Abnormality detection method and system for redundancy-eliminating perception multi-dimensional power utilization data fusion

Abstract

The invention belongs to the field of electricity consumption information acquisition, and provides an anomaly detection method and system for redundancy elimination perception multi-dimensional electricity consumption data fusion, and the method comprises the steps: obtaining historical electricity consumption data, carrying out the preprocessing, and obtaining a time sequence feature vector of the historical electricity consumption data through a trained convolution recursion auto-encoder; based on the preprocessed historical power consumption data, a space mapping function is adopted to obtain a basic feature vector; based on the preprocessed historical power consumption data, obtaining an incidence relation feature vector by adopting a trained graph convolutional neural network model; fusing the time sequence feature vector, the basic feature vector and the incidence relation feature vector by adopting a multi-head attention mechanism to obtain a fused representation vector; filtering redundant information in the fusion representation vector by using the trained redundancy elimination fusion model to obtain a comprehensive representation vector of the time sequence power consumption data; and performing anomaly detection according to the comprehensive representation vector of the time sequence power consumption data to obtain a historical power consumption data evaluation result.

Description

Technical Field

[0001] The present invention belongs to the technical field of electricity consumption information collection, and in particular relates to an anomaly detection method and system for de-redundancy sensing multi-dimensional electricity consumption data fusion. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] With the rapid development of smart grid technology, the amount of data generated by the power system has increased dramatically. In the analysis of time-series power consumption data, anomaly detection aims to identify data points that deviate from the normal pattern, such as sudden load increases and decreases, equipment failures, etc. These anomalies not only affect the safety and operability of the power system, but may also cause economic losses or further damage to the equipment. In this context, how to effectively detect anomalies in time-series power consumption data has become the key to ensuring the stable operation of the power grid and improving the economic benefits of power companies.

[0004] In the power system, data usually exists in the form of time series, covering multiple parameters such as load, voltage, frequency, etc. These data are often affected by multiple factors such as environmental factors, equipment failure or human error, which may lead to abnormal situations. Traditional anomaly detection methods mostly rely on single features for analysis, lack of full mining of the correlation between multi-dimensional features, and therefore cannot fully reflect the complexity of electricity consumption data. In addition, the inventors also found that real-time series electricity consumption data show the characteristics of volatility and approximate periodicity, and electricity consumption data in different industries show different distribution characteristics, and existing methods cannot simultaneously capture the periodicity and volatility in time-series electricity consumption data, resulting in poor anomaly detection performance. Summary of the invention

[0005] In order to solve the above problems, the present invention proposes an anomaly detection method and system for de-redundancy-aware multi-dimensional electricity consumption data fusion. Based on the historical electricity consumption data of the acquisition system, the present invention adopts a convolutional recursive autoencoder to capture the periodicity and dynamics of the electricity consumption data, and obtains the time series feature vector of the electricity consumption data; based on the archival data in the acquisition system, a spatial mapping function is used to obtain the basic feature vector; then, a graph convolutional neural network is used to capture the potential correlation between the time series electricity consumption data, and obtain the correlation feature vector; subsequently, an attention mechanism is used to obtain the fusion representation vector of the time series electricity consumption data; finally, a de-redundancy module is designed to improve the performance of anomaly detection.

[0006] According to some embodiments, a first solution of the present invention provides an anomaly detection method for redundancy-aware multi-dimensional electricity consumption data fusion, which adopts the following technical solutions:

[0007] The anomaly detection method of de-redundancy-aware multi-dimensional electricity consumption data fusion includes:

[0008] Obtain historical electricity consumption data and preprocess it, use the trained convolutional recursive autoencoder to capture the periodicity and dynamics of the historical electricity consumption data, and obtain the time series feature vector of the historical electricity consumption data;

[0009] Based on the preprocessed historical electricity consumption data, a spatial mapping function is used to obtain the basic feature vector;

[0010] Based on the preprocessed historical electricity consumption data, the trained graph convolutional neural network model is used to capture the potential correlation between electricity consumption data and obtain the correlation feature vector;

[0011] A multi-head attention mechanism is used to fuse the time series feature vector, basic feature vector, and association feature vector from different dimensions to obtain a fused representation vector.

