Commercial building energy consumption anomaly detection method based on graph method

Abstract

The invention provides a commercial building energy consumption anomaly detection method based on a graph method, and relates to the technical field of intelligent constructions.The method comprises the steps that a graph structure is constructed based on similarity, nodes are mapped into sub-item and sub-partition sub-tables, and edge weights represent time sequence similarity between the corresponding nodes; generating spectral distribution representing topological characteristics of the graph structures by calculating a Laplacian matrix eigenvalue set of the graph structures, and performing hierarchical clustering on the plurality of graph structures based on a Batta-just distance between the spectral distribution to obtain energy consumption mode clustering results in different time periods; and carrying out contour coefficient analysis on the clustering result, screening a class cluster number corresponding to a maximum contour coefficient value as a corresponding partition, and defining a class cluster set containing historical normal data as a baseline class cluster. According to the invention, the accuracy of building energy consumption anomaly detection can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent buildings, and in particular to a method for detecting abnormal energy consumption in commercial buildings based on a graph method. Background Art

[0002] As energy demand in the building sector increases globally, building energy management has become a key area for improving energy efficiency and reducing carbon emissions. The introduction of sub-meter monitoring technology for commercial buildings provides more granular energy consumption data, but traditional methods have significant limitations when processing such complex data, particularly when identifying anomalous usage patterns. Anomaly detection plays a crucial role in building energy management. However, traditional statistical or supervised learning methods struggle to capture the complex dependencies within sub-meter data.

[0003] Existing technologies have many limitations in detecting anomalies in building energy consumption, and their practicality makes it difficult to effectively adapt to the complex energy consumption patterns in building environments.

[0004] First, existing methods mainly rely on univariate analysis, usually focusing only on the energy consumption measured by a single meter, and are unable to fully analyze the complex relationships between multiple sub-meters. Second, existing technologies are insufficient in responding to dynamic changes in building energy consumption. In actual applications, energy consumption patterns may fluctuate significantly due to changes in tenant behavior, environmental conditions, or equipment operating status, and traditional methods are difficult to flexibly adapt to these dynamic patterns, resulting in a high false alarm rate or serious omissions in anomaly detection. In addition, most existing methods rely on large amounts of labeled data for model training, but in the field of building energy consumption management, it is often very difficult to obtain labeled data. Therefore, existing technologies have obvious defects in unsupervised detection capabilities and are difficult to meet the needs of actual engineering applications. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method for detecting energy consumption anomalies in commercial buildings based on a graph method, which can improve the accuracy of building energy consumption anomaly detection.

[0006] In order to solve the above technical problems, the technical solutions of the present invention are as follows: In a first aspect, a method for detecting abnormal energy consumption in commercial buildings based on a graph method is provided, the method comprising: Collect real-time energy consumption data from the sub-item and sub-area sub-tables within the building to generate multiple sets of energy consumption time series with time series characteristics; Performing dynamic time warping calculation on the energy consumption time series, quantifying the energy consumption time series similarity between each sub-item partition sub-table, and constructing a graph structure based on the similarity, in which nodes are mapped to the sub-item partition sub-tables, and edge weights represent the time series similarity between corresponding nodes; By calculating the Laplace matrix eigenvalue set of the graph structure, a spectral distribution representing the topological characteristics of the graph structure is generated, and hierarchical clustering is performed on multiple graph structures based on the Bhattacharya distance between the spectral distributions to obtain clustering results of energy consumption patterns in different time periods; Performing silhouette coefficient analysis on the clustering results, selecting the number of clusters corresponding to the maximum silhouette coefficient as the corresponding division, and defining the cluster set containing historical normal data as the baseline cluster; Generate a graph structure to be detected for the newly collected energy consumption data, calculate the minimum spectral distance between it and all graph structures in the baseline cluster, and determine it as an abnormal graph structure if the minimum value exceeds a preset threshold; Performing a node connection strength analysis on the abnormal graph structure, calculating the edge weight drop between each node and its adjacent nodes, and marking the sub-item partition sub-table associated with the node with the largest edge weight loss as an abnormal source; Based on the spatial distribution characteristics of the abnormal sources and the time series characteristics of energy consumption mutations, a detection report including equipment maintenance prompts and energy efficiency optimization strategies is generated.

[0007] Furthermore, dynamic time warping calculation is performed on the energy consumption time series to quantify the energy consumption time series similarity between each sub-item partition sub-table, and a graph structure is constructed based on the similarity, in which nodes are mapped to sub-item partition sub-tables, and edge weights represent the time series similarity between corresponding nodes, including: Based on the collected energy consumption data, identify all sub-item and partition sub-meters in the building, where each sub-meter corresponds to an independent energy consumption monitoring unit for a specific area or equipment, and extract the energy consumption time series of each sub-meter; For the energy consumption time series of any two sub-item partitioned sub-tables, dynamic time warping calculation is performed: multiple matching paths are generated through nonlinear time axis alignment. The squared Euclidean distance between adjacent matching elements in each path is calculated. The total cumulative distance is obtained by summing up all the squared local distances on the path. The corresponding path with the smallest total cumulative distance is selected, and the square root of the total cumulative distance of this path is used as the dynamic time warping distance of the two sequences to represent the similarity of the energy consumption time series of the two sub-tables. Mapping the dynamic time warping distance to the interval (0, 1) through a linear normalization function to generate an edge weight value, where the smaller the distance, the closer the similarity score after mapping is to 1, and the larger the distance, the closer the score is to 0; A fully connected graph structure is constructed with the item partitioned sub-tables as nodes and the normalized similarity scores as edge weights, where the weight value of the edge between any two nodes reflects the similarity of the energy consumption time series of the corresponding sub-tables.

