Intelligent economic structure monitoring method and system based on power utilization big data

Abstract

The invention relates to an economic structure intelligent monitoring method and system based on power utilization big data, and the method comprises the following steps: S1, collecting and preprocessing multi-source data of a regional power system; s2, extracting multi-dimensional power features based on the standardized user power consumption behavior data; s3, constructing an industry classification and economic structure recognition model, obtaining the industry affiliation and confidence distribution of each user based on the multi-dimensional power feature vector set, and calculating the power consumption ratio of each industry according to the power consumption of different types of users in the region to form an industrial power consumption structure matrix; s4, according to the industrial power consumption structure matrix, establishing an industrial power consumption-output value regression model in combination with historical economic statistical data; s5, by solving the industry electric quantity-output value regression model, the relative output value proportion of each industry in the region is obtained; and S6, mapping a model solving result into a GIS (Geographic Information System). According to the invention, quasi-real-time intelligent monitoring of the economic structure is realized, and the sensing speed and the response capability of the economic operation situation are greatly accelerated.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring, and in particular to an intelligent monitoring method and system for economic structure based on big data on electricity consumption. Background Technology

[0002] With the advancement of a new wave of informatization and intelligentization, energy data is regarded as an important "barometer" for observing economic and social operations. Electricity, as the most important end-use energy source in modern society, provides consumption data that can highly reflect the production and living activities of various industries, enterprises, and residents within a region. Traditional methods of economic structure monitoring heavily rely on human statistical reporting and periodic sampling, leading to problems such as time lag, coarse granularity, and human error, making it difficult to meet the demands of modern society for real-time, accurate, and dynamic monitoring of economic operations.

[0003] In recent years, the power industry has promoted the widespread adoption of smart meters and the construction of smart distribution network systems, accumulating fine-grained electricity consumption big data covering the entire society and spanning continuous time and space. This data not only includes dynamic indicators such as user-level power load and electricity consumption curves, but also deeply integrates with multi-dimensional information such as geographic, meteorological, and industry-specific tags, providing an extremely rich and accurate data foundation for monitoring regional economic structures. Simultaneously, the rapid development of artificial intelligence, machine learning, and big data mining methods has brought technological breakthroughs to the automated processing, feature extraction, and pattern recognition of power data, achieving efficient mapping between electricity consumption behavior and industry attributes.

[0004] Against this backdrop, how to build an efficient, intelligent, and automated dynamic monitoring method for economic structure based on electricity consumption big data has become a key technical challenge that urgently needs to be addressed in the fields of energy management, government statistics, and regional economic decision-making. Summary of the Invention

[0005] To address the aforementioned issues, the present invention aims to provide an intelligent monitoring method and system for economic structure based on big data on electricity consumption, thereby achieving dynamic perception of economic structure with high frequency and high spatial resolution.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: The intelligent monitoring method for economic structure based on electricity consumption big data includes the following steps: S1: Collect multi-source data from the regional power system and preprocess the collected multi-source data to obtain standardized user electricity consumption behavior data; S2: Based on standardized user electricity consumption behavior data, extract multi-dimensional power features that reflect user behavior and industry attributes; S3: Construct an industry classification and economic structure identification model. Based on a multi-dimensional power feature vector set, obtain the industry affiliation and confidence distribution of each user, and calculate the power consumption ratio of each industry according to the electricity consumption of different categories of users in the region, forming an industrial power consumption structure matrix; S4: Based on the industrial electricity consumption structure matrix and combined with historical economic statistics, establish an industry electricity consumption-output value regression model; S5: By solving the industry electricity consumption-output value regression model, the relative output value share of each industry in the region is obtained; S6: Map the model solution results from S5 onto the GIS geospatial system to generate an economic structure layer based on spatial distribution, enabling intuitive visualization of the differences in industrial structure among different administrative regions.

[0007] Furthermore, multi-source data of the regional power system is collected, specifically as follows: core power indicators at the user level are extracted from the power company's operation database, including daily, monthly, or hourly electricity consumption records for each user, real-time load curve data, power factor change curves, and dynamic characteristics of time-of-use electricity consumption during peak, valley, and normal periods; simultaneously, static attribute information is acquired, including the geographical coordinates of metering points, the topology of the distribution network to which they belong, user capacity levels, and known customer type labels; in addition, external data sources from statistical departments, meteorological departments, and geographic information systems are integrated, including regional GDP, output value of major industries, population density distribution, land use types, temperature and humidity meteorological elements, and spatial reference information of administrative division boundaries.

