Industrial Early Warning and Prevention Methods and Systems Based on Industrial Power Generation Linkage

Abstract

This invention relates to an industrial sector early warning and control method and system based on industry-specific electricity consumption linkage. The method includes: collecting electricity consumption information from various regions and industries; statistically analyzing the electricity consumption information to obtain monthly and weekly electricity consumption data for each region and industry; performing DBSCAN clustering on the monthly electricity consumption data to obtain clustering results for key industrial chains; sequentially performing DBSCAN clustering and K-means clustering based on the DTW algorithm on the weekly electricity consumption data to obtain regional and industry clustering results; using the Louvain algorithm to perform community relationship analysis on the clustering results for key industrial chains and regional and industry clusters to obtain electricity transmission relationships between key industrial chains and between industrial sectors in different regions; and using these electricity transmission relationships to conduct industrial sector early warning and control. This invention effectively improves the accuracy and efficiency of industrial sector early warning and control.

Description

Technical Field

[0001] This invention belongs to the field of energy data analysis and early warning technology, specifically relating to an industrial early warning and prevention method and system based on industrial electricity generation linkage. Background Technology

[0002] The development of the industrial economy is strongly correlated with electricity fluctuation cycles in terms of development speed, amplitude, and correlation. The industrial economy is the mainstay and growth engine of the national economy, and the industrial and supply chains are its lifeblood. In recent years, special events such as logistical disruptions and rising raw material prices have significantly impacted the industrial economy. However, current industrial electricity consumption trend forecasting takes into account the electricity transfer relationships between industries and within the industrial chain, resulting in low accuracy in trend forecasting and an inability to provide accurate data support for early warning in the industrial sector. Summary of the Invention

[0003] Based on the aforementioned shortcomings and deficiencies in the existing technology, one of the objectives of this invention is to at least solve one or more of the aforementioned problems in the existing technology. In other words, one of the objectives of this invention is to provide an industrial early warning and prevention method and system based on industrial power linkage that meets one or more of the aforementioned needs.

[0004] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0005] The industrial early warning and prevention method based on industrial power generation linkage includes the following steps:

[0006] S1. Collect electricity consumption information from various regions and industries;

[0007] S2. Collect monthly and weekly electricity consumption data for each region and industry by statistically analyzing electricity consumption information for each region and industry.

[0008] S3. Perform DBSCAN clustering on the monthly electricity consumption data of various regions and industries to obtain the clustering results of key industrial chains; among them, the clustering results of key industrial chains include several key industrial chains and the industries they contain.

[0009] The weekly electricity consumption data of each region and industry were clustered in two steps to obtain the regional and industry clustering results. The first step of clustering adopted DBSCAN clustering based on the DTW algorithm, and the second step adopted K-means clustering based on the DTW algorithm.

[0010] S4. Using the Louvain algorithm, community relationship analysis was performed on the clustering results of key industrial chains and regional industries to obtain the power transfer relationship between key industrial chains and between industrial sectors in different regions.

[0011] S5. Utilize the power transmission relationships between key industrial chains and between different regions and industrial sectors to conduct early warning and prevention of industrial development.

[0012] As a preferred embodiment, in step S1, the electricity consumption information of each region and industrial sector includes the region, industry name, industry code and its corresponding daily electricity consumption, and is stored in the database.

[0013] As a preferred embodiment, step S2 further includes:

[0014] The monthly and weekly electricity consumption data of various regions and industrial sectors obtained from the statistics were cleaned and normalized.

[0015] As a preferred approach, the data cleaning involves filling in missing data, and interpolation is performed using the interpolate function of the Pandas module.

[0016] As a preferred option, the normalization process employs Z-Score normalization.

[0017] As a preferred embodiment, step S4 further includes:

[0018] Each cluster based on the clustering results of key industrial chains or regional industries is used as a node. Pearson correlation analysis is used to obtain the correlation coefficient and differential correlation coefficient of the rate of change of weekly electricity consumption curves between nodes. The correlation coefficient and differential correlation coefficient are then reduced in dimensionality using PCA, and the resulting one-dimensional coefficient is used as the weight of the edge in the community relationship analysis.

[0019] As a preferred embodiment, the power transfer relationship includes one or more of the following: DTW distance, correlation coefficient, differential correlation coefficient, and each offset step size, number of offsets, and maximum offset step size obtained by the DTW algorithm.