[0012] The de-redundancy fusion model is obtained by combining the gate control unit and the residual network. The redundant information in the fusion representation vector is filtered using the trained de-redundancy fusion model to obtain the comprehensive representation vector of the time series power consumption data.

[0013] Anomaly detection is performed based on the comprehensive representation vector of time-series electricity consumption data to obtain the evaluation results of historical electricity consumption data.

[0014] Furthermore, the trained convolutional recursive autoencoder is used to capture the periodicity and dynamics of the historical electricity consumption data, and the time series feature vector of the historical electricity consumption data is obtained, which is specifically:

[0015] Perform convolution operation based on the preprocessed historical electricity consumption data to extract time series features;

[0016] A convolutional LSTM neural network encoder is used to adaptively mine relevant feature vectors in time series features at different time steps.

[0017] For the relevant feature vectors, the convolution decoder is used to perform feature reconstruction calculation to obtain the time series feature vector of the historical electricity consumption data.

[0018] Furthermore, based on the preprocessed historical electricity consumption data, a spatial mapping function is used to obtain the basic feature vector, specifically:

[0019] Extract inherent features in electricity consumption data based on pre-processed historical electricity consumption data, including basic information of users, electricity consumption addresses and equipment information;

[0020] Based on the product of the spatial mapping function and the inherent features in the electricity usage data, a basic feature vector is determined.

[0021] Furthermore, when using the multi-head attention mechanism for feature fusion, the head decorrelation loss is constructed based on the covariance matrix to minimize the covariance between different attention heads and enhance the ability of fusion representation.

[0022] Furthermore, the covariance matrix-based head decorrelation loss is constructed to minimize the covariance between different attention heads, specifically:

[0023]

[0024] in,‖·‖ F is the Frobenius norm, and the purpose of the diag(·) function is to extract the vectors on the main diagonal of the matrix, the covariance matrix Φ.

[0025] Furthermore, the de-redundancy fusion model is obtained by combining the gating unit and the residual network, specifically:

[0026]

[0027] I=F fusion -[F temporal ⊕F basic ⊕F relationship ]

[0028] F final =norm(γ*F fusion +(1-γ)*I)

[0029] in, represents the activation function, Υ represents the gate vector, W g and b g represents the learnable parameter, I represents the residual vector, norm(·) represents the normalization operation, and F finall A comprehensive representation vector representing time series power consumption data.

[0030] According to some embodiments, the second solution of the present invention provides an anomaly detection system for redundancy-free sensing multi-dimensional electricity consumption data fusion, which adopts the following technical solutions:

[0031] The anomaly detection system for redundancy-free perception and fusion of multi-dimensional electricity consumption data includes:

[0032] A time series feature vector acquisition module is configured to acquire and preprocess historical electricity consumption data, and use a trained convolutional recursive autoencoder to capture the periodicity and dynamics of the historical electricity consumption data to obtain a time series feature vector of the historical electricity consumption data;

[0033] A basic feature vector acquisition module is configured to acquire a basic feature vector using a spatial mapping function based on the preprocessed historical power consumption data;

[0034] The correlation characteristic vector acquisition module is configured to capture the potential correlation between the electricity consumption data based on the preprocessed historical electricity consumption data by using the trained graph convolutional neural network model to obtain the correlation characteristic vector;

[0035] The multi-dimensional feature fusion vector acquisition module is configured to adopt a multi-head attention mechanism to fuse the time series feature vector, the basic feature vector and the association relationship feature vector from different dimensions to obtain a fused representation vector;

[0036] The comprehensive representation vector acquisition module is configured to obtain a de-redundancy fusion model by combining a gating unit and a residual network, and to filter redundant information in the fusion representation vector by using the trained de-redundancy fusion model to obtain a comprehensive representation vector of the time series power consumption data;

[0037] The time series power consumption data anomaly detection module is configured to perform anomaly detection based on the comprehensive representation vector of the time series power consumption data to obtain the historical power consumption data evaluation result.

[0038] According to some embodiments, a third aspect of the present invention provides a computer-readable storage medium.

