Scheduling operation data tracing method and system

Abstract

The invention provides a dispatching operation data traceability method and system, and the method comprises the steps: carrying out the preprocessing of collected power grid data, carrying out the feature extraction of the preprocessed power grid data, generating dispatching data feature information, obtaining a leakage data set under the condition of data leakage, carrying out the feature extraction of the leakage data set, and carrying out the traceability of dispatching operation data. Feature information corresponding to the leaked data set is generated, and then data leakage information of the leaked data set is determined from the scheduling data feature information based on the similarity between the feature information and the scheduling data feature information. The situation that watermarks are added into original data to affect use and sharing of the data is avoided, and meanwhile accurate traceability tracking of data leakage events is achieved.

Description

Technical Field

[0001] The present invention relates to the field of electric power systems, and in particular to a method and system for tracing the source of dispatching operation data. Background Art

[0002] In the power system with the dual carbon goals as the background, a variety of emerging business scenarios involving sources, networks, loads, and storage have emerged. As the digitalization process of dispatching business continues to accelerate, its characteristics such as business complexity, scenario diversification, and blurred boundaries are becoming increasingly prominent, causing the external data security and internal data security threats faced by the power system to continue to escalate. During operation, the power system continuously collects grid operation information and data. These data are sensitive, confidential, and have great application value. When these data are leaked, it is necessary to trace the source and hold accountable in a timely manner.

[0003] In addition, as the interactions between various businesses increase, the power monitoring system needs to continuously collect power grid operation information and data. This type of data is generally unstructured and important. If traditional watermarks are added to the head or tail of the data or to certain fields during use, it will affect the use and sharing of the data, and even affect the performance and stability of the system.

[0004] Data watermarking is a technology that hides identifiers in data through certain rules and is widely used in the field of data traceability. However, in the power system, the data that needs to be collected in real time includes system operation data, power grid model data, load forecast data, maintenance plans, etc. This type of data is usually unstructured and important business data. If data watermarks based on traditional watermarking technology are added to the head and tail of the data or certain fields, it will seriously affect the data use and sharing of the business system, and even affect the performance and stability of the system. If watermarks are not added, data traceability cannot be performed when a data leak occurs, and data security faces severe challenges. In the power system, it is urgent to break through the watermark-free data traceability technology to ensure that the source of the leak can be quickly located after the leak occurs, and provide information for the response to the incident.

[0005] Existing technical solution 1: Patent number CN112069269, "Data tracing method based on big data and multi-dimensional features and big data cloud server", proposes a data tracing method based on big data and multi-dimensional features and a big data cloud server. The characteristic distribution information of the data environment parameters of the multi-dimensional data feature queue is obtained by clustering, and feature correlation identification and data interaction defect identification are performed to obtain data correlation data feature set and defect data feature set. Considering the multi-dimensional characteristics of the data to be traced, a more complete and accurate traceability is achieved.

[0006] Existing technical solution 2: Patent number CN112905857, "A data leakage behavior tracing method and device based on data features", proposes a data leakage behavior tracing method based on data features. First, the characteristics of different data are classified, and the database where the data is located is determined according to the characteristic information of the data. The characteristics of the leaked data are extracted, and then the database to which the leaked data belongs is located, and the operation information related to the leaked data in the leaked database is queried to generate the leakage path, so as to achieve fast and effective tracking.

[0007] Existing technical solution 3: Patent number CN118171249, "A method for data traceability based on database watermark", proposes a method for data traceability based on database watermark. In the process of system open data, the open data is preprocessed, information is embedded or marked. In the process of data use, it is verified whether the detected watermark is consistent with the embedded watermark. If not, the data leakage path is analyzed based on the extracted watermark information.

[0008] Existing technical solution 1 proposes a data tracing method based on big data and multi-dimensional features, which uses a clustering algorithm to obtain data with strong correlation with the characteristics of the leaked data, and then traces the data. However, the computational complexity of the clustering algorithm is relatively high. For a large amount of data in the dispatching operation of the power system, the use of the clustering algorithm may lead to low processing efficiency, which limits the system performance during data analysis, and a more efficient analysis method needs to be considered.

