A thermal energy management system and method based on an industrial green microgrid

Abstract

The present application relates to the technical field of industrial micro-grid thermal energy management, in particular to a thermal energy management system and method based on industrial green micro-grid, comprising: receiving original monitoring information of heat energy generation, consumption and storage units to form a multi-source heterogeneous data set; removing noise and filling information gaps through fusion cleaning to establish a unified formatted energy flow record with time sequence labels; constructing a multi-modal energy flow topology graph based on the record, simulating to generate future multi-time section thermal energy flow state evolution data, extracting abnormal fluctuation characteristics and forming a time-labeled abnormal report; matching historical cases to generate a preliminary strategy, and screening an optimized regulation scheme through forward-looking deduction. The method can realize energy flow dynamic prediction and accurate response to abnormalities, reduce thermal energy loss and network operation risk, and improve the scientificity and effectiveness of industrial green micro-grid thermal energy management.

Description

Technical Field

[0001] This invention relates to the field of thermal energy management technology for industrial microgrids, and in particular to a thermal energy management system and method based on industrial green microgrids. Background Technology

[0002] Thermal management in industrial green microgrids is a crucial aspect of ensuring efficient and stable operation. Current technologies often rely on single types of thermal monitoring data, offering only basic processing for multi-source, heterogeneous signals from thermal generation units, consumption devices, and storage facilities. This results in inadequate data acquisition and display, lacking systematic fusion and cleaning of multi-source data, and failing to establish a unified energy flow record format. Furthermore, conventional thermal management methods primarily employ static topology analysis, reflecting only the current energy flow distribution and failing to effectively simulate or predict future energy flow evolution trends.

[0003] Existing technologies are insufficient in their ability to remove noise data and fill information gaps, easily leading to distortion of energy flow data and affecting the accuracy of management decisions. Static energy flow topologies cannot adapt to the dynamic changes in heat supply and demand in industrial green microgrids, making it difficult to identify potential abnormal heat fluctuations in advance. When faced with abnormal heat events, conventional technologies often adopt preset fixed control strategies, lacking effective utilization of historical operating cases, failing to quickly match control experience for similar events, and failing to conduct forward-looking evaluation of the implementation effect of control strategies, easily resulting in poor adaptability of control schemes and delayed response.

[0004] How to effectively integrate and standardize multi-source heterogeneous thermal energy monitoring data, construct a topology model that can reflect the dynamic energy flow state and simulate future energy flow evolution, and accurately identify abnormal thermal energy fluctuations and form scientific control schemes based on historical cases have become urgent problems to be solved in the thermal energy management of current industrial green microgrids. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a thermal energy management system and method based on an industrial green microgrid.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a thermal energy management method based on an industrial green microgrid, comprising: It receives raw monitoring information from heat energy generation units, heat energy consumption equipment and heat energy storage facilities, and forms a data set containing multi-source heterogeneous signals; The dataset is fused and cleaned to remove noisy data and fill information gaps, and a unified formatted energy flow record containing time-series labels is established. Based on the unified formatted energy flow record, a multimodal energy flow topology graph is constructed; Within the framework of the multimodal energy flow topology, energy flow state evolution simulation is performed to generate thermal energy flow state evolution data containing multiple future time sections; The abnormal thermal energy fluctuation characteristics that violate the preset evolution law are extracted from the thermal energy flow evolution data, and their spatiotemporal attributes are labeled to form an abnormal thermal energy event report with time stamp. The historical running case library is invoked to match and search the time-stamped abnormal thermal event reports, and to identify past event records with similar characteristics; For the identified similar past event records, trace back their corresponding regulatory response sequences to generate a preliminary set of response strategies containing at least one regulatory action; The preliminary response strategy set is input into the multimodal energy flow topology diagram for forward-looking simulation, and the impact trajectory of each regulation action on the overall thermal network state is evaluated. Based on the evaluation results, an optimized control scheme that enables the energy flow network to return to the predetermined stable range was selected.

[0007] As a further aspect of the present invention, the step of receiving raw monitoring information from the heat energy generating unit, the heat energy consuming device, and the heat energy storage facility to form a data set containing multi-source heterogeneous signals specifically includes: Information on heat generation capacity and medium temperature is collected from distributed renewable energy devices; Real-time heat energy demand and medium pressure information are collected from industrial heat-using process nodes; Collect information on storage levels, inlet and outlet temperatures, and flow rates from phase change energy storage tanks or hot water storage tanks; Collect pipeline temperature and pressure information from key nodes of the heat exchange network; The collected information on heat generation power, medium temperature, real-time heat energy demand, medium pressure, inventory level, inlet and outlet temperature, flow rate, pipeline temperature and pressure is associated and packaged according to the source equipment code and timestamp to form an initial data set.

[0008] As a further aspect of the present invention, the data set is fused and cleaned to remove noisy data and fill information gaps, thereby establishing a unified formatted energy flow record containing time-series labels, specifically including: Outlier detection is performed on each signal sequence in the dataset, and a sliding window comparison method is used to identify and remove noisy data points that deviate from the neighborhood mean. For periods of data loss caused by sensor failure or communication interruption, data from nearby functionally similar nodes are used to fill the information gaps through time series interpolation algorithms. Data from different sources, after being cleaned and filled, are transformed and aligned according to a unified time base and physical dimensions; A time-series label is attached to each aligned data point to form a uniformly formatted energy flow record with a consistent structure.

