Energy-saving management system for public institutions based on AI intelligent operation and maintenance

Abstract

The application belongs to the technical field of energy management, and discloses a public institution energy saving management system based on AI intelligent operation and maintenance, which comprises: a multi-source data acquisition module for acquiring building multi-source operation data; a coupling feature analysis module for constructing a building coupling feature data set based on the building multi-source operation data; a cold-heat hedging evaluation module for evaluating cold-heat hedging risks based on the building coupling feature data set and generating R cold-heat hedging risk regions; a hierarchical scheduling generation module for hierarchically sorting the R cold-heat hedging risk regions and generating a cold-heat hedging priority relief list and a cold-heat hedging secondary regulation list; and a cold-heat hedging regulation module for performing cold-heat hedging hierarchical adaptive regulation on the cold-heat hedging priority relief list and the cold-heat hedging secondary regulation list. Through the cold-heat hedging identification and adaptive regulation mechanism, the application solves the problems of inaccurate cold-heat supply-demand matching and lack of targeted regulation strategies, and improves the energy saving effect of building energy management.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy management, more particularly, the present application relates to a public institution energy saving management system based on AI intelligent operation and maintenance. BACKGROUND

[0002] In the process of energy system management of public institution building groups, how to realize the accurate allocation and dynamic balance of cold and heat resources is the key target to guarantee indoor comfort and improve energy utilization efficiency. With the development of large building group construction, the coupling relationship between building sub-regions in air supply path, return air circuit, water system ring connection and building envelope heat transfer is becoming more and more complex, and the dynamic change of cold and heat load presents stronger correlation and propagation. Especially in the background of coexistence of multi-functional areas and significant energy demand difference, cold and heat hedging phenomenon frequently occurs, which not only reduces system energy efficiency, but also easily causes cold and heat fluctuation, frequent start and stop of equipment and other operation and maintenance problems. In order to solve the above problems, the industry gradually introduces an energy saving management system with AI algorithm as the core, tries to combine operation state parameters, and realizes adaptive energy consumption regulation and control based on big data driving.

[0003] However, in the process of identifying and intervening cold and heat hedging, the prior art often relies on static rule configuration or coarse-grained energy consumption aggregation data, which cannot accurately depict the spatial coupling structure and load interaction characteristics between building sub-regions, resulting in coarse cold and heat hedging identification granularity, insufficient timeliness, and difficulty in realizing fine differentiated diagnosis and hierarchical regulation. In addition, the existing scheme lacks systematic evaluation of regional cooling efficiency, actual regulation margin and regulation intention Figure One consistency, and the regulation strategy is prone to problems such as inaccurate matching of cold and heat supply and demand, lack of targetedness of regulation strategy and poor energy saving effect.

[0004] Therefore, it is urgent to develop a graph structured cold and heat hedging diagnosis and regulation system integrating building space structure, operation data and semantic label to realize more targeted and intelligent adaptive regulation of cold and heat hedging. In view of this, the present application provides a public institution energy saving management system based on AI intelligent operation and maintenance to solve the above problems. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art and achieve the above-mentioned purposes, the present application provides the following technical scheme: a public institution energy saving management system based on AI intelligent operation and maintenance, comprising:

[0006] A multi-source data acquisition module for acquiring building multi-source operation data;

[0007] The coupling feature analysis module performs semantic fusion and coupling evaluation based on the building multi-source operation data, and constructs a building coupling feature dataset; the building coupling feature dataset includes a cold-heat hedging difference graph, a regulation intention difference degree matrix, and cold-heat demand directions, dynamic load semantic labels, and regional comprehensive cooling efficiency coefficients corresponding to N building sub-regions;

[0008] The cold-heat hedging evaluation module performs cold-heat hedging risk evaluation based on the building coupling feature dataset, obtains cold-heat hedging risk evaluation scores corresponding to the N building sub-regions, and identifies R cold-heat hedging risk regions;

[0009] The hierarchical scheduling generation module is configured to sort the R cold-heat hedging risk regions according to the functional partition attributes, the building sub-region areas, and the cold-heat hedging duration types, and generate a cold-heat hedging priority relief list and a cold-heat hedging secondary regulation list.

[0010] The cold-heat hedging regulation module is configured to perform cold-heat hedging hierarchical adaptive regulation on the cold-heat hedging priority relief list and the cold-heat hedging secondary regulation list in combination with the air conditioning system operation margin parameter set, the cold-heat hedging difference graph, and the regulation intention difference degree matrix.

[0011] Further, the method of performing cold-heat hedging hierarchical adaptive regulation includes:

[0012] For each building sub-region in the cold-heat hedging priority relief list, denoted as a priority relief building sub-region, a subgraph directly having a spatial coupling relationship with the priority relief building sub-region is obtained from the cold-heat hedging difference graph, denoted as a cold-heat hedging associated subgraph; a matrix element directly associated with the priority relief building sub-region on the graph structure edge is extracted from the regulation intention difference degree matrix, and a regulation intention difference degree set is constructed; the air conditioning system operation margin parameter set, the cold-heat hedging associated subgraph, and the regulation intention difference degree set are input into a priority relief parameter setting model to obtain an instant regulation instruction set; the instant regulation instruction set includes a target air supply amount set value, a terminal water valve target opening degree, a host load instantaneous correction gain, and an energy release instruction of an energy storage unit.

[0013] The instant regulation instruction set is sent to the corresponding control unit for execution.

[0014] When the cold-heat hedging priority relief list is completed, the cold-heat hedging secondary regulation list is updated and cold-heat hedging hierarchical adaptive regulation is performed.

[0015] Further, the method of updating the cold-heat hedging secondary regulation list and performing cold-heat hedging hierarchical adaptive regulation includes:

[0016] For each building sub-area in the cold-heat hedging secondary regulation list, denoted as a secondary mitigation building sub-area, re-evaluate and update the corresponding cold-heat hedging risk assessment score of the secondary mitigation building sub-area;

[0017] Remove the secondary mitigation building sub-area with a cold-heat hedging risk assessment score not greater than a preset cold-heat hedging risk score threshold from the cold-heat hedging secondary regulation list to obtain an updated cold-heat hedging secondary regulation list;

[0018] Obtain a latest air conditioning system operation margin parameter set, and construct a cold-heat hedging risk assessment score set based on the cold-heat hedging risk assessment scores in the updated cold-heat hedging secondary regulation list;

[0019] Input the air conditioning system operation margin parameter set and the cold-heat hedging risk assessment score set into a secondary mitigation parameter setting model to obtain a planned regulation instruction set, and issue the planned regulation instruction set to a corresponding control unit for execution; the planned regulation instruction set is of the same type of regulation instruction as the immediate regulation instruction set.

[0020] Further, the obtaining method of the cold-heat hedging priority mitigation list and the cold-heat hedging secondary regulation list comprises:

[0021] For R cold-heat hedging risk areas, obtain the functional partition attribute, cold-heat hedging duration proportion, building sub-area area, and cold-heat hedging risk assessment score of each cold-heat hedging risk area;

[0022] According to a preset functional partition attribute matching table, convert the functional partition attribute into a functional partition attribute value; compare the cold-heat hedging duration proportion with a preset duration proportion threshold to obtain a cold-heat hedging duration type and convert it into a cold-heat hedging duration type value; the cold-heat hedging duration type is continuous cold-heat hedging or intermittent cold-heat hedging;

[0023] Input the cold-heat hedging risk assessment score, the functional partition attribute value, the cold-heat hedging duration type value, and the building sub-area area into a cold-heat hedging priority assessment model to obtain a corresponding cold-heat hedging priority score;

[0024] Compare the cold-heat hedging priority score with a preset cold-heat hedging priority score threshold to respectively divide to obtain the cold-heat hedging priority mitigation list and the cold-heat hedging secondary regulation list;

[0025] Sort the cold-heat hedging priority mitigation list and the cold-heat hedging secondary regulation list in descending order according to the corresponding cold-heat hedging priority scores to obtain the final cold-heat hedging priority mitigation list and the cold-heat hedging secondary regulation list.

[0026] Further, the obtaining method of the R cold-heat hedging risk areas comprises:

[0027] inputting the building coupling feature dataset into a pre-trained cold-heat hedging risk assessment model to obtain cold-heat hedging risk assessment scores corresponding to the N building sub-regions one by one;

[0028] For the N cold-heat hedging risk assessment scores, the cold-heat hedging risk assessment scores are compared with preset cold-heat hedging risk score thresholds one by one, and if the cold-heat hedging risk assessment score is greater than the cold-heat hedging risk score threshold, the building sub-region corresponding to the cold-heat hedging risk assessment score is marked as a cold-heat hedging risk region;

[0029] After the N building sub-regions are determined, the total number of cold-heat hedging risk regions is summarized, and the total number is denoted as R, thereby obtaining R cold-heat hedging risk regions.

[0030] Further, the method for constructing the building coupling feature dataset comprises:

[0031] Based on temperature and humidity, the cold-heat demand direction corresponding to the N building sub-regions is determined, and the cold-heat demand direction includes cold demand, heat demand, and comfortable demand.

