A building energy-saving optimization strategy generation method based on unmanned aerial vehicle infrared imaging

Abstract

This invention discloses a method for generating building energy-saving optimization strategies based on UAV infrared imaging, specifically relating to the field of building energy-saving management technology. The method includes acquiring infrared images, timestamps, location coordinates, and air conditioning records collected by a UAV during target building inspections; stitching the infrared images together according to the location coordinates; extracting the room activation status and set temperature from the air conditioning records according to the timestamps; and outputting a temperature map and operation table. By collaboratively organizing the infrared inspection results, room air conditioning operation information, and control area constraints at the edge into an action list, room execution sequence, and area execution sequence, energy-saving recommendations can be transformed from static recommendations into strategy results with constraint resolution, prioritization, and batch execution capabilities, relatively improving the problems of recommendation conflicts and implementation difficulties.

Description

Technical Field

[0001] This invention relates to the field of building energy-saving management technology, and more specifically, to a method for generating building energy-saving optimization strategies based on UAV infrared imaging. Background Technology

[0002] In the field of building energy-saving management based on UAV infrared imaging, the mainstream practice in the industry is to solve the problems of identifying thermal anomalies in the building envelope and generating energy-saving improvement suggestions. Typically, UAVs equipped with infrared imaging equipment are used to inspect the building's exterior curtain wall to obtain the surface temperature distribution. Then, combined with the air conditioning start / stop status, temperature setting information or energy consumption records inside the building, areas with abnormal temperature differences are manually interpreted or matched according to rules to output weak insulation locations, abnormal air conditioning operation locations or corresponding energy-saving suggestions. For example, in multi-story inspection scenarios of large office buildings or commercial complexes, drones need to quickly form optimization strategies that can be directly executed by property management or operation and maintenance departments for different oriented curtain walls, different air-conditioning control areas, and rooms of different usage levels under limited flight time, limited edge computing resources, and limited data backhaul conditions. They also need to meet the hard constraints that some key areas cannot be arbitrarily temperature-adjusted, some air-conditioning systems cannot be controlled independently by a single room, and some building envelope defects cannot be repaired immediately. Under this constraint, mainstream practices will consistently reveal bottlenecks such as conflicting energy-saving recommendations, unclear execution priorities, lack of verification order, and difficulty in directly implementing the recommendations. Specifically, multiple actions such as recommending to lower the set temperature, adjust the control sequence, and carry out insulation maintenance will appear simultaneously in the same inspection batch. However, there is a lack of constraints, sequence, and benefit verification basis among these actions, which means that the management side can only continue to rely on manual screening and experience-based decisions. The root cause is that most existing solutions remain at the level of anomaly identification and static recommendation, and fail to further organize the inspection identification results into a sequence of executable energy-saving actions that meet the constraints. The technical problem this application aims to solve is: how to collaboratively transform the results of UAV infrared inspections and building air conditioning operation information into a building energy-saving optimization strategy with constraint resolution, priority ranking, and batch execution capabilities under edge computing conditions. Summary of the Invention

[0003] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for generating building energy-saving optimization strategies based on UAV infrared imaging. This method involves segmenting and identifying the temperature distribution of the exterior curtain wall obtained by UAV infrared inspection and mapping it to rooms, and combining this with room air conditioning operation information to form room anomaly chains, action lists, and regional execution sequences, thereby solving the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for generating building energy-saving optimization strategies based on UAV infrared imaging, comprising: S1. Acquire infrared images, timestamps, location coordinates, and air conditioning records collected by the drone during the inspection of the target building. Perform facade stitching on the infrared images according to the location coordinates. Extract the room activation status and set temperature from the air conditioning records according to the timestamps. Output temperature map and operation table. S2. Obtain the temperature map, divide the temperature map into blocks according to the curtain wall grid, calculate the average temperature, adjacent temperature difference and time-series temperature difference of each block, identify the blocks that deviate as abnormal blocks, and output the abnormal block set. S3. Obtain the abnormal block set, the facade room correspondence table and the operation table. Map each abnormal block to the corresponding room to form a room abnormal record. Write the block location, abnormal time period, room activation status and set temperature into the analysis table and output the analysis table. S4. Obtain the analysis table, merge room abnormal records according to room identifier, count the number of abnormalities, duration of abnormalities and set temperature deviation of each room, and generate temperature adjustment actions, start-stop actions and maintenance actions accordingly, and output the action list. S5. Obtain the action list, sort it in the order of temperature adjustment actions, start-stop actions, and maintenance actions in the same room, group it in the order of the number of rooms in the same control area from fewest to most, generate the sorting and grouping results into a building energy-saving optimization strategy, and output the building energy-saving optimization strategy for the target building.

[0005] In a preferred embodiment, S1 includes: S1-1. Obtain infrared images, timestamps, and location coordinates. Classify the infrared images according to their location coordinates and arrange the infrared images of the same facade in chronological order according to their timestamps. Output the facade image sequence. S1-2. Obtain the facade image sequence, perform temperature point matching and boundary stitching on the overlapping areas of adjacent infrared images in the facade image sequence, and output the temperature map. S1-3. Obtain air conditioning records and timestamps. Extract the room activation status and set temperature corresponding to the time period of the facade image sequence from the air conditioning records according to the timestamps, and write them into the operation table according to the room identifier. Output the operation table.

[0006] In a preferred embodiment, S2 includes: S2-1. Obtain the temperature map, divide the temperature map into blocks according to the curtain wall grid, calculate the average temperature of each block, and output the block temperature table. S2-2. Obtain the block temperature table, calculate the temperature difference between each block and its adjacent blocks, as well as the temperature difference of the same block at different timestamps, and output the block difference table. S2-3. Obtain the block temperature table and block difference table. Identify blocks whose average temperature deviates, whose adjacent block temperature difference deviates, and whose same block temperature difference deviates as abnormal blocks, and output the abnormal block set.

[0007] In a preferred embodiment, S3 includes: S3-1. Obtain the set of abnormal blocks and the facade room correspondence table, map each abnormal block to a room according to the facade location, and output the room mapping table. S3-2. Obtain the room mapping table and the operation table, match the room activation status and set temperature for each abnormal block corresponding to the time period, and output the room matching table.

[0008] In a preferred embodiment, S3 further includes: S3-3. Obtain the room matching table, combine the block location, abnormal time period, room activation status and set temperature of each abnormal block to generate room abnormal records, and output the set of room abnormal records. S3-4. Obtain the set of abnormal room records, write them into the analysis table according to the room identifier, and output the analysis table.

[0009] In a preferred embodiment, S4 includes: S4-1. Obtain the analysis table, merge room anomaly records according to room identifier, and extract the anomaly block location, anomaly time period, room activation status and set temperature corresponding to each room, and output the room merge table. S4-2. Obtain the room merging table, calculate the number of abnormal blocks, duration of abnormality, and set temperature offset according to the room identifier, and merge the abnormal room records with consecutive abnormal block positions, adjacent abnormal time periods, and consistent set temperature offset direction into the same abnormal chain, and output the room abnormal chain list.

[0010] In a preferred embodiment, S4 further includes: S4-3. Obtain the room anomaly chain list, perform temperature adjustment judgment, start / stop judgment and maintenance judgment on each room anomaly chain, write the number of anomaly blocks, the duration of anomaly and the set temperature offset into the corresponding judgment result, and write temperature adjustment token, start / stop token or maintenance token to the room anomaly chain that passes the judgment, and output the room judgment table. S4-4. Obtain the room determination table, perform cross-validation on the same room's abnormal chain that simultaneously writes different tokens, retain the room's abnormal chain with the set temperature offset and abnormal duration as the priority action chain, and write the room's abnormal chain with the set temperature offset and abnormal duration into the conflict flag and output the conflict chain list.