[0009] As a further aspect of the present invention, a multimodal energy flow topology graph is constructed based on the uniformly formatted energy flow record, specifically including: The multimodal energy flow topology diagram reflects the dynamic mapping relationship between heat energy generation, transmission, use and storage; The heat generation unit, heat consumption device, and heat storage facility are abstracted as nodes in the multimodal energy flow topology diagram; The pipes, valves, and heat exchangers connecting the nodes are abstracted as directed edges in the multimodal energy flow topology graph; Assign attributes to each node, including at least node type, rated capacity, and real-time status value; Assign attributes to each directed edge, including at least connectivity, transmission medium, thermal resistivity, and current flow rate; Based on the temporal changes in the uniformly formatted energy flow record, the real-time state values ​​of the nodes and the directed edges are dynamically updated, so that the multimodal energy flow topology graph can reflect the real-time configuration and state of the thermal energy network.

[0010] As a further aspect of the present invention, within the framework of the multimodal energy flow topology diagram, an energy flow state evolution simulation is performed to generate thermal energy flow state evolution data containing multiple future time sections, specifically including: The initial conditions are the real-time state values ​​of each node in the multimodal energy flow topology diagram at the current moment; Based on the predicted output curve of the heat energy generating unit, the production plan of the heat energy consuming equipment, and the ambient temperature forecast, the boundary conditions and disturbance inputs for the future simulation period are set. Based on the fundamental principles of energy conservation and heat transfer, iterative calculations with discrete time steps are performed on the multimodal energy flow topology graph. Simulate the conduction, convection, storage and dissipation of heat energy in the network, and deduce the sequence of changes in the state parameters of each node and directed edge over a future period of time. The snapshots of the entire network state parameters calculated at each discrete time step are stored to form the thermal energy flow evolution data.

[0011] As a further aspect of the present invention, abnormal thermal energy fluctuation characteristics that violate preset evolution laws are extracted from the thermal energy flow evolution data, and their spatiotemporal attributes are labeled to form an abnormal thermal energy event report with time stamps, specifically including: Set safe operating threshold ranges for node temperature change rate, pipeline pressure difference, and energy storage charge / discharge rate; Traverse the thermal energy flow evolution data and compare each parameter with the corresponding safe operation threshold range at each time section; Identify abnormal fluctuations where parameter values ​​continuously exceed the safe operating threshold range or experience short-term, drastic jumps. Record the start time, duration, node and directed edge numbers involved, and specific parameters exceeding the threshold of the abnormal fluctuations; All recorded abnormal fluctuation information is summarized and structured into the aforementioned time-stamped abnormal thermal event report.

[0012] As a further aspect of the present invention, the step of calling the historical running case library to match and search the time-stamped abnormal thermal event reports and identify past event records with similar characteristics specifically includes: The time-stamped abnormal thermal event report is parsed to extract key feature vectors, which include anomaly type, occurrence location, intensity, and duration patterns. In the historical case database, similarity calculation based on multidimensional feature space distance is performed using the key feature vector as the query condition. Retrieve all historical case entries whose distance from the current event's feature vector is less than a preset threshold; The retrieved historical case entries are sorted in descending order of similarity to form a list of similar historical events.

[0013] As a further aspect of the present invention, the step of backtracking the corresponding regulatory response sequence for the identified similar past event records to generate a preliminary response strategy set containing at least one regulatory action specifically includes: Read the complete event processing log from each past event record in the list of similar historical events; Extract all control operation instructions actually executed after the occurrence of the historical event from the event processing log. The control operation instructions include valve opening adjustment, pump start and stop, standby heat source input and load switching. The extracted control operation instructions are combined into a complete control response sequence for historical events according to their execution order. By summarizing the control response sequences corresponding to all similar historical events and removing duplicate or contradictory control actions, a preliminary set of response strategies containing multiple optional control actions is formed.

[0014] As a further aspect of the present invention, the preliminary response strategy set is input into the multimodal energy flow topology graph for prospective extrapolation, and the impact trajectory of each regulation action on the overall thermal network state is evaluated, specifically including: Select one regulatory action to be evaluated from the initial set of response strategies; Based on the current state of the multimodal energy flow topology graph, the control action to be evaluated is applied to update the state parameters of the relevant nodes or directed edges. Using the updated state as the new initial conditions, the energy flow state evolution simulation is re-executed to predict the evolution of the network state over a future period of time. Analyze the evolution process, record the time required for key parameters to recover to the safe operating threshold range, whether new abnormal fluctuations are triggered during the process, and the final stable state of the system; Repeat the steps until every regulatory action in the initial response strategy set has been prospectively analyzed and its impact assessed; Based on the evaluation results, optimized control schemes that enable the energy flow network to return to a predetermined stable range are selected, specifically including: Set optimization objectives, which are to enable the system to recover to stability as quickly as possible, minimize overall adjustment energy consumption, or minimize disturbance to the production process; Compare the forward-looking projection and evaluation results of all control actions, and score and rank each control action according to the optimization objectives. Select the single control action with the highest score, or combine multiple control actions that do not conflict in the evaluation and have synergistic effects to form an optimized control scheme; The optimized control scheme was verified to ensure that all key parameters of the multimodal energy flow topology were within the predetermined stable range at the end of the simulation.

[0015] As a further aspect of the present invention, the present invention also includes a thermal energy management system based on an industrial green microgrid, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, it implements the steps of the thermal energy management method based on an industrial green microgrid as described above.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: Based on the unified formatted energy flow records with time-series labels after fusion and cleaning, a multimodal energy flow topology map is constructed. Within this topology map framework, energy flow state evolution simulation is performed to generate thermal energy flow state evolution data covering multiple future time segments. After fusion and cleaning, the original monitoring information of multi-source heterogeneous thermal energy removes noise data and fills information gaps. The unified formatted energy flow records with time-series labels can clearly present the energy flow change patterns at different time points. The multimodal energy flow topology map integrates the correlation relationships of various units of thermal energy generation, consumption, and storage. Combined with energy flow state evolution simulation, it can capture the energy flow change trends at multiple future time segments, breaking the limitation of conventional technologies that can only monitor the current energy flow state and cannot predict future evolution trends, thus reducing the frequency of abnormal events.