[0032] The air valve opening degree is matched with the pre-constructed air valve opening degree matching table to obtain the air valve opening degree position corresponding to the N building sub-regions; the air valve opening degree position includes low opening degree, medium opening degree, and high opening degree.

[0033] Based on the cold-heat demand direction, the air valve opening degree position, the equipment power consumption, and the personnel density, the dynamic load semantic label corresponding to the N building sub-regions is set;

[0034] According to the difference degree between the spatial coupling relationship of the N building sub-regions and the dynamic load semantic label, a cold-heat hedging difference map and a regulation intention difference degree matrix are constructed;

[0035] Based on the air conditioner air volume, the air conditioner water system temperature difference, and the valve opening degree, the regional comprehensive cooling efficiency coefficient corresponding to the N building sub-regions is constructed;

[0036] The cold-heat hedging difference map, the regulation intention difference degree matrix, and the cold-heat demand direction, the dynamic load semantic label, and the regional comprehensive cooling efficiency coefficient corresponding to the N building sub-regions are constructed into a building coupling feature dataset.

[0037] Further, the method for determining the cold-heat demand direction corresponding to the N building sub-regions based on temperature and humidity comprises:

[0038] For each building sub-region, if the temperature of the building sub-region is higher than the upper limit of the comfortable temperature and the humidity is higher than the upper limit of the comfortable humidity, the cold-heat demand direction is marked as cold demand, and the current process is ended;

[0039] If the temperature of the building sub-area is lower than the lower limit of the comfortable temperature, mark the cold-heat demand direction as heat demand, and end the current process;

[0040] Mark the cold-heat demand direction as comfortable no demand, and end the current process.

[0041] Further, the method for setting the dynamic load semantic labels corresponding to the N building sub-areas comprises:

[0042] For each building sub-area, if the cold-heat demand direction of the building sub-area is cold demand, judge whether the equipment power consumption and the personnel density exceed the corresponding threshold values, if the result of the judgment is that the threshold values are exceeded, judge whether the air valve opening degree is high, if the air valve opening degree is high, mark the dynamic load semantic label as requiring forced refrigeration, if the air valve opening degree is not high, mark the dynamic load semantic label as requiring moderate refrigeration, if the result of the judgment is that the threshold values are not exceeded, mark the dynamic load semantic label as requiring moderate refrigeration;

[0043] If the cold-heat demand direction of the building sub-area is heat demand, that is, the system has excessive cooling, judge whether the air valve opening degree is high, if the air valve opening degree is high, mark the dynamic load semantic label as requiring reduced refrigeration, if the air valve opening degree is not high, mark the dynamic load semantic label as requiring moderate refrigeration;

[0044] If the cold-heat demand direction of the building sub-area is comfortable no demand, mark the dynamic load semantic label as no adjustment demand.

[0045] Further, the method for constructing the cold-heat hedge difference graph and the adjustment intention difference degree matrix comprises:

[0046] Match the dynamic load semantic labels with the pre-constructed dynamic load semantic label matching table to obtain the dynamic load semantic label values corresponding to the N building sub-areas;

[0047] Map and construct the N building sub-areas into a graph structure node set;

[0048] Construct the air supply path coupling edges, the return air convergence coupling edges, the waterway ring connection coupling edges and the shared wall heat transfer edges corresponding to the N building sub-areas into a graph structure edge set;

[0049] For each graph structure edge in the graph structure edge set, construct the graph structure edge and the graph structure edges adjacent to the graph structure edge into a graph structure edge combination, judge whether the dynamic load semantic label values corresponding to the graph structure edge combination are the same, if not, set the adjustment intention difference degree of the graph structure edge combination as the absolute value of the numerical difference between the dynamic load semantic label values corresponding to the graph structure edge combination, if the same, set the adjustment intention difference degree of the graph structure edge combination to zero;

[0050] write the adjustment intention difference into the adjustment intention difference matrix, and the row and column coordinate positions of the adjustment intention difference matrix correspond to the number positions formed by the combination of the graph structure edges;

[0051] When all the graph structure edges in the graph structure edge set are processed, the assigned graph structure edge set is obtained.

[0052] The assigned graph structure edge set and the joint graph structure node set are combined to form a cold-heat arbitrage difference graph.

[0053] Further, the method for constructing the regional comprehensive cooling efficiency coefficient corresponding to the N building sub-areas comprises the following steps:

[0054] For each building sub-area, normalize the air conditioner air volume, air conditioner water system temperature difference and valve opening to a preset normalization interval range, and obtain the air side delivery factor, water side heat exchange factor and valve adjustment factor respectively; based on the air side delivery factor, water side heat exchange factor and valve adjustment factor, the regional comprehensive cooling efficiency coefficient is calculated.

[0055] When all the building sub-areas are processed, the regional comprehensive cooling efficiency coefficient corresponding to the N building sub-areas is obtained.

[0056] Further, the building multi-source operation data comprises temperature, humidity, air conditioner air volume, air conditioner water system temperature difference, valve opening, air valve opening, equipment power consumption and personnel density corresponding to the N building sub-areas.

[0057] Compared with the prior art, the technical effects and advantages of the public institution energy saving management system based on AI intelligent operation and maintenance of the present application are as follows:

[0058] The building coupling feature data set constructed in the present application comprehensively considers the cold and heat demand direction, dynamic load semantic label, regional comprehensive cooling efficiency coefficient, and coupling graph information formed by the spatial path and heat conduction relationship of each building sub-area, so that the evaluation of cold and heat arbitrage not only depends on local indicators, but also can capture the cold and heat imbalance propagation law across regions, significantly improving the accuracy of cold and heat arbitrage identification. By constructing the cold and heat arbitrage difference graph and the adjustment intention difference matrix, the cold and heat supply and demand difference between building sub-areas can be directly quantified, providing data support for the reasonable formulation of subsequent cold and heat conflict regulation strategies.

[0059] Further, the cold-heat hedging risk area is dynamically identified according to the cold-heat hedging risk evaluation score output by the cold-heat hedging risk evaluation model, and combined with key indicators such as functional partition attributes, building sub-area area and cold-heat hedging duration type, a cold-heat hedging priority relief list and a cold-heat hedging secondary regulation list are constructed, so as to realize automatic deployment of differentiated cold-heat hedging regulation strategies. Among them, the priority relief parameter setting model generates immediate regulation instructions and quickly responds to sudden cold-heat hedging problems; the secondary relief parameter setting model formulates a planned regulation strategy based on the updated cold-heat hedging risk level and air conditioning system operation margin parameters, ensuring the stability and energy saving of the overall system regulation.

[0060] In summary, the cold-heat hedging identification and adaptive regulation mechanism driven by data and semantic fusion effectively solves the problems of inaccurate cold-heat supply-demand matching and lack of targeted regulation strategies in traditional methods, improves the operation efficiency, response ability and energy saving effect of the public institution building energy management system, and has good popularization and application value. BRIEF DESCRIPTION OF DRAWINGS

[0061] Figure 1 It is a schematic diagram of the public institution energy saving management system based on AI intelligent operation and maintenance of embodiment 1 of the application;

[0062] Figure 2 It is a flow chart of the public institution energy saving management method based on AI intelligent operation and maintenance of embodiment 2 of the application;

[0063] Figure 3 It is a flow chart of the method for cold-heat hedging hierarchical adaptive regulation;

[0064] Figure 4 It is a flow chart of the method for updating the cold-heat hedging secondary regulation list and performing cold-heat hedging hierarchical adaptive regulation;

[0065] Figure 5 It is a flow chart of the method for obtaining the cold-heat hedging priority relief list and the cold-heat hedging secondary regulation list;

[0066] Figure 6 It is a flow chart of the method for constructing the building coupling feature data set. DETAILED DESCRIPTION

[0067] The technical solutions in the embodiments of the present application will be described in detail, clearly and completely below with reference to the drawings in the embodiments of the present application. It should be particularly noted that the specific embodiments described below are only used to better illustrate and describe the technical solutions of the present application, and are intended to enable those skilled in the art to better understand and implement the present application, and should not be understood as limiting the protection scope of the present application. Those skilled in the art can modify, adjust or equivalently replace the present application according to the content disclosed in the present application without departing from the spirit and essence of the present application, and these should be regarded as the protection scope of the present application.

[0068] Embodiment 1:

[0069] Please refer to Figure 1 As shown in the figure, the embodiment discloses a public institution energy-saving management system based on AI intelligent operation and maintenance, comprising a multi-source data acquisition module, a coupling feature analysis module, a cold and hot hedging evaluation module, a hierarchical scheduling generation module and a cold and hot hedging regulation and control module. Each module is connected by wire and / or wireless to realize data transmission.

[0070] The multi-source data acquisition module is used for acquiring building multi-source operation data; the building multi-source operation data includes temperature, humidity, air conditioning air volume, air conditioning water system temperature difference, valve opening, air valve opening, equipment power consumption and personnel density corresponding to N building sub-regions.

[0071] Specifically, the multi-source data acquisition module is used for performing distributed data collection operation for multiple functional sub-regions in the building operation environment to realize comprehensive perception of building energy consumption state and environmental thermal state. The building multi-source operation data collected by the multi-source data acquisition module specifically includes but is not limited to temperature, humidity, air conditioning air volume, air conditioning water system temperature difference, valve opening, air valve opening, equipment power consumption and personnel density, and each type of data has a clear data source path and system integration interface.