[0011] In a preferred embodiment, S4 further includes: S4-5. Obtain the conflict chain list, perform rollback and re-check on the room exception chain with the conflict mark written, re-execute the branch merge with the other exception chains of the corresponding room after the rollback, update the temperature control token, start / stop token or maintenance token for the merged room exception chain, and output the room action table. S4-6. Obtain the room action table, extract the temperature adjustment action written with the temperature control token, the start / stop action written with the start / stop token, and the maintenance action written with the maintenance token according to the room identifier, and write the room identifier, number of abnormal blocks, duration of abnormality, set temperature offset and action type corresponding to each action into the action list, and output the action list.

[0012] In a preferred embodiment, S5 includes: S5-1. Obtain the action list, classify the temperature control actions, start-up and shutdown actions and maintenance actions according to the room identifier and the control area identifier, and output the room action table and the area action table. S5-2. Obtain the room action table, sort it according to the order of temperature adjustment action, start / stop action, and maintenance action in the same room, write the sorting result into the room execution sequence, and output the room execution table.

[0013] In a preferred embodiment, S5 further includes: S5-3. Obtain the room execution table and the area action table. Group the room execution sequences in order of increasing number of rooms in the same control area, write the grouping results into the area execution sequence, and output the area execution table. S5-4. Obtain the regional execution table, summarize the execution sequences of each region according to the joint control region identifier, generate the building energy-saving optimization strategy, and output the building energy-saving optimization strategy of the target building.

[0014] The technical effects and advantages of this invention are as follows: 1. By coordinating infrared inspection results, room air conditioning operation information, and control area constraints at the edge side into action lists, room execution sequences, and area execution sequences, energy-saving recommendations can be transformed from static recommendations into strategic results with constraint resolution, prioritization, and batch execution capabilities, thereby relatively improving the problems of recommendation conflicts and implementation difficulties. 2. By performing facade classification, temporal arrangement, and temperature stitching of overlapping areas on infrared images, a temperature map consistent with the facade coordinates is formed. This allows subsequent block temperature calculations to be based on a continuous and uniform temperature distribution, thereby relatively reducing the impact of the dispersed location and disordered temporal sequence of the original inspection images on the anomaly identification results. 3. By dividing the temperature map into blocks according to the curtain wall grid, and jointly calculating the average temperature, adjacent temperature difference and time-series temperature difference to determine the abnormal blocks, the anomaly identification can simultaneously reflect the temperature level of the block itself, local spatial differences and changes over time, thereby relatively suppressing misjudgments caused by a single local disturbance. 4. By mapping the abnormal blocks to specific rooms based on the facade room correspondence, and combining them with the operation table to form room abnormality records and analysis tables, a traceable correspondence can be established between the thermal anomalies of the exterior curtain wall and the room air conditioning activation status and set temperature, thereby providing a consistent data foundation for subsequent room-level merging analysis and action generation. 5. By merging the abnormal records of rooms in the same room that are consecutive in location, adjacent in time period and have the same set temperature offset direction into an abnormal chain, and performing temperature adjustment judgment, start-stop judgment and maintenance judgment on the abnormal chain, abnormal manifestations of different causes can be distinguished and processed, thereby relatively improving the targeting of action type recognition. 6. By performing cross-validation, conflict marking, rollback re-checking, and branch merging on the abnormal chain in the same room, and outputting an action list based on this, the conflict relationship in the concurrent scenario of multiple actions can be re-identified and reorganized, thereby relatively improving the problems of unclear action basis and missing execution order. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method steps of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Refer to the instruction manual appendix Figure 1 The present invention provides a method for generating building energy-saving optimization strategies based on UAV infrared imaging, comprising: S1. Acquire infrared images, timestamps, location coordinates, and air conditioning records collected by the drone during the inspection of the target building. Perform facade stitching on the infrared images according to the location coordinates. Extract the room activation status and set temperature from the air conditioning records according to the timestamps. Output temperature map and operation table. In this embodiment, S1 is used to first organize and align the inspection thermal image data and air conditioning operation data at the edge, so that subsequent anomaly block identification, room mapping, and action generation are all based on a unified data foundation. Specifically, the infrared images collected during the UAV inspection are first classified according to the exterior facade of the target building, and a continuous image sequence of the same facade is formed according to the order of acquisition. Then, temperature correspondence and boundary connection are performed on adjacent infrared images within the sequence to generate a temperature map that can directly participate in block calculation. Afterwards, based on the acquisition time corresponding to the facade image sequence, the air conditioning activation status and set temperature of the corresponding room at the same time are extracted from the air conditioning records and organized into an operation table according to room identification, thereby ensuring that the temperature map and the operation table correspond in both time and space dimensions. This implementation process includes the following steps: In S1-1, the infrared images collected by the UAV during the inspection of the target building's exterior facade, the timestamp corresponding to each infrared image, and the position coordinates corresponding to each infrared image are first acquired. Each infrared image is then written as an image record to the edge computing node. The position coordinates use the facade coordinates of the target building's exterior facade, which include at least the facade number, horizontal position, and vertical position to characterize the acquisition position of the infrared image on the corresponding facade. The timestamp is the acquisition time of the infrared image. After reading each image record, the edge computing node first categorizes the infrared images according to their facade numbers. Infrared images with the same facade number are grouped into the same facade set, and then sorted by timestamp from earliest to latest within the same facade set. When two or more infrared images with the same timestamp exist in the same facade set, they are sorted by horizontal position from smallest to largest. If the horizontal positions are the same, they are then sorted by vertical position from smallest to largest to determine the order of acquisition at the same time. After sorting, the infrared images arranged sequentially in the same facade set are written into a corresponding facade image sequence. Each infrared image in the facade image sequence retains its image identifier, timestamp, and position coordinates for subsequent overlapping area determination and stitching. In S1-2, after acquiring the facade image sequence, two adjacent infrared images from the same facade image sequence are read sequentially, and their coverage areas in the facade coordinates are determined based on their corresponding position coordinates. When the coverage areas of two infrared images intersect in the same facade coordinates, the intersection is defined as the overlapping area. When processing the overlapping area, the same position point in the facade coordinates is used as the corresponding reference, and the surface temperature values ​​of the two infrared images at that position point are read respectively, and these two sets of surface temperature values ​​are recorded as a set of temperature correspondence points. When there are multiple position points in the overlapping area, a set of temperature correspondence points is established one by one according to the position points. Then, boundary stitching is performed: for non-overlapping areas, the original infrared image at the corresponding position point is directly retained. The surface temperature value is calculated; for overlapping areas, the arithmetic mean of the two sets of surface temperature values ​​corresponding to the same location point is taken and the average value is written to that location point; then, according to the elevation coordinate order, the non-overlapping area, overlapping area of ​​the previous infrared image, and non-overlapping area of ​​the next infrared image are merged into a continuous temperature map; if the same elevation image sequence contains three or more infrared images, the continuous temperature map obtained in the previous step is used as the preceding stitching result, and the coverage determination, overlapping area extraction, temperature corresponding point generation, and boundary stitching are performed in the same way with the next infrared image until the entire elevation image sequence is processed, and finally the temperature map corresponding to the elevation is output; each location point in the temperature map corresponds to a unique elevation coordinate and a unique surface temperature value; In S1-3, after obtaining the air conditioning records and timestamps, the air conditioning records are first organized into a set of operation records stored according to room identifiers. Each air conditioning record includes at least the room identifier, recording time, room air conditioning activation status, and room air conditioning set temperature. The room air conditioning activation status indicates whether the corresponding room is in an air conditioning-on state at the recording time, and the room air conditioning set temperature indicates the set value for the corresponding room at the recording time. Then, using the timestamps corresponding to each infrared image in the facade image sequence as a benchmark, time matching is performed on the air conditioning records: for each timestamp, the record with a recording time less than or equal to that timestamp and the smallest time difference with that timestamp is found in the air conditioning records, and that record is... The record is identified as the air conditioning record corresponding to the timestamp. When multiple air conditioning records are matched for the same room under different timestamps, they are retained separately according to the timestamp and are not merged. After the time matching is completed, the matched room identifier, timestamp, room air conditioning activation status and room air conditioning set temperature are written into the operation table according to the room identifier. If the same room corresponds to multiple timestamps, they are written into the operation table in the order of the timestamp from earliest to latest. Each record in the operation table corresponds to the specific acquisition time in the facade image sequence. When mapping the abnormal blocks to the room and forming the room abnormal record, the room air conditioning activation status and room air conditioning set temperature of the corresponding room in the abnormal time period can be directly called according to the abnormal time period. Through the above processing, the facade classification, temporal organization, and continuous stitching of infrared images can be completed on the edge computing node first, and the corresponding extraction of air conditioning records and inspection times can be completed simultaneously. This yields a temperature map that can be directly used for subsequent block identification and an operation table that can be directly used for subsequent room matching. This avoids the confusion in facade attribution, acquisition order, and temperature values ​​of the original infrared images, and also avoids the disconnect in subsequent analysis caused by the inability to match air conditioning records with inspection times. It ensures that abnormal blocks, room abnormal records, and action lists in subsequent steps all have clear data sources and relationships. In practical applications: when the UAV inspects the east facade of the target building, the onboard processing unit or accompanying edge gateway first processes the multiple images acquired... Infrared images of the east facade are grouped into the same facade set according to the east facade number, and then sorted by acquisition time to form an east facade image sequence. The surface temperature values ​​of the overlapping areas of adjacent infrared images in the sequence are read point by point and averaged to complete the stitching, resulting in the east facade temperature map. Subsequently, based on the acquisition time of each infrared image in the sequence, the room operation record with the smallest time difference that is less than or equal to the acquisition time is extracted from the air conditioning records exported from the building air conditioning system. For example, if the record for a room at 10:12 is that the air conditioning is on and the set temperature is 24 degrees Celsius, then the room identifier, 10:12, on status, and 24 degrees Celsius are written into the operation table for direct use when mapping the east facade anomaly blocks to that room later.

