Power data indexation method and system based on industry chain synergy effect

Abstract

The application discloses an electric power data indexation method and system based on an industrial chain synergistic effect. The application first collects power consumption time series data and multi-source external data of macro economy, production capacity, logistics and industry and commerce, and obtains a fusion data set through cleaning and space-time alignment fusion; then, an industrial chain topology graph is constructed relying on industry and commerce information, the power consumption correlation weight of an enterprise is calculated through a graph neural network, and a synergistic relationship graph with a causality label is generated in combination with Granger causality test; based on the graph, synergistic efficiency, operation resilience and development vitality indexes are calculated, an electric power-industry synergistic index is obtained by combination, and is periodically updated and smoothed; finally, index data packets are output to each business system to support intelligent decision-making. The application quantifies the power transmission law of upstream and downstream of the industrial chain by fusing multi-source data, constructs an index system with economic explanatory power, improves the real-time performance and accuracy of the operation situation evaluation of the industrial chain, and can support fine decision-making of enterprises, power grids and governments.

Description

Technical Field

[0001] This application relates to the fields of power data analysis and supply chain collaboration technology, specifically to a power data indexation method and system based on supply chain collaboration effects. Background Technology

[0002] The power system has accumulated massive amounts of user electricity consumption data. In traditional scenarios, this type of data is only used for basic business operations such as electricity billing, load forecasting, and fault monitoring.

[0003] Currently, some existing technologies attempt to conduct economic analysis based on electricity data. The conventional approach involves statistically analyzing total regional electricity consumption and changes in electricity load in key industries to help determine regional economic activity and industry operating rates. However, these methods treat electricity data as an independent statistical indicator, and the aforementioned electricity data analysis schemes have three main shortcomings: First, there is a lack of a collaborative analysis and modeling perspective for the industrial chain. Existing solutions only consider the absolute value of electricity consumption of a single enterprise or industry, without taking into account the linkage between upstream and downstream enterprises in production planning, logistics and distribution, inventory management and other aspects. They cannot uncover the dynamic correlation and response patterns of electricity consumption behavior of upstream and downstream enterprises, nor can they quantify the path and intensity of the transmission of upstream electricity consumption fluctuations to downstream.

[0004] Secondly, the economic explanatory power of the data indicators is insufficient. Relying solely on raw electricity consumption figures is insufficient to quantify the three core attributes of the industrial chain: resilience, collaborative efficiency, and development vitality. There is a lack of an index-based indicator system with economic explanatory power to characterize the overall operating status of the industrial chain, which fails to meet the refined and precise decision-making needs of multiple stakeholders, including the government, enterprises, and the power grid.

[0005] Third, the available data dimensions are limited. Existing technologies only use data from within the power system itself, without integrating external data from multiple sources such as macroeconomics, industry capacity, supply chain logistics, and enterprise business registration information to corroborate and interpret electricity consumption fluctuations. As a result, the analysis results have poor robustness, insufficient analytical depth, and are easily affected by data noise.

[0006] Due to the aforementioned shortcomings, existing power data processing methods are unable to transform power data into a proactive and accurate sensing carrier for understanding the operational status of regional industrial chains, making it difficult to support in-depth customer service and government industrial decision-making. Summary of the Invention

[0007] In response to the aforementioned background issues, such as existing technologies focusing only on the absolute value of electricity consumption of a single enterprise or industry, lacking the ability to model the synergistic effects of the industrial chain, and existing power data monitoring methods failing to directly characterize the deep economic attributes of the industrial chain, such as operational resilience, collaborative efficiency, and development vitality, making it difficult to support accurate decision-making for governments, enterprises, and the power grid, and existing solutions mainly relying on internal power system data and failing to fully utilize external data such as macroeconomic data, production capacity data, logistics data, and business information that can corroborate and explain changes in power behavior, the purpose of this invention is to provide a power data indexation method and system based on the synergistic effects of the industrial chain.

[0008] To achieve the objectives of this invention, the technical solution provided by this invention is as follows: First aspect This invention provides a method for indexing power data based on the synergistic effect of the industrial chain, comprising the following steps: Step 101: Collect raw data from multiple data sources, and clean, align and merge the data to generate a merged dataset; Step 102: Based on the fused dataset, construct a collaborative relationship map representing the correlation of electricity consumption behavior among upstream and downstream enterprises in the industrial chain, and quantify the correlation strength; Step 103: Based on the aforementioned collaborative relationship map, calculate the power-industry collaborative index to quantify the overall collaborative status of the industrial chain; Step 104: Regularly acquire new data to update and revise the calculated power-industry synergy index; Step 105: Output the power-industry synergy index to the business application terminal to support intelligent decision-making.

[0009] Further, step 101 includes the following steps: Step 1011: Through the data interface, obtain the cumulative electricity consumption time-series data recorded by multiple users' smart meters from the enterprise's electricity information collection system according to a preset period to obtain the original electricity consumption dataset; Step 1012: Simultaneously obtain multi-source external data from external data servers, including macroeconomic data, key industry capacity data, supply chain logistics data, and enterprise business registration information data; Step 1013: Perform a cleaning operation on the original electricity consumption dataset and the multi-source external data. The cleaning operation includes removing null values ​​and outliers to obtain the cleaned electricity consumption data and external data. Step 1014: Using the user's unique identifier and unified timestamp as the association key, perform spatiotemporal alignment and association fusion of the cleaned electricity consumption data with external data to generate a fused dataset.