[0072] Among them, the temperature is the real-time collection result of the environmental air temperature in each building sub-region, which is obtained by deploying digital temperature sensors in each building sub-region and can be accessed to the public institution building automation control network through wired or wireless mode. Temperature is an important indicator reflecting the state of regional cold load or heat load, and is used to judge the coupling risk caused by uneven cold and heat in the region.

[0073] The humidity is the real-time observation value of the air relative humidity in each building sub-region, which is collected by a humidity sensor and can be installed in the air supply outlet, return air inlet or representative area of the room. Humidity can be used to comprehensively evaluate the thermal comfort index and model correction of condensation risk caused by high humidity, and also participate in the thermal field prediction as a physical constraint.

[0074] Air conditioning air volume reflects the air supply intensity and ventilation capacity of the air supply system of each building sub-region, and is obtained by air volume sensors installed in the air supply pipeline or variable air volume terminal devices VAV. Air conditioning air volume is used to evaluate the air flow delivery efficiency and energy distribution state, and participates in the calculation of air duct heat transfer, which is the core input of regional level heat energy delivery capacity modeling.

[0075] Air conditioning water system temperature difference represents the temperature difference of the chilled water or hot water system of the air conditioner in the inlet and outlet water circuit, which is measured by temperature sensors arranged at the inlet and outlet positions of the chilled water, and the difference is calculated by the control system. The air conditioning water system temperature difference can be used to estimate the cooling and heating load of the system in real time, which is an important parameter to support the building heat load prediction and energy consumption modeling.

[0076] Valve opening degree is used to describe the current opening degree state of the control valve in the terminal control unit in the water system, such as fan coil and cooling unit. It is collected by the position feedback unit carried by the electric regulating valve. The air valve opening degree reflects the current opening degree state of the air valve that adjusts the air volume of the air conditioning system. The collection method is similar to the valve opening degree. The valve opening degree and the air valve opening degree are used to judge the execution strength and coverage area of the current energy regulation strategy, and provide direct support for the physical state modeling of cooling and heating flow regulation.

[0077] Device power consumption reflects the current running power of various energy-consuming devices in the system, such as chiller, heat pump, circulating pump, air supply fan, etc. It is collected through the electric parameter interface of the intelligent electric meter, energy consumption metering unit or variable frequency control unit. Device power consumption is used to describe the energy efficiency state and operation load level of each building sub-region in system energy consumption modeling.

[0078] Personnel density reflects the intensity of personnel activity and heat source distribution in each building sub-region, which is usually estimated by deploying infrared human body sensors, video image recognition systems or positioning beacon devices such as ultra-wideband UWB or Bluetooth low energy BLE. Personnel density can dynamically correct the heat load model of the building sub-region, and is used to construct the unstructured heat source distribution mapping, which is an important input for constructing a real thermal environment.

[0079] Further, after the building multi-source operation data collection is completed, the multi-source data acquisition module enters the data preprocessing stage. According to the time stamp of each sensor device, a global time axis is constructed based on high-precision server time, and the original data of different frequencies and different time domains are aligned to a unified time point window through interpolation or synchronous sampling. For abnormal values or missing values caused by sensor error codes, packet loss or communication interruption, abnormal fluctuation points are identified by setting data integrity detection rules using sliding window statistical method, and missing data is filled in combination with historical data distribution or spatial adjacent point interpolation method to ensure data continuity and physical rationality.

[0080] After completing timing alignment and anomaly correction, the multi-source data acquisition module further performs spatial semantic attribution processing. For data collected from different subsystems, such as the air conditioning system, building automation system, and security system, spatial mapping is performed on each sensor data item based on the pre-integrated Building Information Model (BIM). The data is attributed to specific floors, rooms, or functional areas, achieving a unified spatial identification system. This spatial semantic mapping not only considers the physical location of the sensor, but also its service range, perception coverage boundary, and control link attribution to avoid data offset or misassociation caused by differences in naming conventions across systems.

[0081] Ultimately, the data processed by the multi-source data acquisition module is output as a multi-dimensional feature vector sequence structure, characterized by time synchronization, clear spatial attribution, semantic consistency, and controllable errors. By comprehensively acquiring and real-timely aggregating multi-source building operational data, this application meets the requirements of fusion modeling for spatial state continuity, physical drive consistency, and environmental semantic integrity, effectively supporting the operation of key functional modules such as thermal coupling identification, load drift diagnosis, and multi-objective collaborative optimization.

[0082] The coupling feature analysis module performs semantic fusion and coupling evaluation based on multi-source building operation data to construct a building coupling feature dataset; the building coupling feature dataset includes a cooling and heating hedging difference map, an adjustment intention difference matrix, and the cooling and heating demand directions, dynamic load semantic labels, and regional comprehensive cooling efficiency coefficients corresponding to N building sub-areas.

[0083] like Figure 6 As shown, the method for constructing the building coupling feature dataset includes:

[0084] Based on temperature and humidity, determine the heating and cooling demand directions corresponding to N building sub-areas, where the heating and cooling demand directions include cooling demand, heating demand, and no comfort demand;

[0085] Match the air valve opening with a pre-built air valve opening matching table to obtain the air valve opening gears corresponding to N building sub-areas; the air valve opening gears include low opening, medium opening and high opening;

[0086] Set dynamic load semantic labels for N building sub-areas based on cooling and heating demand direction, air valve opening gear, equipment power consumption, and occupant density;

[0087] According to the difference between the spatial coupling relationship of N building sub-areas and the dynamic load semantic labels, a heat and cold hedging difference map and a regulation intention difference matrix are constructed.

[0088] The N building sub-regions corresponding to the regional comprehensive cooling efficiency coefficient are constructed based on air conditioner air volume, air conditioner water system temperature difference and valve opening degree;

[0089] The cold-heat hedge difference graph, the adjustment intention difference matrix, and the N building sub-regions corresponding to the cold-heat demand direction, the dynamic load semantic label and the regional comprehensive cooling efficiency coefficient are constructed into the building coupling feature data set.

[0090] The method for determining the cold-heat demand direction of the N building sub-regions based on temperature and humidity includes:

[0091] For each building sub-region, if the temperature of the building sub-region is higher than the upper limit of the comfortable temperature, and the humidity is higher than the upper limit of the comfortable humidity, the cold-heat demand direction is marked as cold demand, and the current process is ended;

[0092] If the temperature of the building sub-region is lower than the lower limit of the comfortable temperature, the cold-heat demand direction is marked as heat demand, and the current process is ended;

[0093] The cold-heat demand direction is marked as comfortable no demand, and the current process is ended.

[0094] The method for setting the dynamic load semantic label corresponding to the N building sub-regions includes:

[0095] For each building sub-region, when the cold-heat demand direction of the building sub-region is cold demand, it is determined whether the equipment power consumption and the personnel density exceed the corresponding threshold value, if the result of the determination is that the threshold value is exceeded, it is determined whether the air valve opening degree is high, if the air valve opening degree is high, the dynamic load semantic label is marked as forced cooling is needed; if the air valve opening degree is not high, the dynamic load semantic label is marked as mild cooling is needed; if the result of the determination is that the threshold value is not exceeded, the dynamic load semantic label is marked as mild cooling is needed; the forced cooling is needed label indicates that the current cooling is severely insufficient, and forced cooling is needed; the mild cooling is needed label indicates that the current cooling and heating are basically matched, and mild cooling is needed.

[0096] When the cold-heat demand direction of the building sub-region is heat demand, that is, the system has excessive cooling, it is determined whether the air valve opening degree is high, if the air valve opening degree is high, the dynamic load semantic label is marked as cooling is needed to be reduced, if the air valve opening degree is not high, the dynamic load semantic label is marked as mild cooling is needed;

[0097] When the cold-heat demand direction of the building sub-region is comfortable no demand, the dynamic load semantic label is marked as no adjustment demand.

[0098] It should be noted that, for example, in this application, the threshold value corresponding to the power consumption of the equipment can be set according to the functional partition attribute, for example, the office area is set to 15W / m², the conference room is set to 18W / m², the reception area is set to 10W / m², and the data center room is set to 50W / m².

[0099] For example, the threshold value corresponding to the personnel density can be set according to the functional partition attribute, for example, the office area is set to 10 people / 100m², the conference room is set to 20 people / 100m², the reception area is set to 8 people / 100m², and the data center room is set to 3 people / 100m².

