An industrial power intelligent prediction method and system based on an industry chain portrait

Abstract

The application discloses an industrial electricity intelligent prediction method and system based on an industry chain portrait, relates to the technical field of intelligent prediction, and comprises the following steps: obtaining multi-source operation data covering an upstream raw material end, a midstream manufacturing end and a downstream application end of a target industrial enterprise, arranging and modeling the data, and forming an industry chain portrait reflecting node supply and demand relations; fusing the industry chain portrait based on external factors, enterprise location information, production capacity grades and energy consumption characteristics, and obtaining standardized portrait data; establishing electricity consumption transmission relations between the upstream and the midstream and between the midstream and the downstream, and generating a lag mapping function in combination with a stock turnover period and an order production cycle; constructing an electricity consumption prediction model, performing time series prediction and environment correction on the target enterprise, and generating comprehensive electricity consumption prediction information. The application can realize multi-dimensional energy consumption dynamic prediction of a glass industry chain, and improve prediction precision and energy management level.

Description

Technical Field

[0001] This invention relates to the field of intelligent forecasting technology, specifically to an intelligent forecasting method and system for industrial electricity consumption based on supply chain profiling. Background Technology

[0002] The electricity consumption forecasting method and system for the glass industry chain aims to improve the accuracy and coordination of overall energy consumption control across the entire chain. Against the backdrop of rising energy costs and increasingly stringent requirements for green manufacturing, traditional single-enterprise electricity management methods are insufficient to meet the needs of collaborative optimization across the entire chain. By systematically modeling and predictively analyzing the electricity consumption relationships between upstream raw material supply, midstream manufacturing and processing, and downstream application demands, the energy consumption transmission patterns between each node can be effectively revealed, reducing energy waste, improving the matching degree of production plans, and providing enterprises with a scientific basis for power dispatching and energy-saving decisions. The application of this method helps promote intelligent electricity management and lean production in the glass industry, and is of great significance for the stable operation of the entire chain and the improvement of energy utilization efficiency.

[0003] However, existing methods for forecasting electricity consumption in the glass industry typically rely solely on historical energy consumption data from a single enterprise or node for statistical analysis. This neglects the supply-demand coupling relationship between upstream and downstream industries, as well as the lagged effects of factors such as inventory and production scheduling on electricity consumption changes. Consequently, the forecasts fail to accurately reflect actual energy consumption trends. Furthermore, energy supply levels in different regions, electricity price fluctuations, and policy adjustments also dynamically influence enterprise electricity consumption behavior. Existing methods exhibit significant shortcomings in model adaptability and generalization capabilities. Therefore, a smart industrial electricity consumption forecasting method and system based on a supply chain profile is needed to achieve multi-source data fusion modeling, lagged response analysis, and cross-node electricity consumption forecasting for the glass industry supply chain, thereby addressing the aforementioned problems. Summary of the Invention

[0004] To address the aforementioned technical problems, this paper provides an intelligent forecasting method and system for industrial electricity consumption based on supply chain profiling. This solution overcomes the limitations of existing electricity consumption forecasting methods in the glass industry, which typically rely solely on historical energy consumption data from a single enterprise or node for statistical analysis. These methods neglect the lagged effects of supply and demand relationships between upstream and downstream industries, as well as factors such as inventory and production scheduling on electricity consumption changes, leading to inaccurate forecasts that fail to reflect actual energy consumption trends. Furthermore, the dynamic impact of energy supply levels, electricity price fluctuations, and policy adjustments in different regions on enterprise electricity consumption behavior highlights the significant shortcomings of existing methods in terms of model adaptability and generalization ability.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A smart prediction method for industrial electricity consumption based on supply chain profiling includes:

[0007] Acquire multi-source operational data covering the upstream raw material end, midstream manufacturing end, and downstream application end of the target industrial enterprise. The multi-source operational data includes, but is not limited to, inventory turnover cycle, order scheduling cycle, material input, output, manufacturing load, order volume, and historical electricity consumption.

[0008] The multi-source operational data is organized and structured, and the operational sequence data of the upstream raw material end, midstream manufacturing end and downstream application end are extracted to form a picture of the industrial chain;

[0009] The process involves tracing the industrial chain profile, extracting enterprise location information, production capacity level, and energy consumption characteristic parameters for each node in the industrial chain, and obtaining regional external factor data related to the target industrial enterprise. The energy consumption characteristic parameters include indicators related to unit output energy consumption, equipment load rate, energy utilization efficiency, and peak-valley load ratio.

[0010] Based on regional external factor data, enterprise location information, production capacity level and energy consumption characteristic parameters, the industrial chain profile is integrated to obtain standardized industrial chain profile data.

[0011] Based on standardized supply chain profile data, a first electricity consumption transmission relationship is established between the upstream raw material end and the midstream manufacturing end, and a second electricity consumption transmission relationship is established between the midstream manufacturing end and the downstream application end.

[0012] Based on the inventory turnover cycle and order scheduling cycle in the multi-source operation data, a first lag mapping function is generated for the first electricity consumption transmission relationship between the upstream raw material end and the midstream manufacturing end, and a second lag mapping function is generated for the second electricity consumption transmission relationship between the midstream manufacturing end and the downstream application end.

[0013] Based on the first and second lag mapping functions, an electricity consumption prediction model is established to predict the multi-source operation data of the target industrial enterprises, and to generate comprehensive electricity consumption prediction information by combining regional external factor data.

[0014] In an optional embodiment, the step of organizing and structuring multi-source operational data, and extracting operational sequence data from upstream raw material sources, midstream manufacturing, and downstream applications to form a supply chain profile, specifically includes:

[0015] Extract time stamps, acquisition frequency, and data category information from multi-source operational data;

[0016] The multi-source running data is aligned and completed according to the time stamp to eliminate time offsets and missing data between different sources, and the multi-source running data is synchronized in time.

[0017] The multi-source operational data after time synchronization is grouped according to data categories to obtain data grouping results. The data categories include economic data, material data, and energy consumption data.

[0018] Based on the data grouping results, a structured data table is established, which includes a node identifier field, a time field, and an indicator field.

[0019] Based on the node identifier field, construct node index sets for the upstream raw material end, the midstream manufacturing end, and the downstream application end;

[0020] Based on the node index set, runtime sequence data of each node is extracted from multi-source runtime data. The runtime sequence data includes material input change curves, manufacturing load change trends, and order quantity change sequences.

[0021] Time gradient analysis is performed on the runtime sequence data of each node to obtain the rate of change of input and output between nodes, and the input-output dependency relationship between nodes is determined by correlation matching of the rate of change of different nodes.

[0022] Based on the input-output dependency relationship, the quantitative coupling coefficient between the node's material input and output is analyzed, and the supply and demand mapping relationship between nodes is established.

[0023] Based on the supply and demand mapping relationship and runtime sequence data, a portrait of the industrial chain reflecting the dynamic transmission characteristics of the industrial chain is generated.

[0024] The formula for calculating the quantity coupling coefficient is as follows:

[0025]

[0026] In the formula, For upstream nodes in time The amount of material fed at any given time. For downstream nodes in time Output at any given moment The length of the observation period. This refers to the lag time between upstream material input and downstream output. This represents the average amount of material fed over the observation period. This represents the average output over the observation period.

[0027] In an optional embodiment, the step of fusing data on the industrial chain profile based on regional external factors, enterprise location information, production capacity level, and energy consumption characteristic parameters to obtain standardized industrial chain profile data specifically includes:

[0028] Based on the historical electricity consumption and output of each node in the industrial chain, energy consumption characteristic parameters are obtained. These energy consumption characteristic parameters include indicators related to unit output energy consumption, equipment load rate, energy utilization efficiency, and peak-valley load ratio.