[0010] Further, step 102 includes the following steps: Step 1021: Based on the enterprise business registration information in the fused dataset, identify upstream and downstream enterprise nodes belonging to the same industrial chain, and construct an industrial chain topology diagram with the supply relationship between enterprises as edges. Step 1022: Extract the time-series electricity consumption features of each node enterprise in the industrial chain topology diagram from the fused dataset to form a node feature matrix; Step 1023: Input the aforementioned industry chain topology diagram and node feature matrix into a pre-trained graph neural network model; the training method of this model is as follows: using a historical fusion dataset, with the goal of predicting the electricity consumption changes of the target node, the interaction patterns between adjacent nodes are learned by minimizing the prediction error; Step 1024: The graph neural network model aggregates and updates the features of adjacent nodes, calculates and outputs the weight values ​​of the edges between nodes, and these weight values ​​quantify the degree of influence of changes in the electricity consumption behavior of upstream enterprises on the electricity consumption behavior of downstream enterprises. Step 1025: For node pairs with high-weight edges, the Granger causality test method is used to analyze the electricity consumption time series data of both, identify the statistically significant causal transmission direction, and generate a supply chain collaboration relationship map with causal orientation and intensity labels.

[0011] Further, step 103 includes the following steps: Step 1031: Based on the weight values ​​of all edges in the industrial chain collaboration relationship graph, calculate the weighted average clustering coefficient of the entire industrial chain network to obtain the collaboration efficiency index; Step 1032: Based on the causal transmission path length and strength between nodes in the aforementioned industry chain collaboration relationship graph, calculate the relative change rate of the average path length of the network after simulating the deletion of a key upstream node, and obtain the operational resilience index; Step 1033: Based on the electricity consumption growth rate of each enterprise node in the fused dataset and the influence centrality of each enterprise node in the collaborative relationship graph, calculate the weighted average electricity consumption growth rate to obtain the development vitality index; encapsulate the collaborative efficiency index, operational resilience index and development vitality index into the power-industry collaborative index.

[0012] Further, step 104 includes the following steps: Step 1041: Repeat steps 101 to 103 according to the preset time period, and recalculate the updated power-industry synergy index based on the latest period's fusion dataset and synergy relationship graph. Step 1042: Compare the updated index with the index of the previous period and calculate the index change rate; if the index change rate exceeds a preset threshold, trigger the index correction process, use the index smoothing algorithm to smooth the current power-industry synergy index, and generate the corrected index sequence.

[0013] Further, step 105 includes the following steps: Step 1051: Encapsulate the corrected exponential sequence according to a preset data format to generate an exponential data package; Step 1052: Send the index data packet to the customer service response system, abnormal event early warning system, or government-enterprise collaborative service system to trigger the corresponding business rules or early warning strategies.

[0014] Second aspect This invention provides a power data indexing system based on the synergistic effect of the industrial chain, comprising the following modules: The fused dataset generation module is used to collect raw data from multiple data sources, clean, align, and merge it to generate a fused dataset. The collaborative relationship graph construction module is used to construct a collaborative relationship graph representing the correlation of electricity consumption behavior of upstream and downstream enterprises in the industrial chain based on the fused dataset, and to quantify the correlation strength. The power-industry synergy index calculation module is used to calculate the power-industry synergy index to quantify the overall synergy status of the industrial chain based on the synergy relationship map. The update and correction module is used to periodically acquire new data and update and correct the calculated power-industry synergy index. The output module is used to output the power-industry synergy index to the business application terminal to support intelligent decision-making.

[0015] Furthermore, the fused dataset generation module is used to perform the following steps: Step 1011: Through the data interface, obtain the cumulative electricity consumption time-series data recorded by multiple users' smart meters from the enterprise's electricity information collection system according to a preset period to obtain the original electricity consumption dataset; Step 1012: Simultaneously obtain multi-source external data from external data servers, including macroeconomic data, key industry capacity data, supply chain logistics data, and enterprise business registration information data; Step 1013: Perform a cleaning operation on the original electricity consumption dataset and the multi-source external data. The cleaning operation includes removing null values ​​and outliers to obtain the cleaned electricity consumption data and external data. Step 1014: Using the user's unique identifier and unified timestamp as the association key, perform spatiotemporal alignment and association fusion of the cleaned electricity consumption data with external data to generate a fused dataset.

[0016] Furthermore, the collaborative relationship graph construction module is used to perform the following steps: Step 1021: Based on the enterprise business registration information in the fused dataset, identify upstream and downstream enterprise nodes belonging to the same industrial chain, and construct an industrial chain topology diagram with the supply relationship between enterprises as edges. Step 1022: Extract the time-series electricity consumption features of each node enterprise in the industrial chain topology diagram from the fused dataset to form a node feature matrix; Step 1023: Input the aforementioned industry chain topology diagram and node feature matrix into a pre-trained graph neural network model; the training method of this model is as follows: using a historical fusion dataset, with the goal of predicting the electricity consumption changes of the target node, the interaction patterns between adjacent nodes are learned by minimizing the prediction error; Step 1024: The graph neural network model aggregates and updates the features of adjacent nodes, calculates and outputs the weight values ​​of the edges between nodes, and these weight values ​​quantify the degree of influence of changes in the electricity consumption behavior of upstream enterprises on the electricity consumption behavior of downstream enterprises. Step 1025: For node pairs with high-weight edges, the Granger causality test method is used to analyze the electricity consumption time series data of both, identify the statistically significant causal transmission direction, and generate a supply chain collaboration relationship map with causal orientation and intensity labels.

