Intelligent pressure fluctuation detection and regulation system for building curtain wall

Abstract

The invention provides an intelligent pressure fluctuation detection and regulation system for a building curtain wall, relates to the technical field of intelligent control and wind pressure regulation, and integrates five modules of data acquisition, credibility evaluation, dynamic wind pressure prediction, cooperative regulation and fault-tolerant guarantee to construct a high-precision wind pressure sensing and closed-loop control system. Rejecting failure nodes by evaluating the credibility of real-time data; a wind pressure trend is predicted based on an LSTM model in combination with a typhoon path, a current prediction error parameter Epr is calculated, and an error threshold E is dynamically compared to guarantee prediction reliability; a multi-objective optimization function is adopted to balance related parameters, a regulation and control instruction set is generated, and millisecond-level response and emergency switching are supported; the reconstruction module utilizes Dk1 (t), Dk2 (t) and Drec (t) to repair missing data, and a standby mechanism is started to guarantee continuous operation when necessary; meanwhile, a regulation and control instruction is dynamically and finely adjusted and closed-loop optimization is And the robustness, the response precision and the regulation and control stability of the system in high-humidity and strong-typhoon environments are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control and wind pressure regulation, and in particular to an intelligent pressure fluctuation detection and regulation system for building curtain walls. Background Art

[0002] Intelligent pressure fluctuation detection and control systems for building curtain walls originated in the late 20th century. With the rise of high-rise buildings and the frequent occurrence of extreme weather conditions, the wind pressure resistance and airtightness deficiencies of traditional curtain walls have become increasingly apparent. Early systems relied on mechanical sensors and manual control. However, the integration of the Internet of Things and AI technologies in the early 21st century has driven the development of intelligent systems. By monitoring pressure fluctuations in real time and dynamically adjusting the curtain wall structure, these systems have significantly improved building safety and energy efficiency.

[0003] In the prior art, the publication number is CN118313167A, and the name is an intelligent curtain wall intelligent control and supervision system and method based on the Internet of Things. The invention initializes the construction sequence by calling the intelligent curtain wall three-dimensional diagram generated by BIM simulation, and encodes the gridded curtain wall blocks and structural force detection points presented in the intelligent curtain wall three-dimensional diagram respectively; generates an initialized simulation construction list set and a real-time construction list set; calls the initialized force data cluster at each structural force detection point generated by BIM simulation, generates an initialized detection log set, and based on the Internet of Things technology, synchronously collects the real-time force data cluster at each structural force detection point through the force detection sensor to generate a real-time detection log set; analyzes the first satisfaction state and the second satisfaction state of the real-time control construction method, and performs intelligent control early warning.

[0004] However, when the above technical solution is applied to coastal areas with high humidity and strong typhoon climate, under climatic conditions including strong typhoons, the sensors usually have signal errors or failures due to insufficient environmental adaptability and weak anti-interference capabilities, resulting in the control system being unable to accurately perceive key data such as wind pressure, temperature and humidity, thereby affecting the input accuracy of the prediction model.

[0005] When the perception data is distorted, the control algorithm will generate control instructions based on incomplete or erroneous information, resulting in delayed response or even misadjustment, and failure to respond to instantaneous wind pressure shocks in a timely manner, thereby causing chain safety risks such as local overload of the curtain wall and structural damage.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0007] The purpose of the present invention is to provide an intelligent pressure fluctuation detection and control system for building curtain walls to solve the problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] Building curtain wall intelligent pressure fluctuation detection and control system, including:

[0010] Data acquisition module: used to collect wind pressure data, ambient temperature and humidity data, and structural strain data of each node on the curtain wall surface in real time during the current monitoring period to generate a pre-processed joint data set;

[0011] Credibility calculation module: used to receive the pre-processed joint data set and perform credibility Ckx calculation on the monitoring data of each node in the joint data set through the pre-built multi-dimensional data quality assessment model;

[0012] Intelligent prediction module: This module is used for the dynamic wind pressure prediction model based on the LSTM neural network. It uses the historical wind pressure data on the curtain wall surface and real-time typhoon path information as model input, and uses the wind pressure fluctuation trend as the model output to perform dynamic wind pressure prediction for the next monitoring period.

