An industry chain upstream and downstream identification method and device based on industry power consumption data

Abstract

The application relates to an industry chain upstream and downstream identification method based on industry power consumption data, which comprises the following steps: obtaining power consumption time series of each industry to obtain a sequence set; calculating the time difference correlation coefficient between each two power consumption time series in the sequence set; taking the power consumption time series as nodes and the weight of edges as the time difference correlation coefficient of the two nodes connected by the edges to construct a directed weighted graph; inputting the directed weighted graph to score the power consumption time series through a node importance evaluation algorithm; and determining the level of the corresponding industry according to the score value of the power consumption time series.

Description

Technical Field

[0001] This invention relates to a method for identifying upstream and downstream supply chains based on industry electricity data, belonging to the field of supply chain identification. Background Technology

[0002] The upstream and downstream of an industry chain refers to the relationship between various links or participants within an industry. Based on product flow, it can be divided into upstream and downstream. The close connection between supply and demand between upstream and downstream enterprises promotes technology sharing and innovation. Today, industry chains are becoming increasingly diversified and complex, with growing connections and dependencies between different industries, creating a complex web that requires precise methods for identifying industry chains.

[0003] The current methods for matching and identifying upstream and downstream supply chains include the following:

[0004] One approach involves analyzing industry data across various stages and establishing knowledge graphs to query the upstream and downstream relationships within a company's supply chain. However, knowledge graph analysis of supply chain relationships has limitations in semantic understanding of unstructured or semi-structured data, preventing knowledge graphs from fully comprehending the complex relationships and concepts within certain industry sectors.

[0005] Secondly, by analyzing the flow data of industry operations and order data from both the supply and demand sides, the correlation between industries can be explored to match upstream and downstream sectors, as in the paper "A Smart Matching Method and System for Upstream and Downstream Industries Based on Big Data". However, analyzing only the flow data and order data of industries to analyze their correlation is not comprehensive enough and cannot fundamentally uncover the development characteristics of industries.

[0006] Therefore, a more accurate method for matching and identifying upstream and downstream players in the industrial chain is needed.

[0007] CN114429363A, "A Method for Analyzing the Synergistic Effect of Upstream and Downstream Industrial Chains Based on Electricity Data," discloses a method for analyzing the degree of synergy between upstream and downstream industries by establishing coupling degree analysis models and composite coordination degree models based on electricity consumption evaluation values ​​or electricity consumption ratio evaluation values ​​of upstream and downstream industries. However, the definitions of electricity consumption evaluation values ​​and electricity consumption ratio evaluation values ​​are not clearly defined. Such models are based on a series of assumptions and preconditions, such as linear relationships or stability assumptions. However, in reality, the relationship between upstream and downstream industries may have nonlinear and dynamic characteristics, and the limitations of these assumptions may lead to the limitations of the models. Summary of the Invention

[0008] To overcome the problems existing in the prior art, this invention designs a method and device for identifying upstream and downstream of the industrial chain based on industry electricity data. Based on the electricity consumption time series, it uses time difference correlation coefficient to analyze the relative change trend between industries, uses a directed weighted graph to represent the relationship between industries, and uses the PageRank algorithm to quantify the importance of each industry in the upstream and downstream of the industrial chain, which can more accurately and dynamically reveal the relationship between upstream and downstream of the industrial chain.

[0009] To achieve the above objectives, the present invention adopts the following technical solution:

[0010] Technical Solution 1

[0011] A method for identifying upstream and downstream supply chains based on industry electricity data includes the following steps:

[0012] Obtain the time series of electricity consumption for each industry to obtain a set of sequences;

[0013] For the set of sequences, calculate the time difference correlation coefficient between every two electricity consumption time series;

[0014] Using electricity consumption time series as nodes, a directed weighted graph is constructed, where the weight of an edge is the time difference correlation coefficient between the two nodes it connects;

[0015] Using a directed weighted graph as input, the electricity consumption time series is scored through a node importance evaluation algorithm;

[0016] The corresponding industry level is determined based on the score value of the electricity consumption time series.