[0009] Existing technical solution 2 provides a data leakage tracing method based on data features, which classifies data according to features, determines the leakage database by comparing it with the features of the leaked data, and tracks the leakage path through the related operations of the leaked data. However, there are many types of power system data, and this method may require different classification methods when classifying data, which increases the complexity of the system and may face computing performance bottlenecks when facing large-scale data.

[0010] Existing technical solution 3 describes a data tracing method based on database watermarking, which achieves the tracing of leaked data by embedding information or marking the data in the database. However, for the dispatching and operating data in the power system, the use of traditional data watermarking may affect the use and sharing of data.

[0011] Therefore, it is of great significance to trace the source of data leakage incidents without adding watermarks to the original data. Summary of the invention

[0012] In order to solve the problem of tracing the source of data leakage incidents in the prior art, the present invention proposes a scheduling operation data tracing method and system.

[0013] In a first aspect, a scheduling operation data tracing method is provided, comprising:

[0014] Preprocessing the collected dispatching and operating power grid data to generate preprocessed power grid data;

[0015] Extracting features from the preprocessed power grid data to generate dispatching data feature information;

[0016] In the event of a data leak, obtaining a leaked data set, and performing feature extraction on the leaked data set to generate feature information corresponding to the leaked data set;

[0017] Based on the similarity between the characteristic information and the scheduling data characteristic information, data leakage information of the leakage data set is determined from the scheduling data characteristic information.

[0018] Preferably, said extracting features from said preprocessed power grid data to generate scheduling data feature information includes:

[0019] Inputting the preprocessed power grid data into a pretrained feature extraction model;

[0020] Performing dimensionality reduction compression on the preprocessed power grid data through a convolution encoder in the feature extraction model to generate initial power grid data features;

[0021] Performing a linear transformation on the initial power grid data through a multi-head attention layer in the feature extraction model to generate a power grid data feature vector;

[0022] The dispatching data characteristic information is generated based on the power grid data characteristic vector and the dispatching operation power grid data.

[0023] Preferably, the determining the data leakage information corresponding to the leaked data set from the scheduling data characteristic information based on the similarity between the characteristic information and the scheduling data characteristic information includes:

[0024] Calculate the similarity between the characteristic information and each power grid data characteristic vector in the scheduling data characteristic information based on the cosine similarity calculation method and the mutual information entropy calculation method;

[0025] The scheduling data similar to the characteristic information in the scheduling data characteristic information is determined based on the similarity, and the data leakage information corresponding to the leakage data set is determined based on the scheduling data.

[0026] Preferably, the training process of the feature extraction model includes:

[0027] Acquire historical power grid data and dispatch data feature information corresponding to the power grid data, construct a training set from the power grid data and the dispatch data feature information, use the power grid data in the training set as input and the corresponding dispatch data feature information as output, substitute into a feature extraction model, train the feature extraction model, and obtain a trained feature extraction model;

[0028] Among them, the feature extraction model includes a convolutional encoder and a multi-head attention layer.

[0029] Preferably, the preprocessing of the collected dispatching operation power grid data to generate preprocessed power grid data includes:

[0030] Convert the collected dispatching and operating power grid data into one-dimensional power grid data;

[0031] The one-dimensional power grid data is segmented and filled to generate preprocessed power grid data of a target length.

[0032] Preferably, after generating the scheduling data characteristic information, the method further includes:

[0033] storing the scheduling data characteristic information in a corresponding log file; or,

[0034] The scheduling data feature information is stored in a data feature storage database.

[0035] In the second aspect, a scheduling operation data tracing system is provided, including:

[0036] The processing module is configured to pre-process the collected dispatching operation power grid data to generate pre-processed power grid data;

[0037] An extraction module is configured to extract features from the preprocessed power grid data to generate scheduling data feature information;

[0038] an acquisition module, configured to acquire a leaked data set in the event of a data leak, and perform feature extraction on the leaked data set to generate feature information corresponding to the leaked data set;

[0039] The determination module is configured to determine the data leakage information corresponding to the leakage data set from the scheduling data characteristic information based on the similarity between the characteristic information and the scheduling data characteristic information.