[0029] Acquire regional external factor data corresponding to the target industrial enterprise and its upstream and downstream related enterprises. The external factor data includes regional temperature information, industrial policy indicators, energy supply level and regional electricity price fluctuation information.

[0030] Based on the enterprise location information, spatial registration of external factor data is performed to determine the regional environmental characteristics of each enterprise node.

[0031] Based on the enterprise's production capacity level, enterprises within the same industry node are hierarchically clustered to form a hierarchical node set;

[0032] Based on the energy consumption characteristic parameters, the energy consumption characteristics of the hierarchical node set are processed to obtain the node energy consumption standard parameters;

[0033] The node energy consumption standard parameters are integrated with the regional environmental characteristics of the corresponding node to generate a comprehensive node feature set;

[0034] Based on the node comprehensive feature set, the amplitude of the runtime sequence data is corrected using the node energy consumption standard parameters, and the temporal variation trend of the runtime sequence data is adjusted in combination with the regional environmental characteristics to form a fused temporal feature sequence.

[0035] Based on the fused time-series feature sequences, the industry chain profile is updated to obtain standardized industry chain profile data.

[0036] In an optional embodiment, the step of establishing a first electricity consumption transmission relationship between the upstream raw material end and the midstream manufacturing end, and establishing a second electricity consumption transmission relationship between the midstream manufacturing end and the downstream application end, based on standardized industry chain profile data, specifically includes:

[0037] Obtain data on electricity consumption changes at different points in the industrial chain and their corresponding production load, material input, and order volume.

[0038] Based on the supply and demand mapping relationship between nodes in the standardized industrial chain profile data, the supply and demand response range of upstream raw material nodes to midstream manufacturing nodes, and the load response range of midstream manufacturing nodes to downstream application nodes are determined.

[0039] Cross-correlation function analysis was performed on the electricity consumption change data in each interval to determine the correlation strength of electricity consumption changes under different time lags and obtain the time dependence matrix.

[0040] Based on the time dependency matrix, node pairs with a correlation strength greater than a preset threshold are extracted, and a set of power consumption transmission relationships between nodes is established.

[0041] Based on the transmission relationship set, the node pairs involving the upstream raw material end and the midstream manufacturing end are divided into the first power consumption transmission relationship, and the node pairs involving the midstream manufacturing end and the downstream application end are divided into the second power consumption transmission relationship.

[0042] In an optional embodiment, the step of generating a first lag mapping function for the first electricity consumption transmission relationship between the upstream raw material end and the midstream manufacturing end, and generating a second lag mapping function for the second electricity consumption transmission relationship between the midstream manufacturing end and the downstream application end, based on the inventory turnover cycle and order scheduling cycle in the multi-source operational data, specifically includes:

[0043] Based on the inventory turnover cycle, the transmission relationship of primary electricity consumption between upstream raw material end and midstream manufacturing end is analyzed, and the time series of inventory-driven electricity consumption response under different inventory levels is extracted.

[0044] The time series of electricity consumption response driven by inventory is normalized and smoothed to obtain the time correlation coefficient between inventory changes and electricity consumption changes, and to obtain the inventory delay characteristics of upstream nodes on the electricity consumption of midstream nodes under different inventory states.

[0045] Based on the inventory delay characteristics and time correlation coefficient, an inventory-driven lag response model is established.

[0046] The inventory-driven lag response model is mapped onto the first electricity consumption transmission relationship to form the first lag mapping function that reflects the impact of changes in electricity consumption at upstream nodes on midstream nodes;

[0047] Based on the order scheduling cycle, the transmission relationship of the second electricity consumption between the midstream manufacturing end and the downstream application end is analyzed, and the scheduling-driven electricity consumption response time series under different scheduling rhythms is extracted.

[0048] The time series of power consumption response driven by production scheduling is denoised and standardized to obtain the time response coefficients of production scheduling changes and power consumption changes, and to obtain the scheduling delay characteristics of the power consumption of the downstream node under different production scheduling states.

[0049] Based on the production delay characteristics and time response coefficient, a production scheduling-driven lag response model is established.

[0050] The production scheduling-driven lag response model is mapped onto the second electricity consumption transmission relationship to form a second lag mapping function that reflects the impact of changes in electricity consumption at midstream nodes on downstream nodes;

[0051] The formula for the inventory-driven lag response model is as follows:

[0052]

[0053] In the formula, For the midstream node at time Electricity consumption For the upstream node at time The amount of material to be fed, The time correlation coefficient, This represents the maximum lag time between upstream material input and midstream node electricity consumption. This refers to the lag time between upstream material input and midstream electricity consumption.

[0054] The formula for the production scheduling-driven lag response model is as follows:

[0055]

[0056] In the formula, For downstream nodes at time Electricity consumption For the midstream node at time Manufacturing load, The time response coefficient, This represents the maximum lag time between midstream manufacturing load and downstream electricity consumption. The lag time between midstream manufacturing load and downstream electricity consumption.

[0057] In an optional embodiment, the step of establishing an electricity consumption prediction model based on a first lag mapping function and a second lag mapping function, predicting multi-source operating data of the target industrial enterprise, and generating comprehensive electricity consumption prediction information by combining regional external factor data, specifically includes:

[0058] Based on standardized supply chain profile data and the first and second lag mapping functions, the time series data of the corresponding nodes are mapped to obtain the lag mapping output, and cross-node electricity consumption feature data is extracted to obtain a feature set that can be used for modeling. The node electricity consumption feature data includes the node electricity consumption, material input, manufacturing load, order volume and lag mapping output related features of the upstream raw material end, midstream manufacturing end and downstream application end.

[0059] The feature set is processed to obtain the processed feature data;

[0060] Based on the processed feature data, a distributed lag regression modeling method is adopted. The power consumption, material input, manufacturing load, order volume and lag mapping output of the upstream raw material end, midstream manufacturing end and downstream application end are used as input variables, and the power consumption of the target node is used as the output variable to establish an initial cross-node power consumption prediction model.

[0061] The initial prediction model is iteratively trained to obtain an optimized electricity consumption prediction model that meets the accuracy requirements;

[0062] Acquire historical electricity consumption data of the target industrial enterprise on a daily, weekly, and monthly scale, including electricity consumption records from the past three months to one year;

[0063] Historical data is input into the optimized electricity consumption prediction model for prediction, generating nodal electricity consumption prediction results for the next 24 hours, 7 days and 30 days respectively.

[0064] Environmental corrections are performed on the node power consumption forecast results based on external factor data to obtain corrected forecast results.

[0065] Based on the corrected prediction results, comprehensive electricity consumption prediction information is generated.

[0066] Furthermore, an intelligent industrial electricity consumption forecasting system based on supply chain profiling is proposed to achieve the intelligent forecasting method described above, characterized by comprising:

[0067] A multi-source operation data acquisition module is used to acquire multi-source operation data covering the upstream raw material end, midstream manufacturing end and downstream application end of the target industrial enterprise;

[0068] The industry chain profile building module is used to organize and structure the collected multi-source operation data, and generate an industry chain profile that reflects the dynamic transmission characteristics between nodes.

[0069] The standardized fusion module is used to fuse the industrial chain profile, combining enterprise location information, production capacity level, energy consumption characteristics and regional external factor data to generate standardized industrial chain profile data.

[0070] The electricity consumption transmission analysis module is used to establish cross-node electricity consumption transmission relationships based on standardized industrial chain profile data, and generate a first lag mapping function and a second lag mapping function.

[0071] The electricity consumption forecasting module is used to construct an electricity consumption forecasting model and generate comprehensive electricity consumption forecasting information based on the lag mapping function and standardized industrial chain profile data.

[0072] In an optional embodiment, the multi-source operational data acquisition module includes:

[0073] The data interface unit is used to receive inventory turnover cycle, order scheduling cycle and historical electricity consumption data of the target industrial enterprise and its upstream and downstream enterprises.

