Method and system capable of remotely controlling window in partition mode

Abstract

The invention discloses a method and a system capable of remotely controlling a window in a zoning manner, particularly relates to the field of door and window control, is used for solving the problems of frequent invalid operation, resource waste and low efficiency in window zoning control, and realizes the purpose of controlling the window in a zoning manner by identifying and optimizing window zoning control operation under dynamically changing use requirements. Intelligent management of window opening and closing operation is achieved; firstly, operation behaviors of a user are comprehensively monitored and captured through real-time data collection and analysis, frequent and potential invalid operations are recognized, and resource waste caused by excessive intervention or repeated adjustment is avoided; secondly, quantifying the specific influence of each window operation on the indoor environment and energy consumption in combination with a benefit evaluation model, and identifying and removing low-efficiency or invalid operations in combination with an operation filtering mechanism to ensure that the window control system is always in a high-efficiency running state; finally, the balance between resource optimization and user comfort is realized, and the intelligent degree and sustainability of window control are improved.

Description

Technical Field

[0001] The present invention relates to the field of door and window control, and more particularly to a method and system for remotely controlling windows capable of being partitioned. Background Art

[0002] In the intelligent management of modern buildings, window control systems, as important environmental regulation equipment, are increasingly being used to improve indoor comfort and energy efficiency. With the advancement of intelligent technology, partitioned window control systems have been able to automatically open and close local windows according to user needs, helping to optimize indoor temperature, air circulation and lighting conditions. However, although these systems provide a high degree of flexibility and convenience, in actual use, due to frequent or unreasonable user operations, they often lead to excessive adjustment of window status, which in turn wastes energy, increases system burden and reduces user experience. In particular, when the strategy is not adjusted accurately, the system may over-respond and perform a series of redundant operations, which not only reduces control efficiency, but also poses a challenge to the effective use of energy.

[0003] Most current smart window control systems rely on direct user intervention or preset automatic adjustment rules. However, due to the operating habits, environmental requirements and diversity of window control methods of different users, frequent ineffective adjustments have become a common problem. For example, users may frequently adjust windows when they perceive slight changes in the indoor environment, but these adjustments often do not significantly improve the environmental quality, and may even cause excessive switching of window states, resulting in indoor temperature fluctuations and increased energy consumption. How to identify these unnecessary operations through intelligent data analysis technology and effectively reduce unnecessary adjustments has become the key to improving the efficiency of smart window systems and user satisfaction. The system needs to analyze the user's historical operation data to determine which adjustment behaviors are effective and which are just overreactions, so as to automatically optimize the response strategy, avoid excessive adjustments, reduce resource waste, and improve the system's intelligence level and user experience.

[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 remotely controlling partitioned windows, which realizes the intelligent management of window switch operations under dynamically changing usage requirements by identifying and optimizing window partition control operations; firstly, through real-time data collection and analysis, the user's operating behavior is comprehensively monitored and captured, and frequent and potentially invalid operations are identified to avoid waste of resources caused by excessive intervention or repeated adjustments; secondly, combined with the benefit evaluation model, the specific impact of each window operation on the indoor environment and energy consumption is quantified, and then combined with the operation filtering mechanism, inefficient or invalid operations are identified and removed to ensure that the window control system is always in an efficient operating state; finally, a balance is achieved between resource optimization and user comfort, and the intelligence and sustainability of window control are improved to solve the problems raised in the above-mentioned background technology.

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

[0007] A method for remotely controlling a partitioned window, characterized by comprising the steps of:

[0008] S1. Collect the operation data of window partitions in real time through IoT sensors;

[0009] S2. Extract spatiotemporal features from the collected data, identify the user's operation mode through unsupervised learning, and classify the behavioral characteristics of frequent operations and potential invalid operations;

[0010] S3. Through multivariate regression analysis and causal inference models, quantify the specific impact of each window operation on energy consumption and indoor comfort, and generate benefit evaluation results;

[0011] S4. Based on the benefit evaluation results, use rules and anomaly detection to identify operations that do not meet the contribution standards and filter out unnecessary window adjustments.

[0012] In a preferred embodiment, step S1 includes the following contents:

[0013] Independent sensor nodes are deployed in each window partition, and each sensor node is used to pre-process data, and the processed data is transmitted to the central data acquisition system.

[0014] In a preferred embodiment, step S2 includes the following contents:

[0015] Spatiotemporal feature extraction: First, the time series data is subjected to feature extraction through a convolutional neural network to extract local patterns and structural features in the local time series. The data in each time window is subjected to a convolution operation through a sliding window to extract local operation pattern features.