[0040] Preferably, the extraction module is specifically used for:

[0041] Inputting the preprocessed power grid data into a pretrained feature extraction model;

[0042] Performing dimensionality reduction compression on the preprocessed power grid data through a convolution encoder in the feature extraction model to generate initial power grid data features;

[0043] Performing a linear transformation on the initial power grid data through a multi-head attention layer in the feature extraction model to generate a power grid data feature vector;

[0044] The dispatching data characteristic information is generated based on the power grid data characteristic vector and the dispatching operation power grid data.

[0045] Preferably, the determining module is specifically used to:

[0046] Calculate the similarity between the characteristic information and each power grid data characteristic vector in the scheduling data characteristic information based on the cosine similarity calculation method and the mutual information entropy calculation method;

[0047] The scheduling data similar to the characteristic information in the scheduling data characteristic information is determined based on the similarity, and the data leakage information corresponding to the leakage data set is determined based on the scheduling data.

[0048] Preferably, the system further comprises:

[0049] Model training module, used to:

[0050] Acquire historical power grid data and dispatch data feature information corresponding to the power grid data, construct a training set from the power grid data and the dispatch data feature information, use the power grid data in the training set as input and the corresponding dispatch data feature information as output, substitute into a feature extraction model, train the feature extraction model, and obtain a trained feature extraction model;

[0051] Among them, the feature extraction model includes a convolutional encoder and a multi-head attention layer.

[0052] Preferably, the processing module is specifically used for:

[0053] Convert the collected dispatching and operating power grid data into one-dimensional power grid data;

[0054] The one-dimensional power grid data is segmented and filled to generate preprocessed power grid data of a target length.

[0055] Preferably, the system further comprises:

[0056] Storage module, specifically used for:

[0057] storing the scheduling data characteristic information in a corresponding log file; or,

[0058] The scheduling data feature information is stored in a data feature storage database.

[0059] In a third aspect, the present application further provides an electronic device, comprising: at least one processor and a memory; the memory and the processor are connected via a bus;

[0060] The memory is used to store one or more programs;

[0061] When the one or more programs are executed by the at least one processor, a scheduling operation data tracing method as described above is implemented.

[0062] In a fourth aspect, the present application also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed, a scheduling operation data tracing method as described above is implemented.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] The present invention provides a method and system for tracing the source of dispatching operation data, which pre-processes the collected power grid data, extracts features from the pre-processed power grid data, generates dispatching data feature information, and obtains a leaked data set in the event of data leakage, extracts features from the leaked data set, generates feature information corresponding to the leaked data set, and then determines data leakage information of the leaked data set from the dispatching data feature information based on the similarity between the feature information and the dispatching data feature information. This avoids adding watermarks to the original data to affect the use and sharing of the data, and simultaneously realizes accurate tracing of data leakage events. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 This is a system architecture diagram of the scheduling operation data tracing method of the present invention;

[0066] Figure 2 This is a flow chart of the scheduling operation data tracing method of the present invention;

[0067] Figure 3 A schematic diagram of the data feature extraction algorithm architecture of the scheduling operation data tracing method of the present invention;

[0068] Figure 4 This is a diagram showing an application example of the scheduling operation data tracing method of the present invention;

[0069] Figure 5 It is a structural schematic diagram of the scheduling operation data tracing method of the present invention;

[0070] Figure 6 The figure is a schematic diagram of the structure of an electronic device of the present invention. DETAILED DESCRIPTION

[0071] The present invention proposes a scheduling operation data tracing method, such as Figure 1 As shown in the figure, during the normal business data exchange process, the power data security monitoring platform continuously collects shared exchanged data, extracts and records data features through the Convolutional Auto-Encoder (CAE) and Multi Head Attention mechanism in the data parsing part of the traceability algorithm engine, and records event details at the same time. After the leaked data is found, the data features of the leaked data set are extracted, and a variety of feature similarity calculation methods are used to compare the similarity with the data features collected by the monitoring platform, so as to quickly locate the possible source of the leak, and further analyze the relevant log files to track the possible data leakage path.