[0074] The data preprocessing unit is used to perform time synchronization, missing data completion, and data type grouping processing on the collected data.

[0075] The data storage unit is used to store the processed multi-source operational data and provide it to the industry chain profiling module.

[0076] In an optional embodiment, the industry chain profiling module includes:

[0077] A node timing extraction unit is used to extract runtime timing data from upstream raw material end, midstream manufacturing end and downstream application end from structured multi-source runtime data;

[0078] An input-output dependency analysis unit is used to identify the input-output dependencies between runtime sequence data of each node and to establish a supply-demand mapping relationship between nodes.

[0079] The industrial chain profile generation unit is used to generate an industrial chain profile that reflects the dynamic transmission characteristics of the industrial chain based on the supply and demand mapping relationship and runtime sequence data.

[0080] In an optional embodiment, the electricity consumption prediction module includes:

[0081] A lag mapping generation unit is used to generate a first lag mapping function and a second lag mapping function based on the inventory turnover cycle and the order scheduling cycle.

[0082] The feature extraction and processing unit is used to extract and process cross-node electricity consumption feature data based on standardized industrial chain profile data and hysteresis mapping function;

[0083] A prediction model building unit is used to build an electricity consumption prediction model based on the processed feature data and to make time-series predictions on the historical electricity consumption data of the target industrial enterprise.

[0084] An environmental correction unit is used to correct the prediction results based on external factor data and generate final comprehensive electricity consumption prediction information.

[0085] Compared with the prior art, the beneficial effects of the present invention are:

[0086] This solution proposes an intelligent industrial electricity consumption forecasting method and system based on supply chain profiling. It constructs a supply chain profile reflecting the supply and demand relationships at each node by structurally modeling multi-source operational data covering upstream raw materials, midstream manufacturing, and downstream applications. This profile integrates enterprise location information, production capacity, energy consumption characteristics, and external factors to form standardized supply chain profile data. By establishing the electricity consumption transmission relationship between upstream, midstream, and downstream nodes and generating a lag mapping function based on inventory turnover and order scheduling cycles, the time dependence of electricity consumption transmission between nodes is characterized. This method reveals the correlation patterns of electricity consumption changes in each link of the supply chain, identifies key influencing factors, and constructs a high-precision cross-node electricity consumption forecasting model. This effectively improves the accuracy and real-time performance of electricity consumption forecasting, avoids energy consumption fluctuations or resource waste caused by supply and demand mismatches, and provides technical support for energy optimization scheduling and intelligent decision-making in the glass industry. Attached Figure Description

[0087] Figure 1 This is a flowchart of an intelligent prediction method for industrial electricity consumption based on supply chain profiling proposed in this invention.

[0088] Figure 2 A flowchart for constructing a supply chain profile in this invention;

[0089] Figure 3 This is a flowchart of the cross-node power consumption prediction process in this invention;

[0090] Figure 4 This is a system framework diagram of an intelligent industrial power consumption prediction system based on supply chain profiling proposed in this invention. Detailed Implementation

[0091] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0092] Reference Figure 1 - Figure 4 As shown in the figure, an intelligent prediction method for industrial electricity consumption based on supply chain profiling in an embodiment of the present invention includes:

[0093] Acquire multi-source operational data covering the upstream raw material end, midstream manufacturing end, and downstream application end of the target industrial enterprise. Multi-source operational data includes, but is not limited to, inventory turnover cycle, order scheduling cycle, material input, output, manufacturing load, order volume, and historical electricity consumption.

[0094] The multi-source operational data is organized and structured, and the operational sequence data of the upstream raw material end, midstream manufacturing end and downstream application end are extracted to form a picture of the industrial chain;

[0095] The process involves tracing the industrial chain profile, extracting enterprise location information, production capacity level, and energy consumption characteristic parameters at each node of the industrial chain, and obtaining regional external factor data related to the target industrial enterprise. The energy consumption characteristic parameters include indicators related to unit output energy consumption, equipment load rate, energy utilization efficiency, and peak-valley load ratio.

[0096] Based on regional external factor data, enterprise location information, production capacity level and energy consumption characteristic parameters, the industrial chain profile is integrated to obtain standardized industrial chain profile data.

[0097] Based on standardized supply chain profile data, a first electricity consumption transmission relationship is established between the upstream raw material end and the midstream manufacturing end, and a second electricity consumption transmission relationship is established between the midstream manufacturing end and the downstream application end.

[0098] Based on the inventory turnover cycle and order scheduling cycle in the multi-source operation data, a first lag mapping function is generated for the first electricity consumption transmission relationship between the upstream raw material end and the midstream manufacturing end, and a second lag mapping function is generated for the second electricity consumption transmission relationship between the midstream manufacturing end and the downstream application end.

[0099] Based on the first and second lag mapping functions, an electricity consumption prediction model is established to predict the multi-source operation data of the target industrial enterprises, and to generate comprehensive electricity consumption prediction information by combining regional external factor data.

[0100] Furthermore, the multi-source operational data is organized and structured, and operational sequence data from the upstream raw material end, midstream manufacturing end, and downstream application end are extracted to form a supply chain profile, specifically including:

[0101] Extract time stamps, acquisition frequency, and data category information from multi-source operational data;

[0102] The multi-source running data is aligned and completed according to the time stamp to eliminate time offsets and missing data between different sources, and the multi-source running data is synchronized in time.

[0103] The multi-source operational data after time synchronization is grouped according to data category to obtain data grouping results. The data categories include economic data, material data, and energy consumption data.

[0104] Based on the data grouping results, a structured data table is established, which includes a node identifier field, a time field, and an indicator field.

[0105] Specifically, time stamps, collection frequency, and data category information are extracted from multi-source operational data to ensure the consistency of data from different sources in terms of time dimension and indicator types. Time stamps indicate the collection point of each data point. For example, in the glass industry, raw material companies' records of quartz sand and soda ash inputs are typically sampled in hours, midstream manufacturing data on furnace power consumption and production load are generally sampled in 15-minute intervals, and downstream glass deep-processing companies' order volume and shipment information are updated daily. Identifying the collection frequency helps achieve unified processing of multi-scale time series during data fusion. Data category information is used to classify data with different attributes, mainly including three categories: economic data, material data, and energy consumption data. Economic data reflects the operating rhythm and market fluctuations of the industrial chain, material data describes the flow relationship between raw materials, semi-finished products, and finished products at different stages, and energy consumption data characterizes the changing characteristics of electricity consumption and manufacturing load. Aligning and completing multi-source operational data based on time stamps can be understood as eliminating time offsets and missing data between different sources by unifying the time axis and interpolation correction methods. For example, when raw material input records are missing for certain periods, the system can dynamically complete the missing data by combining changes in production load at the manufacturing end and upstream inventory levels to ensure the integrity and continuity of time-series data. After processing, time-synchronized multi-source operational data is obtained. Subsequently, the system groups the time-synchronized operational data according to data categories to form economic, material, and energy consumption datasets, and builds structured data tables on this basis. The structured data tables include node identifier fields, time fields, and indicator fields. The node identifier field is used to distinguish different enterprise nodes such as upstream raw material end, midstream manufacturing end, and downstream application end. The time field is used to record the sampling time, and the indicator field is used to store core indicators such as input quantity, manufacturing load, order quantity, and electricity consumption, thereby achieving standardized expression of multi-source heterogeneous data in the industrial chain and providing a unified data foundation for the subsequent construction of industrial chain profiles.

[0106] Based on the node identifier field, construct node index sets for the upstream raw material end, the midstream manufacturing end, and the downstream application end;

[0107] Based on the node index set, the runtime sequence data of each node is extracted from multi-source runtime data. The runtime sequence data includes the material input change curve, the manufacturing load change trend, and the order quantity change sequence.

