Building environment dynamic regulation and control method and system based on Internet of Things driving

Abstract

The invention discloses a building environment dynamic regulation and control method and system based on Internet of Things driving, particularly relates to the field of building environment intelligent regulation and control, and is used for solving the problems of multi-region environment dynamic regulation and control and fine management. The environment data of each sub-region of the building is accurately acquired and synchronized, real-time region division is realized in combination with an adaptive clustering algorithm, and the accurate capture of environment features and the reasonability of space division are ensured. On the basis, the independence and interaction strength of the sub-regions are comprehensively evaluated by calculating an environmental independence index, and a scientific basis is provided for refined and coordinated regulation and control. Based on a rule engine and a decision tree reasoning algorithm, an optimal regulation and control strategy is automatically generated, and continuous optimization is performed in combination with a Q learning algorithm, so that the system can adapt to environmental changes, dynamically adjust regulation and control actions, realize intelligent optimization and self-learning of the regulation and control strategy, and remarkably improve the regulation and control efficiency and accuracy of the building environment.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control of building environment, and more specifically, to a method and system for dynamic control of building environment driven by the Internet of Things. Background Art

[0002] In modern building environments, the division of indoor areas often presents complex diversity due to factors such as function, usage requirements, and building structure. Usually, the rooms, corridors, office areas, and other areas of a building form multiple relatively independent sub-areas due to the existence of partitions and partitions. However, these sub-areas are not completely closed, and their internal environmental factors such as temperature, humidity, airflow, air quality, etc. are affected by many aspects and may change significantly due to personnel activities, equipment operation, or changes in the external environment. Especially in large-scale complex buildings, the environmental requirements of different areas are different. For example, the temperature control requirements of conference rooms are completely different from those of office areas; at the same time, dining areas and leisure areas also have unique requirements for air quality control. Therefore, in such a diverse scenario, the building environment control system needs to be able to dynamically respond to the unique needs of each area to meet the comfort and energy efficiency requirements of each area at different time periods.

[0003] Existing environmental control methods mainly rely on preset fixed rules for independent control, but this method shows obvious shortcomings in complex multi-region building environments. Due to the complex environmental conditions, mutual influences and dynamic changes of each sub-region, it is difficult for fixed rules to adapt to the changes in demand in each region in real time, especially when there is a strong environmental interaction between different regions. The control system cannot effectively distinguish which areas should be independently controlled and which areas need to be coordinated. In addition, existing systems usually ignore the complexity of the transmission and interaction of environmental parameters between regions, resulting in the system being unable to respond and control accurately when important parameters such as airflow, temperature and humidity change. Not only does the existing method lag in control and is difficult to adapt to real-time environmental changes, it is also easy to cause over-control or waste of resources in some areas, and it is impossible to achieve refined management and energy efficiency optimization of different areas in the building.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a method and system for dynamic control of building environment driven by the Internet of Things. The environmental data of each sub-area of ​​the building is obtained by multi-type Internet of Things sensors. The dynamic time series data fusion and adaptive clustering algorithm are combined to realize the accurate division of sub-areas, ensure the accurate capture of environmental characteristics and the rationality of regional division, and can reflect the changes in environmental status in real time, providing a solid data foundation for subsequent control. On this basis, by calculating the environmental independence index, the independence and interaction intensity of the sub-areas are evaluated to ensure that the selection of the control mode can adapt to the characteristics of different areas, which not only improves the refined control ability in the complex building environment, but also makes reasonable collaborative control decisions when the regional interaction is significant. The optimal control strategy is automatically generated by the reasoning algorithm based on the rule engine and the decision tree, and the Q learning algorithm in the reinforcement learning is combined for continuous optimization, so that the system can dynamically adjust the control action according to the environmental changes, realize long-term intelligent optimization, ensure that the control efficiency and stability are significantly improved in the complex environment, and finally realize efficient and intelligent dynamic control of the building environment to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] S1, deploys multiple types of IoT sensors in each sub-area of ​​the building, and uses a dynamic time series data fusion algorithm to perform time synchronization and spatial fusion on the collected environmental data to generate an indoor environmental dataset.

[0008] S2, based on environmental data, uses the K-Means clustering algorithm to initially select cluster centers according to the density and distribution of environmental parameters, dynamically adjusts the number of clusters by calculating the variance and silhouette coefficient, and iteratively updates the location of cluster centers until the clustering results converge, completing the adaptive division of sub-areas.

[0009] S3, by calculating the compactness of the internal environmental data points of the sub-region and the differences in environmental parameters between adjacent regions, combined with the weighted processing of internal consistency and boundary interaction strength, the environmental independence index is obtained, and the preset independent distinction threshold is used to determine whether the sub-region generates an independent regulatory signal or a collaborative regulatory signal.

[0010] S4, if a collaborative regulation signal is obtained, then at the boundary of the sub-region, an environmental interaction graph is constructed through a path analysis algorithm based on graph theory, the weights of the airflow and heat conduction paths between regions are calculated, the shortest path algorithm is used to identify regional pairs with strong environmental interactions, and the regional division is modified according to the environmental interaction situation.

[0011] S5, based on the revised regional division results and environmental independence assessment, uses the rule engine and the decision tree-based rule reasoning algorithm to generate an adaptive control strategy, builds a control rule library, uses the decision tree to automatically generate control rules, selects the optimal control action according to the current environmental status, and sends instructions to related devices.

[0012] S6, continuously optimizes environmental changes and control effects through the Q learning algorithm in reinforcement learning, defines the state space and action set, sets the reward function according to the control effect, updates the Q value according to the Q learning update formula, and selects the action with the largest Q value as the current control strategy.

[0013] In a preferred embodiment, step S1 includes the following contents:

[0014] S1.1, evenly deploy temperature and humidity sensors, air quality sensors, light sensors, and micro airflow sensors in each sub-area of ​​the building to form a multi-type sensor network.

[0015] S1.2, the data collected by each sensor node is time synchronized according to its sampling timestamp.

[0016] S1.3, based on the geographical location of the sensors and the spatial layout information of the sub-regions, the multi-type environmental data in each sub-region are fused into a unified environmental data set through a spatial data fusion algorithm.

[0017] S1.4, perform data cleaning on the initially fused environmental data set to detect and exclude abnormal data.

[0018] S1.5, after time synchronization, spatial fusion and data cleaning, the processed data set is input into the data management module to finally generate the indoor environment data set.

[0019] In a preferred embodiment, step S2 includes the following contents:

[0020] S2.1, based on the high-precision indoor environment dataset generated in step S1, extract key environmental parameters and construct the environmental parameters of each sub-area into a multi-dimensional vector.

[0021] S2.2, based on the environmental data after dimensionality reduction, the K-Means clustering algorithm is used to select the initial cluster center.

[0022] S2.3, during the clustering iteration process, the cluster center is dynamically adjusted based on the variance and silhouette coefficient of each clustering result.

[0023] S2.4, automatically adjusts the number of clusters by calculating the variance and silhouette coefficient under different numbers of clusters.

[0024] S2.5, after each iteration, the centroid position of each cluster is recalculated based on the current clustering result, and the distance between the data of each sub-region and the centroid is minimized to ensure that the environmental parameters in the sub-region are similar and meet the standards.

[0025] S2.6, after the clustering results converge, the adaptive division of each sub-region is finally completed.

[0026] In a preferred embodiment, step S3 includes the following contents:

[0027] S3.1, after completing the preliminary division of sub-regions, calculate the compactness of environmental data points within the sub-regions, and use the clustering compactness coefficient to quantify the environmental consistency within the sub-regions; at the same time, calculate the regional interaction effect index, and evaluate the intensity of environmental interaction between regions by analyzing the gradient changes of environmental parameters between sub-regions and adjacent regions.

[0028] The processing process of clustering compactness coefficient:

[0029] a1, for each sub-region C k , including N k data points, calculate the distances between all pairs of points in the sub-region, and construct the distance matrix D k ; The matrix elements are defined as: d ij =||x i -x j ||; where x i and x j Represents sub-region C k The position vectors of the i-th and j-th data points within .