[0008] Furthermore, by calculating the Laplace matrix eigenvalue set of the graph structure, a spectral distribution representing the topological characteristics of the graph structure is generated. Then, hierarchical clustering is performed on multiple graph structures based on the Bhattacharya distance between the spectral distributions to obtain clustering results of energy consumption patterns in different time periods, including: For the graph structure of each time period, a standardized Laplacian matrix is ​​constructed based on the edge weights between nodes, and eigenvalue decomposition is performed on it to extract the eigenvalue set arranged in ascending order, generating a spectral distribution that characterizes the topological characteristics of the graph structure; For the spectral distribution of any two graph structures, calculate their Bhattacharya distance: add up the squares of the eigenvalue differences at corresponding positions of the two eigenvalue sets, and take the square root of the sum. The resulting value is defined as the spectral distance between the two graph structures. Based on the spectral distance matrix between all graph structures, an agglomerative hierarchical clustering algorithm is used to gradually merge the graph structures or clusters with the closest distance using a single-link strategy until a preset number of initial cluster sets are formed. According to the dendrogram generated in the hierarchical clustering process, the hierarchical division results that meet the preset merging threshold are intercepted, and the graph structures of different time periods are classified into corresponding energy consumption pattern clusters to form a clustering result set with time continuity.

[0009] Furthermore, the clustering results are analyzed by silhouette coefficient, and the number of clusters corresponding to the maximum silhouette coefficient is selected as the corresponding division. The cluster set containing historical normal data is defined as the baseline cluster, including: Based on the clustering result set generated by hierarchical clustering, a partitioning scheme is set for each candidate cluster number, and the silhouette coefficients of all data points in each partitioning scheme are calculated: for a single data point, the average distance between it and the remaining data points in its cluster is calculated as the first distance, and the average distance between it and all data points in the nearest remaining cluster is calculated as the second distance. The difference between the second distance and the first distance is divided by the maximum value of the two distances to obtain the silhouette coefficient of the data point; The average value of the silhouette coefficients of all data points in the partitioning scheme corresponding to each candidate cluster number is calculated, and the cluster number with the largest average value is selected as the corresponding partitioning scheme; in the cluster set corresponding to the corresponding partitioning scheme, the clusters containing the graph structure during the historical normal operation period are screened and merged into a baseline cluster set that represents the normal mode of building energy consumption.

[0010] Furthermore, a graph structure to be detected is generated for the newly collected energy consumption data, and the minimum spectral distance between the graph structure and all graph structures in the baseline cluster is calculated. If the minimum value exceeds a preset threshold, it is determined to be an abnormal graph structure, including: Based on the newly collected energy consumption data, the graph structure to be detected is generated, and the eigenvalue set of its normalized Laplace matrix is ​​extracted to generate the corresponding spectral distribution; For each cluster in the baseline cluster set, respectively calculating the Bhattacharya distance between the spectral distribution of the graph structure to be detected and the spectral distributions of all historical graph structures in the cluster, and recording the minimum spectral distance in each cluster; Traverse the minimum spectral distances of all baseline clusters and extract the smallest value as the global minimum spectral distance between the graph structure to be detected and the baseline cluster set; The global minimum spectral distance is compared with a preset abnormality judgment threshold. If the distance exceeds the threshold, the graph structure to be detected is determined to be an abnormal energy consumption mode, the abnormality judgment result is output, and the abnormal graph structure and its corresponding time period are marked as an energy consumption abnormality event.

[0011] Furthermore, the node connection strength analysis is performed on the abnormal graph structure, the decrease in edge weight between each node and its adjacent nodes is calculated, and the sub-item partition sub-table associated with the node with the largest edge weight loss is marked as an abnormal source, including: For each node in the abnormal graph structure, obtain its edge weight data in the historical normal mode, and calculate the absolute value of the difference between the real-time edge weight of the node and all adjacent nodes and the historical edge weight. The difference of all adjacent edges is accumulated to obtain the connection loss of the node; Normalize the connection loss of all nodes to generate a loss intensity index in the interval [0, 1], where the node with the maximum loss is marked as a candidate anomaly source; Based on the edge weight fluctuation range of historical normal data statistics, a dynamic judgment threshold for connection loss is set. If the loss intensity index of a candidate abnormal source exceeds the threshold, its associated sub-item partition sub-table is identified as an abnormal source.

[0012] Furthermore, based on the spatial distribution characteristics of the abnormal source and the time series characteristics of energy consumption mutation, a detection report containing equipment maintenance prompts and energy efficiency optimization strategies is generated, including: The spectral characteristics of the anomaly source location results and the anomaly graph structure are processed to analyze the spatial distribution density of the sub-item partition sub-tables corresponding to the anomaly source and the temporal fluctuation pattern of the energy consumption mutation period to extract the space-time correlation anomaly characteristics. Based on these space-time correlation anomaly characteristics and combined with the fault mode case library in the historical energy consumption event database, the similarity between the anomaly characteristics and known fault modes is matched to infer the type of anomaly cause, including equipment failure, human operation abnormality, or environmental interference. Generate targeted optimization suggestions based on the type of anomaly cause: generate an equipment maintenance list and priority ranking for equipment failure anomalies, generate energy efficiency parameter adjustment strategies for energy consumption pattern mutation anomalies, and associate the spatial location information of the area where the anomaly source is located; process the optimization suggestions, anomaly cause analysis and related data features into a structured detection report.

[0013] In the second aspect, a commercial building energy consumption anomaly detection system based on a graph method includes: An acquisition module is configured to collect real-time energy consumption data from the sub-item partition sub-tables within the building and generate multiple sets of energy consumption time series with time series characteristics; perform dynamic time warping calculations on the energy consumption time series, quantify the energy consumption time series similarity between the sub-item partition sub-tables, and construct a graph structure based on the similarity, in which nodes are mapped to the sub-item partition sub-tables, and edge weights represent the time series similarity between corresponding nodes; A comparison module is configured to generate a spectral distribution representing the topological characteristics of the graph structure by calculating the set of Laplace matrix eigenvalues ​​of the graph structure, and hierarchically cluster multiple graph structures based on the Bhattacharya distance between the spectral distributions to obtain clustering results of energy consumption patterns in different time periods; perform silhouette coefficient analysis on the clustering results, select the number of clusters corresponding to the maximum silhouette coefficient as the corresponding division, and define the cluster set containing historical normal data as the baseline cluster; generate a graph structure to be detected for the newly collected energy consumption data, calculate the minimum value of its spectral distance with all graph structures in the baseline cluster, and determine that the graph structure is abnormal if the minimum value exceeds a preset threshold; The processing module is used to analyze the node connection strength of the abnormal graph structure, calculate the decrease in edge weight between each node and its adjacent nodes, and mark the sub-item partition sub-table associated with the node with the largest edge weight loss as an abnormal source; based on the spatial distribution characteristics and energy consumption mutation time series characteristics of the abnormal source, generate a detection report containing equipment maintenance prompts and energy efficiency optimization strategies.