[0039] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the anomaly detection method for de-redundancy-aware multi-dimensional electricity consumption data fusion as described in the first aspect above.

[0040] According to some embodiments, a fourth aspect of the present invention provides a computer device.

[0041] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the anomaly detection method for de-redundancy-aware multi-dimensional electricity consumption data fusion as described in the first aspect above are implemented.

[0042] A fifth aspect of the present invention provides a computer program product or a computer program.

[0043] The present invention provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the steps in the anomaly detection method for de-redundancy-aware multi-dimensional power consumption data fusion as described in the first aspect above.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] Based on the time-series electricity consumption data, the present invention adopts methods such as convolutional recursive autoencoder, graph convolutional neural network, multi-head attention mechanism, etc. to capture the periodicity and potential correlation of electricity consumption data from different dimensions, deeply explore the potential value behind the electricity consumption data, comprehensively understand the complexity of time-series electricity consumption data, and improve the accuracy and stability of prediction.

[0046] The present invention designs a redundancy removal module by combining the gate control unit and the residual network, which effectively suppresses the redundant information in the process of time series power consumption data fusion, alleviates the gradient vanishing problem in the modeling process, and improves the accuracy of anomaly detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0048] Figure 1 The overall training flow chart of the anomaly detection method for redundancy-free perception multi-dimensional electricity consumption data fusion provided by the embodiment of the present invention;

[0049] Figure 2 A flowchart of an example of abnormal detection of time-series power consumption data provided by an embodiment of the present invention;

[0050] Figure 3 A schematic diagram of the structure of an anomaly detection system for redundancy-free perception and fusion of multi-dimensional electricity consumption data provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0052] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0053] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0054] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0055] Embodiment 1

[0056] The present embodiment provides an anomaly detection method for de-redundancy-aware multi-dimensional electricity consumption data fusion. The present embodiment uses the method applied to a server as an example for illustration. It is understandable that the method can also be applied to a terminal, and can also be applied to a terminal, a server, and a system, and is implemented through the interaction between the terminal and the server. The server can be an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network servers, cloud communications, middleware services, domain name services, security services CDN, and big data and artificial intelligence platforms. The terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited to this. The terminal and the server can be directly or indirectly connected via wired or wireless communication, and this application is not limited here. The present embodiment relies on the Shandong Provincial Key Laboratory of Artificial Intelligence Applications for People's Livelihood Services and the Shandong Provincial General Artificial Intelligence Future Industry Laboratory, and the method includes the following steps:

[0057] Obtain historical electricity consumption data and preprocess it, use the trained convolutional recursive autoencoder to capture the periodicity and dynamics of the historical electricity consumption data, and obtain the time series feature vector of the historical electricity consumption data;

[0058] Based on the preprocessed historical electricity consumption data, a spatial mapping function is used to obtain the basic feature vector;

[0059] Based on the preprocessed historical electricity consumption data, the trained graph convolutional neural network model is used to capture the potential correlation between electricity consumption data and obtain the correlation feature vector;

[0060] A multi-head attention mechanism is used to fuse the time series feature vector, basic feature vector, and association feature vector from different dimensions to obtain a fused representation vector.

[0061] The de-redundancy fusion model is obtained by combining the gate control unit and the residual network. The redundant information in the fusion representation vector is filtered using the trained de-redundancy fusion model to obtain the comprehensive representation vector of the time series power consumption data.

[0062] Anomaly detection is performed based on the comprehensive representation vector of time-series electricity consumption data to obtain the evaluation results of historical electricity consumption data.

[0063] like Figure 1 and Figure 2 As shown, this embodiment provides a training process of an anomaly detection method for de-redundancy perception multi-dimensional electricity consumption data fusion, and takes the data collected in the electricity consumption information collection system of a State Grid company as an example to illustrate the method process described in this embodiment, including:

[0064] A. Taking historical electricity consumption data as the data source, the historical electricity consumption data in the electricity consumption information collection system is collected as sample data, and these sample data are preprocessed, including missing value marking, non-missing value standardization, data definition and storage, etc.

