Energy consumption management and control method and system in multiple scenarios based on digital twinning

Abstract

This invention discloses a method and system for energy consumption management in multiple scenarios based on digital twins. The invention relates to the technical field of digital twins. It identifies multiple key energy consumption nodes based on the identification of various energy consumption paths. A digital twin system for each energy consumption path is constructed based on these key energy consumption nodes, the operating status of energy-consuming equipment, and time points. This determines the current energy consumption level of the energy consumption path, ensuring the accuracy of the digital twin system and improving the accuracy of the current energy consumption level. The corresponding energy allocation mode is determined based on the power grid's energy supply data and energy consumption safety range. The current energy consumption scenario of the enterprise is marked, and an abnormal energy consumption area for the enterprise is determined by combining the energy allocation mode and the real-time load of the power grid. The energy consumption deviation is determined based on the location of the abnormal energy consumption area, the corresponding energy consumption data, and the enterprise's past energy consumption data, improving the accuracy of energy consumption optimization content for the abnormal energy consumption area.

Description

Technical Field

[0001] This invention relates to the technical field of digital twins, and in particular to a method and system for energy consumption management in multiple scenarios based on digital twins. Background Technology

[0002] With the development of technology, enterprises are equipped with multiple energy-consuming devices, including smelting furnaces, transformers, and heating furnaces. For energy consumption management, existing technologies often view each energy-consuming device in isolation or analyze it based solely on static physical connection diagrams. They fail to construct dynamic energy consumption diagrams based on real-time data and overall energy consumption data, and cannot identify and determine the "energy consumption path" and the dynamic direction of energy flow changes along the path in real time. They ignore the digital twin of the energy consumption path, affecting the accuracy of the current energy consumption level of the energy consumption path, resulting in low accuracy of energy consumption optimization content in areas with abnormal energy consumption. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a method and system for energy consumption management in multiple scenarios based on digital twins.

[0004] This invention provides a method for energy consumption management in multiple scenarios based on digital twins, including: Mark each energy-consuming device in the enterprise's distribution map, construct the enterprise's dynamic energy consumption map based on the real-time data of each energy-consuming device, the enterprise's overall energy consumption data and the corresponding distribution map, and determine the energy consumption paths associated with each energy-consuming device; Based on the identification of each energy consumption path, multiple key energy consumption nodes are identified. A digital twin system for the energy consumption path is constructed according to the working status and time nodes of the multiple key energy consumption nodes and energy-consuming equipment to determine the current energy consumption level of the energy consumption path. The energy consumption safety range is determined based on the current energy consumption level of each energy consumption path and the enterprise's past energy consumption data. At the same time, the power grid associated with the enterprise is marked, and the power grid's energy supply data is collected. The corresponding energy allocation mode is determined based on the power grid's energy supply data and the energy consumption safety range. The system marks the current energy consumption scenario of an enterprise and determines the abnormal energy consumption area of ​​the enterprise by combining the energy distribution mode and the real-time load of the power grid. Based on the location of the abnormal energy consumption area, the corresponding energy consumption data and the enterprise's past energy consumption data, the energy consumption deviation is determined, and the energy consumption optimization content of the abnormal energy consumption area is determined by tracing the energy consumption deviation.

[0005] This invention provides a multi-scenario energy consumption management system based on digital twins, which is applied to the aforementioned multi-scenario energy consumption management method based on digital twins; the multi-scenario energy consumption management system based on digital twins includes: The energy consumption path module is used to mark each energy-consuming device in the enterprise's distribution map, construct the enterprise's dynamic energy consumption map based on the real-time data of each energy-consuming device, the enterprise's overall energy consumption data and the corresponding distribution map, and determine the energy consumption path associated with each energy-consuming device. The digital twin module is used to identify multiple key energy consumption nodes based on the identification of each energy consumption path. Based on the multiple key energy consumption nodes, the working status of energy-consuming equipment and time nodes, a digital twin system of the energy consumption path is constructed to determine the current energy consumption level of the energy consumption path. The energy allocation module is used to determine the safe energy consumption range based on the current energy consumption level of each energy consumption path and the enterprise's past energy consumption data. At the same time, it marks the power grid associated with the enterprise, collects the power grid's energy supply data, and determines the corresponding energy allocation mode based on the power grid's energy supply data and the safe energy consumption range. The energy consumption optimization module is used to mark the current energy consumption scenario of an enterprise, and determine the abnormal energy consumption area of ​​the enterprise by combining the energy distribution mode and the real-time load of the power grid. Based on the location of the abnormal energy consumption area, the corresponding energy consumption data and the enterprise's past energy consumption data, the energy consumption deviation is determined, and the energy consumption optimization content of the abnormal energy consumption area is determined by tracing the energy consumption deviation.

[0006] Compared with the prior art, the beneficial effects of the present invention are: (1) Mark each energy-consuming device in the enterprise's distribution map, construct the enterprise's energy consumption dynamic map based on the real-time data of each energy-consuming device, the enterprise's overall energy consumption data and the corresponding distribution map, and determine the energy consumption path associated with each energy-consuming device; determine multiple key energy consumption nodes based on the identification of each energy consumption path, construct the digital twin system of the energy consumption path based on the multiple key energy consumption nodes, the working status of the energy-consuming device and the time node, so as to determine the current energy consumption level of the energy consumption path, introduce multiple key energy consumption nodes, further control the digital twin system of the energy consumption path, ensure the accuracy of the digital twin system, and further improve the accuracy of the current energy consumption level of the energy consumption path.

[0007] (2) Determine the safe energy consumption range based on the current energy consumption level of each energy consumption path and the enterprise's past energy consumption data. At the same time, mark the power grid associated with the enterprise and collect the power grid's energy supply data. Determine the corresponding energy allocation mode based on the power grid's energy supply data and the safe energy consumption range. Mark the enterprise's current energy consumption scenario and determine the enterprise's abnormal energy consumption area by combining the energy allocation mode and the real-time load of the power grid. Determine the energy consumption deviation based on the location of the abnormal energy consumption area, the corresponding energy consumption data, and the enterprise's past energy consumption data. Determine the energy consumption optimization content of the abnormal energy consumption area by tracing the energy consumption deviation. An energy allocation mode is introduced and the abnormal energy consumption area of ​​the enterprise is fully considered. This realizes the control of the energy consumption deviation and improves the accuracy of the energy consumption optimization content of the abnormal energy consumption area. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating the energy consumption management method for multiple scenarios based on digital twins in an embodiment of the present invention. Figure 2 This is a flowchart illustrating step S11 in the energy consumption management method for multiple scenarios based on digital twins in an embodiment of the present invention. Figure 3 This is a flowchart illustrating step S12 in the energy consumption management method for multiple scenarios based on digital twins in an embodiment of the present invention. Figure 4 This is a flowchart illustrating step S13 in the energy consumption management method for multiple scenarios based on digital twins in an embodiment of the present invention. Figure 5 This is a flowchart illustrating step S14 in the energy consumption management method for multiple scenarios based on digital twins in an embodiment of the present invention. Figure 6 This is a schematic diagram of the structural composition of a multi-scenario energy consumption management system based on digital twins in an embodiment of the present invention. Detailed Implementation

[0009] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0010] Please see Figures 1 to 6 A method for energy consumption management in multiple scenarios based on digital twins is proposed and applied to digital twin scenarios. The method includes: Step S11: Mark each energy-consuming device in the enterprise's distribution map, construct the enterprise's dynamic energy consumption map based on the real-time data of each energy-consuming device, the enterprise's overall energy consumption data and the corresponding distribution map, and determine the energy consumption path associated with each energy-consuming device. Step S12: Based on the identification of each energy consumption path, identify multiple key energy consumption nodes, and construct a digital twin system for the energy consumption path according to the working status and time nodes of the multiple key energy consumption nodes and energy consumption equipment to determine the current energy consumption level of the energy consumption path. Step S13: Determine the energy consumption safety range based on the current energy consumption level of each energy consumption path and the enterprise's past energy consumption data. At the same time, mark the power grid associated with the enterprise and collect the power grid's energy supply data. Determine the corresponding energy allocation mode based on the power grid's energy supply data and the energy consumption safety range. Step S14: Mark the current energy consumption scenario of the enterprise, and determine the abnormal energy consumption area of ​​the enterprise by combining the energy distribution mode and the real-time load of the power grid. Determine the energy consumption deviation based on the location of the abnormal energy consumption area, the corresponding energy consumption data and the enterprise's past energy consumption data, and determine the energy consumption optimization content of the abnormal energy consumption area by tracing the energy consumption deviation.