[0030] a2, introduce the compactness mapping function φ(d ij ), defined as follows: Among them: λ>0 is the attenuation coefficient, which controls the rate at which the function decreases; α>1 is the nonlinear coefficient, which adjusts the nonlinearity of the function; the compactness function values ​​of all point pairs are calculated to form a function value matrix.

[0031] a3, multiply the upper triangular elements of the function value matrix to obtain the compactness accumulation Ψ of the sub-region k : Here, the accumulation operation traverses all non-repeated point pairs in the sub-region.

[0032] a4, in order to avoid underflow or overflow in numerical calculation, the compactness accumulation is logarithmically transformed and the logarithmic equalization coefficient η is introduced k : The clustering compactness coefficient CCC is defined as: CCC = -η k ln(Ψ k ).

[0033] In a preferred embodiment, step S3 further includes the following contents:

[0034] Processing process of regional interaction effect index:

[0035] b1, for each sub-region C k and the adjacent sub-region C l First, we need to determine the boundary point sets of the two regions; select each sub-region C k The data point b closest to the boundary k and sub-region C l The boundary data point b in l , forming a boundary point set B kl , contains all data points that interact with adjacent regions on the sub-region boundary, and is defined as follows: kl ={(b k ,b l )|||b k -b l ||<δ}; where b k ∈C k and b l ∈C l , belong to different sub-regions respectively, and δ is a distance threshold used to identify whether two points are located in the boundary interaction region.

[0036] b2, for each pair of boundary points (b k ,b l ), calculate the gradient difference between the two in each environmental parameter, and form the gradient vector G kl , whose components are the differences of various environmental parameters, expressed as: Where: P m (b k ) represents the boundary point b k The value of the mth environmental parameter at ; Yes k and b l The distance between.

[0037] b3, based on the gradient vector calculated in step b2, define the regional interaction strength function ψ(G kl ), which is used to quantify the environmental interaction intensity between sub-regions and is expressed as:

[0038] b4, calculate the interaction strength for all boundary point pairs and accumulate them to obtain the accumulated interaction effect between regions Φ kl :

[0039] b5, logarithm transformation of the interaction effect accumulation and introduction of normalization coefficient ζkl : The final regional interaction effect index IRII is expressed as: IRII = -ζ kl ln(Φ kl ).

[0040] S3.2, the environmental independence index is calculated based on the comprehensive processing of the clustering compactness coefficient and the regional interaction effect index; when the environmental independence index is greater than or equal to the independent distinction threshold, it means that the environmental characteristics of the relevant sub-region are sufficiently independent from the adjacent regions, generating an independent regulation signal; conversely, when the environmental independence index is less than the independent distinction threshold, it means that there is a strong environmental interaction between the relevant sub-region and the adjacent regions, and the independence is insufficient, generating a collaborative regulation signal.

[0041] In a preferred embodiment, step S4 includes the following contents:

[0042] S4.1, for the sub-regions that obtain the synergistic regulation signal, the boundary points between the corresponding sub-regions are used as nodes, and based on the environmental parameters, an environmental interaction graph G = (V, E) is constructed; where the node set V represents the boundary point set B kl , the edge set E represents the interaction path of environmental parameters between sub-areas; by analyzing the airflow velocity, temperature gradient, and heat conductivity, the environmental interaction intensity between each pair of boundary points is calculated and used as the edge weight; the edge weight function is defined as w(b k ,b l ), a nonlinear weight model is established according to the differences in various environmental parameters to indicate the difficulty of interaction between sub-regions.

[0043] S4.2, for each pair of sub-regions, calculate the weights of the airflow path and the heat conduction path between the regions respectively. The weight of the airflow path is w flow (b k ,b l ) is calculated based on the difference in air flow velocity, and the weight of the heat conduction path w heat (b k ,b l ) is calculated based on the temperature change rate and the thermal conductivity coefficient; the formula for the airflow path weight is: Where v(b k ) and v(b l ) represent the air flow velocity at the boundary points respectively; the weight formula of the heat conduction path is: Where κ is the thermal conductivity, T(b k ) and T(b l ) is the temperature value at the boundary point.

[0044] S4.3, use the shortest path algorithm to search for paths on the environmental interaction graph to find the shortest paths on the airflow path and the heat conduction path; the shortest path represents the boundary point pair with the strongest environmental interaction between sub-areas; by identifying the shortest path, determine the sub-area C k and C l The main interaction paths between them are then used to identify the region pairs with strong interactions.

[0045] S4.4, after identifying the main interaction paths of airflow and heat conduction, the initial sub-region division is modified according to the areas with strong environmental interactions; if the interaction intensity of the shortest path exceeds the preset threshold, it means that there is significant environmental interaction between the sub-regions, and the regional boundaries need to be redefined, and the areas with strong interactions need to be merged to form a new regional division.

[0046] In a preferred embodiment, step S5 includes the following contents:

[0047] S5.1, based on the revised regional division result of step S4 and the environmental independence evaluation of step S3, the input data generated by the control strategy is initialized.

[0048] S5.2, based on the environmental characteristics of the building interior and historical environmental data, a control rule base is constructed; the rule base contains predefined control actions under various environmental states; the rule base defines the control behaviors under different environmental states in the form of logical rules, covering different combinations of environmental parameters.

[0049] S5.3, using a decision tree-based rule inference algorithm to generate an adaptive control strategy.

[0050] S5.4, at the end of the decision tree reasoning, each leaf node corresponds to one or more control actions.

[0051] S5.5, the generated control actions will be automatically converted into device instructions by the system and sent to relevant devices through the Internet of Things protocol.

[0052] In a preferred embodiment, step S6 includes the following contents:

[0053] S6.1, for each sub-region, define the state space S in reinforcement learning to describe the current state of environmental regulation.

[0054] S6.2, based on the state space S, define an action set A for each sub-area, and each action in the action set corresponds to a specific control behavior.

[0055] S6.3, according to the control effect, the reward function R(s,a) is set to measure the effect of each action under different environmental conditions; the reward function is designed according to the adjustment range of environmental parameters and the achievement of control goals; the specific reward function form is: Among them, P o (s′) represents the environmental parameter value in the new state s′ after action a is executed in the current state s, is the target environmental parameter value, ΔP o is the allowable parameter deviation; o represents the index of each specific environmental parameter.

[0056] S6.4, after each action is executed, the Q value is updated using the Q value update formula of the Q learning algorithm; the Q value Q(s,a) represents the long-term expected return of selecting action a in state s, and the update formula is: Among them, ρ is the learning rate, which controls the weight of new information at each update; γ is the discount factor, which weighs the relationship between future rewards and current rewards; is the maximum Q value in the next state s′.

[0057] S6.5, after each state transition, select the action a with the largest Q value according to the current state s * , action a * The corresponding Q value Q(s,a * ) is the largest, indicating that the relevant action has the best expected return in the current environment; the selection formula is: After selecting the relevant action, the corresponding control action will be executed and enter the new state s′.

[0058] The building environment dynamic control system driven by the Internet of Things includes: an environmental data acquisition module, an environmental division module, an environmental assessment module, an environmental correction module, a control strategy generation module and a control optimization module.

[0059] The environmental data acquisition module deploys multiple types of IoT sensors in each sub-area of ​​the building, and uses a dynamic time series data fusion algorithm to synchronize and spatially fuse the collected environmental data to generate an indoor environmental data set; the data output is passed to the environmental division module.

[0060] The environmental division module is based on environmental data and uses the K-Means clustering algorithm to initially select cluster centers according to the density and distribution of environmental parameters. The number of clusters is dynamically adjusted by calculating the variance and silhouette coefficient, and the location of the cluster centers is iteratively updated until the clustering results converge, completing the adaptive division of sub-areas. The division results are passed to the environmental assessment module.

[0061] The environmental assessment module calculates the compactness of environmental data points within the sub-region and the differences in environmental parameters between adjacent regions, and combines the weighted processing of internal consistency and boundary interaction strength to derive the environmental independence index. It also determines whether the sub-region generates independent regulation signals or collaborative regulation signals based on the preset independent distinction threshold; the assessment results are transmitted to the environmental correction module.