[0108] Time gradient analysis is performed on the runtime sequence data of each node to obtain the rate of change of input and output between nodes, and the input-output dependency relationship between nodes is determined by correlation matching of the rate of change of different nodes.

[0109] Based on the input-output dependency relationship, the quantitative coupling coefficient between the node's material input and output is analyzed, and the supply and demand mapping relationship between nodes is established.

[0110] Based on the supply and demand mapping relationship and runtime sequence data, a portrait of the industrial chain reflecting the dynamic transmission characteristics of the industrial chain is generated.

[0111] The formula for calculating the quantitative coupling coefficient is as follows:

[0112]

[0113] In the formula, For upstream nodes in time The amount of material fed at any given time. For downstream nodes in time Output at any given moment The length of the observation period. This refers to the lag time between upstream material input and downstream output. This represents the average amount of material fed over the observation period. This represents the average output over the observation period.

[0114] Specifically, after standardizing and organizing multi-source operational data, the system uses the node identifier field in the structured data table to encode and classify each enterprise node in the industry chain to establish a node index system. The node identifier field is used to clarify the enterprise's position attributes and functional characteristics in the production system, thereby achieving accurate matching and hierarchical management of node-level data. For example, in the glass industry chain, raw material suppliers such as quartz sand and soda ash are classified as upstream raw material nodes, float glass manufacturers responsible for melting and forming are classified as midstream manufacturing nodes, and architectural glass, automotive glass, and photovoltaic glass processing companies are classified as downstream application nodes. Based on this index system, the system extracts the operational sequence information of each node from the multi-source operational data, including the upstream material input change curve, the midstream manufacturing load and energy consumption trend, and the downstream order volume change sequence, thereby forming a time-related industry chain operational sequence set.

[0115] Based on this, the system calculates the rate of change of operational indicators (such as input quantity, manufacturing load, output, or order quantity) for each node at each observation time point. This rate of change is the ratio of the difference between indicators at adjacent time points to the time interval, quantifying the trend of node status increase or decrease over time. Subsequently, the system compares the rate of change of input quantity of upstream nodes with the rate of change of output quantity of downstream nodes hourly. By calculating the correlation or matching degree between the two at different time lags, it identifies how long it takes for changes in upstream input to have a significant impact on downstream output, thus determining the input-output dependency relationship between nodes. To quantitatively characterize the degree of supply-demand coupling between different nodes, the system introduces a quantitative coupling coefficient as a key parameter. The quantitative coupling coefficient is calculated by analyzing the relationship between changes in input quantity and output quantity between nodes. The system first compares the changes in input quantity of upstream nodes with the changes in output quantity of midstream nodes, and observes the corresponding relationship between the two at different time lags, thereby identifying the transmission time of upstream changes to midstream output. Then, by statistically analyzing the proportional relationship between changes in input quantity and changes in output quantity at different time periods, the system determines the intensity of the impact of upstream input on midstream output, and uses this to form the coupling coefficient. The same method can be used to analyze the transmission relationship between midstream production load and downstream order volume. For each pair of nodes, the system first determines whether their quantitative coupling coefficient exceeds a preset threshold. If it exceeds the threshold, it is considered that changes in the input of the upstream node have a significant impact on the output of the downstream node. Then, the system records each pair of nodes with significant supply-demand relationships and their corresponding coupling strength, storing them in matrix form. The matrix rows and columns correspond to the node numbers, and the values ​​of the matrix elements represent the supply-demand coupling strength between the nodes. Through this mapping relationship, the system can clearly identify which upstream nodes affect each downstream node and the degree of influence, while revealing the dynamic transmission paths of material flow, energy flow, and information flow between nodes. This provides quantitative and traceable basic data for constructing a supply chain profile and subsequent electricity consumption forecasting models.

[0116] Furthermore, based on regional external factor data, enterprise location information, production capacity level, and energy consumption characteristic parameters, the industrial chain profile is fused to obtain standardized industrial chain profile data, specifically including:

[0117] Based on the historical electricity consumption and output of each node in the industrial chain, energy consumption characteristic parameters are obtained. These energy consumption characteristic parameters include indicators related to unit output energy consumption, equipment load rate, energy utilization efficiency, and peak-valley load ratio.

[0118] Obtain regional external factor data corresponding to the target industrial enterprise and its upstream and downstream related enterprises. The external factor data includes regional temperature information, industrial policy indicators, energy supply level and regional electricity price fluctuation information.

[0119] Based on the enterprise location information, spatial registration of external factor data is performed to determine the regional environmental characteristics of each enterprise node.

[0120] Based on the enterprise's production capacity level, enterprises within the same industry node are hierarchically clustered to form a hierarchical node set;

[0121] Based on the energy consumption characteristic parameters, the energy consumption characteristics of the hierarchical node set are processed to obtain the node energy consumption standard parameters;

[0122] The node energy consumption standard parameters are integrated with the regional environmental characteristics of the corresponding node to generate a comprehensive node feature set;

[0123] Based on the node comprehensive feature set, the amplitude of the runtime sequence data is corrected using the node energy consumption standard parameters, and the temporal variation trend of the runtime sequence data is adjusted in combination with the regional environmental characteristics to form a fused temporal feature sequence.

[0124] Based on the fused time-series feature sequences, the industry chain profile is updated to obtain standardized industry chain profile data.

[0125] Specifically, the system first extracts energy consumption characteristic parameters based on the historical electricity consumption and output of each node in the industrial chain. These parameters include indicators such as energy consumption per unit output, energy utilization efficiency, and peak-valley load ratio, used to characterize the electricity consumption patterns and energy efficiency levels of each node under different production conditions. These characteristic parameters reflect the energy intensity, load fluctuation range, and output efficiency of the nodes, providing a quantitative basis for subsequent analysis of the dynamic characteristics of the industrial chain. Simultaneously, the system acquires regional external factor data corresponding to the target industrial enterprises and their upstream and downstream related enterprises. These external factors include regional temperature changes, industrial policy indicators, energy supply levels, and regional electricity price fluctuations, used to reflect the potential impact of the external environment on enterprise operations and energy consumption. For example, in the glass industry chain, the furnace load and energy consumption of float glass manufacturers are significantly affected by high temperatures, while the production rhythm of photovoltaic glass processing enterprises may be affected by regional electricity price fluctuations.

[0126] Subsequently, the system spatially registers regional external factor data based on the geographical location information of enterprises to determine the regional environmental characteristics of each node. These regional environmental characteristics include indicators such as the average temperature, temperature fluctuation range, industrial policy intensity, energy supply stability, and electricity price fluctuation range of the node's location, used to characterize the potential impact of the external environment on enterprise electricity consumption and production rhythm. For enterprises within the same industry node, the system performs hierarchical clustering according to production capacity level, forming hierarchical node sets to characterize the operational differences of enterprises of different sizes and production capacity levels. The system utilizes the energy consumption characteristic parameters of each node to perform energy consumption characteristic processing on the hierarchical node sets. Specifically, it first compares each node's unit output energy consumption, equipment load rate, energy utilization efficiency, and peak-valley load ratio with other nodes within the same hierarchical node, calculates the relative deviation, and standardizes the indicators so that different indicators can be weighted and summarized under the same dimension. Simultaneously, it adjusts the indicator weights based on the hierarchical clustering results to reflect the impact of different production capacity levels on energy consumption characteristics. Through this processing, the system obtains standard parameters for node energy consumption. These parameters comprehensively quantify the electricity consumption characteristics and energy efficiency level of each node under specific production conditions. They can be directly used to correct the amplitude of runtime sequence data and adjust the trend of time-series changes, making the fused time-series feature sequence more accurately reflect the true dynamic operating characteristics of the nodes. The standard parameters for node energy consumption are combined with regional environmental characteristics to form a comprehensive feature set for nodes. Based on this, the amplitude of historical electricity consumption curves is adjusted, appropriately amplifying the amplitude of nodes with high energy consumption characteristics and appropriately reducing the amplitude of nodes with low energy consumption characteristics. Simultaneously, the system adjusts the trend of time-series data in conjunction with regional environmental characteristics. For example, enterprises in high-temperature areas may experience peak loads during the day and low loads at night. The system uses information such as regional temperature, energy supply stability, and electricity price fluctuations to correct the peaks and troughs of the curves, making the time-series trend more closely match the actual operating environment. After amplitude correction and trend adjustment, a fused time-series feature sequence is formed, and the industry chain profile is updated to obtain standardized industry chain profile data. This allows the dynamic characteristics of each node to simultaneously reflect its own operating rules, production capacity differences, and the impact of the external environment, providing a comparable and reliable data foundation for subsequent electricity consumption forecasting.