[0018] S2. Obtain the temperature map, divide the temperature map into blocks according to the curtain wall grid, calculate the average temperature, adjacent temperature difference and time-series temperature difference of each block, identify the blocks that deviate as abnormal blocks, and output the abnormal block set. In this embodiment, S2 is used to further convert the temperature map output by S1 into a set of abnormal blocks that can participate in room mapping and motion generation. Its focus is not simply on identifying high or low temperature locations from the temperature map, but rather on first establishing unified block units according to the actual grid boundaries of the target building's exterior curtain wall. Then, it calculates the temperature of each block, its temperature relationship with adjacent blocks, and the temperature change relationship of the same block at different acquisition times. Finally, it identifies blocks with abnormal behavior based on comparable results within the same floor and facade. After this processing, the abnormal blocks called in subsequent steps all have clear block locations, clear temperature sources, and clear abnormal criteria, avoiding the problems of unclear boundaries, chaotic values, and unverifiable results caused by making empirical judgments directly on the entire temperature map. This implementation process includes the following steps: In S2-1, the temperature map output by S1 is first obtained, and the curtain wall segmentation data corresponding to the temperature map is obtained simultaneously. The curtain wall segmentation data is used to characterize the segmentation boundaries of the target building's exterior curtain wall in the facade coordinates. It can be derived from the building curtain wall segmentation diagram, facade component layout diagram, or a pre-established facade segmentation table, and each segment corresponds to a unique segmentation identifier, facade number, horizontal boundary, and vertical boundary. After reading the temperature map, the edge computing node extracts the corresponding area in the temperature map according to the horizontal and vertical boundaries of each segment in the curtain wall segmentation data, forming blocks that correspond one-to-one with the actual curtain wall segments. Each block corresponds to a unique block. The system identifies and uniquely positions the facades. When calculating the average temperature for each segment, it reads the surface temperature values ​​of all points within the segment boundary, sums all surface temperature values, and divides the sum by the number of points within the segment to obtain the average temperature of the segment. If there are points without temperature values ​​within a segment boundary, only points with temperature values ​​are included in the summation and counting. After calculating the average temperature of all segments, the segment identifier, facade number, segment location, data collection timestamp, and average temperature are written into the segment temperature table. The segment location is represented by the horizontal and vertical boundaries of the corresponding grid, which are used for subsequent adjacent relationship calculations and room mapping. In S2-2, after obtaining the segmented temperature table, the temperature difference between each segment and its adjacent segments is calculated under the same elevation and the same data collection time stamp. Adjacent segments are defined as two segments sharing a single grid boundary within the same elevation, excluding segments that only touch at corners. During calculation, the average temperature of each segment is read sequentially, along with the average temperature of each adjacent segment sharing the grid boundary. The average temperature of the segment is then subtracted from the average temperature of each adjacent segment to obtain the corresponding adjacent temperature differences. The segment identifier, adjacent segment identifier, data collection time stamp, and adjacent temperature difference are then written into the segmented difference table. Afterwards... The temperature difference of the same block at different timestamps is calculated as follows: the average temperature of the same block at two adjacent collection timestamps is read from early to late, the average temperature of the block at the later timestamp is subtracted from the average temperature of the earlier timestamp to obtain the time-series temperature difference of the block between the two collections, and the block identifier, the earlier timestamp, the later timestamp, and the time-series temperature difference are written into the block difference table. Through the above processing, the block difference table simultaneously saves the horizontal comparison results between blocks and the vertical comparison results of the same block between the collection times. It can be directly used to determine whether there is an abnormal deviation of the block. In S2-3, after obtaining the segmented temperature table and segmented difference table, the segments are grouped according to the facade number and collection timestamp. Within each group, the comparison benchmark for the average temperature is first determined, and then the comparison benchmarks for adjacent temperature differences and time-series temperature differences are determined. For the average temperature, the average temperature of all segments within the same facade, the same collection timestamp, and the same floor is read, and the arithmetic mean of these average temperatures is taken as the group average temperature of the segment. When the average temperature of a segment is higher or lower than the group average temperature, and its deviation direction is consistent with the direction of the time-series temperature difference change of the segment under the previous and next timestamps, the segment is recorded as having a deviation in average temperature. For adjacent temperature differences, all adjacent temperature differences between the segment and each adjacent segment are read. If the direction of the adjacent temperature differences between the segment and two or more adjacent segments is consistent, and the absolute value is greater than the arithmetic mean of the absolute values ​​of all adjacent temperature differences within the group, then the segment is considered to have a deviation in average temperature. If the temperature difference between adjacent blocks deviates, the block is recorded as having a deviation in temperature difference. For time-series temperature difference, the time-series temperature difference of the same block under two adjacent acquisition timestamps is read. If the time-series temperature difference is in the same direction as the time-series temperature difference of the previous time interval of the block and the absolute value increases, the block is recorded as having a deviation in temperature difference within the same block. After completing the above three judgments, the block that simultaneously satisfies the deviation in average temperature, the deviation in temperature difference between adjacent blocks, and the deviation in temperature difference within the same block is identified as an abnormal block. The block identifier, facade number, block location, acquisition timestamp, average temperature, adjacent temperature difference, and time-series temperature difference of the abnormal block are written into the abnormal block set. If a block only satisfies one or two of the deviations, it is not written into the abnormal block set but is retained in the block temperature table and the block difference table for comparison in the next round of acquisition, thereby avoiding the direct formation of abnormal blocks by a single local fluctuation. Through the above processing, the continuous temperature distribution in the temperature map can be transformed into a segmented temperature table, a segmented difference table, and an abnormal segment set corresponding to the actual segmentation of the curtain wall. This ensures that the abnormal segments called in subsequent steps no longer rely on the subjective judgment of the entire thermal image, but are based on the combined effect of three types of results: the average temperature within the segment, the temperature difference between adjacent segments, and the temporal temperature difference of the segment itself. This ensures that the spatial boundary of the abnormal segments is consistent with the actual structural boundary of the curtain wall, and also ensures that the formation process of the abnormal segments has a recalcible comparative basis, thereby reducing the direct impact of sunlight, local reflection, or single acquisition disturbance on the results, and providing clear input for subsequent room mapping and room anomaly record generation. In practical applications: taking the east facade of the target building as an example, the edge computing node first divides the temperature map of the east facade into multiple rectangular segments based on the curtain wall segmentation map of the east facade, and calculates the temperature of each segment separately. Calculate the average surface temperature of all locations within each block and write it into the block temperature table. Then, for a given block, calculate the temperature difference between it and the left, right, and upper blocks, and calculate the temporal temperature difference of that block during two inspections at 10:12 and 10:16. If the average temperature of that block at 10:16 is higher than the average temperature of the east facade group on the same floor, and the adjacent temperature differences between it and the left and right blocks are all positive and their absolute values ​​are all greater than the average absolute value of the adjacent temperature differences in that group, and the temporal temperature difference of that block between 10:12 and 10:16 is positive and further increases compared to the previous temporal interval, then that block is identified as an abnormal block, and its block identifier, location range, 10:16 timestamp, average temperature, adjacent temperature difference, and temporal temperature difference are written into the abnormal block set for subsequent mapping to the corresponding room.