[0014] According to a third aspect, a computing device includes: one or more processors; The storage device is used to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method.

[0015] In a fourth aspect, a computer-readable storage medium stores a program, which implements the method when executed by a processor.

[0016] The above solution of the present invention includes at least the following beneficial effects: This paper uses a dynamic time warping algorithm to calculate the time series similarity of energy consumption data from building sub-items and sub-divisions, and then transforms this data into a graph structure model consisting of nodes and edges. This graph structure can fully capture the complex dependencies between different sub-tables, including temporal dynamics and regional correlations.

[0017] This paper proposes an unsupervised anomaly detection method that identifies both global and point anomalies by analyzing the overall structure and local features of a graph. This method eliminates the need for labeled data, significantly reducing the need for manual intervention and improving the feasibility of practical applications.

[0018] Hierarchical clustering is used to generate baseline clusters of building energy consumption, which represent typical energy consumption patterns of buildings over different time periods. Combined with real-time graph structures, this method supports multi-level anomaly analysis from individual sub-tables to the entire building. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The present invention provides a flowchart of a method for detecting abnormal energy consumption in commercial buildings based on a graph method.

[0020] Figure 2 This is a schematic diagram of a commercial building energy consumption anomaly detection system based on a graph method provided by an embodiment of the present invention.

[0021] Figure 3 It is a schematic diagram of the workflow of the method for detecting abnormal energy consumption of building sub-item and sub-division sub-meters according to the present invention. DETAILED DESCRIPTION

[0022] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0023] like Figure 1 As shown, an embodiment of the present invention proposes a method for detecting abnormal energy consumption in commercial buildings based on a graph method, the method comprising the following steps: Step 11: collect real-time energy consumption data from the sub-item and sub-area sub-tables in the building to generate multiple sets of energy consumption time series with time series characteristics; Step 12: Perform dynamic time warping calculation on the energy consumption time series, quantify the energy consumption time series similarity between each sub-item partition sub-table, and construct a graph structure based on the similarity, in which nodes are mapped to sub-item partition sub-tables, and edge weights represent the time series similarity between corresponding nodes; Step 13, generating a spectral distribution representing the topological characteristics of the graph structure by calculating the Laplace matrix eigenvalue set of the graph structure, and performing hierarchical clustering on multiple graph structures based on the Bhattacharya distance between the spectral distributions to obtain clustering results of energy consumption patterns in different time periods; Step 14: Performing a silhouette coefficient analysis on the clustering results, selecting the number of clusters corresponding to the maximum silhouette coefficient as the corresponding division, and defining the cluster set containing historical normal data as the baseline cluster; Step 15: Generate a graph structure to be detected for the newly collected energy consumption data, calculate the minimum spectral distance between the graph structure and all graph structures in the baseline cluster, and determine it as an abnormal graph structure if the minimum value exceeds a preset threshold; Step 16: Analyze the node connection strength of the abnormal graph structure, calculate the decrease in edge weight between each node and its adjacent nodes, and mark the sub-item partition sub-table associated with the node with the largest edge weight loss as an abnormal source; Step 17: Generate a detection report including equipment maintenance prompts and energy efficiency optimization strategies based on the spatial distribution characteristics of the abnormal sources and the time series characteristics of energy consumption mutations.

[0024] By using sub-tables partitioned by building items, multi-dimensional, high-resolution energy consumption data is obtained, covering real-time power usage in different areas and devices. The DTW algorithm aligns time-shifted, asynchronous sub-table data, accurately capturing similarities in temporal patterns. It transforms abstract energy consumption data into a node-edge graph structure, intuitively reflecting the dependencies between sub-tables. Spectral distance quantifies graph structural differences using the eigenvalues ​​of the graph's Laplacian matrix, capturing overall topological characteristics and avoiding the biased nature of relying solely on local edge weights. The silhouette coefficient automatically selects the optimal number of clusters by evaluating intra-cluster closeness and inter-cluster separation, eliminating the subjectivity of manual selection. By comparing the spectral distance of the real-time graph with the baseline clusters, it quickly identifies deviations from the normal range of overall energy consumption patterns. Detection latency is low, and by calculating changes in edge weights at individual nodes, it can pinpoint specific abnormal devices or areas. Anomaly detection results directly drive optimization measures and are integrated with the building management system in real time, shortening the "detection-response" cycle and reducing energy waste and equipment failure losses.

[0025] In a preferred embodiment of the present invention, the above step 11 includes: Step 111, recording the energy consumption data of each area through the building management system according to the building area where the energy consumption data is collected; Step 112: Summarize the collected energy consumption data of each sub-item partition sub-table to form a preliminary data set; Step 113 , preprocessing the preliminary data set to obtain preprocessed energy consumption data.

[0026] In an embodiment of the present invention, by locating the source of energy consumption and automatically recording the energy consumption data of each area using a building management system, the complexity and errors of data collection are reduced, the efficiency of data collection is improved, the data of each sub-item and partition sub-table is processed to form a comprehensive data set and facilitate data management and analysis, and the preliminary data set is pre-processed to improve data quality and enhance data availability, thereby improving the efficiency of energy consumption management.

[0027] In a specific embodiment of the present invention, the specific implementation steps include: Step 111 , clearly define the building areas where energy consumption data is to be collected, including each floor, room, equipment area, etc. in the building, and set corresponding data collection points in the building management system so that the system can record the energy consumption data of each area.

[0028] Step 112: extract the energy consumption data of each sub-item and partition sub-table from the building management system, perform preliminary cleaning on the extracted data, remove duplicate, erroneous or invalid data records, and aggregate the cleaned data according to dimensions such as region and time to form a preliminary data set.

[0029] Step 113 , standardize the preliminary data set, fill in or delete the missing energy consumption data, and detect and process abnormal values ​​in the data.