[0008] Further preprocessing is performed as follows: Abnormal records are removed using anomaly detection algorithms, and missing data is filled in using the average of similar users. Then, data from different time series and sampling frequencies are time-aligned and spatially registered. Time alignment unifies data from different sampling frequencies to the same time scale through resampling, and a time-series alignment mechanism is established based on standard timestamps to ensure temporal consistency of various data within the same time window. Spatial registration uses a unified geographic coordinate system to spatially register all geographic location information, establishing a precise mapping relationship between users and administrative divisions, power grid topology, and statistical zoning spatial units. Finally, data standardization eliminates the influence of different dimensions and numerical ranges between indicators, resulting in a standardized user electricity consumption behavior database.

[0009] Furthermore, based on standardized user electricity consumption behavior data, multi-dimensional power features reflecting user behavior and industry attributes are extracted, specifically as follows: First, intraday load characteristic indicators, including peak-to-valley ratio, average load factor, and peak load period, are calculated from the perspective of time-series signals to characterize the periodicity and stability of enterprise electricity consumption; second, Fourier transform is used to extract seasonal and periodic fluctuation features to distinguish between production-type and service-type users; subsequently, structural feature vectors are constructed based on power factor, holiday electricity consumption ratio, and variation amplitude; at the user level, the K-Means clustering algorithm is used to cluster similar load curve patterns into one class, initially forming the structure of industry clusters; finally, principal component analysis is used to achieve dimensionality reduction and fusion of high-dimensional features, generating a unified multi-dimensional power feature vector set.

[0010] Furthermore, the industry classification and economic structure identification model adopts a multimodal feature fusion and hierarchical classification architecture, including a feature encoding module, a multimodal fusion layer, and a hierarchical classification network, as follows: the feature encoding module is used to encode time-domain, frequency-domain, geographical, and external statistical features; the multimodal fusion layer fuses the encoded multidimensional features through a self-attention mechanism; the hierarchical classification network includes a two-level structure of industry-level classification and sector-level classification.

[0011] Furthermore, the hierarchical classification network includes a two-level structure of industry-level classification and sector-level classification, as detailed below: Input feature vector h after feature encoding and multimodal fusion i It employs a two-layer fully connected neural network structure: ; in, For activation functions; To achieve an identity mapping at the output layer; and These are the first-layer weight matrix and bias, respectively; Output for hidden layer; and These are the weight matrix and bias of the second layer, respectively; This is the output of a fully connected neural network. The output layer uses the Softmax function to calculate the probability of each industry: ; in, The output of the fully connected neural network for the i-th node in the k-th industry category; Let be the probability of the i-th node in the k-th industry category; And obtain industry forecast tags: ; After determining the industry category, the industry-level classification is responsible for identifying subcategories within that industry. It employs an attention-enhanced graph neural network, utilizing geographical adjacency and enterprise grid relationships to achieve spatially consistent prediction. Each identified object corresponds to a node i, whose input feature is h. i Define a graph of relationships between nodes G=(V,E), with an adjacency matrix A=[A ij The attention layer for each graph is calculated as follows: ; Where a is the attention parameter; h i and h j W represents the feature vectors of nodes i and j, respectively; g The linear transformation weight matrix of the graph attention layer; This involves concatenating vectors. The LeakyReLU activation function; e ij Score the attention from node i to node j; Normalized weights: ; Where k is a neighboring node of i; Let i be the set of neighbors of node i; Node update: ; Among them, W g This is the layer weight matrix; This represents the new feature representation of node i after being updated by the graph attention layer; For activation functions; After passing through the GAT layer, a new node representation is obtained. Enter the category Softmax layer: ; in, This represents the probability that user i belongs to industry m given industry category k; This represents the weight vector of industry m under industry k; This represents the bias term of industry m under industry k; , Let J represent the weight vector and bias term of industry j under industry k, respectively; Industry forecast results: ; The final joint probability is: ; Output industry tags and confidence levels: .

