Adaptive regulation system and method fusing space features and comfort perception

Abstract

The application provides a kind of adaptive regulation system and method of fusing space features and comfort perception, belongs to building energy consumption regulation technical field.Building space is divided into multiple partitions and constructs space feature vector;Real-time collection personnel body state, environment and energy consumption multi-source data, binding space-time coordinates;Based on the length of stay to identify effective users;Individual comfort prototype library is constructed, and individual comfort score is calculated by matching real-time body state vector;Fusion of group preference in subarea generates group consensus comfort distribution and target environment parameter interval;Compare real-time environmental data with target interval to identify deviation dominant dimension and output adjustment direction;Establish energy consumption feature model to extract optimal power interval and lowest energy consumption state;Comfort and energy consumption weight are dynamically balanced based on comfort gap and energy consumption pressure to generate control decision;After being converted into equipment control instruction and executed, the model is updated through feedback residual increment to form a closed loop self-learning optimization.The application realizes the reasonable integration of group comfort demand and space regulation refinement.

Description

Technical Field

[0001] This invention belongs to the field of energy consumption regulation technology for building spaces, specifically relating to an adaptive control system and method that integrates spatial characteristics and comfort perception. Background Technology

[0002] The construction industry is a major source of global energy consumption and carbon emissions. Current statistics show that building energy consumption accounts for approximately 40% of global total energy consumption, with office building energy consumption being the main component, accounting for approximately 35% or more.

[0003] Office buildings are not only major energy consumers but also the spaces where people work daily; people spend 40% to 50% of their day working and staying in office buildings. In this context, indoor comfort is directly related to the physical health, work status, and productivity of employees, and at the same time affects the building's energy consumption level.

[0004] In existing research and patented inventions, achieving high-efficiency building energy conservation without compromising indoor comfort has become one of the core issues in the field of green building. Current traditional research mostly adopts static control strategies for intervention, such as setting air conditioning temperature thresholds and implementing time-sharing energy consumption scheduling. These methods are difficult to adapt to dynamic changes in indoor physical conditions and the diverse needs of people, which can easily lead to energy waste and discomfort. At the same time, these studies still have shortcomings in systematically integrating building space physical parameters, human behavioral characteristics, and system adaptive regulation.

[0005] In recent years, with the rapid development of research in artificial intelligence and Internet of Things technologies, it has become more possible and feasible to balance and regulate the relationship between energy consumption control and comfort management through the intelligent upgrading and renovation of buildings.

[0006] Based on the retrieval and analysis of a large number of relevant patents and literature, the current research on building energy consumption and spatial comfort has the following status and shortcomings.

[0007] Existing research frameworks for comfort standards are relatively singular, with thermal comfort dominating. There is a significant lack of attention to comfort aspects such as indoor air quality, acoustic environment, and lighting environment. Related discussions often stop at simple analysis and standard construction of ambient temperature and air conditioning energy consumption, and research on multi-dimensional and multi-standard comfort indicators is severely lacking. At the same time, research on the correlation between human behavior and the building's physical environment is insufficient, and the correlation between relevant parameters and variables of the building and its users is often ignored during the optimization process.

[0008] Existing research largely focuses on single-point optimization for energy consumption prediction or comfort assessment, lacking systematic modeling of the interactions between changes in different spatial structures within buildings, real-time status of personnel flow and distribution, and environmental control systems. In particular, the coordinated regulation between personnel behavior patterns and spatial zoning energy consumption has not yet been studied. Although user-centered control has been proposed for many years, the closed-loop logic chain from real-world personnel behavior data to adaptive control strategies remains incomplete.

[0009] Existing research lacks in-depth understanding of the potential quantitative relationships between factors such as spatial conditions, functional types, geometric dimensions, and physical characteristics of building spaces and energy consumption and comfort. Current studies primarily focus on the impact of building envelope characteristics on energy consumption, failing to provide differentiated comfort-adaptive control strategies and energy self-regulation bases for people in different spaces. Furthermore, an operational, categorized control logic has not yet been established to address the differences in energy consumption and comfort requirements across different climate zones, time periods, and building spaces.

[0010] Existing energy consumption and comfort self-regulation systems are mostly open-loop control systems or closed-loop control systems based on preset comfort thresholds, rather than true real-time regulation. They have not been able to achieve adaptive comfort adjustment based on the instantaneous physical state feedback and physical perception of the population during operation. Although some studies have attempted to incorporate physical state parameters such as skin temperature and skin conductance into modeling, a reasonable method and theory for integrating these diverse and differentiated signals into the system's real-time control decisions in a real-time, efficient, and logical manner has not yet been developed. At the same time, within the building space, the randomness and individual differences in the behavioral patterns of a large number of people further exacerbate the difficulty of judgment, prediction, and control.

[0011] In summary, existing technologies suffer from the following problems: a lack of systematic research and modeling of the relationship between building space, energy consumption, and human comfort; a lack of a unified prediction and control mechanism that integrates multiple comfort indicators, such as thermal comfort, light comfort, air quality comfort, and acoustic comfort; failure to achieve adaptive and coordinated optimization of energy consumption and comfort in dynamic changes in human populations, such as changes in population distribution, spatial movement of people, and shifts in spatial zoning; a lack of a control mechanism based on real-time feedback of human comfort, resulting in the system's control effectiveness lagging behind actual changes in human perception; the inability to directly generalize comfort analysis models for individual comfort levels to groups, and a lack of integration of different or even conflicting comfort preferences into a single effective control logic when a large number of individuals coexist; and the tendency to control large spaces such as entire rooms or floors, ignoring the subtle environmental zoning within the same space caused by differences in orientation, distance from air conditioning, and population density. Summary of the Invention

[0012] To address the aforementioned issues, this invention provides an adaptive control system and method that integrates spatial characteristics and comfort perception. This system can perceive the differences between group behavior patterns and the spatial environment in which the group is located in real time, and dynamically balance energy consumption and group comfort for adaptive adjustment. At the same time, by constructing an adjustment logic of "group comfort + space + environment", it can achieve refined, personalized and energy-saving control of different functional zones inside the building.

[0013] Firstly, this invention proposes an adaptive control method that integrates spatial features and comfort perception, comprising: dividing the building space into multiple zones and constructing spatial feature vectors for each zone; collecting real-time physical state data of people located in each zone, environmental data of each zone, and energy consumption data, and mapping the collected data to the corresponding zones after binding timestamps and spatial coordinates; identifying effective users in each zone based on personnel location and dwell time; constructing an individual comfort prototype library for each user, and calculating an individual comfort score by matching the real-time physical state vector formed by the real-time collected physical state data with the physical state response anchor points in the individual comfort prototype library; fusing the individual comfort scores and individual preference intervals of each effective user in the same zone to form a group consensus comfort distribution for that zone, and determining the group target environmental parameter interval accordingly; and dividing the corresponding zones into zones. The real-time environmental data of the zone is compared with the target environmental parameter range of the group across multiple environmental dimensions to identify the dominant dimension causing the environmental deviation, and the adjustment direction corresponding to the dominant dimension is output accordingly. Based on the operating power, duration, and historical energy consumption of the equipment in each zone, an energy consumption characteristic model is established, and the optimal power range and the lowest energy consumption state are extracted. Based on the current comfort gap and energy consumption pressure, the comfort weight and energy consumption weight are dynamically weighed to calculate the comprehensive decision score for multiple preset candidate control actions to generate control decisions. The control decisions are converted into control commands for each device in the corresponding zone and executed. The body state data, environmental data, and energy consumption data fed back after execution are obtained, the comfort response residual and energy consumption response residual are calculated, and the individual comfort prototype library, the group consensus comfort distribution, the energy consumption characteristic model, and the comfort weight and the energy consumption weight are incrementally updated accordingly to form a closed-loop self-learning optimization.

[0014] In some examples, dividing the building space into multiple zones and constructing a spatial feature vector for each zone includes: obtaining the orientation, spatial function, window-to-wall ratio, air conditioning coverage efficiency, and spatial scale of each spatial unit constituting the building space, and extracting them as features of each spatial unit; performing cluster analysis on the features of all spatial units using a hierarchical clustering method, and grouping spatial units with similar attributes, similar environmental response characteristics, and consistent equipment control characteristics into the same zone; and extracting statistical values ​​of the features of each spatial unit within the merged zone to form a spatial feature vector uniquely corresponding to that zone.

[0015] In some examples, identifying valid users within each partition based on personnel location and dwell time includes: obtaining real-time location information of personnel and determining their current partition based on the boundaries of the partition; continuously recording the dwell time of the personnel in their current partition; and when the dwell time exceeds a preset threshold, marking the personnel as valid users and establishing their affiliation with their current partition.

[0016] In some examples, calculating the individual comfort score includes: collecting historical physical state data of the user when confirming that they are in a comfortable state, and extracting the center value, fluctuation range, and feature weight of multiple physical state features from the data, combining them to form at least one physical state response anchor point, and all of the physical state response anchor points constitute the user's individual comfort prototype library; matching the real-time physical state vector with each physical state response anchor point in the individual comfort prototype library, and calculating the matching score; and converting the highest matching score into the individual comfort score according to a preset mapping function.

[0017] In some examples, the process of fusing the individual comfort scores and individual preference intervals of each valid user within the same partition to form a group consensus comfort distribution for that partition, and determining the target environmental parameter interval for the group accordingly, includes: determining the individual preference interval and individual preference center for each valid user, and calculating individual weights by combining their individual comfort scores, validity status, and the group density of the partition; using a comfort kernel function, mapping the individual preference interval and individual weight of each valid user to a weighted kernel function in the environmental parameter space; superimposing the weighted kernel functions of all valid users within the partition to form the group consensus comfort distribution; finding the peak point of the group consensus comfort distribution, and selecting a continuous interval that satisfies a preset retention coefficient centered on the peak point as the target environmental parameter interval for the group in that partition.