[0020] As a preferred embodiment, step S5 further includes:

[0021] The clustering results of key industrial chains, regional industry clustering results, electricity transfer relationships between key industrial chains, and electricity transfer relationships between different regions and industrial sectors are merged and summarized according to different industrial chains, different regions, and different industries to form an industry electricity transfer relationship lookup table.

[0022] As a preferred embodiment, the industry electricity consumption relationship query table includes a tag query table and a correlation query table. The tag query table includes regional tags, industry tags, cluster tags, and community tags. The correlation query table includes DTW distance, correlation coefficient, and lead-lag value. The lead-lag value is the offset step size obtained by the DTW algorithm.

[0023] This invention also provides an industrial early warning and prevention system based on industrial power generation linkage, applying the industrial early warning and prevention method described in any of the preceding solutions, wherein the industrial early warning and prevention system includes:

[0024] The data acquisition module is used to collect electricity consumption information from various regions and industries.

[0025] The statistics module is used to collect electricity consumption information from various regions and industries to obtain monthly and weekly electricity consumption data for each region and industry.

[0026] The clustering module is used to perform DBSCAN clustering on monthly electricity consumption data of various regions and industries to obtain clustering results for key industrial chains. The clustering results for key industrial chains include several key industrial chains and the industries they contain. It is also used to perform two-step clustering on weekly electricity consumption data of various regions and industries to obtain regional and industry clustering results. The first step of clustering uses DBSCAN clustering based on the DTW algorithm, and the second step of clustering uses K-means clustering based on the DTW algorithm.

[0027] The community relationship analysis module is used to perform community relationship analysis on the clustering results of key industrial chains and regional industries using the Louvain algorithm, respectively, to obtain the power transfer relationship between key industrial chains and the power transfer relationship between industrial sectors in different regions;

[0028] The early warning and prevention module is used to conduct industrial early warning and prevention by utilizing the power transmission relationships between key industrial chains and between different regions and industrial sectors.

[0029] Compared with the prior art, the beneficial effects of this invention are:

[0030] This invention utilizes the DTW distance algorithm of the DTW algorithm to perform DBSCAN and K-means clustering when clustering weekly electricity data. The DTW distance is used to analyze the transmission relationship of industry electricity in time and space, thereby improving the accuracy of clustering. It also adopts the Louvain community discovery algorithm to construct the industry-region electricity correlation of the industrial sector, providing accurate data support for industrial early warning and prevention, which is conducive to improving the accuracy of industrial early warning and prevention.

[0031] The clustering optimization method in the industrial early warning and prevention method of this invention is nearly 60 times more efficient than the traditional clustering analysis method. Attached Figure Description

[0032] Figure 1 This is a flowchart of the industrial early warning and prevention method based on industrial power generation linkage according to Embodiment 1 of the present invention;

[0033] Figure 2This is a flowchart of a portion of the industrial early warning and prevention method based on industrial power generation linkage according to Embodiment 1 of the present invention.

[0034] Figure 3 This is the key industrial chain clustering result of Embodiment 1 of the present invention;

[0035] Figure 4 This is the clustering result of the first step of clustering in Embodiment 1 of the present invention;

[0036] Figure 5 This is the clustering result of the second step of clustering in Embodiment 1 of the present invention;

[0037] Figure 6 This is a community analysis relationship diagram corresponding to the power transfer relationship between the key industrial chains in Embodiment 1 of the present invention;

[0038] Figure 7 This is a community analysis relationship diagram corresponding to the power transfer relationship between industrial sectors in different regions in Embodiment 1 of the present invention;

[0039] Figure 8 This is a graph showing the changes in electricity consumption in the pharmaceutical manufacturing industry in Xinchang and the rubber and plastic products industry in Fenghua in 2020.

[0040] Figure 9 This is a graph showing the changes in electricity consumption between the general equipment manufacturing industry in Haining and other manufacturing industries in Xiaoshan in 2021.

[0041] Figure 10 This is a diagram illustrating the architecture of an industrial early warning and control system based on industrial power generation linkage, as described in Embodiment 1 of the present invention. Detailed Implementation

[0042] To more clearly illustrate the embodiments of the present invention, specific implementation methods will be described below with reference to the accompanying drawings. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings and other implementation methods can be obtained based on these drawings without any creative effort.

[0043] Example 1:

[0044] like Figure 1 and 2 The industrial early warning and prevention method based on industrial power generation linkage in this embodiment includes the following steps:

[0045] S1. Collect electricity consumption information from various regions and industries.