[0017] Furthermore, the power-industry synergy index calculation module is used to perform the following steps: Step 1031: Based on the weight values ​​of all edges in the industrial chain collaboration relationship graph, calculate the weighted average clustering coefficient of the entire industrial chain network to obtain the collaboration efficiency index; Step 1032: Based on the causal transmission path length and strength between nodes in the aforementioned industry chain collaboration relationship graph, calculate the relative change rate of the average path length of the network after simulating the deletion of a key upstream node, and obtain the operational resilience index; Step 1033: Based on the electricity consumption growth rate of each enterprise node in the fused dataset and the influence centrality of each enterprise node in the collaborative relationship graph, calculate the weighted average electricity consumption growth rate to obtain the development vitality index; encapsulate the collaborative efficiency index, operational resilience index and development vitality index into the power-industry collaborative index.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: The method provided by this invention integrates multi-source data and introduces a collaborative perspective of the industrial chain. It uses graph neural networks and time-series causal inference technology to perform structured modeling of the relationship between electricity flow and industrial flow, which can generate indexed indicators with economic explanatory power.

[0019] This method elevates power data from a passive recording tool to a sensor that actively senses the pulse of the economy, effectively improving the real-time, comparable, and interpretable assessment of the operational status of regional industrial chains. This enables it to better support the intelligent upgrading of core businesses such as customer service response, abnormal event early warning, and government-enterprise collaborative services. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of a power data indexation method based on the synergistic effect of the industrial chain, provided for an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0022] It should be noted that the acquisition of data and collection of information in this application are legal, compliant, or obtained with the consent of the subject of the data collection.

[0023] This application has the following definitions: The Power-Industry Synergy Index is a comprehensive index system used to quantitatively assess the collaborative operation status between upstream and downstream enterprises in a specific region or industrial chain. It includes multiple dimensions such as the synergy efficiency index, operational resilience index, and development vitality index.

[0024] Supply chain synergy effect: In this invention, supply chain synergy effect specifically refers to the time synchronization, causal correlation or response consistency of electricity consumption behavior of related enterprises in the supply chain due to their mutual dependence in production planning, logistics, inventory and other links.

[0025] Multi-source external data refers to a collection of data obtained from external systems, in addition to the user's original electricity consumption data, used to assist in the interpretation and verification of electricity behavior. This data collection includes macroeconomic data, key industry capacity data, supply chain logistics data, and enterprise business registration information.

[0026] like Figure 1 As shown, This application provides a method for indexing power data based on the synergistic effect of the industrial chain, which includes the following steps: Step 101: Collect raw data from multiple data sources, and clean, align and merge it to generate a standardized fused dataset.

[0027] Step 1011: Through the data interface, obtain the cumulative electricity consumption time-series data recorded by multiple users' smart meters from the enterprise's electricity information collection system according to a preset period to obtain the original electricity consumption dataset.

[0028] Among them, the cumulative electricity consumption time series data refers to the cumulative active energy value of the user's smart meter at different time points, which is continuously recorded in chronological order; the original electricity consumption dataset refers to a collection containing user identifier, timestamp, and cumulative electricity consumption fields without any processing.

[0029] In this embodiment, firstly, a data link is established with the electricity information collection system; secondly, data requests are sent to the collection system at fixed time intervals (e.g., 15 minutes, 1 hour, or 1 day); then, the collection system returns the meter readings of multiple users within a specified time range; finally, these readings, along with the user ID and timestamp, are stored as the original electricity data set.

[0030] In practical applications, taking an industrial park containing 50 new energy vehicle companies and their upstream and downstream suppliers as an example, the cumulative active power readings of the smart meters of the 50 companies over the past 24 hours are obtained through the power energy data acquisition interface at a 15-minute interval, resulting in a total of 50 × 96 = 4800 records. Each record contains the company ID, time point (accurate to the minute), and cumulative power consumption (unit: kWh), forming the original power consumption dataset.

[0031] Step 1012: Simultaneously obtain multi-source external data from external data servers. The multi-source external data includes macroeconomic data, key industry capacity data, supply chain logistics data, and enterprise business registration information data.

[0032] Among them, macroeconomic data refers to macroeconomic indicators such as GDP, industrial added value, and price index at the regional level; key industry capacity data refers to data such as output, operating rate, and inventory of specific industries (such as new energy vehicles, steel, and chemicals); supply chain logistics data refers to data such as cargo transportation volume, port throughput, and logistics timeliness related to the industrial chain; and enterprise business information data refers to static information such as the registered address, business scope, registered capital, shareholder structure, and establishment time of enterprises.

[0033] In this embodiment, firstly, data requests are sent to multiple external data sources in parallel; secondly, data files in JSON, XML, or CSV format are received; then, the format of the data returned by each data source is validated, and data blocks with incorrect formats are removed; finally, the validated multi-source data is temporarily stored in memory, awaiting subsequent cleaning.

[0034] In practical applications, the "New Energy Vehicle Industry Policy Index" (range 0-100) for the city where the industrial park is located in the current month is obtained from the government statistics bureau website; the average daily truck traffic volume (unit: vehicle trips) of the main logistics channels of the industrial park in the past week is obtained from the logistics platform API; and the registered capital (unit: RMB 10,000) and industrial chain segment tags (such as "battery production", "motor production", "vehicle assembly") of 50 enterprises are obtained from the enterprise information query platform.

[0035] Step 1013: Perform cleaning operations on the original electricity consumption dataset and multi-source external data. The cleaning operations include removing null values ​​and outliers to obtain cleaned electricity consumption data and external data.

[0036] Among them, null values ​​refer to field values ​​that are missing or NULL in data records; outliers refer to values ​​that clearly do not conform to physical meaning or statistical laws, such as negative electricity consumption, values ​​that exceed the maximum range of the electricity meter, or policy indices that exceed the range of 0 to 100.

[0037] In this embodiment, firstly, the original electricity consumption dataset is traversed, and records with empty or negative cumulative electricity consumption are deleted; secondly, the 3σ principle is used to detect outliers in electricity consumption, and records that deviate from the mean by more than 3 times the standard deviation are marked and removed; next, a similar cleaning process is performed on the external dataset, for example, deleting records where the policy index is not in the range [0,100]; finally, the remaining data after cleaning are stored as the cleaned electricity consumption dataset and the cleaned external dataset, respectively.