[0030] In a preferred embodiment of the present invention, the above step 12 includes: Step 121: Based on the collected energy consumption data, identify all sub-item and partition sub-tables in the building, where each sub-table corresponds to an independent energy consumption monitoring unit of a specific area or equipment, and extract the energy consumption time series of each sub-table; Step 122: Perform dynamic time warping calculations on the energy consumption time series of any two item-partitioned sub-tables: Generate multiple matching paths using nonlinear time axis alignment, calculate the squared Euclidean distance between adjacent matching elements in each path, accumulate all the squared local distances on the path to obtain the total cumulative distance, select the path with the smallest total cumulative distance, and use the square root of the total cumulative distance of this path as the dynamic time warping distance of the two sequences to represent the similarity of the energy consumption time series of the two sub-tables; Step 123: Map the dynamic time warping distance to the interval (0, 1) using a linear normalization function to generate an edge weight value, wherein the smaller the distance, the closer the similarity score after mapping is to 1, and the larger the distance, the closer the score is to 0; In step 124 , a fully connected graph structure is constructed using the item partition sub-tables as nodes and the normalized similarity scores as edge weights, wherein the weight value of the edge between any two nodes reflects the similarity of the energy consumption time series of the corresponding sub-tables.

[0031] In an embodiment of the present invention, by identifying the sub-item partition sub-tables and clarifying the analysis objects, the targeted analysis is improved. The introduction of the dynamic time warping algorithm cleverly captures the dynamic changes in the energy consumption data. The time series similarity is evaluated through a nonlinear alignment path, and the dissimilarity is converted into edge weights. The similarity of the energy consumption time series between sub-tables is quantified. The construction of the graph structure provides key elements, so that the graph structure can accurately reflect the energy consumption relationship between each sub-table, and finally the graph structure is constructed.

[0032] In a specific embodiment of the present invention, the specific implementation steps include: Step 121: Obtain pre-processed energy consumption data, analyze the structure of the energy consumption data, identify different sub-item partition sub-tables, each sub-table usually corresponding to a specific area or equipment in the building, and record the identification (such as name, number) of each sub-table and the area or equipment information it monitors.

[0033] Step 122, pass , calculate and evaluate the similarity of two time series, where, It is a time series X and Y The dynamic time warping distance between X and Y Represents two time series, which are vectors consisting of a series of data points arranged in time order. yes X and Y An alignment path between are all possible alignment paths The smallest path in It is the path All points on (i,j) Perform the summation, Yes X Middle i Elements and Y Middle j Elements The square of the Euclidean distance between It's right (i,j) Is the path An element in i It is a time series X The index in j It is a time series YFor the index in, for the energy consumption time series of any two sub-tables, the DTW algorithm is used to calculate the corresponding alignment path. Through a path with nonlinear alignment on the time axis, the elements of the sequence are matched in order. For all matching elements on the path, the square of the Euclidean distance is calculated, and all the squares of the local distances on the path are accumulated to obtain the total cumulative distance of the path. Among all legal paths, the path with the smallest total cumulative distance is selected as the corresponding alignment path. The square root of the total cumulative distance of the corresponding path is taken to obtain the dynamic time warping distance of the two sequences.

[0034] Step 123, based on the dynamic time warping distance calculated in step 122, determine the dissimilarity between any two sub-tables, by , transform edge weights, where It is a time series and The dynamic time warping similarity between It is a time series and The dynamic time warping distance between the two sub-table energy consumption time series is calculated by Mapped to the interval (0, 1), where ，d is the original DTW distance value, is the minimum value among all calculated DTW distance values, Is the maximum value of all calculated DTW distance values s It is the similarity score after mapping, and its value range is (0, 1). The intuitive similarity score formed is the transformed edge weight, which indicates the closeness of the energy consumption relationship between the two sub-tables.

[0035] In step 124, each sub-item partition sub-table is regarded as a node in the graph structure. According to the edge weight calculated in step 123, a connecting edge is established between the nodes. The edge weight represents the closeness of the energy consumption relationship between the two node sub-tables. All nodes and edges are combined to form a complete graph structure.

[0036] In a preferred embodiment of the present invention, the above step 13 includes: Step 131: For the graph structure of each time period, a standardized Laplacian matrix is ​​constructed based on the edge weights between nodes, and eigenvalue decomposition is performed on it to extract a set of eigenvalues ​​arranged in ascending order to generate a spectral distribution that represents the topological characteristics of the graph structure; Step 132: For the spectral distributions of any two graph structures, calculate their Bhattacharya distance: square the eigenvalue differences at corresponding positions of the two eigenvalue sets and add them together, then take the square root of the sum. The resulting value is defined as the spectral distance between the two graph structures. Step 133: Based on the spectral distance matrix between all graph structures, an agglomerative hierarchical clustering algorithm is used to gradually merge the graph structures or clusters with the closest distance using a single-link strategy until a preset number of initial cluster sets are formed; In step 134 , based on the dendrogram generated during the hierarchical clustering process, hierarchical division results that meet a preset merging threshold are intercepted, and the graph structures of different time periods are classified into corresponding energy consumption pattern clusters to form a clustering result set with temporal continuity.

[0037] In an embodiment of the present invention, the Laplace matrix can transform the topological relationship between nodes and edges into a linear algebraic structure, making it easier to extract the global features of the graph through matrix operations. The spectral characteristics of the graph are captured by the eigenvalue sequence of the Laplace matrix. The spectral distance can accurately measure the global topological differences of different graph structures. The eigenvalues ​​from large to small correspond to the different scale characteristics of the graph. The spectral distance improves the accuracy of clustering by integrating all eigenvalue differences, avoiding the one-sidedness of relying solely on a single feature. The typical graph structure is selected as the initial clustering center through spectral distance to ensure that the baseline cluster can cover the main normal patterns of building energy consumption and avoid clustering offsets caused by random initialization. Through the iterative process of "initial center to expanded cluster", similar graph structures are gradually classified into the same cluster to form a hierarchical baseline pattern system.

[0038] In a specific embodiment of the present invention, the specific implementation steps include: Step 131, construct a Laplace matrix by generating multiple graph structures, each of which represents the relationship between the energy consumption time series of different sub-tables in the building. For each graph structure, define its vertex set V and edge set E, construct the degree matrix D and adjacency matrix A. The Laplace matrix L is defined as .

[0039] Step 132, perform eigenvalue decomposition on each Laplace matrix L to obtain eigenvalues, and then The spectral distance is obtained, where It's a picture and The spectral distance between and Represent two graph structures respectively. and Respectively represent the graph and pictures The Laplace matrix of eigenvalues, k is that the number of eigenvalues ​​is usually equal to the number of vertices of the graph, From 1 to k Sum up all the eigenvalues ​​of i is an index variable, are the eigenvalues ​​of the Laplacian matrix.

