Multi-stream data fusion method and device, electronic equipment and storage medium

Abstract

The invention discloses a multi-stream data fusion method and device, electronic equipment and a storage medium, and belongs to the technical field of multi-stream data processing, and the method comprises the steps: inputting multi-source data into a preset missing value processing model, and outputting complete multi-source data; according to the complete multi-source data, by taking the equipment and the user as nodes and taking the relationship between the equipment and the user as edges, constructing a corresponding knowledge graph; the method comprises the following steps: acquiring feature information corresponding to each node in a knowledge graph, splicing each piece of feature information of each node into a corresponding feature vector, and then aggregating each node according to the feature vector corresponding to each node to obtain feature representation of each node after aggregation; wherein the feature information comprises an electric energy flow feature, an information flow feature and a service flow feature, and by implementing the method and the device, the problem that in the prior art, modeling and fusion are not carried out on the electric energy flow, the information flow and the service flow according to the electric power multi-flow data can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-stream data processing, and in particular to a multi-stream data fusion method, device, electronic device and storage medium. Background Art

[0002] In the context of the rapid development of smart grids, low-voltage distribution networks, as the foundation of power systems, directly impact the operational efficiency and stability of equipment. Real-time monitoring of equipment status and power usage is urgently needed. The integration of multi-stream power data is crucial for this real-time monitoring.

[0003] Energy flow involves the transmission of electricity, including parameters such as voltage, current, and power. Information flow involves device communication and data exchange, impacting the system's response speed and accuracy. Business flow considers diverse user needs, helping to optimize resource allocation and improve power supply reliability. Modeling and integrating energy flow, information flow, and business flow based on multi-flow power data not only enables device status prediction and fault diagnosis, but also significantly improves overall operational efficiency, supports decision-making, and thus comprehensively enhances power supply service quality and promotes a sustainable power supply system. However, currently, no technology exists for modeling and integrating energy flow, information flow, and business flow based on multi-flow power data. Summary of the Invention

[0004] The present invention provides a multi-stream data fusion method, device, electronic device and storage medium, which can solve the technical problem in the prior art that there is no related technology for modeling and fusing electric energy flow, information flow and business flow based on power multi-stream data.

[0005] In order to solve the above technical problems, an embodiment of the present invention provides a multi-stream data fusion method, including:

[0006] Acquire multi-source data of the power system; wherein the multi-source data includes: power load, equipment status and meteorological data;

[0007] Inputting the multi-source data into a preset missing value processing model, so that the missing value processing model extracts the time features and spatial features of the multi-source data, and performs feature fusion on the time features and spatial features to obtain corresponding spatiotemporal features, and then predicts and fills the missing values ​​in the multi-source data according to the spatiotemporal features, and outputs complete multi-source data;

[0008] Based on the complete multi-source data, a corresponding knowledge graph is constructed with devices and users as nodes and the relationships between the devices and users as edges;

[0009] Obtain feature information corresponding to each node in the knowledge graph, splice the feature information of each node into a corresponding feature vector, and then aggregate each node according to the feature vector corresponding to each node to obtain the aggregated feature representation of each node; wherein the feature information includes: power flow features, information flow features, and business flow features.

[0010] As a preferred solution, before inputting the multi-source data into a preset missing value processing model, the method further includes:

[0011] Extracting data fields corresponding to each data in the multi-source data; wherein the data fields include: timestamp, device ID, and measurement value;

[0012] Compare the data fields of each data, and take all data with consistent data fields as a group of duplicate data, then retain any data in each group of duplicate data and delete the rest;

[0013] Each data is compared with a preset data range, and data exceeding the data range is regarded as erroneous data, and all erroneous data are deleted.

[0014] As a preferred solution, the generation of the missing value processing model includes:

[0015] Acquire historical multi-source data of the power system;

[0016] Adding a preset mask matrix to the historical multi-source data, and using the original historical multi-source data and the historical multi-source data after adding the mask matrix as training samples;

[0017] A preset spatiotemporal graph convolutional network model is trained according to the training samples to obtain the missing value processing model.