[0062] If the environmental correction module obtains a collaborative control signal, it will construct an environmental interaction graph at the boundary of the sub-region through a path analysis algorithm based on graph theory, calculate the weights of the airflow and heat conduction paths between regions, use the shortest path algorithm to identify pairs of regions with strong environmental interactions, and correct the regional division according to the environmental interaction situation; the corrected division results are passed to the control strategy generation module.

[0063] Based on the revised regional division results and environmental independence assessment, the control strategy generation module uses a rule engine and a decision tree-based rule inference algorithm to generate an adaptive control strategy, build a control rule library, and use decision trees to automatically generate control rules. It selects the optimal control action according to the current environmental status and sends instructions to related equipment; the generated control strategy is passed to the control optimization module.

[0064] The control optimization module continuously optimizes environmental changes and control effects through the Q-learning algorithm in reinforcement learning, defines the state space and action set, sets the reward function according to the control effect, updates the Q value according to the Q-learning update formula, and selects the action with the largest Q value as the current control strategy; the optimized strategy is finally transmitted to the execution device to realize the control of the building environment.

[0065] The technical effects and advantages of the method and system for dynamically controlling the building environment driven by the Internet of Things of the present invention are as follows:

[0066] 1. The present invention acquires environmental data of each sub-area of ​​a building by deploying multiple types of IoT sensors, combines dynamic time series data fusion algorithms for time synchronization and spatial fusion processing, generates high-precision environmental data sets, and can ensure that the environmental characteristics of each sub-area are accurately captured. Based on the adaptive clustering algorithm of environmental data, the sub-areas are accurately divided by dynamically selecting cluster centers and continuously adjusting the number of clusters, so that the regional division can reflect the changes in environmental status in real time. The comprehensiveness of environmental data and the rationality of spatial division are ensured, the responsiveness to regional differences in the process of dynamic regulation of the building environment is improved, and a solid data foundation is provided for the formulation of subsequent regulation strategies. It can effectively adapt to the complex multi-region environment in the building and realize the efficient operation of refined regulation.

[0067] 2. The present invention dynamically divides the environmental data in the building through an adaptive clustering algorithm, realizes the accurate delineation of sub-regions, and continuously optimizes the regional division results according to the density and distribution of environmental parameters, ensuring that the environmental characteristics of the sub-regions are accurately identified. On this basis, by calculating the compactness within the sub-region and the difference in environmental parameters with adjacent regions, the environmental independence index is obtained, which can accurately evaluate the independence or interaction intensity of each region, thereby providing a reliable basis for the selection of regional control mode. This method of combining regional division with independence assessment not only improves the ability to fine-tune the control of different regions in a complex building environment, but also makes more reasonable collaborative control decisions when there is significant interaction between regions, ensuring that the system can quickly respond to environmental changes, optimize resource utilization efficiency, and ensure the accuracy and sustainability of the control effect.

[0068] 3. The present invention generates an adaptive control strategy through an inference algorithm based on a rule engine and a decision tree, which can automatically select the optimal control action under different environmental conditions, ensuring that the control system responds quickly to real-time environmental changes. At the same time, combined with the Q learning algorithm in reinforcement learning, by continuously optimizing the control strategy and continuously updating the Q value, the system can learn from the actual control effect and gradually improve the control accuracy. This method can not only generate a reasonable control scheme under the initial environmental state, but also dynamically adjust the control action through a self-learning mechanism, realize intelligent optimization in long-term operation, and ensure that in a complex and changeable building environment, the control system can automatically adapt to environmental changes, improve the efficiency and stability of the control, and realize efficient and intelligent dynamic environmental control. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a schematic diagram of the process of the method for dynamic control of building environment driven by the Internet of Things of the present invention;

[0070] Figure 2 It is a structural schematic diagram of the building environment dynamic control system driven by the Internet of Things according to the present invention. DETAILED DESCRIPTION

[0071] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0072] Example 1

[0073] Figure 1 The present invention provides a method for dynamically controlling a building environment based on the Internet of Things, including:

[0074] S1, deploys multiple types of IoT sensors in each sub-area of ​​the building, and uses a dynamic time series data fusion algorithm to perform time synchronization and spatial fusion on the collected environmental data to generate an indoor environmental dataset.

[0075] S2, based on environmental data, uses the K-Means clustering algorithm to initially select cluster centers according to the density and distribution of environmental parameters, dynamically adjusts the number of clusters by calculating the variance and silhouette coefficient, and iteratively updates the location of cluster centers until the clustering results converge, completing the adaptive division of sub-areas.

[0076] S3, by calculating the compactness of the internal environmental data points of the sub-region and the differences in environmental parameters between adjacent regions, combined with the weighted processing of internal consistency and boundary interaction strength, the environmental independence index is obtained, and the preset independent distinction threshold is used to determine whether the sub-region generates an independent regulatory signal or a collaborative regulatory signal.

[0077] S4, if a collaborative regulation signal is obtained, then at the boundary of the sub-region, an environmental interaction graph is constructed through a path analysis algorithm based on graph theory, the weights of the airflow and heat conduction paths between regions are calculated, the shortest path algorithm is used to identify pairs of regions with strong environmental interactions, and the regional division is modified according to the environmental interaction situation.

[0078] S5, based on the revised regional division results and environmental independence assessment, uses the rule engine and the decision tree-based rule reasoning algorithm to generate an adaptive control strategy, builds a control rule library, uses the decision tree to automatically generate control rules, selects the optimal control action according to the current environmental status, and sends instructions to related devices.

[0079] S6, continuously optimizes environmental changes and control effects through the Q learning algorithm in reinforcement learning, defines the state space and action set, sets the reward function according to the control effect, updates the Q value according to the Q learning update formula, and selects the action with the largest Q value as the current control strategy.

[0080] In order to ensure accurate environmental control of each sub-area within the building, it is necessary to first establish a foundation based on comprehensive environmental data collection. By deploying multiple types of IoT sensors in each sub-area, multi-dimensional environmental parameters such as temperature, humidity, air flow velocity, and air quality are monitored in real time to generate high-precision environmental data. By synchronizing and spatially fusion processing these environmental data, not only can the temporal data inconsistency be eliminated, but also the spatial data coverage problem caused by the distribution of sensors can be optimized, thereby generating a complete and high-precision environmental data set. This data set will provide reliable basic data support for subsequent sub-area division, independence assessment, and dynamic control, ensuring that the environmental characteristics within and between regions can be accurately identified and processed.

[0081] Step S1 includes the following contents:

[0082] S1.1, evenly deploy temperature and humidity sensors, air quality sensors, light sensors and micro airflow sensors in each sub-area of ​​the building to form a multi-type sensor network. The sensor layout is carried out according to the pre-planned points, and multiple nodes are deployed in each sub-area to ensure the coverage and accuracy of environmental data in each area. The sensor transmits data through the Internet of Things protocol (such as LoRa or ZigBee), and the collected data includes temperature, humidity, carbon dioxide concentration, PM2.5, light intensity and airflow speed. Each sensor collects data according to a predetermined sampling frequency (such as once every 5 seconds) and transmits it to the central data processing unit in real time through the network.

[0083] S1.2, the data collected by each sensor node is time-synchronized according to its sampling timestamp. Since different sensors may have different sampling frequencies and time accuracy, the timestamp correction algorithm (such as the NTP protocol based on distributed systems) is used to time-align the data of each sensor to ensure the consistency of multiple types of data at the same time. During the time synchronization process, if a delay or loss in data transmission of a sensor is detected, the system automatically fills in the missing values ​​through interpolation or data completion algorithms to ensure the integrity of the time series.

[0084] S1.3, based on the geographical location of the sensor and the spatial layout information of the sub-region, the multi-type environmental data in each sub-region are fused into a unified environmental data set through a spatial data fusion algorithm. The spatial data fusion method based on the weighted Vorono i graph is adopted. First, the Vorono i polygon area is constructed according to the spatial distribution of the sensor, and then the weighted average is performed according to the distance between the sensor and the sub-region boundary to convert the collected data of each sensor into the representative environmental data of the sub-region. This fusion method takes into account the distance factor of each sensor to ensure the reasonable distribution of data weights during the spatial data fusion process.