[0018] In some examples, the real-time environmental data of the corresponding partition is compared with the target environmental parameter range of the group across multiple environmental dimensions to identify the dominant dimension causing the environmental deviation, and the adjustment direction corresponding to the dominant dimension is output accordingly. This includes: performing a Cartesian product operation on the range of the target environmental parameter range of the group across each environmental dimension to construct a multidimensional comfort bounding box; when the real-time environmental data exceeds the multidimensional comfort bounding box, calculating the normalized deviation of the excess portion across each environmental dimension; determining the environmental dimension with the largest normalized deviation as the dominant dimension, and outputting the adjustment direction corresponding to the dominant dimension accordingly.

[0019] In some examples, establishing an energy consumption characteristic model and extracting the optimal power range and the lowest energy consumption state includes: extracting the device's running time, number of start-stop cycles, average power, and peak power according to a preset time window to form a working condition feature vector; performing cluster analysis on the working condition feature vector to obtain multiple operating condition clusters; statistically analyzing the energy consumption distribution within each operating condition cluster, extracting the energy consumption envelope boundary and baseline curve, and determining the lowest energy consumption state and the optimal power range accordingly.

[0020] In some examples, the step of dynamically balancing comfort weights and energy consumption weights based on the current comfort gap and energy consumption pressure to calculate a comprehensive decision score for multiple preset candidate control actions to generate a control decision includes: calculating the difference between the comfort level of the target group and the comfort level of the current group as the comfort gap; comprehensively evaluating the current energy consumption status, external electricity price, power rationing conditions, or energy-saving target constraints to generate the energy consumption pressure; calculating a dynamic gating coefficient based on the comfort gap and the energy consumption pressure, and using the dynamic gating coefficient to determine the comfort weight and the energy consumption weight; calculating a comfort benefit score and an energy cost score for each candidate control action, and using the comfort weight and the energy consumption weight to perform a weighted sum to obtain the comprehensive decision score; and selecting the candidate control action with the highest comprehensive decision score as the control decision.

[0021] Secondly, this invention proposes an adaptive control system that integrates spatial features and comfort perception, comprising: a building space partitioning modeling module, configured to divide the building space into multiple partitions and construct spatial feature vectors for each partition; a multi-source data acquisition module, configured to collect in real time the physical state data of personnel located in each partition, environmental data of each partition, and energy consumption data, and map the collected data to the corresponding partitions after binding timestamps and spatial coordinates; a personnel positioning and partition affiliation identification module, configured to identify effective users in each partition based on personnel location and dwell time; an individual comfort modeling module, configured to construct an individual comfort prototype library for each user, and calculate an individual comfort score by matching the real-time physical state vector formed by the real-time collected physical state data with the physical state response anchor points in the individual comfort prototype library; and a group comfort fusion module, configured to fuse the individual comfort scores and individual preference intervals of each effective user in the same partition to form a group consensus comfort distribution for that partition, and determine the group target environmental parameter interval accordingly; and an environment... The status assessment module is configured to compare the real-time environmental data of the corresponding partition with the target environmental parameter range of the group across multiple environmental dimensions to identify the dominant dimension causing the environmental deviation, and output the adjustment direction corresponding to the dominant dimension accordingly. The energy consumption feature modeling module is configured to establish an energy consumption feature model based on the operating power, duration, and historical energy consumption of the equipment in each partition, and extract the optimal power range and the lowest energy consumption state. The comfort-energy consumption collaborative decision-making module is configured to dynamically weigh comfort weights and energy consumption weights based on the current comfort gap and energy consumption pressure, calculate a comprehensive decision score for multiple preset candidate control actions to generate control decisions. The equipment control execution module is configured to convert the control decisions into control commands for each device in the corresponding partition and execute them. The feedback learning and self-optimization module is configured to acquire the body state data, environmental data, and energy consumption data fed back after execution, calculate the comfort response residual and energy consumption response residual, and incrementally update the individual comfort prototype library, the group consensus comfort distribution, the energy consumption feature model, and the comfort weights and energy consumption weights accordingly to form a closed-loop self-learning optimization.

[0022] Thirdly, the present invention proposes a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the aforementioned adaptive control method for integrating spatial features and comfort perception.

[0023] The present invention has achieved the following significant advancements: Granularization of space and perception. Instead of treating the entire floor or building space as the smallest unit, it evolves and divides it into several smaller granular partitioned spaces based on the inherent properties of the building, and treats the characteristic attributes of each partition as a data unit.

[0024] Real-time acquisition and updating of multi-source data. Three types of data are collected simultaneously: environmental data of the zoned space obtained through environmental monitoring equipment, real-time physical status data and spatial location of each person obtained through wearable devices, and real-time energy consumption power of each zone obtained by connecting to the building's smart power system. These three types of data are aligned and updated according to time and spatial coordinates.

[0025] Offline modeling of individual comfort. For long-term workers, a mapping model can be trained using their recorded historical physical condition data and corresponding environmental data. By inputting their physical condition data and corresponding environmental data, the model outputs a real-time comfort score for each individual. This score is a continuous value on a standard axis, representing the degree from discomfort to comfort to overcomfort, and corresponding to the individual's preferred optimal environmental data range. Using this model, comfort can be extrapolated from physical condition data during subsequent online operations, without relying on environmental data.

[0026] Real-time fusion of crowd comfort levels. For any target zone, the system first identifies the current crowd group within that zone, obtains each individual's real-time comfort score through their individual model, and then determines the adjustment direction based on the proportion of the crowd that is comfortable or not, thus avoiding blind adjustments based on absolute averages.

[0027] Dynamic trade-offs between energy consumption and comfort. An adaptive weighting factor is introduced based on the weighting requirements of comfort and energy consumption. This factor can be adjusted in real time or manually set according to external electricity prices, power grid restrictions, or the building's expected energy consumption limit. When external constraints are relaxed, the weighting can favor comfort; when external constraints are tight and electricity consumption is excessive, the weighting shifts towards energy conservation. Adjustments are made by optimizing the power consumption of equipment, including air conditioning set temperatures, fresh air valve openings, and lighting brightness.

[0028] Distributed execution and real-time closed-loop feedback. Based on the assessment results of existing comfort and environmental data in each zone, customized adjustment decision commands are sent individually to the execution terminals of each zone. After the command is executed, environmental and physical state data are collected again for reassessment and decision-making, forming a continuous feedback control loop at the minute or even second level. Since individual comfort can be directly inferred from physical state signals, direct closed-loop feedback from individual perception to environmental adjustment can be achieved, eliminating the need for indirect inference based on environmental parameters and reducing workflow complexity and latency.

[0029] In summary, the system achieves the following effects: the comfort needs of the group are rationally integrated rather than simply averaged; spatial control units are divided into zoned spaces for greater precision; real-time feedback of physical state data replaces predicted preset thresholds or indirect environmental data inferences; and energy consumption and comfort can be prioritized freely according to external needs. Therefore, based on these foundations, simply put, this invention provides a practically applicable adaptive adjustment logic that shifts from a building-, energy-, or environment-centric approach to a group-centric approach, rather than remaining an ideal or slogan on paper. Furthermore, the operational logic of this model system is not tied to a specific building, possessing portability and replicability. Attached Figure Description

[0030] Figure 1 This is a flowchart of architectural space partitioning modeling according to an embodiment of the present invention.

[0031] Figure 2 This is a flowchart of multi-source data acquisition according to an embodiment of the present invention.

[0032] Figure 3 This is a flowchart of personnel location and zoning identification according to an embodiment of the present invention.

[0033] Figure 4 This is a flowchart of individual comfort modeling according to an embodiment of the present invention.

[0034] Figure 5 This is a flowchart of group comfort fusion according to an embodiment of the present invention.

[0035] Figure 6 This is a flowchart of an environmental status assessment according to an embodiment of the present invention.

[0036] Figure 7 This is a flowchart of energy consumption characteristic modeling according to an embodiment of the present invention.

[0037] Figure 8 This is a flowchart of a comfort-energy consumption coordinated decision-making process according to an embodiment of the present invention.

[0038] Figure 9 This is a flowchart of device control execution according to an embodiment of the present invention.

[0039] Figure 10 This is a flowchart of feedback learning and self-optimization according to an embodiment of the present invention.

[0040] Figure 11 This is a block diagram of an adaptive control system that integrates spatial characteristics and comfort perception according to an embodiment of the present invention. Detailed Implementation

[0041] Taking office buildings as an example, they have characteristics such as high population mobility, irregular population distribution, diverse spatial properties, and complex population comfort (state), which can better illustrate the operating logic of the system of the present invention under complex conditions.

[0042] 1. Spatial characteristics.

[0043] In this invention, the processing of spatial features no longer treats the entire floor as a single control object. Instead, based on the inherent attributes of the building, such as orientation, spatial function, window-to-wall ratio, air conditioning coverage efficiency, distance from air conditioning, and spatial scale, the entire building space is divided into several minimum control zones. Each zone is associated with a set of static inherent attribute data, which serves as the inherent spatial features for subsequent analysis.

[0044] 2. Individual comfort characteristics and scores.