[0016] After the spatiotemporal features are extracted, these features are used to identify the categories of different user operation modes. An autoencoder is used to reduce the dimension of the data and detect abnormal patterns. The autoencoder learns the low-dimensional representation of the data and maps the high-dimensional operation features into a low-dimensional latent space. Different areas of the latent space correspond to different modes of user operation.

[0017] Clustering and unsupervised learning: In the output latent space of the autoencoder, the K-means clustering algorithm is used to cluster data and identify multiple operation categories, including frequent operations and potentially invalid operations. The K-means algorithm maps spatiotemporal features to different cluster centers, uses Euclidean distance to measure the similarity between sample points and cluster centers, and marks each sample to which category based on the clustering results.

[0018] Classification of frequent operations and potentially invalid operations: After obtaining the operation categories through K-means clustering, the propagation and concentration of each operation category are calculated to distinguish between frequent operations and potentially invalid operations.

[0019] The following is a specific process for distinguishing frequent operations from potentially invalid operations:

[0020] The spread degree measures the breadth of influence of a certain operation category in the system, reflecting the diffusion effect of its operation mode. It is defined by calculating the degree of change of each operation category in multiple time windows and the number of window areas involved. The spread degree calculation formula is: Among them, P c is the propagation degree of operation category c; T is the time period of operation data; W is the number of window areas involved in the operation; x t,w is the operation intensity of the window area w at time point t; Δx t is the change in operation at time point t, calculated as: reflects the magnitude of change in the operation at each point in time; is the total change at all time points; It represents the sum of the operation intensity of all window areas at a certain time t, which measures the scope of operation activities at the corresponding time;

[0021] The concentration measures the local concentration of the operation category and reflects the concentration of the operation behavior. The concentration calculation formula is: Among them, C c is the concentration of operation category c.

[0022] The operation index I is obtained by combining the calculation results of the spread and concentration and the weights of the two. c , the operation index is used to distinguish frequent operations from potential invalid operations; the operation index calculation formula is: in, is the historical average operation intensity of the window area w, and the deviation of the historical operation intensity Used to penalize overly volatile operations;

[0023] According to the calculated operation index, it is compared with threshold 1 and threshold 2:

[0024] Frequent operation: If the operation index is greater than the threshold 1, it is determined to be a frequent operation;

[0025] Potentially invalid operation: If the operation index is less than the threshold 2, it is a potentially invalid operation and is determined to be a potentially invalid operation.

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

[0027] S3.1, multiple linear regression is used to analyze the relationship between frequent operations and potential ineffective operations and energy consumption and comfort, and the settings are as follows: E i =β1O i +β2C i +β3T i +β4H i +∈ i ; Among them, E i is the energy consumption of the ith operation, O i is the opening and closing state of the window operation, including frequent operations and potential invalid operations, C i Score the current indoor comfort level, T i is the external temperature, H i is the indoor humidity, ∈ i is the noise term, β1, β2, β3, β4 are regression coefficients;

[0028] S3.2, verify the causal relationship of each operation on energy and comfort through the causal inference model; use the following causal inference formula to calculate the direct and indirect impact of the operation behavior on the final benefit: Among them, P(O i ) is the operation behavior O i The probability of occurrence, ΔE(O i ) is the impact of the corresponding operation on energy consumption;

[0029] S3.3, based on the results of regression analysis and causal inference, generate a benefit evaluation for each operation; specifically, evaluate the difference in benefits between each frequent operation and potential ineffective operation in terms of energy consumption and comfort; calculate the operational benefit index OEI: Among them, β i are the regression coefficients determined for each operation in the regression analysis.

[0030] In a preferred embodiment, step S4 includes the following contents:

[0031] According to the operation benefit index value, the operation categories are initially divided into efficient operations and inefficient operations. If the operation benefit index is greater than the corresponding threshold, it is marked as an efficient operation, otherwise it is marked as an inefficient operation. Next, it is necessary to identify those inefficient operations that show abnormal characteristics in the operation behavior, and use the operation benefit index, propagation, concentration and time series characteristics of the operation as input features to calculate the abnormal score of each operation category. If the abnormal score is greater than the abnormal threshold, the corresponding operation category is marked as an abnormal operation and needs to be filtered.