[0018] As a preferred solution, the process of aggregating the nodes according to the feature vector corresponding to each node to obtain the aggregated feature representation of each node includes:

[0019] For each node, a preset number of neighboring nodes are sampled, and the feature vectors corresponding to the sampled neighboring nodes are aggregated to obtain the aggregated neighbor fusion feature vector;

[0020] The node's own feature vector and the corresponding neighbor fusion feature vector are aggregated to obtain the aggregated feature representation of each node.

[0021] Based on the above embodiment, another embodiment of the present invention provides a multi-stream data fusion device, including: a data acquisition module, a missing value processing module, a knowledge graph construction module, and a node feature aggregation module;

[0022] The data acquisition module is used to acquire multi-source data of the power system; wherein the multi-source data includes: power load, equipment status and meteorological data;

[0023] The missing value processing module is used to input the multi-source data into a preset missing value processing model, so that the missing value processing model extracts the time features and spatial features of the multi-source data, and fuses the time features and spatial features to obtain corresponding spatiotemporal features, and then predicts and fills the missing values ​​in the multi-source data according to the spatiotemporal features, and outputs complete multi-source data;

[0024] The knowledge graph construction module is used to construct a corresponding knowledge graph based on the complete multi-source data, with devices and users as nodes and the relationship between the devices and users as edges;

[0025] The node feature aggregation module is used to obtain the feature information corresponding to each node in the knowledge graph, splice the feature information of each node into a corresponding feature vector, and then aggregate each node according to the feature vector corresponding to each node to obtain the aggregated feature representation of each node; wherein the feature information includes: power flow features, information flow features and business flow features.

[0026] As a preferred solution, before inputting the multi-source data into a preset missing value processing model, the method further includes:

[0027] Extracting data fields corresponding to each data in the multi-source data; wherein the data fields include: timestamp, device ID, and measurement value;

[0028] Compare the data fields of each data, and take all data with consistent data fields as a group of duplicate data, then retain any data in each group of duplicate data and delete the rest;

[0029] Each data is compared with a preset data range, and data exceeding the data range is regarded as erroneous data, and all erroneous data are deleted.

[0030] As a preferred solution, the generation of the missing value processing model includes:

[0031] Acquire historical multi-source data of the power system;

[0032] Adding a preset mask matrix to the historical multi-source data, and using the original historical multi-source data and the historical multi-source data after adding the mask matrix as training samples;

[0033] A preset spatiotemporal graph convolutional network model is trained according to the training samples to obtain the missing value processing model.

[0034] As a preferred solution, the process of aggregating the nodes according to the feature vector corresponding to each node to obtain the aggregated feature representation of each node includes:

[0035] For each node, a preset number of neighboring nodes are sampled, and the feature vectors corresponding to the sampled neighboring nodes are aggregated to obtain the aggregated neighbor fusion feature vector;

[0036] The node's own feature vector and the corresponding neighbor fusion feature vector are aggregated to obtain the aggregated feature representation of each node.

[0037] Based on the above embodiments, another embodiment of the present invention provides an electronic device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the multi-stream data fusion method described in the above embodiments of the invention is implemented.

[0038] Based on the above embodiment, another embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the multi-stream data fusion method described in the above embodiment of the invention.

[0039] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0040] The present invention provides a multi-stream data fusion method to obtain multi-source data of an electric power system; input the multi-source data into a preset missing value processing model so that the missing value processing model extracts the time features and spatial features of the multi-source data, and fuses the time features and spatial features to obtain corresponding spatiotemporal features, and then predicts and fills the missing values ​​in the multi-source data according to the spatiotemporal features, and outputs complete multi-source data; based on the complete multi-source data, a corresponding knowledge graph is constructed with devices and users as nodes and the relationship between the devices and users as edges; feature information corresponding to each node in the knowledge graph is obtained, and the feature information of each node is spliced ​​into a corresponding feature vector, and then the nodes are aggregated according to the feature vector corresponding to each node to obtain the feature representation of each node after aggregation; wherein the feature information includes: power flow features, information flow features and business flow features.