[0012] Furthermore, based on the industrial electricity consumption structure matrix and historical economic statistics, an industry electricity consumption-output value regression model is established, as follows: According to the industrial electricity consumption structure matrix: ; Economic output matrix: ; Among them, E m,t Y represents the electricity consumption of industry m in period t; M represents the total number of industries; T represents the length of the historical time window; m,t The economic output of industry m in period t; And introduce external adjustment variables: ; in, Electricity price index; GDP t Gross Domestic Product (GDP); CPI t Consumer Price Index Policy t For policy dummy variables; Weather t Meteorological factors; Combining the above variables, we obtain the final input variable X. m,t Random forest regression was used: ; Where B is the total number of decision trees in the random forest; Let be the prediction function for the b-th decision tree; Parameter estimation uses an elastic network: ; in, This represents the optimal parameter vector for the industry m-regression model. The square of the L2 norm of the parameter vector; λ1 is the L1 norm of the parameter vector; λ2 is the L2 regularization penalty coefficient, adjusting the strength of parameter contraction; λ3 is the L1 regularization penalty coefficient; Y m,t Let m be the actual output value of industry m at time t.

[0013] Furthermore, by solving the industry electricity consumption-output value regression model, the relative output value share of each industry in the region is obtained, specifically as follows: the industry electricity consumption structure matrix is ​​extracted from the real-time power system, and the parameters estimated in stage S4 are substituted into it. It calculates real-time output forecasts for each industry, performs industry aggregation processing, and outputs the proportion of economic structure.

[0014] Furthermore, the solution results of the S5 model are mapped onto a GIS geospatial system to generate an economic structure layer based on spatial distribution, enabling intuitive visualization of the differences in industrial structure across different administrative regions, as detailed below: The system associates the industrial output share of each administrative unit with its corresponding geographic boundary vector data; each spatial entity in an administrative region is assigned a set of attributes, including the current tertiary industry structure vector, the estimated total output value, and the model confidence level; thematic layers are constructed based on the results of different industry share, and the color intensity is used to express the share of a certain industry, while the tertiary industry structure is displayed using segmented color wheels or nested pie charts; and multiple interactive functions are provided to users, including clicking to query the detailed structure of administrative regions, comparing the structural differences of multiple regions, and exporting spatial analysis reports.

[0015] The economic structure intelligent monitoring system based on electricity consumption big data includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the economic structure intelligent monitoring method based on electricity consumption big data as described above.

[0016] The present invention has the following beneficial effects: 1. This invention combines multi-source data from the regional power system and, through fully automated behavioral data collection and dynamic processing, can reflect industrial electricity consumption and structural changes at a high frequency and seamlessly. Compared with the traditional statistical annual reports and surveys, which have long cycles and strong lag, this invention can achieve near real-time intelligent monitoring of the economic structure, greatly accelerating the perception speed and response capability of economic operation, and meeting the needs of modern macro-control and industrial layout adjustment for real-time data. 2. This invention introduces intelligent models such as deep learning and graph neural networks into the industry classification and economic structure identification process. It accurately portrays energy consumption behavior through multi-dimensional electricity consumption characteristics, realizes efficient mining and confidence quantification of user industry attributes, and automatically establishes an electricity consumption-output regression model by combining historical economic output and electricity consumption data. This enables quantitative estimation of the economic contribution of each industry, significantly improving the scientific nature and automation level of quantitative analysis. 3. This invention seamlessly embeds the model inference results into the GIS geospatial system, realizing the dynamic integration of economic structure data and real geographic space. Through spatial visualization layers, it can not only clearly show the development patterns and changing trends of different regions and industries, but also provide efficient and intuitive data support for regional economic structure optimization, policy simulation and resource allocation. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0018] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: refer to Figure 1 In this embodiment, an intelligent monitoring method for economic structure based on electricity consumption big data is provided, including the following steps: S1: Collect multi-source data from the regional power system and preprocess the collected multi-source data to obtain standardized user electricity consumption behavior data; S2: Based on standardized user electricity consumption behavior data, extract multi-dimensional power features that reflect user behavior and industry attributes; S3: Construct an industry classification and economic structure identification model. Based on a multi-dimensional power feature vector set, obtain the industry affiliation and confidence distribution of each user, and calculate the power consumption ratio of each industry according to the electricity consumption of different categories of users in the region, forming an industrial power consumption structure matrix; S4: Based on the industrial electricity consumption structure matrix and combined with historical economic statistics, establish an industry electricity consumption-output regression model to characterize the correspondence between electricity consumption and economic output. S5: By solving the industry electricity consumption-output value regression model, the relative output value ratio of each industry in the region is obtained, achieving near real-time estimation of the regional economic structure; S6: Map the model solution results from S5 onto the GIS geospatial system to generate an economic structure layer based on spatial distribution, enabling intuitive visualization of the differences in industrial structure among different administrative regions.