[0032] A partitionable remote control window system, comprising: a data acquisition module, a pattern recognition module, a benefit evaluation module and an operation filtering module;

[0033] Data acquisition module: collects the operation data of window partitions in real time through IoT sensors and passes it to the pattern recognition module;

[0034] Pattern recognition module: extracts spatiotemporal features from the collected data, identifies the user's operation pattern through unsupervised learning, classifies the behavioral features of frequent operations and potentially invalid operations, and passes the classification results to the benefit evaluation module;

[0035] Benefit evaluation module: quantify the specific impact of each window operation on energy consumption and indoor comfort through multivariate regression analysis and causal inference model, generate benefit evaluation results, and pass the generated results to the operation filtering module;

[0036] Operation filtering module: Based on the benefit evaluation results, use rules and anomaly detection to identify operations that do not meet the contribution standards and filter out unnecessary window adjustments.

[0037] The technical effects and advantages of the method and system for remotely controlling windows capable of being partitioned are as follows:

[0038] The present invention realizes intelligent management of window switch operations under dynamically changing usage requirements by accurately identifying and optimizing window partition control operations. First, through real-time data collection and analysis, it can comprehensively monitor and capture user operation behaviors, identify frequent and potentially invalid operations, and avoid waste of resources caused by excessive intervention or repeated adjustments. Secondly, combined with a multi-dimensional benefit evaluation model, it can quantify the specific impact of each window operation on the indoor environment and energy consumption, thereby providing data support for subsequent control strategy optimization. In addition, the operation filtering mechanism, through intelligent detection and rule engines, identifies and removes inefficient or invalid operations to ensure that the window control system is always in an efficient operating state. Ultimately, the entire system achieves a balance between resource optimization and user comfort, and improves the intelligence and sustainability of window control. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic flow chart of a method for remotely controlling windows capable of being partitioned according to the present invention;

[0040] Figure 2 The present invention is a schematic structural diagram of a system for remotely controlling windows capable of being partitioned. DETAILED DESCRIPTION

[0041] 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.

[0042] Embodiment 1: Figure 1 The present invention provides a method for remotely controlling a partitioned window, comprising:

[0043] S1. Collect the operation data of window partitions in real time through IoT sensors;

[0044] S2. Extract spatiotemporal features from the collected data, identify the user's operation mode through unsupervised learning, and classify the behavioral characteristics of frequent operations and potential invalid operations;

[0045] S3. Through multivariate regression analysis and causal inference models, quantify the specific impact of each window operation on energy consumption and indoor comfort, and generate benefit evaluation results;

[0046] S4. Based on the benefit evaluation results, use rules and anomaly detection to identify operations that do not meet the contribution standards and filter out unnecessary window adjustments.

[0047] Step S1 includes the following contents:

[0048] To ensure high accuracy and comprehensiveness of data, a variety of IoT sensors are used to monitor window partitions in all directions. Position sensors provide real-time feedback on window status by monitoring the opening and closing angles and position changes of windows; force sensors measure the opening and closing force of windows and provide information on the precision of operation; temperature and humidity sensors are used to monitor changes in the indoor environment and assist in inferring the comfort effect of window operation. Independent sensor nodes are deployed in each window partition, which transmit data to the central data acquisition system through low-power wireless communication protocols (such as Zigbee, Z-Wave, and BluetoothLE).

[0049] Since the window control system involves multiple partitions, an efficient data transmission architecture is required to cope with a large amount of sensor data streams. In the IoT architecture, distributed edge computing technology is used to preprocess data using each sensor node or local gateway. This process includes steps such as denoising, timestamp synchronization, and local data aggregation. In particular, before transmission, the collected raw data is subjected to noise suppression based on weighted median filtering to ensure that abnormal data points do not interfere with subsequent analysis. The specific form of the weighted median filtering algorithm is: y(t) = median(x(tk), x(t-k+1), ..., x(t+k)); where x(t) is the sensor data at time point t and k is the window size. Through this method, high-frequency noise is effectively removed and the reliability of the data is enhanced.

[0050] When sensor data is transmitted through different communication protocols (Zigbee, Z-Wave, Bluetooth LE), adaptive protocol conversion technology between protocols is used to achieve seamless data connection and compatibility. Each sensor node pre-processes the collected raw data locally through edge computing, packages it according to the protocol specifications and sends it to the central server. In a multi-protocol environment, a unified data access layer (DAL) is used to standardize data of different protocols, convert it into a unified format and store it in the database.