[0065] Specifically, the data collected from the power consumption information collection system of a provincial company of State Grid is used as the data set. The data set includes the power consumption data of 996 users (including 126 abnormal users) in a certain area from June to November 2024, and the user power consumption data is collected once every hour.

[0066] B. In this embodiment, the historical electricity consumption data of the electricity consumption information collection system is defined as X = {x 1 ,x 2 ,…,x T}, represents the time series data of T moments, and Y∈{0,1} represents the power consumption behavior of the time series power consumption data, where 0 represents normal power consumption and 1 represents abnormal power consumption behavior.

[0067] C. Obtain historical electricity consumption data and preprocess it, use the trained convolutional recursive autoencoder to capture the periodicity and dynamics of the historical electricity consumption data, and obtain the time series feature vector of the historical electricity consumption data, specifically:

[0068] Perform convolution operation based on the preprocessed historical electricity consumption data to extract time series features;

[0069] A convolutional LSTM neural network encoder is used to adaptively mine relevant feature vectors in time series features at different time steps.

[0070] For the relevant feature vectors, the convolution decoder is used to perform feature reconstruction calculation to obtain the time series feature vector of the historical electricity consumption data.

[0071] The convolutional recursive autoencoder combines the powerful features of convolutional neural networks (CNN) and recursive neural networks (RNN). Its core concept is to extract the local features of time-series power consumption data through multi-scale convolution operations, and use the recursive structure to process the time dependency and periodicity of power consumption data. It mainly consists of two parts: encoder and decoder.

[0072] C1. First, based on the historical power consumption data of the power consumption information collection system and preprocessing, this embodiment uses convolution operation to extract time series features. Assume X t,l-1 represents the feature map of the l-1th layer, and the output vector X of the lth layer t,l for:

[0073] Xt,l =f(W l *X t,l-1 +b l ) (1);

[0074] Among them, * is the convolution operation, f(·) is the activation function, and W l and b l represents learnable parameters.

[0075] C2. The feature vector generated by the convolution operation in step C1 depends on the previous time step in time, but its performance may deteriorate as the sequence length increases. To solve this problem, this embodiment selects the long short-term memory neural network (Long Short-Term Memory, LSTM) in the recursive neural network and designs an LSTM fused with convolution operation, which can adaptively mine relevant feature vectors through different time steps. The specific implementation process is as follows:

[0076]

[0077]

[0078] Among them, * represents the convolution operation, represents the Hadamard product of two vector matrices, and σ represents the activation function. Represents a parameter matrix. and represents the bias matrix. t,l represents the LSTM forget gate state vector, r t,l represents the input gate state vector in LSTM, C t,l represents the hidden state vector in LSTM, O t,l Represents the output gate state vector in LSTM, H t-1,l represents the state vector output at time t-1 before, H t,l Represents the encoder output vector calculated by the convolutional LSTM.

[0079] C3. Based on the vector H calculated by the convolutional LSTM encoder in step C2, this embodiment uses a convolutional decoder to perform feature reconstruction calculation to obtain the time series feature vector F of the electricity consumption data temporal :

[0080]

[0081] in, and represents learnable parameters.

[0082] D. Based on the preprocessed historical electricity consumption data, a spatial mapping function is used to obtain the basic feature vector, specifically:

[0083] Extract inherent features in electricity consumption data based on pre-processed historical electricity consumption data, including basic information of users, electricity consumption addresses and equipment information;

[0084] Based on the product of the spatial mapping function and the inherent features in the electricity usage data, a basic feature vector is determined.

[0085] Basic features refer to the inherent features in electricity consumption data, including basic information of users, electricity consumption addresses, equipment information, etc. Basic feature vector F basic The calculation formula is as follows:

[0086] F basic =X basic ×U(8);

[0087] Among them, X basic represents the inherent characteristics of electricity consumption data, and U represents the spatial mapping function.

[0088] E. Based on the preprocessed historical electricity consumption data, the trained graph convolutional neural network model is used to capture the potential correlation between electricity consumption data, thereby obtaining the correlation representation vector.

[0089] E1. Based on historical electricity consumption data, this embodiment first constructs an electricity knowledge graph in, Representation diagram Node, E is a graph The edge of A is a graph Adjacency matrix. If e ij ∈E, then A ij =1, otherwise A ij = 0. X represents node attribute information.