[0013] Collaborative control module: This module receives the wind pressure fluctuation trend forecast results for the next monitoring period and generates control instructions for curtain wall control parameters using a multi-objective optimization algorithm. Curtain wall control parameters include vent opening, damper stiffness, and shading angle.

[0014] Fault-tolerant guarantee module: It is used to obtain the adjustment margin index of the curtain wall control parameters of each node on the curtain wall surface and the monitoring data credibility Ckx of each monitoring node at the current moment, so as to build a fault-tolerant adjustment model and reconstruct the signal, and at the same time execute the fine-tuning calibration strategy for the control instructions.

[0015] Furthermore, the data acquisition module is used to deploy a weather-resistant sensor array on the curtain wall surface, including fiber optic wind pressure sensors, temperature and humidity sensors coated with a nano-hydrophobic layer, and embedded micro-strain sensors, and to perform preliminary filtering and formatting preprocessing on the collected wind pressure data, ambient temperature and humidity data, and structural strain data to construct a joint data set.

[0016] Furthermore, the credibility calculation module is used to build a multi-dimensional data quality assessment model, and in each data collection cycle, the multi-dimensional data quality assessment model is called to calculate the credibility of the collected data of all current monitoring nodes at the current time j;

[0017] The node sequence set evenly distributed on the curtain wall surface is set to {1, 2, …, k, …, K}, where k is the index of the node and K is the total number of nodes on the curtain wall surface.

[0018] Based on the multi-dimensional data quality assessment model, the physical fluctuation rationality index Ca of various sensors, the adjacent node data consistency index Cb and the historical data trend matching degree Cc are analyzed and quantified in real time;

[0019] After extracting the rationality of physical fluctuations, the consistency of adjacent node data, and the matching degree of historical data trends and performing dimensionless processing, the credibility Ckx of node k is calculated in real time. The specific calculation formula is as follows:

[0020] ;

[0021] Where C i It represents the i-th value after dimensionless processing, i=1 is used to represent the rationality of physical fluctuation Ca; i=2 is used to represent the consistency of adjacent node data Cb; i=3 is used to represent the trend matching degree of historical data Cc, is the credibility of node k Ckx, credibility The value range of is [0,1];

[0022] The credibility of all monitoring nodes Arrange in descending order from high to low to generate a credibility sorted list, where Ckx1≥Ckx2≥...≥Ckx K ;

[0023] Based on historical sensor performance data and system control accuracy requirements, through simulation testing, the lowest data quality level without affecting the prediction accuracy is determined, and the confidence threshold Cth is preset. k , and credibility Conduct comparative assessments;

[0024] When credibility <Confidence threshold Cth k When , the current node data is deleted.

[0025] Furthermore, the intelligent prediction module includes a wind pressure modeling unit;

[0026] The wind pressure modeling unit is connected to the meteorological data platform or meteorological warning platform through the real-time typhoon monitoring and forecast data interface to collect typhoon path information in real time;

[0027] The wind pressure data recorded by each node in the data acquisition module is extracted and used as the input data set along with the typhoon path information. After data normalization and time series processing, the data is input into the wind pressure dynamic prediction model built based on the long-short-term memory neural network.

[0028] The wind pressure dynamic prediction model is trained offline through historical wind pressure events and typical typhoon cases, and has the ability to learn the short-term wind pressure change trend; after the wind pressure dynamic prediction model is run, it generates the wind pressure fluctuation trend prediction value Fyz for the next monitoring time period.

