Power utilization scheme recommendation method, device and equipment based on power consumer group

Abstract

The invention provides a power utilization scheme recommendation method, device and equipment based on a power consumer group, and relates to the technical field of power. The method comprises the steps of obtaining p pieces of multi-modal data of a target power consumer indicated by a group division task, and performing multi-view feature extraction on the p pieces of multi-modal data based on m pre-specified views; constructing m feature matrixes according to the extracted data features of the data of the p modals under the m views; then, a cross-view shared subspace is determined based on the m feature matrixes, and a group division matrix is generated based on the shared subspace. And further, performing group division on the target power users based on the group division matrix to obtain at least one power user group, and determining and recommending a target power utilization scheme corresponding to the power user group. According to the method and the device, group division can be accurately performed on the target power users, so that a highly personalized power utilization scheme is provided for the power user group, and the accuracy of power utilization scheme recommendation is improved.

Description

Technical Field

[0001] The present application relates to the field of electric power technology, and in particular to a method, device and equipment for recommending an electricity usage plan based on an electric power user group. Background Art

[0002] With the rapid development of social economy and the continuous advancement of smart grid technology, the electricity consumption behavior and characteristics of power users are increasingly valued. By analyzing the electricity consumption behavior and characteristics of power users, personalized electricity consumption plans can be recommended for power users, and accurate classification of power user groups is a key step in achieving efficient and accurate electricity consumption plan recommendations.

[0003] Since the electricity consumption behaviors and characteristics of each electricity user group are relatively consistent, by dividing electricity users into groups, more targeted electricity consumption plans can be tailored for each electricity user group. Existing methods for dividing electricity user groups usually rely on single-view data. For example, some methods divide groups based on the historical electricity consumption data of electricity users, such as the electricity consumption and electricity consumption patterns of electricity users in different time periods. Other methods may use basic demographic information, such as the age and income level of electricity users, to make a rougher group division.

[0004] Therefore, the existing technology does not accurately classify electricity users into groups, and is thus unable to provide highly personalized electricity usage plans for electricity user groups, resulting in low accuracy in electricity usage plan recommendations. Summary of the invention

[0005] The present application provides a method, device and equipment for recommending electricity consumption plans based on electricity user groups, so as to solve the technical problem that the defect identification results of power equipment in the prior art are not comprehensive and accurate enough.

[0006] In a first aspect, the present application provides a method for recommending a power consumption plan based on a power user group, comprising:

[0007] In response to a group segmentation task for target power users, multimodal data of the target power users indicated by the group segmentation task is acquired; wherein the multimodal data includes data of p modes, where p is a positive integer greater than 1;

[0008] Based on the pre-specified m views, multi-view feature extraction is performed on the data of the p modalities respectively to obtain data features of the data of the p modalities under the m views; based on the data features of the data of the p modalities under the m views, m feature matrices are constructed; wherein m is a positive integer greater than 1;

[0009] Based on the m feature matrices, a shared subspace across views is determined; based on the shared subspace, a group partition matrix for the target power user is generated; wherein the shared subspace represents a low-dimensional space that can simultaneously represent data features of different modalities and views;

[0010] The target power users are divided into groups based on the group division matrix to obtain at least one power user group; a target power usage plan corresponding to the power user group is determined, and the target power usage plan corresponding to the power user group is recommended to the power user group.

[0011] In a possible design, determining a shared subspace across views based on the m feature matrices includes:

[0012] Determine the contribution of each feature matrix to the group segmentation task, and determine the weight value of each feature matrix based on the contribution; wherein each feature matrix corresponds to a weight value;

[0013] Based on the m feature matrices and the m weight values ​​corresponding to the m feature matrices, a shared subspace across views is determined.

[0014] In a possible design, generating a group partition matrix for the target power user based on the shared subspace includes:

[0015] Based on a preset method, the m feature matrices are mapped to the shared subspace, and a target mapping function is constructed for the mapping process; wherein the mapping process is used to indicate that the m feature matrices are mapped to the shared subspace;

[0016] A first optimization process is performed on the target mapping function, and based on the optimized target mapping function, a group division matrix for the target power users is generated.

[0017] In a possible design, performing a first optimization process on the target mapping function includes:

[0018] Introducing a preset independence criterion into the target mapping function to obtain a corresponding first optimization function;

[0019] Re-mapping the m feature matrices to the shared subspace based on the first optimization function, and calculating feature independence of the m feature matrices in the shared subspace;

[0020] Determine a preset independence threshold of the shared subspace. If the feature independence is greater than or equal to the preset independence threshold, continue to optimize the first optimization function. If the feature independence is less than the preset independence threshold, adjust the preset independence criterion until the feature independence is greater than or equal to the preset independence threshold.

[0021] In a possible design, continuing to optimize the first optimization function includes:

[0022] Introducing a preset local structure constraint into the first optimization function to obtain a second optimization function; determining local structure information of each feature matrix based on the second optimization function;

[0023] A preset graph construction constraint is introduced into the second optimization function to obtain a third optimization function; based on the third optimization function, the m feature matrices are mapped to the shared subspace to retain local structural information of the m feature matrices in the shared subspace.

[0024] In a possible design, generating a group division matrix for the target power users based on the optimized target mapping function includes:

[0025] Obtaining m optimized feature matrices mapped to the shared subspace based on the optimized target mapping function;

[0026] Based on the m optimized feature matrices, a corresponding sparse common graph matrix is ​​generated; and a second optimization process is performed on the sparse common graph matrix to obtain the group partition matrix.

[0027] In a possible design, constructing a target mapping function for the mapping process includes:

[0028] Based on a preset subspace reconstruction technique, determining the number of data features of the m feature matrices mapped to the shared subspace;

[0029] Determining a preset number of feature maps of the shared subspace; if the number of data features is greater than or equal to the preset number of feature maps, constructing a target mapping function for the mapping process;

[0030] If the number of data features is less than the preset number of feature mappings, the preset method is adjusted until the number of data features is greater than or equal to the preset number of feature mappings; based on the adjusted preset method, the m feature matrices are remapped to the shared subspace, and a target mapping function is constructed for the mapping process.