[0019] In this embodiment, multi-source data of the regional power system is collected, specifically as follows: core power indicators at the user level are extracted from the power company's operation database, including daily, monthly, or hourly electricity consumption records for each user, real-time load curve data, power factor change curves, and dynamic characteristics of time-of-use electricity consumption during peak, valley, and normal periods; simultaneously, static attribute information is acquired, including the geographical coordinates of the metering point, the topology of the distribution network to which it belongs, the user capacity level, and known customer type labels; in addition, external data sources from statistical departments, meteorological departments, and geographic information systems are integrated, including regional GDP, output value of major industries, population density distribution, land use type, temperature and humidity meteorological elements, and spatial reference information of administrative division boundaries.

[0020] In this embodiment, the preprocessing is as follows: Abnormal records are removed using an anomaly detection algorithm, and missing data is filled in using the average of similar users. Then, data from different time series and sampling frequencies are time-aligned and spatially registered. The time alignment unifies data from different sampling frequencies (e.g., 15 minutes, 1 hour, 1 day, etc.) to the same time scale through resampling, and establishes a time series alignment mechanism based on standard timestamps to ensure time consistency of various types of data within the same time window. The spatial registration uses a unified geographic coordinate system to spatially register all geographic location information, establishing a precise mapping relationship between users and administrative divisions, power grid topology, and statistical partition spatial units. Finally, data standardization eliminates the influence of different dimensions and numerical ranges between indicators, resulting in a standardized user electricity consumption behavior database.

[0021] In this embodiment, based on standardized user electricity consumption behavior data, multidimensional power features reflecting user behavior and industry attributes are extracted, specifically as follows: First, intraday load characteristic indicators, including peak-to-valley ratio, average load factor, and peak load period, are calculated from the perspective of time-series signals to characterize the periodicity and stability of enterprise electricity consumption; second, Fourier transform is used to extract seasonal and periodic fluctuation features to distinguish between production-type and service-type users; subsequently, a structural feature vector is constructed based on power factor, holiday electricity consumption ratio, and variation amplitude; at the user level, the K-Means clustering algorithm is used to cluster similar load curve patterns into one class, initially forming the structure of industry clusters; finally, principal component analysis is used to achieve dimensionality reduction and fusion of high-dimensional features, generating a unified multidimensional power feature vector set.

[0022] In this embodiment, the industry classification and economic structure identification model adopts a multimodal feature fusion and hierarchical classification architecture, including a feature encoding module, a multimodal fusion layer, and a hierarchical classification network, as follows: the feature encoding module is used to encode time-domain, frequency-domain, geographical, and external statistical features; the multimodal fusion layer fuses the encoded multidimensional features through a self-attention mechanism; the hierarchical classification network includes a two-level structure of industry-level classification and sector-level classification.