[0017] Furthermore, it also includes: preprocessing each electricity consumption time series, wherein the preprocessing is a seasonal adjustment.

[0018] Furthermore, the calculation of the time difference correlation coefficient between pairs of electricity consumption time series specifically involves:

[0019] Calculate the time difference correlation coefficients of the two electricity consumption time series at different periods to obtain H time difference correlation coefficients; take the weighted average of the H time difference correlation coefficients and take the weighted average as the time difference correlation coefficient of the two electricity consumption time series.

[0020] Furthermore, it also includes: if the time difference correlation coefficient is less than a preset threshold, then the weight value of the corresponding edge is set to zero.

[0021] Furthermore, the weight of an edge is set to the absolute value of the time difference correlation coefficient between the two nodes it connects.

[0022] Furthermore, the electricity consumption time series is scored using a node importance assessment algorithm, specifically as follows:

[0023] The directed weighted graph is represented as a matrix W;

[0024] Normalize matrix W column by column to obtain matrix H as the transition matrix;

[0025] Construct a scoring vector, where each component of the scoring vector corresponds to a time series of electricity consumption;

[0026] Based on matrix H, the time series score of each electricity consumption is iteratively calculated until the convergence condition is met, expressed by the formula:

[0027] score (k+1) =(1-α)e N,1 +αH·score (k)

[0028] In the formula, α>0 is the damping coefficient; e N,1 H is an N-order column vector with all components equal to 1; H is the transition matrix; score (k) Let k be the scoring vector, and k be the number of iterations.

[0029] Furthermore, based on the score, the corresponding industry level is determined, specifically as follows:

[0030] S1. Traverse the sequences of each undetermined level and classify at least one sequence whose score value meets the condition into the i-th level; reconstruct the directed weighted graph and calculate its score value based on the remaining undetermined levels of the sequences.

[0031] S2. Repeat step S1 until the hierarchy of all sequences is obtained.

[0032] Furthermore, the conditions are expressed by the following formula:

[0033]

[0034] In the formula, The score represents the rating of sequence u; U represents the set of sequences with undetermined hierarchies; K represents the set of all sequences; H u,i H represents the normalized value of the time difference correlation coefficient between sequence u and sequence i; u,k This represents the normalized value of the time difference correlation coefficient between sequence u and sequence k. This is the score value for the current traversed sequence.

[0035] Technical Solution Two

[0036] A supply chain upstream and downstream identification device based on industry electricity data includes a processor and a memory for storing executable instructions of the processor; the processor is used to read the executable instructions from the memory and execute the instructions to achieve the following steps:

[0037] Obtain the time series of electricity consumption for each industry to obtain a set of sequences;

[0038] For the set of sequences, calculate the time difference correlation coefficient between every two electricity consumption time series;

[0039] Using electricity consumption time series as nodes, a directed weighted graph is constructed, where the weight of an edge is the time difference correlation coefficient between the two nodes it connects;

[0040] Using a directed weighted graph as input, the electricity consumption time series is scored through a node importance evaluation algorithm;

[0041] The corresponding industry level is determined based on the score value of the electricity consumption time series.

[0042] Furthermore, it also includes: preprocessing each electricity consumption time series, wherein the preprocessing is a seasonal adjustment.

[0043] Furthermore, the calculation of the time difference correlation coefficient between pairs of electricity consumption time series specifically involves:

[0044] Calculate the time difference correlation coefficients of the two electricity consumption time series at different periods to obtain H time difference correlation coefficients; take the weighted average of the H time difference correlation coefficients and take the weighted average as the time difference correlation coefficient of the two electricity consumption time series.

[0045] Furthermore, it also includes: if the time difference correlation coefficient is less than a preset threshold, then the weight value of the corresponding edge is set to zero.

[0046] Furthermore, the weight of an edge is set to the absolute value of the time difference correlation coefficient between the two nodes it connects.