[0011] refer to Figure 2 In step S11, the specific steps are as follows: S111: Collect the name of the enterprise and determine the distribution map of the enterprise based on the traceability of the enterprise name. The distribution map of the enterprise covers building information and geographical information, and marks each energy-consuming device; determine the real-time data of each energy-consuming device based on the real-time monitoring of each energy-consuming device. S112: In the enterprise's distribution diagram, each energy-consuming device is abstracted as a device node and edge in the topology. Device nodes are mapped to energy-consuming devices at each level, and edges represent the physical medium connection of energy flow. Real-time data of each energy-consuming device is mapped to the corresponding device node, and the transient changes of each device node are captured in real time. Based on the transient changes and the overall energy consumption data of the enterprise, the enterprise's energy consumption dynamic diagram is constructed. Based on the identification of the energy consumption dynamic diagram, the direction of energy consumption change is determined, and the energy consumption path associated with each energy-consuming device is determined by tracing along each direction of energy consumption change.

[0012] In the embodiments of this application, the name of the enterprise is collected, and the distribution map of the enterprise is determined by tracing the enterprise name. The distribution map of the enterprise covers building information and geographical information, and marks each energy-consuming device. The real-time data of each energy-consuming device is determined based on the real-time monitoring of each energy-consuming device, which is compatible with the overall consideration of real-time monitoring of each energy-consuming device and ensures the accuracy of the real-time data of each energy-consuming device.

[0013] At this point, the system receives the company name as a unique search key and performs reverse tracing in the background database (such as ERP or asset management system) through the Enterprise Service Bus (ESB) or unified identity authentication system. This process is not only to obtain the name, but also to verify the company's unique identifier (UID), thereby calling the associated geographic information database and engineering drawing database.

[0014] The system will automatically integrate Building Information Modeling (BIM) and Geographic Information System (GIS) data; BIM data provides micro-level building information such as the internal structure, floor height, load-bearing capacity, and pipeline routing of the factory; while GIS data provides macro-level geographic information such as the absolute location of the factory in the Earth coordinate system, altitude, and surrounding topography.

[0015] On the merged digital base map, the system uses computer vision technology or manual annotation interface to perform semantic recognition and marking of all physical energy consumption entities. This is not just about marking points, but about giving each device a digital object containing spatial coordinates (XYZ), asset number and device type, ensuring that physical devices have a unique "identity ID" in digital space.

[0016] For energy-consuming equipment from different eras and manufacturers, the system supports multiple industrial communication protocols (such as Modbus TCP, OPCUA, MQTT, IEC104, etc.) through the Industrial Internet of Things Gateway (IIoTGateway); at the edge, the collected raw signals are preprocessed (such as filtering, noise reduction, and aggregation) to reduce network bandwidth pressure and ensure data quality.

[0017] The collected data is not limited to basic electrical parameters (voltage, current, power), but also extends to process parameters (pressure, flow rate, speed, temperature) and environmental parameters (ambient temperature, humidity). At the same time, the system adopts a high-precision clock synchronization protocol (such as PTP or NTP) to ensure that the data collected from different devices are strictly aligned on the time axis, avoiding logical deviations in energy consumption analysis due to timing differences.

[0018] Specifically, after the system inputs "enterprise", it automatically traces and loads the digital model of its main factory located in the high-tech park; in the distribution map, the building information of the "first joint factory building" can be clearly seen: 200 meters long and 80 meters wide, including a steel structure roof and an underground pipe gallery layer; at the same time, combined with GIS information, it accurately shows that the factory building is located in the southeast corner of the park, and the elevation is +45.5 meters (geographic information).

[0019] Based on this, the system meticulously marked the equipment in the workshop: at coordinates (X:15.5, Y:30.2, Z:0.0), it was marked as "No. 1 high-energy-consuming smelting furnace"; at coordinates (X:15.6, Y:30.2, Z:8.5), it was marked as "furnace exhaust gas treatment fan"; in the power distribution room area, "10kV high-voltage incoming line cabinet" and "low-voltage capacitor compensation cabinet" were marked. These marked points formed a physical node network for enterprise energy consumption management in the distribution map, providing spatial anchor points for subsequent dynamic topology connections.

[0020] For the previously marked "No. 1 high-energy-consuming smelting furnace", the system establishes a connection with its PLC controller through the OPCUA protocol and captures the following data in real time: Electrical parameters: A-phase current fluctuates between 850A and 860A, and the real-time value of active power is 480kW; Process parameters: The current temperature inside the furnace is 1180℃, and the cooling water flow rate is 120L / min; Status parameters: The equipment operation status word is "Running".

[0021] Meanwhile, for the "furnace exhaust gas treatment fan", the system collected data showing that the frequency of its inverter was 45Hz and the instantaneous power was 22kW. All of this data was encapsulated into data packets with nanosecond-level timestamps and mapped in real time to the corresponding device nodes on the enterprise's digital twin distribution map. At this time, on the enterprise's digital monitoring screen, engineers saw not only static device icons, but also real-time power flow, temperature curves and operating status jumping next to each icon.

[0022] Furthermore, in the enterprise's distribution diagram, each energy-consuming device is abstracted as a device node and edge in the topology. Device nodes are mapped to energy-consuming devices at each level, and edges represent the physical medium connecting energy flow. Real-time data of each energy-consuming device is mapped to the corresponding device node, and the transient changes of each device node are captured in real time. Based on these transient changes and the enterprise's overall energy consumption data, an energy consumption dynamic diagram of the enterprise is constructed. Based on the identification of this energy consumption dynamic diagram, the direction of energy consumption change is determined, and the energy consumption path associated with each energy-consuming device is determined by tracing along each direction of energy consumption change. This approach incorporates the overall considerations for identifying the energy consumption dynamic diagram and ensures the accuracy of the direction of energy consumption change.

[0023] At this point, the energy-consuming equipment at each level marked in S111 (such as transformers, motors, heating furnaces, and lighting groups) are abstracted as "nodes" in the topology; each node is given multi-dimensional attributes, including equipment type, rated capacity, energy efficiency level, and its hierarchical relationship in space (such as which workshop or production line it belongs to).

[0024] The physical media connecting various devices (cables, busbars, steam pipes, compressed air pipelines) are abstracted as "edges" of the connection nodes. "Edges" not only represent the connection relationship, but also carry physical attributes, such as impedance (affecting power transmission loss), pipe diameter (limiting fluid flow), and the unidirectional or bidirectional flow characteristics of the transmission medium. Through this abstraction, the complex physical facilities of an enterprise are reconstructed into a directed weighted graph composed of nodes and edges to provide the underlying data structure.

[0025] The real-time data stream (voltage, current, power, temperature, etc.) collected by S111 is injected into the corresponding topology nodes in real time; the node attributes change from static "rated values" to dynamic "real-time values"; the system does not only collect numerical values, but also monitors the transient rate of change of data in real time through sliding window method or edge computing nodes, that is, it captures differential characteristics such as power surge, voltage drop, frequency fluctuation, etc. These transient characteristics often reflect the start-up and shutdown of equipment, operating condition switching or fault precursors.

[0026] By combining node-level transient changes with the enterprise's overall energy consumption data (such as the total incoming power factor), the system updates the state of the entire topology graph in real time. For example, the weight of the edge can be dynamically changed based on the heat loss calculated from the real-time current, and the color of the node can be rendered in real time based on the load rate, thus forming a "dynamic energy consumption graph" that evolves over time.

[0027] Based on the constructed energy consumption dynamic graph, the system uses directed graph analysis to identify the propagation direction of transient energy changes. For example, is it the upstream grid voltage fluctuation that causes the downstream equipment malfunction, or is it the downstream equipment startup that causes the grid voltage drop? The system performs a depth-first search (DFS) or breadth-first search (BFS) along the identified energy flow direction. By backtracking, it traces backward from the end device node where the transient change occurs to the energy source node, passing through all intermediate connecting edges and nodes, thereby locking in a complete "energy consumption path" with strong causal correlation.

[0028] Specifically, in the enterprise's digital twin system, the physical facilities within the first joint plant were topologically processed: Nodes: "High-energy-consuming smelting furnace No. 1" was abstracted as node N-01, "10kV transformer No. 2" was abstracted as node T-02, and "low-voltage distribution cabinet A3" was abstracted as node P-03; Edges: The copper busbar connecting transformer T-02 and distribution cabinet P-03 was abstracted as edge E-01, and assigned physical attributes such as rated current carrying capacity (3000A) and resistance value (0.0005Ω); The power cable connecting distribution cabinet P-03 and smelting furnace N-01 was abstracted as edge E-02. At this point, the enterprise's energy consumption management interface no longer only displays a three-dimensional model, but also shows the logical topology links connecting these core devices through a perspective view, providing a clear framework for analyzing energy flow.