[0072] In order to better understand the present invention, the content of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0073] Embodiment 1:

[0074] A scheduling operation data tracing method, such as Figure 2 As shown, including:

[0075] Step 1: preprocessing the collected dispatching operation power grid data to generate preprocessed power grid data;

[0076] Step 2: Extract features from the preprocessed power grid data to generate dispatch data feature information.

[0077] In this embodiment, in step 1, the collected dispatching operation power grid data is preprocessed to generate the preprocessed power grid data, including:

[0078] Convert the collected dispatching and operating power grid data into one-dimensional power grid data;

[0079] The one-dimensional power grid data is segmented and filled to generate preprocessed power grid data of a target length.

[0080] In this embodiment, in step 2, the process of extracting features from the preprocessed power grid data to generate dispatch data feature information includes:

[0081] Inputting the preprocessed power grid data into a pretrained feature extraction model;

[0082] Performing dimensionality reduction compression on the preprocessed power grid data through a convolution encoder in the feature extraction model to generate initial power grid data features;

[0083] Performing a linear transformation on the initial power grid data through a multi-head attention layer in the feature extraction model to generate a power grid data feature vector;

[0084] The dispatching data characteristic information is generated based on the power grid data characteristic vector and the dispatching operation power grid data.

[0085] In this embodiment, the feature extraction model is a pre-trained model, and its corresponding training process includes: obtaining historical power grid data and dispatching data feature information corresponding to the power grid data, constructing a training set from the power grid data and the dispatching data feature information, taking the power grid data in the training set as input and the corresponding dispatching data feature information as output, substituting them into the feature extraction model, training the feature extraction model, and obtaining a trained feature extraction model; wherein the feature extraction model includes a convolutional encoder and a multi-head attention layer.

[0086] In this embodiment, after generating the scheduling data characteristic information, it also includes: storing the scheduling data characteristic information in a corresponding log file; or, storing the scheduling data characteristic information in a data characteristic storage database, so as to subsequently obtain the scheduling data characteristic information from the corresponding log file or data characteristic storage database.

[0087] like Figure 3 As shown, Figure 3 This is a schematic diagram of the data feature extraction algorithm architecture of the scheduling operation data tracing method of the present invention, combined with Figure 3 The data feature extraction algorithm of the present invention is introduced.

[0088] The feature extraction algorithm proposed in the present invention includes four steps: first, collect business data (i.e. the aforementioned dispatching and operating power grid data), record the specific information of the business processing (including collection time, operator ID, operation type, data forwarding time, data forwarding content, etc.), and extract the original data of the business processing; second, pre-process the original data, convert the original data into one-dimensional data, and divide it into equal-sized fragments to meet the input requirements of the convolutional neural network, and complete it if the size is insufficient; then, combine the convolutional autoencoder with the multi-head attention mechanism to extract features from the power grid data; finally, store the specific information of the business processing and the extracted data features in the form of logs or databases; wherein convolution represents convolution, Attention score represents attention score, concat represents function, linear represents linear function, and Relu represents linear rectification function. The specific steps include:

[0089] 1) Data collection

[0090] In the power system, the data collected in real time often has the characteristics of time series. For example, voltage, current, load and other data will change over time. Flattening these data into one-dimensional data for processing is more helpful to capture the characteristics of time. In addition, the computational cost of one-dimensional data is lower, and the computing resources required for storage, processing and analysis are smaller. It is also more efficient to process in the real-time power monitoring system.

[0091] For the power grid data being shared, the data probe is used to collect the specific information of the business and extract the original data. The specific information of the business includes: the timestamp of data collection, the operator's ID, the type of operation (read, modify, delete, transfer), the timestamp of data transmission, the content summary of data transmission, and the target user ID of data transmission. The last three items are related information of transmission. If the operation type is not transmission, they are null values.