[0017] This method extracts spatiotemporally attributed anomalous thermal energy fluctuation features from multi-time-section thermal energy flow evolution data, generating time-stamped anomalous thermal energy event reports. It then uses a historical case library to match similar past event records, retrospectively analyzes corresponding control response sequences to generate a preliminary response strategy set. This preliminary strategy set is then input into a multimodal energy flow topology map for forward-looking simulation, evaluating the impact trajectory of each control action on the overall thermal network state, and ultimately selecting the optimal control scheme. The spatiotemporally attributed anomalous fluctuation features and time-stamped anomalous event reports accurately pinpoint the specific location and time of anomalies, avoiding the problems of fuzzy anomaly identification and inaccurate location in conventional technologies. The application of the historical case library allows for the reference of control experience from similar anomalous events, reducing the blindness of control strategies. Forward-looking simulation can grasp the overall impact of each control action on the thermal network, selecting control schemes suitable for the current energy flow state. This overcomes the limitations of fixed control and poor targeting in conventional technologies, enabling rapid response to anomalous events, restoring the thermal network to a stable range, and reducing heat loss and network operation risks. Attached Figure Description

[0018] Figure 1 This is a flowchart of a thermal energy management method based on an industrial green microgrid according to the present invention; Figure 2 This is a flowchart illustrating the process of collecting raw monitoring information and forming a data set. Figure 3 A flowchart for constructing a multimodal energy flow topology graph; Figure 4 A comparison chart of recovery time and energy consumption increase for industrial green microgrid control strategies; Figure 5 This is a graph showing the evolution trend of outlet temperature of the thermal storage unit and pressure at key nodes of the pipeline network. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] See Figure 1 The system receives raw monitoring information from heat generation units, heat consumption devices, and heat storage facilities, forming a dataset containing multi-source heterogeneous signals. This dataset is then fused and cleaned to remove noise and fill information gaps, creating a uniformly formatted energy flow record with time-series labels. Based on this record, a multimodal energy flow topology map is constructed, depicting the dynamic mapping relationships between heat generation, transmission, use, and storage. Within this multimodal topology map, energy flow state evolution simulation is performed, generating heat flow state evolution data across multiple future time segments. Abnormal heat energy fluctuation characteristics that violate preset evolutionary patterns are extracted from the data, and their spatiotemporal attributes are labeled, forming time-stamped abnormal heat energy event reports. A historical case library is used to match and retrieve these time-stamped abnormal heat energy event reports, identifying past event records with similar characteristics. For these identified similar past event records, their corresponding regulatory response sequences are traced back, generating a preliminary response strategy set containing at least one regulatory action. A preliminary set of response strategies is input into the multimodal energy flow topology diagram for prospective simulation, evaluating the impact trajectory of each regulatory action on the overall thermal network state. Based on the evaluation results, an optimized regulatory scheme that restores the energy flow network to a predetermined stable range is selected.

[0022] In one embodiment of the invention, raw monitoring information from heat generation units, heat consumption devices, and heat storage facilities is received to form a data set containing multi-source heterogeneous signals. This is implemented in a chemical plant thermal microgrid that includes a solar collector field, process heat from fermenters, and steam storage tanks. The monitoring system for the solar collector field collects heat generation power and heat transfer oil outlet temperature information at a frequency of once per minute. (See also...) Figure 2The distributed control system of the fermenter collects real-time information on heat demand and steam pressure entering the tank. Sensors deployed on the steam storage tank collect information on storage level, inlet steam temperature, outlet steam temperature, and circulating water flow rate. At specific nodes of the main steam pipeline in the plant area, pressure and temperature transmitters continuously collect pipeline temperature and pressure information. It can be understood that the collected information on heat production power, medium temperature, real-time heat demand, medium pressure, storage level, inlet and outlet temperatures, flow rate, pipeline temperature, and pressure all carry a unique device code and a timestamp accurate to milliseconds. The data acquisition gateway associates and packages this information according to a preset protocol, forming an initial, structured data set.

[0023] The dataset is fused and cleaned to remove noisy data and fill information gaps, establishing a uniformly formatted energy flow record with time-series labels. In practice, outlier detection is performed independently for each signal sequence in the dataset. For the heat production power sequence of the solar collector field, a sliding window with a width of 10 sampling points is used. The mean and standard deviation of the data within the window are calculated. When a data point deviates from the window mean by more than 3 times the standard deviation, the sliding window comparison method identifies that point as a noisy data point and removes it. For the 5-minute data gap caused by a temporary failure of the fermenter pressure sensor, the pressure data sequence of another adjacent fermenter node with similar processes is used to fill in the missing data using a Lagrange interpolation algorithm, generating a continuous time series. Optionally, the heterogeneous data from solar collectors, fermenters, steam storage tanks, and pipelines, after being cleaned and filled, are uniformly converted to a timeline based on Coordinated Universal Time (UTC). All energy-related physical quantities are uniformly converted to megajoules, all temperature quantities to Kelvin, and all pressure quantities to Pascals. Based on this, strict alignment of the different source data is achieved in the time series. Each converted and aligned data point is appended with a time-series label containing year, month, day, hour, minute, second, and millisecond, ultimately forming a uniformly formatted energy flow record that is continuous in time, consistent in dimensions, and structurally regular.