[0046] Specifically, the electricity consumption information for each region and industrial sector includes the region, industry name, industry code, and corresponding daily electricity consumption, and is stored in the database for subsequent data statistics and processing.

[0047] S2. Collect statistics on electricity consumption information of various regions and industries to obtain monthly and weekly electricity consumption data for each region and industry.

[0048] Specifically, the study focuses on industrial electricity consumption in Zhejiang Province. The research sample consists of two parts: monthly electricity consumption data and weekly electricity consumption data.

[0049] In addition, after obtaining the monthly and weekly electricity consumption data for various regions and industries, data cleaning and normalization were performed on the collected data. Specifically, for data that is missing or zero, data cleaning was required, i.e., filling in the missing data. This was done using the `interpolate` function of the Pandas module.

[0050] In this embodiment, after data cleaning is completed, normalization is performed, specifically using Z-Score normalization.

[0051] S3. Perform DBSCAN clustering on the monthly electricity consumption data of various regions and industries to obtain the clustering results of key industrial chains; among them, the clustering results of key industrial chains include several key industrial chains and the industries they contain.

[0052] Specifically, the DBSCAN algorithm is a density-based clustering algorithm that defines a cluster as the largest set of density-connected points. Two important parameters in DBSCAN are eps and min_samples. A larger eps result in fewer clusters; min_samples needs to be selected appropriately based on the actual characteristics of the data and the target result.

[0053] The DBSCAN algorithm is used to analyze the consistency of monthly electricity consumption data curves and to identify key industrial chains, serving as the first layer (industry layer) for subsequent analysis of industrial relationships. For example... Figure 3 As shown, by adjusting the parameters eps=2.5 and min_samples=3, the final clustering result was obtained, with 34 industrial sectors clustered into 7 major industrial chains.

[0054] In addition, the weekly electricity consumption data of each region and industry were clustered in two steps to obtain the regional and industry clustering results. The first step of clustering adopted DBSCAN clustering based on the DTW algorithm, and the second step adopted K-means clustering based on the DTW algorithm.

[0055] Specifically, this embodiment employs a two-step clustering method to analyze the consistency of weekly electricity consumption data curves. The first step, clustering, aims to identify industries with similar electricity consumption changes and eliminate those with obvious inconsistencies. DBSCAN clustering based on the DTW algorithm is used. The custom distance function DTW algorithm calculates the minimum distance between weekly electricity consumption curves when different time-series shifts occur, meeting the consistency requirements when curves shift. It can also calculate the step size, number of shifts, and maximum step size for each shift. Specific calculations can be found in existing DTW algorithms and will not be elaborated here. The second step, clustering, aims to split the large clusters that performed poorly in the first step. K-means clustering based on the DTW algorithm is used, comprehensively considering the silhouette coefficient, to split the large clusters as much as possible.

[0056] The two-step clustering result in this embodiment is the second-level structure (regional layer) for subsequent analysis of industrial relationships.

[0057] When the DBSCAN algorithm parameters eps and min_sample are selected as 0.8 and 4 respectively, the first step of clustering results is optimal. Figure 4 As shown; when the maximum initial number of centers m = 18 in the K-means algorithm, the splitting effect of the second-step clustering result is the best, as shown. Figure 5 As shown.

[0058] S4. Using the Louvain algorithm, community relationship analysis is performed on the clustering results of key industrial chains and regional industries to obtain the power transfer relationship between key industrial chains and between industrial sectors in different regions.

[0059] In addition, this embodiment uses each cluster of key industrial chain clustering results or regional industry clustering results as a node, uses Pearson correlation analysis to obtain the correlation coefficient and differential correlation coefficient of the rate of change of weekly electricity consumption curves between each node, and uses PCA to reduce the dimensionality of the correlation coefficient and differential correlation coefficient, and the resulting one-dimensional coefficient is used as the weight of the edge in the community relationship analysis.

[0060] Specifically, the Louvain algorithm is used for social network analysis. Its core principle is to determine the importance of a node within a "community" based on the concept of centrality, thereby assessing the overall concentration and distribution of the network. Modularity is a parameter that measures the performance of the Louvain model, used to determine whether the algorithm's partitioning result has reached the optimal solution. Modularity ranges from -0.5 to 1. The optimal solution of the Louvain algorithm is achieved when node similarity is highest within a community and lowest outside the community. The Louvain community detection method focuses on finding connections between targets in the network, accurately uncovering hierarchical communities.