[0041] Through the present invention, missing value processing can be performed on the multi-source processing of the power system, and then a corresponding knowledge graph can be constructed based on the processed multi-source data. Then, the power flow characteristics, information flow characteristics and business flow characteristics corresponding to each node in the knowledge graph are obtained, the multi-stream characteristics are fused, and then the nodes are aggregated according to the fused feature vector to obtain the final feature representation of each node, thereby completing the multi-stream data fusion of each node. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flow chart of a multi-stream data fusion method provided by one embodiment of the present invention;

[0043] Figure 2 It is the overall implementation flow chart of the present invention;

[0044] Figure 3 It is a flow chart of data integrity processing by spatiotemporal feature extraction;

[0045] Figure 4 This is a flowchart of multi-stream information fusion in graph neural networks;

[0046] Figure 5 The figure is a structural diagram of a multi-stream data fusion device provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0047] To make the objectives, technical solutions, and advantages of this application more clear, the technical solutions in this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this application.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned figure descriptions are intended to cover non-exclusive inclusions.

[0049] In the description of the embodiments of this application, the technical terms "first" and "second" are used only to distinguish different objects and should not be understood to indicate or imply relative importance or implicitly specify the quantity, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, the meaning of "plurality" is more than two, unless otherwise clearly and specifically defined.

[0050] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0051] In the description of the embodiments of this application, the term "and / or" is simply a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent the following three situations: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0052] In the description of the embodiments of the present application, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).

[0053] In the description of the embodiments of the present application, unless otherwise expressly specified or limited, technical terms such as "installed," "connected," "connected," and "fixed" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integration; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; internal connections between two components or interactions between two components. Those skilled in the art can understand the specific meanings of the above terms in the embodiments of the present application based on specific circumstances.

[0054] Example 1

[0055] Please refer to Figure 1 To address the problem that there is no existing technology for modeling and fusing power flow, information flow, and business flow based on multi-stream power data, an embodiment of the present invention provides a flowchart of a multi-stream data fusion method, including the following specific steps:

[0056] S1. Acquire multi-source data of the power system; wherein the multi-source data includes: power load, equipment status, and meteorological data;

[0057] Preferably, before inputting the multi-source data into a preset missing value processing model, it also includes: extracting the data fields corresponding to each data in the multi-source data; wherein the data fields include: timestamp, device ID and measurement value; comparing the data fields of each data, treating all data with consistent data fields as a group of duplicate data, and then retaining any data in each group of duplicate data and deleting the remaining data; comparing each data with a preset data range, treating data exceeding the data range as erroneous data, and deleting all erroneous data.

[0058] Specifically, by extracting temporal and spatial features, this invention can effectively capture how the operating status of different devices changes over time and space, enhancing the ability to handle missing data. Combining temporal and spatial features with graph structure allows for better prediction of missing data, intelligently filling in missing data using information from surrounding nodes, and improving data integrity and reliability. Furthermore, graph neural networks can model complex relationships between devices, including adjacency and interactions, and integrate feature information from multiple streams to more comprehensively capture the system's operating status and environmental impacts, thereby improving model accuracy.

[0059] In multi-stream data modeling, this invention mainly addresses the incompleteness of multi-stream data, including duplication, missing and other issues, and fuses multi-stream data to improve the performance of data modeling. It proposes a multi-stream fusion method that combines spatiotemporal feature extraction and graph neural network. Figure 2 , which is the overall implementation flow chart of the present invention, is divided into three parts: data information collected and generated by various electrical equipment, data integrity processing by spatiotemporal feature extraction, and multi-stream information fusion by graph neural network.

[0060] Please refer to Figure 3 , which is a flow chart for data integrity processing for spatiotemporal feature extraction, including the following specific steps:

[0061] Data collection and data preprocessing:

[0062] (1) Data collection: Collect multi-source data such as power load, equipment status, and meteorological data.