[0029] The system maps the real-time power data of the enterprise's "No. 1 high-energy-consuming smelting furnace" (node ​​N-01) (currently surging from 480kW to 550kW) onto the topology map; at the same time, the system captures the transient changes of this node: the power increased by 14% within 2 seconds, accompanied by a slight distortion of the current waveform.

[0030] Based on the enterprise's overall energy consumption data at this time (it was found that the total incoming load also generated a peak in the same direction), the digital twin platform immediately updated the energy consumption dynamic graph: the color of node N-01 and node T-02 changed from green (normal) to yellow (high load warning), and the line connecting them, edge E-02, was thickened, intuitively showing the intensity of energy flow on this path at this moment.

[0031] Based on the sudden increase in power at node N-01 in the enterprise's energy consumption dynamic diagram, the system identifies the current energy consumption change direction as an inrush current flowing from the upstream power supply side to the smelting furnace load side. The system immediately initiates a path tracing mechanism: Starting point: taking the currently abnormally active node N-01 (smelting furnace No. 1) as the starting point; Backtracking: along the reverse direction of energy flow, sequentially passing through side E-02 (power cable) > node P-03 (low-voltage distribution cabinet A3) > side E-01 (bus trunking) > node T-02 (transformer No. 2); End point: finally tracing back to the 10kV high-voltage side incoming line point. Through this process, the system determines a critical energy consumption path: 10kV grid side > transformer No. 2 > low-voltage distribution cabinet A3 > smelting furnace No. 1. Based on this, the system judges that the current surge in energy consumption along this path is caused by the change in the operating conditions of the end smelting furnace, and no other branch anomalies are found. This clear path provides a precise logical object for targeted energy consumption level assessment and safety range definition in subsequent step S13.

[0032] refer to Figure 3 In step S12, the specific steps are as follows: S121: Input each energy consumption path into the deep learning network and perform deep learning in the deep learning network to determine multiple primary energy consumption nodes. Determine the energy consumption priority of multiple primary energy consumption nodes based on the node position and corresponding energy consumption data combination of multiple primary energy consumption nodes, and combine the overall energy consumption of the enterprise to determine the node screening mechanism to screen out multiple key energy consumption nodes. S122: Collect the working status of energy-consuming devices, and construct corresponding digital twin frameworks based on the working status of multiple key energy-consuming nodes and energy-consuming devices. Perform multi-factor fusion on the environmental data and time nodes of the digital twin framework and energy-consuming devices, and perform multi-scenario simulations in the fused state to output the digital twin system of the energy consumption path. Determine the current energy consumption level of the energy consumption path through the real-time identification of the digital twin system. The current energy consumption level of the energy consumption path covers the high-efficiency level, the economic level, and the early warning level.

[0033] In the embodiments of this application, each energy consumption path is input into a deep learning network, and deep learning is performed in the deep learning network to determine multiple primary energy consumption nodes. The energy consumption priority of multiple primary energy consumption nodes is determined based on the node positions of multiple primary energy consumption nodes and the corresponding energy consumption data combinations. Combined with the enterprise's overall energy consumption to determine the node screening mechanism, multiple key energy consumption nodes are screened out. This approach takes into account the overall consideration of the node positions of multiple primary energy consumption nodes and the corresponding energy consumption data combinations, ensuring the accuracy of the energy consumption priority of multiple primary energy consumption nodes.

[0034] At this point, the sequence of device nodes, historical operating data streams, and connections between nodes involved in the energy consumption path are transformed into high-dimensional feature vectors and input into the network. The network learns the behavior patterns of devices in complex energy consumption networks through training, automatically identifies those nodes that play a key role in the energy transmission or conversion process, and filters out micro nodes that only play a switching role or consume very little energy, so as to output a set of "primary energy consumption nodes". These nodes are potential objects of interest, but have not yet been finally prioritized.

[0035] The analysis focuses on the centrality and betweenness of nodes in the topology graph. Nodes located in the backbone or aggregation layer (such as transformers) have higher weights, while those at the edge have relatively lower weights. Based on real-time data, the load factor, power factor, and energy consumption fluctuation of nodes are calculated. Nodes with large power bases and significant fluctuations are assigned higher priority scores. Simultaneously, linear weighting or the Analytic Hierarchy Process (AHP) is used to integrate location weights with data weights to generate an "energy consumption priority" value for each primary energy consumption node.

[0036] Obtain the company's current overall energy consumption data (such as total active power and peak load); set an adaptive threshold mechanism; for example, use the Pareto principle (80 / 20 rule) to screen out the set of nodes whose cumulative energy consumption accounts for 80%; or lower the screening threshold according to the company's current load pressure (such as during the "peak summer" period) to include more potential risk nodes in the monitoring scope; if the company's overall load is low, the screening mechanism will be tightened, focusing only on core large users; if the overall load is close to the safe limit, the screening mechanism will be relaxed.

[0037] The priority values ​​of primary energy consumption nodes are compared with the thresholds in the screening mechanism; nodes that meet the screening criteria are officially identified as "critical energy consumption nodes". These selected nodes will be used as core objects and passed to the subsequent step S122 to build a high-fidelity digital twin system and achieve refined management.

[0038] Specifically, in the enterprise's system, the critical path determined by S112—10kV grid side > Transformer No. 2 > Low-voltage distribution cabinet A3 > Smelting furnace No. 1—is input into a pre-trained graph convolutional neural network (GCN). After forward propagation analysis, the network automatically identifies the following in this path: Transformer No. 2: Involves high-voltage transformation and is the core of energy conversion, so it is marked as a primary node; Low-voltage distribution cabinet A3: Involves power distribution and is marked as a primary node; Smelting furnace No. 1: As a terminal high-power load, it is marked as a primary node. However, some non-critical small detection instruments or auxiliary relays in the path are automatically filtered by the network due to their low weight and are not included in the list of primary nodes.

[0039] For the primary nodes of the enterprise, the system performs the following calculations: For “Transformer No. 2”: its location is at the source side, and its location weight score is 0.9; its real-time load rate is 75%, and its data weight score is 0.8; after comprehensive calculation, its energy consumption priority is rated as “high (0.85)”; For “Smelting Furnace No. 1”: its location is at the end, and its location weight score is 0.6; however, its real-time power is as high as 550kW and it is in a high-frequency fluctuation state, and its data weight score is 0.95; after comprehensive calculation, its energy consumption priority is rated as “extremely high (0.88)”; For “Low-voltage Distribution Cabinet A3”: its location and data performance are both medium, and its priority is rated as “medium (0.6)”; optional, the above data are preset data and are for reference only.

[0040] The system detected that the company's overall energy consumption was operating at a high level (total load reached 8500kW, close to the warning value). Based on this background, the node screening mechanism automatically adjusted to "strict monitoring mode". This mechanism sets dual screening conditions: the node priority value must be greater than 0.5; the instantaneous power of the node must account for more than 2% of the company's total power. This mechanism ensures that under the current high load operation, no node that causes system fluctuations will be missed.

[0041] Using the above screening mechanism, the system finally identified two key energy-consuming nodes from the primary nodes: Key Node 1: No. 1 high-energy-consuming smelting furnace; Reason: Priority 0.88 (extremely high), and large power proportion, it is the main source of the current high load; Key Node 2: No. 2 transformer; Reason: Priority 0.85 (high), it is the energy hub of the entire region, and its stability is directly related to the power supply security of the enterprise.

[0042] Although "low-voltage distribution cabinet A3" has a priority of 0.6, it has not been included in the core node (or monitored as an auxiliary node) of the twin system in this round of S12 construction because it has not reached the more stringent cumulative proportion threshold or specific risk indicators under the high load screening criteria. This has achieved the optimized allocation of computing resources.

[0043] Furthermore, the working status of energy-consuming devices is collected, and a corresponding digital twin framework is constructed based on the working status of multiple key energy-consuming nodes and devices. Multi-factor fusion is performed on the environmental data and time nodes of the digital twin framework and energy-consuming devices, and multi-scenario simulations are conducted in the fused state to output a digital twin system for the energy consumption path. The current energy consumption level of the energy consumption path is determined through real-time identification of this digital twin system. The current energy consumption level of the energy consumption path covers high-efficiency, economic, and early warning levels, incorporating a holistic consideration of the working status of multiple key energy-consuming nodes and devices, ensuring the accuracy of the corresponding digital twin framework. Simultaneously, multiple key energy-consuming nodes are introduced to further control the digital twin system of the energy consumption path, ensuring the accuracy of the digital twin system and further improving the accuracy of the current energy consumption level of the energy consumption path.