[0061] This embodiment uses the Louvain algorithm to analyze the community correlation of weekly electricity consumption data curves, including three steps:

[0062] 1) Node selection and edge weight calculation:

[0063] Each industry cluster (classification) in the above clustering results is taken as node i; the correlation coefficient and differential correlation coefficient of the weekly electricity consumption curves between each node are calculated using their rate of change. The two correlation coefficients are reduced to one dimension by principal component analysis (PCA), and this data is used as the weight k of the edge.

[0064] 2) Modularity Q-optimization and network cohesion stage:

[0065] Iterate through all nodes i in each sub-industry and calculate the change in modularity ΔQ:

[0066]

[0067] Σtot is the sum of the weights on the links connected to nodes in C;

[0068] k i It is the sum of the weights of the links associated with node i;

[0069] k i,in It is the sum of the links from node i to nodes in C;

[0070] m is the sum of the weights of all links in the network;

[0071] The node corresponding to the increase in ΔQ is taken as the new community label, and this process is iterated and updated continuously until the overall network no longer changes its community label and the modularity ΔQ no longer increases. The above is a brief explanation of the Louvain algorithm; for the specific process, please refer to existing technologies, which will not be elaborated here.

[0072] The results of the community relations analysis are as follows:

[0073] like Figure 6 As shown, community relationship analysis at the industry level was conducted using monthly electricity consumption data, with a maximum module value of 0.59, representing the community relationship results for key industrial chains. (Example...) Figure 7 As shown, a regional-level community relationship analysis was conducted using weekly electricity consumption data, with a maximum modularity of 0.72. The following results can be obtained from this:

[0074] Result 1: The greater the distance between points, the smaller the correlation between industries. Figure 6 The seven key industrial chains within the framework exhibit significant intra-chain clustering. For example, the agricultural and food production industrial chain (… Figure 6 The four industries in the lower right corner are relatively independent in the industrial sector, far removed from most industrial sectors, and have weak industrial linkages.

[0075] Result 2: The industry with the most connections to other industries indicates a higher industry concentration. The centrality of the industry can be calculated based on the number of connections and the weight k. For example, in the community of the paper-related industry chain ( Figure 7 (Nos. 10, 21), Paper and Paper Products Industry ( Figure 7 The industry with the highest centrality is number 15.

[0076] S5. Utilize the power transfer relationships between key industrial chains and between different regions and industrial sectors for industrial early warning and prevention. The power transfer relationships include one or more of the following: DTW distance, correlation coefficient, differential correlation coefficient, and the offset step size, number of offsets, and maximum offset step size obtained from the DTW algorithm.

[0077] The DTW (Dynamic Time Warping) distance algorithm in this embodiment is a method for measuring the similarity between two time series. It calculates the similarity between two time series by extending, shortening, and shifting the time series.

[0078] Based on the constructed community relationship results, this embodiment calculates the correlation between the electricity consumption of key industries in each industry chain corresponding to each community; it uses the DTW algorithm to calculate the sequence distance, slices the data, and continuously adjusts the offset to minimize the distance between the two, thereby mining the potential number of offsets and offset step size between the electricity consumption of industries, and further clarifying the changing patterns of industry relationships in different regions and times within the industry chain; it also compares the occurrence time of special factors such as logistics disruptions and orderly electricity use to explore the connection between special factors and the changing patterns of relationships.

[0079] The following calculations use DTW distance to reconstruct the transmission relationships of monthly electricity consumption data for various industrial sectors in Zhejiang Province from 2019 to 2021. The specific results are as follows:

[0080] (1) The pharmaceutical manufacturing industry and the rubber and plastic products industry under the influence of external factors

[0081] There are potential offset times and offset steps between the pharmaceutical manufacturing industry in Xinchang and the rubber and plastic products industries in Changxing, Lanxi, Fenghua, Jiashan, Yongkang, and Dongyang. See Table 1 for details.