[0051] Ultimately, the data after edge computing and preprocessing will be uniformly transmitted to the cloud system. In the data integration process, data from different sources (such as opening and closing status, force, and environmental data) are merged in time and space to form a unified operation data stream. These data streams not only support the current window status monitoring, but also provide reference for future control decisions.

[0052] Step S2 includes the following contents:

[0053] In the remote control system of window partitions, the core of identifying user operation patterns is to extract key features from a large amount of operation data and cluster and classify them through unsupervised learning algorithms. This not only involves simple recognition of operation behavior, but also requires deep learning processing of multi-dimensional data such as operation frequency, intensity and time changes to further identify the behavioral characteristics of frequent operations and potential invalid operations.

[0054] Spatiotemporal feature extraction: The window operation data obtained during the data collection phase includes multiple dimensions, such as operation time, operation frequency, force, window status, etc. These data have obvious spatiotemporal characteristics. Therefore, the time series data is firstly subjected to feature extraction through a convolutional neural network. Convolutional neural networks are particularly suitable for extracting local patterns and structural features in local time series. For example, the convolution kernel can perform convolution operations on the data in each time window through a sliding window to extract local operation pattern features.

[0055] After the spatiotemporal features are extracted, these features need to be used to identify the categories of different user operation modes. Autoencoders are used for data dimensionality reduction and abnormal pattern detection. Autoencoders can learn low-dimensional representations of data and map high-dimensional operation features into low-dimensional latent spaces. Different areas of the latent space correspond to different modes of user operation.

[0056] Clustering and unsupervised learning: In the output latent space of the autoencoder, the K-means clustering algorithm is used to cluster data and identify multiple operation categories, including frequent operations and potentially invalid operations. The K-means algorithm maps spatiotemporal features to different cluster centers, uses Euclidean distance to measure the similarity between sample points and cluster centers, and marks each sample to which category based on the clustering results. By adjusting the number of cluster centers k, the accuracy and range of classification can be flexibly adjusted.

[0057] Classification of frequent operations and potentially invalid operations: After obtaining the operation categories through K-means clustering, frequent operations and potentially invalid operations can be further distinguished by calculating the spread and concentration of each operation category. Frequent operations usually have a higher cluster concentration and a lower abnormality, while potentially invalid operations are characterized by low frequency, highly unstable or large-scale operations.

[0058] The following is a specific process for distinguishing frequent operations from potentially invalid operations:

[0059] The spread degree measures the breadth of an operation category's influence in the system, reflecting the diffusion effect of its operation mode. It can be defined by calculating the degree of change of each operation category in multiple time windows and the number of window areas involved. The higher the spread degree, the wider the influence of the operation category in the spatial and temporal dimensions, which may be related to frequent operations.

[0060] The formula for calculating the spread is:

[0061] in:

[0062] P c is the propagation degree of operation category c.

[0063] T is the number of time cycles for operating data.

[0064] W is the number of window areas involved in the operation.

[0065] x t,w is the operation intensity (such as the degree of window opening) of the window area w corresponding to the time point t.

[0066] Δx t is the change in operation at time point t, calculated as: It reflects the magnitude of change in the operation at each point in time. A larger magnitude of change indicates a more extensive operation, which may affect more areas.

[0067] is the total change at all time points.

[0068] It represents the sum of the operation intensity of all window areas at a certain time t, and measures the scope of operation activities at the corresponding time.

[0069] After normalizing these factors, the spread degree calculates the spread breadth of each operation category in the time and space dimensions.

[0070] The concentration measures the local concentration of the operation category and reflects the concentration of the operation behavior. If the operation shows a strong concentration trend in a specific time and window area, the operation category may be a frequent operation. Otherwise, if the operation distribution is scattered or too volatile, it may be an invalid operation.

[0071] Concentration calculation formula: Among them, C c is the concentration of operation category c.

[0072] The greater the calculated concentration, the more concentrated the operation is, which may be a frequent and effective operation; conversely, if the concentration is low, it means that the operation behavior is more dispersed, which may be an invalid or meaningless operation.

[0073] The operation index I is obtained by combining the calculation results of the spread and concentration and the weights of the two. c The operation index is used to distinguish between frequent operations and potentially invalid operations. The higher the index, the more frequent and concentrated the operation behavior is, and the system response may be more effective; otherwise, it is an invalid operation.

[0074] Operation index calculation formula: in, is the historical operation average intensity of the window area w.