[0044] At this time, the system collects the working status of the key energy-consuming equipment screened by S121 at high frequency, including not only electrical parameters (voltage, current, power factor), but also non-electrical parameters (equipment vibration spectrum, speed, pressure, medium flow) and control command status (such as PID set value, valve opening).

[0045] Based on a combination of physical mechanism modeling (such as Maxwell's equations and fluid dynamics equations) and data-driven modeling, a high-fidelity twin framework is constructed for key energy consumption nodes. This framework includes the device's geometric model, physical property model (such as resistance, inductance, and moment of inertia), and behavioral logic model; it defines the transfer function relationship between inputs (voltage, control signals) and outputs (power, speed, and losses).

[0046] Data such as temperature, humidity, altitude, and air pressure of the equipment's environment are mapped into the twin framework. For example, an increase in ambient temperature directly leads to poor heat dissipation in the transformer, thereby changing its internal thermal resistance parameters and affecting the efficiency of the cooling system. Absolute time (year, month, day, hour) and relative time (equipment running time, production cycle phase) are introduced. Time nodes are not only used to mark data but also to adjust model parameters, such as simulating the degree of equipment aging (the longer the running time, the greater the friction coefficient) or responding to time-of-use electricity pricing policies. Using a multi-dimensional feature fusion network, these external factors are used as correction coefficients to adjust the boundary conditions in the twin framework in real time, transforming the model from "ideal operating conditions" to "current actual operating conditions".

[0047] With the above factors integrated, the system runs multiple simulation scenarios in parallel, such as normal operation simulation, fault simulation, and energy efficiency boundary simulation. The solver (such as a finite element analysis (FEM) or finite volume method (FVM) solver) calculates the current physical field distribution (electric field, magnetic field, temperature field) in real time and outputs a complete digital twin system of the energy consumption path. This includes not only real-time data but also derived indicators calculated by simulation, such as real-time energy efficiency ratio (COP), theoretical optimal energy consumption value, expected temperature rise trend of the equipment, and prediction of remaining service life. This system can reflect the real physical behavior of the equipment under the current environment and operating conditions.

[0048] The system compares the real-time operating indicators (such as real-time energy efficiency, load rate, and safety margin) output by the digital twin system with preset benchmark thresholds. The level determination logic is as follows: High efficiency level: The equipment operates in the optimal operating range, the energy efficiency ratio is in the theoretical optimal range, and the loss is extremely low; Economic level: The equipment operates in a non-optimal operating range but within an acceptable range, or although the energy efficiency is not at its highest, the economic benefits are balanced when considering electricity price and output; Warning level: The equipment operating parameters exceed the safety boundary, or the energy efficiency is extremely low (there is huge waste), or the equipment health indicators deteriorate, and there is a risk of failure.

[0049] Specifically, for the enterprise's key nodes, "No. 1 high-energy-consuming smelting furnace" and "No. 2 transformer," the system collected their current operating status: the smelting furnace induction coil current is 1200A, and the transformer tap is at the 3rd tap. Based on this, the system constructed a twin framework containing the following: Smelting furnace model: including an electromagnetic induction circuit model and a heat loss calculation model; Transformer model: including an iron core hysteresis loss model and a coil copper loss model; Coupling model: connecting the two and defining the matching relationship between the transformer output impedance and the smelting furnace input impedance. This framework is the "virtual prototype" of this energy consumption path for the enterprise in the digital world.

[0050] The system injects real-time environmental data of the enterprise into the twin framework: Environmental fusion: The current workshop ambient temperature is 38℃ (25℃ higher than the standard). The system automatically increases the thermal resistance parameter in the twin model of "Transformer No. 2" to simulate its reduced heat dissipation capacity at high temperatures. At the same time, the inlet water temperature of the cooling tower is corrected to 30℃. Time fusion: The current time is 14:30 (peak summer electricity consumption), and the smelting furnace has been running continuously for 4 hours. According to the time node, the system increases the thermal accumulation effect coefficient of the winding in the model and takes into account the impact of slight fluctuations in grid voltage (-2%). At this time, the twin framework is a dynamic operating entity full of high temperature and thermal accumulation effect.

[0051] After integrating environmental and temporal data, the enterprise's digital twin system initiated multi-scenario simulations: the system calculated the hot spot temperature distribution of "Transformer No. 2" at an ambient temperature of 38℃, and the results showed that its top oil temperature was 12℃ higher than that under standard environmental conditions, approaching the alarm threshold; the system calculated the energy conversion efficiency of the "smelting furnace" and found that due to the increase in the cooling water inlet temperature, the heat dissipation efficiency decreased, resulting in an increase of 3.5% in active power loss; the system output a visualized digital twin energy consumption path system with temperature field cloud maps, which intuitively showed the energy transmission efficiency and thermal risk distribution from the transformer to the smelting furnace.

[0052] The twin data shows that although "Smelting Furnace No. 1" is operating normally, the copper loss of "Transformer No. 2" has increased sharply due to overload operation in a high-temperature environment, causing the overall energy efficiency ratio (COP) of this path to drop to 0.82 (the baseline is 0.92). The transformer hot spot temperature is predicted to exceed 90°C in 15 minutes (the safety limit is 95°C). Although no tripping fault has occurred yet, the high loss and high temperature risk have seriously affected safety and economy. The system has determined the current energy consumption level of this energy consumption path to be "warning level". This reminds the company's operation and maintenance personnel that they must take cooling measures or adjust the production load immediately, otherwise they will face the risk of equipment damage or high energy consumption penalties.

[0053] refer to Figure 4 In step S13, the specific steps are as follows: S131: Based on the detection of the enterprise's database, determine the enterprise's past energy consumption data, perform correlation analysis between the current energy consumption level of each energy consumption path and the enterprise's past energy consumption data, and combine deep clustering to determine the distribution information of the enterprise's optimal energy consumption range under different production conditions. Based on the distribution information of the optimal energy consumption range and the matching of the time dimension, determine the energy consumption fluctuation range of the energy consumption path, and combine the enterprise's preset safe energy consumption level to determine the corresponding safe energy consumption range. S132: Collect data from the power grid associated with the enterprise, and collect multiple energy supply data from the power grid in all aspects, including power grid frequency data, voltage deviation data, and real-time electricity price data; unify the energy supply data of the power grid side with the enterprise's energy consumption safety range into a unified model, and perform multi-objective solution to determine the corresponding multiple energy allocation contents, and determine the corresponding energy allocation mode based on the multiple fusion of multiple energy allocation contents.

[0054] In the embodiments of this application, the company's past energy consumption data is determined based on the detection of the company's database. The current energy consumption level of each energy consumption path is correlated with the company's past energy consumption data. Deep clustering is used to determine the distribution information of the company's optimal energy consumption range under different production conditions. The energy consumption fluctuation range of the energy consumption path is determined based on the matching of the distribution information of the optimal energy consumption range and the time dimension. The corresponding safe energy consumption range is determined in combination with the company's preset safe energy consumption level. This approach takes into account the overall consideration of the matching of the distribution information of the optimal energy consumption range and the time dimension, ensuring the accuracy of the energy consumption fluctuation range of the energy consumption path.

[0055] At this point, the system performs a full scan of the enterprise's historical database, executes ETL (Extract, Transform, Load) operations, removes outliers caused by sensor malfunctions through noise reduction, and supplements missing data fragments through interpolation to ensure the integrity and accuracy of historical energy consumption data.

[0056] Data from different sources (such as production logs, energy consumption records, and environmental monitoring) are strictly aligned based on a unified timestamp to construct a multi-dimensional time-series database. Statistical features (such as mean, peak value, and variance) and time-domain features (such as ramp rate and fluctuation frequency) are extracted from the raw data to form the company's historical energy consumption feature library.

[0057] The "current energy consumption level" (e.g., "warning level") output by S122 and its corresponding real-time operating feature vector (e.g., load rate, temperature, power) are matched with similar features in the historical database using similarity matching (e.g., cosine similarity calculation). Using association rule methods (e.g., Apriori method), the potential relationship between the current energy consumption level and specific historical events (e.g., raw material changes, environmental changes, equipment aging) is analyzed. The deviation between the current real-time data and the historical data of the same period and operating conditions is calculated to determine whether the current "warning" is a normal process fluctuation or an abnormal energy efficiency degradation.