[0063] (2) Data cleaning: Perform simple identification on the collected data. For duplicate data, compare the key fields of the data record (timestamp, device ID, measurement value, etc.) to identify duplicate data. If two data D1 = {t1, ID1, V1} and D2 = {t2, ID2, V2} satisfy:

[0064] t1=t2, ID1=ID2, V1=V2;

[0065] D1 and D2 are considered duplicate data, and one copy of the valid data will be retained while the other duplicate data will be deleted.

[0066] At the same time, the collected data may contain erroneous data. Set range constraints on the data (voltage, current, etc.). If the collected data does not meet the range, it will be considered as erroneous data and deleted. Use the statistical method Z-Score formula:

[0067]

[0068] Where X is the data value, μ is the mean, and σ is the standard deviation. If |Z| > 3 (i.e., it deviates from the mean by 3 standard deviations), X is considered an outlier. Remove the outlier to ensure a reasonable data distribution.

[0069] (3) Feature extraction: Obtain timestamp-related features from the data, such as hours, weeks, holidays, etc. Use the GRU time series model to extract time features, including: long-term dependency of data, short-term change rate, and time periodicity. In addition, timestamp features can be used as additional input to enhance the GRU model's perception of periodic and static time information. Extract spatial feature information and use GCN to aggregate the node's neighbor information. The information of neighbor nodes is fused through the aggregation operation of GCN to capture the direct or indirect connections between nodes. This feature reflects the position of a node in the entire graph structure, as well as its similarity or difference in attributes with other nodes. GCN can handle spatial associations between different streams in multi-stream data.

[0070] Spatiotemporal Graph Convolutional Network (ST-GCN): This constructs a graph convolutional layer and a temporal convolutional layer to process data. The graph convolutional layer processes the data and extracts spatial features. The temporal convolutional layer processes the data and extracts temporal features.

[0071] S2. Inputting the multi-source data into a preset missing value processing model, so that the missing value processing model extracts the temporal features and spatial features of the multi-source data, and performs feature fusion on the temporal features and spatial features to obtain corresponding spatiotemporal features, and then predicts and fills the missing values ​​in the multi-source data according to the spatiotemporal features, and outputs complete multi-source data;

[0072] Preferably, the generation of the missing value processing model includes: obtaining historical multi-source data of the power system; adding a preset mask matrix to the historical multi-source data, and using the original historical multi-source data and the historical multi-source data after adding the mask matrix as training samples; training a preset spatiotemporal graph convolutional network model according to the training samples to obtain the missing value processing model.

[0073] Specifically, after data collection and data preprocessing, missing value prediction is performed:

[0074] (1) Data input: The user's energy flow (e.g., current, voltage, power), information flow (e.g., communication delay, bandwidth), and service flow (e.g., user request frequency, service type) are used as the network input parameters. The format of the input data is: X = N × T × F, where N is the number of nodes (e.g., users or devices). T is the number of time steps. F is the feature dimension of each node. A graph G is constructed: consisting of nodes (users, devices) and edges (connection relationships). Missing values ​​are marked with the special symbol NaN and handled by a mask matrix during training.

[0075] Mask matrix: Define a binary mask matrix M with the same shape as the input data X.

[0076] If the data is missing, set M[i][t][f] = 0; if the data is complete, set M[i][t][f] = 1.

[0077] (2) Network structure:

[0078] For temporal feature extraction (GRU):

[0079] Input: Time series features X of each node.

[0080] By learning short-term fluctuations and long-term trends through the forget gate and update gate, the hidden state is updated:

[0081] z t =σ(W z ·[h t-1 ,x t ]+b z );

[0082]

[0083] Among them, h t is the time feature (hidden state of the current time step), z t is the update gate, used to fuse the current input with the past memory, is a candidate state, used to represent the input features of the current time step.

[0084] Output: Time characteristic: H time ={h1,h2,...,h T};

[0085] For spatial feature extraction (GCN):

[0086] Input: Time feature H time And graph structure G (adjacency matrix A).