[0024] In some embodiments, the threshold coefficient for identifying outliers in the sliding window comparison method can be dynamically adjusted based on data stationarity, and the calculation formula is as follows: in: Represents a dynamic threshold. This represents the arithmetic mean of the data within the sliding window. This represents the standard deviation of the data within the sliding window. This represents the configurable sensitivity coefficient. It is understood that time series interpolation algorithms are not limited to Lagrange interpolation; linear interpolation or spline interpolation methods can also be used in specific implementations. The fusion cleaning process is entirely executed automatically by a pre-set program, and uniformly formatted energy flow records are persistently stored in the form of database tables or files in a specific format.

[0025] In one embodiment of the present invention, a multimodal energy flow topology graph is constructed based on uniformly formatted energy flow records. These records originate from a thermal energy network in an industrial park comprising a biomass boiler, a drying production line, and a high-temperature thermal storage tank. The multimodal energy flow topology graph reflects the dynamic mapping relationships between heat generation, transmission, use, and storage. The construction process is based on graph theory; see [reference needed]. Figure 3 This approach abstracts a biomass boiler in the physical world as a heat source node in a multimodal energy flow topology graph, three parallel drying production lines as three independent load nodes, a high-temperature heat storage tank as a heat storage node, and steam pipes, regulating valves, and plate heat exchangers connecting these devices as directed edges connecting the corresponding nodes. The direction of the directed edges represents the mainstream heat flow direction under the design conditions. Each node is assigned a set of attributes, including at least node type, rated capacity, and real-time status values. The node type distinguishes between "heat source," "load," and "heat storage." Rated capacity represents the maximum output of a heat source node, its maximum demand for a load node, and its maximum heat storage for a heat storage node. Real-time status values ​​are obtained from a uniformly formatted energy flow record; for example, the real-time status value of a biomass boiler node includes the current output thermal power and outlet steam temperature. Each directed edge is also assigned attributes. The attribute set includes at least the connection relationship, transmission medium, thermal resistance coefficient, and current flow rate. The connection relationship defines the starting and ending node numbers to which the edge is connected. The transmission medium is recorded as "superheated steam" or "heat transfer oil". The thermal resistance coefficient is a fixed parameter calculated based on the pipe material, length, and insulation. The current flow rate value is obtained by mapping the corresponding pipe flow meter data in the uniformly formatted energy flow record.

[0026] The multimodal energy flow topology graph reflects the real-time configuration and state of the thermal energy network. Based on the temporal changes in the uniformly formatted energy flow records, the real-time state values ​​of nodes and directed edges are dynamically updated programmatically. In specific implementations, whenever the latest uniformly formatted energy flow record for a given time slice is received, the system traverses all nodes and directed edges in the multimodal energy flow topology graph, searches for the latest data matching the device code in the uniformly formatted energy flow records, and uses this data to overwrite the real-time state values ​​of the corresponding nodes or directed edges, completing one state refresh of the multimodal energy flow topology graph. For example, the real-time state value "current heat demand" of the load node of the drying production line is updated to the latest reported real-time heat demand of the production line in the uniformly formatted energy flow record, and the attribute "current flow rate" of the directed edge connecting the biomass boiler and the main pipeline is updated to the latest reading of the corresponding pipeline flow meter in the uniformly formatted energy flow record. In some embodiments, the dynamic updating of the real-time state values ​​of nodes follows a unified mapping and assignment rule, the relationship of which can be expressed as: in: This represents the real-time state value vector of a node v in the multimodal energy flow topology at time t. This represents the latest set of data in the uniformly formatted energy flow record at time t that is associated with the physical device corresponding to node v. This represents a predefined mapping function from raw data to node state attributes. It can be understood that the real-time state value updates of directed edges follow the same logic. Optionally, the structure of the multimodal energy flow topology graph, i.e., the addition and removal of nodes and edges, is defined during initial system configuration and manually updated and maintained by administrators during equipment maintenance or network topology modifications. The multimodal energy flow topology graph is stored and computed in memory using an adjacency list or attribute graph data structure.

[0027] In one embodiment of the present invention, energy flow state evolution simulation is performed within the framework of a multimodal energy flow topology graph to generate thermal energy flow state evolution data containing multiple future time segments. The scenario is set as a regional heating microgrid integrating a geothermal source, multiple process workshops, and a large water-based thermal storage unit. The initial conditions are the real-time state values ​​of each node in the current multimodal energy flow topology graph. These real-time state values ​​include the current output of the geothermal pump, the instantaneous load of each process workshop, the water level and temperature of the thermal storage unit, and the pressure and temperature of each node in the pipeline network. Based on the predicted output curve of the thermal energy generating unit, the production plan of the thermal energy consuming equipment, and the ambient temperature forecast, boundary conditions and disturbance inputs for the future simulation period are set. The predicted output curve of the geothermal pump is derived from the periodic production plan of the geothermal well. The production plan of each process workshop provides a list of thermal energy consuming equipment to be turned on or off in each workshop within the next 8 hours and their rated power. The ambient temperature forecast is obtained from the meteorological service interface for the temperature change data of the next 24 hours. Based on the fundamental principles of energy conservation and heat transfer, iterative calculations with discrete time steps are performed on a multimodal energy flow topology graph. The time step is set to 1 minute. In each iteration step, based on the connections between nodes and the properties of directed edges, the conduction, convection, storage, and dissipation processes of heat energy in the network are calculated. This simulates the dynamic process of heat energy flowing from geothermal source nodes through the pipeline network to load nodes or thermal storage nodes, while simultaneously calculating pipeline heat loss and the self-cooling of thermal storage units. The simulation process continues until the preset future simulation period ends, thus obtaining a sequence of changes in the state parameters of each node and directed edge over a future period. A snapshot of the entire network state parameters calculated at each discrete time step, including the temperature, load, and energy storage level of all nodes, as well as the flow rate and pressure drop of all directed edges, is stored chronologically, ultimately forming structured thermal energy flow evolution data, which is a multidimensional time series array.