[0058] Deep embedded clustering (DEC) or density-based clustering (DBSCAN) is used to perform cluster analysis on historical multidimensional energy consumption data; the massive data is divided into several typical production condition clusters (such as: light load standby, ramp-up start, full load stable production, and overload sprint).

[0059] Within each cluster, a subset of data with the best energy efficiency indicators (such as the lowest unit consumption and the highest power factor) is selected; the statistical distribution (mean and confidence interval) of these subsets is calculated to determine the "optimal energy consumption range" under this operating condition, which represents the energy consumption level that the equipment should achieve under ideal conditions.

[0060] The system identifies the attributes of the current time node (season, day of the week, time period) and searches for the energy consumption fluctuation patterns of the same period in history. For example, the grid voltage fluctuates greatly in summer, or there are process differences between night shift and day shift. Based on the optimal energy consumption range, the system superimposes the inertial fluctuation factor of the same period in history (such as the allowable ±3% process fluctuation), and predicts the reasonable energy consumption fluctuation range of the energy consumption path at the current moment through time series analysis (such as ARIMA model). This is a dynamic frequency band that includes upper and lower limits.

[0061] The system retrieves the enterprise's preset "safe energy consumption level" parameters (such as the rated capacity of transformers, the thermal stability limit of cables, and the upper limit of grid demand assessment), which are hard constraints; it takes the intersection of the above "energy consumption fluctuation range" and "preset safety threshold"; if the lower limit of the fluctuation range is lower than the minimum safe operating requirements (such as the minimum load to prevent boiler shutdown), the lower limit is raised; if the upper limit of the fluctuation range exceeds the equipment safety red line (such as the transformer overload threshold), the upper limit is forcibly lowered; and generates an "energy consumption safety range" that meets both economic requirements (close to the optimal range) and absolutely satisfies safety requirements.

[0062] Specifically, for the enterprise's "Transformer No. 2 > Smelting Furnace No. 1" path, the system called up the historical database of the same period over the past three years; the detection found that some data was missing in the summer of 2023; the system automatically used the Lagrange interpolation method to repair the missing power values; and determined the complete energy consumption data of this path in previous production cycles, including the daily start and stop times, the power curve during peak production periods, and the base load during standby, and constructed a historical energy consumption benchmark map of this path.

[0063] S122 determines that the path is currently in the "warning level"; the system performs correlation analysis between the current operating conditions (ambient temperature 38℃, load 850kW) and the historical database; the analysis found that although there were also high-temperature periods in the historical data, under the same ambient temperature and the same production cycle, the historical average load was only 810kW, and the energy consumption level was mostly "economic level"; the deviation calculated by the system was +4.9%, which indicates that although the company's current energy consumption status has not yet tripped, it has deviated significantly from the historical normal level, and there is a strong correlation with abnormal energy efficiency.

[0064] The system performed deep clustering on millions of historical data points of the enterprise's smelting furnace, identifying three main operating conditions: "furnace preheating", "feeding and melting", and "heat holding and refining". For the current "feeding and melting" operating condition, the system extracted the top 10% of historical energy efficiency data segments from the clusters. The calculation results show that under this operating condition, the enterprise's optimal energy consumption range should be between 780kW and 820kW. This means that if the system is operating well, the current power should be controlled within this range.

[0065] The current time is "14:30 in the summer afternoon," which is the time of day with the highest temperature and the most unstable grid voltage. Based on historical data, the system found that during this period, due to the impact of reduced cooling efficiency and voltage drop, the power of the smelting furnace is usually allowed to fluctuate positively. Therefore, the system corrected the optimal range of 780kW-820kW for the time dimension and determined that the energy consumption fluctuation range at the current moment is 775kW to 840kW, which gives the equipment a reasonable tolerance for environmental factors.

[0066] The system performs a final verification of the enterprise's parameters: fluctuation range: 775kW-840kW; preset safety constraints: the rated capacity of transformer No. 2 is 1000kW, but the enterprise's "safe energy consumption level" is set as 90% of the rated capacity as the alarm line (i.e. 900kW), and in order to prevent long-term overheating and aging, the recommended operating limit is 860kW; after fusion calculation, the upper limit of the fluctuation range of 840kW does not exceed the recommended upper limit of 860kW, so it is retained; the system determines that the energy consumption path has a safe energy consumption range of [775kW, 840kW] at the current moment.

[0067] Furthermore, data is collected from the power grid associated with the enterprise, and multiple energy supply data from the power grid are collected comprehensively, including power grid frequency data, voltage deviation data, and real-time electricity price data. The energy supply data from the power grid side is modeled in a unified manner with the enterprise's energy consumption safety range, and multi-objective solutions are performed to determine the corresponding multiple energy allocation contents. Based on the multiple fusions of multiple energy allocation contents, the corresponding energy allocation mode is determined, which is compatible with the overall consideration of the multiple fusions of multiple energy allocation contents and ensures the accuracy of the corresponding energy allocation mode.

[0068] At this point, the system identifies and connects to the monitoring point of the upstream substation supplying power to the enterprise through a communication interface (such as IEC61850 or ModbusTCP), introducing grid frequency data, voltage deviation data, and real-time electricity price data. Grid frequency data: Real-time monitoring of the system frequency (e.g., 50Hz) reflects the supply and demand balance of the grid. A large frequency deviation means poor grid stability, requiring the enterprise to provide primary frequency regulation services or reduce load. Voltage deviation data: Real-time monitoring of the voltage amplitude at the grid connection point. Voltage deviation (e.g., within ±10%) directly affects equipment operating efficiency; excessively high or low voltage will lead to increased energy consumption or even equipment damage. Real-time electricity price data: Accessing the electricity market trading interface to obtain real-time electricity price (RTP) or time-of-use (TOU) signals, which are the core variables for calculating economic costs.

[0069] By integrating external and internal constraints, a mathematical model for a multivariate constrained optimization problem is constructed. The decision variables of the model include: the power Pgrid obtained from the grid, the discharge power Pess of the enterprise's internal energy storage system, and the output Pgen of the self-provided generator set. Internal constraint: the total load of the enterprise, Pload = Pgrid + Pess + Pgen, must fall within the energy consumption safety range [775kW, 840kW] determined by S131. External constraint: Pgrid must meet the maximum power demand limit of the power grid company. The voltage deviation range limits the reactive power adjustment space of the equipment, and a multi-vector objective function is established that includes the lowest operating cost (electricity expenditure), the best power quality (minimum voltage deviation), and the smoothest grid interaction (avoiding power surges).

[0070] The solution employs a particle swarm optimization (PSO), genetic algorithm (GA), or mixed integer linear programming (MILP) solver to find a Pareto optimal solution between conflicting objectives (such as cost saving vs. voltage stabilization). The solver outputs a series of discrete control commands, i.e., "energy allocation content," which includes: how much power needs to be absorbed from the grid, how much energy storage system needs to discharge, and how much reactive power support needs to be provided by the reactive power compensation equipment (SVG / SVC).

[0071] The specific allocation contents (energy storage discharge, grid power limiting, reactive power regulation) obtained from the above solution are logically integrated and packaged; based on the characteristics of the integrated strategy, the current "energy allocation mode" is defined. This mode is not only a set of instructions, but also a label of the current operating status of the system, which is convenient for dispatchers to understand and for system log recording; common mode types include "peak shaving and valley filling mode", "economic optimization mode", "voltage protection mode", "island operation mode", etc.

[0072] Specifically, the company's energy management system is connected to the data interface of the regional substation via a dedicated network. The key energy supply data collected at the moment are as follows: grid frequency: 49.95Hz (slightly lower than the standard value, indicating a heavy grid load); voltage deviation: -5.5% (bus voltage is slightly low, at the lower edge of the allowable range); real-time electricity price: 1.35 yuan / kWh (during the summer peak electricity price period). These data indicate that the current external power grid is in a state of "both expensive and unstable", which is an important basis for subsequent strategy formulation.

[0073] The system integrates collected external data with the enterprise's internal status into a unified model; Load constraints: The model mandates that the total load must be controlled between [775kW, 840kW] to eliminate the previous "warning" state; Voltage constraints: Due to the current voltage deviation of -5.5%, the model limits the reactive power extracted from the grid to prevent further voltage collapse; Cost objectives: Substituting the electricity price coefficient of 1.35 yuan / kWh into the cost function, the growth of Pgrid is suppressed with a large weight. This model forms a complex mathematical programming problem, guiding how to rationalize electricity use while meeting safety red lines.