[0100] The construction method of the cold-heat hedging difference graph and the adjustment intention difference degree matrix includes:

[0101] The dynamic load semantic label is matched with the pre-constructed dynamic load semantic label matching table to obtain the dynamic load semantic label value corresponding to the N building sub-areas;

[0102] The N building sub-areas are mapped and constructed as a graph structure node set;

[0103] The supply air path coupling edge, the return air convergence coupling edge, the water loop ring connection coupling edge, and the shared wall heat transfer edge corresponding to the N building sub-areas are constructed into a graph structure edge set;

[0104] It should be noted that the supply air path coupling edge means that multiple building sub-areas share the same supply air branch pipe or the same variable air volume terminal device VAV; the return air convergence coupling edge means that the return air pipes flow together in the convergence node; the water loop ring connection coupling edge means that the terminal coil shares a water supply and return loop; and the shared wall heat transfer edge means that two building sub-areas are separated only by a lightweight wall or a glass curtain wall. In this application, VAV is a variable air volume terminal device, which mainly functions to adjust the air flow into the area according to the temperature demand in the area in the air conditioning system to achieve fine control of indoor temperature. The variable air volume terminal device is usually arranged at the end of the supply air branch pipe and serves one or more adjacent areas, with the functions of air valve adjustment, supply air control, and partial reheat. Therefore, when two sub-areas share the same variable air volume terminal device, it means that they have a coupling relationship in terms of supply air adjustment behavior, which is suitable as a supply air path coupling basis for cold-heat hedging risk identification.

[0105] For each graph structure edge in the graph structure edge set, the graph structure edge and the graph structure edge adjacent to the graph structure edge are constructed into a graph structure edge combination, it is judged whether the dynamic load semantic label values corresponding to the graph structure edge combination are same, if not same, the adjustment intention difference degree of the graph structure edge combination is set as the absolute value of the numerical difference between the dynamic load semantic label values corresponding to the graph structure edge combination; if same, the adjustment intention difference degree of the graph structure edge combination is set as zero; the adjustment intention difference degree is written into the adjustment intention difference degree matrix, the row and column coordinate positions in the adjustment intention difference degree matrix are corresponding positions of the numbers formed by the graph structure edge combination; for example, the numbers of the two edges in the graph structure edge combination are 1 and 3 respectively, the coordinate position formed by the graph structure edge combination is (1, 3), that is, the position in the adjustment intention difference degree matrix is the first row and the third column.

[0106] When all the graph structure edges in the graph structure edge set are processed, the graph structure edge set after assignment is obtained;

[0107] The joint graph structure node set and the graph structure edge set after assignment are constructed into a cold-heat arbitrage difference graph.

[0108] It should be noted that in the preferred embodiment of the present application, the constructed cold-heat arbitrage difference graph and the adjustment intention difference degree matrix are the core data structures in the present application, which are mainly used to reflect the real-time contradiction relationship between the building sub-regions in the cold-heat adjustment behavior, and further support the execution of key functions such as multi-region collaborative control, cold-heat load adjustment optimization and cold-heat arbitrage risk identification. Specifically, the adjustment intention difference degree matrix takes the graph structure edge combination as the index unit, expresses the inconsistency of the adjustment direction and adjustment intensity of the adjacent sub-regions at the current time through the numerical difference, and significantly improves the quantifiability and identifiability of the cold-heat arbitrage behavior.

[0109] Further, the cold-heat arbitrage difference graph realizes the graphical modeling of the cold-heat adjustment contradiction path in the building by retaining the building physical topology structure and weighting according to the difference degree of each graph structure edge. The cold-heat arbitrage difference graph not only reflects the existence of the building energy transmission channel, i.e. the air supply path coupling edge, the return air convergence coupling edge, the waterway ring connection coupling edge and the shared wall heat transfer edge in the present application, but also integrates the adjustment state difference on each channel, which can accurately locate the spatial region and cause path of local cold-heat arbitrage, i.e. quantify the potential cold-heat arbitrage risk path between adjacent regions, and provide visual analysis support for the control strategy. The cold-heat arbitrage risk region can be identified based on the cold-heat arbitrage difference graph, a cold-heat arbitrage priority relief list and a cold-heat arbitrage secondary adjustment list are constructed, directional intervention and strategy adjustment on the cold-heat arbitrage risk region are realized, thereby effectively improving the pertinence of system response and the accuracy of energy saving control.

[0110] The method for constructing the regional comprehensive cooling efficiency coefficient corresponding to the N building sub-areas comprises the following steps:

[0111] For each building sub-area, the air conditioning air volume, the air conditioning water system temperature difference and the valve opening degree are normalized to a preset normalization interval range, and the air side delivery and distribution factor, the water side heat exchange factor and the valve adjustment factor are obtained respectively; the regional comprehensive cooling efficiency coefficient is calculated based on the air side delivery and distribution factor, the water side heat exchange factor and the valve adjustment factor.

[0112] When all the building sub-areas are processed, the regional comprehensive cooling efficiency coefficients corresponding to the N building sub-areas are obtained.

[0113] The method for calculating the regional comprehensive cooling efficiency coefficient comprises the following steps:

[0114]

[0115] is the regional comprehensive cooling efficiency coefficient corresponding to the nth building sub-area, is the air side delivery and distribution factor corresponding to the nth building sub-area, is the water side heat exchange factor corresponding to the nth building sub-area, is the valve adjustment factor corresponding to the nth building sub-area. and are corresponding weight coefficients, and The sum of the above is 1. In the preferred embodiment of the present application, the value of may be set to 0.4, the value of may be set to 0.3, and the value of may be set to 0.3.

[0116] It should be noted that in the cold-heat hedging diagnosis, it is difficult to accurately determine whether the "insufficient cooling" is caused by the limitation of the system delivery and distribution capacity or by the hedging of adjacent areas. Therefore, the present application further introduces the regional comprehensive cooling efficiency coefficient, which quantifies the instantaneous cooling efficiency that can be actually obtained by the building sub-area under the current working condition by simultaneously using three operation data of the air conditioning air volume, the air conditioning water system temperature difference and the valve opening degree, provides a more physically constrained criterion for subsequent adjustment decision, realizes real-time quantification of the regional cooling potential, greatly improves the accuracy of the cold-heat hedging reason determination, and provides an interpretable and quantifiable cooling capacity constraint for the subsequent control strategy, thereby further improving the accuracy of energy-saving adjustment.

[0117] ​​​​Further, the air conditioning air volume is normalized to a normalized interval range to obtain an air side distribution factor, the air conditioning water system temperature difference is normalized to a normalized interval range to obtain a water side heat exchange factor, and the valve opening is normalized to a normalized interval range to obtain a valve adjustment factor. For example, the normalized interval range can be set as or In the preferred embodiment of the present application, the normalized interval range is set as .

[0118] An example of the air valve opening matching table is shown in Table 1:

[0119] Table 1 Air valve opening matching table

[0120]

[0121] It should be noted that in the embodiments of the present application, the air valve opening matching table is used to grade the current air supply adjustment behavior, thereby assisting in determining whether the air supply system has effectively responded to the current cold and heat load. Specifically, the system compares the air valve opening value of each building sub-region with the preset air valve opening matching table, and divides it into low opening, medium opening and high opening. Among them, the low opening indicates that the air supply strategy of the current building sub-region is in a passive inhibition state, which is usually used in the load reduction stage; the medium opening is the baseline air supply intensity when the system does not adjust; and the high opening indicates that the system has responded to the regional cold / heat demand.

[0122] An example of the dynamic load semantic label matching table is shown in Table 2:

[0123] Table 2 Dynamic load semantic label matching table

[0124]

[0125] It should be noted that in the preferred embodiment of the present application, in order to realize the quantitative analysis of the cold and heat regulation behavior between the building sub-regions and the identification of the hedging risk, the present application constructs a numerical mapping mechanism based on dynamic load semantic labels. Specifically, the present application generates a dynamic load semantic label for each building sub-region, and assigns the label values of 4, 3, 2, and 1 to the forced cooling demand, the mild cooling demand, the no adjustment demand, and the reduced cooling demand, respectively. The above-mentioned label values not only reflect the directionality of the cold and heat demand, but also embody the relative level of the adjustment intensity. Based on the label value system, for any pair of building sub-regions adjacent through the air supply path, the air return path, the water circuit or the physical wall, the absolute value of the difference between the label values is calculated as the adjustment intention difference of the building sub-region pair. The greater the adjustment intention difference, the more the adjacent building sub-regions present a state of opposite direction or extremely inconsistent adjustment amplitude in the air conditioning system, and the higher the cold and heat hedging risk, thereby realizing efficient identification of the cold and heat flow contradiction in the building and decision support for optimization of control.

[0126] The cold and heat hedging evaluation module evaluates the cold and heat hedging risk based on the building coupling feature data set, obtains the cold and heat hedging risk evaluation scores corresponding to the N building sub-regions, and identifies R cold and heat hedging risk regions.

[0127] The method for obtaining the R cold and heat hedging risk regions comprises:

[0128] The building coupling feature data set is input into a pre-trained cold and heat hedging risk evaluation model to obtain the cold and heat hedging risk evaluation scores corresponding to the N building sub-regions one by one;

[0129] For the N cold and heat hedging risk evaluation scores, the cold and heat hedging risk evaluation scores are compared with the preset cold and heat hedging risk score threshold one by one, and if the cold and heat hedging risk evaluation score is greater than the cold and heat hedging risk score threshold, the building sub-region corresponding to the cold and heat hedging risk evaluation score is marked as a cold and heat hedging risk region;

[0130] It should be noted that the building sub-regions not marked as cold and heat hedging risk regions can be marked as non-cold and heat hedging risk regions. In the preferred embodiment of the present application, the value range of the cold and heat hedging risk evaluation score is limited to The closer the cold and heat hedging risk evaluation score is to 1, the higher the possibility of the cold and heat hedging risk occurring in the current operating state of the building sub-region, and the more urgent the adjustment demand; on the contrary, the closer the cold and heat hedging risk evaluation score is to 0, the better the matching degree of the cold and heat supply, and the more stable the operating state. The cold and heat hedging risk score threshold is set to 0.75, and in specific implementation, the present application supports configuring the cold and heat hedging risk score threshold as two modes of dynamic adjustment or static setting to adapt to the fine energy-saving strategy requirements in different scenarios.