[0090] E2. In order to fully explore the potential correlation between electricity consumption data, based on the power knowledge graph constructed in step E1, this embodiment uses a graph convolutional network to capture the correlation between electricity consumption data and obtain the electricity consumption data correlation relationship representation vector F relationship The graph convolution operation of the node is as follows:

[0091]

[0092] Among them, H τ Represents the node output of the τth layer. ∈ represents the deviation matrix. represents a diagonal matrix, W τrepresents the parameter matrix of the τth layer. δ(·) represents the activation function.

[0093] F. Based on the time series feature vector F obtained in step C temporal , the basic feature vector F obtained in step D basic And step E to obtain the association feature vector F relationship , a decorrelated multi-head attention mechanism is used to achieve multi-dimensional feature fusion from multiple levels to obtain the fusion representation vector F of the time series power consumption data fusion The design of multi-head attention is as follows:

[0094] Q=F temporal (10);

[0095] K=F basic (11);

[0096] V=F relationship (12);

[0097]

[0098] F fusion =Concat(α 1 ,α 2 ,…,α n )W o (14);

[0099] Where Q, K and V represent the query vector, key vector and value vector respectively. i represents the attention score. k W represents the embedding dimension. i Q , W i K , W i V , W o are all weight matrices. Concat(·) represents a concatenation operation. fusion Represents the fused vector obtained through the multi-head attention mechanism.

[0100] The multi-head attention mechanism captures different aspects of their dependencies through the heads of self-attention. However, in practice, the inventors found that the heads of this mechanism may be more inclined to focus on similar dependencies. Therefore, this embodiment uses a head decorrelation module to improve the ability of the multi-head attention mechanism to learn different aspects of features, automatically reducing the influence of redundant information and irrelevant features when learning features, thereby enhancing the fusion representation ability of the model.

[0101] According to the above formula, we can get the output F of the multi-head attention mechanism fusion In order to simplify the complexity of the following formula, this embodiment replaces F with Ω.fusion The covariances between all activation pairs i and j in Ω form a matrix Φ.

[0102]

[0103] in, It defines the sample mean, Λ is the batch size, It is defined as the i of Ω in this batch th Activation status.

[0104] Then, based on the covariance matrix Φ, a head decorrelation loss is constructed To minimize the covariance between different attention heads, thereby enhancing the ability of fusion representation.

[0105]

[0106] in,‖·‖ F is the Frobenius norm, and the purpose of the diag(·) function is to extract the vector of the main diagonal of the matrix.

[0107] G. Based on the fusion representation vector F obtained in step E fusion Since it is difficult to objectively quantify the redundant information in the fusion vector, this embodiment merges the gate control unit into the residual structure and designs a redundancy removal fusion to remove the possible repeated semantic information, thereby obtaining a comprehensive representation vector of the time series power consumption data. The definition of the redundancy removal module is as follows:

[0108]

[0109] I=F fusion -[F temporal ⊕F basic ⊕F relationship ] (18);

[0110] F final =norm(γ*F fusion +(1-γ)*I) (19);

[0111] in, represents the activation function; γ represents the gate vector, W g and b g represents the learnable parameter, I represents the residual vector, norm(·) represents the normalization operation, and F finall A comprehensive representation vector representing time series power consumption data.

[0112] H. Comprehensively represent the time series power consumption data obtained in step G as vector F final , input the softmax function to detect anomalies in time series electricity consumption data:

[0113] Y = softmax(W final F final +b final ) (20);

[0114] Among them, W final and b final is the learning parameter of the softmax function, and Y represents the anomaly detection value output by the model.

[0115] I. Use cross entropy as the loss function for model discrimination. If Y ′ is the true category distribution, then the loss function is defined as follows:

[0116]

[0117] J. Through steps F and I, the overall loss function of the model is defined as follows:

[0118]

[0119] Among them, L total is the overall loss function value of the model. 1 and β 2 is a hyperparameter. Represents a set of parameters. ‖·‖ 2 It represents L2 regularization, which can effectively prevent the model from overfitting. Through the iterative solution of step F and step I, L is continuously minimized. total , thereby obtaining the learning parameters of the model proposed in this embodiment.