[0085] S1.4, perform data cleaning on the initially fused environmental data set to detect and exclude abnormal data. First, use the 3σ rule to detect outliers on the data and remove abnormal data far from the average value; then use wavelet transform to reduce noise in the time series to ensure data stability. During the data cleaning process, if continuous data anomalies or sensor failures are detected, the system automatically marks the data source of the sensor as unreliable and triggers alternative sensors for data compensation to ensure the integrity of the overall data set.

[0086] S1.5, after time synchronization, spatial fusion and data cleaning, the processed data set is input into the data management module to finally generate a high-precision indoor environment data set. The data set covers multi-dimensional environmental parameters such as temperature, humidity, air quality, light and airflow in each sub-area, and has corrected and removed noise and abnormal data. This high-precision data set provides basic data support for subsequent regional division and regulation decisions.

[0087] After obtaining and processing comprehensive environmental data, the next step is to reasonably divide the sub-areas inside the building. The purpose of sub-area division is to find out the environmental characteristics of different areas based on the similarity and spatial distribution characteristics of environmental data. By using an adaptive clustering algorithm, the compactness of data points within each sub-area is first analyzed to identify areas with strong environmental consistency. Subsequently, the spatial range of each sub-area is accurately divided based on the boundary characteristics and environmental parameter differences between regions. This step not only ensures the scientific nature of the regional division, but also provides clear regional boundaries and characteristic descriptions for the subsequent environmental independence assessment, laying a solid foundation for differentiated regulation between regions.

[0088] Step S2 includes the following contents:

[0089] S2.1, based on the high-precision indoor environment dataset generated in step S1, extract key environmental parameters, including temperature, humidity, air quality, light intensity, air flow velocity, etc., and construct the environmental parameters of each sub-area into a multidimensional vector. Through the principal component analysis (PCA) method, the high-dimensional environmental parameters are reduced in dimension, the main components that have the greatest impact on environmental differences are retained, data redundancy is reduced, and clustering efficiency is improved.

[0090] S2.2, based on the environmental data after dimensionality reduction, the K-Means clustering algorithm is used to select the initial cluster center. The initial selection of the cluster center is based on the density distribution. Specifically, the relative distance and similarity between sub-regions are calculated through the distance matrix, and the points where the environmental parameters between sub-regions meet the standard are selected as the initial cluster center to ensure that the initial cluster center distribution is reasonable and can cover different environmental states.

[0091] S2.3, during the clustering iteration process, the cluster center is dynamically adjusted based on the variance and silhouette coefficient of each clustering result. The variance is used to measure the discreteness of the environmental parameters in each sub-region, and the silhouette coefficient is used to evaluate the separation between sub-regions. When the variance value of a sub-region is too high or the silhouette coefficient is too low, the position of the cluster center is dynamically adjusted. By redistributing the members of each sub-region, the cluster center is moved to a more representative sub-region position to improve the clustering accuracy.

[0092] S2.4, automatically adjusts the number of clusters by calculating the variance and silhouette coefficient under different numbers of clusters. The initial number of clusters is set by empirical values. As the iteration proceeds, if the current number of clusters fails to meet the requirements of regional division, the system automatically increases or decreases according to the peak position of the silhouette coefficient until the silhouette coefficient reaches the local optimal value to determine the optimal number of clusters. This process dynamically adjusts the number of clusters to make the regional division more consistent with the actual distribution of environmental data.

[0093] S2.5, after each iteration, the centroid position of each cluster is recalculated based on the current clustering result, and the distance between the data and the centroid of each sub-region is minimized to ensure that the environmental parameters in the sub-region are similar and meet the standards. The location of the cluster center is continuously updated iteratively until the centroid positions of all clusters converge, that is, all centroid positions are stable and the data in the region no longer changes significantly, and the regional division is completed.

[0094] S2.6, after the clustering results converge, the adaptive division of each sub-region is finally completed. The environmental parameters in each sub-region are similar and significantly different from those in other regions. The regional boundaries are automatically determined based on the clustering results to ensure that each region has independent environmental characteristics.

[0095] The present invention acquires environmental data of each sub-area of ​​a building by deploying multiple types of IoT sensors, combines time synchronization and spatial fusion processing with a dynamic time series data fusion algorithm, generates a high-precision environmental data set, and can ensure that the environmental characteristics of each sub-area are accurately captured. The adaptive clustering algorithm based on environmental data dynamically selects cluster centers and continuously adjusts the number of clusters to accurately divide sub-areas, so that regional division can reflect changes in environmental status in real time. The comprehensiveness of environmental data and the rationality of spatial division are ensured, and the responsiveness to regional differences in the dynamic regulation of the building environment is improved, providing a solid data foundation for the subsequent formulation of regulatory strategies, which can effectively adapt to the complex multi-region environment in the building and achieve efficient operation of refined regulation.

[0096] After completing the sub-region division, it is necessary to further evaluate the environmental independence of each sub-region in order to provide a basis for subsequent dynamic regulation. The core of environmental independence assessment is to combine the environmental consistency within the sub-region and the interaction intensity on the regional boundary, and obtain the environmental independence index of each sub-region through the analysis of internal tightness and comprehensive processing of differences in environmental parameters of adjacent regions. This index is used to quantify the independence performance of the sub-region in the overall environment, and help the system determine whether the region has the necessity of independent regulation based on the preset independent distinction threshold. If the internal environment of the sub-region is relatively stable and the interaction with the adjacent region is weak, an independent regulation signal will be generated; if the internal environment of the region fluctuates greatly and there is a strong interaction with the adjacent region, a coordinated regulation signal will be generated. This evaluation provides a scientific basis and data support for the implementation of precise regulation strategies.

[0097] Step S3 includes the following contents:

[0098] S3.1, after completing the preliminary division of sub-regions, calculate the compactness of environmental data points within the sub-regions, and use the cluster compactness coefficient to quantify the environmental consistency within the sub-regions. At the same time, calculate the regional interaction effect index, and evaluate the intensity of environmental interaction between regions by analyzing the gradient changes of environmental parameters between sub-regions and adjacent regions. These two indicators are used to comprehensively evaluate the environmental independence of sub-regions after weighted processing, providing basic data for the subsequent generation of regulatory signals.

[0099] The processing process of clustering compactness coefficient:

[0100] a1, for each sub-region C k , including N k data points, calculate the distances between all pairs of points in the sub-region, and construct the distance matrix D k The matrix elements are defined as: ij =||x i -x j ||; where x i and x j Represents sub-region C k The position vectors of the i-th and j-th data points in , ||·|| represents the norm of the vector, and the Euclidean distance is used or a suitable distance metric is selected according to actual needs.

[0101] a2, introduce the compactness mapping function φ(d ij ), the function must satisfy the requirement that it takes a larger function value when the distance is small, and the function value decreases rapidly when the distance increases. The definition is as follows:

[0102] in:

[0103] λ>0 is the attenuation coefficient, which controls the rate at which the function decreases.

[0104] α>1 is the nonlinear coefficient, which adjusts the nonlinearity of the function.

[0105] Calculate the compactness function values ​​of all point pairs and construct a function value matrix, where the matrix elements are φ(d ij ).

[0106] a3, multiply the upper triangular elements of the function value matrix to obtain the compactness accumulation Ψ of the sub-region k : Here, the accumulation operation traverses all non-repeated point pairs in the sub-region.

[0107] a4, in order to avoid underflow or overflow in numerical calculation, the compactness accumulation is logarithmically transformed and the logarithmic equalization coefficient η is introduced k :

[0108] The clustering compactness coefficient CCC is defined as: CCC = -η k ln(Ψ k ).