[0029] Furthermore, the intelligent prediction module also includes an error evaluation unit;

[0030] After extracting the real-time wind pressure value Fa and ambient relative humidity Fb from the joint dataset and performing dimensionless processing, the ambient corrected wind pressure value Fre is calculated. The specific calculation formula is as follows:

[0031] ;

[0032] Based on the environmentally corrected wind pressure value Fre, a real-time error evaluation is performed on the wind pressure fluctuation trend forecast value Fyz, and the current forecast error parameter Epr is obtained using the following formula:

[0033] ;

[0034] Where, e represents a small correction to prevent the denominator from being zero under zero wind pressure conditions;

[0035] The preset prediction error threshold E is compared with the current prediction error parameter Epr. The specific contents are as follows:

[0036] If the current prediction error parameter Epr ≤ the prediction error threshold E, the wind pressure fluctuation trend prediction result is judged to be normal, and the control instruction is continued to be executed based on the output of the current wind pressure dynamic prediction model;

[0037] If the current prediction error parameter Epr> the prediction error threshold E, then the wind pressure fluctuation trend prediction result is judged to be abnormal, and the self-check mechanism of the wind pressure dynamic prediction model is triggered.

[0038] Furthermore, the collaborative control module includes an instruction generation unit;

[0039] The instruction generation unit is used to receive the wind pressure fluctuation trend prediction value Fyz and the input of the multi-objective optimization algorithm;

[0040] Based on the wind pressure fluctuation trend prediction value Fyz, the optimal control objective function including the wind pressure safety threshold Zf, energy efficiency Xn and comfort Sn is constructed through a multi-objective optimization algorithm. The specific function is expressed as follows:

[0041] ;

[0042] In the formula, α, β, and γ are the weighted coefficients of wind pressure safety threshold, energy efficiency, and comfort, respectively, and are adjusted according to actual needs;

[0043] Based on the optimal regulation objective function, a set of regulation instructions including the opening degree of the ventilation opening, the stiffness of the damper, and the shading angle is generated.

[0044] Furthermore, the collaborative regulation module further includes a real-time response unit;

[0045] The real-time response unit is used to deploy edge computing nodes at key positions of the building curtain wall and directly connect them to the sensor array; use the edge computing nodes to receive the set of regulation instructions and perform real-time regulation operations, including adjusting the opening degree of the ventilation opening, the stiffness of the damper, and the shading angle;

[0046] When the current prediction error parameter Epr is greater than the prediction error threshold E, an emergency response mechanism is triggered to perform millisecond-level wind pressure regulation.

[0047] Furthermore, the collaborative regulation module further includes an effect feedback unit;

[0048] The effect feedback unit is used to graphically feedback the evaluation result of the current prediction error parameter Epr to the management terminal, providing abnormal prompts and regulation suggestions;

[0049] When the evaluation result is that the current prediction error parameter Epr is greater than the prediction error threshold E, the prediction deviation abnormality is prompted in forms including interface warning and abnormal marking of the data curve, and regulation optimization suggestions are submitted, including selecting to update the model or perform error correction instructions according to the self-check mechanism;

[0050] When the evaluation result is that the current prediction error parameter Epr is less than or equal to the prediction error threshold E, the current stable operation state of the system is synchronously displayed, and trend visualization analysis is provided for operation and maintenance personnel to make auxiliary decisions and policy corrections.

[0051] Furthermore, the fault tolerance guarantee module includes a signal reconstruction unit;

[0052] The signal reconstruction unit is used to identify it as a failed node when the credibility of the k-th node is lower than the threshold Cth k At this time, use the real-time wind pressure data Dk1(t) and Dk2(t) of the two nearest valid nodes k1 < k err < k2 on both sides of the failed node, combine the wind pressure dynamic prediction model to extract the wind pressure time series characteristics, and use the interpolation algorithm to obtain the missing data at the current moment t for data reconstruction, and generate a prediction compensation value Drec(t);

[0053] Among them, k1 represents the number of the valid node adjacent to the previous one of the failed node kerr and with a current credibility higher than the threshold Cth k The kerr represents the node number in the failed state, that is, the k-th node, whose current credibility is lower than the set Cth kThreshold; k2 indicates the next adjacent node after the failed node kerr, and its current credibility is higher than the threshold Cth k A valid node number;