[0120] Conduct risk assessment on the operating status of the electricity consumption information collection system for the test samples, push the assessment results, and compare them with the thresholds set by the actual electricity consumption information collection system, and feedback to update the underlying data information, so as to continuously optimize the data weight values ​​in the model and continuously improve the anomaly detection of time-series electricity consumption data.

[0121] Embodiment 2

[0122] This embodiment provides an anomaly detection system for redundancy-free perception and fusion of multi-dimensional electricity consumption data, including:

[0123] A time series feature vector acquisition module is configured to acquire and preprocess historical electricity consumption data, and use a trained convolutional recursive autoencoder to capture the periodicity and dynamics of the historical electricity consumption data to obtain a time series feature vector of the historical electricity consumption data;

[0124] A basic feature vector acquisition module is configured to acquire a basic feature vector using a spatial mapping function based on the preprocessed historical power consumption data;

[0125] The correlation characteristic vector acquisition module is configured to capture the potential correlation between the electricity consumption data based on the preprocessed historical electricity consumption data by using the trained graph convolutional neural network model to obtain the correlation characteristic vector;

[0126] The multi-dimensional feature fusion vector acquisition module is configured to adopt a multi-head attention mechanism to fuse the time series feature vector, the basic feature vector and the association relationship feature vector from different dimensions to obtain a fused representation vector;

[0127] The comprehensive representation vector acquisition module is configured to obtain a de-redundancy fusion model by combining a gating unit and a residual network, and to filter redundant information in the fusion representation vector by using the trained de-redundancy fusion model to obtain a comprehensive representation vector of the time series power consumption data;

[0128] The time series power consumption data anomaly detection module is configured to perform anomaly detection based on the comprehensive representation vector of the time series power consumption data to obtain the historical power consumption data evaluation result.

[0129] The examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the contents disclosed in the above embodiment 1. It should be noted that the above modules as part of the system can be executed in a computer system such as a set of computer executable instructions.

[0130] The description of each embodiment in the above embodiments has different emphases. For parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0131] The proposed system can be implemented in other ways. For example, the system embodiment described above is only illustrative, and the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0132] Embodiment 3

[0133] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the anomaly detection method for de-redundancy-aware multi-dimensional electricity consumption data fusion as described in the first embodiment above are implemented.

[0134] Embodiment 4

[0135] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the anomaly detection method for de-redundancy-aware multi-dimensional electricity consumption data fusion as described in the first embodiment are implemented.

[0136] Embodiment 5

[0137] This embodiment provides a computer program product or a computer program, which includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the steps in the anomaly detection method for de-redundancy-aware multi-dimensional power consumption data fusion described in the first embodiment.

[0138] 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 hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0139] 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.

[0140] 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.

[0141] 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 in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0142] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0143] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. An anomaly detection method for de-redundancy-aware multi-dimensional electricity consumption data fusion, characterized in that: include: Obtain historical electricity consumption data and preprocess it, use the trained convolutional recursive autoencoder to capture the periodicity and dynamics of the historical electricity consumption data, and obtain the time series feature vector of the historical electricity consumption data; Based on the preprocessed historical electricity consumption data, a spatial mapping function is used to obtain the basic feature vector; Based on the preprocessed historical electricity consumption data, the trained graph convolutional neural network model is used to capture the potential correlation between electricity consumption data and obtain the correlation feature vector; A multi-head attention mechanism is used to fuse the time series feature vector, basic feature vector, and association feature vector from different dimensions to obtain a fused representation vector. The de-redundancy fusion model is obtained by combining the gate control unit and the residual network. The redundant information in the fusion representation vector is filtered using the trained de-redundancy fusion model to obtain the comprehensive representation vector of the time series power consumption data. Anomaly detection is performed based on the comprehensive representation vector of time-series electricity consumption data to obtain the evaluation results of historical electricity consumption data.