[0074] In response to the company's current overload of 850kW, after multiple iterations, the following energy allocation was proposed: Energy storage action: The company's industrial and commercial energy storage system was activated to discharge, with an output set at 60kW; Grid absorption: The power drawn from the grid was reduced and limited to 790kW; Reactive power support: The SVC static reactive power compensation device was put into operation, outputting 50kvar of inductive reactive power to support the voltage level deteriorated due to the reduction of grid active power; Verification: 790kW (grid) + 60kW (energy storage) = 850kW (total load), which meets production needs; and 790kW falls within the safe range [775kW, 840kW], while significantly reducing high electricity costs.

[0075] The system integrates three allocation components: energy storage discharge, grid rationing, and reactive power support. Since the current core decision-making motivation is to avoid peak high electricity prices (1.35 yuan / kWh) and address low voltage issues, while also considering the need for safe load reduction, the system ultimately determines the company's current energy allocation model as: "Electricity Price Response-Based Power Quality Governance Model" (or simply "Economic Peak Shaving Model"). This model means that the company is utilizing internal energy storage resources to replace expensive grid power, while simultaneously assisting the grid in stabilizing voltage through reactive power regulation, achieving a win-win situation for both the company's economic benefits and the grid's safety and stability.

[0076] refer to Figure 5 In step S14, the specific steps are as follows: S141: Collect the current energy consumption scenarios of the enterprise, which include full-load production scenarios, holiday supply guarantee scenarios, and partial maintenance scenarios; determine the baseline curve based on the traceability of the current energy consumption scenarios of the enterprise; integrate the current energy consumption scenarios of the enterprise, the energy allocation mode of the enterprise, and the real-time load of the power grid, and construct the corresponding actual energy consumption curve in the fusion; compare the actual energy consumption curve with the baseline curve, determine the corresponding curve difference area, trace along the curve difference area to determine the corresponding energy-consuming equipment, and mark the abnormal energy consumption area of ​​the enterprise. S142: Based on the identification of the enterprise's abnormal energy consumption areas, determine the corresponding locations and energy consumption data. Perform point-by-point difference calculations between the energy consumption data of the abnormal energy consumption areas and the enterprise's past energy consumption data to determine the energy consumption deviation. At the same time, use a digital twin system to control the energy consumption deviation and trace back upstream along the data chain of the energy consumption deviation to reproduce the root cause analysis results of the anomaly. Based on the identification of the root cause analysis results, determine the energy consumption optimization content of the abnormal energy consumption areas. The energy consumption optimization content includes specific operational parameters.

[0077] In the embodiments of this application, the current energy consumption scenario of the enterprise is collected, which includes full-load production scenario, holiday supply guarantee scenario, and partial maintenance scenario; a baseline curve is determined based on the traceability of the current energy consumption scenario of the enterprise; the current energy consumption scenario of the enterprise, the energy distribution mode of the enterprise, and the real-time load of the power grid are fused into multiple factors, and the corresponding actual energy consumption curve is constructed in the fusion; the actual energy consumption curve is compared with the baseline curve, and the corresponding curve difference area is determined; the corresponding energy-consuming equipment is determined by tracing along the curve difference area, and the abnormal energy consumption area of ​​the enterprise is marked, which is compatible with the overall consideration of the traceability of the current energy consumption scenario of the enterprise and ensures the accuracy of the baseline curve.

[0078] At this point, the system connects to the enterprise's Manufacturing Execution System (MES) and Equipment Management System (EAM) to extract production plans, shift schedules, and maintenance work orders. Using a rule engine or classification method, it automatically identifies the current energy consumption scenario label: whether it is "full-load production" (all equipment in the plant is operating at high load), "holiday supply guarantee" (only basic guarantee load is operated), or "partial maintenance" (some areas are shut down, while other areas are operating). Based on the identified scenario label, it retrieves the corresponding typical operating characteristics and constraints in the knowledge base, providing contextual boundaries for subsequent benchmark matching.

[0079] The system retrieves historical records from the historical database that have the same characteristics as the current scenario (both are operating at full capacity), the same product model, and similar environmental parameters (such as similar temperatures). It extracts the power load curves of these high-quality historical records, aggregates and smooths them to form a theoretically achievable baseline energy consumption curve. This curve represents the energy consumption trajectory of the enterprise's equipment when it is operating at its best in the current scenario.

[0080] The system takes the enterprise's current energy consumption scenario (as a weighting coefficient), energy allocation mode (the strategy determined by S132, such as energy storage discharge), and real-time grid load (such as changes in motor efficiency caused by grid voltage fluctuations) as input variables; using Kalman filtering or moving average, it integrates the real-time collected power data to generate an actual energy consumption curve that includes real noise, strategy adjustments, and external interference. This curve shows the actual energy consumption situation of the current equipment.

[0081] The system calculates the difference (residual) between the actual energy consumption curve and the benchmark energy consumption curve at each time point; sets a dynamic threshold; if the residual exceeds the threshold and the duration exceeds the set time window (e.g., 5 consecutive minutes), the area is determined to be a curve difference area; the system marks the characteristics of the difference area, such as "positive offset" (excessive energy consumption) and "violent fluctuation" (poor contact or control oscillation).

[0082] The system traces downstream from the total metering point (difference detection point) along the energy consumption path established by S112; using correlation analysis or causal inference, it calculates the similarity between the real-time load changes of each sub-node (branch, device) and the waveform of the total difference area; the node with the highest similarity is the source device causing the difference; on the digital twin interface, the corresponding physical device or area is highlighted (e.g., marked in red), marked as "energy consumption abnormal area", and an abnormality report is pushed.

[0083] Specifically, the company's digital twin system read the MES production schedule for the day and, combined with the current calendar time (14:30 on a weekday), identified the current "full-load production scenario" through semantic analysis. The system traced back and found that this scenario meant that the company's core production line (including the No. 1 smelting furnace) had to operate at full speed, auxiliary equipment (such as dust removal and cooling) had to be fully operational, and unplanned shutdowns were not allowed. The determination of this scenario provided a basis for setting the "high standard" of the baseline curve.

[0084] For the company's "full-load production scenario", the system matched the best operating data from the same period last year in the database; based on this, the system generated the current benchmark energy consumption curve: Start-up phase (08:00-09:00): power climbs linearly, peaking at 750kW; steady-state production phase (09:00-17:00): power fluctuates around 800kW (standard deviation ±5kW). This 800kW steady-state line is the company's current energy consumption benchmark.

[0085] When constructing the actual energy consumption curve of the enterprise, the system integrated multiple factors: Scenario impact: Full-load production leads to a high base load; Energy distribution mode impact: Since S132 is implementing "economic peak shaving mode", the energy storage system is discharging (-60kW), which to some extent reduces the total incoming power, but the total process load remains unchanged; Grid load impact: At this time, the grid voltage is low (-5.5%), which causes asynchronous motors (such as the stirring fan of the smelting furnace) to increase current in order to maintain output, resulting in increased copper loss. The generated actual energy consumption curve shows that although energy storage shavings the peak, due to the efficiency loss caused by the low grid voltage, the actual curve of the enterprise's total load does not reach the ideal flatness, but shows a slow upward trend of 820kW, superimposed with high-frequency fluctuation noise.

[0086] The system overlays and compares the company's actual curve (820kW trend) with the benchmark curve (800kW straight line). The analysis results show that the two curves are significantly separated between 14:25 and 14:35. The actual curve is consistently about 20kW higher than the benchmark curve and exhibits non-periodic sawtooth fluctuations. The system determines that this period is a significant curve difference area, characterized by "continuous positive deviation accompanied by high-frequency fluctuations", which means that there is unexplained energy loss.

[0087] Faced with a positive deviation of 20kW, the system immediately initiated a tracing along the path: 10kV power grid > Transformer No. 2 > Low-voltage distribution cabinet A3 > Smelting furnace No. 1. Analysis revealed that among all lower-level branches, the power fluctuation curve of the "hydraulic station circuit of smelting furnace No. 1" was highly consistent with the fluctuation waveform of the total difference area (correlation coefficient > 0.95). Simultaneously, further analysis showed that due to the low grid voltage, the main oil pump motor of the hydraulic station automatically increased the current to maintain system pressure, resulting in additional heat loss. The system immediately marked the "hydraulic station area" located in a corner of the workshop as a red "abnormal energy consumption area" in the enterprise's digital twin 3D map and prompted: "Due to the low grid voltage, the hydraulic pump efficiency has decreased, and the energy consumption is abnormally high by 20kW."