[0040] In step 133 , the spectral distances between all graph structures are calculated through step 132 to form a spectral distance matrix. A number of initial cluster centers are selected based on the spectral distance matrix. The cluster centers represent the initial baseline clusters.

[0041] Step 134 : For each remaining graph structure, calculate its spectral distance from all initial cluster centers through step 132 , and assign each graph structure to the cluster center with the smallest spectral distance to form a larger cluster.

[0042] In a preferred embodiment of the present invention, the above step 14 includes: Step 141: Based on the clustering result set generated by hierarchical clustering, a partitioning scheme is set for each candidate cluster number, and the silhouette coefficients of all data points in each partitioning scheme are calculated: for a single data point, the average distance between it and the remaining data points in its cluster is calculated as a first distance, and the average distance between it and all data points in the nearest remaining cluster is calculated as a second distance. The difference between the second distance and the first distance is divided by the maximum value of the two distances to obtain the silhouette coefficient of the data point; Step 142: Count the average values ​​of the silhouette coefficients of all data points in the partitioning scheme corresponding to each candidate cluster number, and select the cluster number with the largest average value as the corresponding partitioning scheme; in the cluster set corresponding to the corresponding partitioning scheme, filter the clusters containing the graph structure during the historical normal operation period, and merge them into a baseline cluster set that represents the normal mode of building energy consumption.

[0043] In an embodiment of the present invention, the similarity of the graph structure within the cluster to which it belongs is measured to ensure that the energy consumption patterns within the same cluster are highly consistent. The difference between the graph structure and other clusters is measured to avoid different patterns being mistakenly classified into the same cluster, thereby improving the discrimination of baseline clusters. The rationality of clustering is quantified by mathematical formulas to avoid the subjectivity of manual experience judgment. The silhouette coefficient is calculated independently for each graph structure to identify "fuzzy samples" at the edge of the cluster. If the number of clusters is too large, the baseline pattern will be fragmented, and normal fluctuations will be mistakenly judged as abnormal. The silhouette coefficient can suppress such situations through overall scoring. If the number of clusters is too small, the real differences will be masked, resulting in a decrease in the sensitivity of abnormality detection. The silhouette coefficient can automatically identify pattern divergence points. The calculation complexity of the silhouette coefficient is linearly related to the sample size, and is suitable for clustering optimization of large-scale historical energy consumption data in commercial buildings.

[0044] In a specific embodiment of the present invention, the specific implementation steps include: Step 141, for each cluster, , calculate the silhouette coefficient, where and They represent the average distance from the graph to other graphs in the same cluster (intra-cluster closeness) and the average distance to the nearest neighbor cluster (inter-cluster separation), i Is an index variable, and finding its average value gives the overall silhouette coefficient under the number of clusters.

[0045] Step 142, after calculating the corresponding silhouette coefficient according to the number of clusters, the silhouette coefficient is an indicator used to evaluate the clustering effect, which measures the degree of closeness of any graph within its cluster and the degree of separation from the nearest neighbor cluster. The value range of the silhouette coefficient is ,in: A value close to 1 indicates that the graph is well classified, the cluster to which it belongs is densely packed, and it is clearly distinguishable from other clusters.

[0046] A value close to 0 indicates that the graph is located at the cluster boundary and may belong to multiple clusters, so the clustering effect is not obvious.

[0047] A value less than 0 indicates that the graph may be misclassified and belong to the wrong cluster.

[0048] By averaging the silhouette coefficients of all graphs, we can obtain an overall clustering quality score for that number of clusters. Next, we traverse multiple possible numbers of clusters, calculate the corresponding average silhouette coefficients, and plot the change curve. Finally, we select the number of clusters with the highest average silhouette coefficient. A higher silhouette coefficient indicates closer clusters and more separated data points between clusters, meaning better clustering results.

[0049] In a preferred embodiment of the present invention, the above step 15 generates a graph structure to be detected for the newly collected energy consumption data, calculates the minimum spectral distance between the graph structure and all graph structures in the baseline cluster, and determines it as an abnormal graph structure if the minimum value exceeds a preset threshold, including: Step 151: Based on the newly collected energy consumption data, a graph structure to be detected is generated, and the eigenvalue set of its normalized Laplace matrix is ​​extracted to generate the corresponding spectral distribution; for each cluster in the baseline cluster set, the Bhattacharya distance between the spectral distribution of the graph structure to be detected and the spectral distributions of all historical graph structures in the cluster is calculated, and the minimum spectral distance in each cluster is recorded; the minimum spectral distances of all baseline clusters are traversed, and the smallest value is extracted as the global minimum spectral distance between the graph structure to be detected and the baseline cluster set; the global minimum spectral distance is compared with a preset abnormality judgment threshold. If the distance exceeds the threshold, the graph structure to be detected is determined to be an abnormal energy consumption mode, the abnormality judgment result is output, and the abnormal graph structure and its corresponding time period are marked as energy consumption abnormality events.

[0050] In a specific example, the specific implementation process of step 151 is as follows: First, the newly collected energy consumption data is converted into a specific graph structure that reflects the inherent connections and characteristics of the energy consumption data. This newly generated graph is called the graph structure to be detected. Next, key mathematical features are extracted from this graph structure to be detected. This involves calculating all the eigenvalues ​​of its normalized Laplacian matrix. These eigenvalues ​​are combined to form a spectral distribution that describes the characteristics of the graph structure. This spectral distribution acts like a "fingerprint" of the graph structure to be detected, representing its unique structural features.

[0051] The "fingerprint" of the graph structure to be tested is then compared with each cluster in the baseline cluster set. The baseline cluster set contains multiple pre-established clusters representing normal energy consumption patterns. Each cluster contains many historical graph structures. These historical graph structures are derived from normal energy consumption data collected in the past, and each has a corresponding spectral distribution.