[0092] 2) Data preprocessing

[0093] First, the original data is converted into one-dimensional data. Since the length of the data is inconsistent, it is preprocessed into a fixed length through segmentation and padding. Finally, several one-dimensional data with a length of 256 are obtained to meet the input requirements of the convolutional neural network.

[0094] 3) Data feature extraction

[0095] After segmentation, the data will first pass through the convolution encoder to compress and reconstruct the features of the data, and then pass through the multi-head attention layer to output features of unified dimensions. The training process updates the network by comparing the decoder output with the original data to calculate the loss, and optimizes the attention layer structure by calculating the attention mechanism loss.

[0096] An autoencoder is an unsupervised neural network model, which consists of two parts: an encoder and a decoder. The encoder is responsible for compressing the input data into a lower dimension, and the decoder is responsible for reconstructing the original data based on the low-dimensional data. During the training process, the network is updated with the goal of minimizing the loss between the decoder output data and the original data. Compared with traditional autoencoders, convolutional autoencoders use convolutional layers to reduce and compress data, retaining important features of the data. The encoder used in the present invention contains two one-dimensional convolutions. The size of the corresponding convolution kernel in the two convolutional layers is kernel_size=3, the stride size is stride=2, and padding=1. After two layers of convolution, an output with 32 channels and a dimension of 1×32 is obtained. The decoder uses two transposed convolutional layers symmetrical to it. The training process uses the mean square error loss function to train the neural network, that is:

[0097]

[0098] Among them, L MSEis the mean square error loss function, N is the number of sample data, x i is the true value of the i-th sample data, is the predicted value of the i-th sample data output by the model, and i is the index variable used to traverse all sample data. Sample data refers to the data extracted from the overall data during the training process.

[0099] The core idea of ​​the multi-head attention mechanism is to divide the attention mechanism into multiple heads, each head calculates attention independently, and then splices the results together. This allows the model to focus on different parts of the input at the same time and capture more features. The multi-head attention mechanism first maps the input to the vector space of query, key, and value through linear transformation. For each head, there is the following linear transformation:

[0100] Q=XW Q , K = XW K , V = XW V

[0101] Among them, W Q , W K , W V is the learned weight matrix, Q is the query, K is the key, and V is the value.

[0102] Each head calculates attention weights and uses these weights to perform a weighted summation of the value vectors, calculated as:

[0103]

[0104] Among them, Attention(Q,K,V) is the attention weight, Softmax represents the normalized exponential function, and d k The outputs of all heads are concatenated and finally mapped through a linear layer to output a one-dimensional feature vector of length 32.

[0105] The present invention combines convolutional autoencoders with attention mechanisms to extract features from data shared by power grid exchanges. Convolutional autoencoders use convolutional layers to encode and decode input data and extract local features of the data, while the multi-head attention mechanism can capture the long-distance dependencies of the entire data. Combining the two technologies enables the model to learn local and global features of the input data.

[0106] 4) Data feature recording

[0107] After collecting the data features, the data features are combined with the specific information of the corresponding business and appended to the log file, or inserted into the database related to the monitoring platform, so that the data features can be retrieved in time for tracing when data leakage occurs in the future.

[0108] Step 3: In the event of data leakage, obtain a leaked data set, perform feature extraction on the leaked data set, and generate feature information corresponding to the leaked data set;

[0109] Step 4: Based on the similarity between the characteristic information and the scheduling data characteristic information, determine the data leakage information of the leakage data set from the scheduling data characteristic information

[0110] In this embodiment, in step 4, based on the similarity between the characteristic information and the characteristic information of the scheduling data, the process of determining the data leakage information corresponding to the leakage data set from the characteristic information of the scheduling data includes:

[0111] Calculate the similarity between the characteristic information and each power grid data characteristic vector in the scheduling data characteristic information based on the cosine similarity calculation method and the mutual information entropy calculation method;

[0112] The scheduling data similar to the characteristic information in the scheduling data characteristic information is determined based on the similarity, and the data leakage information corresponding to the leakage data set is determined based on the scheduling data.