[0082] Table 1. Electricity transfer relationship between the pharmaceutical manufacturing industry in Xinchang and downstream rubber and plastic products industries in various regions (DTW results)

[0083]

[0084] Taking the link between the pharmaceutical manufacturing industry in Xinchang and the rubber and plastic products industry in Fenghua as an example, the analysis process is explained: each offset step size [1,1,2] represents a total of 3 offsets between the two curves (the first two lags are 1 month, and the third lag is 2 months). Figure 8 As shown, the electricity consumption curves for Xinchang Pharmaceutical and Fenghua Rubber exhibit a clear sequential relationship in terms of electricity consumption during March-April 2020, July-August 2020, and August-October 2020. The Xinchang pharmaceutical industry experienced an increase in electricity consumption in March 2020 due to external factors. This signal, transmitted over a month, reached the Fenghua rubber industry (pharmaceutical packaging requires a large amount of rubber products), causing its electricity consumption to rise in April. Under the continued influence of this upward signal, the peak monthly electricity consumption for the Fenghua rubber industry was delayed until August 2020 (compared to July 2019). Therefore, based on the early warning and prevention measures outlined in this example, the upward signal in electricity consumption from the Xinchang pharmaceutical industry at the beginning of 2020 is expected to extend to the rubber industries in Changxing, Lanxi, Fenghua, Jiashan, Yongkang, and Dongyang.

[0085] (2) The impact of logistics disruptions on the general equipment manufacturing industry and other equipment manufacturing industries

[0086] There are potential offset times and offset steps between the general equipment manufacturing industry in Haining and other equipment manufacturing industries in Xiaoshan District of Hangzhou City, Huzhou City, and Changxing County, as shown in Table 2.

[0087] Table 2. Electricity transfer relationship between Haining's general equipment manufacturing industry and other downstream equipment manufacturing industries. (DTW results)

[0088]

[0089]

[0090] Taking the link between the general equipment manufacturing industry in Haining and other equipment manufacturing industries in Xiaoshan, Hangzhou as an example, this illustrates the process applied to logistics disruption analysis: The electricity consumption curves of Haining general equipment and Xiaoshan other equipment show a clear sequential relationship between April-June 2021 and October-November 2020. The main reason is the ship grounding incident in the Suez Canal in early March 2021, causing global logistics disruptions. Affected by this, other manufacturing industries in Xiaoshan experienced a large backlog of products, leading to a continuous decline in electricity consumption in April 2021. This signal was transmitted to the downstream Haining general equipment manufacturing industry, where electricity consumption began to decline in May. Figure 9 As shown.

[0091] (3) Weekly electricity transfer relationship in the automotive industry

[0092] The weekly patterns of inter-industry linkages were analyzed, taking the automotive-related industry as an example, as shown in Table 3.

[0093] Table 3. Electricity Transfer Relationship between Cixi Automobile Manufacturing and Downstream Automobile Power Sales (DTW Results)

[0094]

[0095] Therefore, a monthly electricity consumption correction coefficient can be constructed based on the DTW distance algorithm to correct the monthly electricity consumption forecast of downstream industries; and the month in which the maximum monthly electricity consumption of downstream industries occurs can be estimated based on the number of offsets and the offset step size.

[0096] Step S5 in this embodiment further includes:

[0097] The clustering results of key industrial chains, regional industry clustering results, electricity transfer relationships between key industrial chains, and electricity transfer relationships between different regions and industrial sectors are merged and summarized according to different industrial chains, different regions, and different industries to form an industry electricity transfer relationship lookup table.

[0098] The industry electricity consumption relationship query table includes a tag query table and a correlation query table. The tag query table includes regional tags, industry tags, cluster tags, and community tags. The correlation query table includes DTW distance, correlation coefficient, and lead-lag value. The lead-lag value is the offset step size obtained by the DTW algorithm.

[0099] In summary, the above-mentioned industrial early warning and prevention method in this embodiment can be further applied in areas such as load forecasting, industrial planning, and monitoring and early warning.

[0100] Furthermore, based on the above-described industrial early warning and prevention method of this embodiment, such as Figure 10 As shown, this embodiment also provides an industrial early warning and prevention system based on industrial power generation linkage, including a data acquisition module, a statistics module, a clustering module, a community relationship analysis module, and an early warning and prevention module.

[0101] The data acquisition module in this embodiment is used to collect electricity consumption information from various regions and industries.

[0102] Specifically, the electricity consumption information for each region and industrial sector includes the region, industry name, industry code, and corresponding daily electricity consumption, and is stored in the database for subsequent data statistics and processing.

[0103] The statistics module in this embodiment is used to collect electricity consumption information from various regions and industries to obtain monthly and weekly electricity consumption data for each region and industry.

[0104] Specifically, the study focuses on industrial electricity consumption in Zhejiang Province. The research sample consists of two parts: monthly electricity consumption data and weekly electricity consumption data.