[0047] Furthermore, the electricity consumption time series is scored using a node importance assessment algorithm, specifically as follows:

[0048] The directed weighted graph is represented as a matrix W;

[0049] Normalize matrix W column by column to obtain matrix H as the transition matrix;

[0050] Construct a scoring vector, where each component of the scoring vector corresponds to a time series of electricity consumption;

[0051] Based on matrix H, the time series score of each electricity consumption is iteratively calculated until the convergence condition is met, expressed by the formula:

[0052] score (k+1) =(1-α)e N,1 +αH·score (k)

[0053] In the formula, α>0 is the damping coefficient; e N,1H is an N-order column vector with all components equal to 1; H is the transition matrix; score (k) Let k be the scoring vector, and k be the number of iterations.

[0054] Furthermore, based on the score, the corresponding industry level is determined, specifically as follows:

[0055] S1. Traverse the sequences of each undetermined level and classify at least one sequence whose score value meets the condition into the i-th level; reconstruct the directed weighted graph and calculate its score value based on the remaining undetermined levels of the sequences.

[0056] S2. Repeat step S1 until the hierarchy of all sequences is obtained.

[0057] Furthermore, the conditions are expressed by the following formula:

[0058]

[0059] In the formula, The score represents the rating of sequence u; U represents the set of sequences with undetermined hierarchies; K represents the set of all sequences; H u,i H represents the normalized value of the time difference correlation coefficient between sequence u and sequence i; u,k This represents the normalized value of the time difference correlation coefficient between sequence u and sequence k. This is the score value for the current traversed sequence.

[0060] Compared with the prior art, the present invention has the following features and beneficial effects:

[0061] 1. This invention is based on electricity consumption time series, uses time difference correlation coefficient to analyze the relative changing trends between industries, uses directed weighted graphs to represent the correlation between industries, and uses the PageRank algorithm to quantify the importance of each industry in the upstream and downstream of the industrial chain, which can more accurately and dynamically reveal the upstream and downstream relationships of the industrial chain.

[0062] 2. This invention uses the weighted average of the time difference correlation coefficients under different periods as the final time difference correlation coefficient, rather than the maximum time difference correlation coefficient, thus avoiding abnormal correlation coefficients under certain periods.

[0063] 3. This invention takes into account that the precedence and lag relationships of different industries may vary under different circumstances. It constructs a directed weighted graph to represent industry relationships, and the nodes have bidirectional logical relationships. That is, it allows bidirectional causal relationships between electricity consumption time symbol sequences, which is more in line with the actual industry development situation and the upstream and downstream identification results are more accurate.

[0064] 4. This invention sets the weight as the absolute value of the time difference correlation coefficient to represent the correlation between two time series, eliminating the interference caused by the leading or lagging relationship between time series, and paying more attention to the strength of industry correlation.

[0065] 5. This invention sets weight values ​​below a threshold to zero, making the matrix sparse. Correlation exists only between time series with high correlation, reducing computational and storage complexity and removing noise to eliminate unreliable correlations caused by noise, thereby improving the accuracy of sorting.

[0066] 6. The formula of this invention sets the ratio of the degree of leadership of sequence u over sequence i to the degree of leadership of sequence u over other sequences as the weight, and performs a weighted summation of the score values ​​of sequences with undetermined levels to select the sequence as the i-th level, which can better determine the priority of the industry corresponding to the sequence in the industrial chain. Attached Figure Description

[0067] Figure 1 This is a flowchart of the present invention;

[0068] Figure 2-5 This is a schematic diagram of the upstream and downstream identification results. Detailed Implementation

[0069] The present invention will now be described in more detail with reference to the embodiments.

[0070] In one implementation, a method for matching and identifying upstream and downstream players in the industrial chain based on industry electricity data sorting includes the following steps:

[0071] Obtain the time series of electricity consumption for each industry to get the sequence set S.

[0072] For a set of sequences S, calculate the time difference correlation coefficient between every two electricity consumption time series. For example, for sequences x and y, there are time difference correlation coefficients for sequence x relative to sequence y and time difference correlation coefficients for sequence y relative to sequence x.

[0073] Using electricity consumption time series as nodes, and the weight of each edge as the time difference correlation coefficient between the two nodes it connects, a directed weighted graph is constructed.