[0052] For each baseline cluster, we need to calculate the similarity between the spectral distribution of the graph structure to be tested and the spectral distribution of each historical graph structure within that cluster. The similarity metric used here is the Bhattacharya distance, which measures the degree of similarity between two distributions. The smaller the distance, the more similar the two distributions are, which means the corresponding graph structures are closer. After calculating the Bhattacharya distance between the graph structure to be tested and all historical graph structures within the cluster, we only need to record the smallest distance value, as this minimum distance represents the proximity of the graph structure to the most similar historical graph structure within the cluster.

[0053] After processing all baseline clusters, each cluster has a corresponding minimum spectral distance. Next, we need to find the smallest value among all the minimum spectral distances of all clusters. This value is the global minimum spectral distance between the graph structure under test and the entire set of baseline clusters, reflecting the closeness of the graph structure under test to the most similar case among all normal energy consumption patterns.

[0054] Finally, this global minimum spectral distance is compared with a pre-set anomaly threshold. If this distance exceeds the threshold, it indicates that the graph structure under test is not close enough to the most similar cases of normal energy consumption patterns, and the difference is significant. In this case, the graph structure under test can be determined to represent an abnormal energy consumption pattern. Once an anomaly is determined, the anomaly determination result is output, and the abnormal graph structure and its corresponding time period are marked as an energy consumption anomaly event for further analysis and processing.

[0055] In a preferred embodiment of the present invention, the above step 16 includes: Step 161: For each node in the abnormal graph structure, obtain its edge weight data under the historical normal mode, and calculate the absolute value of the difference between the real-time edge weight of the node and all adjacent nodes and the historical edge weight. The difference of all adjacent edges is accumulated to obtain the connection loss of the node; Step 162: normalize the connection loss of all nodes to generate a loss intensity index in the interval [0, 1], where the node with the maximum loss is marked as a candidate anomaly source; Step 163: Based on the edge weight fluctuation range of historical normal data statistics, a dynamic judgment threshold for the connection loss amount is set. If the loss intensity index of the candidate abnormal source exceeds the threshold, the associated item partition sub-table is located as the abnormal source.

[0056] In an embodiment of the present invention, by defining the node connection loss as the sum of the edge weight changes with all neighboring nodes, starting from the association relationship between nodes, the specific node with abnormal energy consumption can be accurately located. In commercial buildings, the energy consumption data of different areas or equipment are interrelated. The specific node with the abnormality is identified from the overall graph structure to achieve refined positioning of the abnormality. The connection loss is calculated using the absolute value of the difference between the historical edge weight and the real-time edge weight. The data comparison method objectively reflects the changes in the node connection relationship, avoids subjective judgment, analyzes according to the actual changes in the data, accurately captures the subtle changes in the node connection relationship, and effectively identifies energy consumption anomalies caused by factors such as equipment failure and operation mode changes. Thresholds are set based on historical data, flexibly adapt to different building characteristics and operation laws, enhance the versatility of the method, timely mark abnormal nodes, achieve preventive maintenance, and reduce the probability of equipment failure.

[0057] In a specific embodiment of the present invention, the specific implementation steps include: Step 161: Based on the identified abnormal graph structure, for each node in the graph structure, Calculate the connection loss, where f It is used to identify a specific node in the graph structure. The index of r It is used to traverse the nodes The index of the neighbor node of yes t Time Node With neighboring nodes The weight of the edge between It's a historic moment node With neighboring nodes The weight of the edge between yes is a node Neighbor nodes of is a node The connection loss of each node is defined as the sum of the edge weight changes between the node and all neighboring nodes. Neighboring nodes refer to nodes that are directly connected to the node.

[0058] Step 162, according to each node in the abnormal graph structure, obtain the historical edge weight and real-time edge weight. The historical edge weight is the weight value of the edge between the node and the neighboring node under normal energy consumption mode, and the real-time edge weight is the edge weight value corresponding to the current moment. The node connection loss is calculated through the steps.

[0059] Step 163: Perform statistical analysis on the node connection loss based on the historically collected energy consumption data and the corresponding graph structure information. For example, analyze the distribution of node connection loss in different time periods and under different working conditions, find the fluctuation range of connection loss under normal circumstances, set a connection loss threshold, and compare the connection loss calculated for each node with the set threshold. If the connection loss of a node exceeds the set threshold, it is determined that the connection relationship of the node has changed and the node is marked as an abnormal node.

[0060] In a preferred embodiment of the present invention, the above step 17 includes: Step 171: Process the spectral characteristics of the anomaly source location results and the anomaly graph structure, analyze the spatial distribution density of the sub-item partition sub-table corresponding to the anomaly source and the temporal fluctuation pattern of the energy consumption mutation time period, and extract the space-time correlation anomaly characteristics. Based on the space-time correlation anomaly characteristics and in combination with the fault mode case library in the historical energy consumption event database, match the anomaly characteristics with the similarity of known fault modes, and infer the cause of the anomaly, including equipment failure, human operation abnormality, or environmental interference. Step 172: Generate targeted optimization suggestions based on the type of abnormality cause: generate an equipment maintenance list and priority ranking for equipment failure type abnormalities, generate an energy efficiency parameter adjustment strategy for energy consumption pattern mutation type abnormalities, and associate the spatial location information of the area where the abnormality source is located; process the optimization suggestions, abnormality cause analysis and related data features into a structured detection report.

[0061] In an embodiment of the present invention, the overall anomaly is judged by combining the spectral distance with the confidence interval threshold, and the node connection loss is used to locate the point anomaly, so as to achieve accurate anomaly diagnosis, in-depth analysis of the anomaly, and accurate attribution based on historical and real-time data to avoid superficial problem solving. Optimization suggestions are generated based on scientific analysis to help managers make reasonable decisions and reduce costs. By real-time docking with the building management system, the level of management intelligence and automation is enhanced, and anomaly handling is timely and efficient. Its closed-loop mode of detection, analysis, suggestion, and execution facilitates experience accumulation and strategy optimization, so that the system can adapt to the operating characteristics of the building.

[0062] In a specific embodiment of the present invention, the specific implementation steps include: Step 171, by combining historical energy consumption data and corresponding graph structure information such as energy consumption patterns and node connection relationships in different time periods in history, and the current real-time collected energy consumption data and graph structure, identify the type of anomaly, such as whether the energy consumption anomaly is caused by equipment failure, improper operating parameter settings, changes in external environmental factors, etc., and find out the root cause of the anomaly.