[0109] The cluster compactness coefficient is used to indicate the compactness of data points within a sub-region. By calculating the distance between all pairs of data points within the sub-region and their nonlinear mapping, it reflects the similarity and consistency of environmental parameters within the sub-region. The smaller the cluster compactness coefficient value, the more compact the distribution of data points within the sub-region, the more consistent the environmental parameters, the smaller the environmental changes within the region, and the stronger the regional independence; the larger the value, the more dispersed the distribution of data points within the sub-region, the greater the difference in environmental parameters, the more environmental fluctuations within the region, and the possibility of strong regional interaction or being affected by the external environment, and the weaker independence. Therefore, the cluster compactness coefficient is used to evaluate the consistency and compactness of the internal environment of the sub-region, so as to judge its environmental independence.

[0110] Processing process of regional interaction effect index:

[0111] b1, for each sub-region C k and the adjacent sub-region C l First, we need to determine the boundary points of the two regions. Select each sub-region C k The data point b closest to the boundary k and sub-region C l The boundary data point b in l , forming a boundary point set B kl , contains all data points that interact with adjacent regions on the sub-region boundary. It is defined as follows: B kl ={(b k ,b l )|||b k -b l ||<δ}; where b k ∈C k and b l ∈C l , belong to different sub-regions respectively, and δ is a distance threshold used to identify whether two points are located in the boundary interaction region.

[0112] b2, for each pair of boundary points (b k ,b l ), calculate the gradient difference between the two in each environmental parameter, and form the gradient vector G kl , whose components are the differences of various environmental parameters, expressed as:

[0113] in:

[0114] P m (b k ) represents the boundary point b k The value of the mth environmental parameter at .

[0115] Yes k and b l The distance between.

[0116] Gradient vector G kl It reflects the rate of change of each environmental parameter at the boundary between the sub-region and the adjacent sub-region.

[0117] b3, based on the gradient vector calculated in step b2, define the regional interaction strength function ψ(G kl ), which is used to quantify the environmental interaction intensity between sub-regions. It is expressed as:

[0118] This nonlinear function produces smaller values ​​when the environmental parameters differ greatly, indicating weaker interaction strength; it produces larger values ​​when the parameters differ less, indicating stronger interaction strength.

[0119] b4, calculate the interaction strength for all boundary point pairs and accumulate them to obtain the accumulated interaction effect between regions Φ kl :

[0120] b5, in order to avoid values ​​that are too large or too small, the accumulated interaction effect is logarithmically transformed and a normalization coefficient ζ is introduced kl :

[0121] The final regional interaction effect index IRII is expressed as: IRII = -ζ kl ln(Φ kl ).

[0122] The regional interaction effect index is used to represent and quantify the intensity of environmental interaction between adjacent sub-regions, specifically reflecting the degree of difference in environmental parameters between sub-region boundary points. The regional interaction effect index evaluates the independence between adjacent regions by calculating the gradient change of environmental parameters and the interaction intensity function. A larger value indicates that the environmental interaction between sub-regions is weaker, the boundary environment difference is larger, and the regions have stronger independence; a smaller value indicates that the interaction between regions is stronger, the environmental parameters change less at the boundary, there is a higher mutual influence between regions, and the independence is weaker. Therefore, this index is used to quantitatively evaluate whether the sub-regions have relatively independent environmental characteristics and provide a basis for environmental regulation strategies.

[0123] S3.2, the environmental independence index is calculated based on the comprehensive processing of the clustering compactness coefficient and the regional interaction effect index. For example, it can be calculated by the following calculation formula:

[0124] The environmental independence index is used to represent and evaluate the independence of sub-regions in environmental parameters, and comprehensively reflects the compactness of environmental data points within the sub-region and the intensity of interaction with adjacent regions. The environmental independence index can comprehensively evaluate the independence of sub-regions by measuring the environmental consistency within the region through the cluster compactness coefficient, and quantifying the environmental interaction between the sub-region and the adjacent regions in combination with the regional interaction effect index. The larger the value, the more consistent the environmental parameters within the sub-region, and the weaker the interaction with the adjacent regions, and the stronger the independence of the region; the smaller the value, the greater the environmental fluctuation within the sub-region, and the stronger the environmental interaction with the adjacent regions, and the weaker the regional independence. Therefore, the environmental independence index is used to determine whether the sub-region needs independent regulation, and provides a scientific basis for the formulation of environmental regulation strategies.

[0125] The rationality of selecting the cluster compactness coefficient and the regional interaction effect index to participate in the calculation of the environmental independence index lies in that these two indicators provide a complete evaluation basis from two key dimensions: the consistency of the internal environment of the region and the interaction intensity between regions. The cluster compactness coefficient can quantify the similarity and volatility of environmental parameters in the sub-region by calculating the compactness of the environmental data points within the region, reflecting the stability and consistency within the region. The tighter the data points within the region, the smaller the change of environmental parameters and the stronger the regional independence; on the contrary, if the internal data points are loosely distributed, it means that there are more environmental fluctuations in the region and the independence is weak. The regional interaction effect index reflects the mutual influence and interaction intensity between regions by analyzing the gradient changes of environmental parameters between sub-regions and adjacent regions. The interaction effect index can evaluate the environmental interaction between the boundaries of regions. The greater the gradient change, the weaker the interaction intensity and the more independent the sub-region. On the contrary, the interaction intensity is strong and the independence is poor. Combining the cluster compactness coefficient and the regional interaction effect index can comprehensively evaluate the environmental independence of the sub-region from both the internal and external perspectives of the region, taking into account both the stability of the internal environment and the mutual influence between regions, providing a more comprehensive and scientific quantitative basis for the environmental independence index. Therefore, selecting these two indicators for calculation can maximize the capture of the independence characteristics of the sub-region and ensure that the regional regulation strategy is based on sufficient and accurate environmental data.

[0126] When the environmental independence index is greater than or equal to the independent distinction threshold, it means that the environmental characteristics of the sub-region are sufficiently independent from the adjacent regions, the consistency of the environmental parameters within the region is high, and the interaction with the adjacent regions is weak, which is suitable for independent regulation and generation of independent regulation signals. On the contrary, when the environmental independence index is less than the independent distinction threshold, it means that there is a strong environmental interaction between the sub-region and the adjacent regions, the environmental parameters within the region fluctuate greatly, and the independence is insufficient. It is suitable for coordinated regulation with the adjacent regions to generate coordinated regulation signals. This judgment provides a basis for the regulation strategy, ensuring that the regulation effect is more accurate and efficient.

[0127] After completing the environmental independence assessment, for those sub-regions that are judged to be insufficiently independent, it is necessary to further optimize the control strategy by in-depth analysis of the environmental interactions between regions. By constructing an environmental interaction graph based on graph theory, analyzing the airflow, heat conduction and other paths at the sub-region boundaries, calculating the interaction intensity, and identifying the paths with the strongest interactions between regions through the shortest path algorithm. This step aims to accurately identify the main environmental interaction paths between sub-regions, and dynamically correct the regional divisions based on the interaction intensity, providing a basis for subsequent coordinated control and ensuring the effectiveness and rationality of the control.

[0128] Step S4 includes the following contents:

[0129] S4.1, for the sub-regions that obtain the coordinated control signal, the boundary points between the corresponding sub-regions are used as nodes, and based on the environmental parameters such as airflow, temperature and heat conduction between the regions, an environmental interaction graph G = (V, E) is constructed. Among them, the node set V represents the boundary point set B kl , the edge set E represents the interaction path of environmental parameters between sub-areas. By analyzing parameters such as airflow velocity, temperature gradient, and thermal conductivity, the environmental interaction intensity between each pair of boundary points is calculated and used as the edge weight. The edge weight function is defined as w(b k ,b l ), a nonlinear weight model is established according to the differences in various environmental parameters (such as airflow and temperature change rate) to indicate the difficulty of interaction between sub-areas.

[0130] S4.2, for each pair of sub-regions, calculate the weights of the airflow path and the heat conduction path between the regions respectively. The weight of the airflow path is w flow (b k ,b l ) is calculated based on the difference in air flow velocity, and the weight of the heat conduction path w heat (b k ,b l ) is calculated based on the temperature change rate and thermal conductivity. The formula for the airflow path weight is: Where v(b k ) and v(b l) represent the air flow velocity at the boundary point. The weight formula of the heat conduction path is: Where κ is the thermal conductivity, T(b k ) and T(b l ) is the temperature value at the boundary point. The two weights reflect the difficulty of air flow transmission and heat exchange between sub-areas, respectively, providing a basis for subsequent path analysis.