[0087] Normalize the adjacency matrix of the graph to aggregate the information of neighbor nodes.

[0088]

[0089] Where I is the identity matrix, which means adding self-loops to each node. D is the degree matrix (a diagonal matrix, indicating the degree of each node).

[0090] Combine the aggregated neighbor information with the node's own features and update them into the spatial features of the current node:

[0091]

[0092] Among them, H (k) i represents the spatial feature representation of node i at the kth layer.

[0093] Output: H space Indicates the attribute association between nodes.

[0094] Spatiotemporal feature fusion: Fusion of temporal features: H time and spatial characteristics H space :

[0095] H spatio-temporal =W fusion CONCAT(H time ,H space );

[0096] Among them, W fusion is the weight matrix of spatiotemporal fusion.

[0097] Finally, a fully connected layer is used to transform the spatiotemporal features H spatio-temporal Prediction results mapped to missing values

[0098] Network training:

[0099] (1) Initialization: Initialize the parameters of GRU and GCN, initialize the learning rate and optimizer.

[0100] (2) Forward propagation:

[0101] Input data X and graph structure G;

[0102] Temporal feature extraction (GRU): GRU is used to extract the temporal feature H of each node time .

[0103] Spatial feature extraction (GCN): GCN is used to combine neighbor node information to extract spatial features H space .

[0104] Spatiotemporal feature fusion: Fusion of temporal features: H time and spatial characteristics H space .

[0105] Predict missing values: Use a fully connected layer or regression layer to predict missing values

[0106] (3) Calculation error:

[0107] The predicted value is calculated using the mask mean square error loss function L The difference between the true value X.

[0108] (4) Directional propagation:

[0109] Calculate the gradient through back propagation and optimize the parameters of GRU and GCN:

[0110]

[0111] Where η is the learning rate.

[0112] (5) Iterative training:

[0113] Repeat forward propagation and backward propagation for multiple iterations (epochs) until the loss function converges.

[0114] S3. Based on the complete multi-source data, a corresponding knowledge graph is constructed with devices and users as nodes and the relationships between the devices and users as edges;

[0115] For details, please refer to Figure 4 , which is a flowchart of multi-stream information fusion in graph neural networks, specifically includes the following steps:

[0116] Graph construction:

[0117] Node V: defines the nodes in the graph, including power equipment (such as transformers, distribution boxes), users, communication equipment, etc.

[0118] Edge E: defines the relationship between nodes, such as the connection of power flow (power transmission between devices), the communication link of information flow (information transmission between devices), and the interaction of business flow (service request between users and devices).

[0119] S4. Obtain feature information corresponding to each node in the knowledge graph, splice the feature information of each node into a corresponding feature vector, and then aggregate each node according to the feature vector corresponding to each node to obtain the feature representation of each node after aggregation; wherein the feature information includes: power flow features, information flow features and business flow features.

[0120] Preferably, the process of aggregating each node according to the feature vector corresponding to each node to obtain the aggregated feature representation of each node includes: for each node, sampling a preset number of neighboring nodes, and aggregating the feature vectors corresponding to the sampled neighboring nodes to obtain the aggregated neighbor fusion feature vector; aggregating the node's own feature vector and the corresponding neighbor fusion feature vector to obtain the aggregated feature representation of each node.

[0121] Specifically, after building the knowledge graph, continue with feature engineering: define feature vectors for each node. Features can include:

[0122] (1) Characteristics of electrical energy flow: current, voltage, power, etc.;

[0123] (2) Information flow characteristics: transmission delay, bandwidth utilization, packet loss rate, etc.

[0124] (3) Business flow characteristics: user request frequency, service response time, business type, etc.

[0125] Multi-source information fusion:

[0126] (1) Concatenate the three types of data stream features into a feature vector:

[0127]

[0128] Among them, h power is the characteristic of electric energy flow, h info is the information flow feature, h business It is the business flow feature.