[0074] Using a directed weighted graph as input, the electricity consumption time series is scored using a node importance evaluation algorithm; based on the score of the electricity consumption time series, the corresponding industry level is determined. Node importance evaluation algorithms include, but are not limited to, PageRank, Google Out-degree algorithm, and HITS algorithm.

[0075] The advancement of this embodiment lies in the following: The inventors of this invention considered the high correlation between electricity consumption and national economic growth. Taking 1949-2006 as an example, the correlation coefficient between electricity consumption and GDP in the United States reached 0.9856, while in my country it reached 0.993, reflecting the core position of industrial electricity consumption in industrial development. Therefore, this embodiment can reveal the essential industrial structure and upstream and downstream relationships in the industrial chain by analyzing industrial electricity consumption data.

[0076] In one implementation, the electricity consumption time series is seasonally adjusted before calculating the time difference correlation coefficient to eliminate trend and seasonal components in the time series data, thereby highlighting the true attributes of the industry itself.

[0077] In one implementation, the time difference correlation coefficient is calculated:

[0078] Calculate the time difference correlation coefficients of the two electricity consumption time series at different periods to obtain H time difference correlation coefficients; take the weighted average of the H time difference correlation coefficients and take the weighted average as the time difference correlation coefficient of the two electricity consumption time series.

[0079] In this embodiment, the weight is set to 1. Specifically, it can be expressed as a formula:

[0080]

[0081]

[0082]

[0083] In the formula, h is a natural number, representing the lead or lag period; x and y represent two sequences, with sequence y serving as the baseline value; r x,y (h) represents the time difference correlation coefficient between sequence x and sequence y, where x leads or lags by h periods; N is the number of data points in the sequence; x n This represents the nth data point in sequence x.

[0084] In existing technologies, the time difference correlation coefficients of sequence x relative to y under different h values ​​are calculated, and the one with the largest absolute value is taken as the final time difference correlation coefficient of sequences x and y, expressed by the formula:

[0085] r x,y (h * ) = max h＞0 |r x,y (h)|,h * >0

[0086] If |r x,y |>|r y,x |, then x is considered to be ahead of y, h * r is the number of periods that x leads y.x,y (h * ) is the time difference correlation coefficient of x leading y.

[0087] The improvement of this embodiment lies in the use of a weighted average of the time difference correlation coefficients under different periods, rather than the maximum time difference correlation coefficient, which avoids abnormal correlation coefficients under certain leading periods.

[0088] In one implementation, an industry is selected as the identification target. For that industry, its upstream and downstream industries are identified, and electricity consumption time series whose time difference correlation coefficient with the industry's electricity consumption time series is lower than a threshold are deleted. For example, if the real estate industry is the identification target, and the time difference correlation coefficient between the financial industry's electricity consumption time series and the real estate industry's electricity consumption time series is lower than a preset threshold, then the financial industry's electricity consumption time series is deleted.

[0089] In one implementation, a directed weighted graph G = (S, W) is constructed using electricity consumption time series as nodes and the weight of an edge as the time difference correlation coefficient between the two nodes it connects. In a computing device, the directed weighted graph is stored and processed as a matrix W. Matrix element W n,m and W m,n These represent the time difference correlation coefficients of sequence n relative to m and sequence m relative to n, respectively.

[0090] The advancement of this embodiment lies in the fact that, considering that the lead-lag relationship may differ among industries under different circumstances, a directed weighted graph is constructed to represent industry relationships. There are bidirectional logical relationships between nodes, that is, bidirectional causal relationships are allowed between the electricity consumption time symbol sequences, which is more in line with the actual industry development situation and the upstream and downstream identification results are more accurate.

[0091] In one implementation, the edge weights in the directed weighted graph are set to the absolute values ​​of the time difference correlation coefficients, expressed by the formula: A threshold l>0 is set. If the weight value of an edge is less than the threshold l, the weight value of that edge is set to zero.

[0092] The advantage of this embodiment is that:

[0093] The weights are set to the absolute value of the time difference correlation coefficient to represent the correlation between two time series, eliminating the interference caused by the leading or lagging relationship between time series, and focusing more on the strength of industry correlation.