[0054] Based on the maximum tolerable error boundary between the reconstruction deviation range of historical node failure data and the control accuracy requirement, a prediction compensation threshold D is preset and compared with the prediction compensation value Drec(t) to evaluate the effectiveness of the reconstructed data. The specific contents are as follows:

[0055] If the predicted compensation value Drec(t) ≤ the predicted compensation threshold D, the current reconstructed data is considered valid. At this time, the current reconstructed data replaces the failed node and the instruction is adjusted;

[0056] If the predicted compensation value Drec(t) is greater than the predicted compensation threshold D, the current reconstructed data is considered invalid and the backup actuator is triggered;

[0057] Furthermore, the fault-tolerance guarantee module also includes an instruction fine-tuning unit;

[0058] The instruction fine-tuning unit evaluates the error between the predicted compensation value Drec(t) and the actual feedback state of the system, including the vent opening, damper stiffness, and shading angle, and the predicted control target value to obtain the current control deviation value Ectrl. The specific calculation formula is as follows:

[0059]

[0060] Among them, Ureal represents the actual control state value, and Uexp represents the predicted control target value;

[0061] Based on key performance indicators including wind pressure response time and structural response stability, simulation analysis is performed on the maximum allowable error range under the most unfavorable climatic conditions, and the control deviation threshold Ec is preset;

[0062] The control deviation threshold Ec is compared with the current control deviation value Ectrl for evaluation. The specific evaluation contents are as follows:

[0063] If the current control deviation value Ectrl ≤ the control deviation threshold Ec, the current control state is determined to be normal and the original control instruction is maintained;

[0064] If the current control deviation value Ectrl>the control deviation threshold Ec, the current control state is determined to be abnormal, indicating that there is a deviation in the current control. At this time, the instruction correction mechanism is activated to fine-tune the control parameters according to the degree of deviation and historical trends.

[0065] Compared with the existing technology, the beneficial effects of the present invention are: through the organic cooperation of the data acquisition module, the credibility calculation module, the intelligent prediction module, the coordinated control module and the fault-tolerant guarantee module, it is possible to realize high-precision real-time acquisition and quality assessment of curtain wall surface wind pressure, ambient temperature and humidity, and structural strain data under high humidity and strong typhoon climate conditions, avoiding data errors and failure problems caused by insufficient environmental adaptability of sensors; in particular, the credibility calculation module performs a comprehensive credibility Ckx assessment based on the physical fluctuation rationality Ca, the consistency Cb of adjacent node data, and the historical data trend matching degree Cc, and sets a credibility threshold Cth k Screening is carried out to effectively eliminate invalid node data, improve the quality of data input, and lay an accurate and reliable foundation for subsequent wind pressure fluctuation trend prediction and control instruction generation; through the intelligent prediction module, historical wind pressure data and typhoon path information are combined, and a long-short-term memory neural network is used to construct a wind pressure dynamic prediction model, and the prediction error parameter Epr is calculated in real time and compared with the preset prediction error threshold E to ensure the credibility and stability of the prediction output.

[0066] The present invention further uses a collaborative control module to dynamically generate control instructions for vent opening, damper stiffness and shading angle under the guidance of a multi-objective optimization function of wind pressure safety threshold Zf, energy efficiency Xn and comfort Sn, and performs millisecond-level responses on edge computing nodes based on a real-time response unit. At the same time, emergency control is triggered under abnormal conditions, significantly improving the dynamic response speed and accuracy of the system in extreme climates such as strong typhoons; the prediction error parameter Epr is visually monitored through an effect feedback unit, and warnings and strategy correction suggestions are provided in case of abnormalities, further improving the intelligent operation and maintenance level of the system; in the fault-tolerant guarantee module, the signal reconstruction unit uses an interpolation algorithm to calculate the predicted compensation value Drec(t) of the missing data based on the real-time wind pressure data Dk1(t) and Dk2(t) of the valid nodes on both sides of the failed node and the time series features extracted by the wind pressure dynamic prediction model, and verifies the validity of the reconstructed data by comparing it with the predicted compensation threshold D. If it is invalid, the standby actuator is switched in time to ensure that the system can continue to operate stably even when some nodes fail.