[0031] In a second aspect, the present application provides a device for recommending electricity consumption plans based on electricity user groups, comprising:

[0032] An acquisition module, configured to, in response to a group segmentation task for target power users, acquire multimodal data of target power users indicated by the group segmentation task; wherein the multimodal data includes data of p modes, where p is a positive integer greater than 1;

[0033] A processing module, configured to perform multi-view feature extraction on the data of the p modalities based on the pre-specified m views, respectively, to obtain data features of the data of the p modalities under the m views; and construct m feature matrices based on the data features of the data of the p modalities under the m views; wherein m is a positive integer greater than 1;

[0034] A determination module, configured to determine a shared subspace across views based on the m feature matrices;

[0035] The processing module is further used to generate a group partition matrix for the target power user based on the shared subspace; wherein the shared subspace represents a low-dimensional space that can simultaneously represent data features of different modalities and views;

[0036] A division module, configured to divide the target power users into groups based on the group division matrix to obtain at least one power user group;

[0037] The determination module is further used to determine a target electricity consumption plan corresponding to the electricity user group;

[0038] The recommendation module is used to recommend a target electricity usage plan corresponding to the electricity user group to the electricity user group.

[0039] In a possible design, the determining module is further used to:

[0040] Determine the contribution of each feature matrix to the group segmentation task, and determine the weight value of each feature matrix based on the contribution; wherein each feature matrix corresponds to a weight value;

[0041] Based on the m feature matrices and the m weight values ​​corresponding to the m feature matrices, a shared subspace across views is determined.

[0042] In a possible design, the processing module further includes: a mapping module and a building module.

[0043] The mapping module is used to map the m feature matrices to the shared subspace based on a preset method;

[0044] The construction module is used to construct a target mapping function for a mapping process; wherein the mapping process is used to indicate mapping the m feature matrices to the shared subspace;

[0045] The processing module is further used to perform a first optimization process on the target mapping function, and generate a group division matrix for the target power users based on the optimized target mapping function.

[0046] In a possible design, the processing module further includes: an introduction module, configured to introduce a preset independence criterion into the target mapping function to obtain a corresponding first optimization function;

[0047] The mapping module is further used to re-map the m feature matrices to the shared subspace based on the first optimization function;

[0048] The processing module further includes: a calculation module, configured to calculate the feature independence of the m feature matrices in the shared subspace;

[0049] The determination module is further used to determine a preset independence threshold of the shared subspace;

[0050] The processing module is also used to continue optimizing the first optimization function if the feature independence is greater than or equal to the preset independence threshold; if the feature independence is less than the preset independence threshold, adjust the preset independence criterion until the feature independence is greater than or equal to the preset independence threshold.

[0051] In a possible design, the introducing module is further used to introduce a preset local structural constraint into the first optimization function to obtain a second optimization function;

[0052] The determination module is further used to determine the local structure information of each feature matrix based on the second optimization function;

[0053] The introducing module is further used to introduce a preset graph construction constraint into the second optimization function to obtain a third optimization function;

[0054] The mapping module is further used to map the m feature matrices to the shared subspace based on the third optimization function, so as to retain local structural information of the m feature matrices in the shared subspace.

[0055] In a possible design, the acquisition module is further used to acquire m optimized feature matrices mapped to the shared subspace based on the optimized target mapping function;

[0056] The processing module is further used to generate a corresponding sparse common graph matrix based on the m optimized feature matrices; and perform a second optimization process on the sparse common graph matrix to obtain the group partition matrix.

[0057] In a possible design, the determination module is further used to determine the number of data features of the m feature matrices mapped to the shared subspace based on a preset subspace reconstruction technique; determine the number of preset feature mappings of the shared subspace;

[0058] The construction module is further configured to construct a target mapping function for the mapping process if the number of data features is greater than or equal to the preset number of feature mappings;

[0059] The processing module is also used to adjust the preset method if the number of data features is less than the preset feature mapping number until the number of data features is greater than or equal to the preset feature mapping number; re-map the m feature matrices to the shared subspace based on the adjusted preset method, and construct a target mapping function for the mapping process.

[0060] In a third aspect, an embodiment of the present application provides an electronic device, comprising: at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method described in the first aspect above and various possible designs.

[0061] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the method described in the first aspect and various possible designs is implemented.

[0062] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the method described in the first aspect and various possible designs of the first aspect.

[0063] The method, device and equipment for recommending a power consumption plan based on a power user group provided by the present application respond to a group segmentation task for a target power user, and obtain p multimodal data of the target power user indicated by the group segmentation task. Wherein, p is a positive integer greater than 1. Multimodal data can comprehensively consider different types of data, can capture complex relationships that cannot be identified by a single type of data, and more comprehensively reflect the power consumption behavior and characteristics of the target power user. Next, based on the pre-specified m views, multi-view feature extraction is performed on the data of the p modes respectively, and the data features of the data of the p modes under the m views are obtained. Wherein, m is a positive integer greater than 1. Multi-view feature extraction involves analyzing and processing data from different perspectives or dimensions. Multi-view feature extraction based on multimodal data can further enhance the ability of data analysis and application, and provide a basis for the subsequent accurate group segmentation of target power users. Afterwards, m feature matrices are constructed based on the data features of the p modal data under the m views. Based on the constructed m feature matrices, a shared subspace across views is determined, and based on the shared subspace, a group segmentation matrix for the target power user is generated. Further, the target power users are divided into groups based on the group division matrix to obtain at least one power user group. The target power usage plan corresponding to the power user group is determined, and the corresponding target power usage plan is recommended to the power user group. The present application can accurately divide the target power users into groups, thereby providing a highly personalized power usage plan for the power user group, and improving the accuracy of the power usage plan recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0065] Figure 1 Schematic diagram of the process of the method for recommending power consumption plans based on power user groups provided in the embodiment of the present application Figure 1 ;

[0066] Figure 2 A schematic diagram of the structure of a method for recommending a power consumption plan based on a power user group provided in an embodiment of the present application;

[0067] Figure 3 Schematic diagram of the process of the method for recommending power consumption plans based on power user groups provided in the embodiment of the present application Figure 2 ;

[0068] Figure 4 A schematic diagram of the structure of a device for recommending electricity consumption plans based on electricity user groups provided in an embodiment of the present application;

[0069] Figure 5A hardware structure diagram of an electronic device provided in an embodiment of the present application.