[0045] Regarding crowd comfort, within each spatial zone, individuals will have different comfort standards and scores due to the inherent differences in the zone's attributes. The optimal score is set as the origin of the score. This involves collecting body state data through wearable devices (skin conductance signal acquisition systems). Each individual's optimal comfort corresponds to unique optimal body state data: body temperature, blood pressure, skin conductance, and fatigue level, as well as unique optimal environmental data: temperature, humidity, air pressure, carbon dioxide concentration, illuminance, and noise. These two types of data are only associated with the individual, space, and time period. Through timed and fixed-point data measurement and recording, a complete data raster covering all weather conditions, all seasons, and all climate periods can be formed for continuous self-optimization.

[0046] 3. Characteristics of crowd comfort.

[0047] Within each zone, the wristband's positioning and receiving functions can be used to locate the wearer's position and the space they are in in real time, identifying the group of people currently in that zone. Users who stay in each zone for more than 5 minutes are included as valid users. The optimal comfort standard for each valid user in that zone is statistically summarized in real time, and the optimal group comfort in that zone is obtained using an asymmetric Pareto optimization rule. This comfort level corresponds to real-time and unique optimal group physical state data and optimal group environmental data, and this is used as the optimal standard. At the same time, the real-time individual comfort data of each valid user in the space is statistically summarized based on the wristband's real-time measurements to obtain real-time group physical state data, and real-time group environmental data is measured using environmental monitoring equipment.

[0048] Real-time group physical condition data and real-time group environmental data are compared with the optimal group physical condition data and optimal group environmental data for that partition to determine group comfort and environmental status, and to determine whether spatial environmental control equipment (air conditioning, fresh air system, ventilation ducts, etc.) needs to be adjusted and in what direction. For example, priority is given to identifying the proportion of people with comfort scores below a threshold (significant discomfort); if this proportion exceeds a set threshold, priority control is triggered. Otherwise, when most people are in a neutral or comfortable range, a target environmental parameter range preferred by the group is output, such as the acceptable range of temperature and humidity.

[0049] 4. Energy consumption characteristics.

[0050] Based on historical data of equipment power consumption and personnel office usage in each space, the general power consumption curves of each space within the building can be analyzed by using the real-time power (kWh) and cumulative energy consumption of the air conditioning (HVAC), lighting, and fresh air systems in each zone, and the corresponding unique optimal equipment power consumption and minimum energy consumption can be obtained.

[0051] 5. Balanced regulation.

[0052] In the adjustment of step 3, the energy consumption curve of step 4 can be introduced. When determining whether to adjust and the direction of adjustment, the power consumption and running time of the equipment are adjusted and the start and stop are controlled based on the unique optimal power consumption and minimum energy consumption of the equipment corresponding to each zone space.

[0053] The optimal comfort level and its corresponding data for each individual can be correlated with the spatial data of each zone. By recording and analyzing data over multiple time periods, the correlation between the inherent attribute data of the spatial zone and the comfort data can be established. This enables the prediction of comfort-related data by inputting the inherent attribute data of the zone, thus forming a predictive forecast and a complete system.

[0054] 1. The present invention needs to record the following relevant data and content.

[0055] (1) Inherent attribute data of the partition space: orientation, space function, window-to-wall ratio (%), air conditioning coverage efficiency (%), distance from workstation to air conditioner (m), space scale (m³).

[0056] (2) Comfort characteristics judgment - corresponding physical condition data and environmental data: physical condition data: body temperature (°C), blood pressure (kPa), skin conductance (µS), fatigue (%) (this type of data directly determines the comfort rating system and judges the status of personnel); environmental data: temperature (°C), humidity (%), air pressure (mPa), carbon dioxide concentration (%), illuminance (lux), noise (dB).

[0057] (3) Building energy consumption data: equipment power (KW), historical energy consumption (kWh), real-time energy consumption (kWh).

[0058] 2. Recording methods and formats.

[0059] (1) The inherent attribute data of the partition space is obtained by measuring and designing the space. It can be measured using tools such as laser rangefinders and anemometers, and modeled with spatial modeling software. It belongs to the inherent attributes that will not change or change, and does not require multiple measurements; only a single measurement is needed.

[0060] (2) Body status data needs to be monitored and measured by wearing a leather electric wristband. The wristband can complete the relevant calculation of body status data and achieve real-time spatial positioning in conjunction with a signal receiver. It can be further displayed in the spatial model to show the flow of people and individuals in real time.

[0061] The measurement and recording of physical condition data are divided into the following: a. Optimal physical condition data under the best comfort state in each zone space. Through the experimental phase, the activity state of each individual in each zone space is experimentally measured. This measurement can be carried out and recorded multiple times according to different time periods and climate seasons; b. Real-time physical condition data of each individual in the zone space is recorded and analyzed in real time through the wristband and uploaded to the system database.

[0062] (3) Environmental data can be monitored and measured using environmental monitoring instruments, enabling real-time data recording and uploading, and real-time monitoring of the environment in each zone.

[0063] The environmental data measurement and recording are divided into the following: a. The acquisition of the optimal environmental data of the partition space corresponds to the experimental stage in (2)a and is carried out simultaneously; b. The real-time environmental data of each partition space is recorded and analyzed in real time through the environmental monitoring instrument and uploaded to the system database.

[0064] (4) Building energy consumption data can be queried and analyzed through the smart power system of the building or the power grid.

[0065] Example 1: An adaptive control method that integrates spatial characteristics and comfort perception.

[0066] This embodiment of the adaptive control method integrating spatial features and comfort perception includes: dividing the building space into multiple zones and constructing spatial feature vectors for each zone; collecting real-time physical state data of people in each zone, environmental data of each zone, and energy consumption data, and mapping the collected data to the corresponding zones after binding timestamps and spatial coordinates; identifying effective users in each zone based on personnel location and dwell time; constructing an individual comfort prototype library for each user, and calculating an individual comfort score by matching the real-time physical state vector formed by the real-time collected physical state data with the physical state response anchor points in the individual comfort prototype library; fusing the individual comfort scores and individual preference intervals of each effective user in the same zone to form a group consensus comfort distribution for that zone, and determining the group target environmental parameter interval accordingly; and converting the real-time physical state vector of each effective user in the same zone into a spatial feature vector for each zone. The method compares real-time environmental data with the target environmental parameter range of the group across multiple environmental dimensions to identify the dominant dimension causing environmental deviation, and outputs the corresponding adjustment direction for that dominant dimension. Based on the operating power, duration, and historical energy consumption of equipment within each partition, an energy consumption characteristic model is established, and the optimal power range and lowest energy consumption state are extracted. Based on the current comfort gap and energy consumption pressure, comfort weights and energy consumption weights are dynamically balanced to calculate a comprehensive decision score for multiple preset candidate control actions to generate control decisions. The control decisions are then converted into control commands for each device within the corresponding partition and executed. Post-execution feedback data on body state, environment, and energy consumption are obtained, and comfort response residuals and energy consumption response residuals are calculated. Based on these, the individual comfort prototype library, the group consensus comfort distribution, the energy consumption characteristic model, and the comfort weights and energy consumption weights are incrementally updated to form a closed-loop self-learning optimization. The following details this method.

[0067] Step 1: Model the spatial partitioning of the building.

[0068] The entire building or floor is divided into several minimum controllable spatial units, which serve as the spatial basis for subsequent comfort assessment, group preference integration, and energy consumption optimization control. The inherent attributes of the building space are collected, including but not limited to spatial orientation, function type, window-to-wall ratio, air conditioning coverage efficiency, distance from air conditioning units, and spatial scale parameters. These attributes are then uniformly encoded to construct corresponding spatial feature vectors.

[0069] A hierarchical clustering partitioning method based on architectural spatial feature vectors is adopted to perform cluster analysis on the feature similarity of each spatial unit. Spatial units with similar spatial attributes, similar environmental response characteristics, and consistent equipment control characteristics are grouped into the same partition, thus forming multiple partitions with high control consistency. Furthermore, a unique identifier (ID) is assigned to each partition, and a corresponding static attribute file is established. This file records the spatial characteristics, boundary information, and basic control attributes of the partition. Environmental response characteristics refer to the variation patterns and responsiveness of physical environmental parameters such as indoor temperature, carbon dioxide concentration, air pressure, and brightness of each spatial unit under the same external climatic conditions or internal control effects. To a certain extent, this reflects the sensitivity, reaction speed, and reaction patterns of each spatial unit to changes in environmental parameters.

[0070] The specific processing procedure is as follows: Figure 1 As shown, the process includes: first, acquiring architectural design drawings, BIM / spatial modeling data, and on-site measurement data; then, extracting features such as orientation, function type, window-to-wall ratio, air conditioning coverage efficiency, distance from air conditioning units, and spatial scale from the data, and standardizing them to generate spatial feature vectors; subsequently, performing hierarchical clustering operations on the spatial feature vectors to obtain several spatial clustering results; next, determining the spatial boundaries of each partition based on the clustering results; then, establishing a unique number and attribute record for each partition; finally, outputting a partition spatial attribute library, which provides basic data support for subsequent comfort modeling and energy consumption control.

[0071] Hierarchical clustering groups areas with similar spatial attributes, thermal environment responses, and equipment control characteristics into the same partition, which can effectively reduce the differences within the partition, improve the consistency and stability of partition control, and further enhance the accuracy of subsequent comfort control and energy consumption control.

[0072] Step 2: Multi-source data acquisition.

[0073] The system collects three core types of data related to building zoning in real time: personnel physical condition data, environmental data, and energy consumption data, providing a unified data foundation for subsequent comfort assessments and energy consumption optimization. Specifically, it acquires data related to the operational status of each zoning area from wearable wristbands, environmental sensors, and the building energy management system, and manages data from different sources, frequencies, and spatial locations in a unified manner.