[0131] After the determination of N building sub-areas is completed, the total number of cold-heat hedging risk areas is summarized and recorded as R, thereby obtaining R cold-heat hedging risk areas.

[0132] The training method of the cold-hot hedging risk assessment model includes:

[0133] A hot-cold hedging risk assessment dataset is pre-constructed, wherein the hot-cold hedging risk assessment dataset includes Y groups of hot-cold hedging risk assessment data and hot-cold hedging risk assessment scores corresponding to the Y groups of hot-cold hedging risk assessment data, where Y is a positive integer; the hot-cold hedging risk assessment data includes a building coupling feature dataset; the hot-cold hedging risk assessment dataset is divided into a training set and a validation set, wherein the training set is used for parameter learning of the hot-cold hedging risk assessment model, and the validation set is used for real-time monitoring of the generalization performance and overfitting degree of the hot-cold hedging risk assessment model;

[0134] A node-output graph neural network architecture is used as the cold-hot hedging risk assessment model. The cold-hot hedging risk assessment data is standardized and vectorized and then input into the node-output graph neural network architecture. The node-output graph neural network architecture consists of an input layer, a hidden layer, and an output layer. Each hidden layer uses a nonlinear activation function to extract features, and the output layer uses a Sigmoid activation function to obtain the probability distribution corresponding to each cold-hot hedging risk assessment score. Finally, the cold-hot hedging risk assessment score corresponding to the maximum probability is taken as the prediction result of the cold-hot hedging risk assessment model. During the training process, the cross-entropy loss function is used as the optimization target, and a gradient descent optimization algorithm is used to update the network weights, and an early stopping strategy is set: when the prediction accuracy on the validation set reaches or exceeds the preset threshold, the cold-hot hedging risk assessment model is determined to have converged and the training is terminated.

[0135] It should be noted that within the technical framework of this invention, the building coupling feature dataset provides a structured representation of the operating status and mutual coupling relationships of N building sub-areas. This includes both node-level features (i.e., the heating and cooling demand directions, dynamic load semantic labels, and regional comprehensive cooling efficiency coefficients in this application), as well as edge-level difference information aggregated through graph convolution (i.e., the heating and cooling hedging difference graph and the adjustment intention difference matrix in this application). Therefore, the building coupling feature dataset can simultaneously reveal the instantaneous heating and cooling load conditions of building sub-areas and the intensity of coupling contradictions between these building sub-areas in the overall energy network, meeting the input dimensionality and semantic integrity required for heating and cooling hedging risk assessment.

[0136] Node-level features provide the heat load intensity signal within the building sub-areas for the heat-cold hedging risk assessment model. Edge-level dissimilarity information propagates coupling conflict information to associated nodes via the graph convolutional layer. The heat-cold hedging risk assessment model, combining both node-level features and edge-level dissimilarity information, can comprehensively determine the probability of each building sub-area triggering heat-cold hedging at the current moment. Therefore, using the building coupling feature dataset as the overall input to the heat-cold hedging risk assessment model, a heat-cold hedging risk assessment score corresponding to each of the N building sub-areas can be obtained.

[0137] This application identifies R hot and cold hedging risk areas, and has significant rationality and technical advantages. Specifically, this application avoids relying solely on static rules based on thresholds and can capture complex, multi-source nonlinear interaction effects. It utilizes the local aggregation mechanism of the node-output graph neural network architecture to make the edge-level difference directly contribute to the hot and cold conflict risk assessment of adjacent nodes, significantly improving the local hedging positioning accuracy. The node-level independent output format allows the assessment results to be directly mapped to regional risk alerts, facilitating the generation and execution of subsequent partition adjustment strategies. Therefore, it meets the technical requirements of this application to achieve refined hot and cold hedging diagnosis and adaptive energy-saving control.

[0138] The hierarchical scheduling generation module is used to hierarchically sort the R hot and cold hedging risk areas according to functional zoning attributes, building sub-area areas and hot and cold hedging duration types, and generate a hot and cold hedging priority mitigation list and a hot and cold hedging secondary adjustment list, thereby supporting the hierarchical response and differentiated resource allocation of the energy-saving control system.

[0139] like Figure 5 As shown, the method for obtaining the hot and cold hedging priority mitigation list and the hot and cold hedging secondary adjustment list includes:

[0140] S100: let the initial value of r be 1, and the value range of r be 1 to R;

[0141] S101: Obtain the functional zoning attribute of the rth cold-heat arbitrage risk area, match the functional zoning attribute with a pre-constructed functional zoning attribute matching table to obtain a corresponding functional zoning attribute value; record the proportion of the monitored cold-heat arbitrage duration in a unit time as the cold-heat arbitrage duration proportion, and determine whether the cold-heat arbitrage duration proportion exceeds a preset cold-heat arbitrage duration proportion threshold; if the determination result is yes, the cold-heat arbitrage duration type of the rth cold-heat arbitrage risk area is marked as continuous cold-heat arbitrage; if the determination result is no, the cold-heat arbitrage duration type of the rth cold-heat arbitrage risk area is marked as intermittent cold-heat arbitrage; the cold-heat arbitrage duration type is numerized as a cold-heat arbitrage duration type value; for example, the cold-heat arbitrage duration type value corresponding to the continuous cold-heat arbitrage can be marked as 10, and the cold-heat arbitrage duration type value corresponding to the intermittent cold-heat arbitrage can be marked as 11. For example, in the preferred embodiment of the present application, the cold-heat arbitrage duration proportion threshold can be set to 70%.

[0142] S102: Input the cold-heat arbitrage risk assessment score, the functional zoning attribute value, the cold-heat arbitrage duration type value and the building sub-area area of the rth cold-heat arbitrage risk area into a cold-heat arbitrage priority assessment model to obtain a corresponding cold-heat arbitrage priority score;

[0143] S103: Determine whether the cold-heat arbitrage priority score is greater than a preset cold-heat arbitrage priority score threshold; if the determination result is yes, the rth cold-heat arbitrage risk area is added to a cold-heat arbitrage priority mitigation list; if the determination result is no, the rth cold-heat arbitrage risk area is added to a cold-heat arbitrage secondary adjustment list;

[0144] S104: Let r=r+1; if r is less than or equal to R, return to S101 for continuous execution; if r is greater than R, sort the cold-heat arbitrage priority mitigation list and the cold-heat arbitrage secondary adjustment list in descending order according to the corresponding cold-heat arbitrage priority scores to obtain the final cold-heat arbitrage priority mitigation list and the cold-heat arbitrage secondary adjustment list.

[0145] The training method of the cold-heat arbitrage priority assessment model comprises:

[0146] A hot and cold hedging priority evaluation dataset is pre-constructed, wherein the hot and cold hedging priority evaluation dataset includes the hot and cold hedging priority evaluation data of the PG group and the hot and cold hedging priority scores corresponding to the hot and cold hedging priority evaluation data of the PG group, where PG is a positive integer; the hot and cold hedging priority evaluation data includes the hot and cold hedging risk assessment score, the functional zoning attribute value, the hot and cold hedging persistence type value, and the building sub-area area; the hot and cold hedging priority evaluation dataset is divided into a training set and a validation set, wherein the training set is used for learning the parameters of the hot and cold hedging priority evaluation model, and the validation set is used for real-time monitoring of the generalization performance and overfitting degree of the hot and cold hedging priority evaluation model;

[0147] A deep neural network with a multi-layer perceptron as the core is used as the hot and cold hedging priority evaluation model. The hot and cold hedging priority evaluation data is input into the deep neural network after standardization and vectorization processing. The deep neural network consists of an input layer, a hidden layer and an output layer; each hidden layer uses a nonlinear activation function to extract features, and the output layer uses a softmax activation function to obtain the probability distribution corresponding to each hot and cold hedging priority score. Finally, the hot and cold hedging priority score corresponding to the maximum probability is taken as the prediction result of the hot and cold hedging priority evaluation model; during the training process, the cross entropy loss function is used as the optimization target, and the gradient descent optimization algorithm is used to update the network weights, and an early stopping strategy is set: when the prediction accuracy on the validation set reaches or exceeds the preset threshold, it is determined that the hot and cold hedging priority evaluation model has converged and the training is terminated.

[0148] It should be noted that in this application, the hot / cold hedging priority assessment model outputs a corresponding hot / cold hedging priority score based on input features such as the hot / cold hedging risk assessment score of the building sub-area, the functional zoning attribute value, the hot / cold hedging persistence type value, and the sub-area area. The hot / cold hedging priority score is used to quantify the urgency and response priority of each hot / cold hedging risk area during the operation and maintenance adjustment process.