[0129] (2) GraphSAGE realizes multi-source information fusion through neighbor sampling, aggregation, and feature updating.

[0130] a) Neighbor sampling:

[0131] Each node v samples a fixed number of neighboring nodes Used for local aggregation.

[0132]

[0133] Among them, K is the number of sampled neighbor nodes.

[0134] b) Feature aggregation:

[0135] Aggregate the features of the sampled neighbor nodes:

[0136]

[0137] Among them, AGGREGATE is an aggregation function that uses LSTM aggregation to perform sequence processing on neighbor features.

[0138] c) Feature Updates:

[0139] Combine the node's own features with the aggregated neighbor features, and generate a new feature representation through linear transformation and activation function:

[0140]

[0141] Among them, W (k) is the weight matrix of the current layer, and σ is the activation function.

[0142] d) Final feature representation:

[0143] After K layers, the final feature representation of each node v is:

[0144]

[0145] Among them, h fused It is a comprehensive feature that integrates power flow, information flow and business flow.

[0146] LSTM aggregation process:

[0147] The features of neighbor nodes Treated as an input sequence:

[0148]

[0149] Recursively process the neighbor feature sequence:

[0150] Forget gate: controls the degree to which information from the previous state is retained in the current state.

[0151] f t =σ(W i ·[h t-1 ,x t ]);

[0152] Among them, f t is the output of the forget gate, h t-1 is the hidden state of the previous time step, x t is the current input.

[0153] Input gate: controls the degree of influence of the current input on the unit state:

[0154] i t =σ(W i ·[h t-1 ,x t ]);

[0155]

[0156] Unit state update: Combine the forget gate and input gate to update the unit state:

[0157]

[0158] Among them, C t is the current cell state, C t-1 It is the previous unit status.

[0159] Output gate: controls the degree of influence of the current cell state on the output:

[0160] o t =σ(W o ·[h t-1 ,x t ]);

[0161] Current hidden status:

[0162] h t =o t ⊙tanh(C t );

[0163] Feature aggregation:

[0164] Final hidden state of LSTM Used as an aggregate representation of neighbor features:

[0165]

[0166] It can be seen that the present invention provides a multi-stream data fusion method, which can achieve the following beneficial effects:

[0167] (1) Improve data integrity and accurately fill missing values: By combining spatiotemporal feature extraction with graph convolutional neural networks (GCN), the temporal and spatial correlations in the data can be effectively captured, thereby accurately filling missing values ​​and improving data integrity and quality.

[0168] (2) Powerful feature extraction capabilities: spatiotemporal feature integration: The time series model can capture the temporal dependency of data, while GCN can effectively mine the spatial relationship between nodes. The combination of the two can comprehensively extract useful features and enhance the expressiveness of the model.

[0169] (3) Effective multi-stream data fusion: Information sharing: Through graph neural networks (GNNs), data features of different streams can be shared between nodes, promoting information fusion and improving the model's adaptability to complex scenarios. Handling complex relationships: GNNs can handle complex relationships between nodes, which is suitable for the interaction between various devices in the power system and improves the effect of multi-stream data fusion.

[0170] (4) Enhanced model performance: Improved accuracy: Multi-stream data fusion based on processed complete data can reduce noise and redundancy, improving the model's prediction accuracy, especially in applications such as power load forecasting. Adaptability and flexibility: The solution can dynamically adjust to changes in new data, maintaining the model's high adaptability and adapting to the ever-changing power system environment.

[0171] (5) Scalability: Modular design: The design of the solution allows for the flexible addition of new data streams or features to adapt to the needs of different scenarios and achieve scalability. Cross-domain application: Although the solution focuses on power systems, its methodology and model architecture can be applied to other fields such as transportation and environmental monitoring.