[0074] A multi-source data acquisition method based on a unified spatiotemporal index is adopted. By binding a timestamp and spatial coordinates to each piece of acquired data, personnel physical status data, environmental data, and energy consumption data are uniformly mapped to corresponding partitions, realizing integrated acquisition and association of multi-source data. The unified spatiotemporal index is used to characterize the temporal and spatial location of data generation, thereby enabling various types of data to form corresponding relationships within the same partition and at the same time.

[0075] The specific processing procedure is as follows: Figure 2 As shown, the process includes: first, collecting physical status data of individuals from wearable wristbands, including one or more of body temperature, blood pressure, skin conductance, and fatigue level; second, collecting one or more environmental data from environmental sensors, including temperature, humidity, air pressure, carbon dioxide concentration, and light intensity; third, collecting equipment power, operating status, and energy consumption data from the building energy management system; subsequently, attaching a unified timestamp and spatial coordinate information to all types of data; next, mapping the data to corresponding spatial units according to building zoning information; and finally forming real-time physical status data streams, real-time environmental data streams, and real-time energy consumption data streams for subsequent comfort assessment and energy consumption control.

[0076] By using a unified spatiotemporal index, data from different sources can be aligned in the same partition and at the same time, avoiding correlation errors caused by inconsistent sampling frequencies, scattered data sources, or unclear spatial locations, thereby improving the accuracy and stability of subsequent comfort modeling and energy consumption decisions.

[0077] Step 3: Personnel location and zone affiliation identification.

[0078] The system determines the zone where the user wearing the wristband is located and identifies the valid users within the zone, providing a data foundation for subsequent group comfort assessment and zone control; it receives the location information of the wristband wearer, building zone boundary information, and dwell time data, identifies and counts the space where people are located, and forms the population distribution in each zone.

[0079] A zone-based location identification method based on dwell time constraints is adopted to continuously distinguish the location trajectory of the bracelet wearer: when the dwell time at a certain location exceeds a set threshold, the person is identified as a valid user, and their zone is determined, establishing a correspondence between the person and the zone. Dwell time, as a judgment condition, aims to distinguish between people who pass through briefly and people who actually stay in the space; through this process, the interference caused by short-term movement can be reduced, making the identification results of people within the zone closer to real-world usage scenarios.

[0080] like Figure 3 As shown, the processing flow is as follows: obtain the real-time location information of the wristband wearer and synchronously read the building zoning boundary data; determine the zoning of the person based on the location coordinates and continuously record the length of time the person stays in the zoning; when the length of stay exceeds the set threshold, mark the person as a valid user and establish their zoning affiliation; summarize the people in each zoning to form a zoning population set and population density, and use the relevant results for subsequent comfort assessment.

[0081] Based on this method, by introducing dwell time as a criterion, individuals who pass through briefly, stay temporarily, or only traverse a route can be excluded, thereby reducing the interference of such individuals on the data. The processed zoning personnel information is more in line with actual usage conditions, which is beneficial for maintaining stability in subsequent comfort assessments and environmental control, and improves the accuracy of zoning control results.

[0082] Step 4: Individual comfort modeling.

[0083] Figure 4 This paper demonstrates the process of individual comfort modeling. This step aims to characterize the individual comfort state of users under different spatial, temporal, and environmental conditions, providing a basis for subsequent personalized environmental control. The processing method no longer relies on fixed environmental parameter thresholds for comfort judgment, but instead estimates the current comfort level directly based on human body state response data. Then, it infers the corresponding environmental parameter range based on changes in body state, allowing the results to adjust automatically according to individual differences. This method can construct comfort models that more closely reflect individual experiences, providing more specific data support for subsequent environmental control.

[0084] Collect physical condition data such as body temperature, blood pressure, skin conductance, and fatigue level for each user, as well as environmental data such as temperature, humidity, air pressure, carbon dioxide concentration, light intensity, and noise. Combine this with the user's initial comfort confirmation information to build an individual comfort prototype library.

[0085] This step constructs an individual comfort prototype mapping method based on body state response anchors. Body state response anchors refer to a set of relatively stable body state signal features formed when a user confirms they are in a comfortable state, used to characterize the typical response pattern of the human body in a comfortable state. These anchors are not single body state values, but rather composed of multiple body state features, encompassing the central level, fluctuation range, and weighting of each feature within the overall representation. Using this combined representation, the body state performance in a comfortable state can be presented more completely.

[0086] In its implementation, the system selects several typical environmental scenarios in the initial stage and sends comfort confirmation requests to the user. The user confirms their current state one or a few times, and this confirmation is presented as at least one of the status labels: "comfortable," "neutral," or "uncomfortable." The system uses the physical state data and environmental data corresponding to the time period when the user confirms being in a comfortable state as the initial comfort sample.

[0087] Suppose that the body state vector collected at a certain moment is:

[0088] in, to These correspond to one or more data points from body temperature, blood pressure, skin conductance, and fatigue level. After screening historical comfort samples, the system calculates the central value, fluctuation range, and feature weight for each body state feature in a comfortable state. The central value characterizes the stability level of the body state feature in a comfortable state; the fluctuation range reflects the allowable normal range of variation in a comfortable state; and the feature weight reflects the relative importance of the body state feature to the comfort level assessment.

[0089] This leads to the formation of individual physical state response anchor points:

[0090] in: For the first The central value of each physical condition characteristic in a comfortable state =1,2,…,n; This corresponds to the fluctuation range; These are the corresponding feature weights.

[0091] When a user exists in multiple stable and comfortable states under different environmental conditions, the system can generate multiple bodily state response anchor points and collectively form an individual comfort prototype library:

[0092] The aforementioned individual comfort prototype library is used to describe the various comfort response patterns of the same user in different spatial environments.

[0093] During system operation, this step collects current body state data in real time and converts it into a real-time body state vector; it then matches this real-time body state vector with various body response anchor points in the individual comfort prototype library and calculates the corresponding matching score. For any body state response anchor point, its matching score is defined as:

[0094] in: For the current state and the first Matching score between each comfort anchor point; These are real-time physical state characteristic values; This corresponds to the center value of the comfort anchor point; This corresponds to the fluctuation range; For the corresponding feature weights; To prevent extremely small constants with a denominator of zero, the higher the matching score, the closer the current physical state is to the user's historical comfort state.

[0095] The system uses the comfort anchor point with the highest matching score as the current comfort reference state, and derives the optimal environmental parameter range for the current user based on the environmental parameter range corresponding to the anchor point; it calculates the individual comfort score based on the matching score and constructs the corresponding individual comfort preference model.

[0096] The system converts the current matching score into an individual comfort score based on a preset mapping function. This mapping function is a monotonically increasing function, ensuring that a higher matching score leads to an individual comfort score closer to the upper limit representing complete comfort. In an optional implementation, the mapping function is a linear mapping, i.e., individual comfort score = current matching score × preset scaling factor.

[0097] Based on the individual comfort prototype library and its real-time matching results, the system constructs an individual comfort preference model corresponding to the user. This model includes at least the user's individual comfort prototype library, historical matching statistics for each anchor point, and mapping function parameters from the matching score to the comfort score. During operation, the system continuously collects the user's real-time physical state data and environmental data. When preset anchor point update trigger conditions are met, the system incrementally updates the physical state response anchor points in the individual comfort prototype library to achieve adaptive iteration of the individual comfort preference model.

[0098] The anchor point update triggering conditions include at least one of the following: (1) receiving a comfort confirmation command actively issued by the user; (2) during the process of the system actively adjusting environmental parameters to explore different environmental states, monitoring that the matching score between the current body state vector and a specific body state response anchor point exceeds the historical average matching score of the anchor point within a consecutive preset number of sampling periods. In response to satisfying any of the above triggering conditions, the system uses body state data and environmental data within a preset time period before and after the triggering time to recalculate the center value, fluctuation range, and feature weight of each body state feature in the anchor point, and completes the anchor point update.

[0099] The human body's response to environmental changes often precedes subjective perception. Using the method described above, there's no longer a need for long-term subjective feedback through questionnaires or other means of data input into the system. Instead, it dynamically infers the current comfort state based on the degree of matching between real-time physical status and historical comfort prototypes. Compared to traditional methods based on fixed air conditioning temperature thresholds, this method can adapt to individual differences among users and improve the real-time performance, stability, and personalization of comfort assessment.

[0100] Step 5: Group comfort integration.

[0101] This step integrates the comfort preferences of multiple individuals within a partition into a unified control target at the group level, providing a basis for group comfort in subsequent partition control. Specifically, this step summarizes the individual comfort scores, individual preference ranges, and group size and density of all valid users within the partition, forming a foundation for group comfort integration, and then outputting environmental parameter targets that reflect the overall preferences of the partition.

[0102] Figure 5 The process of group comfort fusion is demonstrated. In its implementation, a group comfort fusion method based on comfort consensus kernel density is adopted. This method does not simply average individual preferences, but transforms the individual preference interval of each effective user into a comfort kernel function. At the group level, all comfort kernel functions are density-superimposed to identify the environmental parameter interval with the highest group comfort consensus. Here, the comfort kernel function is used to characterize the strength of an individual's comfort contribution to the environmental parameter within its preference interval, and the group consensus comfort distribution is used to characterize the degree of clustering of individual preferences in the environmental parameter space within the partition.

[0103] The individual preference interval refers to the range of optimal environmental parameters output by the individual comfort preference model in step 4 for each valid user, enabling them to reach or approach a comfortable state. The lower and upper limits of this interval are denoted as follows: and .