[0131] S4.3, use the shortest path algorithm (such as Dijkstra algorithm or Floyd-Warshall algorithm) to search for the shortest path on the airflow path and the heat conduction path on the environmental interaction graph. The selection of the shortest path is based on the path weight calculated above, and the path length is the difficulty of the interaction of environmental parameters between sub-areas. The shortest path represents the boundary point pair with the strongest environmental interaction between sub-areas. By identifying the shortest path, the sub-area C is determined. k and C l The main interaction paths between them are then used to identify the region pairs with strong interactions.

[0132] S4.4, after identifying the main interaction paths of airflow and heat conduction, the initial sub-region division is revised according to the areas with strong environmental interactions. If the interaction intensity of the shortest path exceeds the preset threshold, it means that there is significant environmental interaction between sub-regions, and the regional boundaries need to be redefined, and the areas with strong interactions need to be merged to form a new regional division to ensure that environmental interference and resource waste between sub-regions are reduced during environmental control.

[0133] The present invention dynamically divides the environmental data in the building through an adaptive clustering algorithm, realizes the accurate delineation of sub-regions, and continuously optimizes the regional division results according to the density and distribution of environmental parameters, ensuring that the environmental characteristics of the sub-regions are accurately identified. On this basis, by calculating the compactness within the sub-region and the difference in environmental parameters with adjacent regions, the environmental independence index is obtained, which can accurately evaluate the independence or interaction intensity of each region, thereby providing a reliable basis for the selection of regional control mode. This method of combining regional division with independence evaluation not only improves the ability to fine-tune the control of different regions in a complex building environment, but also makes more reasonable collaborative control decisions when there is significant interaction between regions, ensuring that the system can quickly respond to environmental changes, optimize resource utilization efficiency, and ensure the accuracy and sustainability of the control effect.

[0134] After completing the correction of regional division and environmental independence assessment, it is necessary to formulate adaptive control strategies for each sub-region. In order to ensure that the control strategy can dynamically adapt to the current environmental status, a control strategy generation mechanism based on rule engine and decision tree inference algorithm is adopted to automatically generate the optimal control plan by analyzing the environmental data of the sub-region and combining with the predefined rule base. The goal of this step is to select the optimal control action based on the independence and interaction strength of the current sub-region, and convert it into specific device instructions to ensure the real-time and accuracy of environmental control.

[0135] Step S5 includes the following contents:

[0136] S5.1, based on the revised regional division results of step S4 and the environmental independence evaluation of step S3, the input data for generating the control strategy is initialized. The boundary information, environmental independence index and regional interaction of each sub-region are imported into the control system. The input data provides a basis for rule generation and ensures the accuracy and real-time performance of the control strategy.

[0137] S5.2, build a control rule base based on the environmental characteristics and historical environmental data inside the building. The rule base contains predefined control actions under various environmental conditions, such as temperature adjustment, airflow control, humidity management, etc. The rule base defines the control behavior under different environmental conditions in the form of logical rules, such as "when the temperature exceeds the set threshold and the airflow is insufficient, start the cooling device", etc., covering different combinations of environmental parameters.

[0138] S5.3, uses a rule-based reasoning algorithm based on a decision tree to generate an adaptive control strategy. By analyzing the environmental status of the current sub-area and matching it with the conditions in the rule base, the decision tree structure performs branch judgments from top to bottom and automatically generates a control strategy. Each node represents an environmental parameter or device state. When the current parameter meets the node condition, the system automatically selects the optimal control action based on the branch path. The specific form is that the root node branches for the first time based on key environmental parameters (such as temperature or humidity), and the internal nodes make the next judgment based on more detailed state parameters (such as airflow velocity, heat conduction) until the leaf node forms a specific control action.

[0139] S5.4, at the end of the decision tree reasoning, each leaf node corresponds to one or more control actions. The system automatically selects the optimal control action based on the environmental parameters of the current sub-region. The selection of the optimal control action is based on a comprehensive judgment of the environmental independence index and the regional interaction intensity. For example, for regions with strong environmental independence, the system gives priority to independent control schemes; while for regions with strong interactions, the system chooses collaborative control schemes.

[0140] S5.5, the generated control actions will be automatically converted into device instructions by the system and sent to related devices (such as air conditioners, ventilation systems, humidifiers, etc.) through the Internet of Things protocol. The system monitors the execution of the equipment in real time, and adjusts the control strategy in time according to the operating status data fed back by the equipment to ensure that the environmental parameters are maintained within the set comfort range. The equipment instructions include startup, adjustment parameters, etc., to ensure that the environmental control of each sub-area can dynamically match the current environmental status.

[0141] In the aforementioned control strategy generation and execution process, in order to further improve the intelligent level of environmental control, it is necessary to continuously optimize the control effect through self-learning algorithms. By introducing the Q learning algorithm in reinforcement learning, the system can autonomously adjust and improve the control strategy in a constantly changing environment. By defining the environmental state space and the set of control actions, and combining the reward function to provide feedback on the effect of each control action, the system optimizes the control strategy according to the Q value update mechanism, and gradually learns and selects the optimal control action. This process ensures that the control system can dynamically adapt to complex environmental changes and improve the efficiency and accuracy of the overall control.

[0142] Step S6 includes the following contents:

[0143] S6.1, for each sub-region, define the state space S in reinforcement learning to describe the current state of environmental control. The dimensions of the state space are composed of key environmental parameters in the sub-region, such as temperature, humidity, air flow velocity, air quality, etc. These parameters represent the environmental state of the current region, and each state point s∈S is a specific combination of these environmental parameters. By reasonably defining the state space, ensure that all possible environmental conditions are covered so that the reinforcement learning algorithm can fully explore different control effects.

[0144] S6.2, based on the state space S, define an action set A for each sub-area. Each action in the action set corresponds to a specific control behavior, such as turning on the air conditioner, adjusting the humidity, changing the airflow direction, etc. The dimension of the action set is related to the device type. The action set of each sub-area may contain multiple control dimensions. For example, temperature control may include different cooling or heating levels, and humidity control may include multiple humidity adjustment actions. The choice of each action a∈A will directly affect the change of the current environmental state.

[0145] S6.3, according to the control effect, set the reward function R(s,a) to measure the effect of each action under different environmental conditions. The reward function is designed according to the adjustment range of the environmental parameters and the achievement of the control target. For example, if the control action makes the environmental parameters tend to the target value, the reward value is high; if the action causes the environmental parameters to deviate from the target value, the reward value is low or negative. The specific reward function form is: Among them, Po (s′) represents the environmental parameter value in the new state s′ after action a is executed in the current state s, is the target environmental parameter value, ΔP o is the allowed parameter deviation. The reward function reflects the degree of deviation between the control action and the target state; o represents the index of each specific environmental parameter.

[0146] S6.4, after each action is executed, the Q value is updated using the Q value update formula of the Q learning algorithm. The Q value Q(s,a) represents the long-term expected return of selecting action a in state s, and the update formula is: Among them, ρ is the learning rate, which controls the weight of new information at each update; γ is the discount factor, which weighs the relationship between future rewards and current rewards; is the maximum Q value in the next state s′. Through the iterative update of Q value, the reinforcement learning algorithm can continuously optimize the control strategy and maximize the expected return.

[0147] S6.5, after each state transition, select the action a with the largest Q value according to the current state s * , action a * The corresponding Q value Q(s,a * ) is the largest, indicating that the action has the best expected return in the current environment. The selection formula is: After selecting the action, the control action will be executed and enter the new state s'. This process ensures that the environmental parameters of each sub-area gradually approach the target value and optimizes the control effect by continuously selecting the optimal action.