[0149] For example, the value range of the thermal hedge priority score is set to In the continuous interval of , the higher the cold-heat hedging priority score value, the more serious the cold-heat hedging phenomenon corresponding to the building sub-area is, the greater the impact on the energy system loss, the more significant the threat to the indoor environmental comfort, and the higher the priority of the adjustment response.

[0150] In order to achieve hierarchical regulation of hot and cold hedging risk areas, a hot and cold hedging priority score threshold is set in this application to divide the priority mitigation area and the secondary regulation area. The hot and cold hedging priority score threshold can be dynamically set based on historical operation data and actual operation and maintenance experience, and preferably a numerical range between 0.70 and 0.80 is used for parameter optimization. In the implementation example, the hot and cold hedging priority score threshold is preferably set to 0.75, that is, when the priority score of a hot and cold hedging risk area is greater than 0.75, it is determined that the area should be included in the hot and cold hedging priority mitigation list; if it is lower than or equal to 0.75, it is included in the hot and cold hedging secondary regulation list.

[0151] By setting the hot and cold hedging priority scoring threshold, not only the response classification of the hot and cold hedging areas is achieved, but also the unnecessary tilt of scheduling resources is effectively avoided, and the overall system's coordination scheduling capabilities and energy-saving response accuracy when facing large-scale hot and cold hedging risks are improved, reflecting the innovative value of this application in intelligent operation control and energy efficiency collaborative management.

[0152] An example of the functional partition attribute matching table is shown in Table 3:

[0153] Table 3 Functional zoning attribute matching table

[0154]

[0155] It should be noted that in this application, for example, combined with the functional zoning attribute matching table set in the embodiment, the four typical areas of "office", "conference room", "reception area" and "data center room" are numerically modeled according to the sensitivity of the area to the indoor hot and cold environment, and on the basis of the hot and cold hedging risk assessment, a hot and cold hedging persistence identification mechanism is introduced to construct a hot and cold hedging priority relief list and a hot and cold hedging secondary adjustment list to achieve hierarchical scheduling control. The hierarchical scheduling strategy of this application not only reflects the priority level of different functional areas in operation and maintenance adjustment, but also introduces the dynamic characteristics of hot and cold hedging persistence, which improves the pertinence of the policy response and the practicality of energy-saving measures.

[0156] Specifically, for example, data center rooms are assigned the highest level in the numerical setting of functional zoning attributes due to their high density of equipment, heavy operating load, and extremely high requirements for temperature control accuracy. Their adjustment priority is significantly higher than areas with a higher tolerance for temperature and humidity fluctuations, such as conference rooms and reception areas. At the same time, by monitoring the duration of the hot and cold hedging phenomenon per unit time, we can further distinguish between continuous hot and cold hedging and intermittent hot and cold hedging. Among them, continuous hot and cold hedging means that the imbalance between hot and cold in the area exists for a long time. If it is not handled in time, it will cause continuous energy waste or local thermal comfort collapse, and it needs to be included in the hot and cold hedging priority mitigation list; intermittent hot and cold hedging means that the hot and cold hedging fluctuates periodically or sporadically, and can be postponed for processing and included in the hot and cold hedging secondary adjustment list, thereby achieving a phased response.

[0157] By integrating the hot and cold hedging risk assessment scores, functional zoning attribute values, hot and cold hedging duration type values ​​and building sub-area areas into modeling, and using the priority scoring results to drive the scheduling order allocation of hot and cold hedging areas, it is possible to achieve dynamic optimization matching of resource regulation intensity, response rhythm and target areas, thereby enhancing the efficiency and accuracy of the system's energy-saving response.

[0158] The hot and cold hedging control module combines the air conditioning system's operating margin parameter set, the hot and cold hedging difference map, and the adjustment intention difference matrix to perform hierarchical and adaptive hot and cold hedging control on the hot and cold hedging priority relief list and the hot and cold hedging secondary adjustment list. The air conditioning system operating margin parameter set describes the air conditioning system's remaining adjustment headroom without overloading, that is, its current effective output capacity.

[0159] The air conditioning system operation margin parameter set includes real-time air supply volume margin, cold water flow regulation margin, hot water flow regulation margin, terminal valve stroke remaining percentage, host frequency conversion adjustment range, cold storage unit remaining energy release capacity and heat storage unit remaining energy release capacity.

[0160] It should be noted that the real-time air supply margin is calculated by obtaining the current air supply data of the terminal fan coil unit or variable air volume box terminal device and calculating the difference between it and the designed maximum air supply volume. The real-time air supply volume can be collected by the air volume sensor or the built-in feedback unit of the variable air volume terminal device; the designed maximum air supply volume is obtained based on the initial system settings. The difference between the two is the current air supply adjustment margin.

[0161] The cold water flow regulation margin and hot water flow regulation margin represent the remaining adjustable flow capacity in the system's cold water and hot water branches, respectively. These margins are determined by comparing the current flow rate with the pump's variable frequency control upper flow limit, using electromagnetic flowmeters installed at the outlets of the cold and hot water pumps. For pump stations with bidirectional regulation capabilities, a comprehensive upper and lower bound assessment can also be performed in conjunction with a set minimum flow threshold.

[0162] The end valve stroke remaining percentage reflects the adjustable space of the end regulating valve of the current building sub-area. The current opening of the valve is obtained in real time through the position feedback unit built in the valve electric actuator, and the remaining percentage is calculated in combination with the maximum stroke of the valve structure. The end valve stroke remaining percentage can be used to determine whether the end regulating capacity is saturated, thereby affecting the mitigation strategy of the local cold-heat hedge problem.

[0163] The host variable frequency regulating range represents the remaining space of the current operating frequency of the water chilling unit or heat source host from the upper and lower limits of the allowed operating frequency. The host variable frequency regulating range is obtained from the operating frequency feedback value provided by the host control system, and is calculated after being compared with the maximum / minimum operating frequency set by the host. The end valve stroke remaining percentage is used to evaluate the regulating response capacity at the system level.

[0164] The cold storage unit remaining discharging capacity and the heat storage unit remaining discharging capacity are used to indicate the cold / heat energy that can be released by the current centralized energy storage device. The temperature, volume and energy storage state of the current energy storage medium are obtained in real time through the energy management controller of the cold storage / heat storage device, and the remaining discharging capacity is converted based on the heat capacity calculation formula.

[0165] Through the collection and summary of the above-mentioned various operating margin parameters, the application can dynamically construct an operating margin parameter set reflecting the overall regulating capacity of the air conditioning system, and provide multi-dimensional and real-time resource support basis for cold-heat hedge regulation.

[0166] As shown in Figure 3 The method for performing cold-heat hedge hierarchical adaptive regulation includes:

[0167] For each building sub-area in the cold-heat hedge priority mitigation list, denoted as a priority mitigation building sub-area, a subgraph directly having a spatial coupling relationship with the priority mitigation building sub-area is obtained from the cold-heat hedge difference graph, denoted as a cold-heat hedge associated subgraph. The matrix elements directly associated with the priority mitigation building sub-area on the graph structure edge are extracted from the regulating intention difference degree matrix to construct a regulating intention difference degree set. The air conditioning system operating margin parameter set, the cold-heat hedge associated subgraph and the regulating intention difference degree set are input into a priority mitigation parameter setting model to obtain an instant regulation instruction set. The instant regulation instruction set includes a target air supply amount set value, an end water valve target opening, a host load instantaneous correction gain and an energy release instruction of the energy storage unit.

[0168] The instant adjustment instruction set is written into the building automation system control bus according to the node address mapping, the instructions are protocol encapsulated according to the equipment type, and are respectively sent to the corresponding control unit for execution. For example, the target air supply amount set value is written into the air volume set register of the variable air volume terminal device or the fan coil module, the terminal water valve target opening is written into the target stroke register of the electric regulating valve driver, the host load instantaneous correction gain is written into the variable frequency control interface of the chiller / boiler, and the energy release instruction of the energy storage unit is sent to the energy management controller of the cold storage / heat storage device.

[0169] When the cold-heat hedging priority relief list is completed, the cold-heat hedging secondary adjustment list is updated and the cold-heat hedging hierarchical adaptive regulation and control is performed.

[0170] As shown in Figure 4 The method for updating the cold-heat hedging secondary adjustment list and performing the cold-heat hedging hierarchical adaptive regulation and control includes:

[0171] For each building sub-area in the cold-heat hedging secondary adjustment list, denoted as a secondary relief building sub-area, the cold-heat hedging risk assessment score corresponding to the secondary relief building sub-area is re-evaluated and updated;

[0172] The secondary relief building sub-area with a cold-heat hedging risk assessment score not greater than a preset cold-heat hedging risk score threshold is removed from the cold-heat hedging secondary adjustment list to obtain an updated cold-heat hedging secondary adjustment list; and it is ensured that the cold-heat hedging secondary adjustment list only retains cold-heat hedging risk areas that still have adjustment necessity at present.