[0023] In this embodiment, the hierarchical classification network includes a two-level structure of industry-level classification and sector-level classification, as detailed below: The industry-level classification is used to identify the macro-level affiliation of "primary industry / secondary industry / tertiary industry"; the input of this layer is the feature vector h obtained by feature encoding and multimodal fusion. i It employs a two-layer fully connected neural network structure: ; in, For activation functions; To achieve an identity mapping at the output layer; and These are the first-layer weight matrix and bias, respectively; Output for hidden layer; and These are the weight matrix and bias of the second layer, respectively; This is the output of a fully connected neural network. The output layer uses the Softmax function to calculate the probability of each industry: ; in, The output of the fully connected neural network for the i-th node in the k-th industry category; Let be the probability of the i-th node in the k-th industry category; And obtain industry forecast tags: ; After determining the industry categories (primary, secondary, and tertiary industries), the industry-level classification is responsible for identifying subcategories within that industry. It employs an attentional graph neural network (AGNN) and leverages geographical adjacency and enterprise-grid relationships to achieve spatially consistent predictions. Each identified object corresponds to a node i, whose input feature is h. i Define a graph of relationships between nodes G=(V,E), with an adjacency matrix A=[A ij The attention layer for each graph is calculated as follows: ; Where a is the attention parameter; h i and h j W represents the feature vectors of nodes i and j, respectively; g The linear transformation weight matrix of the graph attention layer; This involves concatenating vectors. The LeakyReLU activation function; e ij Score the attention from node i to node j; Normalized weights: ; Where k is a neighboring node of i; Let i be the set of neighbors of node i; Node update: ; Among them, W g This is the layer weight matrix; This represents the new feature representation of node i after being updated by the graph attention layer; For activation functions; After passing through the GAT layer, a new node representation is obtained. Enter the category Softmax layer: ; in, This represents the probability that user i belongs to industry m given industry category k; This represents the weight vector of industry m under industry k; This represents the bias term of industry m under industry k; , Let J represent the weight vector and bias term of industry j under industry k, respectively; Industry forecast results: ; The final joint probability is: ; Output industry tags and confidence levels: .

[0024] In this embodiment, based on the industrial electricity consumption structure matrix and combined with historical economic statistics, an industry electricity consumption-output value regression model is established, as follows: According to the industrial electricity consumption structure matrix: ; Economic output matrix: ; Among them, E m,t Y represents the electricity consumption of industry m in period t; M represents the total number of industries; T represents the length of the historical time window; m,t The economic output of industry m in period t; And introduce external adjustment variables: ; in, Electricity price index; GDP t Gross Domestic Product (GDP); CPI t Consumer Price Index Policy t For policy dummy variables; Weather t Meteorological factors; Combining the above variables, we obtain the final input variable X. m,t Random forest regression was used: ; Where B is the total number of decision trees in the random forest; Let be the prediction function for the b-th decision tree; Parameter estimation uses an elastic network: ; in, This represents the optimal parameter vector for the industry m-regression model. The square of the L2 norm of the parameter vector; λ1 is the L1 norm of the parameter vector; λ2 is the L2 regularization (ridge regression) penalty coefficient, which adjusts the strength of parameter contraction; λ3 is the L1 regularization (LASSO) penalty coefficient; Y m,t Let m be the actual output value of industry m at time t.

[0025] In this embodiment, the relative output value proportion of each industry in the region is obtained by solving the industry electricity consumption-output value regression model, specifically as follows: the industry electricity consumption structure matrix is ​​extracted from the real-time power system, and the parameters estimated in stage S4 are substituted into it. It calculates real-time output forecasts for each industry, performs industry aggregation processing, and outputs the proportion of economic structure.

[0026] In this embodiment, the model solution results of S5 are mapped onto a GIS geospatial system to generate an economic structure layer based on spatial distribution, thereby achieving intuitive visualization of the differences in industrial structure among different administrative regions. Specifically, the industrial output value ratio (ρ) of each administrative unit is plotted as follows: k,t The system associates k=1,2,3 with the primary, secondary, and tertiary industries, respectively, with their corresponding geographic boundary vector data. Each administrative region's spatial entity is assigned a set of attributes, including the current tertiary industry structure vector, estimated total output value, and model confidence level. Thematic layers are constructed based on the proportion of different industries, using varying shades of color to represent the proportion of a particular industry, and the tertiary industry structure is displayed using segmented color wheels or nested pie charts. The system provides users with various interactive functions, including clicking to query detailed administrative region structures, comparing structural differences across multiple regions, and exporting spatial analysis reports. For macro-level decision-makers, this GIS platform can quickly locate key areas with significant structural changes and identify target objects or weak links for structural adjustments. For park managers and research institutions, it can also achieve precise regional industrial profiling and policy simulation based on spatial big data.