[0094] Setting weights below a threshold to zero makes the matrix sparse, ensuring connections only exist between time series with high correlation, thus reducing computational and storage complexity. Simultaneously, noise is removed, eliminating potentially low and unreliable correlations caused by noise or randomness, thereby improving ranking accuracy.

[0095] In one implementation, the PageRank algorithm is used to iteratively score each electricity consumption time series and determine the corresponding industry tier, specifically:

[0096] Normalize matrix W column by column to obtain matrix H as the transition matrix;

[0097] Construct a score vector (k) Each component of the scoring vector corresponds to a time series of electricity consumption, and the scoring vector is initialized with a value of "score". (0) ;

[0098] Set the value of the damping coefficient α;

[0099] Let matrix H, score (0) α is input into the PageRank algorithm for iterative calculation, and the PageRank algorithm outputs a score. (+∞) Expressed as a formula:

[0100] score (k+1) =(1-α)e N,1 +αH·score (k)

[0101] In the formula, α>0 is a given damping coefficient; e N,1 H is an N-order column vector with all components equal to 1; H is the transition matrix; score (k) Let k be the scoring vector, and k be the number of iterations.

[0102] When α is sufficiently small, the scoring vector will inevitably converge to a value equal to the initial score. (0) irrelevant vector value score (+∞) score (+∞) =(1-α)(I-αH) -1 e N,1 All that is needed is to ensure that the spectral radius of αH is less than 1. (+∞) Each component is a score for each electricity consumption time series.

[0103] Sort the sequence set S from highest to lowest according to the score.

[0104] Iterate through each sequence and check if each sequence satisfies the condition, which is expressed by the formula:

[0105]

[0106] In the formula, Let U represent the score of sequence u; U is the set of sequences with undetermined levels; K represents the set of all sequences; H represents the score of sequence u. u,iLet H be the matrix element in the u-th row and i-th column of the transition matrix, representing the normalized value of the time difference correlation coefficient between sequence u and sequence i; similarly, H... u,k This represents the normalized value of the time difference correlation coefficient between sequence u and sequence k. This is the score value for the current traversed sequence.

[0107] The sequence that meets the condition is considered the first subset of the entire sequence set to receive the "information", it is marked as the first level, and it is removed from the sequence set S;

[0108] The remaining sequences are treated as a new sequence set. The directed weighted graph G is reconstructed and scored to obtain the second set of sequences that do not meet the conditions. This set is then used as the second-level information receiving subset of the entire sequence set. This process is repeated until the sequence set S is divided into multiple levels.

[0109] For example, sorting the sequence set S from high to low. Check whether each sequence meets the requirements; for example, for the sequence

[0110]

[0111] Hypothetical sequence The requirements are met; the scores of the remaining sequences are recalculated and a directed weighted graph is constructed; the sequence set S is sorted from highest to lowest according to the newly calculated scores. Check whether each sequence meets the requirements; for example, for the sequence

[0112]

[0113] From the above description of the embodiments, those skilled in the art will clearly understand that the embodiments described herein can be implemented in hardware, software, firmware, middleware, code, or any suitable combination thereof. For hardware implementation, the processor can be implemented in one or more of the following units: Application-Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field-Programmable Gate Array (FPGA), processor, controller, microcontroller, microprocessor, other electronic units designed to implement the functions described herein, or combinations thereof. For software implementation, some or all of the processes of the embodiments can be implemented by a computer program instructing the associated hardware. During implementation, the program can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of a computer program from one place to another. Storage media can be any available medium accessible to a computer. Computer-readable media can include, but is not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code having the form of instructions or data structures and accessible to a computer.