2. The method for detecting anomalies by de-redundancy sensing multi-dimensional electricity consumption data fusion according to claim 1, characterized in that: The trained convolutional recursive autoencoder is used to capture the periodicity and dynamics of the historical electricity consumption data, and the time series feature vector of the historical electricity consumption data is obtained, which is specifically: Perform convolution operation based on the preprocessed historical electricity consumption data to extract time series features; A convolutional LSTM neural network encoder is used to adaptively mine relevant feature vectors in time series features at different time steps. For the relevant feature vectors, the convolution decoder is used to perform feature reconstruction calculation to obtain the time series feature vector of the historical electricity consumption data.

3. The method for detecting anomalies by de-redundancy sensing multi-dimensional electricity consumption data fusion according to claim 1, characterized in that: Based on the preprocessed historical electricity consumption data, the spatial mapping function is used to obtain the basic feature vector, specifically: Extract inherent features in electricity consumption data based on pre-processed historical electricity consumption data, including basic information of users, electricity consumption addresses and equipment information; Based on the product of the spatial mapping function and the inherent features in the electricity usage data, a basic feature vector is determined.

4. The method for detecting anomalies by de-redundancy sensing multi-dimensional electricity consumption data fusion according to claim 1, characterized in that: When using the multi-head attention mechanism for feature fusion, the head decorrelation loss is constructed based on the covariance matrix to minimize the covariance between different attention heads and enhance the ability of fusion representation.

5. The method for detecting anomalies by de-redundancy sensing multi-dimensional electricity consumption data fusion according to claim 4, characterized in that: The head decorrelation loss is constructed based on the covariance matrix to minimize the covariance between different attention heads, specifically: in,‖·‖ F is the Frobenius norm, and the purpose of the diag(·) function is to extract the vectors on the main diagonal of the matrix, the covariance matrix Φ.

6. The method for detecting anomalies by de-redundancy sensing multi-dimensional electricity consumption data fusion according to claim 1, characterized in that: The de-redundancy fusion model is obtained by combining the gate control unit and the residual network, specifically: I=F fusion -[F temporal ⊕F basic ⊕F relationship ] F final =norm(Y*F fusion +(1-Y)*I) in, represents the activation function, Υ represents the gate vector, W g and b g represents the learnable parameter, I represents the residual vector, norm(·) represents the normalization operation, and F finall A comprehensive representation vector representing time series power consumption data.

7. Redundancy-removing perception multi-dimensional electricity consumption data fusion anomaly detection system, characterized by: include: A time series feature vector acquisition module is configured to acquire and preprocess historical electricity consumption data, and use a trained convolutional recursive autoencoder to capture the periodicity and dynamics of the historical electricity consumption data to obtain a time series feature vector of the historical electricity consumption data; A basic feature vector acquisition module is configured to acquire a basic feature vector using a spatial mapping function based on the preprocessed historical power consumption data; The correlation characteristic vector acquisition module is configured to capture the potential correlation between the electricity consumption data based on the preprocessed historical electricity consumption data by using the trained graph convolutional neural network model to obtain the correlation characteristic vector; The multi-dimensional feature fusion vector acquisition module is configured to adopt a multi-head attention mechanism to fuse the time series feature vector, the basic feature vector and the association relationship feature vector from different dimensions to obtain a fused representation vector; The comprehensive representation vector acquisition module is configured to obtain a de-redundancy fusion model by combining a gating unit and a residual network, and to filter redundant information in the fusion representation vector by using the trained de-redundancy fusion model to obtain a comprehensive representation vector of the time series power consumption data; The time series power consumption data anomaly detection module is configured to perform anomaly detection based on the comprehensive representation vector of the time series power consumption data to obtain the historical power consumption data evaluation result.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the anomaly detection method for de-redundancy-aware multi-dimensional electricity consumption data fusion as described in any one of claims 1 to 6 are implemented.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the anomaly detection method for de-redundancy-aware multi-dimensional electricity consumption data fusion as described in any one of claims 1-6 are implemented.

10. A computer program product, characterized in that The computer program product includes a computer program, and when the computer program is executed by a processor, the steps in the anomaly detection method for de-redundancy-aware multi-dimensional electricity consumption data fusion as described in any one of claims 1-6 are implemented.

Images