[0070] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0071] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0072] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in sequences other than those illustrated or described herein, for example.

[0073] In the embodiments of the present application, the words "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0074] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0075] With the rapid development of social economy and the advancement of smart grid technology, the electricity consumption behavior of power users has become more complex and diversified. Since understanding the electricity consumption behavior and characteristics of power users can better meet user needs, the electricity consumption behavior and characteristics of power users have received more and more attention.

[0076] By analyzing the electricity usage behavior and characteristics of electricity users, personalized electricity usage plans can be provided to electricity users. This means recommending the most appropriate electricity usage plans or energy-saving measures to them based on their specific needs and habits, thereby improving their satisfaction and electricity usage efficiency.

[0077] The electricity consumption behaviors and needs of power users may vary greatly. In order to effectively provide each user with a suitable electricity consumption plan, it is necessary to divide the power users into groups. The group division of power users can identify the specific needs and behavior patterns of different power users, and accurate power user group division is the basis for recommending personalized power consumption plans. Because the electricity consumption behaviors and characteristics of each power user group are relatively consistent, accurate power user group division can tailor more targeted power consumption plans for each power user group.

[0078] Existing methods for segmenting power user groups usually rely on a single type of data view. For example, some methods segment power users based only on their historical power consumption data, such as their power consumption in different time periods and their power consumption patterns.

[0079] Although this method can provide certain electricity consumption behaviors and characteristics, it cannot fully capture the multi-dimensional characteristics of electricity users by relying solely on historical electricity consumption data, and the understanding of electricity users' electricity consumption behaviors is not comprehensive enough, making it difficult to meet the requirements of modern power systems for electricity consumption behavior analysis.

[0080] The electricity consumption behavior of electricity users may be affected by many factors, such as seasonal changes, changes in family members, changes in lifestyle, etc. These factors are difficult to fully reflect only by historical electricity consumption data.

[0081] Some other methods use basic demographic information, such as the age and income level of electricity users, to divide electricity users into groups. Although this method can provide some user background information, it is too rough and cannot reflect the specific electricity usage habits and needs of electricity users.

[0082] Due to the limitations of existing methods, the group classification of power users is often not accurate enough. This inaccuracy makes it impossible to provide power users with highly personalized power consumption plans, which in turn leads to low accuracy in power consumption plan recommendations, and the recommended power consumption plans may not be suitable for users' actual needs.

[0083] In view of the above technical problems, in order to solve the problems of inaccurate group division of power users caused by relying on single view data and low accuracy of power consumption plan recommendation, the inventors thought of collecting multimodal data of power users to comprehensively reflect the power consumption behavior and characteristics of power users. At the same time, in order to further improve the accuracy of group division of power users, the inventors considered extracting multi-view features from the collected multimodal data to obtain data features of multiple modal data under multiple views. Then, the power users are accurately divided into groups based on the extracted data features, and accurate power consumption plan recommendations are further achieved based on the group division results.

[0084] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0085] The embodiment of the present application provides a method for recommending electricity usage plans based on electricity user groups. Figure 1 Schematic diagram of the process of the method for recommending power consumption plans based on power user groups provided in the embodiment of the present application Figure 1 ,like Figure 1 As shown, the method for recommending a power consumption plan based on power user groups includes:

[0086] S101. In response to a group segmentation task for target power users, multimodal data of the target power users indicated by the group segmentation task is acquired; wherein the multimodal data includes data of p modes, and p is a positive integer greater than 1.

[0087] Among them, multimodal data refers to different types of data, such as image data, text data, video data, etc. Multimodal data integrates multiple forms of information to provide more comprehensive and multi-dimensional user information.

[0088] There are many ways to obtain the multimodal data of the target power users, for example, using Python crawler technology to obtain the multimodal data of the target power users, or directly exporting the multimodal data of the target power users from the relevant application background, etc. The specific acquisition method is determined by actual needs and is not specifically limited here.

[0089] S102. Based on the pre-specified m views, perform multi-view feature extraction on the data of p modalities respectively to obtain data features of the data of p modalities under the m views; and construct m feature matrices based on the data features of the data of p modalities under the m views.

[0090] Wherein, m is a positive integer greater than 1. In a specific example, if the acquired multimodal data includes image data and text data, it is necessary to perform feature extraction of m views of the image data and the text data. The image data may be images of power equipment (such as electric meters, household appliances, etc.) used by the target power user, exterior and interior photos of the residence, and images of the surrounding environment (including commercial areas, industrial areas, and residential areas, etc.).

[0091] The text data can be the basic information of the target power user, such as age, gender, family size, occupation and other demographic information, which is used to provide the basic background characteristics of the target power user. It can also be the power consumption record of the target power user, such as historical power consumption data, power bills, power consumption habit descriptions, etc., which is used to reflect the power consumption behavior of the target power user. It can also be the feedback and comments of the target power user, such as feedback, suggestions, user experience and other text information on power services, which is used to understand the needs and satisfaction of the target power user. In addition, the descriptions and opinions of the target power user in social media or questionnaires can also be collected to reflect the target power user's views on power consumption and power services.

[0092] Specifically, multi-view feature extraction is performed on image data, including but not limited to the following views: local binary patterns (LBP), global image features (GIST), scale invariant feature transform (SIFT), histograms of oriented gradients (HOG), color texture moments (CTM), and learning image features from image data through models such as convolutional neural networks (CNN).

[0093] Perform multi-view feature extraction on text data, including but not limited to the following views: term frequency (TF), n-gram, word vector (Word2Vec), text topic, text sentiment analysis, and learning text features from text data through models such as Recurrent Neural Network (RNN). Among them, n-gram divides text data into n consecutive words and counts the frequency of n words as feature representation. Word2Vec maps the words in text data to a low-dimensional vector space, connects the words in the context, and uses vectors to represent the semantic information of the words. Text topics such as latent dirichlet allocation (LDA) regard text data as a probability distribution composed of multiple topics, and use topic distribution as text features. Text sentiment analysis extracts positive and negative emotions as text features by analyzing the emotional color of text data.