[0173] The latest air conditioning system operation margin parameter set is obtained, and the cold-heat hedging risk assessment scores in the updated cold-heat hedging secondary adjustment list are constructed into a cold-heat hedging risk assessment score set;

[0174] The air conditioning system operation margin parameter set and the cold-heat hedging risk assessment score set are input into a secondary relief parameter setting model to obtain a planned adjustment instruction set, and the adjustment instructions in the planned adjustment instruction set are respectively sent to the corresponding control unit for execution; the adjustment instructions in the planned adjustment instruction set are of the same type as the adjustment instructions in the instant adjustment instruction set. It is ensured that the cold-heat hedging phenomenon in the secondary cold-heat hedging risk area is smoothly relieved within a controllable range, while the energy saving benefit and system operation safety are maximized, so as to realize adaptive regulation and control and energy efficiency optimization of the cold-heat hedging phenomenon.

[0175] The training method of the priority relief parameter setting model includes:

[0176] A priority relief parameter setting dataset is constructed in advance, the priority relief parameter setting dataset includes YX sets of priority relief parameter setting data and a corresponding set of immediate adjustment instructions for the YX sets of priority relief parameter setting data, YX is a positive integer; the priority relief parameter setting data includes a set of air conditioning system operation margin parameters, a cold-heat hedging correlation subgraph, and a set of adjustment intention difference degrees; the priority relief parameter setting dataset is divided into a training set and a verification set, the training set is used for priority relief parameter setting model parameter learning, and the verification set is used for real-time monitoring of the generalization performance and overfitting degree of the priority relief parameter setting model.

[0177] A node output type graph neural network architecture is used as the priority relief parameter setting model, the priority relief parameter setting data is input into the node output type graph neural network architecture after being standardized and vectorized, the node output type graph neural network architecture is composed of an input layer, hidden layers, and an output layer; each hidden layer uses a nonlinear activation function to extract features, and the output layer uses a Sigmoid activation function to obtain a probability distribution corresponding to each set of immediate adjustment instructions, and finally the set of immediate adjustment instructions corresponding to the maximum probability is taken as the prediction result of the priority relief parameter setting model; a cross-entropy loss function is used as the optimization objective in the training process, a gradient descent type optimization algorithm is used to update the network weights, and an early stopping strategy is set: when the prediction accuracy on the verification set reaches or exceeds a preset threshold, it is determined that the priority relief parameter setting model has converged and the training is terminated.

[0178] The training method of the secondary relief parameter setting model includes:

[0179] A secondary relief parameter setting dataset is constructed in advance, the secondary relief parameter setting dataset includes CJ sets of secondary relief parameter setting data and a corresponding set of planned adjustment instructions for the CJ sets of secondary relief parameter setting data, CJ is a positive integer; the secondary relief parameter setting data includes a set of air conditioning system operation margin parameters and a set of cold-heat hedging risk assessment scores; the secondary relief parameter setting dataset is divided into a training set and a verification set, the training set is used for secondary relief parameter setting model parameter learning, and the verification set is used for real-time monitoring of the generalization performance and overfitting degree of the secondary relief parameter setting model.

[0180] The deep neural network taking the multilayer perceptron as the core is used as the secondary relief parameter setting model. The secondary relief parameter setting data is input into the deep neural network after being standardized and vectorized. The deep neural network is composed of an input layer, a hidden layer and an output layer. Each hidden layer uses a nonlinear activation function to extract features, and the output layer uses a Softmax activation function to obtain a probability distribution corresponding to each set of plan adjustment instructions. Finally, the set of plan adjustment instructions corresponding to the maximum probability is taken as the prediction result of the secondary relief parameter setting model. In the training process, the cross-entropy loss function is used as the optimization objective, the gradient descent optimization algorithm is used to update the network weights, and the early stopping strategy is set: when the prediction accuracy on the validation set reaches or exceeds the preset threshold, it is determined that the secondary relief parameter setting model has converged and the training is terminated.

[0181] It should be noted that in the overall control framework of the present application, the cold-heat hedging risk area is divided into a cold-heat hedging priority relief list and a cold-heat hedging secondary adjustment list. Different control schemes are adopted for the two types of lists. The technical considerations of this hierarchical design are specifically in terms of differences in risk urgency and business continuity. The building sub-areas in the cold-heat hedging priority relief list mostly belong to data machine rooms or personnel-intensive areas, and their sensitivity to temperature and humidity stability and energy consumption impact is much higher than that of general office or public reception areas. The continuous existence of any cold-heat hedging may directly affect core business or cause comfort complaints. Therefore, the priority relief scheme aims to instantaneously suppress the conflict, adopts an immediate adjustment strategy of seconds to minutes, and emphasizes fast and effective cold / heat compensation in the instruction parameters, such as rapid increase of target air supply, large adjustment of valve travel, sudden increase of main refrigeration gain, and immediate energy release of energy storage unit. After adjustment, high-frequency closed-loop feedback is configured to ensure that the cold-heat risk is suppressed in the shortest time.

[0182] In terms of differences in system resource scheduling and energy saving coordination, the cold-heat hedging risk of the building sub-areas covered by the cold-heat hedging secondary adjustment list is relatively low, and adjustment lag or time period processing can be tolerated. To avoid system load peaks or frequent start-stop impacts on equipment caused by large-scale simultaneous adjustment, the secondary adjustment scheme aims to smooth the conflict and save energy, and adopts a planned and batched scheduling strategy of hours. The secondary relief parameter setting model emphasizes resource allocation and load balancing more, and sets a slow gradient for the instruction parameters, such as slow increase of air supply curve, valve fine tuning, smooth correction of main refrigeration base load, and batched charging and discharging of energy storage. In addition, the application can also realize peak shaving and energy efficiency optimization in combination with off-peak electricity prices and night maintenance windows.

[0183] Based on the above differences, this application constructs a priority mitigation parameter setting model and a secondary mitigation parameter setting model respectively. The priority mitigation parameter setting model focuses on real-time margins and calculates the adjustment amplitude with high weights; the secondary mitigation parameter setting model focuses on full-time margins and optimizes the adjustment parameters in a time-sharing manner. Both the priority mitigation parameter setting model and the secondary mitigation parameter setting model are based on the air-conditioning system operation margin parameter set, and are respectively adapted to the technical scenarios of immediate conflict suppression and planned energy-saving scheduling, thereby achieving differentiated and precise control of areas with different risk levels.

[0184] Example 2:

[0185] See also Figure 2 As shown, this embodiment provides a method for energy conservation management of public institutions based on AI intelligent operation and maintenance, including:

[0186] Obtain multi-source building operation data;

[0187] Based on semantic fusion and coupling evaluation of multi-source building operation data, a building coupling feature dataset is constructed. The building coupling feature dataset includes a cooling and heating hedging difference map, an adjustment intention difference matrix, and the cooling and heating demand directions, dynamic load semantic labels, and regional comprehensive cooling efficiency coefficients corresponding to N building sub-areas.

[0188] Based on the building coupling feature dataset, a heat-cold hedging risk assessment is performed to obtain the heat-cold hedging risk assessment scores corresponding to N building sub-areas, and R heat-cold hedging risk areas are identified.

[0189] It is used to hierarchically sort the R hot and cold hedging risk areas according to functional zoning attributes, building sub-area areas, and hot and cold hedging persistence types, and generate a hot and cold hedging priority mitigation list and a hot and cold hedging secondary adjustment list;

[0190] Combined with the air-conditioning system operating margin parameter set, the hot and cold hedging difference map and the adjustment intention difference matrix, the hot and cold hedging priority relief list and the hot and cold hedging secondary adjustment list are adaptively controlled by hot and cold hedging hierarchical control.

[0191] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

[0192] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. The energy conservation management system for public institutions based on AI intelligent operation and maintenance is characterized by: include: Multi-source data acquisition module, used to obtain multi-source building operation data; The coupling feature analysis module performs semantic fusion and coupling evaluation based on multi-source building operation data to construct a building coupling feature dataset; The building coupling feature dataset includes a cooling and heating hedging difference map, an adjustment intention difference matrix, and cooling and heating demand directions, dynamic load semantic labels, and regional comprehensive cooling efficiency coefficients corresponding to N building sub-areas; The method for constructing the building coupling feature dataset includes: Based on temperature and humidity, determine the heating and cooling demand directions corresponding to N building sub-areas, where the heating and cooling demand directions include cooling demand, heating demand, and no comfort demand; Match the air valve opening with a pre-built air valve opening matching table to obtain the air valve opening gears corresponding to N building sub-areas; the air valve opening gears include low opening, medium opening and high opening; Set dynamic load semantic labels for N building sub-areas based on cooling and heating demand direction, air valve opening gear, equipment power consumption, and occupant density; According to the difference between the spatial coupling relationship of N building sub-areas and the dynamic load semantic labels, a heat and cold hedging difference map and a regulation intention difference matrix are constructed. Based on the air volume, temperature difference of the air conditioning water system and valve opening, the regional comprehensive cooling efficiency coefficient corresponding to N building sub-areas is constructed; The building coupling feature dataset is constructed by combining the heat and cold hedging difference map, the adjustment intention difference matrix, the heat and cold demand directions, dynamic load semantic labels, and regional comprehensive cooling efficiency coefficients corresponding to N building sub-areas. The heat-cold hedging assessment module performs heat-cold hedging risk assessment based on the building coupling feature dataset, obtains heat-cold hedging risk assessment scores corresponding to N building sub-areas, and identifies R heat-cold hedging risk areas; A hierarchical scheduling generation module is used to hierarchically sort the R hot and cold hedging risk areas according to functional zoning attributes, building sub-area areas, and hot and cold hedging persistence types, and generate a hot and cold hedging priority mitigation list and a hot and cold hedging secondary adjustment list; The hot and cold hedging control module is used to combine the air conditioning system operating margin parameter set, the hot and cold hedging difference map, and the adjustment intention difference matrix to perform hot and cold hedging hierarchical adaptive control on the hot and cold hedging priority relief list and the hot and cold hedging secondary adjustment list. The method for performing hot and cold hedging hierarchical adaptive control includes: For each building sub-area in the priority relief list for cold and hot hedging, which is recorded as a priority relief building sub-area, a sub-graph with a direct spatial coupling relationship with the priority relief building sub-area is obtained from the cold and hot hedging difference graph, which is recorded as a cold and hot hedging associated sub-graph; the matrix elements directly associated with the priority relief building sub-area on the graph structure edge are extracted from the adjustment intention difference matrix to construct an adjustment intention difference set; the air-conditioning system operation margin parameter set, the cold and hot hedging associated sub-graph, and the adjustment intention difference set are input into the priority relief parameter setting model to obtain an immediate adjustment instruction set; the immediate adjustment instruction set includes the target air supply volume set value, the terminal water valve target opening, the host load instantaneous correction gain, and the energy storage unit energy release instruction; Send the real-time adjustment instruction set to the corresponding control unit for execution; When the hot and cold hedge priority relief list is completed, the hot and cold hedge secondary adjustment list is updated and hot and cold hedge hierarchical adaptive regulation is performed.