[0038] In practical application, a record of -5.2 kWh for a certain enterprise's electricity consumption was found, which was identified as an outlier and the record was deleted; another record of a policy index of 105, exceeding the upper limit, was also deleted. After cleaning, 4780 of the original 4800 electricity consumption records remained, achieving a 98% integrity rate for external data records.

[0039] Step 1014: Using the user's unique identifier and unified timestamp as the association key, perform spatiotemporal alignment and association fusion with multi-source external data to generate a fused dataset. The fused dataset serves as the input for the next step.

[0040] Among them, a unique user identifier refers to a code that can uniquely identify an electricity user, such as an enterprise's unified social credit code or an electricity meter asset number; a unified timestamp refers to converting the time field of data from different sources into a time point with the same time zone and the same precision (such as accurate to the minute); spatiotemporal alignment refers to associating the electricity consumption data of the same user at the same time point with external data into the same record; and a fused dataset refers to a wide table that includes user ID, timestamp, electricity consumption, and external multi-source fields.

[0041] In this embodiment, firstly, the cleaned electricity consumption dataset and the external dataset are initially grouped according to the "Enterprise ID" field; secondly, for each enterprise, the timestamp of the enterprise's electricity consumption data is matched with the timestamp of the external data to find the external record with the smallest time difference; then, the successfully matched electricity consumption record and the external record are merged into a new record; finally, all the new records are summarized to generate a fused dataset, where each row of the fused dataset corresponds to the complete information of an enterprise at a certain point in time.

[0042] In practical applications, the "Unified Social Credit Code of Enterprise" and "Data Collection Time (Year, Month, Day, Hour)" are used as the joint primary key. The cleaned electricity consumption data of 50 enterprises are associated with the corresponding policy index, truck traffic volume, registered capital and other information into the same record, generating a fusion dataset containing fields such as timestamp, enterprise ID, electricity consumption, industrial chain link, policy index, and logistics index. It contains 30 days of data from 50 enterprises, totaling 1,500 records.

[0043] Step 102: Based on the fused dataset, construct a collaborative relationship map representing the correlation of electricity consumption behavior among upstream and downstream enterprises in the industrial chain, and quantify the correlation strength.

[0044] Step 1021: Based on the enterprise business registration information in the fused dataset, identify upstream and downstream enterprise nodes belonging to the same industrial chain, and construct an industrial chain topology diagram with the supply relationship between enterprises as the edge.

[0045] The industry chain topology diagram is a directed graph. Nodes in the graph represent enterprises, and directed edges represent material or semi-finished product supply relationships from upstream suppliers to downstream manufacturers. Upstream and downstream enterprise nodes refer to related enterprises determined based on their business scope, industry classification, and procurement contracts or public supply chain relationships between enterprises in business registration information.

[0046] In this embodiment, firstly, the "industrial chain link" field of all enterprises is extracted from the fused dataset; secondly, the upstream and downstream nodes of each enterprise are determined according to the preset industrial chain knowledge graph (e.g., the hierarchical relationship of "battery production → motor production → vehicle assembly" in the new energy vehicle industrial chain); then, a directed edge is created for each supply relationship; finally, all nodes and edges are combined into a directed graph, i.e., the industrial chain topology graph.

[0047] In practical applications, based on the "industry chain link" field in the fused dataset, 15 "battery supplier" nodes, 10 "motor supplier" nodes, and 25 "vehicle assembly plant" nodes were identified. According to the actual production process of new energy vehicles, the supply relationships from each "battery supplier" and "motor supplier" to each "vehicle assembly plant" were defined as edges in the graph, ultimately constructing an industry chain topology graph containing 50 nodes and 15×25+10×25=625 directed edges.

[0048] Step 1022: Extract the time-series electricity consumption features of each node enterprise in the industrial chain topology diagram from the fused dataset to form a node feature matrix.

[0049] Among them, the electricity consumption time series feature refers to the electricity consumption sequence of each enterprise within a continuous time window, such as the total daily electricity consumption over the past 30 days; the node feature matrix is ​​a two-dimensional array, with rows corresponding to nodes (enterprises) and columns corresponding to time points, and the matrix elements being the electricity consumption value of the enterprise at the corresponding time point.

[0050] In this embodiment, firstly, based on the node ID list in the industry chain topology diagram, the electricity consumption records of these enterprises are filtered out from the fused dataset; secondly, the electricity consumption of each enterprise is sorted according to a unified time order to form a time series; then, the time series of all enterprises are aligned to the same set of time points; finally, the aligned sequences are stacked row by row to generate a node feature matrix.

[0051] In practical applications, the total daily electricity consumption of 50 companies over the past 30 days is extracted to form a matrix with 50 rows and 30 columns. .use The index representing the enterprise (value range 1 to 50) is used. The index representing the number of days (values ​​range from 1 to 30) then Indicates the first The company in the Daily electricity consumption, measured in kWh.

[0052] Step 1023: Input the industry chain topology diagram and node feature matrix into a pre-trained graph neural network model. The training method of the graph neural network model is as follows: using a historical fusion dataset, with the goal of predicting the electricity consumption changes of the target node, the interaction patterns between adjacent nodes are learned by minimizing the prediction error.

[0053] Among them, the graph neural network model is a deep learning model that can process graph structure data. It uses a message passing mechanism to aggregate information from neighboring nodes to update the representation of the current node. The training method refers to the process of iteratively optimizing model parameters using historical data. The training objective is to minimize the mean square error between the predicted electricity consumption and the actual electricity consumption.