[0088] Furthermore, based on the identification of energy consumption anomalies in the enterprise, the corresponding locations and energy consumption data are determined. A point-by-point difference calculation is performed between the energy consumption data of the anomaly areas and the enterprise's historical energy consumption data to determine the energy consumption deviation. Simultaneously, a digital twin system is used to control the energy consumption deviation, and the data chain of the deviation is traced upstream to reproduce the root cause analysis results. Based on the identification of the root cause analysis results, energy consumption optimization content for the anomaly areas is determined. This optimization content includes specific operational parameters and incorporates the overall considerations of the root cause analysis results, ensuring the accuracy of the energy consumption optimization content for the anomaly areas. Additionally, an energy allocation mode is introduced, fully considering the enterprise's energy consumption anomalies, thus achieving control over energy consumption deviation and improving the accuracy of energy consumption optimization content for the anomaly areas.

[0089] At this point, the system accurately locates the spatial coordinates and corresponding logical measurement points of the energy consumption anomaly area marked by S141 (such as a specific workshop or a specific circuit); extracts the current real-time energy consumption time series data Preal(t) (power, current, flow, etc.) of the area; calls the historical best energy consumption data Pbase(t) of the device / area under the same operating conditions and time period in the enterprise database; and performs a point-by-point difference operation of ΔP(t)=Preal(t)−Pbase(t), which is not only a numerical subtraction but also a vector difference process to calculate the power deviation at each sampling time.

[0090] By integrating or weighting the difference sequence ΔP(t) over the abnormal duration period and supplementing it with root mean square (RMS) calculation, a statistically significant energy consumption deviation (unit: kWh or instantaneous kW) is determined. This deviation excludes normal fluctuations and only reflects the pure increment caused by the anomaly.

[0091] The calculated energy consumption deviation (+6.5kW) is used as an input parameter and injected back into the digital twin system constructed by S122. The digital twin simulates the operating state of the device in virtual space. The system attempts to adjust various physical parameters in the model (such as friction coefficient, resistance value, and pressure setpoint) to observe which parameter change causes the virtual model output to match the actual deviation. At the same time, it traces back upwards along the data chain of "deviation source > actuator > control logic > sensor feedback". When the virtual model is adjusted to a certain state (e.g., the system resistance increases abnormally), the simulation output completely coincides with the actual data. At this time, the system can reproduce the mechanism of the anomaly and lock in the root cause analysis results.

[0092] Based on the root cause analysis results, the system performs reverse optimization testing in the digital twin; that is, it determines which parameters can be adjusted to eliminate the deviation and restore the equipment to the optimal energy consumption range. The system calculates specific adjustment values, which are no longer qualitative suggestions (such as "adjusting the valve"), but quantitative specific operating parameters, including the correction values ​​of PID parameters, the valve opening angle, the frequency set value, etc., and generates an energy consumption optimization instruction set containing the target equipment, the object to be adjusted, the target parameter values, and the expected effects.

[0093] Specifically, the system identified the company's "hydraulic station area" as an abnormal region; extracted the real-time power curve of the hydraulic station's main pump during the current time period (14:25-14:35), with an average value of 45.0kW; retrieved the best operating data for the same process section (full-load smelting) from the historical database, with an average power of 38.5kW; performed point-by-point difference calculation, and found that the real-time power was continuously oscillating above the baseline; calculated that the energy consumption deviation during the current time period was +6.5kW (instantaneous increment), which is converted into an abnormal power loss of 1.08kWh, indicating that the hydraulic station consumed approximately 1 kWh more electricity due to abnormal reasons in the past 10 minutes.

[0094] The enterprise's digital twin system starts a high-fidelity model of the hydraulic station. The system asks, "What caused the main pump to output an extra 6.5kW of power?" The model, through fluid dynamics simulation, finds that in order to maintain the normal system pressure (120Bar), the pump's output flow rate did not increase significantly, but the motor input power increased. The system traces along the data chain: motor ammeter (correct) > frequency converter (correct) > hydraulic circuit (anomaly found). The simulation shows that if the return oil filter is clogged, causing the back pressure to increase, or if the proportional valve opening is insufficient, this phenomenon will occur.

[0095] Further analysis of the low grid voltage (-5.5%) data collected by S132 was conducted using the model: Under low voltage, the motor's electromagnetic torque decreases and the slip increases, leading to a sharp increase in rotor copper losses; at the same time, the PID controller automatically increases the current setpoint to maintain a constant speed; the system ultimately reproduced the abnormal mechanism, and the root cause analysis result was determined to be: "Due to the grid voltage drop, the motor efficiency characteristics deteriorate, coupled with a slight increase in back pressure in the hydraulic pipeline, causing the motor to dig deep in the inefficient zone."

[0096] To address the inefficiency of the enterprise's hydraulic power station caused by low voltage, a digital twin system was used for optimization simulation. Since the grid voltage could not be changed and mechanical back pressure could not be eliminated online, the system calculated that the current process pressure (120 Bar) was slightly higher than the minimum requirement for the actual smelting process (115 Bar). Appropriately reducing the system pressure could decrease the motor load torque, thereby offsetting the losses caused by low voltage and reducing energy consumption deviation. The system calculated optimized operating parameters. Optimization output: Target object: Hydraulic power station proportional pressure valve; Current setpoint: 120 Bar; Specific operating parameters (after optimization): Adjusted to 115 Bar; Auxiliary parameters: The inverter torque boost parameter was adjusted from the current "1.0" to "1.2" to enhance the starting torque under low voltage; Expected effect: The main pump power is expected to decrease by 6.0 kW, eliminating energy consumption deviation without affecting production safety.

[0097] Please see Figure 6 , Figure 6 This is a schematic diagram of the structural composition of a digital twin-based multi-scenario energy consumption management system according to an embodiment of the present invention; the digital twin-based multi-scenario energy consumption management system is applied to the above-mentioned digital twin-based multi-scenario energy consumption management method; the digital twin-based multi-scenario energy consumption management system includes: The energy consumption path module 21 is used to mark each energy-consuming device in the enterprise's distribution map, construct the enterprise's dynamic energy consumption map based on the real-time data of each energy-consuming device, the enterprise's overall energy consumption data and the corresponding distribution map, and determine the energy consumption path associated with each energy-consuming device. The digital twin module 22 is used to identify multiple key energy consumption nodes based on the identification of each energy consumption path, and to construct a digital twin system of the energy consumption path based on the working status and time nodes of the multiple key energy consumption nodes and energy consumption equipment in order to determine the current energy consumption level of the energy consumption path. The energy distribution module 23 is used to determine the energy consumption safety range based on the current energy consumption level of each energy consumption path and the enterprise's past energy consumption data. At the same time, it marks the power grid associated with the enterprise, collects the power grid's energy supply data, and determines the corresponding energy distribution mode based on the power grid's energy supply data and the energy consumption safety range. The energy consumption optimization module 24 is used to mark the current energy consumption scenario of the enterprise, and determine the energy consumption abnormal area of ​​the enterprise by combining the energy distribution mode and the real-time load of the power grid. Based on the location of the energy consumption abnormal area, the corresponding energy consumption data and the enterprise's past energy consumption data, the energy consumption deviation is determined, and the energy consumption optimization content of the energy consumption abnormal area is determined by tracing the energy consumption deviation.

[0098] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A multi-scenario energy consumption management method based on digital twinning, characterized in that, include: Mark each energy-consuming device in the enterprise's distribution map, construct the enterprise's dynamic energy consumption map based on the real-time data of each energy-consuming device, the enterprise's overall energy consumption data and the corresponding distribution map, and determine the energy consumption paths associated with each energy-consuming device; Based on the identification of each energy consumption path, multiple key energy consumption nodes are identified. A digital twin system for the energy consumption path is constructed according to the working status and time nodes of the multiple key energy consumption nodes and energy-consuming equipment to determine the current energy consumption level of the energy consumption path. The energy consumption safety range is determined based on the current energy consumption level of each energy consumption path and the enterprise's past energy consumption data. At the same time, the power grid associated with the enterprise is marked, and the power grid's energy supply data is collected. The corresponding energy allocation mode is determined based on the power grid's energy supply data and the energy consumption safety range. The system marks the current energy consumption scenario of an enterprise and determines the abnormal energy consumption area of ​​the enterprise by combining the energy distribution mode and the real-time load of the power grid. Based on the location of the abnormal energy consumption area, the corresponding energy consumption data and the enterprise's past energy consumption data, the energy consumption deviation is determined, and the energy consumption optimization content of the abnormal energy consumption area is determined by tracing the energy consumption deviation.