[0104] Let the partition be the first i Individual comfort ratings from valid users Its preference range is The corresponding individual preference centers are:

[0105] The corresponding interval half-width is:

[0106] in, To prevent extremely small constants with a denominator of zero, individual weights are further determined based on individual comfort scores, effectiveness status, and group density. The individual weight is used to characterize the degree of influence of the user on the group's control objectives. Preferably, it can be determined using the following normalization method:

[0107] in, This indicates a valid user flag, which is 1 if the user meets the valid user criteria, and 0 otherwise. This is for zoning group density. This weighting method allows users who are highly comfortable, have effective dwell time, and are in a reasonable density to contribute more to the group's objectives.

[0108] In the environmental parameter space, the first The preference range of each user is mapped to the comfort kernel function. ,in This represents an environmental parameter vector, which includes one or more of temperature, humidity, air pressure, carbon dioxide concentration, light intensity, and noise. Preferably, the comfort kernel function can be expressed as:

[0109] in, This represents the distance between the current environmental parameter vector and the individual's preference center. This comfort kernel function (also known as the weighted kernel function) has a higher value near the individual's preference center and decays smoothly as the degree of deviation increases, used to describe the individual's comfort support for the environmental parameters.

[0110] Subsequently, the comfort kernel functions of all valid users within the partition are weighted and superimposed to form a group consensus comfort distribution:

[0111] Where N represents the number of effective users. The group consensus comfort distribution reflects the degree of co-aggregation of preferences among multiple individuals within the partition in the environmental parameter space. The system searches for the peak point in the group consensus comfort distribution. And taking the peak point as the center, take the condition that satisfies

[0112] The continuous interval is used as the range of target environmental parameters for the group, where This is the preset peak value retention coefficient. The target environmental parameter range for this group is the unified control target for the current partition.

[0113] Simultaneously, the system calculates the proportions of people in comfortable, neutral, and uncomfortable states to characterize the comfort state distribution within the current partition. Specifically, an individual's comfort score is classified as comfortable if it exceeds a first threshold; neutral if it falls between the first and second thresholds; and uncomfortable if it falls below the second threshold. By statistically analyzing the proportions of individuals in each of these three states, the group comfort structure of the partition can be obtained.

[0114] Furthermore, the system calculates a group conflict level index based on the dispersion of the group consensus comfort distribution and the degree of preference splitting. Preferably, the group conflict level index can be expressed as:

[0115] in, Indicates the degree of dispersion of individual preference centers. Indicates the degree of preference split. and These are the weighting coefficients. It can be represented by the ratio of peak height to total weight, that is:

[0116] The higher the value of this indicator, the more obvious the conflict of individual preferences within the partition, which also means that the consistency of group comfort is weaker.

[0117] Finally, this step outputs three indicators: a group comfort score, a range of group target environmental parameters, and a group conflict level index. The group comfort score represents the overall comfort level of the group, and this score is based on the aforementioned group consensus comfort distribution. A scalar value is calculated; specifically, it can be based on the peak consensus strength of the distribution. Obtained by normalization.

[0118] Using the above method, group comfort will no longer be equivalent to the simple arithmetic average of individual comfort in previous studies, but will instead be defined by a high-density overlapping region formed by the superposition of multiple preferences. By employing the comfort consensus kernel density method, we can identify environmental parameter ranges with high consistency within the group while preserving individual preference differences, reducing the interference of extreme individuals on the results, thereby more accurately handling individual preference conflicts and improving the stability and rationality of zonal control objectives.

[0119] Step 6: Environmental status assessment.

[0120] This step is used to determine whether the environment of each zone can meet the comfort needs of the group and to provide an environmental status basis for subsequent adjustment and control. This step standardizes the real-time environmental parameters and compares them with the target environmental range of individuals or groups to identify whether the current zone is in an unfavorable state such as being too cold, too hot, too stuffy, too dark, or too noisy. At the same time, based on the comparison results, it outputs the degree of environmental deviation and the direction of adjustment.

[0121] During implementation, an environmental state identification method based on a multidimensional comfort bounding box and a deviation-dominant dimension is adopted. The multidimensional comfort bounding box refers to the set of multidimensional closed intervals formed by setting acceptable ranges for targets in multiple environmental dimensions, which is used to characterize the environmental boundary of the partition that meets the comfort needs of the group. The deviation-dominant dimension refers to the dimension with the largest normalized deviation among all environmental dimensions when the real-time environmental data exceeds the multidimensional comfort bounding box. This dimension corresponds to the main source of discomfort in the current environment.

[0122] Figure 6The process of environmental status assessment is illustrated. The system first acquires real-time environmental data, including one or more of temperature, humidity, air pressure, carbon dioxide concentration, light intensity, and noise; simultaneously, it acquires the target environmental range for individuals and groups. The target environmental range can be represented as the comfort zone corresponding to each environmental dimension, i.e.:

[0123] in, and They represent the first The lower and upper bounds of each environmental dimension (d=0,1,…) n ), n This indicates the number of environmental dimensions. When constructing a multidimensional comfort bounding box based on the above target environmental range, the comfort intervals of each dimension are directly combined using a Cartesian product to form a multidimensional acceptable domain:

[0124] Here, B represents the multidimensional comfort bounding box, which is used to determine whether the current real-time environmental point is within the acceptable range for the group.

[0125] Before making a judgment, the system standardizes real-time environmental data to compare environmental parameters of different dimensions on a uniform scale. After standardization... d The values ​​for each environmental dimension are denoted as follows: The system determines the real-time environment vector. Does it fall within the multidimensional comfort enclosure box B? If all dimensions are satisfied... If the current environment vector is within the bounding box, the comfort state label is output; if the real-time environment vector exceeds the bounding box, the normalized deviation of each environment dimension is calculated.

[0126] For the d Each environmental dimension has a deviation that can be defined as the distance of the current value relative to the target interval, specifically including the following two cases: when When this occurs, it indicates that the dimension is too low, and the deviation is expressed as:

[0127] when When this occurs, it indicates that the dimension is too high, and the deviation is expressed as:

[0128] when When this occurs, the deviation in that dimension is recorded as 0.

[0129] in, To prevent division by zero of extremely small constants, the system can obtain the normalized bias set for each environmental dimension through the above method.

[0130] Subsequently, the system selects the dimension with the largest deviation as the dominant deviation dimension:

[0131] It then outputs the current environmental status label based on the dominant dimension. For example, when the dominant dimension is temperature and... When the dominant dimension is temperature, the output is "slightly cold"; when the dominant dimension is temperature and When the dominant dimension is carbon dioxide concentration and the deviation is too large, the output is "too hot"; when the dominant dimension is light and the deviation is too large, the output is "too dark"; when the dominant dimension is noise and the deviation is too large, the output is "too high noise".

[0132] Simultaneously, the system outputs a total deviation, which characterizes the overall degree to which the current environment deviates from the group's comfort target. Preferably, the total deviation can be represented by a weighted sum of the deviations of each dimension:

[0133] in, For the first The weighting coefficients for each dimension are used to reflect the degree of influence of that dimension on the comfort state. The weighting coefficients can be preset according to the group comfort model.

[0134] Regarding the output of the adjustment direction, the system generates the control direction based on the dominant dimension of the deviation and its sign. Specifically, when the dominant dimension is lower than the target lower limit, the system outputs the "increase direction" corresponding to that dimension; when the dominant dimension is higher than the target upper limit, the system outputs the "decrease direction" corresponding to that dimension. For example: when the temperature is too low, the adjustment direction is to increase the set temperature or decrease the supply air cooling capacity; when the temperature is too high, the adjustment direction is to decrease the set temperature or increase the supply air cooling capacity; when the carbon dioxide concentration is too high, the adjustment direction is to increase the fresh air volume; when the light intensity is too low, the adjustment direction is to increase the lighting output; when the noise level is too high, the adjustment direction is to reduce the noise source output or enhance sound insulation control. The adjustment direction is used to output clear execution instructions to subsequent control modules.

[0135] Environmental comfort is not determined by a single temperature, but is the result of the combined effects of multiple environmental dimensions. A multidimensional comfort bounding box can provide a unified description of the target environmental range for a group, avoiding the simplification of complex environmental states to a single threshold judgment. The deviation-dominant dimension method can accurately identify the most significant sources of discomfort in the current zone and output the corresponding adjustment direction, thereby improving the accuracy of environmental state judgment and the targetedness and stability of subsequent adjustment and control.

[0136] Step 7: Energy consumption characteristic modeling.

[0137] This step establishes the correlation between the power, operating time, historical energy consumption, and comfort effects of equipment in each zone, providing a quantitative basis for subsequent trade-offs between comfort and energy consumption. Specifically, it statistically analyzes the real-time power, historical energy consumption, and operating status of equipment such as HVAC, lighting, and fresh air systems, extracts general power consumption curves for each zone, identifies the lowest energy consumption state and the optimal operating range, and finally outputs an energy consumption benchmark model.

[0138] In implementation, an energy consumption envelope baseline modeling method based on operating condition clusters is adopted. The "operating condition cluster" refers to several operating mode categories formed based on the similarity of equipment operating characteristics within a certain time window, used to characterize the typical operating behavior of similar equipment under different control states. The "energy consumption envelope baseline" refers to the energy consumption distribution boundary and its baseline curve obtained statistically from each operating condition cluster, used to characterize the lowest stable energy consumption level and reasonable power range of the partition under different operating modes. This method combines real-time power, runtime, and state sequences of the equipment into operating condition feature vectors, clusters different equipment operating modes within the partition to form several operating condition clusters, and then extracts the envelope boundary and baseline curve from the energy consumption distribution in each operating condition cluster, thereby establishing a partition energy consumption characteristic model.