[0054] In this embodiment, firstly, the pre-trained graph neural network model parameters are loaded; secondly, the graph structure obtained in step 1021 is... and the feature matrix obtained in step 1022 The first feature is used as the input to the model; then, the model performs multi-layer graph convolution operations, where each node in each layer aggregates the features of its neighboring nodes; finally, the model outputs the updated feature vector for each node, which contains information about industry chain collaboration.

[0055] In practical applications, a two-layer graph convolutional network model is loaded. This model is trained for 100 epochs using historical data from 200 companies. The training objective is to predict the electricity consumption of downstream companies for the next day based on the electricity consumption of upstream companies over the past week. The prediction error on the final validation set is 5.2%. The graph and feature matrix of the current 50 nodes are input into the model, and the model outputs 50 128-dimensional feature vectors.

[0056] Step 1024: The graph neural network model aggregates and updates the features of adjacent nodes, calculates and outputs the weight values ​​of the edges between nodes, and quantifies the impact of changes in the electricity consumption behavior of upstream enterprises on the electricity consumption behavior of downstream enterprises. The weight values ​​serve as input for the next step.

[0057] The weight of an edge is a real number between 0 and 1. The larger the value, the stronger the impact of changes in electricity consumption by upstream enterprises on electricity consumption by downstream enterprises. Aggregation and updating refer to the process in a graph neural network where nodes update their own features by weighted summation of the features of their neighboring nodes and then through nonlinear transformation.

[0058] In this embodiment, firstly, the model combines the learned node feature vectors in pairs. For each directed edge, it calculates the cosine similarity between the source node features and the target node features or obtains the initial weights through a learnable attention function. Secondly, the model normalizes the weights of all incoming edges for each target node. Next, the model outputs the normalized weight values ​​as the final edge weights. Finally, the node ID, edge direction, and weight values ​​are saved as an edge weight list.

[0059] In practical applications, the model calculates the edge weight from "Battery Supplier A" to "Vehicle Assembly Plant B" to be 0.85, and the edge weight from "Motor Supplier C" to "Vehicle Assembly Plant B" to be 0.32. These two edges and their corresponding weights are recorded to form an edge weight list.

[0060] Step 1025: For node pairs with high-weight edges, the Granger causality test method is used to analyze the electricity consumption time series data of the two nodes, identify the statistically significant causal transmission direction, and generate a supply chain collaboration relationship map with causal orientation and intensity labels.

[0061] Among them, the Granger causality test is a statistical hypothesis testing method used to determine whether one time series is helpful in predicting another time series; a high-weighted edge refers to an edge whose weight value is greater than a preset threshold (e.g., 0.6); the causal transmission direction refers to the direction from the cause sequence to the result sequence determined when the p-value of the F statistic in the Granger test is less than the significance level (e.g., 0.05).

[0062] In this embodiment, firstly, the electricity consumption time series data of the two enterprises corresponding to the high-weight edges are extracted from the fused dataset; secondly, a stationarity test is performed on each pair of time series data, and if it is not stationary, first-order differencing is performed; then, a lag period is constructed as follows: The Granger causal regression model was used to calculate the F-statistic and p-value. Finally, edges with p-values ​​less than 0.05 were marked as edges with causal relationships, and the original weights were retained as strength labels to update the original industrial chain synergy relationship map.

[0063] In practical applications, Granger causality tests were performed on the daily electricity consumption series of "Battery Supplier A" and "Vehicle Assembly Plant B" over the past 30 days. With a maximum lag order of 3, the calculated F-statistic was 8.32, corresponding to a p-value of 0.012, which is less than 0.05, thus confirming a unidirectional causal transmission from "Battery Supplier A" to "Vehicle Assembly Plant B". This edge was marked in the original graph as "Causal direction: A→B, Influence strength: 0.85".

[0064] Step 103: Based on the aforementioned industrial chain collaboration relationship map, calculate the power-industry collaboration index to quantify the overall collaboration status of the industrial chain. Step 1031: Calculate the weighted average clustering coefficient of the entire industrial chain network based on the weight values ​​of all edges in the collaborative relationship graph, and obtain the collaborative efficiency index.

[0065] Among them, the weighted average clustering coefficient is an indicator that measures the tendency of nodes in the network to form close clusters (triangular structures), taking into account the weight of the edges; the higher the collaborative efficiency index value, the higher the efficiency of information or material transmission in the industrial chain, and the closer the local collaboration between enterprises.

[0066] In this embodiment, firstly, each node in the collaborative relationship graph is traversed, and the weighted local clustering coefficient of each node is calculated; secondly, the local clustering coefficients of all nodes are summed; finally, the weighted average clustering coefficient, i.e., the collaborative efficiency index, is obtained by dividing by the total number of nodes.

[0067] For nodes The formula for calculating the weighted local clustering coefficient is: in, It is a node The number of neighboring nodes, It is a node with neighbors The normalized weight of the edges between them (weight divided by the maximum weight). Traversing nodes All different neighbor pairs.

[0068] In practical applications, a collaborative efficiency index is calculated for the weighted graph generated in step 1025. Assuming the weighted average clustering coefficient of 50 nodes is 0.72, this value is used as the collaborative efficiency index of the current industrial park, indicating that the local collaboration of the industrial chain is close and the information transmission efficiency is high.

[0069] Step 1032: Based on the causal transmission path length and strength between nodes in the collaborative relationship graph, calculate the relative change rate of the average path length of the network after simulating the deletion of a key upstream node, and obtain the operational resilience index.