[0027] The intelligent economic structure monitoring system based on electricity consumption big data includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps of the intelligent economic structure monitoring method based on electricity consumption big data as described above. Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0028] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0029] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0030] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0031] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An intelligent monitoring method for economic structure based on big data on electricity consumption, characterized in that, Includes the following steps: S1: Collect multi-source data from the regional power system and preprocess the collected multi-source data to obtain standardized user electricity consumption behavior data; S2: Based on standardized user electricity consumption behavior data, extract multi-dimensional power features that reflect user behavior and industry attributes; S3: Construct an industry classification and economic structure identification model. Based on a multi-dimensional power feature vector set, obtain the industry affiliation and confidence distribution of each user, and calculate the power consumption ratio of each industry according to the electricity consumption of different categories of users in the region, forming an industrial power consumption structure matrix; S4: Based on the industrial electricity consumption structure matrix and combined with historical economic statistics, establish an industry electricity consumption-output value regression model; S5: By solving the industry electricity consumption-output value regression model, the relative output value share of each industry in the region is obtained; S6: Map the model solution results from S5 onto the GIS geospatial system to generate an economic structure layer based on spatial distribution, enabling intuitive visualization of the differences in industrial structure among different administrative regions.

2. The intelligent monitoring method for economic structure based on big data of electricity consumption according to claim 1, characterized in that, The collection of multi-source data from the regional power system is specifically as follows: Core power indicators at the user level are extracted from the power company's operational database, including daily, monthly, or hourly electricity consumption records for each user, real-time load curve data, power factor change curves, and dynamic characteristics of time-of-use electricity consumption during peak, valley, and normal periods. Simultaneously, static attribute information is acquired, including the geographical coordinates of metering points, the topology of their respective distribution networks, user capacity levels, and known customer type labels. Furthermore, external data sources from statistical departments, meteorological departments, and geographic information systems are integrated, including regional GDP, output value of major industries, population density distribution, land use types, temperature and humidity meteorological elements, and spatial reference information for administrative boundaries.

3. The intelligent monitoring method for economic structure based on big data of electricity consumption according to claim 2, characterized in that, The preprocessing is as follows: Abnormal records are removed using an anomaly detection algorithm, and missing data is filled in using the average of similar users. Then, data from different time series and sampling frequencies are time-aligned and spatially registered. Time alignment unifies data from different sampling frequencies to the same time scale through resampling, and a time-series alignment mechanism is established based on standard timestamps to ensure time consistency of various types of data within the same time window. Spatial registration uses a unified geographic coordinate system to spatially register all geographic location information, establishing a precise mapping relationship between users and administrative divisions, power grid topology, and statistical partition spatial units. Finally, data standardization eliminates the influence of different dimensions and numerical ranges between indicators, resulting in a standardized user electricity consumption behavior database.

4. The intelligent monitoring method for economic structure based on big data of electricity consumption according to claim 1, characterized in that, Based on standardized user electricity consumption data, multi-dimensional power characteristics reflecting user behavior and industry attributes are extracted, as follows: First, daily load characteristic indicators, including peak-valley ratio, average load rate, and maximum load period, are calculated from the perspective of time series signals to characterize the periodicity and stability of the enterprise's electricity consumption. Secondly, Fourier transform is used to extract seasonal and cyclical fluctuation features to distinguish between production-type and service-type users. Subsequently, structural feature vectors are constructed based on power factor, holiday electricity consumption ratio and variation. At the user level, the K-Means clustering algorithm is used to cluster similar load curve patterns into one class, initially forming the structure of industry clusters. Finally, principal component analysis is used to achieve dimensionality reduction and fusion of high-dimensional features, generating a unified multi-dimensional power feature vector set.

5. The intelligent monitoring method for economic structure based on big data of electricity consumption according to claim 1, characterized in that, The industry classification and economic structure identification model adopts a multimodal feature fusion and hierarchical classification architecture. It includes a feature encoding module, a multimodal fusion layer, and a hierarchical classification network, as follows: The feature encoding module is used to encode time-domain, frequency-domain, geographic, and external statistical features; the multimodal fusion layer fuses the encoded multidimensional features through a self-attention mechanism; The hierarchical classification network comprises a two-level structure: industry-level classification and sector-level classification.