[0028] Abnormal thermal energy fluctuations that violate preset evolution patterns are extracted from thermal energy flow evolution data, and their spatiotemporal attributes are labeled to generate time-stamped abnormal thermal energy event reports. In specific implementation, safe operating threshold ranges are set for key parameters such as node temperature change rate, pipeline pressure difference, and energy storage charge / discharge rate. For example, the safe operating threshold range for the pressure difference of the main transmission and distribution pipeline is specified as 0.2 MPa to 0.8 MPa, and the safe operating threshold range for the charge rate of the thermal storage unit is specified as 0 to 50 MW. The thermal energy flow evolution data is traversed, and each parameter is compared with its corresponding safe operating threshold range at each time section. The system program sequentially reads the full network status snapshot at each simulation time point, and checks whether the value of each parameter in the snapshot falls within the preset safe operating threshold range. The system identifies abnormal fluctuations where parameter values ​​consistently exceed safe operating thresholds or experience short-term, drastic jumps. For example, in the simulation data, at minute 120, the pressure in a branch pipeline of a workshop suddenly drops from 0.5 MPa to 0.15 MPa within two simulation steps, which is identified as a short-term, drastic jump. From minute 300 onwards, the heat release rate of the thermal storage unit remains at 55 MW for 10 minutes, exceeding the upper limit of 50 MW, which is identified as a sustained over-limit. The system records the start time, duration, involved node and directed edge numbers, and specific parameters exceeding the threshold for each abnormal fluctuation. For the pressure drop anomaly, the start time is recorded as minute 120 of the simulation, the duration as 2 minutes, involving the directed edge "pipeline P-203," and the parameter "pressure" value as 0.15 MPa (below the lower limit of 0.2 MPa). In some embodiments, the threshold for determining short-term, drastic jumps can be calculated by comparing the absolute value of the rate of change of the parameter between adjacent time steps with a set threshold. The calculation formula is as follows: in: This is a Boolean condition indicating whether a transition occurred. and These represent the parameters at two consecutive simulation time steps. and The value, Represents the simulation time step. This represents the preset rate of change threshold for specific parameters. Optionally, all recorded abnormal fluctuation information is summarized and organized into a time-stamped abnormal thermal event report according to a predefined structured format. The report includes fields such as event ID, trigger time, list of involved components, description of abnormal parameters, out-of-limit values, and duration. It can be understood that the abnormal thermal event report is automatically generated after the simulation and is output as an intermediate result.

[0029] In one embodiment of the present invention, the historical operation case library is invoked to match and retrieve time-stamped abnormal thermal energy event reports, and past event records with similar characteristics are identified. A newly generated abnormal thermal energy event report records an abnormal event that occurred at "Chemical Industrial Park B Zone Steam Branch P-207". This time-stamped abnormal thermal energy event report is analyzed to extract key feature vectors. The key feature vectors include the anomaly type, occurrence location, intensity, and duration pattern. The anomaly type extracted from the report is "sudden drop in pipeline pressure", the occurrence location is the directed edge "E-207" between pipeline nodes "N-15" and "N-16", the intensity feature is that the pressure drops from 0.62 MPa to 0.18 MPa, and the duration pattern is "rapid drop followed by sustained low level for 120 seconds". In the historical case database, using key feature vectors as query criteria, similarity calculations are performed based on multidimensional feature space distance. The database stores all recorded abnormal events and their feature vectors from the past three years. The comprehensive distance between the current event's feature vector and the feature vector of each case in the historical database is calculated across four dimensions: "anomaly type," "location topological distance," "intensity deviation," and "duration ratio." All historical case entries whose distance to the current event's feature vector is less than a preset threshold of 0.15 are retrieved. These historical case entries are then sorted in descending order of similarity to form a list of similar historical events. This list contains five historical cases with case numbers C-1012, C-0883, C-1205, C-0941, and C-1120. It can be understood that the feature vector structure of the cases stored in the historical case database is uniform, and the formula for calculating the multidimensional feature space distance is: in: Represents the overall distance. and The exception type codes representing the current event and historical events are respectively, and the function... Calculate the difference in types (0 for the same, 1 for different). and They represent positional encoding and function, respectively. Calculate the shortest path distance based on the pipeline network topology map. and These represent the intensity quantization value and the function, respectively. Calculate the relative deviation of strength. and Representing duration and function respectively Calculate the duration ratio difference. to These are the weight coefficients corresponding to each dimension. See Table 1, which shows the feature vectors and distance calculation results for the current event and some historical cases.

[0030] Table 1: Example Table of Abnormal Event Feature Matching Retrieval For the identified similar past event records, the corresponding control response sequence is traced back to generate a preliminary response strategy set containing at least one control action. In specific implementation, the complete event processing log is read from each past event record in the list of similar historical events. For case C-1012, its event processing log is stored in the associated field of the database, recording the entire process of text and timestamp information from event occurrence, alarm, manual confirmation to execution of control operation. From the event processing log, all control operation instructions actually executed after the historical event occurred are extracted. Control operation instructions include valve opening adjustment, pump start / stop, standby heat source input, and load switching. Three operation instructions are extracted from the log of case C-1012: Instruction 1 (30 seconds after the event) is "Close the regulating valve V-207B, adjust the opening from 80% to 50%", Instruction 2 (45 seconds after the event) is "Start the standby booster pump P-B2", and Instruction 3 (180 seconds after the event) is "Switch the load of drying line H-3 from the main network". The extracted control operation commands are combined into a complete control response sequence for historical events according to their execution order. The complete control response sequence for case C-1012 is [command 1, command 2, command 3]. By summarizing the control response sequences corresponding to all similar historical events and removing duplicate or contradictory control actions, a preliminary response strategy set containing multiple optional control actions is formed. After summarizing the control response sequences of cases C-1012, C-0883, and C-0941, multiple control actions are obtained, including "closing the regulating valve V-207B slightly," "starting the standby booster pump P-B2," "cutting off the load of drying line H-3 from the main grid," "opening the main pipeline regulating valve V-Main significantly," and "switching to the standby steam source S-2." Among these, "starting the standby booster pump P-B2" appears in multiple cases and is retained as one of them. However, "opening the main pipeline regulating valve V-Main" and "closing the regulating valve V-207B slightly" have opposite effects on the same parameter in physical logic and are considered contradictory control actions. In the preliminary response strategy set, they are retained as different options for subsequent evaluation. In some embodiments, the parsing and instruction extraction of the event processing log are completed through natural language processing template matching or direct parsing of structured operation records. Optionally, the initial response strategy set is stored in the form of a list, where each item contains a description of a regulatory action, the historical case number from which the action originated, and the execution order of the action in the original sequence.