[0019] S3. Obtain the abnormal block set, the facade room correspondence table and the operation table. Map each abnormal block to the corresponding room to form a room abnormal record. Write the block location, abnormal time period, room activation status and set temperature into the analysis table and output the analysis table. In this embodiment, S3 is used to establish a traceable correspondence between the abnormal block set obtained in S2 and the room operation information, so that the abnormal manifestations on the exterior curtain wall are converted into room abnormal records that can be analyzed by room, and further form an analysis table for subsequent abnormal chain merging and action generation. This process does not simply point the abnormal block directly to a certain room, but first completes spatial mapping based on the facade room correspondence, then completes time matching based on the operation table, and then organizes the block location, abnormal time period, room air conditioning activation status and set temperature into room abnormal records according to a unified field, and finally writes them into the analysis table according to the room identifier, so as to ensure that the number of abnormalities, abnormal duration and set temperature offset in subsequent steps have a clear source. This implementation process includes the following steps: In S3-1, the abnormal block set and the facade-room correspondence table are first obtained. Each abnormal block record in the abnormal block set includes at least the block identifier, facade number, block location, collection timestamp, average temperature, adjacent temperature difference, and time-series temperature difference. The facade-room correspondence table represents the correspondence between the target building's exterior curtain wall block locations and rooms. This table can be pre-generated based on the building facade grid diagram, floor plan, and room number table, and each correspondence includes at least the facade number, block boundary, room identifier, and room outer boundary range. After reading each abnormal block record, the edge computing node first searches for candidate rooms under the same facade in the facade-room correspondence table according to the facade number of that abnormal block, and then... The location of the abnormal block is compared with the outer boundary of each candidate room. When the outer boundary of a room covers the boundary of the abnormal block, that room is identified as the corresponding room of the abnormal block. If an abnormal block corresponds to multiple rooms, the overlap length or overlap area between the boundary of the abnormal block and the outer boundary of each room is calculated, and the room with the largest overlap is identified as the corresponding room of the abnormal block. If the overlap results are the same, the room with the smaller room number is retained as the corresponding room to eliminate mapping ambiguity. After the above mapping is completed, the block identifier, facade number, block location, collection timestamp, and corresponding room identifier are written into the room mapping table so that each abnormal block corresponds to a unique room identifier. In S3-2, after obtaining the room mapping table and the operation table, operation status matching is performed on each mapping record in the room mapping table. Each record in the operation table includes at least a room identifier, a matching timestamp, the room's air conditioning activation status, and the room's air conditioning set temperature. After the edge computing node reads the corresponding room identifier and collection timestamp from a room mapping record, it searches the operation table for an operation record with the same room identifier and a matching timestamp that matches the collection timestamp. The room's air conditioning activation status and room's air conditioning set temperature in this operation record are then used as the operation status of the abnormal segment at the collection time. If no record with the exact same timestamp exists in the operation table, the node searches for an operation record with the same room identifier, a matching timestamp less than the collection timestamp, and the smallest time difference between the matching timestamp and the collection timestamp. Record the operation and identify the corresponding operation record. If the same block identifier appears in the room mapping table under multiple consecutive collection timestamps, first arrange the multiple mapping records corresponding to the block in order of collection timestamp from earliest to latest, and then merge the adjacent mapping records with the same room identifier into the same abnormal time period. The start time of the abnormal time period is taken from the collection timestamp of the first mapping record, and the end time is taken from the collection timestamp of the last mapping record. If a block appears only under a single collection timestamp, then that collection timestamp is used as both the start and end time of the abnormal time period. After completing the operation status matching, write the block identifier, facade number, block location, corresponding room identifier, abnormal time period, room air conditioning activation status, and room air conditioning set temperature into the room matching table. In S3-3, after obtaining the room matching table, the records in the room matching table are read one by one according to the room identifier and the block identifier. The block position, abnormal time period, room air conditioner activation status and set temperature in each record are combined to generate the corresponding room abnormal record. The block position follows the block boundary corresponding to the abnormal block, the abnormal time period follows the start time and end time formed in S3-2, and the room air conditioner activation status and set temperature adopt the operating status corresponding to the abnormal time period. If there are multiple different block identifiers under the same room identifier, multiple room abnormal records are generated respectively. If there is no time interval between two abnormal time periods for the same room identifier and the same block identifier, the two abnormal time periods are merged into a continuous abnormal time period, and then a room abnormal record is generated. Each room abnormal record includes at least the room identifier, block identifier, block position, abnormal start time, abnormal end time, room air conditioner activation status and set temperature. After generation, all room abnormal records are written into the room abnormal record set. Each record in the room abnormal record set directly corresponds to an abnormal block behavior of a room in an abnormal time period. In S3-4, after obtaining the set of room anomaly records, all room anomaly records are written and organized according to room identifiers. Specifically, an analysis table indexed by room identifiers is first established, then records in the set of room anomaly records are read one by one, and room anomaly records with the same room identifier are continuously written into the same room record area in the analysis table. Each record in the analysis table retains the room identifier, block identifier, block position, anomaly start time, anomaly end time, room air conditioning status, and set temperature. Different blocks or different time periods are not summarized in this step to ensure that S4 can directly merge, form anomaly chains, and determine actions based on the original room anomaly records. If multiple room anomaly records correspond to the same room, they are written into the analysis table in order of anomaly start time from earliest to latest. If the anomaly start time is the same, they are written into the analysis table in order of block position from smallest to largest horizontal position on the facade. After all writing is completed, the analysis table is output so that each record in the analysis table can be traced back to the specific anomaly block and specific operating status. Through the above processing, the abnormal curtain wall results represented by facade location in the abnormal block set can be converted into an analysis table organized by room identifier. This ensures that each room abnormal record retains spatial location, time range, and operational status information simultaneously, providing a direct basis for subsequent room-based statistics of abnormal block quantity, calculation of abnormal duration, and formation of set temperature offset. This avoids problems such as unclear mapping relationship between abnormal blocks and rooms, unclear source of abnormal time period, and incompatibility of operational status. It also ensures that subsequent temperature adjustment actions, start-stop actions, and maintenance actions can be traced back to specific rooms, specific blocks, and specific abnormal time periods. In practical application: taking the east facade of the target building as an example, if the facade number of a certain abnormal block record in the abnormal block set is east facade, the block position is column 8, row 12, and the collection timestamp is 10:16, the edge computing node first looks up the room corresponding to column 8, row 12 of the east facade in the facade room correspondence table. If the main To correspond to room 1208, room 1208 is identified as the corresponding room and written into the room mapping table. Then, the operation table is searched for the operation record corresponding to room 1208 at 10:16. If a record for 10:16 exists in the operation table, the room's air conditioning status and set temperature are directly read. If not, the record earlier than 10:16 with the smallest time difference is read. If the abnormal block appears consecutively at 10:12, 10:16, and 10:20, and the corresponding room obtained from all three mappings is room 1208, then 10:12 is taken as the abnormal start time, and 10:20 as the abnormal end time, forming the abnormal time period corresponding to this block. The block location, 10:12 to 10:20, the air conditioning status of room 1208, and the set temperature of 24 degrees Celsius are combined into a room abnormal record, and then written into the analysis table according to the room identifier 1208, for subsequent abnormal chain merging and action generation for room 1208.