[0070] Among them, the key upstream node refers to the node with the largest degree (the number of edges pointing downstream); the average path length of the network refers to the average of the shortest path lengths between all node pairs; the relative rate of change = (average path length after deletion - original average path length) / original average path length; the larger the operational resilience index, the stronger the dependence of the industrial chain on key nodes and the worse its resilience.

[0071] In this embodiment of the application, firstly, the shortest path length between all node pairs in the original cooperative relationship graph is calculated, and then the average is obtained. Next, identify the node with the largest out-degree as the key upstream node, and delete this node and all its edges from the graph. Then, calculate the average shortest path length between all pairs of nodes remaining in the graph after the deletion. Finally, calculate the relative rate of change. As an index of operational resilience.

[0072] In practical applications, the average path length of the original graph is 2.1, and the out-degree of the key upstream node "Battery Supplier A" is 25 (pointing to 25 vehicle assembly plants). After deleting this node, the average path length of the remaining 49 nodes becomes 3.4, thus improving the operational resilience index. That is, 61.9%. This high index indicates that the industry chain is heavily reliant on "battery supplier A" and has poor resilience.

[0073] Step 1033: Based on the electricity consumption growth rate of each enterprise node in the fused dataset and the influence centrality of each enterprise node in the collaborative relationship graph, calculate the weighted average electricity consumption growth rate to obtain the development vitality index.

[0074] Among them, the electricity consumption growth rate refers to the percentage change in the electricity consumption of an enterprise in the current period compared to the previous period; the influence centrality refers to the degree centrality (i.e., the sum of the in-degree and out-degree of a node) or betweenness centrality of a node in the graph, reflecting the importance of a node in the network; the weighted average electricity consumption growth rate is the result of weighting the electricity consumption growth rates of each enterprise with centrality as the weight. The higher the value, the stronger the overall development vitality of the industrial chain.

[0075] In this embodiment of the application, firstly, the total electricity consumption of each enterprise for the two most recent time periods (such as this month and last month) is extracted from the fused dataset, and the month-on-month growth rate is calculated. Secondly, calculate the degree centrality of each node in the collaborative relationship graph. ,in Represents a node in-degree (pointing to the node) (number of edges) Represents a node out-degree (from node) (Number of starting edges).

[0076] Next, calculate the weighted average growth rate: in, The total number of nodes. For nodes degree centrality, For nodes The growth rate of electricity consumption.

[0077] In practical applications, the degree center of the core enterprise "vehicle assembly plant B" is... The company's electricity consumption growth rate was Degree center of a small "battery supplier" The company's electricity consumption growth rate was Assume there are 48 other firms, each with known degree centrality and growth rate. Calculate the degree centrality and growth rate of all firms. The sum, then divided by the total number of enterprises. The sum of these yields the development vitality index. 15%, indicating that the entire industrial chain is showing positive growth momentum.

[0078] Step 104: Regularly acquire new data to update and revise the calculated indices to maintain their timeliness and accuracy. Step 1041: Repeat steps 101 to 103 according to the preset time period, and recalculate the updated power-industry synergy index based on the latest period's fusion dataset and synergy relationship graph.

[0079] The preset time period can be daily, weekly, or monthly, depending on the business's timeliness requirements; the updated index refers to the collaborative efficiency index, operational resilience index, and development vitality index, which are recalculated based on the latest data.

[0080] In this embodiment of the application, firstly, a timer triggers an update task; secondly, new electricity consumption data and external data since the previous calculation are acquired; then, the data acquisition, cleaning, and fusion in step 101, the map construction and update in step 102, and the index calculation in step 103 are repeated; finally, a new set of index values ​​is output, covering the results of the previous cycle.

[0081] In practical applications, it is set to automatically perform an update once a day at 2:00 AM. For example, if the index set (0.72, 0.62, 0.15) is calculated on July 1, the new data for the entire day of July 1 is obtained at 2:00 AM on July 2, and the updated index (0.70, 0.65, 0.16) is obtained after recalculation.

[0082] Step 1042: Compare the updated index with the index of the previous period and calculate the rate of change of the index. If the rate of change exceeds a preset threshold, trigger the index correction process, use the index smoothing algorithm to smooth the current index, and generate a corrected index sequence.

[0083] The rate of change of the index refers to the absolute value of the difference between the updated index and the previous period's index divided by the previous period's index; the preset threshold can be set according to business sensitivity, for example, 0.1 (i.e., 10%); the exponential smoothing algorithm adopts the first exponential smoothing formula.

[0084] In this embodiment of the application, firstly, the rate of change of each sub-index is calculated. Secondly, judgment Is it greater than the threshold? (e.g., 0.1); if greater than, then... Perform exponential smoothing; the smoothing formula is: .in, For the first The exponential value after periodic smoothing; For the first The original index value obtained from the period calculation (i.e., the updated index) ); For the first The smoothed exponent value (i.e., the smoothed exponent value of the previous period; if there is no smoothed value of the previous period, then take the previous period's value). ); The smoothing factor ranges from 0 to 1, typically between 0.2 and 0.3. Finally, the smoothed exponent is used to replace the original updated value, generating the corrected exponent sequence.

[0085] In practical applications, the smoothed value of the resilience index from the previous cycle... The updated original index value The rate of change This triggers a correction. The smoothing factor is then calculated. The corrected exponent is The final operational resilience index was output as 0.6589.

[0086] Step 105: Output the power-industry synergy index to the business application terminal to support intelligent decision-making. Step 1051: Encapsulate the power-industry synergy index according to the preset data format to generate an index data package.

[0087] The preset data format can be a structured format such as JSON, XML, or Protobuf; the index data packet is a data unit containing fields such as timestamp, index name, index value, and validity period.