[0067] In the instruction fine-tuning unit, the present invention calculates the control deviation value Ectrl between the actual control state value Ureal and the predicted control target value Uexp, and compares it with the control deviation threshold Ec set by simulation in real time, and dynamically adjusts the control parameters including the vent opening, damper stiffness and shading angle according to the degree of deviation and the change trend, forming a closed-loop feedback continuous optimization process; when the control deviation value Ectrl exceeds the control deviation threshold Ec, the correction mechanism is immediately activated to ensure that the system control output quickly approaches the predicted target, and takes into account the wind pressure control accuracy and indoor comfort requirements, ultimately significantly improving the robustness, stability and safety of the curtain wall system in extreme climatic environments, and effectively solving the problems of prediction deviation and control failure caused by perception distortion. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a schematic diagram of the framework flow of the overall system of the present invention. DETAILED DESCRIPTION

[0069] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to specific embodiments.

[0070] It should be noted that, unless otherwise defined, the technical or scientific terms used in the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0071] Example 1:

[0072] See also Figure 1 The present invention provides a building curtain wall intelligent pressure fluctuation detection and control system, comprising:

[0073] Data acquisition module: used to collect wind pressure data, ambient temperature and humidity data, and structural strain data of each node on the curtain wall surface in real time during the current monitoring period to generate a pre-processed joint data set;

[0074] Credibility calculation module: used to receive the pre-processed joint data set and perform credibility Ckx calculation on the monitoring data of each node in the joint data set through the pre-built multi-dimensional data quality assessment model;

[0075] Intelligent prediction module: This module is used for the dynamic wind pressure prediction model based on the LSTM neural network. It uses the historical wind pressure data on the curtain wall surface and real-time typhoon path information as model input, and uses the wind pressure fluctuation trend as the model output to perform dynamic wind pressure prediction for the next monitoring period.

[0076] Collaborative control module: This module receives the wind pressure fluctuation trend forecast results for the next monitoring period and generates control instructions for curtain wall control parameters using a multi-objective optimization algorithm. Curtain wall control parameters include vent opening, damper stiffness, and shading angle.

[0077] Fault-tolerant guarantee module: It is used to obtain the adjustment margin index of the curtain wall control parameters of each node on the curtain wall surface and the monitoring data credibility Ckx of each monitoring node at the current moment, so as to build a fault-tolerant adjustment model and reconstruct the signal, and at the same time execute the fine-tuning calibration strategy for the control instructions.

[0078] In this embodiment, the data acquisition module deploys weather-resistant fiber-optic wind pressure sensors, temperature and humidity sensors coated with a nano-hydrophobic layer, and embedded micro-strain sensors on the curtain wall surface. This allows for real-time collection of wind pressure data, ambient temperature and humidity data, and structural strain data. After preliminary filtering and formatting, a high-quality joint data set is generated, improving the accuracy and integrity of the monitoring data from the source.

[0079] The credibility calculation module uses a multi-dimensional data quality assessment model, combining the physical fluctuation rationality index Ca, the adjacent node data consistency index Cb, and the historical trend matching index Cc to perform dimensionless analysis and calculate the credibility Ckx, realizing the timely identification and elimination of abnormal data nodes to ensure the quality of data input;

[0080] The intelligent prediction module is based on a long short-term memory neural network (LSTM) wind pressure dynamic prediction model. Combining historical wind pressure data with real-time typhoon path information, it can predict the wind pressure fluctuation trend Fyz within the next monitoring period in advance, improving the system's ability to respond in advance to strong wind impacts.

[0081] The collaborative control module constructs the optimal control objective function based on the wind pressure safety threshold Zf, energy efficiency Xn, and comfort Sn. It dynamically adjusts the vent opening, damper stiffness, and shading angle through a multi-objective optimization algorithm, accurately generating a control instruction set that takes into account structural safety, energy conservation, and indoor comfort.