[0105] In addition, after obtaining the monthly and weekly electricity consumption data for various regions and industries, an additional data preprocessing module was used to clean and normalize the data. Specifically, for missing or zero data, data cleaning was performed to fill in the missing data, using the `interpolate` function of the Pandas module.

[0106] In this embodiment, after data cleaning is completed, normalization is performed, specifically using Z-Score normalization.

[0107] The clustering module in this embodiment is used to perform DBSCAN clustering on monthly electricity consumption data of various regions and industries to obtain key industrial chain clustering results; wherein, the key industrial chain clustering results include several key industrial chains and the industries they contain.

[0108] Specifically, the DBSCAN algorithm is a density-based clustering algorithm that defines a cluster as the largest set of density-connected points. Two important parameters in DBSCAN are eps and min_samples. A larger eps result in fewer clusters; min_samples needs to be selected appropriately based on the actual characteristics of the data and the target result.

[0109] The DBSCAN algorithm is used to analyze the consistency of monthly electricity consumption data curves and to identify key industrial chains, serving as the first layer (industry layer) for subsequent analysis of industrial relationships. For example... Figure 3 As shown, by adjusting the parameters eps=2.5 and min_samples=3, the final clustering result was obtained, with 34 industrial sectors clustered into 7 major industrial chains.

[0110] In addition, the clustering module in this embodiment is also used to perform two-step clustering on the weekly electricity consumption data of each region and industry to obtain the regional and industry clustering results; wherein, the first step of clustering adopts DBSCAN clustering based on the DTW algorithm, and the second step of clustering adopts K-means clustering based on the DTW algorithm.

[0111] Specifically, this embodiment employs a two-step clustering method to analyze the consistency of weekly electricity consumption data curves. The first step, clustering, aims to identify industries with similar electricity consumption changes and eliminate those with obvious inconsistencies. DBSCAN clustering based on the DTW algorithm is used. The custom distance function DTW algorithm calculates the minimum distance between weekly electricity consumption curves when different time-series shifts occur, meeting the consistency requirements when curves shift. It can also calculate the step size, number of shifts, and maximum step size for each shift. Specific calculations can be found in existing DTW algorithms and will not be elaborated here. The second step, clustering, aims to split the large clusters that performed poorly in the first step. K-means clustering based on the DTW algorithm is used, comprehensively considering the silhouette coefficient, to split the large clusters as much as possible.

[0112] The two-step clustering result in this embodiment is the second-level structure (regional layer) for subsequent analysis of industrial relationships.

[0113] When the DBSCAN algorithm parameters eps and min_sample are selected as 0.8 and 4 respectively, the first step of clustering results is optimal. Figure 4 As shown; when the maximum initial number of centers m = 18 in the K-means algorithm, the splitting effect of the second-step clustering result is the best, as shown. Figure 5 As shown.

[0114] The community relationship analysis module in this embodiment uses the Louvain algorithm to perform community relationship analysis on the clustering results of key industrial chains and regional industry clustering results, respectively, to obtain the electricity transfer relationships between key industrial chains and between industrial sectors in different regions. In addition, the weight calculation module in this embodiment uses each cluster of key industrial chain clustering results or regional industry clustering results as a node, and uses Pearson correlation analysis to obtain the correlation coefficient and differential correlation coefficient of the rate of change of weekly electricity curves between each node. Then, PCA is used to reduce the dimensionality of the correlation coefficient and differential correlation coefficient, and the resulting one-dimensional coefficient is used as the weight of the edges in the community relationship analysis.

[0115] Specifically, the Louvain algorithm is used for social network analysis. Its core principle is to determine the importance of a node within a "community" based on the concept of centrality, thereby assessing the overall concentration and distribution of the network. Modularity is a parameter that measures the performance of the Louvain model, used to determine whether the algorithm's partitioning result has reached the optimal solution. Modularity ranges from -0.5 to 1. The optimal solution of the Louvain algorithm is achieved when node similarity is highest within a community and lowest outside the community. The Louvain community detection method focuses on finding connections between targets in the network, accurately uncovering hierarchical communities.

[0116] This embodiment uses the Louvain algorithm to analyze the community correlation of weekly electricity consumption data curves, including three steps:

[0117] 1) Node selection and edge weight calculation:

[0118] Each industry cluster (classification) in the above clustering results is taken as node i; the correlation coefficient and differential correlation coefficient of the weekly electricity consumption curves between each node are calculated using their rate of change. The two correlation coefficients are reduced to one dimension by principal component analysis (PCA), and this data is used as the weight k of the edge.