[0094] It should be noted that any number of image feature extraction views and any number of text feature extraction views can be selected, but the total number of image feature extraction views and text feature extraction views must be m.

[0095] For the convenience of description, the image features extracted from the image data and the text features extracted from the text data are collectively referred to as the multi-view dataset X = {X 1 v ,X 2 v ,…,X n v}∈R dv×n . Among them, 1≤v≤m, X v represents the feature of the vth view, n represents the number of target power users, dv represents the feature dimension of the vth view, and m represents the number of views.

[0096] S103 . Determine a shared subspace across views based on the m feature matrices; and generate a group partition matrix for target power users based on the shared subspace.

[0097] It should be noted that each feature matrix is ​​useful for the task of group segmentation for target power users, but the contribution of each feature matrix to the task of group segmentation is different. Therefore, it is necessary to determine the contribution of each feature matrix to the task of group segmentation, and determine the weight value of each feature matrix based on the contribution. Among them, each feature matrix corresponds to a weight value. After that, the shared subspace across views can be determined based on the m feature matrices and the m weight values ​​corresponding to the m feature matrices.

[0098] Explanatory, the shared subspace across views is a low-dimensional space that can simultaneously represent data features of different modalities and views, aiming to capture and represent common features from different modalities or views. By projecting data features of different modalities or views into the shared subspace, alignment and comparison between different modalities or views can be achieved.

[0099] In a specific example, matrix decomposition technology can be used to mine the shared subspace across views from m feature matrices. Matrix decomposition technology is a technology that decomposes a matrix into the product of multiple matrices, which is used for dimensionality reduction, feature extraction and data compression. Common matrix decomposition technologies include singular value decomposition, non-negative matrix decomposition, principal component analysis, etc.

[0100] After determining the shared subspace, the m feature matrices can be mapped to the shared subspace based on a preset method. It can be understood that since the m feature matrices come from different modalities or views, each feature matrix usually has a different feature space and distribution. By constructing a target mapping function for the mapping process, the m feature matrices can be unified into the shared subspace for comparison and analysis.

[0101] It should be explained that the purpose of the target mapping function is to transform the m feature matrices under different modalities or views into a unified shared subspace. In order to ensure the effectiveness of the target mapping function and the quality of the shared subspace, the target mapping function needs to be tested.

[0102] In a possible implementation, based on a preset subspace reconstruction technique, the number of data features of the m feature matrices mapped to the shared subspace is determined. Among them, the subspace reconstruction technique is a common technique for evaluating the effectiveness of dimensionality reduction methods. Afterwards, the preset number of feature mappings of the shared subspace is determined. If the number of quantitative features is greater than or equal to the preset number of feature mappings, a target mapping function is constructed for the mapping process. If the number of data features is less than the preset number of feature mappings, the preset method is adjusted until the number of data features is greater than or equal to the preset number of feature mappings. Among them, the preset method is a method used to map m feature matrices to a shared subspace. Further, based on the adjusted preset method, the m feature matrices are re-mapped to the shared subspace, and a target mapping function is constructed for the mapping process.

[0103] It should be understood that the target mapping function may not be able to fully capture the complexity and diversity of the m feature matrices, so it is necessary to optimize the target mapping function to better map and represent the data features in the feature matrix, which is crucial for accurate group segmentation of power users. Based on the optimized target mapping function, a group segmentation matrix for target power users can be generated.

[0104] It should be noted that the optimization process of the target mapping function includes multiple optimization steps. Next, the optimization process steps of the target mapping function will be described in detail.

[0105] Optimization step 1:

[0106] A preset independence criterion is introduced into the target mapping function to obtain a corresponding first optimization function. Based on the first optimization function, the m feature matrices are remapped to a shared subspace, and the feature independence of the m feature matrices in the shared subspace is calculated.

[0107] Determine a preset independence threshold of the shared subspace. If the feature independence is greater than or equal to the preset independence threshold, continue to optimize the first optimization function; if the feature independence is less than the preset independence threshold, adjust the preset independence criterion until the feature independence is greater than or equal to the preset independence threshold.

[0108] For example, the independence of feature matrices in a shared subspace can be evaluated and ensured by introducing the Hilbert-Schmidt Independence Criterion (HSIC), which is a statistical tool for measuring the independence between two random variables or feature sets.

[0109] Optimization step 2:

[0110] A preset local structure constraint is introduced into the first optimization function to obtain a second optimization function; based on the second optimization function, the local structure information of each feature matrix is ​​determined.

[0111] The local structure information refers to the proximity relationship and local geometric shape of the data features in the feature matrix. Exemplarily, a local structure regularization constraint can be introduced into the first optimization function to determine the local structure information of each feature matrix.

[0112] Optimization step three:

[0113] A preset graph construction constraint is introduced into the second optimization function to obtain a third optimization function. Based on the third optimization function, the m feature matrices are mapped to a shared subspace to retain local structural information of the m feature matrices in the shared subspace.

[0114] Explanatory, by introducing a preset graph construction constraint in the second optimization function, we further ensure that the local structural information of the m feature matrices is preserved in the shared subspace. The graph construction constraint involves building a graph structure (such as a k-NN graph) and using the graph to maintain the proximity relationship between data features in the feature matrix during the mapping process.

[0115] By combining the local structure constraints and graph construction constraints, the local structure information of the m feature matrices can be retained, and the m feature matrices mapped to the shared subspace still reflect the geometric characteristics of the original m feature matrices, thereby enhancing the effectiveness of feature representation.

[0116] At this point, the optimization of the target mapping function is completed. After that, m optimized feature matrices mapped to the shared subspace based on the optimized target mapping function are obtained, and the corresponding sparse public graph matrix is ​​generated based on the m optimized feature matrices. Further optimization of the sparse public graph matrix can obtain the group partition matrix.