[0063] Step 172 , based on the analysis of the cause of the abnormality, formulate corresponding optimization suggestions. For example, if the HVAC system power consumption in a certain area is abnormally high, the system will recommend adjusting the air conditioning operating parameters; if certain equipment consumes too much power, it may be recommended to arrange maintenance or replace the equipment.

[0064] like Figure 2 As shown, an embodiment of the present invention further provides a commercial building energy consumption anomaly detection system based on a graph method, comprising: Acquisition module 21 is used to collect real-time energy consumption data from the sub-item partition sub-tables within the building and generate multiple sets of energy consumption time series with time series characteristics; perform dynamic time warping calculations on the energy consumption time series, quantify the energy consumption time series similarity between each sub-item partition sub-table, and construct a graph structure based on the similarity, in which nodes are mapped to the sub-item partition sub-tables, and edge weights represent the time series similarity between corresponding nodes; The comparison module 22 is configured to generate a spectral distribution representing the topological characteristics of the graph structure by calculating the set of eigenvalues ​​of the Laplace matrix of the graph structure, and hierarchically cluster the multiple graph structures based on the Bhattacharya distance between the spectral distributions to obtain clustering results of energy consumption patterns in different time periods; perform silhouette coefficient analysis on the clustering results, select the number of clusters corresponding to the maximum silhouette coefficient as the corresponding division, and define the cluster set containing historical normal data as the baseline cluster; generate a graph structure to be detected for the newly collected energy consumption data, calculate the minimum value of its spectral distance with all graph structures in the baseline cluster, and determine that the graph structure is abnormal if the minimum value exceeds a preset threshold; Processing module 23 is used to analyze the node connection strength of the abnormal graph structure, calculate the decrease in edge weight between each node and its adjacent nodes, and mark the sub-item partition sub-table associated with the node with the largest edge weight loss as an abnormal source; based on the spatial distribution characteristics and energy consumption mutation time series characteristics of the abnormal source, generate a detection report containing equipment maintenance prompts and energy efficiency optimization strategies.

[0065] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for detecting abnormal building energy consumption based on a graph method, characterized in that: The method comprises: Collect real-time energy consumption data from the sub-item and sub-area sub-tables within the building to generate multiple sets of energy consumption time series with time series characteristics; Performing dynamic time warping calculation on the energy consumption time series, quantifying the energy consumption time series similarity between each sub-item partition sub-table, and constructing a graph structure based on the similarity, in which nodes are mapped to the sub-item partition sub-tables, and edge weights represent the time series similarity between corresponding nodes; By calculating the Laplace matrix eigenvalue set of the graph structure, a spectral distribution representing the topological characteristics of the graph structure is generated, and hierarchical clustering is performed on multiple graph structures based on the Bhattacharya distance between the spectral distributions to obtain clustering results of energy consumption patterns in different time periods; Performing silhouette coefficient analysis on the clustering results, selecting the number of clusters corresponding to the maximum silhouette coefficient as the corresponding division, and defining the cluster set containing historical normal data as the baseline cluster; Generate a graph structure to be detected for the newly collected energy consumption data, calculate the minimum spectral distance between it and all graph structures in the baseline cluster, and determine it as an abnormal graph structure if the minimum value exceeds a preset threshold; Performing a node connection strength analysis on the abnormal graph structure, calculating the edge weight drop between each node and its adjacent nodes, and marking the sub-item partition sub-table associated with the node with the largest edge weight loss as an abnormal source; Based on the spatial distribution characteristics of the abnormal sources and the time series characteristics of energy consumption mutations, a detection report including equipment maintenance prompts and energy efficiency optimization strategies is generated.

2. The method for detecting abnormal energy consumption in commercial buildings based on graph method according to claim 1, characterized in that: Perform dynamic time warping calculation on the energy consumption time series, quantify the energy consumption time series similarity between each sub-item partition sub-table, and construct a graph structure based on the similarity, in which nodes are mapped to sub-item partition sub-tables, and edge weights represent the time series similarity between corresponding nodes, including: Based on the collected energy consumption data, identify all sub-item and partition sub-meters in the building, where each sub-meter corresponds to an independent energy consumption monitoring unit for a specific area or equipment, and extract the energy consumption time series of each sub-meter; For the energy consumption time series of any two sub-item partitioned sub-tables, dynamic time warping calculation is performed: multiple matching paths are generated through nonlinear time axis alignment. The squared Euclidean distance between adjacent matching elements in each path is calculated. The total cumulative distance is obtained by summing up all the squared local distances on the path. The corresponding path with the smallest total cumulative distance is selected, and the square root of the total cumulative distance of this path is used as the dynamic time warping distance of the two sequences to represent the similarity of the energy consumption time series of the two sub-tables. Mapping the dynamic time warping distance to the interval (0, 1) through a linear normalization function to generate an edge weight value, where the smaller the distance, the closer the similarity score after mapping is to 1, and the larger the distance, the closer the score is to 0; A fully connected graph structure is constructed with the item partitioned sub-tables as nodes and the normalized similarity scores as edge weights, where the weight value of the edge between any two nodes reflects the similarity of the energy consumption time series of the corresponding sub-tables.

3. The method for detecting abnormal energy consumption in commercial buildings based on graph method according to claim 2, characterized in that: By calculating the Laplace matrix eigenvalue set of the graph structure, a spectral distribution representing the topological characteristics of the graph structure is generated. Multiple graph structures are hierarchically clustered based on the Bhattacharya distance between the spectral distributions to obtain clustering results of energy consumption patterns in different time periods, including: For the graph structure of each time period, a standardized Laplacian matrix is ​​constructed based on the edge weights between nodes, and eigenvalue decomposition is performed on it to extract the eigenvalue set arranged in ascending order, generating a spectral distribution that characterizes the topological characteristics of the graph structure; For the spectral distribution of any two graph structures, calculate their Bhattacharya distance: add up the squares of the eigenvalue differences at corresponding positions of the two eigenvalue sets, and take the square root of the sum. The resulting value is defined as the spectral distance between the two graph structures. Based on the spectral distance matrix between all graph structures, an agglomerative hierarchical clustering algorithm is used to gradually merge the graph structures or clusters with the closest distance using a single-link strategy until a preset number of initial cluster sets are formed. According to the dendrogram generated in the hierarchical clustering process, the hierarchical division results that meet the preset merging threshold are intercepted, and the graph structures of different time periods are classified into corresponding energy consumption pattern clusters to form a clustering result set with time continuity.