[0119] 2) Modularity Q-optimization and network cohesion stage:

[0120] Iterate through all nodes i in each sub-industry and calculate the change in modularity ΔQ:

[0121]

[0122] Σtot is the sum of the weights on the links connected to nodes in C;

[0123] k i It is the sum of the weights of the links associated with node i;

[0124] k i,in It is the sum of the links from node i to nodes in C;

[0125] m is the sum of the weights of all links in the network;

[0126] The node corresponding to the increase in ΔQ is taken as the new community label, and this process is iterated and updated continuously until the overall network no longer changes its community label and the modularity ΔQ no longer increases. The above is a brief explanation of the Louvain algorithm; for the specific process, please refer to existing technologies, which will not be elaborated here.

[0127] The results of the community relations analysis are as follows:

[0128] like Figure 6 As shown, community relationship analysis at the industry level was conducted using monthly electricity consumption data, with a maximum module value of 0.59, representing the community relationship results for key industrial chains. (Example...) Figure 7 As shown, a regional-level community relationship analysis was conducted using weekly electricity consumption data, with a maximum modularity of 0.72. The following results can be obtained from this:

[0129] Result 1: The greater the distance between points, the smaller the correlation between industries. Figure 6 The seven key industrial chains within the framework exhibit significant intra-chain clustering. For example, the agricultural and food production industrial chain (… Figure 6 The four industries in the lower right corner are relatively independent in the industrial sector, far removed from most industrial sectors, and have weak industrial linkages.

[0130] Result 2: The industry with the most connections to other industries indicates a higher industry concentration. The centrality of the industry can be calculated based on the number of connections and the weight k. For example, in the community of the paper-related industry chain ( Figure 7 (Nos. 10, 21), Paper and Paper Products Industry ( Figure 7 The industry with the highest centrality is number 15.

[0131] The early warning and control module in this embodiment is used to conduct industrial early warning and control by utilizing the power transfer relationships between key industrial chains and between different regions and industrial sectors. The power transfer relationships include one or more of the following: DTW distance, correlation coefficient, differential correlation coefficient, and the offset step size, number of offsets, and maximum offset step size obtained from the DTW algorithm.

[0132] The DTW (Dynamic Time Warping) distance algorithm in this embodiment is a method for measuring the similarity between two time series. It calculates the similarity between two time series by extending, shortening, and shifting the time series.

[0133] Based on the constructed community relationship results, this embodiment calculates the correlation between the electricity consumption of key industries in each industry chain corresponding to each community; it uses the DTW algorithm to calculate the sequence distance, slices the data, and continuously adjusts the offset to minimize the distance between the two, thereby mining the potential number of offsets and offset step size between the electricity consumption of industries, and further clarifying the changing patterns of industry relationships in different regions and times within the industry chain; it also compares the occurrence time of special factors such as logistics disruptions and orderly electricity use to explore the connection between special factors and the changing patterns of relationships.

[0134] The above case studies demonstrate that the transmission of electricity consumption between industries exhibits a pattern of "temporal consistency and spatial transmissibility." Further in-depth analysis of the leading factors in industry electricity consumption reveals specific indicators (such as oil imports and road freight volume) that influence future electricity consumption trends across the entire industrial chain. Based on these indicators, leading indices can be constructed, allowing for timely issuance of early warning information when similar special events occur. This guides management forces to intervene in the industrial chain in advance, reducing material and energy waste caused by information asymmetry.

[0135] This invention combines factors such as industry restructuring, output changes, or industrial relocation in a region to reflect the impact of these changes on electricity consumption in a timely manner in industry electricity consumption trend analysis. As a supplement to traditional electricity consumption forecasting methods, it can reduce forecasting errors caused by changes in the industrial chain.

[0136] In addition, this invention can be used as a "plug-and-play" application to mine the impact of special events on the "linkage" patterns of the industry using electricity data, analyze the leading, consistent, and lagging characteristics of electricity consumption in the industry in terms of time, as well as the transmissibility in geographical space; construct a power industry chain map from the perspective of electricity, provide decision-making basis, and expand the external application value of energy data.

[0137] The above description is merely a detailed explanation of preferred embodiments and principles of the present invention. For those skilled in the art, there may be changes in specific implementation methods based on the ideas provided by the present invention, and these changes should also be considered within the scope of protection of the present invention.