[0127] Furthermore, based on standardized supply chain profiling data, a primary electricity consumption transmission relationship is established between upstream raw material suppliers and midstream manufacturing, and a secondary electricity consumption transmission relationship is established between midstream manufacturing and downstream application users. Specifically, this includes:

[0128] Obtain data on electricity consumption changes at different points in the industrial chain and their corresponding production load, material input, and order volume.

[0129] Based on the supply and demand mapping relationship between nodes in the standardized industrial chain profile data, the supply and demand response range of upstream raw material nodes to midstream manufacturing nodes, and the load response range of midstream manufacturing nodes to downstream application nodes are determined.

[0130] Cross-correlation function analysis was performed on the electricity consumption change data in each interval to determine the correlation strength of electricity consumption changes under different time lags and obtain the time dependence matrix.

[0131] Based on the time dependency matrix, node pairs with a correlation strength greater than a preset threshold are extracted, and a set of power consumption transmission relationships between nodes is established.

[0132] Based on the transmission relationship set, the system classifies node pairs involving upstream raw materials and midstream manufacturing into the first power consumption transmission relationship, and node pairs involving midstream manufacturing and downstream applications into the second power consumption transmission relationship. Specifically, the system first acquires power consumption change data for each node in the industrial chain at different time periods, along with corresponding production load, raw material input, and order volume information, to characterize the energy consumption and operational characteristics of each node under different production conditions. Combining the supply and demand mapping relationship in the standardized industrial chain profile, the system can determine the response range of upstream raw material nodes to midstream manufacturing nodes, and the load response range of midstream manufacturing nodes to downstream application nodes. The response range represents the time interval within which the production activities of upstream or midstream nodes affect the power consumption or output of downstream nodes after a certain time delay. For example, in the glass industry chain, when the amount of quartz sand or soda ash added increases, the load of the float glass melting furnace usually rises after a few hours to a day. This period is the supply and demand response range of the upstream raw material end to the midstream manufacturing end. The change in the midstream manufacturing load will affect the electricity consumption and production rhythm of downstream architectural glass or photovoltaic glass processing enterprises after a certain lag. This lag time is the load response range of the midstream to the downstream.

[0133] After determining the response interval, the system performs cross-correlation function analysis on the electricity consumption change data within each interval to quantify the correlation strength between nodes. This includes smoothing and standardizing the electricity consumption sequences of upstream and downstream nodes, and then calculating the cross-correlation values ​​of the two sequences at different lag times to measure the degree of response of upstream node changes to downstream nodes. The peak value of the cross-correlation function and the corresponding lag time are used to identify the most significant energy consumption transmission period between nodes and form a time dependence matrix. Each element in the matrix represents the correlation strength of electricity consumption changes between a pair of nodes under a specific lag condition. Subsequently, the system extracts node pairs in the matrix whose correlation strength exceeds a preset threshold to establish a set of electricity consumption transmission relationships between nodes. The transmission relationship reflects the impact of the electricity consumption or production behavior of upstream nodes on the electricity consumption and load of downstream nodes after a certain time delay. In the glass industry chain, the node pairs of upstream raw material end and midstream manufacturing end form the first electricity consumption transmission relationship, reflecting the energy consumption transmission law of raw material input to melting production; the node pairs of midstream manufacturing end and downstream application end form the second electricity consumption transmission relationship, reflecting the transmission effect of production load changes on the electricity demand of terminal processing enterprises. Through cross-correlation function analysis, the system can accurately characterize the energy flow and load response between various links in the industrial chain, providing a data foundation for the generation of lag mapping functions and electricity consumption forecasting.

[0134] Furthermore, based on the inventory turnover cycle and order scheduling cycle in the multi-source operational data, a first lag mapping function is generated for the first electricity consumption transmission relationship between the upstream raw material end and the midstream manufacturing end, and a second lag mapping function is generated for the second electricity consumption transmission relationship between the midstream manufacturing end and the downstream application end, specifically including:

[0135] Based on the inventory turnover cycle, the transmission relationship of primary electricity consumption between upstream raw material end and midstream manufacturing end is analyzed, and the time series of inventory-driven electricity consumption response under different inventory levels is extracted.

[0136] The time series of electricity consumption response driven by inventory is normalized and smoothed to obtain the time correlation coefficient between inventory changes and electricity consumption changes, and to obtain the inventory delay characteristics of upstream nodes on the electricity consumption of midstream nodes under different inventory states.

[0137] Based on the inventory delay characteristics and time correlation coefficient, an inventory-driven lag response model is established.

[0138] The inventory-driven lag response model is mapped onto the first electricity consumption transmission relationship to form the first lag mapping function that reflects the impact of changes in electricity consumption at upstream nodes on midstream nodes;

[0139] Understandably, the system first constructs time series of upstream raw material inputs and midstream electricity consumption based on observation points, and then divides the time periods into three categories: low inventory, medium inventory, and high inventory, according to upstream raw material inventory levels. At each inventory level, the system performs sliding window processing and normalization on the time series, calculating the rate of change in upstream inventory and midstream electricity consumption within each window. Subsequently, at different lag times, the system compares the rate of change series of upstream raw material inputs with the rate of change series of midstream electricity consumption hourly, calculating their time correlation coefficient. This correlation coefficient reflects the strength of the response of the upstream input change rate to the midstream electricity consumption change rate over time, thus identifying the delayed characteristics of the impact of upstream inputs on midstream electricity consumption. Based on the inventory delay characteristics and the time correlation coefficient, the system establishes an inventory-driven lag response model to describe the time-varying response of upstream input changes to midstream electricity consumption under different inventory states. The lag mapping function then mathematically maps the model results to the supply chain transmission relationship, quantifying the specific impact of upstream electricity consumption changes on midstream nodes. For example, in the glass industry chain, when the raw material inventory of float glass manufacturers is at a high level, their furnace load and energy consumption will increase by about 10% after 6 to 12 hours. When the inventory is at a low level, the response lag time may be extended to 24 hours with a smaller increase. This delay characteristic and change magnitude are quantified by the first lag mapping function, thereby providing accurate time response information for subsequent cross-node power consumption forecasting.

[0140] Based on the order scheduling cycle, the transmission relationship of the second electricity consumption between the midstream manufacturing end and the downstream application end is analyzed, and the scheduling-driven electricity consumption response time series under different scheduling rhythms is extracted.

[0141] The time series of power consumption response driven by production scheduling is denoised and standardized to obtain the time response coefficients of production scheduling changes and power consumption changes, and to obtain the scheduling delay characteristics of the power consumption of the downstream node under different production scheduling states.

[0142] Based on the production delay characteristics and time response coefficient, a production scheduling-driven lag response model is established.