[0075] Deviations from historical operating intensity It is used to penalize operations with excessive fluctuations. If the window area of ​​a certain operation category frequently deviates from its historical average, it may indicate that the operation is unstable and potentially invalid. Therefore, the denominator in the index formula is used to penalize such fluctuations, so that the value of the operation index more truly reflects the effectiveness of the operation.

[0076] According to the calculated operation index, it is compared with threshold 1 and threshold 2:

[0077] Frequent operation: If the operation index is greater than the threshold of 1, it is determined to be a frequent operation. In this case, the spread and concentration of the operation are high, indicating that users have a high dependence or demand on this operation category. A high value of the operation index represents the expansion effect of the operation mode in space and time, and has a high concentration in a local area, indicating that the operation is highly responsive to the environment or user needs. Therefore, the system will respond to these operations first to maintain the efficiency and accuracy of the window control system.

[0078] If windows are frequently opened and closed in a certain area, and this operation occurs in multiple areas and also shows a high frequency and concentration in time, a high value of the operation index means that the user's demand at this time is relatively stable and strong, and the system needs to continue to maintain this operation.

[0079] Potentially invalid operation: If the operation index is less than the threshold 2, it is a potentially invalid operation and is determined to be a potentially invalid operation.

[0080] At this time, the spread is low, indicating that the operation has a small impact range and low concentration, indicating that the operation may be an occasional behavior that occurs in a local time and region, lacking continuity and consistency. More importantly, when the operation fluctuates greatly, or the performance of the operation at multiple time points deviates greatly from the historical norm, the operation index will be severely punished, further reflecting that the operation may be unnecessary and inefficient. At this time, the system will reduce the frequency of response to these operations, avoid unnecessary adjustments, save energy and reduce user interference.

[0081] If the user occasionally adjusts the windows in a certain area, and such adjustments only occur during certain specific time periods, while the windows are basically in a stable state at other times, and the operation index is low, the system will identify such operations as potentially invalid operations, thereby avoiding excessive response.

[0082] Among them, threshold 1 and threshold 2 are thresholds that are dynamically adjusted based on historical data. These thresholds can be optimized and adjusted through reinforcement learning or adaptive algorithms to accurately classify different operation categories.

[0083] By combining spatiotemporal feature extraction with deep learning models, this step can not only extract implicit patterns from complex operation data, but also efficiently classify frequent operations and potentially invalid operations using unsupervised learning and clustering algorithms. The application of autoencoders and K-means clustering in spatiotemporal data analysis further enhances the system's insight into operation behaviors, realizes intelligent and automated operation pattern recognition and classification, and provides a scientific basis for subsequent optimization strategies.

[0084] Step S3 includes the following contents:

[0085] In step S2, through spatiotemporal feature extraction and unsupervised learning, the system has successfully divided the operation behaviors into two categories: frequent operations and potentially invalid operations. Frequent operations represent the operation behaviors preferred by users in actual use, while potentially invalid operations may be inefficient behaviors caused by user misoperation, over-adjustment, etc. At this point, the operation data is no longer a simple opening and closing state, but an accurate classification obtained after deep learning and cluster analysis. The next task is to conduct an in-depth analysis of these classification results, quantify their impact on energy consumption and indoor comfort, and generate corresponding benefit evaluation results.

[0086] Frequent operations and potential invalid operations are taken as key inputs and data preprocessing is performed first. Frequently operated data will be given a higher analysis weight because it has a greater impact on indoor comfort and energy efficiency. Potentially invalid operation data will be denoised and data smoothing technology will be used to reduce its adverse interference to the final model.

[0087] Data smoothing techniques include methods such as locally weighted regression (LWR) to reduce the impact of fluctuations in potential ineffective operations on the regression model, ensuring that the analysis more accurately reflects the optimization effect of efficient operations on energy efficiency and comfort.

[0088] S3.1, the goal of regression analysis is to quantify the impact of each operation on energy consumption and indoor comfort. Multiple linear regression is used to analyze the relationship between frequent operations and potential ineffective operations and energy consumption and comfort.

[0089] Assume that: E i =β1O i +β2C i +β3T i +β4H i +∈ i ;

[0090] Among them, E i is the energy consumption of the ith operation, O i is the opening and closing state of the window operation, including frequent operations and potential invalid operations, C i Score the current indoor comfort level, Ti is the external temperature, H i is the indoor humidity, ∈ i is the noise term, β1, β2, β3, β4 are the regression coefficients.