4. The method for detecting abnormal energy consumption in commercial buildings based on graph method according to claim 3, characterized in that: Perform silhouette coefficient analysis on the clustering results, select the number of clusters corresponding to the maximum silhouette coefficient as the corresponding division, and define the cluster set containing historical normal data as the baseline cluster, including: Based on the clustering result set generated by hierarchical clustering, a partitioning scheme is set for each candidate cluster number, and the silhouette coefficients of all data points in each partitioning scheme are calculated: for a single data point, the average distance between it and the remaining data points in its cluster is calculated as the first distance, and the average distance between it and all data points in the nearest remaining cluster is calculated as the second distance. The difference between the second distance and the first distance is divided by the maximum value of the two distances to obtain the silhouette coefficient of the data point; The average value of the silhouette coefficients of all data points in the partitioning scheme corresponding to each candidate cluster number is calculated, and the cluster number with the largest average value is selected as the corresponding partitioning scheme; in the cluster set corresponding to the corresponding partitioning scheme, the clusters containing the graph structure during the historical normal operation period are screened and merged into a baseline cluster set that represents the normal mode of building energy consumption.

5. The method for detecting abnormal energy consumption in commercial buildings based on graph method according to claim 4, characterized in that: Generate a graph structure to be detected for the newly collected energy consumption data, calculate the minimum spectral distance between it and all graph structures in the baseline cluster, and determine it as an abnormal graph structure if the minimum value exceeds a preset threshold, including: Based on the newly collected energy consumption data, the graph structure to be detected is generated, and the eigenvalue set of its normalized Laplace matrix is ​​extracted to generate the corresponding spectral distribution; For each cluster in the baseline cluster set, respectively calculating the Bhattacharya distance between the spectral distribution of the graph structure to be detected and the spectral distributions of all historical graph structures in the cluster, and recording the minimum spectral distance in each cluster; Traverse the minimum spectral distances of all baseline clusters and extract the smallest value as the global minimum spectral distance between the graph structure to be detected and the baseline cluster set; The global minimum spectral distance is compared with a preset abnormality judgment threshold. If the distance exceeds the threshold, the graph structure to be detected is determined to be an abnormal energy consumption mode, the abnormality judgment result is output, and the abnormal graph structure and its corresponding time period are marked as an energy consumption abnormality event.

6. The method for detecting abnormal energy consumption in commercial buildings based on graph method according to claim 5, characterized in that: Perform node connection strength analysis on the abnormal graph structure, calculate the edge weight drop between each node and its adjacent nodes, and mark the sub-item partition sub-table associated with the node with the largest edge weight loss as an abnormal source, including: For each node in the abnormal graph structure, obtain its edge weight data in the historical normal mode, and calculate the absolute value of the difference between the real-time edge weight of the node and all adjacent nodes and the historical edge weight. The difference of all adjacent edges is accumulated to obtain the connection loss of the node; Normalize the connection loss of all nodes to generate a loss intensity index in the interval [0, 1], where the node with the maximum loss is marked as a candidate anomaly source; Based on the edge weight fluctuation range of historical normal data statistics, a dynamic judgment threshold for connection loss is set. If the loss intensity index of a candidate abnormal source exceeds the threshold, its associated sub-item partition sub-table is identified as an abnormal source.

7. The method for detecting abnormal energy consumption in commercial buildings based on graph method according to claim 6, characterized in that: Based on the spatial distribution characteristics of the abnormal sources and the time series characteristics of energy consumption mutations, a detection report containing equipment maintenance prompts and energy efficiency optimization strategies is generated, including: The spectral characteristics of the anomaly source location results and the anomaly graph structure are processed to analyze the spatial distribution density of the sub-item partition sub-tables corresponding to the anomaly source and the temporal fluctuation pattern of the energy consumption mutation period to extract the space-time correlation anomaly characteristics. Based on these space-time correlation anomaly characteristics and combined with the fault mode case library in the historical energy consumption event database, the similarity between the anomaly characteristics and known fault modes is matched to infer the type of anomaly cause, including equipment failure, human operation abnormality, or environmental interference. Generate targeted optimization suggestions based on the type of anomaly cause: generate an equipment maintenance list and priority ranking for equipment failure anomalies, generate energy efficiency parameter adjustment strategies for energy consumption pattern mutation anomalies, and associate the spatial location information of the area where the anomaly source is located; process the optimization suggestions, anomaly cause analysis and related data features into a structured detection report.

8. A commercial building energy consumption anomaly detection system based on a graph method, the system implementing the method according to any one of claims 1 to 7, characterized in that: include: The acquisition module is used to collect real-time energy consumption data from the sub-item and partition sub-tables in the building and generate multiple sets of energy consumption time series with time series characteristics; Performing dynamic time warping calculation on the energy consumption time series, quantifying the energy consumption time series similarity between each sub-item partition sub-table, and constructing a graph structure based on the similarity, in which nodes are mapped to the sub-item partition sub-tables, and edge weights represent the time series similarity between corresponding nodes; A comparison module is configured to generate a spectral distribution representing the topological characteristics of the graph structure by calculating the set of Laplace matrix eigenvalues ​​of the graph structure, and hierarchically cluster multiple graph structures based on the Bhattacharya distance between the spectral distributions to obtain clustering results of energy consumption patterns in different time periods; perform silhouette coefficient analysis on the clustering results, select the number of clusters corresponding to the maximum silhouette coefficient as the corresponding division, and define the cluster set containing historical normal data as the baseline cluster; generate a graph structure to be detected for the newly collected energy consumption data, calculate the minimum value of its spectral distance with all graph structures in the baseline cluster, and determine that the graph structure is abnormal if the minimum value exceeds a preset threshold; A processing module is used to analyze the node connection strength of the abnormal graph structure, calculate the edge weight drop between each node and its adjacent nodes, and mark the sub-item partition sub-table associated with the node with the largest edge weight loss as an abnormal source; Based on the spatial distribution characteristics of the abnormal sources and the time series characteristics of energy consumption mutations, a detection report including equipment maintenance prompts and energy efficiency optimization strategies is generated.

9. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images