[0143] The production scheduling-driven lag response model is mapped onto the second electricity consumption transmission relationship to form a second lag mapping function that reflects the impact of changes in electricity consumption at midstream nodes on downstream nodes;

[0144] Understandably, the system analyzes the secondary electricity consumption transmission relationship between the midstream manufacturing end and the downstream application end based on the order scheduling cycle of the midstream manufacturing nodes. The system divides time periods into three categories: slow, medium, and fast, according to different scheduling rhythms. Under each scheduling rhythm, the system constructs time series for the output of the midstream nodes and the electricity consumption of the downstream nodes at the observation time points, and performs denoising and standardization processing on the time series to eliminate data noise and dimensional differences. Subsequently, under different lag times, the system compares the rate of change sequence of the output of the midstream nodes with the rate of change sequence of the electricity consumption of the downstream nodes hourly, calculating their time response coefficients. This coefficient reflects the response strength of midstream scheduling changes to downstream electricity consumption changes after a certain time, thereby identifying the delay characteristics of midstream output to downstream electricity consumption. Based on the scheduling delay characteristics and time response coefficients, the system establishes a scheduling-driven lag response model to describe the time-varying response law of midstream node output changes to downstream node electricity consumption under different scheduling states. The second lag mapping function mathematically maps the model results to the transmission relationship in the industrial chain, quantifying the specific impact of changes in electricity consumption at midstream nodes on downstream nodes. For example, in the glass industry chain, when float glass manufacturers are operating at high production levels, the energy consumption of downstream architectural glass processing companies will increase by about 8% after 2 to 4 hours. However, when midstream production is slower, the response lag may extend to 6 hours with a smaller increase. This delay characteristic and response magnitude are quantified through the second lag mapping function, thus providing accurate time response information and inter-node transmission patterns for subsequent cross-node electricity consumption forecasting.

[0145] The formula for the inventory-driven lag response model is as follows:

[0146]

[0147] In the formula, For the midstream node at time Electricity consumption For the upstream node at time The amount of material to be fed, The time correlation coefficient, This represents the maximum lag time between upstream material input and midstream node electricity consumption. This refers to the lag time between upstream material input and midstream electricity consumption.

[0148] The formula for the production scheduling-driven lag response model is as follows:

[0149]

[0150] In the formula, For downstream nodes at time Electricity consumption For the midstream node at time Manufacturing load, The time response coefficient, This represents the maximum lag time between midstream manufacturing load and downstream electricity consumption. The lag time between midstream manufacturing load and downstream electricity consumption.

[0151] Furthermore, based on the first and second lag mapping functions, an electricity consumption forecasting model is established to predict the multi-source operating data of the target industrial enterprises. This model is then combined with regional external factor data to generate comprehensive electricity consumption forecasting information, specifically including:

[0152] Based on standardized supply chain profile data and the first and second lag mapping functions, the time series data of the corresponding nodes are mapped to obtain the lag mapping output. Cross-node electricity consumption feature data is extracted to obtain a feature set that can be used for modeling. The node electricity consumption feature data includes the node electricity consumption, material input, manufacturing load, order volume and lag mapping output related features of the upstream raw material end, midstream manufacturing end and downstream application end.

[0153] The feature set is processed to obtain the processed feature data;

[0154] Based on the processed feature data, a distributed lag regression modeling method is adopted. The power consumption, material input, manufacturing load, order volume and lag mapping output of the upstream raw material end, midstream manufacturing end and downstream application end are used as input variables, and the power consumption of the target node is used as the output variable to establish an initial cross-node power consumption prediction model.

[0155] The initial prediction model is iteratively trained to obtain an optimized electricity consumption prediction model that meets the accuracy requirements;

[0156] Specifically, based on standardized industry chain profile data and previously established first and second lag mapping functions, the system maps the time-series data of nodes at the upstream raw material end, midstream manufacturing end, and downstream application end to obtain lag mapping outputs. The mapping process essentially transforms the electricity transmission relationships and time delay characteristics between different nodes into calculable numerical outputs through functions, thereby reflecting the energy consumption transfer patterns between nodes. Based on the lag mapping outputs, the system further extracts cross-node electricity consumption characteristic data to form a feature set that can be used for modeling. Subsequently, the system performs data processing on the feature set, including missing value imputation, normalization, denoising, and outlier correction, to obtain more accurate and stable processed feature data, providing reliable input for training the prediction model. For example, in the glass industry chain, this feature set may include changes in the feed rate and load of float glass melting furnaces and their lag response characteristics to the electricity consumption of architectural glass processing enterprises. Based on the processed feature data, the system uses a distributed lag regression method to establish a cross-node electricity consumption prediction model. In this model, the electricity consumption, material input, manufacturing load, and order volume of nodes at the upstream raw material end, midstream manufacturing end, and downstream application end, as well as the output of the lag mapping function, are used as input variables, while the electricity consumption of the target node is used as the output variable. The model continuously optimizes its parameters through iterative training, enabling the prediction results to accurately reflect the energy transmission and time delay effects between different nodes. After training, the initial model yields an optimized prediction model, which can be used to predict the changes in electricity consumption of the target node over different time periods.

[0157] Acquire historical electricity consumption data of the target industrial enterprise on a daily, weekly, and monthly scale, including electricity consumption records from the past three months to one year;

[0158] Historical data is input into the optimized electricity consumption prediction model for prediction, generating nodal electricity consumption prediction results for the next 24 hours, 7 days and 30 days respectively.

[0159] Environmental corrections are performed on the node power consumption forecast results based on external factor data to obtain corrected forecast results.

[0160] Based on the corrected prediction results, comprehensive electricity consumption prediction information is generated.

[0161] Understandably, the system first acquires historical electricity consumption data of the target industrial enterprise at different time periods, including electricity consumption records from the past three months to one year, reflecting the enterprise's actual electricity consumption under different production conditions and external environments. Subsequently, this historical data is input into an optimized electricity consumption forecasting model for prediction, yielding electricity consumption forecasts for the next 24 hours, 7 days, and 30 days. These forecasts can initially reflect the energy consumption trends and load levels of the target enterprise and its upstream and downstream nodes under future production rhythms, but do not consider the potential impact of external environmental factors on electricity consumption. To improve forecast accuracy, the system further incorporates external factors such as regional temperature, energy supply levels, electricity price fluctuations, and industrial policy indicators to perform environmental correction on the forecast results. The environmental correction process includes adjusting the time series of predicted electricity consumption based on external factors, correcting deviations that may arise from temperature changes, policy adjustments, or electricity price fluctuations, thereby obtaining corrected forecast results that are closer to actual operating conditions. Based on this, the system generates comprehensive electricity consumption forecast information, which not only includes the energy consumption level trends of the target industrial enterprise at different time periods but also the electricity load distribution at different production nodes or stages. For example, in the glass industry chain, comprehensive electricity consumption forecast information can show the energy consumption trend of float glass melting furnaces under different feeding and load conditions, and at the same time show the electricity consumption distribution of downstream architectural glass or photovoltaic glass processing enterprises, providing a basis for decision-making in energy dispatch, production planning and energy-saving optimization.

[0162] Furthermore, an intelligent industrial electricity consumption forecasting system based on supply chain profiling is proposed to achieve the intelligent forecasting method described above, characterized by comprising:

[0163] The multi-source operation data acquisition module is used to acquire multi-source operation data covering the upstream raw material end, midstream manufacturing end and downstream application end of the target industrial enterprise.

[0164] The industry chain profiling module is used to organize and structure the collected multi-source operational data, and generate an industry chain profile that reflects the dynamic transmission characteristics between nodes.

[0165] The standardization fusion module is used to fuse the industry chain profile, combining enterprise location information, production capacity level, energy consumption characteristics and regional external factor data to generate standardized industry chain profile data.

[0166] The electricity consumption transmission analysis module is used to establish cross-node electricity consumption transmission relationships based on standardized industrial chain profile data, and generate the first lag mapping function and the second lag mapping function.