[0139] like Figure 7 As shown, the specific processing steps include: First, acquiring the real-time power, historical energy consumption, and operating status of HVAC, lighting, and fresh air equipment; second, extracting characteristic parameters such as equipment runtime, start / stop count, average power, and peak power according to time windows; then combining these characteristic parameters into a working condition feature vector to uniformly represent the equipment's operating behavior within that time window; next, clustering the working condition feature vector to form operating condition clusters, where samples within the same cluster have similar operating states and energy consumption performance; subsequently, statistically analyzing the energy consumption distribution within each cluster, extracting the median energy consumption, lower envelope boundary, and upper envelope boundary to form the energy consumption baseline and fluctuation range for the corresponding operating condition; further, determining the minimum energy consumption state based on the low-energy-consumption and stable operating condition clusters, where the minimum energy consumption state refers to the equipment operating state with low energy consumption and small fluctuations while meeting basic operating requirements; then generating the optimal power range based on the energy consumption envelope boundary, where the optimal power range refers to the range of equipment power values ​​that balances stability and economy under the corresponding operating condition; finally, outputting the partitioned energy consumption curve, the optimal power range, and the minimum energy consumption reference value.

[0140] Similar equipment exhibits a stable power-time-energy consumption relationship under different operating conditions. By employing operating condition clusters, discrete energy consumption data can be summarized into several interpretable operating modes, avoiding fluctuations caused by simple statistical analysis of the original energy consumption sequence. Furthermore, by extracting the minimum energy consumption reference state and optimal power range through the energy consumption envelope baseline, a stable and quantifiable basis can be provided for subsequent trade-offs between comfort and energy consumption, thereby improving the accuracy of zonal energy consumption analysis and the feasibility of control decisions.

[0141] Step 8: Comfort-Energy Consumption Co-decision Decision.

[0142] This step is used to dynamically balance comfort and energy conservation, integrating group comfort status, environmental status, and energy consumption status to generate specific control strategies for zoned control. Specifically, this step receives group comfort level, environmental deviation, energy consumption status, and external constraint information, including electricity price, power curtailment conditions, and energy conservation targets, and outputs control decisions, weight parameters, and execution priorities accordingly.

[0143] This step employs a dynamic weighted collaborative decision-making method based on joint gating of comfort gap and energy consumption pressure. The "comfort gap" refers to the degree of deviation of the current group's comfort state from the target comfort state, representing the difference between the current environment and the group's desired comfort state. The "energy consumption pressure" refers to the urgency of the current energy consumption state relative to the energy-saving target, representing the energy-saving pressure of the current system under constraints such as electricity prices, power rationing, or energy-saving targets. The "gating coefficient" refers to an adjustment parameter dynamically generated based on the comfort gap and energy consumption pressure, used to continuously switch between comfort priority and energy-saving priority, rather than using fixed, hard-coded control logic.

[0144] like Figure 8 As shown, the specific processing steps include: first, acquiring group comfort, environmental deviation, energy consumption status, and external constraints; second, converting group comfort into a comfort gap, which characterizes the degree of deviation of the current state from the group's target comfort state; then, converting energy consumption status and external constraints into energy consumption pressure, which characterizes the urgency of energy conservation in the current state; subsequently, calculating dynamic gating coefficients based on comfort gaps and energy consumption pressures; then generating comfort weights and energy consumption weights based on the dynamic gating coefficients; next, constructing a control candidate set, which includes one or more of the following: air conditioning setpoint, fresh air volume, lighting output, fan start / stop, or other zone control actions; then calculating the comfort benefit score and energy consumption cost score for each candidate control action; then, weighting the comfort benefit score and energy consumption cost score according to dynamic weights to form a comprehensive decision score; finally, selecting the control action with the highest comprehensive decision score as the control decision, and outputting the corresponding weight parameters and execution priority.

[0145] The comfort gap, energy consumption pressure, dynamic gating coefficient, and comprehensive decision score can be represented as follows: The comfort gap can be represented as:

[0146] in, For comfort gap, For the comfort of the target group, This represents the current group comfort level. This expression reflects the degree to which the current group comfort level is insufficient compared to the target state.

[0147] Energy consumption pressure can be expressed as:

[0148] in, Due to energy consumption pressure, This represents the current energy consumption status. As the current electricity price factor, As a power rationing or energy-saving constraint factor, These are the corresponding weighting coefficients. This expression reflects the urgency of the current system's energy consumption.

[0149] The dynamic gating coefficient can be expressed as:

[0150] in, For dynamic gating coefficients, To prevent extremely small constants from being divided by zero, the dynamic gating coefficient is used to control the allocation ratio of comfort weights and energy consumption weights. When the comfort gap is large, With increased power, the system prioritizes comfort; however, when energy consumption is high... The system is being reduced in size, prioritizing energy conservation.

[0151] Comfort weight and energy consumption weight can be expressed as follows:

[0152] in, For comfort weight, Energy consumption weight.

[0153] For any candidate control action Its comprehensive decision score can be expressed as:

[0154] in, For action The contribution score to comfort benefits. For action The contribution score to energy consumption cost; the system selects from all candidate control actions. The largest action serves as the final control decision. This decision-making method can dynamically balance comfort benefits and energy costs, and the weights can change in real time with the current environment and constraints.

[0155] In terms of output, this step provides three types of parameters and instructions: control decisions, weight parameters, and execution priorities. Control decisions are used to instruct subsequent steps on the specific adjustments to equipment such as air conditioning, fresh air, lighting, or fans. Weight parameters reflect the degree of emphasis between comfort priority and energy saving priority at the current moment. Execution priorities specify the order and response level of each control action.

[0156] The advantage of this method is that comfort and energy saving are not fixed opposites that can be mutually exclusive, but rather dynamically adjusted according to changes in personnel distribution, external electricity prices, power rationing conditions, and target strategies. By implementing joint gating of comfort gaps and energy consumption pressures, the weights can be continuously and dynamically adjusted according to the environment and constraints, which can avoid insufficient comfort and constrain energy consumption levels, thereby improving the adaptability, interpretability, and enforceability of control decisions.

[0157] Step 9: Equipment control execution.

[0158] This step transforms the control strategy generated in step 8 (comfort-energy consumption collaborative decision-making) into actual equipment control commands and sends them to the corresponding zone terminal devices to achieve automatic adjustment and execution of the building environment. Unlike unified broadcast control, this step performs independent control based on building zones, enabling different zones to execute different control actions and supporting distributed device collaborative execution, thereby improving the precision and efficiency of control.

[0159] This step receives the control commands output from step 8 and the current status of the zoned devices, and generates corresponding execution commands for devices such as air conditioners, fresh air systems, exhaust systems, lighting, fans, and valves. These control commands are used to adjust parameters such as the set temperature, airflow, valve opening, lighting brightness, and start / stop time of the devices. Simultaneously, it receives execution result feedback from each device to complete the control loop.

[0160] In implementation, this step employs a distributed instruction execution method based on a partition control mapping table. The "partition control mapping table" is a data mapping structure that associates partition IDs, device IDs, device types, adjustable parameters, and execution permissions, used to determine which devices within each partition need to perform which control actions. "Distributed instruction execution" means that each partition device independently receives and executes control commands based on its local mapping relationship, rather than having a central controller synchronously execute commands for all devices. Through this method, a one-to-one mapping is established between control actions and specific devices, allowing different partitions to independently execute different control strategies without blocking each other.

[0161] like Figure 9 As shown, the specific processing steps include: First, acquiring control decisions, which at least include target temperature, fresh air volume, lighting brightness, fan status, valve opening degree, and execution priority; second, acquiring the status of zoned devices, which includes current operating status, online status, current power, current setpoint, and executable permissions, wherein executable permissions are used to determine whether the target device is allowed to execute the control action; then, constructing a zoned control mapping table, whereby the system establishes a corresponding relationship based on zone ID, device ID, device type, adjustable parameters, and control permissions.

[0162] in Indicates the partition number. Indicates the equipment number. Indicates control parameters, This represents the set of permitted actions. The mapping table is used to determine which device in which partition can perform which control actions.

[0163] When generating device execution commands, the system converts control objectives into device-level execution commands based on the control decision and partition control mapping table. Preferably, the device action mapping function can be expressed as:

[0164] in, Indicates the target device The execution command, Indicates the control objective. This represents the device status, and P represents the zoning strategy. Using this mapping function, the system can translate control objectives into specific device actions: temperature targets correspond to the air conditioner's set temperature, carbon dioxide control targets correspond to the fresh air system's airflow, and light targets correspond to the lighting system's brightness.

[0165] During the command issuance process, the system independently issues control commands to devices within each partition, allowing for parallel execution without blocking. Higher-priority actions are prioritized for issuance and adjustment to ensure that critical control needs are responded to first. The execution priority can be represented as follows:

[0166] Where U represents the level of comfort urgency, E represents the level of energy consumption impact, and R represents the execution priority. and This is the weighting coefficient; this priority function is used to determine the execution order and response level of control actions.

[0167] After the control is executed, the system obtains the result receipts returned by each device. The receipts include whether the execution was successful, the current device status, the current operating parameters, the execution delay, and other abnormal statuses. Based on this, the system outputs further device execution commands and generates control result receipts and execution status records for subsequent status updates and fault handling.

[0168] The core of this method lies not in unified control, but in refined execution by zone. Its advantage is that traditional building control is mostly broadcast-style unified regulation, which makes it difficult to adjust each zone independently. This step uses a zone control mapping table to establish the correspondence between zone, equipment and action, so that each area can execute differentiated control strategies, thereby reducing ineffective control and control conflicts, improving response speed, enhancing system scalability, and supporting distributed control execution in large-scale buildings.

[0169] Step 10: Feedback learning and self-optimization.