[0148] The present invention generates an adaptive control strategy through an inference algorithm based on a rule engine and a decision tree, and can automatically select the optimal control action under different environmental conditions to ensure that the control system responds quickly to real-time environmental changes. At the same time, combined with the Q learning algorithm in reinforcement learning, the control strategy is continuously optimized and the Q value is continuously updated, so that the system can learn from the actual control effect and gradually improve the control accuracy. This method can not only generate a reasonable control scheme under the initial environmental state, but also dynamically adjust the control action through a self-learning mechanism to achieve intelligent optimization in long-term operation, ensuring that in a complex and changeable building environment, the control system can automatically adapt to environmental changes, improve the efficiency and stability of the control, and achieve efficient and intelligent dynamic environmental control.

[0149] Example 2

[0150] Figure 2 The invention provides a building environment dynamic control system based on the Internet of Things drive, comprising: an environment data acquisition module, an environment division module, an environment assessment module, an environment correction module, a control strategy generation module and a control optimization module.

[0151] The environmental data acquisition module deploys multiple types of IoT sensors in each sub-area of ​​the building, and uses a dynamic time series data fusion algorithm to synchronize and spatially fuse the collected environmental data to generate an indoor environmental data set; the data output is passed to the environmental division module.

[0152] The environmental division module is based on environmental data and uses the K-Means clustering algorithm to initially select cluster centers according to the density and distribution of environmental parameters. The number of clusters is dynamically adjusted by calculating the variance and silhouette coefficient, and the location of the cluster centers is iteratively updated until the clustering results converge, completing the adaptive division of sub-areas. The division results are passed to the environmental assessment module.

[0153] The environmental assessment module calculates the compactness of environmental data points within the sub-region and the differences in environmental parameters between adjacent regions, and combines the weighted processing of internal consistency and boundary interaction strength to derive the environmental independence index. It also determines whether the sub-region generates independent regulation signals or collaborative regulation signals based on the preset independent distinction threshold; the assessment results are transmitted to the environmental correction module.

[0154] If the environmental correction module obtains a collaborative control signal, it will construct an environmental interaction graph at the boundary of the sub-region through a path analysis algorithm based on graph theory, calculate the weights of the airflow and heat conduction paths between regions, use the shortest path algorithm to identify pairs of regions with strong environmental interactions, and correct the regional division according to the environmental interaction situation; the corrected division results are passed to the control strategy generation module.

[0155] Based on the revised regional division results and environmental independence assessment, the control strategy generation module uses a rule engine and a decision tree-based rule inference algorithm to generate an adaptive control strategy, build a control rule library, and use decision trees to automatically generate control rules. It selects the optimal control action according to the current environmental status and sends instructions to related equipment; the generated control strategy is passed to the control optimization module.

[0156] The control optimization module continuously optimizes environmental changes and control effects through the Q-learning algorithm in reinforcement learning, defines the state space and action set, sets the reward function according to the control effect, updates the Q value according to the Q-learning update formula, and selects the action with the largest Q value as the current control strategy; the optimized strategy is finally transmitted to the execution device to realize the control of the building environment.

[0157] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0158] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0159] It should be noted that, in this article, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including a..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0160] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for dynamic control of building environment driven by the Internet of Things, characterized in that: Includes steps: S1, deploys multiple types of IoT sensors in each sub-area of ​​the building, synchronizes the collected environmental data in time and space through a dynamic time series data fusion algorithm to generate an indoor environmental data set; S2, based on environmental data, using the K-Means clustering algorithm, initially selects cluster centers according to the density and distribution of environmental parameters, dynamically adjusts the number of clusters by calculating the variance and silhouette coefficient, and iteratively updates the location of cluster centers until the clustering results converge, completing the adaptive division of sub-areas; S3, by calculating the compactness of the environmental data points within the sub-region and the difference in environmental parameters between adjacent regions, combined with the weighted processing of internal consistency and boundary interaction strength, the environmental independence index is obtained, and the pre-set independent distinction threshold is used to determine whether the sub-region generates an independent regulation signal or a collaborative regulation signal; S4, if the coordinated control signal is obtained, then at the boundary of the sub-region, an environmental interaction graph is constructed through a path analysis algorithm based on graph theory, the weights of the airflow and heat conduction paths between regions are calculated, the shortest path algorithm is used to identify regional pairs with strong environmental interactions, and the regional division is modified according to the environmental interaction situation; S5, based on the revised regional division results and environmental independence assessment, the rule engine and the rule reasoning algorithm based on the decision tree are used to generate the adaptive control strategy, build the control rule library, use the decision tree to automatically generate the control rules, select the optimal control action according to the current environmental status, and send instructions to the relevant equipment; S6 , through the Q learning algorithm in reinforcement learning, the environmental changes and control effects are continuously optimized, the state space and action set are defined, the reward function is set according to the control effect, the Q value is updated according to the Q learning update formula, and the action with the largest Q value is selected as the current control strategy.

2. The method for dynamic control of building environment based on Internet of Things drive according to claim 1 is characterized in that: Step S1 includes the following contents: S1.1, evenly deploy temperature and humidity sensors, air quality sensors, light sensors, and micro airflow sensors in each sub-area of ​​the building to form a multi-type sensor network; S1.2, the data collected by each sensor node is time synchronized according to its sampling timestamp; S1.3, based on the geographical location of the sensors and the spatial layout information of the sub-regions, the multi-type environmental data in each sub-region are fused into a unified environmental data set through a spatial data fusion algorithm; S1.4, perform data cleaning on the initially fused environmental data set, detect and exclude abnormal data; S1.5, after time synchronization, spatial fusion and data cleaning, the processed data set is input into the data management module to finally generate the indoor environment data set.

3. The method for dynamic control of building environment based on Internet of Things according to claim 2 is characterized in that: Step S2 includes the following contents: S2.1, based on the high-precision indoor environment dataset generated in step S1, extract key environmental parameters and construct the environmental parameters of each sub-area into a multi-dimensional vector; S2.2, based on the environmental data after dimensionality reduction, the K-Means clustering algorithm is used to select the initial cluster center; S2.3, during the clustering iteration process, the cluster center is dynamically adjusted based on the variance and silhouette coefficient of each clustering result; S2.4, automatically adjusts the number of clusters by calculating the variance and silhouette coefficient under different numbers of clusters; S2.5, after each iteration, the centroid position of each cluster is recalculated based on the current clustering result, and the distance between the data of each sub-region and the centroid is minimized to ensure that the environmental parameters in the sub-region are similar and meet the standards; S2.6, after the clustering results converge, the adaptive division of each sub-region is finally completed.

4. The method for dynamic control of building environment based on Internet of Things according to claim 3 is characterized in that: Step S3 includes the following contents: S3.1, after completing the preliminary division of sub-regions, calculate the compactness of environmental data points within the sub-regions, and use the clustering compactness coefficient to quantify the environmental consistency within the sub-regions; at the same time, calculate the regional interaction effect index, and evaluate the intensity of environmental interaction between regions by analyzing the gradient changes of environmental parameters between sub-regions and adjacent regions; The processing process of clustering compactness coefficient: a1, for each sub-region C k , including N k data points, calculate the distances between all pairs of points in the sub-region, and construct the distance matrix D k ; The matrix elements are defined as: d ij =||x i -x j ||; where x i and x j Represents sub-region C k The position vectors of the i-th and j-th data points in ; a2, introduce the compactness mapping function φ(d ij ), defined as follows: Among them: λ>0 is the attenuation coefficient, which controls the rate at which the function decreases; α>1 is the nonlinear coefficient, which adjusts the nonlinearity of the function; the compactness function values ​​of all point pairs are calculated to form a function value matrix; a3, multiply the upper triangular elements of the function value matrix to obtain the compactness accumulation Ψ of the sub-region k : Here, the cumulative multiplication operation traverses all non-repeated point pairs in the sub-region; a4, in order to avoid underflow or overflow in numerical calculation, the compactness accumulation is logarithmically transformed and the logarithmic equalization coefficient η is introduced k : The clustering compactness coefficient CCC is defined as: CCC = -η k ln(Ψ k ).