[0031] See Figure 4In the assessment of recovery time and energy consumption increase in industrial green microgrid control strategies, the performance of various control actions can be quantitatively analyzed using dual-axis indicators. The bar chart represents recovery time (unit: seconds), reflecting the system's response speed from an abnormal state back to a predetermined stable range; the line represents energy consumption increase (unit: %), measuring the additional energy consumption resulting from the execution of control actions. From the perspective of recovery time, the longest recovery time was for "cutting off the load of drying line H-3," reaching 100 seconds, indicating a wide impact on the system state and a long adjustment cycle. The shortest recovery time was for "starting the standby booster pump P-B2," at only 45 seconds, demonstrating its advantage in rapid response to pressure anomalies. From the perspective of energy consumption increase, the largest energy consumption increase was for "starting the standby booster pump P-B2," reaching 15%, due to the high energy consumption characteristics of pump startup and operation. The smallest energy consumption increase was for "closing the regulating valve V-207B," at only 5%, indicating that the energy cost of valve throttling actions is relatively controllable. Considering both indicators, "closing the control valve V-207B" offers moderate recovery time (85 seconds) and energy consumption increase (5%), making it a sound choice that balances efficiency and cost. "Opening the main pipeline control valve V-Main" has a recovery time of 90 seconds and an energy consumption increase of 7%, showing a relatively balanced performance. "Switching to the backup steam source S-2" has a recovery time of 60 seconds and an energy consumption increase of 12%, offering an advantage in recovery speed but at a higher energy cost. The figure provides a quantitative basis for selecting optimized control schemes, allowing for prioritization and combination of various control actions based on preset optimization goals such as "fastest recovery to stability" or "lowest overall control energy consumption."

[0032] In one embodiment of the present invention, a preliminary set of response strategies is input into a multimodal energy flow topology graph for prospective extrapolation to evaluate the impact trajectory of each control action on the overall thermal network state. The preliminary set of response strategies originates from the case matching results of an abnormal event of "exceeding the limit of the outlet temperature of the thermal storage unit". It includes four candidate control actions: "opening the bypass regulating valve V-Bypass", "starting the circulating pump P-Circ", "putting the standby electric heater E-H1 into operation", and "reducing the set value of the process load L-5". One control action to be evaluated is selected from the preliminary set of response strategies. For example, the action "opening the bypass regulating valve V-Bypass" is selected first. Based on the current state of the multimodal energy flow topology graph, the control action "opening the bypass regulating valve V-Bypass" to be evaluated is applied to update the state parameters of the relevant nodes or directed edges. Specifically, the "valve opening" attribute of the directed edge corresponding to "regulating valve V-Bypass" in the multimodal energy flow topology graph is updated from the current 40% to 70%, and the "local resistance coefficient" of the directed edge is updated synchronously according to the valve characteristic curve. Using the updated state as the new initial conditions, the energy flow state evolution simulation was re-executed to predict the evolution of the network state over a future period. The simulation duration was 60 minutes, with a time step of 1 minute. The simulation calculated the temporal changes in the outlet temperature of the thermal storage unit, the pressure of key nodes in the pipeline network, and the overall thermal balance of the heating network under the new valve opening. The evolution process obtained from the simulation was analyzed, recording the time required for key parameters to recover to the safe operating threshold range, whether any new abnormal fluctuations were triggered during the process, and the final stable state of the system. The analysis showed that after applying the "open the bypass regulating valve V-Bypass" action, the outlet temperature of the thermal storage unit fell back below the safe threshold 18 minutes after the start of the simulation. No other node pressure exceeded the limit during the process, and the system finally reached a new equilibrium with the outlet temperature slightly below the upper limit. Repeat the above steps of "selecting actions - updating status - resimulating - analyzing records" until each control action in the initial response strategy set has completed forward-looking simulation and impact assessment. Then, perform the same assessment process for the three actions of "starting the circulating pump P-Circ", "putting the standby electric heater E-H1 into operation", and "reducing the set value of the process load L-5" in sequence, and record the assessment results for each.