[0170] This step aims to continuously revise the comfort model, energy consumption model, and control strategy based on the feedback data generated after control execution, so as to achieve closed-loop self-learning of the system. Compared with one-time open-loop control, this step obtains feedback again after each round of control execution and continuously optimizes the model parameters and control strategy accordingly, forming a continuous closed-loop iterative mechanism.

[0171] This step receives feedback data on body status, environment, energy consumption, historical model parameters, and historical control results after control execution to evaluate the effectiveness of the previous control execution. Specifically, it includes determining whether comfort has improved, whether energy consumption has decreased, revising the individual comfort model, revising the group comfort model, revising the energy consumption model, and outputting the updated comfort model, the updated energy consumption model, and a new round of control strategy to optimize the control parameters for the next round.

[0172] In implementation, this step employs a closed-loop incremental self-learning method based on comfort response residuals. The "comfort response residual" refers to the difference between the actual comfort state after control execution and the predicted comfort state before control, used to characterize the deviation between the system's prediction and the actual human feedback. The "closed-loop incremental self-learning" means that after each round of control, the system only uses newly added feedback data to locally update the model parameters, rather than retraining the entire model. Through this method, the system can continuously learn its own control errors, the model gradually adapts to changes in user and environment over time, and the control strategy is continuously corrected, thus possessing long-term adaptive capabilities.

[0173] like Figure 10As shown, the specific processing steps include: First, acquiring feedback data after execution, including individual physical state data, group comfort state, real-time environmental data, equipment energy consumption data, and current control results. This feedback data is used to evaluate the control effect of this round. Then, generating the actual comfort state. The system recalculates the actual comfort score based on post-execution body temperature, skin conductance, blood pressure, fatigue level, and environmental parameters to reflect the true human comfort state after the control action is executed. Next, calculating the comfort response residual. The system compares the predicted comfort level before control with the actual comfort level after control and defines the comfort response residual as:

[0174] in, Indicates the comfort response residual. Indicates actual comfort level. This represents the predicted comfort level. The comfort response residual is used to measure the model prediction error; the larger the residual value, the greater the deviation between the model prediction and the actual feedback.

[0175] Simultaneously calculate the energy consumption response residual, which is defined as:

[0176] in, This represents the residual in the energy consumption response. Indicates actual energy consumption. This represents the predicted energy consumption. The energy consumption response residual is used to evaluate the deviation of the energy-saving effect after control.

[0177] After completing the residual calculation, the system performs an incremental model update. This incremental model update means that the system does not retrain the entire model, but instead uses only the newly added feedback data to make local corrections to the existing model.

[0178] Specifically, this includes: correcting individual comfort anchor points; correcting group comfort kernel density distribution; correcting energy consumption envelope baseline; and correcting dynamic control weight parameters in comfort-energy consumption collaborative decision-making.

[0179] Preferably, the model parameter update process can be expressed as:

[0180] in, Indicates the current model parameters. This represents the updated model parameters, and E represents the feedback residual. This represents the learning rate. The learning rate controls the magnitude of parameter updates; the larger the residual, the greater the magnitude of model parameter correction.

[0181] After the model update is completed, the system regenerates the next round of control strategy based on the updated comfort model and energy consumption model, including re-outputting the group comfort target, energy consumption target, dynamic control weights, and new control actions.

[0182] A closed-loop control chain is formed, so that the control result of each round can affect the control strategy of the next round, and continuous self-learning optimization is achieved.

[0183] The beneficial effects of the above method are as follows: human comfort and building energy consumption continuously change with time, personnel, seasons, and usage habits, making it difficult for traditional static models to adapt to real-world scenarios in the long term. This module, however, constructs a closed-loop mechanism through post-execution feedback, uses comfort response residuals to correct prediction errors, and employs incremental updates to reduce model training costs. This allows the system to continuously adapt to changes in users and the environment, thereby improving comfort prediction accuracy, energy-saving control stability, and long-term control performance. The core of this method lies in the system's ability to continuously learn its own control errors, rather than relying on a fixed model that remains unchanged over time.

[0184] Example 2: An adaptive control system that integrates spatial characteristics and comfort perception.

[0185] like Figure 11 As shown, the adaptive control system integrating spatial characteristics and comfort perception in this embodiment includes: a building space zoning modeling module, a multi-source data acquisition module, a personnel positioning and zoning identification module, an individual comfort modeling module, a group comfort fusion module, an environmental state assessment module, an energy consumption characteristic modeling module, a comfort-energy consumption collaborative decision-making module, an equipment control execution module, and a feedback learning and self-optimization module.

[0186] The architectural space zoning modeling module is used to divide the architectural space into multiple zones and construct spatial feature vectors for each zone. This module acquires the orientation, spatial function, window-to-wall ratio, air conditioning coverage efficiency, distance from air conditioning units, and spatial dimensions of each spatial unit constituting the architectural space, extracting these as features for each spatial unit. A hierarchical clustering method is used to perform cluster analysis on the features of all spatial units, grouping spatial units with similar attributes, similar environmental response characteristics, and consistent equipment control characteristics into the same zone. Statistical values ​​of the features of each spatial unit within the merged zone are extracted to form a spatial feature vector uniquely corresponding to that zone.

[0187] The multi-source data acquisition module is used to collect real-time physical status data of personnel located in each of the aforementioned zones, as well as environmental data and energy consumption data for each zone. The collected data is then mapped to the corresponding zone after being bound with timestamps and spatial coordinates. The physical status data includes one or more of body temperature, blood pressure, skin conductance, and fatigue level, collected via a wearable wristband. The environmental data includes one or more of temperature, humidity, air pressure, carbon dioxide concentration, illuminance, and noise, collected via environmental sensors. The energy consumption data includes equipment power, operating status, and historical energy consumption, collected via a building energy management system.

[0188] The personnel location and partition affiliation identification module is used to identify valid users within each partition based on personnel location and dwell time. This module acquires real-time location information of personnel and determines their current partition based on the partition boundaries; it continuously records the dwell time of the personnel within their current partition; when the dwell time exceeds a preset threshold, it marks the personnel as valid users and establishes their affiliation with their current partition.

[0189] The individual comfort modeling module is used to build an individual comfort prototype library for each user. It calculates an individual comfort score by matching a real-time body state vector formed from real-time collected body state data with body state response anchor points in the individual comfort prototype library. This module collects historical body state data when the user confirms they are in a comfortable state, and extracts the center value, fluctuation range, and feature weights of multiple body state features from this data. These are combined to form at least one body state response anchor point, and all of these anchor points constitute the user's individual comfort prototype library. The real-time body state vector is matched with each body state response anchor point in the individual comfort prototype library to calculate a matching score. The highest matching score is then converted into the individual comfort score according to a preset mapping function.

[0190] The group comfort fusion module is used to fuse the individual comfort scores and individual preference intervals of each effective user within the same partition, forming a group consensus comfort distribution and determining the group target environmental parameter interval. This module determines the individual preference interval and individual preference center for each effective user, and calculates individual weights by combining their individual comfort score, effectiveness status, and the group density of the partition; using a comfort kernel function, it maps the individual preference interval and individual weight of each effective user to a weighted kernel function in the environmental parameter space; it superimposes the weighted kernel functions of all effective users within the partition to form the group consensus comfort distribution; it finds the peak point of the group consensus comfort distribution, and selects a continuous interval centered on this peak point that satisfies a preset retention coefficient as the group target environmental parameter interval for that partition.

[0191] The environmental status assessment module compares the real-time environmental data of the corresponding partition with the target environmental parameter range of the group across multiple environmental dimensions to identify the dominant dimension causing the environmental deviation, and outputs the adjustment direction corresponding to the dominant dimension. This module performs a Cartesian product operation on the range of the target environmental parameter range of the group across each environmental dimension to construct a multidimensional comfort bounding box; when the real-time environmental data exceeds the multidimensional comfort bounding box, it calculates the normalized deviation of the excess portion across each environmental dimension; the environmental dimension with the largest normalized deviation is determined as the dominant deviation dimension, and the adjustment direction corresponding to the dominant dimension is output accordingly.

[0192] The energy consumption characteristic modeling module is used to establish an energy consumption characteristic model based on the operating data of the equipment in each partition, and to extract the optimal power range and the lowest energy consumption state. This module extracts the operating time, start-stop frequency, average power, and peak power of the equipment according to a preset time window, forming an operating condition feature vector; it performs cluster analysis on the operating condition feature vector to obtain multiple operating condition clusters; it statistically analyzes the energy consumption distribution within each operating condition cluster, extracts the energy consumption envelope boundary and baseline curve, and determines the lowest energy consumption state and the optimal power range accordingly.

[0193] The comfort-energy consumption collaborative decision-making module dynamically weighs comfort weights and energy consumption weights based on the current comfort gap and energy consumption pressure, and generates control decisions. This module calculates the difference between the target group's comfort level and the current group's comfort level as the comfort gap; it comprehensively evaluates the current energy consumption status, external electricity prices, power rationing conditions, or energy-saving target constraints to generate the energy consumption pressure; it calculates a dynamic gating coefficient based on the comfort gap and the energy consumption pressure, and uses the dynamic gating coefficient to determine the comfort weights and the energy consumption weights; it calculates a comfort benefit score and an energy cost score for each candidate control action, and uses the comfort weights and energy consumption weights to perform a weighted sum to obtain a comprehensive decision score; finally, it selects the candidate control action with the highest comprehensive decision score as the control decision.

[0194] The equipment control execution module is used to convert the control decisions into control commands for each device within the corresponding zone and execute them. The devices include one or more of air conditioning, fresh air systems, exhaust systems, lighting systems, and fans; the control commands are used to adjust one or more of the following: set temperature, air volume, valve opening, lighting brightness, and start / stop status.