[0167] The electricity consumption forecasting module is used to construct an electricity consumption forecasting model and generate comprehensive electricity consumption forecasting information based on the lag mapping function and standardized industry chain profile data.

[0168] Furthermore, the multi-source operational data acquisition module includes:

[0169] The data interface unit is used to receive inventory turnover cycle, order scheduling cycle and historical electricity consumption data of the target industrial enterprise and its upstream and downstream enterprises.

[0170] The data preprocessing unit is used to perform time synchronization, missing data completion, and data type grouping on the collected data.

[0171] The data storage unit is used to store the processed multi-source operational data and provide it to the industry chain profiling module.

[0172] Furthermore, the industry chain profiling module includes:

[0173] The node timing extraction unit is used to extract runtime timing data from upstream raw material end, midstream manufacturing end and downstream application end from structured multi-source runtime data;

[0174] The input-output dependency analysis unit is used to identify the input-output dependencies between runtime sequence data of each node and to establish supply and demand mapping relationships between nodes.

[0175] The industry chain profile generation unit is used to generate industry chain profiles that reflect the dynamic transmission characteristics of the industry chain based on supply and demand mapping relationships and runtime sequence data.

[0176] Furthermore, the electricity consumption prediction module includes:

[0177] The lag mapping generation unit is used to generate a first lag mapping function and a second lag mapping function based on the inventory turnover cycle and the order scheduling cycle.

[0178] The feature extraction and processing unit is used to extract and process cross-node electricity consumption feature data based on standardized industrial chain profile data and hysteresis mapping function;

[0179] The prediction model building unit is used to build an electricity consumption prediction model based on the processed feature data and to make time-series predictions on the historical electricity consumption data of the target industrial enterprise.

[0180] The environmental correction unit is used to correct the prediction results based on external factor data and generate the final comprehensive electricity consumption prediction information.

[0181] In summary, the advantages of this invention are as follows: By constructing a complete industrial chain profile covering upstream raw materials, midstream manufacturing, and downstream applications, and combining inventory turnover cycles and order scheduling cycles to generate a lag mapping function, dynamic modeling and time response analysis of the electricity transmission relationship between nodes in the industrial chain can be achieved. This reveals the correlation patterns of electricity consumption changes in different links of the industrial chain, identifies key factors causing energy consumption fluctuations, and determines whether production rhythm or resource allocation needs to be adjusted to optimize energy utilization efficiency. Simultaneously, through in-depth modeling of lag response characteristics, the accuracy and stability of electricity consumption forecasting can be effectively improved, avoiding energy waste caused by uneven supply and demand or forecasting deviations, and providing reliable data support for energy scheduling and intelligent decision-making in the glass industry.

[0182] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for intelligent prediction of industrial electricity consumption based on supply chain profiling, characterized in that, include: Acquire multi-source operational data covering the upstream raw material end, midstream manufacturing end, and downstream application end of the target industrial enterprise. The multi-source operational data includes, but is not limited to, inventory turnover cycle, order scheduling cycle, material input, output, manufacturing load, order volume, and historical electricity consumption. The multi-source operational data is organized and structured, and the operational sequence data of the upstream raw material end, midstream manufacturing end and downstream application end are extracted to form a picture of the industrial chain; The process involves tracing the industrial chain profile, extracting enterprise location information, production capacity level, and energy consumption characteristic parameters for each node in the industrial chain, and obtaining regional external factor data related to the target industrial enterprise. The energy consumption characteristic parameters include indicators related to unit output energy consumption, equipment load rate, energy utilization efficiency, and peak-valley load ratio. Based on regional external factor data, enterprise location information, production capacity level and energy consumption characteristic parameters, the industrial chain profile is integrated to obtain standardized industrial chain profile data. Based on standardized supply chain profile data, a first electricity consumption transmission relationship is established between the upstream raw material end and the midstream manufacturing end, and a second electricity consumption transmission relationship is established between the midstream manufacturing end and the downstream application end. Based on the inventory turnover cycle and order scheduling cycle in the multi-source operation data, a first lag mapping function is generated for the first electricity consumption transmission relationship between the upstream raw material end and the midstream manufacturing end, and a second lag mapping function is generated for the second electricity consumption transmission relationship between the midstream manufacturing end and the downstream application end. Based on the first and second lag mapping functions, an electricity consumption prediction model is established to predict the multi-source operation data of the target industrial enterprises, and to generate comprehensive electricity consumption prediction information by combining regional external factor data.

2. The intelligent prediction method for industrial electricity consumption based on supply chain profiling as described in claim 1, characterized in that, The process of organizing and structuring multi-source operational data, and extracting operational sequence data from upstream raw material sources, midstream manufacturing, and downstream applications to form a supply chain profile, specifically includes: Extract time stamps, acquisition frequency, and data category information from multi-source operational data; The multi-source running data is aligned and completed according to the time stamp to eliminate time offsets and missing data between different sources, and the multi-source running data is synchronized in time. The multi-source operational data after time synchronization is grouped according to data categories to obtain data grouping results. The data categories include economic data, material data, and energy consumption data. Based on the data grouping results, a structured data table is established, which includes a node identifier field, a time field, and an indicator field. Based on the node identifier field, construct node index sets for the upstream raw material end, the midstream manufacturing end, and the downstream application end; Based on the node index set, runtime sequence data of each node is extracted from multi-source runtime data. The runtime sequence data includes material input change curves, manufacturing load change trends, and order quantity change sequences. Time gradient analysis is performed on the runtime sequence data of each node to obtain the rate of change of input and output between nodes, and the input-output dependency relationship between nodes is determined by correlation matching of the rate of change of different nodes. Based on the input-output dependency relationship, the quantitative coupling coefficient between the node's material input and output is analyzed, and the supply and demand mapping relationship between nodes is established. Based on the supply and demand mapping relationship and runtime sequence data, a portrait of the industrial chain that reflects the dynamic transmission characteristics of the industrial chain is generated.

3. The intelligent prediction method for industrial electricity consumption based on supply chain profiling as described in claim 2, characterized in that, The process involves fusing data on regional external factors, enterprise location information, production capacity levels, and energy consumption characteristics to obtain standardized industrial chain profile data. This includes: Based on the historical electricity consumption and output of each node in the industrial chain, energy consumption characteristic parameters are obtained. These energy consumption characteristic parameters include indicators related to unit output energy consumption, equipment load rate, energy utilization efficiency, and peak-valley load ratio. Acquire regional external factor data corresponding to the target industrial enterprise and its upstream and downstream related enterprises. The external factor data includes regional temperature information, industrial policy indicators, energy supply level and regional electricity price fluctuation information. Based on the enterprise location information, spatial registration of external factor data is performed to determine the regional environmental characteristics of each enterprise node. Based on the enterprise's production capacity level, enterprises within the same industry node are hierarchically clustered to form a hierarchical node set; Based on the energy consumption characteristic parameters, the energy consumption characteristics of the hierarchical node set are processed to obtain the node energy consumption standard parameters; The node energy consumption standard parameters are integrated with the regional environmental characteristics of the corresponding node to generate a comprehensive node feature set; Based on the node comprehensive feature set, the amplitude of the runtime sequence data is corrected using the node energy consumption standard parameters, and the temporal variation trend of the runtime sequence data is adjusted in combination with the regional environmental characteristics to form a fused temporal feature sequence. Based on the fused time-series feature sequences, the industry chain profile is updated to obtain standardized industry chain profile data.