[0020] S4. Obtain the analysis table, merge room abnormal records according to room identifier, count the number of abnormalities, duration of abnormalities and set temperature deviation of each room, and generate temperature adjustment actions, start-stop actions and maintenance actions accordingly, and output the action list. In this embodiment, S4 is used to further convert the room anomaly records in the analysis table into executable energy-saving action results. Its focus is not on providing conclusions for each individual room anomaly record, but rather on first merging them according to room identifiers, then forming anomaly chains around the anomaly locations, anomaly periods, and set temperature changes within the same room. Subsequently, based on the spatial continuity, temporal continuity, and set temperature offset continuity of the anomaly chains, it distinguishes between three types of action sources: temperature adjustment, start / stop, and maintenance. It also performs cross-validation, conflict rollback, and re-merging on multiple types of actions that may correspond to the anomaly chain in the same room, ultimately forming an action list. This process avoids directly interpreting unclosed anomaly processes within the same room as a single action, and also avoids different action types overlapping or directly conflicting within the same room. This implementation process includes the following steps: In S4-1, the analysis table is first obtained, and all room anomaly records are read according to the room identifier. Each room anomaly record in the analysis table includes at least the room identifier, block identifier, block location, anomaly start time, anomaly end time, room air conditioning status, and set temperature. The edge computing node establishes a room merging table using the room identifier as an index, and writes room anomaly records with the same room identifier in the analysis table into the same room merging unit. During the writing process, the anomaly block location, anomaly time period, room air conditioning status, and set temperature corresponding to the room are extracted one by one, where the anomaly time period consists of the anomaly start time and the anomaly end time. If there are multiple room anomaly records under the same room... For inter-room anomaly records, they are first sorted by the anomaly start time from earliest to latest. When the anomaly start times are the same, they are sorted by the horizontal position of the block on the corresponding facade from smallest to largest. If the horizontal positions are the same, they are then sorted by the vertical position from smallest to largest. After merging, the room identifier, the corresponding positions of all anomaly blocks, the corresponding time periods, the corresponding room air conditioning status, and the corresponding set temperature are written into the room merging table. This ensures that each room merging unit in the room merging table retains all the basic fields of the original room anomaly records. When calculating the number of anomaly blocks, the duration of anomalies, and the set temperature offset, this table is directly called, and the analysis table is not read repeatedly. In S4-2, after obtaining the room merging table, the room merging units are read one by one according to the room identifier. For each room, the number of abnormal blocks, the duration of abnormality, and the set temperature offset are calculated separately. The number of abnormal blocks is the number of different block identifiers appearing in the current merging unit for that room. If the same block identifier appears repeatedly in multiple abnormal periods, it is only counted as one abnormal block. The duration of abnormality is the sum of the durations of each segment formed by subtracting the start time of abnormality from the end time of all abnormal records in that room. If the abnormal periods of two consecutive abnormal records are consecutive, they are considered as consecutive abnormal periods before calculating the consecutive duration. The set temperature offset is obtained by reading the set temperature in all abnormal records of that room according to the room identifier, and subtracting the set temperature of the earliest abnormal record in the analysis table for that room from the set temperature of each abnormal record to obtain the set temperature offset of that abnormal record. If the difference is positive, it is recorded as a positive offset; if the difference is negative, it is recorded as a negative offset; if the difference is zero, it is recorded as no offset. After completing the above calculations, an anomaly chain is formed within the same room: The sorted room anomaly records are read one by one. When the anomaly block position of the subsequent room anomaly record shares a boundary or corresponding consecutive number with the anomaly block position of the preceding room anomaly record on the facade grid, they are recorded as consecutively connected. When the start time of the anomaly in the subsequent room anomaly record is equal to the end time of the anomaly in the preceding room anomaly record, or the time interval between them is equal to one sampling interval, they are recorded as adjacent anomaly periods. When the set temperature offsets corresponding to the two room anomaly records are both positive or both negative, they are recorded as having the same set temperature offset direction. Only when all three conditions are met simultaneously are the two room anomaly records merged into the same anomaly chain. If subsequent room anomaly records within the same room continue to meet the above conditions, they are then merged into the anomaly chain. After merging, the room identifier, anomaly chain identifier, number of anomaly blocks, anomaly duration, set temperature offset, the range of block positions constituting the anomaly chain, and the range of anomaly periods are written into the room anomaly chain list. In S4-3, after obtaining the room anomaly chain, temperature adjustment judgment, start / stop judgment, and maintenance judgment are performed on each room anomaly chain. During temperature adjustment judgment, the room air conditioner's on / off status, set temperature offset, and anomaly duration corresponding to the room anomaly chain are read. If the room air conditioner remains on during the anomaly period corresponding to the anomaly chain, the set temperature offset is not zero, and the direction of the set temperature offset remains consistent throughout the entire anomaly chain, the room anomaly chain is determined to meet the temperature adjustment judgment, and a temperature adjustment token is written. During start / stop judgment, the room air conditioner's on / off status at the start and end points of the anomaly period corresponding to the room anomaly chain is read. If the room air conditioner's on / off status changes from off to on or from on to off during the formation of the anomaly chain, and the time of this change falls within the anomaly period range of the anomaly chain, then... The abnormal chain in a room is determined to meet the start / stop criteria, and a start / stop token is written. During maintenance, the duration of the abnormality, the number of abnormal blocks, and the set temperature offset of the abnormal chain in that room are read. When the duration of the abnormality spans two or more consecutive abnormal periods, the number of abnormal blocks remains unchanged, and the set temperature offset is zero or there is no directional change between abnormal records in previous and subsequent rooms, the abnormal chain in that room is determined to meet the maintenance criteria, and a maintenance token is written. After completing the three types of criteria, the number of abnormal blocks, duration of the abnormality, set temperature offset, temperature adjustment criteria, start / stop criteria, and maintenance criteria corresponding to the abnormal chain in that room are written to the corresponding criteria, and the temperature adjustment token, start / stop token, or maintenance token is written to the room criteria table. If an abnormal chain in a room does not meet the corresponding criteria, no token is written under that criteria. In S4-4, after obtaining the room determination table, the system searches for the same room's abnormal chain that has different tokens written simultaneously, based on the room identifier and the abnormal chain identifier, and performs cross-validation on that room's abnormal chain. During cross-validation, the system first reads the set temperature offset change sequence and the abnormal duration change sequence of the room's abnormal chain. The set temperature offset change sequence is arranged according to the chronological order of the abnormal records in each room within the abnormal chain, and the abnormal duration change sequence is arranged according to the cumulative duration of each consecutive abnormal period within the abnormal chain. When the set temperature offset change sequence maintains the same offset direction and does not reverse direction or interrupt as the abnormal duration increases, the room's abnormal chain is recorded as having the same set temperature offset and abnormal duration. When the set temperature offset change sequence reverses direction, returns to zero, and then offsets again... If the duration of the abnormality is interrupted but the direction of the set temperature offset still changes, the abnormal chain for that room is recorded as having an inconsistent set temperature offset and duration of abnormality. For consistent abnormal chains for that room, the action type corresponding to the first token written is retained, and the abnormal chain for that room is written as the priority action chain. If both temperature control tokens and start / stop tokens exist in the same abnormal chain for that room, the temperature control token is retained first. If maintenance tokens and other tokens exist, the other tokens before the maintenance token are retained first. The maintenance token is retained only if neither temperature control tokens nor start / stop tokens exist. For inconsistent abnormal chains for that room, a conflict flag is written to the abnormal chain record for that room, and the room identifier, abnormal chain identifier, type of token written, sequence of changes in set temperature offset, and sequence of changes in duration of abnormality are written to the conflict chain list. In S4-5, after obtaining the conflict chain list, a rollback and re-examination is performed on each room exception chain marked with a conflict, based on the exception time period. During the rollback and re-examination, the original room exception records that make up the exception chain are traced back based on the room exception chain identifier, and the previously merged room exception chain is split back into multiple room exception records arranged by exception time period. Then, using each room exception record as a basic unit, its corresponding exception block position, exception start time, exception end time, room air conditioning status and set temperature are re-examined, and a rollback exception chain is re-established based on the shorter exception time period. If the action types corresponding to the two split rollback exception chains are the same, and the exception end time of the previous rollback exception chain is the same as the exception start time of the next rollback exception chain, the rollback exception chain is re-established. If the timings are consecutive, or if the block positions of two rollback anomaly chains are consecutively connected, then they are re-executed and merged into a new merged anomaly chain. If the action types are different, they are retained as independent branches. After completing the rollback anomaly chain reconstruction and branch merging, the temperature adjustment judgment, start / stop judgment, and maintenance judgment in S4-3 are re-executed on the merged room anomaly chain, and the temperature adjustment token, start / stop token, or maintenance token is updated according to the re-judgment results. Finally, the room identifier, anomaly chain identifier, updated token type, number of anomaly blocks, anomaly duration, and set temperature offset are written into the room action table. Through this process, the action conflicts caused by the excessively wide merging of anomaly chains can be decomposed and the source of the action can be restored at a finer granular level. In S4-6, after obtaining the room action table, extract the room anomaly chains containing temperature control tokens, start / stop tokens, and maintenance tokens one by one according to the room identifier, and generate temperature control actions, start / stop actions, and maintenance actions respectively. When generating a temperature control action, write the room identifier, the number of anomaly blocks in the room's anomaly chain, the anomaly duration, the set temperature offset, and the action type "temperature control action" into the action list. When generating a start / stop action, write the room identifier, the corresponding number of anomaly blocks, the anomaly duration, the set temperature offset, and the action type "start / stop action" into the action list. When generating a maintenance action, ... Write the room identifier, the corresponding number of abnormal blocks, the duration of the abnormality, the set temperature offset, and the action type "maintenance action" into the action list. If there are multiple different abnormal chains in the room action table for the same room and they correspond to the same action type, write multiple action records for each chain. If there are multiple different abnormal chains in the same room and they have different action types, retain them all and write them into the action list separately, without sorting them in this step. After completing all extraction and writing, output the action list so that each action record in the action list can be traced back to the specific room abnormal chain and its corresponding number of abnormal blocks, duration of the abnormality, and set temperature offset. Through the above processing, the room anomaly records in the analysis table can be first organized into a calculable, comparable, and rollback-capable room anomaly chain. This chain is then further transformed into an action list with clearly defined action types and justifications. This solves the problem of a single room anomaly record being difficult to directly interpret as an energy-saving action, as well as the problem of cross-conflict when multiple types of actions occur simultaneously in the same room. Furthermore, by using conflict chain rollback and branch merging, it ensures that each action in the final output action list has a clear data source, a clear formation process, and a clear action attribution. In practical applications: For example, room 1208 has three abnormal room records in the analysis table. The first record corresponds to the 8th column, 12th row of the east facade, with an abnormal period from 10:12 to 10:20. The room's air conditioning was on, and the set temperature was 24 degrees Celsius. The second record corresponds to the adjacent 9th column, 12th row, with an abnormal period from 10:20 to 10:28. The room's air conditioning was still on, and the set temperature was 25 degrees Celsius. The third record corresponds to the same area, with an abnormal period from 10:28 to 10:36. The activation status is changed to deactivated, while the set temperature remains at 25 degrees Celsius. The edge computing node first merges the three room anomaly records according to room identifier 1208, calculating the number of anomaly blocks as 2, the cumulative anomaly duration as 24 minutes, and the set temperature offset of the latter two records relative to the earliest set temperature of 24 degrees Celsius as a positive offset of 1 degree Celsius. Since the first two room anomaly records are consecutive in block position, adjacent in time period, and have the same set temperature offset direction, they are merged into the same anomaly chain and the temperature adjustment token is written first. The third room anomaly record... The start / stop token is written when state switching is enabled; then cross-validation is performed on the abnormal chain of the room. If the set temperature offset is found to remain positive while the duration of the abnormality is continuously increasing, the temperature adjustment token is retained as the priority action chain; if the set temperature offset direction changes repeatedly and the duration of the abnormality is interrupted in another room, it is written into the conflict chain and the original room abnormality record is removed and re-examined; finally, the temperature adjustment action of room 1208 or the start / stop action and maintenance action of other rooms can be obtained in the action list, which can be used for subsequent sorting and grouping to generate building energy-saving optimization strategies.