[0117] Explanatory, a sparse common graph matrix is ​​a matrix used to represent the relationships between data points, which combines the characteristics of sparsity and commonality. Sparsity means that most elements in the sparse common graph matrix are zero, and only a few elements are non-zero, that is, there are only a few direct connections between nodes. Commonality means that the sparse common graph matrix reflects the shared structure or commonality between data from multiple modalities or views.

[0118] Typically, a sparse common graph matrix is ​​generated by computing the similarities between data features in a feature matrix in a shared subspace. In the sparse common graph matrix, nodes represent data features and the weights of edges reflect the similarities between nodes.

[0119] The optimization of the sparse common graph matrix is ​​to improve its effectiveness in the segmentation of power user groups, which may involve adjusting weights, removing noise edges, or enhancing important connections. The optimization of the sparse common graph matrix can use regularization techniques, such as Laplace rank constraints, to ensure the smoothness and connectivity of the sparse common graph matrix. The optimized sparse common graph matrix can be regarded as the final group segmentation matrix.

[0120] Next, the generation process of the group division matrix for target power users is explained in detail through a specific example. In a specific example, the image data and text data of the target power user are obtained, and the image data and text data are respectively extracted under m views. Based on the acquired data of the two modes and the data features under m views, m feature matrices are constructed, where each feature matrix is ​​the data features under different modes and views.

[0121] Next, the matrix decomposition technique is used to mine the shared subspace across views from the constructed m feature matrices, and its data formula is defined as:

[0122]

[0123] Among them, U v ∈R dv×rRepresents the individual data features contained in the feature matrix under the vth view, R∈R r×n is the shared subspace information matrix across views, which is used to encode the complementary information of different views. r represents the latent vector dimension, that is, the number of latent features. 1 Is a non-negative number used to balance the regularization term of the generated error and the penalty subspace. ||·|| 2 F Represents the square of the Frobenius norm of the characteristic matrix.

[0124] It should be noted that, considering the influence of the weight value of each feature matrix on the shared subspace, the weight of the v-th view is expressed as ω v , that is, the contribution of the vth view to the group segmentation task. Then, formula (1-1) can be re-expressed as:

[0125]

[0126] Among them, λ 2 is a non-negative number.

[0127] Furthermore, in order to increase the diversity of personality data features contained in the feature matrix under different views, HSIC, which is used to evaluate the dependencies between data features, is introduced to approximately quantify the diversity. The mathematical formula of HSIC is defined as:

[0128] HSIC(U v ,U h )=(n-1) -2 Tr(K v HK h H) (1-3)

[0129] Among them, K v It's U v The corresponding kernel matrix, K h It's U h The corresponding kernel matrix, the constituent elements of matrix H are h ij =γ ij -1 / n. If i=j, then γ ij =1, otherwise γ ij =0.

[0130] It needs to be explained that due to the inner product kernel function K v =(U v ) T U v and K h =(U h ) T U hAs the kernel function in HSIC, the overall HSIC of the feature matrices under m views is maximized to reduce the dependency between feature matrices:

[0131]

[0132] in, Tr(·) is the trace function, which is used to find the trace of the characteristic matrix. T Represents the transposed matrix of the feature matrix X.

[0133] In addition, in order to utilize the local structural information of the feature matrix, a local structural regularization constraint is introduced to mine the local structural information of the feature matrix. Specifically, the affinity matrix W of the data is used to represent the local structural information, and its mathematical formula can be defined as:

[0134]

[0135] Among them, L = DW is the Laplace matrix, D is the degree matrix, which is a diagonal matrix with diagonal elements r i is the shared subspace information matrix R∈R across views r×n The i-th column of .

[0136] After determining the local structural information of the feature matrix, we can combine the subspace reconstruction technique and the k-NN graph to learn the sparse common graph matrix. The form of the sparse common graph matrix is ​​as follows:

[0137]

[0138] Among them, S∈R n×n is a sparse binary indicator matrix, R T R∈R n×n Reconstruct the matrix for the subspace, each element in the matrix represents the similarity between two corresponding data features. Adopt the idea of ​​k-NN graph construction and use constraints

[0139]

[0140] The canonical binary indicator matrix S has only k non-zero elements in each row. Therefore, this constraint ensures the sparsity of S, and for each node in the graph, it retains the k neighbors with the largest weight. In order to ensure the maximum weight in W, it is necessary to maximize

[0141]

[0142] Therefore, a minus sign is used in formula (1-6).

[0143] After obtaining the sparse common graph matrix, it can be optimized to generate the final group partition matrix. Specifically, the Laplace rank constraint can be introduced to make the sparse common graph matrix have idealized properties, and then generate the corresponding partition matrix:

[0144]

[0145] According to Ky Fans theorem, the Laplace rank constraint can be transformed, so we can get

[0146]

[0147] Among them, σ i (L S ) is L S The i-th smallest eigenvalue, L S The first k smallest eigenvalues ​​of are all 0, that is

[0148]

[0149] So that rank(L S )=nk holds. Therefore, the above problem can be transformed into the following form:

[0150]

[0151] Among them, I k ∈R k×k represents the identity matrix, F∈R n×k Represents a partition matrix.

[0152] By combining formula (1-2), formula (1-4), formula (1-5) and formula (1-8), the final group division matrix for target power users is obtained:

[0153]

[0154] S104, grouping the target power users based on the group division matrix to obtain at least one power user group; determining a target power usage plan corresponding to the power user group, and recommending the target power usage plan corresponding to the power user group to the power user group.

[0155] It can be understood that each divided power user group represents a group of target power users with similar power consumption behaviors or characteristics. According to the power consumption behaviors or characteristics of each power user group, the most appropriate target power consumption plan is designed or selected for each power user group, which may include time period electricity prices, energy-saving suggestions, load management strategies, etc.

[0156] Once the target electricity consumption scheme suitable for each power user group is determined, these target electricity consumption schemes can be recommended to the corresponding power user groups. In this way, the target electricity consumption scheme received by each power user group is tailored to its specific needs, thereby improving the effectiveness of the target electricity consumption scheme and the acceptance of the power user group.