2. The digital-twin-based energy consumption management method under multiple scenarios according to claim 1, characterized in that, The process involves marking each energy-consuming device on the enterprise's distribution map, constructing a dynamic energy consumption map of the enterprise based on real-time data from each energy-consuming device, the enterprise's overall energy consumption data, and the corresponding distribution map, and determining the energy consumption paths associated with each energy-consuming device, including: The system collects the company's name and traces its distribution map to identify the company. This map includes building and geographic information and marks each energy-consuming device. Real-time data for each energy-consuming device is then determined based on its real-time monitoring.

3. The digital-twin-based energy consumption management method under multiple scenarios according to claim 2, characterized in that, The step of marking each energy-consuming device in the enterprise's distribution map, constructing a dynamic energy consumption map of the enterprise based on the real-time data of each energy-consuming device, the enterprise's overall energy consumption data, and the corresponding distribution map, and determining the energy consumption paths associated with each energy-consuming device, also includes: In the enterprise's distribution diagram, each energy-consuming device is abstracted as a device node and edge in the topology. Device nodes are mapped to energy-consuming devices at each level, and edges represent the physical medium connecting energy flow. Real-time data of each energy-consuming device is mapped to the corresponding device node, and the transient changes of each device node are captured in real time. Based on the transient changes and the overall energy consumption data of the enterprise, an energy consumption dynamic diagram of the enterprise is constructed. Based on the identification of this energy consumption dynamic diagram, the direction of energy consumption change is determined, and the energy consumption path associated with each energy-consuming device is determined by tracing along each direction of energy consumption change.

4. The digital-twin-based energy consumption management method under multiple scenarios according to claim 1, characterized in that, The process involves identifying multiple key energy consumption nodes based on the identification of various energy consumption paths, and constructing a digital twin system for each energy consumption path based on these key energy consumption nodes, the operating status of energy-consuming devices, and time points. This system aims to determine the current energy consumption level of the energy consumption path, including: Each energy consumption path is input into a deep learning network, and deep learning is performed in the deep learning network to determine multiple primary energy consumption nodes. The energy consumption priority of multiple primary energy consumption nodes is determined based on the node position and corresponding energy consumption data combination. Combined with the overall energy consumption of the enterprise, a node screening mechanism is determined to screen out multiple key energy consumption nodes.

5. The digital-twin-based energy consumption management method under multiple scenarios according to claim 4, characterized in that, The process of identifying multiple key energy consumption nodes based on the identification of each energy consumption path, and constructing a digital twin system for the energy consumption path based on the operating status and time points of the multiple key energy consumption nodes and energy-consuming equipment to determine the current energy consumption level of the energy consumption path, also includes: The system collects the operating status of energy-consuming devices and constructs corresponding digital twin frameworks based on the operating status of multiple key energy-consuming nodes and devices. It then performs multi-factor fusion of environmental data and time nodes of the digital twin framework and energy-consuming devices, and conducts multi-scenario simulations in the fused state to output a digital twin system for the energy consumption path. The system also determines the current energy consumption level of the energy consumption path through real-time identification of the digital twin system. The current energy consumption level of the energy consumption path covers high-efficiency level, economic level, and early warning level.

6. The energy consumption management method for multiple scenarios based on digital twins according to claim 1, characterized in that, The process involves determining the energy consumption safety range based on the current energy consumption level of each energy consumption path and the enterprise's historical energy consumption data. Simultaneously, it involves marking the power grid associated with the enterprise and collecting the power grid's supply data. Based on the power grid's supply data and the energy consumption safety range, a corresponding energy allocation mode is determined, including: Based on the detection of the enterprise's database, the enterprise's past energy consumption data is determined. The current energy consumption level of each energy consumption path is correlated with the enterprise's past energy consumption data. Combined with deep clustering, the distribution information of the enterprise's optimal energy consumption range under different production conditions is determined. Based on the distribution information of the optimal energy consumption range and the matching of the time dimension, the energy consumption fluctuation range of the energy consumption path is determined. Combined with the enterprise's preset safe energy consumption level, the corresponding safe energy consumption range is determined.

7. The energy consumption management method for multiple scenarios based on digital twins according to claim 6, characterized in that, The process of determining the energy consumption safety range based on the current energy consumption level of each energy consumption path and the enterprise's historical energy consumption data, while simultaneously marking the power grid associated with the enterprise and collecting the power grid's energy supply data, and determining the corresponding energy allocation mode based on the power grid's energy supply data and the energy consumption safety range, also includes: Collect data from the power grid associated with the enterprise, and comprehensively collect multiple energy supply data from the power grid, including power grid frequency data, voltage deviation data, and real-time electricity price data. Unify the energy supply data from the power grid side with the enterprise's energy consumption safety range, and perform multi-objective solutions to determine the corresponding multiple energy allocation contents. Based on the multi-fusion of multiple energy allocation contents, determine the corresponding energy allocation mode.

8. The energy consumption management method for multiple scenarios based on digital twins according to claim 1, characterized in that, The system identifies the current energy consumption scenario of the marked enterprise and determines the enterprise's abnormal energy consumption area by combining the energy distribution mode and the real-time load of the power grid. Based on the location of the abnormal energy consumption area, the corresponding energy consumption data, and the enterprise's historical energy consumption data, it determines the energy consumption deviation. Then, by tracing the energy consumption deviation, it determines the energy consumption optimization content for the abnormal energy consumption area, including: The system collects data on the enterprise's current energy consumption scenarios, including full-load production, holiday supply guarantee, and partial maintenance. A baseline curve is determined based on the traceability of the enterprise's current energy consumption scenarios. The system integrates the enterprise's current energy consumption scenarios, energy allocation mode, and real-time load of the power grid, and constructs the corresponding actual energy consumption curve during the integration process. The actual energy consumption curve is compared with the baseline curve to identify the corresponding curve difference area. The system traces along the curve difference area to identify the corresponding energy-consuming equipment and marks the enterprise's abnormal energy consumption areas.

9. The energy consumption management method for multiple scenarios based on digital twins according to claim 8, characterized in that, The system identifies the current energy consumption scenario of the marked enterprise and determines the enterprise's abnormal energy consumption area by combining the energy distribution mode and the real-time load of the power grid. Based on the location of the abnormal energy consumption area, the corresponding energy consumption data, and the enterprise's historical energy consumption data, it determines the energy consumption deviation. Furthermore, it traces the energy consumption deviation to determine the energy consumption optimization content for the abnormal energy consumption area, and also includes: Based on the identification of abnormal energy consumption areas in the enterprise, the corresponding locations and energy consumption data are determined. The energy consumption data of the abnormal energy consumption areas are compared with the enterprise's previous energy consumption data point by point to determine the amount of energy consumption deviation. At the same time, the energy consumption deviation is controlled using a digital twin system, and the data chain of energy consumption deviation is traced back upwards to reproduce the root cause analysis results of the anomaly. Based on the identification of the root cause analysis results, the energy consumption optimization content of the abnormal energy consumption areas is determined, and the energy consumption optimization content includes specific operational parameters.

10. A multi-scenario energy consumption management system based on digital twins, characterized in that, The energy consumption management system based on digital twins in multiple scenarios is applied to the energy consumption management method based on digital twins in multiple scenarios as described in any one of claims 1-9; The energy consumption management system based on digital twins for multiple scenarios includes: The energy consumption path module is used to mark each energy-consuming device in the enterprise's distribution map, construct the enterprise's dynamic energy consumption map based on the real-time data of each energy-consuming device, the enterprise's overall energy consumption data and the corresponding distribution map, and determine the energy consumption path associated with each energy-consuming device. The digital twin module is used to identify multiple key energy consumption nodes based on the identification of each energy consumption path. Based on the multiple key energy consumption nodes, the working status of energy-consuming equipment and time nodes, a digital twin system of the energy consumption path is constructed to determine the current energy consumption level of the energy consumption path. The energy allocation module is used to determine the safe energy consumption range based on the current energy consumption level of each energy consumption path and the enterprise's past energy consumption data. At the same time, it marks the power grid associated with the enterprise, collects the power grid's energy supply data, and determines the corresponding energy allocation mode based on the power grid's energy supply data and the safe energy consumption range. The energy consumption optimization module is used to mark the current energy consumption scenario of an enterprise, and determine the abnormal energy consumption area of ​​the enterprise by combining the energy distribution mode and the real-time load of the power grid. Based on the location of the abnormal energy consumption area, the corresponding energy consumption data and the enterprise's past energy consumption data, the energy consumption deviation is determined, and the energy consumption optimization content of the abnormal energy consumption area is determined by tracing the energy consumption deviation.

Images