[0021] S5. Obtain the action list, sort it in the order of temperature adjustment actions, start-stop actions, and maintenance actions in the same room, group it in the order of the number of rooms in the same control area from few to many, generate the sorting and grouping results into a building energy-saving optimization strategy, and output the building energy-saving optimization strategy of the target building. In this embodiment, S5 is used to further organize the action list output by S4 into a building energy-saving optimization strategy that can be directly sent to the building management terminal. Its purpose is not simply to summarize temperature control actions, start-up / stop actions, and maintenance actions, but rather to first establish action classification relationships by room and control area, then form a room execution sequence with a sequential order within each room, and finally group each room execution sequence according to the actual number of rooms involved in the actions within the control area, ultimately generating a building energy-saving optimization strategy with area affiliation, execution order, and batch structure. Since multiple rooms within the same control area often share the same air conditioning control unit, if the room internal sequence is not first formed and then the area execution sequence is further formed, problems such as overlapping action execution objects, mutual interference of actions within the area, and chaotic strategy distribution order can easily occur. This implementation process includes the following steps: In S5-1, the action list is first obtained, and the correspondence between rooms and control areas is simultaneously acquired. Each action record in the action list includes at least a room identifier, the number of abnormal blocks, the duration of the abnormality, the set temperature offset, and the action type, which can be a temperature adjustment action, a start / stop action, or a maintenance action. The correspondence between rooms and control areas is provided by the building air conditioning zoning configuration table. Each configuration record includes at least a room identifier and a control area identifier, used to represent the air conditioning linkage control area to which the room belongs. The edge computing node reads the action records in the action list one by one, first writing the action records into the room action table based on the room identifier, and then looking up the corresponding control area identifier in the building air conditioning zoning configuration table based on the room identifier. The action record, along with the control area identifier, is written into the area action table. If multiple action records exist for the same room, all are written into the room action table, and the same control area identifier is retained in the area action table. After writing, each record in the room action table includes at least the room identifier, action type, number of abnormal blocks, duration of abnormality, and set temperature offset. Each record in the area action table includes at least the control area identifier, room identifier, action type, number of abnormal blocks, duration of abnormality, and set temperature offset. Through this process, action records can be simultaneously organized into a room action table read by room and an area action table read by control area, providing a unified input for subsequent room internal sorting and area grouping. In S5-2, after obtaining the room action table, all action records corresponding to the same room are sorted according to the room identifier. During sorting, all action types corresponding to the room are read first, and then temperature adjustment actions are arranged in a predetermined order, followed by start / stop actions and maintenance actions. This order is used because temperature adjustment actions directly correspond to the set temperature adjustment, have low execution costs, and can be restored immediately; start / stop actions involve adjustments to the air conditioning operating period, requiring coordination with the current operating status; and maintenance actions involve on-site handling of the building envelope, have a longer execution cycle, and are usually not completed in the immediate control chain. Therefore, the room's internal execution sequence is formed by first ordering temperature adjustment actions, then start / stop actions, and finally maintenance actions. If two or more action records of the same action type exist in the same room, they are sorted from longest to shortest abnormal duration; if the abnormal duration is the same, they are sorted from most to least abnormal block count; if the abnormal block count is still the same, they are sorted from largest to smallest absolute value of set temperature offset. After sorting, all action records in the room are written into the room execution sequence according to the sorting result, and the room identifier, sequence position, action type, number of abnormal blocks, abnormal duration, and set temperature offset are written into the room execution table. For each room, only one room execution sequence is generated, and the action order in the room execution sequence is the internal execution order when the room participates in the subsequent area grouping. In S5-3, after obtaining the room execution table and the region action table, the room execution sequences are first grouped according to the joint control region identifier, and room execution sequences belonging to the same joint control region are grouped into the same region candidate set. The number of rooms within the same joint control region is determined by the actual number of rooms in that region with action records under the current action list, not by the total number of rooms configured in that region. The edge computing nodes count the number of rooms with action records in the region candidate sets of each joint control region, and sort the joint control regions from fewest to most rooms. When two joint control regions have the same number of rooms, they are sorted from smallest to largest by the joint control region identifier. Subsequently, the execution sequences are processed sequentially according to the sorted joint control regions. The system reads the room execution sequences corresponding to each control area and combines the room execution sequences within the same control area into a single area execution sequence. Within the same control area, if multiple room execution sequences exist, they are arranged in ascending order of room identifier. If no identical room identifiers exist, no further decision-making is made. After grouping, the control area identifier, room identifier, sequence position in the room execution sequence, action type, number of abnormal blocks, abnormal duration, and set temperature offset are written into the area execution sequence, and the area execution table is output. This process prioritizes control areas with fewer rooms involved in the action, thereby reducing the impact of simultaneous adjustments in multiple rooms within the same area. In S5-4, after obtaining the regional execution table, the execution sequences of each region are read sequentially according to the joint control region identifier, and the execution sequences of each region are summarized in the sorted regional order to generate a building energy-saving optimization strategy. Specifically, during generation, a regional strategy record is first created for each joint control region, and then all regional execution sequence contents corresponding to that joint control region are written into that regional strategy record. Each regional strategy record includes at least the joint control region identifier, room identifier, action type, sequence position, number of abnormal blocks, abnormal duration, set temperature offset, and strategy generation time. Once all regional strategy records are summarized... Then, the building energy-saving optimization strategy corresponding to the target building is generated, and the building energy-saving optimization strategy is written into the strategy output area or sent to the building management terminal. The building energy-saving optimization strategy is not a single action suggestion, but an ordered set of results composed of multiple regional strategy records. After reading it, the building management terminal can display or distribute it in batches according to the control area. Moreover, each regional strategy record can be traced back to the corresponding room execution sequence and original action record. In this way, the final output building energy-saving optimization strategy retains the room-level action basis and has the organizational structure of the control area, which is convenient for subsequent regional execution, review and write-back. Through the above processing, the results of various actions in the action list can be first implemented at the room and control area levels. Then, a clear sequence of room execution is formed within each room, and a region execution sequence ordered by the actual number of rooms performing the actions is formed at the control area level. This ultimately yields a building energy-saving optimization strategy with region affiliation, room affiliation, action sequence, and batch structure. This ensures that multiple actions within the same room are not issued haphazardly, and that actions in multiple rooms within the same control area are organized in an orderly manner according to their regional impact, thereby improving the executability of the strategy results and its direct usability to the management side. In practical applications: for example, if the action list includes temperature control and start / stop actions for room 1208, maintenance actions for room 1210, and temperature control actions for room 1503, and the building air conditioning zoning configuration table shows that rooms 1208 and 1210 belong to control area A, while room 1503 belongs to control area B, then the edge computing node will first implement the actions for rooms 1208 and 1210... The actions of room 1503 and room 1208 are written into the room action table, and then into the area action table according to the building air conditioning zoning configuration table. Subsequently, a room execution sequence is formed for room 1208 with temperature adjustment actions first and start / stop actions last. A room execution sequence is formed for room 1210 with maintenance actions retained, and a room execution sequence is formed for room 1503 with temperature adjustment actions retained. Then, the number of rooms with action records in control area A is counted as 2, and the number of rooms with action records in control area B is counted as 1. The execution sequence of room 1503 corresponding to control area B is first written into the area execution sequence, and then the execution sequences of rooms 1208 and 1210 corresponding to control area A are written into the area execution sequence. Finally, these are summarized to form the building energy-saving optimization strategy for the target building, which includes at least the control area identifier, room identifier, action type, execution order, number of abnormal blocks, duration of abnormality, set temperature offset, and strategy generation time, for the building management terminal to call in batches by area.