[0195] The feedback learning and self-optimization module is used to calculate residuals based on post-execution feedback data on physical state, environment, and energy consumption, and to incrementally update the model and control parameters. This module acquires the post-execution feedback data on physical state, environment, and energy consumption; calculates the difference between the actual and predicted comfort levels after control execution as the comfort response residual, and calculates the difference between the actual and predicted energy consumption as the energy consumption response residual; using the comfort response residual and the energy consumption response residual, it incrementally updates the individual comfort prototype library, the group consensus comfort distribution, the energy consumption feature model, and the comfort weights and energy consumption weights, forming a closed-loop self-learning optimization.

[0196] Example 3: A computer system.

[0197] The computer system includes a processor and memory. The memory and processor can be interconnected via a bus system and / or other forms of connection. The processor can include any one or more processors such as a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP). The memory can include volatile memory, such as random access memory (RAM). The memory can also include non-volatile memory. volatile memory), such as read-only memory (read-only memory). ROM (Mobile Only Memory), flash memory, hard disk drive (HDD), or solid state drive (SSD). The memory stores executable program code, which the processor executes to implement the adaptive control method for integrating spatial features and comfort perception as described in Embodiment 1.

[0198] Example 4: A computer-readable storage medium. For example, the computer-readable storage medium may be a read-only memory (ROM). Read-only memory (ROM), random access memory (RAM), and compact disc read-only memory (CD-ROM). Only memory, CD ROM, magnetic tape, floppy disk, and optical data storage devices, etc. This computer-readable storage medium is used to store computer-readable instructions, which, when executed by a computer, can implement the adaptive control method for integrating spatial characteristics and comfort perception described in Embodiment 1.

Claims

1. An adaptive control method integrating spatial characteristics and comfort perception, characterized in that, include: The building space is divided into multiple zones, and a spatial feature vector is constructed for each zone. Real-time data collection of physical status of personnel in each of the partitions, environmental data and energy consumption data of each partition, and mapping of the collected data to the corresponding partitions after binding timestamps and spatial coordinates; Identify valid users within each of the aforementioned zones based on user location and dwell time; An individual comfort prototype library is built for each user. The individual comfort score is calculated by matching the real-time body state vector formed by the real-time collected body state data with the body state response anchor points in the individual comfort prototype library. The individual comfort scores and individual preference ranges of each valid user in the same partition are merged to form the group consensus comfort distribution of the partition, and the group target environment parameter range is determined accordingly. The real-time environmental data of the corresponding partition is compared with the target environmental parameter range of the group on multiple environmental dimensions to identify the dominant dimension causing the environmental deviation, and the adjustment direction corresponding to the dominant dimension is output accordingly. Based on the operating power, duration, and historical energy consumption of the equipment in each partition, an energy consumption characteristic model is established, and the optimal power range and the lowest energy consumption state are extracted. Based on the current comfort gap and energy consumption pressure, the comfort weight and energy consumption weight are dynamically weighed to calculate the comprehensive decision score for multiple preset candidate control actions to generate control decisions. The control decisions are translated into control commands for each device within the corresponding partition and executed. After execution, the system acquires body state data, environmental data, and energy consumption data, calculates comfort response residuals and energy consumption response residuals, and incrementally updates the individual comfort prototype library, the group consensus comfort distribution, the energy consumption feature model, the comfort weights, and the energy consumption weights to form a closed-loop self-learning optimization.

2. The method according to claim 1, characterized in that, The process of dividing the building space into multiple zones and constructing spatial feature vectors for each zone includes: The orientation, spatial function, window-to-wall ratio, air conditioning coverage efficiency, and spatial scale of each spatial unit constituting the building space are obtained and extracted as features of each spatial unit. Hierarchical clustering is used to perform cluster analysis on the features of all the spatial units, and spatial units with similar attributes, similar environmental response characteristics and consistent equipment control characteristics are grouped into the same partition. The statistical values ​​of the features of each spatial unit within the merged partition are extracted to form the spatial feature vector that uniquely corresponds to the partition.

3. The method according to claim 1, characterized in that, The method of identifying valid users within each partition based on user location and dwell time includes: Obtain the real-time location information of personnel and determine their current partition based on the boundaries of the partition; Continuously record the duration of time the person stays in the current partition; When the dwell time exceeds a preset threshold, the person is marked as a valid user and their affiliation with the current partition is established.

4. The method according to claim 1, characterized in that, The calculation of individual comfort scores includes: Collect historical physical state data of users when they confirm that they are in a comfortable state, and extract the center value, fluctuation range and feature weight of multiple physical state features from them, and combine them to form at least one physical state response anchor point. All of the physical state response anchor points constitute the user's individual comfort prototype library. The real-time body state vector is matched with each body state response anchor point in the individual comfort prototype library, and a matching score is calculated. The highest matching score is converted into the individual's comfort score according to a preset mapping function.

5. The method according to claim 1, characterized in that, The process of fusing the individual comfort scores and their individual preference ranges of all valid users within the same partition to form a group consensus comfort distribution for that partition, and determining the group's target environmental parameter range accordingly, includes: Determine the individual preference range and individual preference center for each valid user, and calculate the individual weight by combining their individual comfort score, validity status, and the group density of that partition; Using the comfort kernel function, the individual preference interval and individual weight of each effective user are mapped to a weighted kernel function in the environmental parameter space; The weighted kernel function of all valid users in the partition is superimposed to form the comfortable distribution of the group consensus. Find the peak point of the comfortable distribution of the group consensus, and select a continuous interval that satisfies the preset retention coefficient as the group target environment parameter interval of the partition.

6. The method according to claim 1, characterized in that, The real-time environmental data of the corresponding partition is compared with the target environmental parameter range of the group across multiple environmental dimensions to identify the dominant dimension causing the environmental deviation, and the adjustment direction corresponding to the dominant dimension is output accordingly, including: The range of the target environmental parameters of the group in each environmental dimension is subjected to Cartesian product operation to construct a multidimensional comfort bounding box. When the real-time environmental data exceeds the multidimensional comfort bounding box, the normalized deviation of the excess portion in each environmental dimension is calculated. The environmental dimension with the largest normalized deviation is determined as the dominant dimension, and the adjustment direction corresponding to the dominant dimension is output accordingly.

7. The method according to claim 1, characterized in that, The establishment of the energy consumption characteristic model and the extraction of the optimal power range and the lowest energy consumption state include: Extract the equipment's running time, number of start-stop cycles, average power, and peak power according to a preset time window to form an operating condition feature vector; Cluster analysis is performed on the feature vectors of the operating conditions to obtain multiple clusters of operating conditions; The energy consumption distribution within each operating condition cluster is statistically analyzed, and the energy consumption envelope boundary and baseline curve are extracted to determine the minimum energy consumption state and the optimal power range.

8. The method according to claim 1, characterized in that, The process involves dynamically balancing comfort and energy consumption weights based on the current comfort gap and energy pressure, calculating a comprehensive decision score for multiple preset candidate control actions to generate a control decision, including: The difference between the comfort level of the target group and the comfort level of the current group is calculated as the comfort gap; The energy consumption pressure is generated by comprehensively assessing the current energy consumption status, external electricity prices, power rationing conditions, or energy-saving target constraints. The dynamic gating coefficient is calculated based on the comfort gap and the energy consumption pressure, and the comfort weight and the energy consumption weight are determined using the dynamic gating coefficient. For each candidate control action, calculate the comfort benefit score and the energy consumption cost score, and then use the comfort weight and the energy consumption weight to perform a weighted sum to obtain the comprehensive decision score. The candidate control action with the highest comprehensive decision score is selected as the control decision.

9. An adaptive control system integrating spatial characteristics and comfort perception, characterized in that, include: The building space partitioning modeling module is configured to divide the building space into multiple partitions and construct spatial feature vectors for each partition. The multi-source data acquisition module is configured to collect in real time the physical status data of people located in each of the partitions, the environmental data and energy consumption data of each partition, and map the collected data to the corresponding partitions after binding the timestamps and spatial coordinates. The personnel location and partition affiliation identification module is configured to identify valid users within each partition based on personnel location and dwell time; The individual comfort modeling module is configured to build an individual comfort prototype library for each user. It calculates the individual comfort score by matching the real-time body state vector formed by real-time collected body state data with the body state response anchor points in the individual comfort prototype library. The group comfort fusion module is configured to fuse the individual comfort scores and individual preference ranges of each valid user in the same partition to form the group consensus comfort distribution of the partition, and determine the group target environment parameter range accordingly. The environmental status assessment module is configured to compare the real-time environmental data of the corresponding partition with the target environmental parameter range of the group on multiple environmental dimensions to identify the dominant dimension causing the environmental deviation, and output the adjustment direction corresponding to the dominant dimension accordingly. The energy consumption feature modeling module is configured to establish an energy consumption feature model based on the operating power, duration and historical energy consumption of the devices in each partition, and extract the optimal power range and the lowest energy consumption state. The comfort-energy consumption collaborative decision-making module is configured to dynamically weigh comfort weights and energy consumption weights based on the current comfort gap and energy consumption pressure, and calculate a comprehensive decision score for multiple preset candidate control actions to generate control decisions; The device control execution module is configured to convert the control decision into control instructions for each device in the corresponding partition and execute them. The feedback learning and self-optimization module is configured to acquire body state data, environmental data, and energy consumption data after execution, calculate comfort response residuals and energy consumption response residuals, and incrementally update the individual comfort prototype library, the group consensus comfort distribution, the energy consumption feature model, the comfort weights, and the energy consumption weights accordingly, forming a closed-loop self-learning optimization.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the adaptive control method for integrating spatial features and comfort perception as described in any one of claims 1 to 8.

Images