[0172] Example 2

[0173] Please refer to Figure 5 , is a schematic diagram of the structure of a multi-stream data fusion device provided by an embodiment of the present invention, the device comprising: a data acquisition module, a missing value processing module, a knowledge graph construction module, and a node feature aggregation module;

[0174] The data acquisition module is used to acquire multi-source data of the power system; wherein the multi-source data includes: power load, equipment status and meteorological data;

[0175] The missing value processing module is used to input the multi-source data into a preset missing value processing model, so that the missing value processing model extracts the time features and spatial features of the multi-source data, and fuses the time features and spatial features to obtain corresponding spatiotemporal features, and then predicts and fills the missing values ​​in the multi-source data according to the spatiotemporal features, and outputs complete multi-source data;

[0176] The knowledge graph construction module is used to construct a corresponding knowledge graph based on the complete multi-source data, with devices and users as nodes and the relationship between the devices and users as edges;

[0177] The node feature aggregation module is used to obtain the feature information corresponding to each node in the knowledge graph, splice the feature information of each node into a corresponding feature vector, and then aggregate each node according to the feature vector corresponding to each node to obtain the aggregated feature representation of each node; wherein the feature information includes: power flow features, information flow features and business flow features.

[0178] Preferably, before inputting the multi-source data into a preset missing value processing model, it also includes: extracting the data fields corresponding to each data in the multi-source data; wherein the data fields include: timestamp, device ID and measurement value; comparing the data fields of each data, treating all data with consistent data fields as a group of duplicate data, and then retaining any data in each group of duplicate data and deleting the remaining data; comparing each data with a preset data range, treating data exceeding the data range as erroneous data, and deleting all erroneous data.

[0179] Preferably, the generation of the missing value processing model includes: obtaining historical multi-source data of the power system; adding a preset mask matrix to the historical multi-source data, and using the original historical multi-source data and the historical multi-source data after adding the mask matrix as training samples; training a preset spatiotemporal graph convolutional network model according to the training samples to obtain the missing value processing model.

[0180] Preferably, the process of aggregating each node according to the feature vector corresponding to each node to obtain the aggregated feature representation of each node includes: for each node, sampling a preset number of neighboring nodes, and aggregating the feature vectors corresponding to the sampled neighboring nodes to obtain the aggregated neighbor fusion feature vector; aggregating the node's own feature vector and the corresponding neighbor fusion feature vector to obtain the aggregated feature representation of each node.

[0181] It should be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive effort.

[0182] Those skilled in the art will clearly understand that for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0183] Example 3

[0184] Accordingly, an embodiment of the present invention provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the multi-stream data fusion method described in the above embodiment of the invention is implemented.

[0185] The electronic device may be a computing device such as a desktop computer, a notebook computer, a PDA, a cloud server, etc. The device may include, but is not limited to, a processor and a memory.

[0186] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the device and connects various parts of the entire device using various interfaces and lines.

[0187] Example 4

[0188] Accordingly, an embodiment of the present invention provides a storage medium, which includes a stored computer program. When the computer program is executed, the device where the storage medium is located is controlled to execute the multi-stream data fusion method described in the above embodiment of the invention.

[0189] The memory can be used to store the computer program, and the processor realizes various functions of the device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0190] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, it can implement the steps of the above-mentioned various method embodiments. The computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0191] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A multi-stream data fusion method, characterized in that: include: Acquiring multi-source data of the power system; wherein the multi-source data includes: power load, equipment status and meteorological data; Inputting the multi-source data into a preset missing value processing model, so that the missing value processing model extracts the time features and spatial features of the multi-source data, and performs feature fusion on the time features and spatial features to obtain corresponding spatiotemporal features, and then predicts and fills the missing values ​​in the multi-source data according to the spatiotemporal features, and outputs complete multi-source data; Based on the complete multi-source data, a corresponding knowledge graph is constructed with devices and users as nodes and the relationships between the devices and users as edges; Obtain feature information corresponding to each node in the knowledge graph, splice the feature information of each node into a corresponding feature vector, and then aggregate each node according to the feature vector corresponding to each node to obtain the aggregated feature representation of each node; wherein the feature information includes: power flow features, information flow features, and business flow features.