[0022] Working principle: First, a drone equipped with infrared imaging equipment inspects the building's exterior curtain wall, acquiring infrared images with timestamps and location coordinates. The images are then categorized, sequentially arranged, and stitched together along the edges of the facade to create temperature maps for each facade. Simultaneously, the system extracts the air conditioning status and set temperature of the rooms corresponding to the inspection time, creating an operation table. Next, the system divides the temperature maps into blocks according to the curtain wall's structural sections, calculating the average temperature of each block, the temperature difference with adjacent blocks, and the temperature changes over time, identifying abnormal blocks. Finally, based on the correspondence between rooms on the facade, the abnormal blocks are mapped to specific... The system collects room-specific data and combines it with operation tables to form room anomaly records and analysis tables. Based on this, the system merges anomaly records by room, calculates the number of anomaly blocks, the duration of anomalies, and the set temperature offset. It further merges anomaly records that are spatially continuous, temporally continuous, and have the same temperature offset direction within the same room into anomaly chains. Then, it determines whether the anomaly chain is more suitable to be interpreted as a temperature control problem, a start-up / stop problem, or a building envelope maintenance problem. Finally, the obtained actions are sorted, grouped, and summarized by room and control area to generate energy-saving optimization strategies that can be directly provided to the building management terminal. For example, in an office building, when a drone inspects the east facade and finds several areas of the curtain wall that are consistently hot, the system doesn't immediately conclude that this is due to poor insulation. Instead, it first identifies which rooms these abnormal areas correspond to, and then checks whether the air conditioning in those rooms was on or off during the same time period, and whether the set temperature was raised or lowered. If abnormal areas appear consecutively in a room, and the air conditioning is constantly on with the set temperature continuously deviating, the system will prioritize generating a temperature adjustment action. If the abnormality occurs while the air conditioning is switching on and off, a start / stop adjustment action will be generated. If the abnormality persists for a long time, but the air conditioning settings remain largely unchanged, the system is more likely to generate a building envelope maintenance action. Finally, these actions are organized according to the order of temperature adjustment, start / stop, and maintenance within the same room, and then grouped and output by control area. This way, property management or maintenance personnel can know which rooms to adjust first, which areas need to be grouped for processing, and which issues belong to subsequent maintenance items.