[0114] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should be able to analyze that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. A method for identifying upstream and downstream supply chains based on industry electricity data, characterized in that, Includes the following steps: Obtain the time series of electricity consumption for each industry to obtain a set of sequences; For the set of sequences, calculate the time difference correlation coefficient between every two electricity consumption time series; Using electricity consumption time series as nodes, a directed weighted graph is constructed, where the weight of an edge is the time difference correlation coefficient between the two nodes it connects; Using a directed weighted graph as input, the electricity consumption time series is scored through a node importance evaluation algorithm, specifically: The directed weighted graph is represented as ; right Normalize by column to obtain the matrix As the transition matrix; Construct a scoring vector, where each component of the scoring vector corresponds to a time series of electricity consumption; According to the matrix The time series score of each electricity consumption is iteratively calculated until the convergence condition is met, which can be expressed by the following formula: In the formula, the damping coefficient ; for A column vector of order whose components are all 1; Here is the transition matrix; Let k be the scoring vector, and k be the number of iterations. Based on the scoring values ​​of electricity consumption time series, the corresponding industry level is determined as follows: S1. Traverse the sequences of each undetermined level and classify at least one sequence whose score value meets the condition into the i-th level; reconstruct the directed weighted graph and calculate its score value based on the remaining undetermined levels of the sequences. The condition is expressed by the following formula: In the formula, This represents the score value of sequence u; U represents the set of sequences with undetermined hierarchies. Represents the set of all sequences; This represents the normalized value of the time difference correlation coefficient between sequence u and sequence i; This represents the normalized value of the time difference correlation coefficient between sequence u and sequence k. This is the score value for the current traversed sequence; S2. Repeat step S1 until the hierarchy of all sequences is obtained.

2. The method for identifying upstream and downstream supply chains based on industry electricity data according to claim 1, characterized in that, Also includes: Each electricity consumption time series is preprocessed, and the preprocessing is a seasonal adjustment.

3. The method for identifying upstream and downstream supply chains based on industry electricity data according to claim 1, characterized in that, The calculation of the time difference correlation coefficient between the two electricity consumption time series is specifically as follows: Calculate the time difference correlation coefficient between each pair of electricity consumption time series at different periods to obtain H time difference correlation coefficients; take the weighted average of the H time difference correlation coefficients and take the weighted average as the time difference correlation coefficient between the two electricity consumption time series.

4. The method for identifying upstream and downstream supply chains based on industry electricity data according to claim 1, characterized in that, Also includes: If the time difference correlation coefficient is less than the preset threshold, the weight value of the corresponding edge will be set to zero.

5. The method for identifying upstream and downstream supply chains based on industry electricity data according to claim 1, characterized in that, The weight of an edge is set to the absolute value of the time difference correlation coefficient between the two nodes it connects.

6. A supply chain upstream and downstream identification device based on industry electricity data, characterized in that, The system includes a processor and a memory for storing executable instructions; the processor is configured to read the executable instructions from the memory and execute the instructions to perform the following steps: Obtain the time series of electricity consumption for various industries to obtain a set of sequences. For the set of sequences, calculate the time difference correlation coefficient between every two electricity consumption time series; Using electricity consumption time series as nodes, and the weight of each edge as the time difference correlation coefficient between the two nodes it connects, a directed weighted graph is constructed. Using a directed weighted graph as input, the electricity consumption time series is scored through a node importance evaluation algorithm, specifically: The directed weighted graph is represented as ; right Normalize by column to obtain the matrix As the transition matrix; Construct a scoring vector, where each component of the scoring vector corresponds to a time series of electricity consumption; According to the matrix The time series score of each electricity consumption is iteratively calculated until the convergence condition is met, which can be expressed by the following formula: In the formula, the damping coefficient ; for A column vector of order whose components are all 1; Here is the transition matrix; Let k be the scoring vector, and k be the number of iterations. The industry tier is determined based on the scoring values ​​of electricity consumption time series; specifically: S1. Traverse the sequences of each undetermined level and classify at least one sequence whose score value meets the condition into the i-th level; reconstruct the directed weighted graph and calculate its score value based on the remaining undetermined levels of the sequences. The condition is expressed by the following formula: In the formula, This represents the score value of sequence u; U represents the set of sequences with undetermined hierarchies. Represents the set of all sequences; This represents the normalized value of the time difference correlation coefficient between sequence u and sequence i; This represents the normalized value of the time difference correlation coefficient between sequence u and sequence k. This is the score value for the current traversed sequence; S2. Repeat step S1 until the hierarchy of all sequences is obtained.

Images