2. The public institution energy conservation management system based on AI intelligent operation and maintenance according to claim 1 is characterized in that: The method for updating the cold and hot hedging secondary adjustment list and performing cold and hot hedging hierarchical adaptive control includes: For each building sub-area in the secondary adjustment list for cold and heat hedging, record it as a secondary mitigation building sub-area, and re-evaluate and update the corresponding cold and heat hedging risk assessment score of the secondary mitigation building sub-area; Remove the secondary mitigation building sub-areas whose cold and heat hedging risk assessment scores are not greater than the preset cold and heat hedging risk score threshold from the cold and heat hedging secondary adjustment list to obtain an updated cold and heat hedging secondary adjustment list; Obtain the latest air conditioning system operating margin parameter set, and construct a heat and cold hedging risk assessment score set based on the heat and cold hedging risk assessment scores in the updated heat and cold hedging secondary adjustment list; The air-conditioning system operating margin parameter set and the cold and hot hedging risk assessment score set are input into the secondary mitigation parameter setting model to obtain a planned adjustment instruction set, which is then sent to the corresponding control unit for execution; the planned adjustment instruction set is of the same type as the adjustment instruction in the immediate adjustment instruction set.

3. The public institution energy conservation management system based on AI intelligent operation and maintenance according to claim 1 is characterized in that: Methods for obtaining the hot and cold hedge priority mitigation list and the hot and cold hedge secondary adjustment list include: For R cold-heat hedging risk areas, obtain the functional zoning attributes, cold-heat hedging duration ratio, building sub-area area, and cold-heat hedging risk assessment score of each cold-heat hedging risk area; According to the preset functional zoning attribute matching table, the functional zoning attributes are converted into functional zoning attribute values; the hot and cold hedging duration ratio is compared with the preset duration ratio threshold to obtain the hot and cold hedging duration type, and converted into the hot and cold hedging duration type value; the hot and cold hedging duration type is continuous hot and cold hedging or intermittent hot and cold hedging; Input the heat and cold hedging risk assessment score, functional zoning attribute value, heat and cold hedging persistence type value and building sub-area area into the heat and cold hedging priority assessment model to obtain the corresponding heat and cold hedging priority score; Compare the hot and cold hedge priority scores with the preset hot and cold hedge priority score thresholds to obtain a hot and cold hedge priority mitigation list and a hot and cold hedge secondary adjustment list respectively; The hot and cold hedge priority relief list and the hot and cold hedge secondary adjustment list are sorted in descending order according to the corresponding hot and cold hedge priority scores to obtain the final hot and cold hedge priority relief list and the hot and cold hedge secondary adjustment list.

4. The public institution energy conservation management system based on AI intelligent operation and maintenance according to claim 1 is characterized in that: The method for obtaining R cold and hot hedging risk areas includes: Inputting the building coupling feature dataset into a pre-trained heat-cold hedging risk assessment model to obtain heat-cold hedging risk assessment scores corresponding to N building sub-areas; For N cold-heat hedging risk assessment scores, compare each score with a preset cold-heat hedging risk score threshold. If the score is greater than the threshold, mark the building sub-area corresponding to the score as a cold-heat hedging risk area. After the determination of N building sub-areas is completed, the total count of the cold-heat hedging risk areas is summarized and recorded as R, thereby obtaining R cold-heat hedging risk areas.

5. The public institution energy conservation management system based on AI intelligent operation and maintenance according to claim 1 is characterized in that: Methods for determining the cooling and heating demand directions corresponding to N building sub-areas based on temperature and humidity include: For each building sub-area, if the temperature of the building sub-area is higher than the upper limit of the comfortable temperature and the humidity is higher than the upper limit of the comfortable humidity, the heating and cooling demand direction is marked as a cooling demand and the current process ends; If the temperature of the building sub-area is lower than the lower limit of the comfort temperature, the cooling and heating demand direction is marked as the heat demand and the current process ends; Mark the heating and cooling demand direction as comfortable and no demand, and end the current process.

6. The public institution energy conservation management system based on AI intelligent operation and maintenance according to claim 1 is characterized in that: The method for setting the dynamic load semantic labels corresponding to N building sub-areas includes: For each building sub-area, when the heating and cooling demand direction of the building sub-area is cooling demand, determine whether the equipment power consumption and personnel density exceed the corresponding thresholds. If the judgment result exceeds the threshold, determine whether the air valve opening gear is high. If the air valve opening gear is high, the dynamic load semantic label is recorded as forced cooling required; if the air valve opening gear is not high, the dynamic load semantic label is recorded as warm cooling required; if the judgment result does not exceed the threshold, the dynamic load semantic label is recorded as warm cooling required; When the cooling and heating demand direction of the building sub-area is heat demand, that is, the system has excessive cooling, it is determined whether the air valve opening gear is high. If the air valve opening gear is high, the dynamic load semantic label is recorded as needing to reduce cooling. If the air valve opening gear is not high, the dynamic load semantic label is recorded as needing warm cooling. When the cooling and heating demand direction of the building sub-area is comfortable with no demand, the dynamic load semantic label is recorded as no adjustment demand.

7. The public institution energy conservation management system based on AI intelligent operation and maintenance according to claim 1 is characterized in that: The construction method of the hot-cold hedging difference map and the adjustment intention difference matrix includes: Match the dynamic load semantic label with the pre-built dynamic load semantic label matching table to obtain the dynamic load semantic label values ​​corresponding to N building sub-areas; Mapping N building sub-areas into a set of graph structure nodes; The air supply path coupling edges, return air confluence coupling edges, water loop coupling edges and shared wall heat transfer edges corresponding to N building sub-areas are constructed into a graph structure edge set; For each graph structure edge in the graph structure edge set, construct a graph structure edge combination with the graph structure edge adjacent to the graph structure edge, and determine whether the dynamic load semantic label values ​​corresponding to the graph structure edge combination are the same. If they are not the same, set the adjustment intention difference of the graph structure edge combination to the absolute value of the difference between the dynamic load semantic label values ​​corresponding to the graph structure edge combination; if they are the same, set the adjustment intention difference of the graph structure edge combination to zero; The adjustment intention difference is written into the adjustment intention difference matrix, and the row and column coordinate positions in the adjustment intention difference matrix are the numbered corresponding positions formed by the edge combination of the graph structure; When all graph structure edges in the graph structure edge set are processed, the assigned graph structure edge set is obtained; The cold-hot hedge difference graph is constructed by combining the graph structure node set and the assigned graph structure edge set.

8. The public institution energy conservation management system based on AI intelligent operation and maintenance according to claim 1 is characterized in that: The method for constructing the regional comprehensive cooling efficiency coefficient corresponding to N building sub-areas includes: For each building sub-area, the air conditioning air volume, air conditioning water system temperature difference, and valve opening are normalized to a preset normalized range, and the air-side distribution factor, water-side heat exchange factor, and valve adjustment factor are obtained respectively. The regional comprehensive cooling efficiency coefficient is calculated based on the air-side distribution factor, water-side heat exchange factor, and valve adjustment factor. When all building sub-areas are processed, the regional comprehensive cooling efficiency coefficients corresponding to the N building sub-areas are obtained.

9. The public institution energy conservation management system based on AI intelligent operation and maintenance according to claim 1 is characterized in that: The multi-source building operation data includes temperature, humidity, air-conditioning air volume, air-conditioning water system temperature difference, valve opening, air valve opening, equipment power consumption and personnel density corresponding to N building sub-areas.

Images