[0023] The above description is merely 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 within the protection scope of the present invention.

Claims

1. A method for generating building energy-saving optimization strategies based on UAV infrared imaging, characterized in that, include: S1. Acquire infrared images, timestamps, location coordinates, and air conditioning records collected by the drone during the inspection of the target building. Perform facade stitching on the infrared images according to the location coordinates. Extract the room activation status and set temperature from the air conditioning records according to the timestamps. Output temperature map and operation table. S2. Obtain the temperature map, divide the temperature map into blocks according to the curtain wall grid, calculate the average temperature, adjacent temperature difference and time-series temperature difference of each block, identify the blocks that deviate as abnormal blocks, and output the abnormal block set. S3. Obtain the abnormal block set, the facade room correspondence table and the operation table. Map each abnormal block to the corresponding room to form a room abnormal record. Write the block location, abnormal time period, room activation status and set temperature into the analysis table and output the analysis table. S4. Obtain the analysis table, merge room abnormal records according to room identifier, count the number of abnormalities, duration of abnormalities and set temperature deviation of each room, and generate temperature adjustment actions, start-stop actions and maintenance actions accordingly, and output the action list. S5. Obtain the action list, sort it in the order of temperature adjustment actions, start-stop actions, and maintenance actions in the same room, group it in the order of the number of rooms in the same control area from fewest to most, generate the sorting and grouping results into a building energy-saving optimization strategy, and output the building energy-saving optimization strategy for the target building.

2. The method for generating building energy-saving optimization strategies based on UAV infrared imaging according to claim 1, characterized in that: S1 includes: S1-1. Obtain infrared images, timestamps, and location coordinates. Classify the infrared images according to their location coordinates and arrange the infrared images of the same facade in chronological order according to their timestamps. Output the facade image sequence. S1-2. Obtain the facade image sequence, perform temperature point matching and boundary stitching on the overlapping areas of adjacent infrared images in the facade image sequence, and output the temperature map. S1-3. Obtain air conditioning records and timestamps. Extract the room activation status and set temperature corresponding to the time period of the facade image sequence from the air conditioning records according to the timestamps, and write them into the operation table according to the room identifier. Output the operation table.

3. The method for generating building energy-saving optimization strategies based on UAV infrared imaging according to claim 2, characterized in that: S2 includes: S2-1. Obtain the temperature map, divide the temperature map into blocks according to the curtain wall grid, calculate the average temperature of each block, and output the block temperature table. S2-2. Obtain the block temperature table, calculate the temperature difference between each block and its adjacent blocks, as well as the temperature difference of the same block at different timestamps, and output the block difference table. S2-3. Obtain the block temperature table and block difference table. Identify blocks whose average temperature deviates, whose adjacent block temperature difference deviates, and whose same block temperature difference deviates as abnormal blocks, and output the abnormal block set.

4. The method for generating building energy-saving optimization strategies based on UAV infrared imaging according to claim 3, characterized in that... : S3 includes: S3-1. Obtain the set of abnormal blocks and the facade room correspondence table, map each abnormal block to a room according to the facade location, and output the room mapping table. S3-2. Obtain the room mapping table and the operation table, match the room activation status and set temperature for each abnormal block corresponding to the time period, and output the room matching table.

5. The method for generating building energy-saving optimization strategies based on UAV infrared imaging according to claim 4, characterized in that... : S3 further includes: S3-3. Obtain the room matching table, combine the block location, abnormal time period, room activation status and set temperature of each abnormal block to generate room abnormal records, and output the set of room abnormal records. S3-4. Obtain the set of abnormal room records, write them into the analysis table according to the room identifier, and output the analysis table.

6. The method for generating building energy-saving optimization strategies based on UAV infrared imaging according to claim 5, characterized in that: S4 includes: S4-1. Obtain the analysis table, merge room anomaly records according to room identifier, and extract the anomaly block location, anomaly time period, room activation status and set temperature corresponding to each room, and output the room merge table. S4-2. Obtain the room merging table, calculate the number of abnormal blocks, duration of abnormality, and set temperature offset according to the room identifier, and merge the abnormal room records with consecutive abnormal block positions, adjacent abnormal time periods, and consistent set temperature offset direction into the same abnormal chain, and output the room abnormal chain list.

7. The method for generating building energy-saving optimization strategies based on UAV infrared imaging according to claim 6, characterized in that: S4 further includes: S4-3. Obtain the room anomaly chain list, perform temperature adjustment judgment, start / stop judgment and maintenance judgment on each room anomaly chain, write the number of anomaly blocks, the duration of anomaly and the set temperature offset into the corresponding judgment result, and write temperature adjustment token, start / stop token or maintenance token to the room anomaly chain that passes the judgment, and output the room judgment table. S4-4. Obtain the room determination table, perform cross-validation on the same room's abnormal chain that simultaneously writes different tokens, retain the room's abnormal chain with the set temperature offset and abnormal duration as the priority action chain, and write the room's abnormal chain with the set temperature offset and abnormal duration into the conflict flag and output the conflict chain list.

8. The method for generating building energy-saving optimization strategies based on UAV infrared imaging according to claim 7, characterized in that: S4 further includes: S4-5. Obtain the conflict chain list, perform rollback and re-check on the room exception chain with the conflict mark written, re-execute the branch merge with the other exception chains of the corresponding room after the rollback, update the temperature control token, start / stop token or maintenance token for the merged room exception chain, and output the room action table. S4-6. Obtain the room action table, extract the temperature adjustment action written with the temperature control token, the start / stop action written with the start / stop token, and the maintenance action written with the maintenance token according to the room identifier, and write the room identifier, number of abnormal blocks, duration of abnormality, set temperature offset and action type corresponding to each action into the action list, and output the action list.

9. The method for generating building energy-saving optimization strategies based on UAV infrared imaging according to claim 8, characterized in that: S5 includes: S5-1. Obtain the action list, classify the temperature control actions, start-up and shutdown actions and maintenance actions according to the room identifier and the control area identifier, and output the room action table and the area action table. S5-2. Obtain the room action table, sort it according to the order of temperature adjustment action, start / stop action, and maintenance action in the same room, write the sorting result into the room execution sequence, and output the room execution table.

10. The method for generating building energy-saving optimization strategies based on UAV infrared imaging according to claim 9, characterized in that: S5 also includes: S5-3. Obtain the room execution table and the area action table. Group the room execution sequences in order of increasing number of rooms in the same control area, write the grouping results into the area execution sequence, and output the area execution table. S5-4. Obtain the regional execution table, summarize the execution sequences of each region according to the joint control region identifier, generate the building energy-saving optimization strategy, and output the building energy-saving optimization strategy of the target building.

Images