[0157] The method, device and equipment for recommending electricity consumption plans based on electricity user groups provided in the present application respond to the group segmentation task for target electricity users, and obtain p multimodal data of the target electricity users indicated by the group segmentation task. Wherein, p is a positive integer greater than 1. Multimodal data can comprehensively consider different types of data, can capture complex relationships that cannot be identified by a single type of data, and more comprehensively reflect the electricity consumption behavior and characteristics of the target electricity users. Next, based on the pre-specified m views, multi-view feature extraction is performed on the data of the p modes respectively to obtain the data features of the data of the p modes under the m views. Wherein, m is a positive integer greater than 1. Multi-view feature extraction involves analyzing and processing data from different perspectives or dimensions. Performing multi-view feature extraction on the basis of multimodal data can further enhance the ability of data analysis and application, and provide a basis for the subsequent accurate group segmentation of target electricity users. Afterwards, m feature matrices are constructed based on the data features of the data of the p modes under the m views. Determine the contribution of each feature matrix to the group segmentation task, and determine the weight value of each feature matrix based on the contribution, and further determine the shared subspace across views based on the m feature matrices and the corresponding m weight values. Next, the m feature matrices are mapped to the shared subspace based on the preset method, and the target mapping function is constructed for the mapping process. The target mapping function is optimized by sequentially introducing the preset independence criterion, the preset local structure constraint, and the preset graph construction constraint in the target mapping function. Obtain m optimized feature matrices mapped to the shared subspace based on the optimized target mapping function to generate the corresponding sparse public graph matrix, and then optimize the sparse public graph matrix to obtain the final group segmentation matrix. Further, the target power users are grouped based on the group segmentation matrix to obtain at least one power user group. Determine the target power consumption plan corresponding to the power user group, and recommend the corresponding target power consumption plan to the power user group. The present application can accurately group the target power users, thereby providing a highly personalized power consumption plan for the power user group and improving the accuracy of the power consumption plan recommendation.

[0158] Next, a specific embodiment is used to Figure 1 The specific process of the method for recommending electricity consumption plans based on electricity user groups is summarized in the following. Figure 2 A schematic diagram of a method for recommending a power consumption plan based on a power user group provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the method for recommending electricity consumption plans based on electricity user groups includes: a data acquisition unit, a feature extraction unit, a shared subspace learning unit, a sparse common graph matrix construction unit, a sparse common graph matrix optimization unit, and an electricity user group division unit.

[0159] The data acquisition unit includes an image data acquisition subunit and a text data acquisition subunit. The feature extraction unit includes an image feature extraction subunit and a text feature extraction subunit. Specifically, the image feature extraction subunit is used to extract the intrinsic features of the image data, and the text feature extraction subunit is used to extract the intrinsic features of the text data.

[0160] The shared subspace learning unit includes a matrix decomposition subunit, a Hilbert-Schmidt independence criterion subunit, a local structure constraint subunit, and a graph construction constraint subunit. Specifically, the matrix decomposition subunit is used to mine the cross-view shared subspace from the feature matrix; the Hilbert-Schmidt independence criterion subunit is used to enhance the diversity of the individual data features contained in the feature matrix; the local structure constraint subunit is used to determine the local structure information of the feature matrix; and the graph construction constraint subunit is used to retain the local structure information of the feature matrix in the shared subspace.

[0161] The sparse common graph matrix construction unit is used to learn the sparse common graph matrix. The sparse common graph matrix optimization unit includes a Laplace rank constraint subunit, wherein the Laplace rank constraint subunit is used to make the sparse common graph matrix have idealized properties and generate a corresponding partitioning matrix. The power user group partitioning unit is used to divide the target power users into groups.

[0162] It should be noted that the data acquisition unit is connected to the feature extraction unit, and the feature extraction unit is connected to the matrix decomposition subunit in the shared subspace learning unit. After that, the matrix decomposition subunit is sequentially connected to the Hilbert-Schmidt independence criterion subunit, the local structure constraint subunit, and the graph construction constraint subunit. The graph construction constraint subunit is connected to the sparse common graph construction unit, the sparse common graph matrix construction unit is connected to the Laplace rank constraint subunit in the sparse common graph matrix optimization unit, and the Laplace rank constraint subunit is connected to the power user group division unit.

[0163] based on Figure 2 The structure shown, Figure 3 Schematic diagram of the process of the method for recommending power consumption plans based on power user groups provided in the embodiment of the present application Figure 2 ,like Figure 3 As shown, the method for recommending a power consumption plan based on a power user group specifically includes the following steps:

[0164] S301, acquiring image data and text data of a target power user through a data acquisition unit;

[0165] S302, performing multi-view feature extraction on the image data and the text data through a feature extraction unit, and determining a feature matrix under multiple views; wherein the feature matrix includes an image feature matrix and a text feature matrix;

[0166] S303, mining a shared subspace across views from the feature matrix through a matrix decomposition subunit;

[0167] S304, optimizing the shared subspace through the Hilbert-Schmidt independence criterion subunit, the local structure constraint subunit, and the graph construction constraint subunit;

[0168] S305, generating a sparse public graph matrix through a sparse public graph matrix construction unit based on the optimized shared subspace;

[0169] S306, generating a group partition matrix for target power users through a Laplace rank constraint subunit based on the sparse common graph matrix;

[0170] S307 , based on the group division matrix, group the target power users by using the power user group division unit.

[0171] By performing multi-view feature extraction on image data and text data, the influence of data features under multiple views on the group division of target power users is fully considered. The shared subspace across views is mined from the feature matrix corresponding to the data features, and the consensus information of multiple views is fully utilized. It is considered that different views share potential consensus representations. It can be seen that by comprehensively considering the multi-perspectives of multimodal data, the feature differences and mutual correlations of different power user groups are explained, and the comprehensiveness and interpretability of the power user group division are improved. The final group division matrix can accurately divide the target power users into groups, thereby providing highly personalized power consumption plans for power user groups and improving the accuracy of power consumption plan recommendations.