[0113] In one embodiment, after the characteristics of the power grid exchange shared data are collected, when a data leakage event occurs, it is necessary to use a suitable method to compare the characteristics of the leaked data set with the characteristics of the collected data to quickly locate the possible source of the leakage. The present invention uses a variety of feature similarity calculation methods to calculate the correlation between the characteristics of the leaked data set and the characteristics in the collected database, determines the possible source of the leakage according to the size of the correlation, and further analyzes the relevant business information and logs to track the leakage path. Specifically:

[0114] 1) Cosine similarity

[0115] Cosine Similarity is used to measure the similarity between two vectors. Its value is between -1 and 1. The closer the value is to 1, the more similar they are. When a data leak occurs, cosine similarity can compare the similarity between the features of the leaked dataset and the features of the collected dataset to help locate the source of the leak. The formula is as follows:

[0116]

[0117] in, represents the dot product of two eigenvectors, and Represents the eigenvectors and Model.

[0118] 2) Mutual Information Entropy

[0119] Mutual Information (MI) is used to measure the degree of mutual dependence between two random variables. The formula is as follows:

[0120]

[0121] Among them, I(A; B) represents the mutual information between random variables A and B, p(a,b) represents the joint probability distribution between A and B, and p(a) and p(b) represent the marginal probability distributions of A and B respectively.

[0122] Combining the cosine similarity and mutual information entropy methods, first standardize both to the interval [0,1], and then perform weighted summation of the standardized results to obtain the comprehensive similarity. The formula is as follows:

[0123] S = α·cos+(1-α)·I(A;B)

[0124] Among them, α is the weight coefficient, which is used to adjust the contribution of the two methods to the similarity. Cosine similarity can capture the linear similarity between data sets, while mutual information entropy can capture the nonlinear relationship between data sets and evaluate the overall dependence between two variables. Combining cosine similarity and mutual information entropy can provide a more robust data similarity assessment method, especially when the data contains noise or outliers. Cosine similarity and mutual information entropy each have different sensitivities to different types of noise and outliers. When used in combination, they can complement each other and reduce errors. Using comprehensive similarity can more comprehensively evaluate the similarity between power grid data and leakage data sets, thereby improving the accuracy and robustness of traceability.

[0125] After obtaining a data set with high similarity to the leaked data set, we further analyze the log files corresponding to the relevant data set, trace possible leakage paths, and locate the source of the leakage. When the size of the leaked data set is large, it can provide richer leaked data features, reduce the impact of noise on the analysis results, and help improve the accuracy of data tracing.

[0126] The present invention is applicable to the traceability algorithm engine of the power data monitoring platform. In the power data sharing and exchange platform, the data of the dispatch operation is collected through the data probe, and the feature extraction technology combining the convolutional autoencoder and the multi-head attention mechanism is used to collect the data involved in the business for feature extraction, and the data features and business information are recorded in the log or database. When a data leakage event occurs, the leaked data set is submitted to the security monitoring platform, and the data features of the leaked data set are extracted in the same way, and the similarity is compared with the collected data features to locate the possible source of the leakage.

[0127] Before deployment, a large amount of actual business data needs to be collected to train the parameters in the feature extraction model so that the model can extract data features more accurately. After the training is completed, a data leakage incident is simulated to verify the model to ensure its robustness. Finally, the trained and verified model is deployed on the power data monitoring platform.

[0128] After the above solution is deployed, data probes are used to continuously monitor and collect business processing data. The feature extraction model processes this data in real time and records the extracted data features and business information in logs or databases. This mechanism ensures that a quick response and effective tracing can be carried out when a data leak occurs.

[0129] This embodiment relates to a feature-based watermark-free tracing mechanism applicable to scheduling operation data. This mechanism extracts data features for similarity comparison, thereby tracing the leaked data without changing the original data. The specific structure is as follows Figure 4 As shown, API stands for Application Programming Interface.