[0088] In this embodiment, firstly, the current timestamp and index validity period (e.g., until the next update) are obtained; secondly, the collaborative efficiency index, operational resilience index, development vitality index, and comprehensive index (which can be defined as the weighted sum of the three sub-indices) are organized in key-value pair format; then, metadata (such as data version and calculation cycle) is added; finally, the entire dictionary is serialized into a JSON string to generate an index data package.

[0089] Step 1052: Send the index data packet to the customer service response system, abnormal event early warning system, or government-enterprise collaborative service system to trigger the corresponding business rules or early warning strategies.

[0090] Among them, the customer service response system is used to proactively push service suggestions to enterprise customers based on index changes; the abnormal event early warning system is used to issue supply chain risk warnings when the operational resilience index exceeds the threshold; and the government-enterprise collaborative service system is used to provide supply chain operation reports to government economic departments.

[0091] In this embodiment, firstly, a list of target systems for the index data packet is determined according to the configured business rules; secondly, the data packet is sent to each target system in parallel via a message queue or HTTP API; then, after receiving the data packet, each target system parses the index value and matches it with its internal rule engine; finally, the rule engine executes corresponding actions, such as sending an alert email, generating a work order, or updating the dashboard.

[0092] In practical applications, after receiving the index data packet, the abnormal event early warning system detects that the operational resilience index of 0.6589 is greater than the preset threshold of 0.5, triggering the "supply chain risk early warning" rule, automatically generating an early warning work order and pushing it to the account manager responsible for the industrial park, while displaying the early warning information on the enterprise's large screen monitoring system.

[0093] Through the steps described above, this application deeply integrates massive amounts of raw power data with multi-source industrial information, and uses graph neural networks and causal inference techniques for structured modeling. Ultimately, it generates an index system that can reflect the collaborative efficiency, operational resilience, and development vitality of the industrial chain in real time and quantitatively, effectively supporting the power sector's accurate perception and proactive service of the industrial economy.

[0094] In addition, corresponding to the above method, the present invention also provides a power data indexing system based on the synergistic effect of the industrial chain, comprising the following modules: The fused dataset generation module is used to collect raw data from multiple data sources, clean, align, and merge it to generate a fused dataset. The collaborative relationship graph construction module is used to construct a collaborative relationship graph representing the correlation of electricity consumption behavior of upstream and downstream enterprises in the industrial chain based on the fused dataset, and to quantify the correlation strength. The power-industry synergy index calculation module is used to calculate the power-industry synergy index to quantify the overall synergy status of the industrial chain based on the synergy relationship map. The update and correction module is used to periodically acquire new data and update and correct the calculated power-industry synergy index. The output module is used to output the power-industry synergy index to the business application terminal to support intelligent decision-making.

[0095] Finally, it should be noted that the above embodiments are merely illustrative and explanatory of the present invention, and are not intended to limit the present invention to the scope of the described embodiments. Furthermore, those skilled in the art will understand that the present invention is not limited to the above embodiments, and many more variations and modifications can be made based on the teachings of the present invention, all of which fall within the scope of protection claimed by the present invention.

Claims

1. A method for indexing electricity data based on the synergistic effect of the industrial chain, characterized in that, Includes the following steps: Step 101: Collect raw data from multiple data sources, and clean, align and merge the data to generate a merged dataset; Step 102: Based on the fused dataset, construct a collaborative relationship map representing the correlation of electricity consumption behavior among upstream and downstream enterprises in the industrial chain, and quantify the correlation strength; Step 103: Based on the aforementioned collaborative relationship map, calculate the power-industry collaborative index to quantify the overall collaborative status of the industrial chain; Step 104: Regularly acquire new data to update and revise the calculated power-industry synergy index; Step 105: Output the power-industry synergy index to the business application terminal to support intelligent decision-making.

2. The power data indexing method based on the synergistic effect of the industrial chain according to claim 1, characterized in that, Step 101 includes the following steps: Step 1011: Through the data interface, obtain the cumulative electricity consumption time-series data recorded by multiple users' smart meters from the enterprise's electricity information collection system according to a preset period to obtain the original electricity consumption dataset; Step 1012: Simultaneously obtain multi-source external data from external data servers, including macroeconomic data, key industry capacity data, supply chain logistics data, and enterprise business registration information data; Step 1013: Perform a cleaning operation on the original electricity consumption dataset and the multi-source external data. The cleaning operation includes removing null values ​​and outliers to obtain the cleaned electricity consumption data and external data. Step 1014: Using the user's unique identifier and unified timestamp as the association key, perform spatiotemporal alignment and association fusion of the cleaned electricity consumption data with external data to generate a fused dataset.

3. The power data indexing method based on the synergistic effect of the industrial chain according to claim 1, characterized in that, Step 102 includes the following steps: Step 1021: Based on the enterprise business registration information in the fused dataset, identify upstream and downstream enterprise nodes belonging to the same industrial chain, and construct an industrial chain topology diagram with the supply relationship between enterprises as edges. Step 1022: Extract the time-series electricity consumption features of each node enterprise in the industrial chain topology diagram from the fused dataset to form a node feature matrix; Step 1023: Input the aforementioned industry chain topology diagram and node feature matrix into a pre-trained graph neural network model; the training method of this model is as follows: using a historical fusion dataset, with the goal of predicting the electricity consumption changes of the target node, the interaction patterns between adjacent nodes are learned by minimizing the prediction error; Step 1024: The graph neural network model aggregates and updates the features of adjacent nodes, calculates and outputs the weight values ​​of the edges between nodes, and these weight values ​​quantify the degree of influence of changes in the electricity consumption behavior of upstream enterprises on the electricity consumption behavior of downstream enterprises. Step 1025: For node pairs with high-weight edges, the Granger causality test method is used to analyze the electricity consumption time series data of both, identify the statistically significant causal transmission direction, and generate a supply chain collaboration relationship map with causal orientation and intensity labels.