[0091] The process of obtaining the comfort score is based on real-time monitoring of indoor environmental parameters (such as temperature, humidity, air circulation, etc.), and is obtained through deviation analysis from the optimal comfort range. First, key environmental data such as indoor temperature and humidity are collected and compared with the preset optimal comfort range (that is, the ideal temperature and humidity range for user comfort, usually set based on ergonomics and thermal comfort models). Specifically, the deviation between the current temperature and humidity values ​​and the temperature and humidity thresholds in the optimal comfort range is calculated, and the comfort score is adjusted according to different temperature and humidity deviations in combination with the weighted model. Generally speaking, when the temperature and humidity are close to the optimal value, the comfort score is higher. If the deviation increases, the comfort score decreases. The score can be obtained through a linear or nonlinear mapping relationship, such as a comprehensive weighted function based on temperature and humidity, to obtain a specific comfort index, which is then used to optimize the opening and closing state of the window and the indoor environment adjustment strategy.

[0092] For frequent operations, the impact of the operation on comfort and energy efficiency may be positive (improving comfort or optimizing energy consumption). Regression analysis will assign specific coefficients to each operation, quantifying its impact on the target variable (such as energy efficiency).

[0093] For potentially ineffective operations, the regression coefficients of these operations may be low or negative, indicating that they have a weak impact on benefits or cause waste of resources.

[0094] S3.2, further verify the causal relationship of each operation on energy and comfort through causal inference models (such as causal graph models or propensity score matching methods PSM). This step can effectively avoid confounding factors in regression analysis and ensure the accuracy of evaluation results.

[0095] For example, through the causal graph model, it is possible to analyze whether the occurrence of a frequent operation causes an increase in energy consumption, or whether a potentially ineffective operation fails to lead to the expected improvement in comfort, thereby accurately quantifying the benefit contribution of each operation.

[0096] The following causal inference formula is used to calculate the direct and indirect impact of operational behavior on the final benefit:

[0097] Among them, P(O i ) is the operation behavior O i The probability of occurrence, ΔE(O i ) is the impact of the operation on energy consumption. Through the causal inference model, the key influencing factors can be identified more accurately from the operation behavior.

[0098] S3.3, based on the results of regression analysis and causal inference, generate a benefit evaluation for each operation. Specifically, evaluate the difference in benefits between each frequent operation and potentially ineffective operation in terms of energy consumption and comfort. For frequent operations, if their energy efficiency and comfort benefits are high, they will be confirmed as efficient operations, and will continue to be retained and strengthened; for potentially ineffective operations, if their benefits are low or negative, unnecessary operations will be filtered out through dynamic strategies to reduce ineffective interventions.

[0099] To further optimize decision making, calculate the operational effectiveness index OEI: Among them, β i are the regression coefficients determined for each operation in the regression analysis.

[0100] Among them, OEI stands for Operational Effectiveness Index, which reflects the contribution of operational behavior to resource optimization. A higher value of the Operational Effectiveness Index indicates that the operation is an efficient operation, while a lower value of the Operational Effectiveness Index indicates a potential ineffective operation.

[0101] Step S3 combines multivariate regression analysis and causal inference models, and provides a quantitative basis for the system by deeply analyzing the benefits of frequent operations and potential ineffective operations, ensuring that window partition control can accurately and intelligently respond to user needs. In this process, through regression analysis and causal inference, the system not only identifies efficient operations, but also eliminates inefficient operations in real time, achieving the optimal balance between energy and comfort, minimizing resource waste and optimizing user experience.

[0102] In step S3, through multivariate regression analysis and causal inference models, the system has quantified the specific impact of each window operation on energy consumption and indoor comfort, and generated an operation benefit index. These indices not only reflect the effectiveness of the operation behavior, but also reveal which operations contribute more to the overall benefit of the system. In order to further optimize the window control system, it is necessary to identify and filter out those operations that do not meet the contribution standards to reduce energy waste and user interference. Therefore, the core task of step S4 is to use rules and anomaly detection to conduct in-depth analysis of the benefit evaluation results and intelligently filter out inefficient or invalid window adjustment operations.

[0103] Step S4 includes the following contents:

[0104] According to the operation benefit index value, the operation category is initially divided into efficient operation and inefficient operation. If the operation benefit index is greater than the corresponding threshold, it is marked as efficient operation, otherwise it is marked as inefficient operation. Efficient operation will be retained and responded to first, while inefficient operation needs further screening and processing.

[0105] Next, it is necessary to identify inefficient operations that exhibit abnormal characteristics in their operating behavior, especially those operating patterns that frequently fluctuate or emergent.