[0082] The fault-tolerant guarantee module monitors the node credibility Ckx and the control parameter adjustment margin in real time, uses the signal reconstruction algorithm to generate the predicted compensation value Drec(t), and calculates the control deviation value Ectrl based on the actual control state value Ureal and the predicted control target value Uexp. It performs fine-tuning calibration within the maximum tolerance error range to ensure that even in the event of partial node failure or data anomalies, the overall control system still maintains an efficient, accurate and robust operating state.

[0083] Example 2

[0084] The data acquisition module is used to perform preliminary filtering and formatting preprocessing on the collected wind pressure data, ambient temperature and humidity data, and structural strain data to construct a joint data set.

[0085] Furthermore, the data acquisition module is used to deploy a weather-resistant sensor array on the curtain wall surface, including a fiber optic wind pressure sensor, a temperature and humidity sensor coated with a nano-hydrophobic layer, and an embedded micro-strain sensor;

[0086] Furthermore, an adaptive Kalman filter algorithm was used to denoise the raw data, eliminating signal fluctuations caused by environmental interference. Timestamp alignment technology was used to ensure the spatiotemporal synchronization of multi-source data. Finally, the pre-processed standardized data was integrated into a joint dataset with a unified spatiotemporal benchmark, providing high-quality input for subsequent credibility assessment and predictive modeling. This processing flow significantly improved the reliability and consistency of data collection under extreme climate conditions and solved the problem of cumulative prediction errors caused by poor data quality in traditional systems.

[0087] The credibility calculation module is used to build a multi-dimensional data quality assessment model and call it in each data collection cycle to calculate the credibility of the data collected by all monitoring nodes at the current time j.

[0088] Furthermore, in the construction of the multi-dimensional data quality assessment model, the input includes three key parameters: real-time physical quantity fluctuation data collected by sensors, including wind pressure, temperature, humidity, and strain; synchronized monitoring data from adjacent monitoring nodes; and historical monitoring data trends of the monitoring nodes. The output is a dimensionless comprehensive credibility score Ckx. This score quantitatively analyzes the degree of consistency in three dimensions: the rationality of physical fluctuations, the consistency of adjacent data, and the matching degree of historical trends. It ultimately outputs a standardized credibility index in the range of 0-1 for subsequent data weighted fusion and anomaly detection.

[0089] The purpose of quantitatively analyzing the rationality of physical fluctuations is to analyze whether they conform to the laws of fluid mechanics, the purpose of analyzing the consistency of adjacent data is to evaluate spatial correlation, and the purpose of analyzing the matching degree of historical trends is to evaluate temporal correlation;

[0090] The node sequence set evenly distributed on the curtain wall surface is set to {1, 2, …, k, …, K}, where k is the index of the node and K is the total number of nodes on the curtain wall surface.

[0091] Based on the multi-dimensional data quality assessment model, the physical fluctuation rationality index Ca of various sensors, the adjacent node data consistency index Cb and the historical data trend matching degree Cc are analyzed and quantified in real time;

[0092] Furthermore, the physical fluctuation rationality index Ca, the adjacent node data consistency index Cb, and the historical data trend matching degree Cc are calculated using the following formula:

[0093]

[0094] Where, is the standard deviation of the sensor measurements within the sampling segment; It is the mean of the measured values ​​within the sampling period; the value range of the physical fluctuation rationality index Ca is [0,1]. The closer it is to 1, the more reasonable the fluctuation is, and the closer it is to 0, the more abnormal the fluctuation is.

[0095]

[0096] Where, and It is the mean of the two adjacent nodes during the monitoring period. The closer the adjacent node data consistency index Cb is to 1, the better the adjacent consistency is. The greater the deviation, the lower the index.