2. The multi-stream data fusion method according to claim 1, wherein: Before inputting the multi-source data into a preset missing value processing model, the method further includes: Extracting data fields corresponding to each data in the multi-source data; wherein the data fields include: timestamp, device ID, and measurement value; Compare the data fields of each data, and take all data with consistent data fields as a group of duplicate data, then retain any data in each group of duplicate data and delete the rest; Each data is compared with a preset data range, and data exceeding the data range is regarded as erroneous data, and all erroneous data are deleted.

3. The multi-stream data fusion method according to claim 1, wherein: The generation of the missing value processing model includes: Acquire historical multi-source data of the power system; Adding a preset mask matrix to the historical multi-source data, and using the original historical multi-source data and the historical multi-source data after adding the mask matrix as training samples; A preset spatiotemporal graph convolutional network model is trained according to the training samples to obtain the missing value processing model.

4. The multi-stream data fusion method according to claim 1, wherein: The step of aggregating the nodes according to the feature vector corresponding to each node to obtain the aggregated feature representation of each node includes: For each node, a preset number of neighboring nodes are sampled, and the feature vectors corresponding to the sampled neighboring nodes are aggregated to obtain the aggregated neighbor fusion feature vector; The node's own feature vector and the corresponding neighbor fusion feature vector are aggregated to obtain the aggregated feature representation of each node.

5. A multi-stream data fusion device, characterized in that: include: Data acquisition module, missing value processing module, knowledge graph construction module and node feature aggregation module; The data acquisition module is used to acquire multi-source data of the power system; wherein the multi-source data includes: power load, equipment status and meteorological data; The missing value processing module is used to input the multi-source data into a preset missing value processing model, so that the missing value processing model extracts the time features and spatial features of the multi-source data, and fuses the time features and spatial features to obtain corresponding spatiotemporal features, and then predicts and fills the missing values ​​in the multi-source data according to the spatiotemporal features, and outputs complete multi-source data; The knowledge graph construction module is used to construct a corresponding knowledge graph based on the complete multi-source data, with devices and users as nodes and the relationship between the devices and users as edges; The node feature aggregation module is used to obtain the feature information corresponding to each node in the knowledge graph, splice the feature information of each node into a corresponding feature vector, and then aggregate each node according to the feature vector corresponding to each node to obtain the aggregated feature representation of each node; wherein the feature information includes: power flow features, information flow features and business flow features.

6. The multi-stream data fusion device according to claim 5, characterized in that: Before inputting the multi-source data into a preset missing value processing model, the method further includes: Extracting data fields corresponding to each data in the multi-source data; wherein the data fields include: timestamp, device ID, and measurement value; Compare the data fields of each data, and take all data with consistent data fields as a group of duplicate data, then retain any data in each group of duplicate data and delete the rest; Each data is compared with a preset data range, and data exceeding the data range is regarded as erroneous data, and all erroneous data are deleted.

7. The multi-stream data fusion device according to claim 5, characterized in that: The generation of the missing value processing model includes: Acquire historical multi-source data of the power system; Adding a preset mask matrix to the historical multi-source data, and using the original historical multi-source data and the historical multi-source data after adding the mask matrix as training samples; A preset spatiotemporal graph convolutional network model is trained according to the training samples to obtain the missing value processing model.

8. The multi-stream data fusion device according to claim 5, wherein: The step of aggregating the nodes according to the feature vector corresponding to each node to obtain the aggregated feature representation of each node includes: For each node, a preset number of neighboring nodes are sampled, and the feature vectors corresponding to the sampled neighboring nodes are aggregated to obtain the aggregated neighbor fusion feature vector; The node's own feature vector and the corresponding neighbor fusion feature vector are aggregated to obtain the aggregated feature representation of each node.

9. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the multi-stream data fusion method according to any one of claims 1 to 4 when executing the computer program.

10. A storage medium, characterized in that: The storage medium includes a stored computer program, wherein when the computer program is executed, the device where the storage medium is located is controlled to execute the multi-stream data fusion method according to any one of claims 1 to 4.

Images