5. The method for dynamic control of building environment based on Internet of Things according to claim 4 is characterized in that: Step S3 also includes the following: Processing process of regional interaction effect index: b1, for each sub-region C k and the adjacent sub-region C l First, we need to determine the boundary point sets of the two regions; select each sub-region C k The data point b closest to the boundary k and sub-region C l The boundary data point b in l , forming a boundary point set B kl , contains all data points that interact with adjacent regions on the sub-region boundary, and is defined as follows: kl ={(b k ,b l )|||b k -B l ||<δ}; where b k ∈C k and b l ∈C l , belong to different sub-regions respectively, and δ is a distance threshold used to identify whether two points are located in the boundary interaction region; b2, for each pair of boundary points (b k ,b l ), calculate the gradient difference between the two in each environmental parameter, and form the gradient vector G kl , whose components are the differences of various environmental parameters, expressed as: Where: P m (b k ) represents the boundary point b k The value of the mth environmental parameter at ; Yes k and b l The distance between b3, based on the gradient vector calculated in step b2, define the regional interaction strength function ψ(G kl ), which is used to quantify the environmental interaction intensity between sub-regions and is expressed as: b4, calculate the interaction strength for all boundary point pairs and accumulate them to obtain the accumulated interaction effect between regions Φ kl : b5, logarithm transformation of the interaction effect accumulation and introduction of normalization coefficient ζ kl : The final regional interaction effect index IRII is expressed as: IRII = -ζ kl ln(Φ kl ); S3.2, the environmental independence index is calculated based on the comprehensive processing of the clustering compactness coefficient and the regional interaction effect index; when the environmental independence index is greater than or equal to the independent distinction threshold, it means that the environmental characteristics of the relevant sub-region are sufficiently independent from the adjacent regions, generating an independent regulation signal; conversely, when the environmental independence index is less than the independent distinction threshold, it means that there is a strong environmental interaction between the relevant sub-region and the adjacent regions, and the independence is insufficient, generating a coordinated regulation signal.

6. The method for dynamic control of building environment based on Internet of Things according to claim 5 is characterized in that: Step S4 includes the following contents: S4.1, for the sub-regions that obtain the synergistic regulation signal, the boundary points between the corresponding sub-regions are used as nodes, and based on the environmental parameters, an environmental interaction graph G = (V, E) is constructed; where the node set V represents the boundary point set B kl , the edge set E represents the interaction path of environmental parameters between sub-areas; by analyzing the airflow velocity, temperature gradient, and heat conductivity, the environmental interaction intensity between each pair of boundary points is calculated and used as the edge weight; the edge weight function is defined as w(b k ,b l ), a nonlinear weight model is established according to the differences in various environmental parameters to indicate the difficulty of interaction between sub-regions; S4.2, for each pair of sub-regions, calculate the weights of the airflow path and the heat conduction path between the regions respectively. The weight of the airflow path is w flow (b k ,b l ) is calculated based on the difference in air flow velocity, and the weight of the heat conduction path w heat (b k ,b l ) is calculated based on the temperature change rate and the thermal conductivity coefficient; the formula for the airflow path weight is: Where v(b k ) and v(b l ) represent the air flow velocity at the boundary points respectively; the weight formula of the heat conduction path is: Where κ is the thermal conductivity, T(b k ) and T(b l ) is the temperature value at the boundary point; S4.3, use the shortest path algorithm to search for paths on the environmental interaction graph to find the shortest paths on the airflow path and the heat conduction path; the shortest path represents the boundary point pair with the strongest environmental interaction between sub-areas; by identifying the shortest path, determine the sub-area C k and C l The main interaction paths between them are used to identify pairs of regions with strong interactions. S4.4, after identifying the main interaction paths of airflow and heat conduction, the initial sub-region division is modified according to the areas with strong environmental interactions; if the interaction intensity of the shortest path exceeds the preset threshold, it means that there is significant environmental interaction between the sub-regions, and the regional boundaries need to be redefined, and the areas with strong interactions need to be merged to form a new regional division.

7. The method for dynamic control of building environment based on Internet of Things according to claim 6 is characterized in that: Step S5 Includes the following: S5.1, based on the revised regional division result in step S4 and the environmental independence evaluation in step S3, initializing the input data generated by the control strategy; S5.2, based on the environmental characteristics of the building and historical environmental data, a control rule base is constructed; the rule base contains predefined control actions under various environmental states; the rule base defines the control behaviors under different environmental states in the form of logical rules, covering different combinations of environmental parameters; S5.3, using a decision tree-based rule inference algorithm to generate an adaptive control strategy; S5.4, at the end of the decision tree reasoning, each leaf node corresponds to one or more regulatory actions; S5.5, the generated control actions will be automatically converted into device instructions by the system and sent to relevant devices through the Internet of Things protocol.

8. The method for dynamic control of building environment based on Internet of Things according to claim 7 is characterized in that: Step S6 includes the following contents: S6.1, for each sub-region, define the state space S in reinforcement learning to describe the current state of environmental regulation; S6.2, based on the state space S, define an action set A for each sub-area, where each action in the action set corresponds to a specific control behavior; S6.3, according to the control effect, the reward function R(s,a) is set to measure the effect of each action under different environmental conditions; the reward function is designed according to the adjustment range of environmental parameters and the achievement of control goals; the specific reward function form is: Among them, P o (s′) represents the environmental parameter value in the new state s′ after action a is executed in the current state s, is the target environmental parameter value, ΔP o is the allowable parameter deviation; o represents the index of each specific environmental parameter; S6.4, after each action is executed, the Q value is updated using the Q value update formula of the Q learning algorithm; the Q value Q(s,a) represents the long-term expected return of selecting action a in state s, and the update formula is: Among them, ρ is the learning rate, which controls the weight of new information at each update; γ is the discount factor, which weighs the relationship between future rewards and current rewards; is the maximum Q value in the next state s′; S6.5, after each state transition, select the action a with the largest Q value according to the current state s * , action a * The corresponding Q value Q(s,a * ) is the largest, indicating that the relevant action has the best expected return in the current environment; the selection formula is: After selecting the relevant action, the corresponding control action will be executed and enter the new state s′.

9. A building environment dynamic control system driven by the Internet of Things, used to implement the building environment dynamic control method driven by the Internet of Things according to any one of claims 1 to 8, characterized in that: include: Environmental data acquisition module, environmental division module, environmental assessment module, environmental correction module, control strategy generation module and control optimization module; The environmental data acquisition module deploys multiple types of IoT sensors in each sub-area of ​​the building, and uses a dynamic time series data fusion algorithm to synchronize and spatially fuse the collected environmental data to generate an indoor environmental data set. The data output is transmitted to the environmental division module. The environment division module uses the K-Means clustering algorithm based on the environment data to initially select cluster centers according to the density and distribution of environmental parameters, dynamically adjust the number of clusters by calculating the variance and silhouette coefficient, and iteratively update the location of the cluster centers until the clustering results converge, completing the adaptive division of sub-areas. The division results are transferred to the environmental assessment module; The environmental assessment module calculates the compactness of the environmental data points within the sub-region and the difference in environmental parameters between adjacent regions, and combines the weighted processing of internal consistency and boundary interaction strength to obtain the environmental independence index, and determines whether the sub-region generates independent regulation signals or collaborative regulation signals based on the preset independent distinction threshold; the assessment results are transmitted to the environmental correction module; If the environmental correction module obtains a collaborative control signal, it will construct an environmental interaction graph at the sub-region boundary through a path analysis algorithm based on graph theory, calculate the weights of the airflow and heat conduction paths between regions, use the shortest path algorithm to identify regional pairs with strong environmental interactions, and correct the regional division according to the environmental interaction situation; the corrected division result is transmitted to the control strategy generation module; The control strategy generation module generates adaptive control strategies based on the revised regional division results and environmental independence assessment, adopts the rule engine and the rule reasoning algorithm based on the decision tree, builds a control rule library, automatically generates control rules using the decision tree, selects the optimal control action according to the current environmental status, and sends instructions to related devices; The generated control strategy is passed to the control optimization module; The control optimization module continuously optimizes environmental changes and control effects through the Q learning algorithm in reinforcement learning, defines the state space and action set, sets the reward function according to the control effect, updates the Q value according to the Q learning update formula, and selects the action with the largest Q value as the current control strategy; The optimized strategy is ultimately transmitted to the execution device to achieve regulation of the building environment.

Images