[0106] The operation benefit index, spread, concentration, and time series characteristics of the operation (such as the operation frequency change rate and the operation intensity fluctuation rate) are used as input features to calculate the anomaly score of each operation category. For example, the anomaly score AS of each operation category is calculated by the following composite scoring formula:

[0107] in:

[0108] λ1 and λ2 are weight coefficients, reflecting the importance of propagation and concentration in anomaly scoring.

[0109] γ is a nonlinear exponent that increases sensitivity to high deviation operations.

[0110] Deviations of the operating intensity from the historical mean are calculated for each window region, and extreme deviations are emphasized using an exponential transformation.

[0111] If the anomaly score is greater than the anomaly threshold, the corresponding operation category is marked as an abnormal operation and needs to be filtered.

[0112] Step S4 systematically identifies and filters out window adjustment operations that do not meet the standards for energy consumption and indoor comfort based on the Operational Effectiveness Index (OEI), propagation, and concentration by combining rules and anomaly detection. Specifically, inefficient operation categories are preliminarily marked based on preset thresholds and operation characteristics; then, anomaly detection further verifies the ineffectiveness of these operations through an anomaly scoring mechanism. Finally, the window control strategy is automatically adjusted through an adaptive controller to eliminate or reduce unnecessary operations, thereby optimizing resource utilization and improving overall efficiency and user experience. This process realizes the intelligent screening and dynamic optimization of operation behaviors, ensuring that the window control system continues to operate on an efficient and accurate track.

[0113] Embodiment 2: Figure 2 The present invention provides a partitionable remote control window system, comprising: a data acquisition module, a pattern recognition module, a benefit evaluation module and an operation filtering module;

[0114] Data acquisition module: collects the operation data of window partitions in real time through IoT sensors and passes it to the pattern recognition module;

[0115] Pattern recognition module: extracts spatiotemporal features from the collected data, identifies the user's operation pattern through unsupervised learning, classifies the behavioral features of frequent operations and potentially invalid operations, and passes the classification results to the benefit evaluation module;

[0116] Benefit evaluation module: quantify the specific impact of each window operation on energy consumption and indoor comfort through multivariate regression analysis and causal inference model, generate benefit evaluation results, and pass the generated results to the operation filtering module;

[0117] Operation filtering module: Based on the benefit evaluation results, use rules and anomaly detection to identify operations that do not meet the contribution standards and filter out unnecessary window adjustments.

[0118] 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.

[0119] 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.

[0120] 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.

[0121] 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 remotely controlling windows in partitions, characterized in that: Includes steps: S1. Collect the operation data of window partitions in real time through IoT sensors; S2. Extract spatiotemporal features from the collected data, identify the user's operation mode through unsupervised learning, and classify the behavioral characteristics of frequent operations and potential invalid operations; S3. Through multivariate regression analysis and causal inference models, quantify the specific impact of each window operation on energy consumption and indoor comfort, and generate benefit evaluation results; S4. Based on the benefit evaluation results, use rules and anomaly detection to identify operations that do not meet the contribution standards and filter out unnecessary window adjustments.

2. A method for remotely controlling a partitionable window according to claim 1, characterized in that: Step S1 includes the following contents: Independent sensor nodes are deployed in each window partition, and each sensor node is used to pre-process data, and the processed data is transmitted to the central data acquisition system.

3. A method for remotely controlling a partitionable window according to claim 2, characterized in that: Step S2 includes the following contents: Spatiotemporal feature extraction: First, the time series data is subjected to feature extraction through a convolutional neural network to extract local patterns and structural features in the local time series. The data in each time window is subjected to a convolution operation through a sliding window to extract local operation pattern features. After the spatiotemporal features are extracted, these features are used to identify the categories of different user operation modes. An autoencoder is used to reduce the dimension of the data and detect abnormal patterns. The autoencoder learns the low-dimensional representation of the data and maps the high-dimensional operation features into a low-dimensional latent space. Different areas of the latent space correspond to different modes of user operation. Clustering and unsupervised learning: In the output latent space of the autoencoder, the K-means clustering algorithm is used to cluster data and identify multiple operation categories, including frequent operations and potentially invalid operations. The K-means algorithm maps spatiotemporal features to different cluster centers, uses Euclidean distance to measure the similarity between sample points and cluster centers, and marks each sample to which category based on the clustering results. Classification of frequent operations and potentially invalid operations: After obtaining the operation categories through K-means clustering, the propagation and concentration of each operation category are calculated to distinguish between frequent operations and potentially invalid operations.

