, and then transmit the ciphertext data through a quantum channel via the sending end; The fifth module is used to associate the decrypted data with the real-time security level of the data after the receiving end decrypts the encrypted data to form a decryption log. The decryption log includes a data identifier, a timestamp, the real-time security level of the data, a decryption success indicator or a decryption failure indicator, decryption time, and a quantum bit error rate. The sixth module is used to generate a driving signal based on the decryption log to drive the preset intelligent data recognition model to perform online incremental updates.

9. An electronic device, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images