[0033] Based on the evaluation results, optimized control schemes that enable the energy flow network to recover to a predetermined stable range are selected. In specific implementation, optimization objectives are set: to restore the system to stability as quickly as possible, to minimize overall regulation energy consumption, or to minimize disturbance to the production process. In this example, a multi-objective weighted approach is adopted, and the optimization objective is set as "minimizing the weighted sum of regulation energy consumption and production disturbance while ensuring recovery." The forward-looking simulation evaluation results of all control actions are compared, and each control action is scored and ranked according to the optimization objective. The score is calculated based on the specific numerical values ​​in the simulation results of each action. A comprehensive score is calculated using an optimization objective function and then ranked. The single control action with the highest score, or a combination of multiple control actions that do not conflict in the evaluation and have synergistic effects, is selected to constitute an optimized control scheme. The scoring ranking shows that the action "starting the circulating pump P-Circ" has the highest comprehensive score, while the action "opening the bypass regulating valve V-Bypass" does not conflict with it in the simulation and can accelerate the initial temperature drop. Therefore, these two actions are combined to constitute an optimized control scheme. In some embodiments, the optimization objective function can be formally expressed as: in: This represents a comprehensive evaluation score for a regulatory action (the lower the score, the better). The function represents the recovery time caused by the action. Map it to a time cost score. The function represents the adjustment energy consumed by this action. Map it to an energy cost score. The function represents the production load disturbance caused by this action. Map it to a disturbance cost score. These are the weighting coefficients for each component cost. The verification and optimization control scheme ensures that all key parameters of the multimodal energy flow topology remain within the predetermined stable range at the end of the simulation. A final overall simulation verification of the combined schemes "starting the circulating pump P-Circ" and "opening the bypass regulating valve V-Bypass" is conducted. Simulation results show that the outlet temperature of the thermal storage unit returns to a safe range 15 minutes after the scheme is implemented, and by the end of the simulation (60 minutes), the temperature and pressure parameters of all nodes in the entire network remain within their respective predetermined stable ranges.

[0034] See Figure 5During the temperature and pressure trend monitoring in the combined scheme verification phase, the evolution of the thermal storage unit outlet temperature and the pressure at key nodes in the pipeline network directly reflects the implementation effect of the optimized control scheme. Specifically, the thermal storage unit outlet temperature (solid line) decreased linearly from an initial 95℃, rapidly falling back to below the temperature safety threshold of 85℃ (dashed line) after 15 minutes, and stabilizing at around 55℃ after 60 minutes; the pressure at key nodes in the pipeline network (dashed line with dots) showed a single-peak pattern of first rising and then falling, reaching a peak of approximately 1.31MPa around 25 minutes, and then gradually falling back to around 1.10MPa, without triggering pressure over-limits throughout the process. This combined approach involves the coordinated action of "starting the circulating pump P-Circ" and "opening the bypass regulating valve V-Bypass." Its control logic can be verified by the curve characteristics: opening the bypass valve directly increases the flow rate in the heat dissipation circuit, accelerating the initial drop in the outlet temperature of the heat storage unit; starting the circulating pump enhances the overall heat exchange efficiency of the pipeline network, further increasing the temperature drop rate, while optimizing the flow field distribution to avoid local pressure anomalies. From a temporal perspective, the temperature recovers to below the safe threshold within 15 minutes, significantly shorter than the recovery time of a single-action approach (such as "opening the bypass regulating valve V-Bypass," which requires 18 minutes). Furthermore, the pressure curve is smooth without abrupt changes, verifying the effectiveness of this combined approach under the multi-objective optimization of "fastest recovery to stability, minimizing energy consumption and production disturbance."

[0035] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A thermal energy management method based on an industrial green microgrid, characterized in that, include: It receives raw monitoring information from heat energy generation units, heat energy consumption equipment and heat energy storage facilities, and forms a data set containing multi-source heterogeneous signals; The dataset is fused and cleaned to remove noisy data and fill information gaps, and a unified formatted energy flow record containing time-series labels is established. Based on the unified formatted energy flow record, a multimodal energy flow topology graph is constructed; Within the framework of the multimodal energy flow topology, energy flow state evolution simulation is performed to generate thermal energy flow state evolution data containing multiple future time sections; The abnormal thermal energy fluctuation characteristics that violate the preset evolution law are extracted from the thermal energy flow evolution data, and their spatiotemporal attributes are labeled to form an abnormal thermal energy event report with time stamp. The historical running case library is invoked to match and search the time-stamped abnormal thermal event reports, and to identify past event records with similar characteristics; For the identified similar past event records, trace back their corresponding regulatory response sequences to generate a preliminary set of response strategies containing at least one regulatory action; The preliminary response strategy set is input into the multimodal energy flow topology diagram for forward-looking simulation, and the impact trajectory of each regulation action on the overall thermal network state is evaluated. Based on the evaluation results, an optimized control scheme that enables the energy flow network to return to the predetermined stable range was selected.

2. The thermal energy management method based on an industrial green microgrid according to claim 1, characterized in that, The process of receiving raw monitoring information from heat generation units, heat consumption devices, and heat storage facilities to form a data set containing multi-source heterogeneous signals specifically includes: Information on heat generation capacity and medium temperature is collected from distributed renewable energy devices; Real-time heat energy demand and medium pressure information are collected from industrial heat-using process nodes; Collect information on storage levels, inlet and outlet temperatures, and flow rates from phase change energy storage tanks or hot water storage tanks; Collect pipeline temperature and pressure information from key nodes of the heat exchange network; The collected information on heat generation power, medium temperature, real-time heat energy demand, medium pressure, inventory level, inlet and outlet temperature, flow rate, pipeline temperature and pressure is associated and packaged according to the source equipment code and timestamp to form an initial data set.

3. The thermal energy management method based on an industrial green microgrid according to claim 2, characterized in that, The dataset is fused and cleaned to remove noisy data and fill information gaps, creating a uniformly formatted energy flow record containing time-series labels. Specifically, this includes: Outlier detection is performed on each signal sequence in the dataset, and a sliding window comparison method is used to identify and remove noisy data points that deviate from the neighborhood mean. For periods of data loss caused by sensor failure or communication interruption, data from nearby functionally similar nodes are used to fill the information gaps through time series interpolation algorithms. Data from different sources, after being cleaned and filled, are transformed and aligned according to a unified time base and physical dimensions; A time-series label is attached to each aligned data point to form a uniformly formatted energy flow record with a consistent structure.