[0172] Figure 4 A schematic diagram of the structure of a device for recommending electricity consumption plans based on electricity user groups provided in an embodiment of the present application, such as Figure 4 As shown, the power consumption plan recommendation device 400 based on power user groups includes: an acquisition module 401, a processing module 402, a determination module 403, a division module 404, and a recommendation module 405;

[0173] The acquisition module 401 is used to respond to the group segmentation task for the target power users and acquire the multimodal data of the target power users indicated by the group segmentation task; wherein the multimodal data includes data of p modes, where p is a positive integer greater than 1;

[0174] The processing module 402 is used to perform multi-view feature extraction on the data of p modalities based on the pre-specified m views, respectively, to obtain data features of the data of the p modalities under the m views; based on the data features of the data of the p modalities under the m views, construct m feature matrices; wherein m is a positive integer greater than 1;

[0175] A determination module 403, configured to determine a shared subspace across views based on the m feature matrices;

[0176] The processing module 402 is further used to generate a group partition matrix for target power users based on the shared subspace; wherein the shared subspace represents a low-dimensional space that can simultaneously represent data features of different modalities and views;

[0177] A division module 404 is used to divide the target power users into groups based on the group division matrix to obtain at least one power user group;

[0178] The determination module 403 is further used to determine a target electricity consumption plan corresponding to the electricity user group;

[0179] The recommendation module 405 is used to recommend a target electricity usage plan corresponding to the electricity user group to the electricity user group.

[0180] In a possible design, the determination module 403 is further configured to:

[0181] Determine the contribution of each feature matrix to the group segmentation task, and determine the weight value of each feature matrix based on the contribution; wherein each feature matrix corresponds to a weight value;

[0182] Based on the m feature matrices and the m weight values ​​corresponding to the m feature matrices, a shared subspace across views is determined.

[0183] In a possible design, the processing module 402 further includes: a mapping module 406 and a construction module 407.

[0184] A mapping module 406, used to map the m feature matrices to a shared subspace based on a preset method;

[0185] A construction module 407 is used to construct a target mapping function for a mapping process; wherein the mapping process is used to indicate mapping m feature matrices to a shared subspace;

[0186] The processing module 402 is further used to perform a first optimization process on the target mapping function, and generate a group division matrix for target power users based on the optimized target mapping function.

[0187] In a possible design, the processing module 402 further includes: an introduction module 408, configured to introduce a preset independence criterion into the target mapping function to obtain a corresponding first optimization function;

[0188] The mapping module 406 is further used to re-map the m feature matrices to the shared subspace based on the first optimization function;

[0189] The processing module 402 further includes: a calculation module 409, which is used to calculate the feature independence of m feature matrices in the shared subspace;

[0190] The determination module 403 is further used to determine a preset independence threshold of the shared subspace;

[0191] The processing module 402 is also used to continue optimizing the first optimization function if the feature independence is greater than or equal to the preset independence threshold; if the feature independence is less than the preset independence threshold, adjust the preset independence criterion until the feature independence is greater than or equal to the preset independence threshold.

[0192] In a possible design, the introducing module 408 is further used to introduce a preset local structural constraint into the first optimization function to obtain a second optimization function;

[0193] The determination module 403 is further used to determine the local structure information of each feature matrix based on the second optimization function;

[0194] The introduction module 408 is further used to introduce the preset graph construction constraint into the second optimization function to obtain a third optimization function;

[0195] The mapping module 406 is further configured to map the m feature matrices to the shared subspace based on the third optimization function, so as to retain the local structural information of the m feature matrices in the shared subspace.

[0196] In a possible design, the acquisition module 401 is further used to obtain m optimized feature matrices mapped to the shared subspace based on the optimized target mapping function;

[0197] The processing module 402 is further used to generate a corresponding sparse common graph matrix based on the m optimized feature matrices; and perform a second optimization process on the sparse common graph matrix to obtain a group partition matrix.

[0198] In one possible design, the determination module 403 is further used to determine the number of data features of the m feature matrices mapped to the shared subspace based on a preset subspace reconstruction technique; determine the number of preset feature mappings of the shared subspace;

[0199] The construction module 407 is further used to construct a target mapping function for the mapping process if the number of data features is greater than or equal to the preset number of feature mappings;

[0200] The processing module 402 is also used to adjust the preset method if the number of data features is less than the preset number of feature mappings until the number of data features is greater than or equal to the preset number of feature mappings; re-map the m feature matrices to the shared subspace based on the adjusted preset method, and construct a target mapping function for the mapping process.

[0201] The device for recommending electricity usage plans based on electricity user groups provided in the embodiment of the present application can be used to execute the method for recommending electricity usage plans based on electricity user groups in any of the above embodiments. The implementation principle and technical effects thereof are similar and will not be described in detail herein.

[0202] It should be noted that it should be understood that the division of the various modules of the above device is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software calling through processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software calling through processing elements, and some modules can be implemented in the form of hardware. In addition, all or part of these modules can be integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software.

[0203] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 5 As shown, the electronic device may include: a transceiver 51 , a processor 52 , and a memory 53 .

[0204] The processor 52 executes the computer execution instructions stored in the memory, so that the processor 52 executes the scheme in the above embodiment. The processor 52 can be a general-purpose processor, including a central processing unit CPU, a network processor (NP), etc.; it can also be a digital signal processor DSP, an application-specific integrated circuit ASIC, a field programmable gate array FPGA or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.

[0205] The memory 53 is connected to the processor 52 via a system bus and completes communication between them. The memory 53 is used to store computer program instructions.

[0206] The transceiver 51 may be used to communicate with other devices.

[0207] The system bus can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The system bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The transceiver is used to realize the communication between the database access device and other computers (such as clients, read-write libraries, and read-only libraries). The memory may include random access memory (RAM) and may also include non-volatile memory.

[0208] The electronic device provided in the embodiments of the present application can be used to execute the method provided in any of the above embodiments. The implementation principles and technical effects are similar and will not be repeated here.

[0209] An embodiment of the present application further provides a computer-readable storage medium, in which computer instructions are stored. When the computer instructions are executed on a computer, the computer executes the method provided in any of the above embodiments.