4. A method for remotely controlling a partitionable window according to claim 3, characterized in that: The following is a specific process for distinguishing frequent operations from potentially invalid operations: The spread degree measures the breadth of influence of a certain operation category in the system, reflecting the diffusion effect of its operation mode. It is defined by calculating the degree of change of each operation category in multiple time windows and the number of window areas involved. The spread degree calculation formula is: Among them, P c is the propagation degree of operation category c; T is the time period of operation data; W is the number of window areas involved in the operation; x t,w is the operation intensity of the window area w at time point t; Δx t is the change in operation at time point t, calculated as: reflects the magnitude of change in the operation at each point in time; is the total change at all time points; It represents the sum of the operation intensity of all window areas at a certain time t, which measures the scope of operation activities at the corresponding time; The concentration measures the local concentration of the operation category and reflects the concentration of the operation behavior. The concentration calculation formula is: Among them, C c is the concentration of operation category c.

5. A method for remotely controlling a partitionable window according to claim 4, characterized in that: The operation index I is obtained by combining the calculation results of the spread and concentration and the weights of the two. c , the operation index is used to distinguish frequent operations from potential invalid operations; the operation index calculation formula is: in, is the historical average operation intensity of the window area w, and the deviation of the historical operation intensity Used to penalize overly volatile operations; According to the calculated operation index, it is compared with threshold 1 and threshold 2: Frequent operation: If the operation index is greater than the threshold 1, it is determined to be a frequent operation; Potentially invalid operation: If the operation index is less than the threshold 2, it is a potentially invalid operation and is determined to be a potentially invalid operation.

6. A method for remotely controlling a partitionable window according to claim 5, characterized in that: Step S3 includes the following contents: S3.1, multiple linear regression is used to analyze the relationship between frequent operations and potential ineffective operations and energy consumption and comfort, and the settings are as follows: E i =β1O i +β2C i +β3T i +β4H i +∈ i ; Among them, E i is the energy consumption of the ith operation, O i is the opening and closing state of the window operation, including frequent operations and potential invalid operations, C i Score the current indoor comfort level, T i is the external temperature, H i is the indoor humidity, ∈ i is the noise term, β1, β2, β3, β4 are regression coefficients; S3.2, verify the causal relationship of each operation on energy and comfort through the causal inference model; use the following causal inference formula to calculate the direct and indirect impact of the operation behavior on the final benefit: Among them, P(O i ) is the operation behavior O i The probability of occurrence, ΔE(O i ) is the impact of the corresponding operation on energy consumption; S3.3, based on the results of regression analysis and causal inference, generate a benefit evaluation for each operation; specifically, evaluate the difference in benefits between each frequent operation and potential ineffective operation in terms of energy consumption and comfort; calculate the operational benefit index OEI: Among them, β i are the regression coefficients determined for each operation in the regression analysis.

7. A method for remotely controlling a partitionable window according to claim 6, characterized in that: Step S4 includes the following contents: According to the operation benefit index value, the operation categories are initially divided into efficient operations and inefficient operations. If the operation benefit index is greater than the corresponding threshold, it is marked as an efficient operation, otherwise it is marked as an inefficient operation. Next, it is necessary to identify those inefficient operations that show abnormal characteristics in the operation behavior, and use the operation benefit index, propagation, concentration and time series characteristics of the operation as input features to calculate the abnormal score of each operation category. If the abnormal score is greater than the abnormal threshold, the corresponding operation category is marked as an abnormal operation and needs to be filtered.

8. A partitionable remote control window system, used to implement a partitionable remote control window method as claimed in any one of claims 1 to 7, characterized in that: include: Data collection module, pattern recognition module, benefit evaluation module and operation filtering module; Data acquisition module: collects the operation data of window partitions in real time through IoT sensors and passes it to the pattern recognition module; Pattern recognition module: extracts spatiotemporal features from the collected data, identifies the user's operation pattern through unsupervised learning, classifies the behavioral features of frequent operations and potentially invalid operations, and passes the classification results to the benefit evaluation module; Benefit evaluation module: quantify the specific impact of each window operation on energy consumption and indoor comfort through multivariate regression analysis and causal inference model, generate benefit evaluation results, and pass the generated results to the operation filtering module; Operation filtering module: Based on the benefit evaluation results, use rules and anomaly detection to identify operations that do not meet the contribution standards and filter out unnecessary window adjustments.

Images