6. The intelligent monitoring method for economic structure based on big data of electricity consumption according to claim 5, characterized in that, The hierarchical classification network comprises a two-level structure: industry-level classification and sector-level classification, as detailed below: Input feature vector h after feature encoding and multimodal fusion i It employs a two-layer fully connected neural network structure: ； in, For activation functions; To achieve an identity mapping at the output layer; and These are the first-layer weight matrix and bias, respectively; Output for hidden layer; and These are the second-layer weight matrix and bias, respectively; This is the output of a fully connected neural network. The output layer uses the Softmax function to calculate the probability of each industry: ； in, The output of the fully connected neural network for the i-th node in the k-th industry category; Let be the probability of the i-th node in the k-th industry category; And obtain industry forecast tags: ; After determining the industry category, the industry-level classification is responsible for identifying subcategories within that industry. It employs an attention-enhanced graph neural network, utilizing geographical adjacency and enterprise grid relationships to achieve spatially consistent prediction. Each identified object corresponds to a node i, whose input feature is h. i Define a graph of relationships between nodes G=(V,E), with an adjacency matrix A=[A ij The attention layer for each graph is calculated as follows: ； Where a is the attention parameter; h i and h j The feature vectors of nodes i and j are respectively; W g The linear transformation weight matrix of the graph attention layer; This involves concatenating vectors. The LeakyReLU activation function; e ij Score the attention from node i to node j; Normalized weights: ； Where k is a neighboring node of i; Let i be the set of neighbors of node i; Node update: ； Among them, W g This is the layer weight matrix; This represents the new feature representation of node i after being updated by the graph attention layer; For activation functions; After passing through the GAT layer, a new node representation is obtained. Enter the category Softmax layer: ； in, This represents the probability that user i belongs to industry m given industry category k; This represents the weight vector of industry m under industry k; This represents the bias term of industry m under industry k; , Let J represent the weight vector and bias term of industry j under industry k, respectively; Industry forecast results: ; The final joint probability is: ； Output industry tags and confidence levels: 。 7. The intelligent monitoring method for economic structure based on big data of electricity consumption according to claim 1, characterized in that, Based on the industrial electricity consumption structure matrix and combined with historical economic statistics, an industry electricity consumption-output value regression model is established, as follows: According to the industrial electricity consumption structure matrix: ； Economic output matrix: ； Among them, E m,t Y represents the electricity consumption of industry m in period t; M represents the total number of industries; T represents the length of the historical time window; m,t The economic output of industry m in period t; And introduce external adjustment variables: ； in, Electricity price index; GDP t Gross Domestic Product (GDP); CPI t Consumer Price Index Policy t For policy dummy variables; Weather t Meteorological factors; Combining the above variables, we obtain the final input variable X. m,t Random forest regression was used: ； Where B is the total number of decision trees in the random forest; Let be the prediction function for the b-th decision tree; Parameter estimation uses an elastic network: ； in, This represents the optimal parameter vector for the industry m-regression model. The square of the L2 norm of the parameter vector; λ1 is the L1 norm of the parameter vector; λ2 is the L2 regularization penalty coefficient, adjusting the strength of parameter contraction; λ3 is the L1 regularization penalty coefficient; Y m,t Let m be the actual output value of industry m at time t.

8. The intelligent monitoring method for economic structure based on big data of electricity consumption according to claim 7, characterized in that, The process involves solving the industry-electricity consumption-output value regression model to obtain the relative output value proportion of each industry in the region. Specifically, the industry electricity consumption structure matrix is ​​extracted from the real-time power system and substituted with the parameters estimated in stage S4. It calculates real-time output forecasts for each industry, performs industry aggregation processing, and outputs the proportion of economic structure.

9. The intelligent monitoring method for economic structure based on big data of electricity consumption according to claim 1, characterized in that, The process involves mapping the solution results of the S5 model onto a GIS geospatial system to generate an economic structure layer based on spatial distribution, thereby achieving intuitive visualization of the differences in industrial structure across different administrative regions. The details are as follows: The industrial output value of each administrative unit is associated with its corresponding geographic boundary vector data; each spatial entity of the administrative region is assigned a set of attributes, including the current three-industry structure vector, the estimated total output value, and the model confidence level; thematic layers are constructed based on the results of different industry proportions, and the proportion of a certain industry is expressed by color depth, and the three-industry structure is displayed by segmented color wheels or nested pie charts. It also provides users with a variety of interactive functions, including clicking to query the detailed structure of administrative regions, comparing the structural differences between multiple regions, and exporting spatial analysis reports.

10. An intelligent monitoring system for economic structure based on big data on electricity consumption, characterized in that: It includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the intelligent monitoring method for economic structure based on big data of electricity consumption as described in any one of claims 1-9.

Images