Claims

1. An industrial early warning and prevention method based on industrial power generation linkage, characterized in that, Includes the following steps: S1. Collect electricity consumption information from various regions and industries; S2. Collect monthly and weekly electricity consumption data for each region and industry by statistically analyzing electricity consumption information for each region and industry. S3. Perform DBSCAN clustering on the monthly electricity consumption data of various regions and industries to obtain the clustering results of key industrial chains; among them, the clustering results of key industrial chains include several key industrial chains and the industries they contain. The weekly electricity consumption data of each region and industry were clustered in two steps to obtain the regional and industry clustering results. The first step of clustering adopted DBSCAN clustering based on the DTW algorithm, and the second step adopted K-means clustering based on the DTW algorithm. S4. Using the Louvain algorithm, community relationship analysis was performed on the clustering results of key industrial chains and regional industries to obtain the power transfer relationship between key industrial chains and between industrial sectors in different regions. S5. Utilize the power transmission relationships between key industrial chains and between different regions and industrial sectors to conduct early warning and prevention of industrial development.

2. The industrial early warning and prevention method according to claim 1, characterized in that, In step S1, the electricity consumption information of each region and industrial sector includes the region, industry name, industry code and its corresponding daily electricity consumption, and is stored in the database.

3. The industrial early warning and prevention method according to claim 2, characterized in that, Step S2 further includes: The monthly and weekly electricity consumption data of various regions and industrial sectors obtained from the statistics were cleaned and normalized.

4. The industrial early warning and prevention method according to claim 3, characterized in that, The data cleaning process involves filling in missing data, using the interpolate function of the Pandas module for interpolation.

5. The industrial early warning and prevention method according to claim 4, characterized in that, The normalization process uses Z-Score normalization.

6. The industrial early warning and prevention method according to claim 5, characterized in that, Step S4 further includes: Each cluster based on the clustering results of key industrial chains or regional industries is used as a node. Pearson correlation analysis is used to obtain the correlation coefficient and differential correlation coefficient of the rate of change of weekly electricity consumption curves between nodes. The correlation coefficient and differential correlation coefficient are then reduced in dimensionality using PCA, and the resulting one-dimensional coefficient is used as the weight of the edge in the community relationship analysis.

7. The industrial early warning and prevention method according to claim 6, characterized in that, The power transfer relationship includes one or more of the following: DTW distance, correlation coefficient, differential correlation coefficient, and each offset step size, number of offsets, and maximum offset step size obtained by the DTW algorithm.

8. The industrial early warning and prevention method according to any one of claims 1-7, characterized in that, Step S5 further includes: The clustering results of key industrial chains, regional industry clustering results, electricity transfer relationships between key industrial chains, and electricity transfer relationships between different regions and industrial sectors are merged and summarized according to different industrial chains, different regions, and different industries to form an industry electricity transfer relationship lookup table.

9. The industrial early warning and prevention method according to claim 8, characterized in that, The industry electricity consumption relationship query table includes a tag query table and a correlation query table. The tag query table includes regional tags, industry tags, cluster tags, and community tags. The correlation query table includes DTW distance, correlation coefficient, and lead-lag value. The lead-lag value is the offset step size obtained by the DTW algorithm.

10. An industrial early warning and prevention system based on industrial power generation linkage, employing the industrial early warning and prevention method as described in any one of claims 1-8, characterized in that, The industrial early warning and prevention system includes: The data acquisition module is used to collect electricity consumption information from various regions and industries. The statistics module is used to collect electricity consumption information from various regions and industries to obtain monthly and weekly electricity consumption data for each region and industry. The clustering module is used to perform DBSCAN clustering on monthly electricity consumption data of various regions and industries to obtain clustering results for key industrial chains. The clustering results for key industrial chains include several key industrial chains and the industries they contain. It is also used to perform two-step clustering on weekly electricity consumption data of various regions and industries to obtain regional and industry clustering results. The first step of clustering uses DBSCAN clustering based on the DTW algorithm, and the second step of clustering uses K-means clustering based on the DTW algorithm. The community relationship analysis module is used to perform community relationship analysis on the clustering results of key industrial chains and regional industries using the Louvain algorithm, respectively, to obtain the power transfer relationship between key industrial chains and the power transfer relationship between industrial sectors in different regions; The early warning and prevention module is used to conduct industrial early warning and prevention by utilizing the power transmission relationships between key industrial chains and between different regions and industrial sectors.

Images