[0130] The dispatching operation data tracing method provided by the present invention provides an effective tracing algorithm engine for the power data monitoring platform, ensuring that possible data leakage sources can be efficiently traced without modifying the original data, thereby improving the security and reliability of the power system during data exchange and sharing.

[0131] Embodiment 2:

[0132] The present invention based on the same inventive concept also provides a scheduling operation data tracing system, such as Figure 5 As shown, including:

[0133] The processing module is configured to pre-process the collected dispatching operation power grid data to generate pre-processed power grid data;

[0134] An extraction module is configured to extract features from the preprocessed power grid data to generate scheduling data feature information;

[0135] an acquisition module, configured to acquire a leaked data set in the event of a data leak, and perform feature extraction on the leaked data set to generate feature information corresponding to the leaked data set;

[0136] The determination module is configured to determine the data leakage information corresponding to the leakage data set from the scheduling data characteristic information based on the similarity between the characteristic information and the scheduling data characteristic information.

[0137] Preferably, the extraction module is specifically used for:

[0138] Inputting the preprocessed power grid data into a pretrained feature extraction model;

[0139] Performing dimensionality reduction compression on the preprocessed power grid data through a convolution encoder in the feature extraction model to generate initial power grid data features;

[0140] Performing a linear transformation on the initial power grid data through a multi-head attention layer in the feature extraction model to generate a power grid data feature vector;

[0141] The dispatching data characteristic information is generated based on the power grid data characteristic vector and the dispatching operation power grid data.

[0142] Preferably, the determining module is specifically used to:

[0143] Calculate the similarity between the characteristic information and each power grid data characteristic vector in the scheduling data characteristic information based on the cosine similarity calculation method and the mutual information entropy calculation method;

[0144] The scheduling data similar to the characteristic information in the scheduling data characteristic information is determined based on the similarity, and the data leakage information corresponding to the leakage data set is determined based on the scheduling data.

[0145] Preferably, the system further comprises:

[0146] Model training module, used to:

[0147] Acquire historical power grid data and dispatch data feature information corresponding to the power grid data, construct a training set from the power grid data and the dispatch data feature information, use the power grid data in the training set as input and the corresponding dispatch data feature information as output, substitute into a feature extraction model, train the feature extraction model, and obtain a trained feature extraction model;

[0148] Among them, the feature extraction model includes a convolutional encoder and a multi-head attention layer.

[0149] Preferably, the processing module is specifically used for:

[0150] Convert the collected dispatching and operating power grid data into one-dimensional power grid data;

[0151] The one-dimensional power grid data is segmented and filled to generate preprocessed power grid data of a target length.

[0152] Preferably, the system further comprises:

[0153] Storage module, specifically used for:

[0154] storing the scheduling data characteristic information in a corresponding log file; or,

[0155] The scheduling data feature information is stored in a data feature storage database.

[0156] Example 3

[0157] like Figure 6 As shown, the present invention also provides an electronic device, which may be a computer device, a single-chip device, an intelligent mobile device, etc. The electronic device in this embodiment may include a processor, a memory, a transceiver component, etc. The memory, the processor, and the transceiver component are connected via a bus; the memory may be used to store an execution program, and an exemplary execution program may include instructions; the processor is used to execute the instructions stored in the memory. The memory may also be used to store data, which may be called and / or modified when the instructions are executed.

[0158] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in a storage medium to implement the corresponding method flow or corresponding functions, so as to realize the steps of a scheduling operation data tracing method in the above embodiment.

[0159] Example 4

[0160] Based on the same inventive concept, the present invention also provides a readable storage medium, specifically an electronic device readable storage medium (Memory), which is a memory device in an electronic device for storing programs and data. It can be understood that the storage medium here can include both built-in storage media in electronic devices and extended storage media supported by electronic devices. The storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more execution programs (including program codes). It should be noted that the storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage. The processor loads and executes one or more instructions stored in the storage medium, which can implement the steps of a scheduling operation data tracing method in the above embodiment.