4. A thermal energy management method based on an industrial green microgrid according to claim 3, characterized in that, Based on the unified formatted energy flow record, a multimodal energy flow topology graph is constructed, specifically including: The multimodal energy flow topology diagram reflects the dynamic mapping relationship between heat energy generation, transmission, use and storage; The heat generation unit, heat consumption device, and heat storage facility are abstracted as nodes in the multimodal energy flow topology diagram; The pipes, valves, and heat exchangers connecting the nodes are abstracted as directed edges in the multimodal energy flow topology graph; Assign attributes to each node, including at least node type, rated capacity, and real-time status value; Assign attributes to each directed edge, including at least connectivity, transmission medium, thermal resistivity, and current flow rate; Based on the temporal changes in the uniformly formatted energy flow record, the real-time state values ​​of the nodes and the directed edges are dynamically updated, so that the multimodal energy flow topology graph can reflect the real-time configuration and state of the thermal energy network.

5. A thermal energy management method based on an industrial green microgrid according to claim 4, characterized in that, Within the framework of the multimodal energy flow topology, energy flow state evolution simulation is performed to generate thermal energy flow state evolution data containing multiple future time segments, specifically including: The initial conditions are the real-time state values ​​of each node in the multimodal energy flow topology diagram at the current moment; Based on the predicted output curve of the heat energy generating unit, the production plan of the heat energy consuming equipment, and the ambient temperature forecast, the boundary conditions and disturbance inputs for the future simulation period are set. Based on the fundamental principles of energy conservation and heat transfer, iterative calculations with discrete time steps are performed on the multimodal energy flow topology graph. Simulate the conduction, convection, storage and dissipation of heat energy in the network, and deduce the sequence of changes in the state parameters of each node and directed edge over a future period of time. The snapshots of the entire network state parameters calculated at each discrete time step are stored to form the thermal energy flow evolution data.

6. A thermal energy management method based on an industrial green microgrid according to claim 5, characterized in that, The abnormal thermal energy fluctuation characteristics that violate the preset evolution law are extracted from the thermal energy flow evolution data, and their spatiotemporal attributes are labeled to form a time-stamped abnormal thermal energy event report, specifically including: Set safe operating threshold ranges for node temperature change rate, pipeline pressure difference, and energy storage charge / discharge rate; Traverse the thermal energy flow evolution data and compare each parameter with the corresponding safe operation threshold range at each time section; Identify abnormal fluctuations where parameter values ​​continuously exceed the safe operating threshold range or experience short-term, drastic jumps. Record the start time, duration, node and directed edge numbers involved, and specific parameters exceeding the threshold of the abnormal fluctuations; All recorded abnormal fluctuation information is summarized and structured into the aforementioned time-stamped abnormal thermal event report.

7. A thermal energy management method based on an industrial green microgrid according to claim 6, characterized in that, The process of calling the historical running case library to match and search the time-stamped abnormal thermal event reports, and identifying past event records with similar characteristics, specifically includes: The time-stamped abnormal thermal event report is parsed to extract key feature vectors, which include anomaly type, occurrence location, intensity, and duration patterns. In the historical case database, similarity calculation based on multidimensional feature space distance is performed using the key feature vector as the query condition. Retrieve all historical case entries whose distance from the current event's feature vector is less than a preset threshold; The retrieved historical case entries are sorted in descending order of similarity to form a list of similar historical events.

8. A thermal energy management method based on an industrial green microgrid according to claim 7, characterized in that, The step involves tracing back the corresponding regulatory response sequences of the identified similar past event records to generate a preliminary set of response strategies containing at least one regulatory action, specifically including: Read the complete event processing log from each past event record in the list of similar historical events; Extract all control operation instructions actually executed after the occurrence of the historical event from the event processing log. The control operation instructions include valve opening adjustment, pump start and stop, standby heat source input and load switching. The extracted control operation instructions are combined into a complete control response sequence for historical events according to their execution order. By summarizing the control response sequences corresponding to all similar historical events and removing duplicate or contradictory control actions, a preliminary set of response strategies containing multiple optional control actions is formed.

9. A thermal energy management method based on an industrial green microgrid according to claim 8, characterized in that, The preliminary response strategy set is input into the multimodal energy flow topology graph for prospective extrapolation, evaluating the impact trajectory of each regulatory action on the overall thermal network state, specifically including: Select one regulatory action to be evaluated from the initial set of response strategies; Based on the current state of the multimodal energy flow topology graph, the control action to be evaluated is applied to update the state parameters of the relevant nodes or directed edges. Using the updated state as the new initial conditions, the energy flow state evolution simulation is re-executed to predict the evolution of the network state over a future period of time. Analyze the evolution process, record the time required for key parameters to recover to the safe operating threshold range, whether new abnormal fluctuations are triggered during the process, and the final stable state of the system; Repeat the steps until every regulatory action in the initial response strategy set has been prospectively analyzed and its impact assessed; Based on the evaluation results, optimized control schemes that enable the energy flow network to return to a predetermined stable range are selected, specifically including: Set optimization objectives, which are to enable the system to recover to stability as quickly as possible, minimize overall adjustment energy consumption, or minimize disturbance to the production process; Compare the forward-looking projection and evaluation results of all control actions, and score and rank each control action according to the optimization objectives. Select the single control action with the highest score, or combine multiple control actions that do not conflict in the evaluation and have synergistic effects to form an optimized control scheme; The optimized control scheme was verified to ensure that all key parameters of the multimodal energy flow topology were within the predetermined stable range at the end of the simulation.

10. A thermal energy management system based on an industrial green microgrid, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the thermal energy management method based on an industrial green microgrid as described in any one of claims 1 to 9.

Images