4. The intelligent prediction method for industrial electricity consumption based on supply chain profiling as described in claim 3, characterized in that, Based on standardized supply chain profiling data, a first electricity consumption transmission relationship is established between the upstream raw material end and the midstream manufacturing end, and a second electricity consumption transmission relationship is established between the midstream manufacturing end and the downstream application end. Specifically, this includes: Obtain data on electricity consumption changes at different points in the industrial chain and their corresponding production load, material input, and order volume. Based on the supply and demand mapping relationship between nodes in the standardized industrial chain profile data, the supply and demand response range of upstream raw material nodes to midstream manufacturing nodes, and the load response range of midstream manufacturing nodes to downstream application nodes are determined. Cross-correlation function analysis was performed on the electricity consumption change data in each interval to determine the correlation strength of electricity consumption changes under different time lags and obtain the time dependence matrix. Based on the time dependency matrix, node pairs with a correlation strength greater than a preset threshold are extracted, and a set of power consumption transmission relationships between nodes is established. Based on the transmission relationship set, the node pairs involving the upstream raw material end and the midstream manufacturing end are divided into the first power consumption transmission relationship, and the node pairs involving the midstream manufacturing end and the downstream application end are divided into the second power consumption transmission relationship.

5. The intelligent prediction method for industrial electricity consumption based on supply chain profiling according to claim 4, characterized in that, Based on the inventory turnover cycle and order scheduling cycle in the multi-source operational data, a first lag mapping function is generated for the first electricity consumption transmission relationship between the upstream raw material end and the midstream manufacturing end, and a second lag mapping function is generated for the second electricity consumption transmission relationship between the midstream manufacturing end and the downstream application end. Specifically, this includes: Based on the inventory turnover cycle, the transmission relationship of primary electricity consumption between upstream raw material end and midstream manufacturing end is analyzed, and the time series of inventory-driven electricity consumption response under different inventory levels is extracted. The time series of electricity consumption response driven by inventory is normalized and smoothed to obtain the time correlation coefficient between inventory changes and electricity consumption changes, and to obtain the inventory delay characteristics of upstream nodes on the electricity consumption of midstream nodes under different inventory states. Based on the inventory delay characteristics and time correlation coefficient, an inventory-driven lag response model is established. The inventory-driven lag response model is mapped onto the first electricity consumption transmission relationship to form the first lag mapping function that reflects the impact of changes in electricity consumption at upstream nodes on midstream nodes; Based on the order scheduling cycle, the transmission relationship of the second electricity consumption between the midstream manufacturing end and the downstream application end is analyzed, and the scheduling-driven electricity consumption response time series under different scheduling rhythms is extracted. The time series of power consumption response driven by production scheduling is denoised and standardized to obtain the time response coefficients of production scheduling changes and power consumption changes, and to obtain the scheduling delay characteristics of the power consumption of the downstream node under different production scheduling states. Based on the production delay characteristics and time response coefficient, a production scheduling-driven lag response model is established. The production scheduling-driven lag response model is mapped onto the second electricity consumption transmission relationship to form a second lag mapping function that reflects the impact of changes in electricity consumption at midstream nodes on downstream nodes; The formula for the inventory-driven lag response model is as follows: In the formula, For the midstream node at time Electricity consumption For the upstream node at time The amount of material to be fed, The time correlation coefficient, This represents the maximum lag time between upstream material input and midstream node electricity consumption. This refers to the lag time between upstream material input and midstream electricity consumption. The formula for the production scheduling-driven lag response model is as follows: In the formula, For downstream nodes at time Electricity consumption For the midstream node at time Manufacturing load, The time response coefficient, This represents the maximum lag time between midstream manufacturing load and downstream electricity consumption. The lag time between midstream manufacturing load and downstream electricity consumption.

6. The intelligent prediction method for industrial electricity consumption based on supply chain profiling according to claim 5, characterized in that, The electricity consumption forecasting model, based on the first and second lag mapping functions, is established to predict the multi-source operating data of the target industrial enterprises and, combined with regional external factor data, generates comprehensive electricity consumption forecasting information. Specifically, this includes: Based on standardized supply chain profile data and the first and second lag mapping functions, the time series data of the corresponding nodes are mapped to obtain the lag mapping output, and cross-node electricity consumption feature data is extracted to obtain a feature set that can be used for modeling. The node electricity consumption feature data includes the node electricity consumption, material input, manufacturing load, order volume and lag mapping output related features of the upstream raw material end, midstream manufacturing end and downstream application end. The feature set is processed to obtain the processed feature data; Based on the processed feature data, a distributed lag regression modeling method is adopted. The power consumption, material input, manufacturing load, order volume and lag mapping output of the upstream raw material end, midstream manufacturing end and downstream application end are used as input variables, and the power consumption of the target node is used as the output variable to establish an initial cross-node power consumption prediction model. The initial prediction model is iteratively trained to obtain an optimized electricity consumption prediction model that meets the accuracy requirements; Acquire historical electricity consumption data of the target industrial enterprise on a daily, weekly, and monthly scale, including electricity consumption records from the past three months to one year; Historical data is input into the optimized electricity consumption prediction model for prediction, generating nodal electricity consumption prediction results for the next 24 hours, 7 days and 30 days respectively. Environmental corrections are performed on the node power consumption forecast results based on external factor data to obtain corrected forecast results. Based on the corrected prediction results, comprehensive electricity consumption prediction information is generated.

7. An intelligent industrial electricity consumption forecasting system based on supply chain profiling, used to implement the intelligent forecasting method as described in any one of claims 1-6, characterized in that, include: A multi-source operation data acquisition module is used to acquire multi-source operation data covering the upstream raw material end, midstream manufacturing end and downstream application end of the target industrial enterprise; The industry chain profile building module is used to organize and structure the collected multi-source operation data, and generate an industry chain profile that reflects the dynamic transmission characteristics between nodes. The standardized fusion module is used to fuse the industrial chain profile, combining enterprise location information, production capacity level, energy consumption characteristics and regional external factor data to generate standardized industrial chain profile data. The electricity consumption transmission analysis module is used to establish cross-node electricity consumption transmission relationships based on standardized industrial chain profile data, and generate a first lag mapping function and a second lag mapping function. The electricity consumption forecasting module is used to construct an electricity consumption forecasting model and generate comprehensive electricity consumption forecasting information based on the lag mapping function and standardized industrial chain profile data.

8. The intelligent industrial power consumption forecasting system based on supply chain profiling according to claim 7, characterized in that, The multi-source operational data acquisition module includes: The data interface unit is used to receive inventory turnover cycle, order scheduling cycle and historical electricity consumption data of the target industrial enterprise and its upstream and downstream enterprises. The data preprocessing unit is used to perform time synchronization, missing data completion, and data type grouping processing on the collected data. The data storage unit is used to store the processed multi-source operational data and provide it to the industry chain profiling module.

9. The intelligent industrial power consumption forecasting system based on supply chain profiling according to claim 7, characterized in that, The industry chain profiling module includes: A node timing extraction unit is used to extract runtime timing data from upstream raw material end, midstream manufacturing end and downstream application end from structured multi-source runtime data; An input-output dependency analysis unit is used to identify the input-output dependencies between runtime sequence data of each node and to establish a supply-demand mapping relationship between nodes. The industrial chain profile generation unit is used to generate an industrial chain profile that reflects the dynamic transmission characteristics of the industrial chain based on the supply and demand mapping relationship and runtime sequence data.

10. The intelligent industrial power consumption forecasting system based on supply chain profiling according to claim 7, characterized in that, The electricity consumption prediction module includes: A lag mapping generation unit is used to generate a first lag mapping function and a second lag mapping function based on the inventory turnover cycle and the order scheduling cycle. The feature extraction and processing unit is used to extract and process cross-node electricity consumption feature data based on standardized industrial chain profile data and hysteresis mapping function; A prediction model building unit is used to build an electricity consumption prediction model based on the processed feature data and to make time-series predictions on the historical electricity consumption data of the target industrial enterprise. An environmental correction unit is used to correct the prediction results based on external factor data and generate final comprehensive electricity consumption prediction information.

Images