[0161] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, 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 disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0162] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, 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 flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0163] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate 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 process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0164] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0165] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are included in the scope of the claims of the present invention to be approved.

Claims

1. A scheduling operation data tracing method, characterized in that: include: Preprocessing the collected dispatching and operating power grid data to generate preprocessed power grid data; Extracting features from the preprocessed power grid data to generate dispatching data feature information; In the event of a data leak, obtaining a leaked data set, and performing feature extraction on the leaked data set to generate feature information corresponding to the leaked data set; Based on the similarity between the characteristic information and the scheduling data characteristic information, data leakage information of the leakage data set is determined from the scheduling data characteristic information.

2. The method according to claim 1, characterized in that The extracting features of the preprocessed power grid data to generate scheduling data feature information includes: Inputting the preprocessed power grid data into a pretrained feature extraction model; Performing dimensionality reduction compression on the preprocessed power grid data by using a convolution encoder in the feature extraction model to generate initial power grid data features; Performing a linear transformation on the initial power grid data through a multi-head attention layer in the feature extraction model to generate a power grid data feature vector; The dispatching data characteristic information is generated based on the power grid data characteristic vector and the dispatching operation power grid data.

3. The method according to claim 2, characterized in that The determining, based on the similarity between the characteristic information and the scheduling data characteristic information, the data leakage information corresponding to the leakage data set from the scheduling data characteristic information includes: Calculate the similarity between the characteristic information and each power grid data characteristic vector in the scheduling data characteristic information based on the cosine similarity calculation method and the mutual information entropy calculation method; The scheduling data similar to the characteristic information in the scheduling data characteristic information is determined based on the similarity, and the data leakage information corresponding to the leakage data set is determined based on the scheduling data.

4. The method according to claim 2, characterized in that: The training process of the feature extraction model includes: Acquire historical power grid data and dispatch data feature information corresponding to the power grid data, construct a training set from the power grid data and the dispatch data feature information, use the power grid data in the training set as input and the corresponding dispatch data feature information as output, substitute into a feature extraction model, train the feature extraction model, and obtain a trained feature extraction model; Among them, the feature extraction model includes a convolutional encoder and a multi-head attention layer.

5. The method according to claim 1, characterized in that The preprocessing of the collected dispatching operation power grid data to generate preprocessed power grid data includes: Convert the collected dispatching and operating power grid data into one-dimensional power grid data; The one-dimensional power grid data is segmented and filled to generate preprocessed power grid data of a target length.

6. The method according to claim 1, characterized in that After the scheduling data characteristic information is generated, the method further includes: storing the scheduling data characteristic information in a corresponding log file; or, The scheduling data feature information is stored in a data feature storage database.

7. A scheduling operation data tracing system, characterized in that: include: The processing module is configured to pre-process the collected dispatching operation power grid data to generate pre-processed power grid data; An extraction module is configured to extract features from the preprocessed power grid data to generate scheduling data feature information; an acquisition module, configured to acquire a leaked data set in the event of a data leak, and perform feature extraction on the leaked data set to generate feature information corresponding to the leaked data set; The determination module is configured to determine the data leakage information corresponding to the leakage data set from the scheduling data characteristic information based on the similarity between the characteristic information and the scheduling data characteristic information.

8. The system according to claim 7, characterized in that The extraction module is specifically used for: Inputting the preprocessed power grid data into a pretrained feature extraction model; Performing dimensionality reduction compression on the preprocessed power grid data by using a convolution encoder in the feature extraction model to generate initial power grid data features; Performing a linear transformation on the initial power grid data through a multi-head attention layer in the feature extraction model to generate a power grid data feature vector; The dispatching data characteristic information is generated based on the power grid data characteristic vector and the dispatching operation power grid data.

9. An electronic device, characterized in that: include: at least one processor and memory; The memory and the processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, the scheduling operation data tracing method as described in any one of claims 1 to 6 is implemented.

10. A readable storage medium, characterized in that: An execution program is stored thereon, and when the execution program is executed, the scheduling operation data tracing method as described in any one of claims 1 to 6 is implemented.

Images