[0210] An embodiment of the present application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When at least one processor executes the computer program, the method provided in any of the above embodiments can be implemented.

[0211] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of modules is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms.

[0212] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to implement the solution of this embodiment.

[0213] In addition, each functional module in each embodiment of the present application can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The above-mentioned module-composed unit can be implemented in the form of hardware or in the form of hardware plus software functional units.

[0214] The above-mentioned integrated module implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform some steps of the methods of various embodiments of the present application.

[0215] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the invention may be directly implemented as being executed by a hardware processor, or may be implemented by a combination of hardware and software modules in the processor.

[0216] The memory may include a high-speed RAM memory, and may also include a non-volatile storage NVM, such as at least one disk memory, and may also be a USB flash drive, a mobile hard disk, a read-only memory, a magnetic disk or an optical disk, etc.

[0217] The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of the present application is not limited to only one bus or one type of bus.

[0218] The above storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The storage medium can be any available medium that can be accessed by a general or special purpose computer.

[0219] An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the storage medium can also exist as discrete components in an electronic control unit or a main control device.

[0220] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0221] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for recommending electricity consumption plans based on electricity user groups, characterized in that: include: In response to a group segmentation task for target power users, multimodal data of the target power users indicated by the group segmentation task is acquired; wherein the multimodal data includes data of p modes, where p is a positive integer greater than 1; Based on the pre-specified m views, multi-view feature extraction is performed on the data of the p modalities respectively to obtain data features of the data of the p modalities under the m views; based on the data features of the data of the p modalities under the m views, m feature matrices are constructed; wherein m is a positive integer greater than 1; Based on the m feature matrices, a shared subspace across views is determined; based on the shared subspace, a group partition matrix for the target power user is generated; wherein the shared subspace represents a low-dimensional space that can simultaneously represent data features of different modalities and views; The target power users are divided into groups based on the group division matrix to obtain at least one power user group; a target power usage plan corresponding to the power user group is determined, and the target power usage plan corresponding to the power user group is recommended to the power user group.

2. The method according to claim 1, characterized in that The determining of a shared subspace across views based on the m feature matrices includes: Determine the contribution of each feature matrix to the group segmentation task, and determine the weight value of each feature matrix based on the contribution; wherein each feature matrix corresponds to a weight value; Based on the m feature matrices and the m weight values ​​corresponding to the m feature matrices, a shared subspace across views is determined.

3. The method according to claim 1, characterized in that The generating, based on the shared subspace, a group partitioning matrix for the target power user comprises: Based on a preset method, the m feature matrices are mapped to the shared subspace, and a target mapping function is constructed for the mapping process; wherein the mapping process is used to indicate that the m feature matrices are mapped to the shared subspace; A first optimization process is performed on the target mapping function, and based on the optimized target mapping function, a group division matrix for the target power users is generated.

4. The method according to claim 3, characterized in that The performing a first optimization process on the target mapping function comprises: Introducing a preset independence criterion into the target mapping function to obtain a corresponding first optimization function; Re-mapping the m feature matrices to the shared subspace based on the first optimization function, and calculating feature independence of the m feature matrices in the shared subspace; Determine a preset independence threshold of the shared subspace. If the feature independence is greater than or equal to the preset independence threshold, continue to optimize the first optimization function. If the feature independence is less than the preset independence threshold, adjust the preset independence criterion until the feature independence is greater than or equal to the preset independence threshold.

5. The method according to claim 4, characterized in that The continuing to optimize the first optimization function includes: Introducing a preset local structure constraint into the first optimization function to obtain a second optimization function; determining local structure information of each feature matrix based on the second optimization function; A preset graph construction constraint is introduced into the second optimization function to obtain a third optimization function; based on the third optimization function, the m feature matrices are mapped to the shared subspace to retain local structural information of the m feature matrices in the shared subspace.

6. The method according to any one of claims 3 to 5, characterized in that The generating a group division matrix for the target power users based on the optimized target mapping function includes: Obtaining m optimized feature matrices mapped to the shared subspace based on the optimized target mapping function; Based on the m optimized feature matrices, a corresponding sparse common graph matrix is ​​generated; and a second optimization process is performed on the sparse common graph matrix to obtain the group partition matrix.

7. The method according to claim 3, characterized in that The mapping process constructs a target mapping function, including: Based on a preset subspace reconstruction technique, determining the number of data features of the m feature matrices mapped to the shared subspace; Determining a preset number of feature maps of the shared subspace; if the number of data features is greater than or equal to the preset number of feature maps, constructing a target mapping function for the mapping process; If the number of data features is less than the preset number of feature mappings, the preset method is adjusted until the number of data features is greater than or equal to the preset number of feature mappings; based on the adjusted preset method, the m feature matrices are remapped to the shared subspace, and a target mapping function is constructed for the mapping process.

8. A device for recommending electricity consumption plans based on electricity user groups, characterized in that: include: An acquisition module, configured to, in response to a group segmentation task for target power users, acquire multimodal data of target power users indicated by the group segmentation task; wherein the multimodal data includes data of p modes, where p is a positive integer greater than 1; A processing module, configured to perform multi-view feature extraction on the data of the p modalities based on the pre-specified m views, respectively, to obtain data features of the data of the p modalities under the m views; and construct m feature matrices based on the data features of the data of the p modalities under the m views; wherein m is a positive integer greater than 1; A determination module, configured to determine a shared subspace across views based on the m feature matrices; The processing module is further used to generate a group partition matrix for the target power user based on the shared subspace; wherein the shared subspace represents a low-dimensional space that can simultaneously represent data features of different modalities and views; A division module, configured to divide the target power users into groups based on the group division matrix to obtain at least one power user group; The determination module is further used to determine a target electricity consumption plan corresponding to the electricity user group; The recommendation module is used to recommend a target electricity usage plan corresponding to the electricity user group to the electricity user group.

9. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method for recommending electricity usage plans based on electricity user groups according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method for recommending electricity usage plans based on electricity user groups according to any one of claims 1 to 7.

Images