4. The power data indexing method based on the synergistic effect of the industrial chain according to claim 1, characterized in that, Step 103 includes the following steps: Step 1031: Based on the weight values ​​of all edges in the industrial chain collaboration relationship graph, calculate the weighted average clustering coefficient of the entire industrial chain network to obtain the collaboration efficiency index; Step 1032: Based on the causal transmission path length and strength between nodes in the aforementioned industry chain collaboration relationship graph, calculate the relative change rate of the average path length of the network after simulating the deletion of a key upstream node, and obtain the operational resilience index; Step 1033: Based on the electricity consumption growth rate of each enterprise node in the fused dataset and the influence centrality of each enterprise node in the collaborative relationship graph, calculate the weighted average electricity consumption growth rate to obtain the development vitality index; encapsulate the collaborative efficiency index, operational resilience index and development vitality index into the power-industry collaborative index.

5. The power data indexing method based on the synergistic effect of the industrial chain according to claim 1, characterized in that, Step 104 includes the following steps: Step 1041: Repeat steps 101 to 103 according to the preset time period, and recalculate the updated power-industry synergy index based on the latest period's fusion dataset and synergy relationship graph. Step 1042: Compare the updated index with the index of the previous period and calculate the index change rate; if the index change rate exceeds a preset threshold, trigger the index correction process, use the index smoothing algorithm to smooth the current power-industry synergy index, and generate the corrected index sequence.

6. The power data indexation method based on the synergistic effect of the industrial chain according to claim 1, characterized in that, Step 105 includes the following steps: Step 1051: Encapsulate the corrected exponential sequence according to a preset data format to generate an exponential data package; Step 1052: Send the index data packet to the customer service response system, abnormal event early warning system, or government-enterprise collaborative service system to trigger the corresponding business rules or early warning strategies.

7. A power data indexing system based on the synergistic effect of the industrial chain, characterized in that, Includes the following modules: The fused dataset generation module is used to collect raw data from multiple data sources, clean, align, and merge it to generate a fused dataset. The collaborative relationship graph construction module is used to construct a collaborative relationship graph representing the correlation of electricity consumption behavior of upstream and downstream enterprises in the industrial chain based on the fused dataset, and to quantify the correlation strength. The power-industry synergy index calculation module is used to calculate the power-industry synergy index to quantify the overall synergy status of the industrial chain based on the synergy relationship map. The update and correction module is used to periodically acquire new data and update and correct the calculated power-industry synergy index. The output module is used to output the power-industry synergy index to the business application terminal to support intelligent decision-making.

8. The power data indexing system based on the synergistic effect of the industrial chain according to claim 7, characterized in that, The fused dataset generation module is used to perform the following steps: Step 1011: Through the data interface, obtain the cumulative electricity consumption time-series data recorded by multiple users' smart meters from the enterprise's electricity information collection system according to a preset period to obtain the original electricity consumption dataset; Step 1012: Simultaneously obtain multi-source external data from external data servers, including macroeconomic data, key industry capacity data, supply chain logistics data, and enterprise business registration information data; Step 1013: Perform a cleaning operation on the original electricity consumption dataset and the multi-source external data. The cleaning operation includes removing null values ​​and outliers to obtain the cleaned electricity consumption data and external data. Step 1014: Using the user's unique identifier and unified timestamp as the association key, perform spatiotemporal alignment and association fusion of the cleaned electricity consumption data with external data to generate a fused dataset.

9. A power data indexing system based on the synergistic effect of the industrial chain according to claim 7, characterized in that, The collaborative relationship graph construction module is used to perform the following steps: Step 1021: Based on the enterprise business registration information in the fused dataset, identify upstream and downstream enterprise nodes belonging to the same industrial chain, and construct an industrial chain topology diagram with the supply relationship between enterprises as edges. Step 1022: Extract the time-series electricity consumption features of each node enterprise in the industrial chain topology diagram from the fused dataset to form a node feature matrix; Step 1023: Input the aforementioned industry chain topology diagram and node feature matrix into a pre-trained graph neural network model; the training method of this model is as follows: using a historical fusion dataset, with the goal of predicting the electricity consumption changes of the target node, the interaction patterns between adjacent nodes are learned by minimizing the prediction error; Step 1024: The graph neural network model aggregates and updates the features of adjacent nodes, calculates and outputs the weight values ​​of the edges between nodes, and these weight values ​​quantify the degree of influence of changes in the electricity consumption behavior of upstream enterprises on the electricity consumption behavior of downstream enterprises. Step 1025: For node pairs with high-weight edges, the Granger causality test method is used to analyze the electricity consumption time series data of both, identify the statistically significant causal transmission direction, and generate a supply chain collaboration relationship map with causal orientation and intensity labels.

10. A power data indexing system based on the synergistic effect of the industrial chain according to claim 7, characterized in that, The power-industry synergy index calculation module is used to perform the following steps: Step 1031: Based on the weight values ​​of all edges in the industrial chain collaboration relationship graph, calculate the weighted average clustering coefficient of the entire industrial chain network to obtain the collaboration efficiency index; Step 1032: Based on the causal transmission path length and strength between nodes in the aforementioned industry chain collaboration relationship graph, calculate the relative change rate of the average path length of the network after simulating the deletion of a key upstream node, and obtain the operational resilience index; Step 1033: Based on the electricity consumption growth rate of each enterprise node in the fused dataset and the influence centrality of each enterprise node in the collaborative relationship graph, calculate the weighted average electricity consumption growth rate to obtain the development vitality index; The collaborative efficiency index, operational resilience index, and development vitality index are packaged into a power-industry collaborative index.

Images