[0097]

[0098] Where, It is a data sequence of multiple sampling points in the current time period; It is a data sequence of multiple sampling points under the same historical working conditions; Represents the Pearson correlation coefficient of two sets of data; the value range of the historical data trend matching degree Cc is [-1, 1]. Usually only the positive correlation part is used, that is, the part where the historical data trend matching degree Cc>0. The higher the correlation, the higher the index;

[0099] After extracting the rationality of physical fluctuations, the consistency of adjacent node data, and the matching degree of historical data trends and performing dimensionless processing, the credibility Ckx of node k is calculated in real time. The specific calculation formula is as follows:

[0100]

[0101] Where C irepresents the i-th value after dimensionless processing, including physical fluctuation rationality Ca, adjacent node data consistency Cb and historical data trend matching Cc, with a value range of [0,1];

[0102] Furthermore, the cube root is taken to maintain dimensional consistency and a stable scoring range in [0,1], to avoid the value range being too small or deviating from the distribution after direct multiplication;

[0103] The credibility of all monitoring nodes Arrange in descending order from high to low to generate a credibility sorted list, where Ckx1≥Ckx2≥...≥Ckx K ;

[0104] Furthermore, the credibility threshold Cth k Based on the historical sensor performance database, the long-term operating parameters of node k are extracted, including failure rate, drift coefficient and environmental adaptability. Combined with the system control accuracy requirements, a mapping relationship with the control failure probability is established through Monte Carlo simulation. When the control failure probability exceeds the safety threshold P_, the corresponding minimum credibility threshold Cth is reversely derived. k , and finally generate a dynamic threshold set {Cth1, Cth2...Cth K}, achieving precision and differentiation of data quality control;

[0105] Preset credibility threshold Cth k , and credibility Conduct comparative assessments;

[0106] When credibility <Confidence threshold Cth k , remove the current node data.

[0107] This embodiment, unlike existing technologies, deploys weather-resistant fiber-optic wind pressure sensors, temperature and humidity sensors coated with a nano-hydrophobic layer, and embedded micro-strain sensors on the curtain wall surface. This allows for real-time collection of wind pressure data, ambient temperature and humidity data, and structural strain data, followed by preliminary filtering and formatting preprocessing. This creates a highly consistent and timely joint data set, thereby improving the quality of the underlying monitoring data.

[0108] By establishing a multi-dimensional data quality assessment model, we systematically introduce the physical fluctuation rationality index Ca, the adjacent node data consistency index Cb, and the historical data trend matching index Cc. These three are dimensionlessly processed and combined to generate the credibility Ckx. This model can accurately quantify the reliability of each monitoring node's data, dynamically eliminate abnormal or failed nodes, and ensure the credibility and stability of the input data for subsequent wind pressure forecasting and coordinated control algorithms.

[0109] Among them, the collection of wind pressure data is used to reflect the external wind load on the curtain wall in real time, the collection of temperature and humidity data is used to evaluate the impact of environmental changes on the physical response of the curtain wall, and the collection of strain data is used to capture the deformation state of the curtain wall structure under the action of dynamic wind pressure. Each of them undertakes the basic task of perceiving the physical state of the monitoring system; the physical fluctuation rationality index Ca is used to test whether the single-point data conforms to the basic laws of fluid mechanics and material mechanics, the adjacent node data consistency index Cb is used to quantify the mutual consistency of spatial distribution data, and the historical trend matching index Cc is used to measure the consistency of the data evolution trajectory. Through the joint evaluation of the above-mentioned various indicators, the accuracy of anomaly recognition and overall robustness of the system in complex environments can be greatly improved.

[0110] Example 3

[0111] The intelligent prediction module includes a wind pressure modeling unit;

[0112] The wind pressure modeling unit is connected to the meteorological data platform or meteorological warning platform through the real-time typhoon monitoring and forecast data interface to collect typhoon path information in real time;

[0113] The wind pressure data recorded by each node in the data acquisition module is extracted and used as the input data set along with the typhoon path information. After data normalization and time series processing, the data is input into the wind pressure dynamic prediction model built based on the long-short-term memory neural network.

[0114] The wind pressure dynamic prediction model is trained offline through historical wind pressure events and typical typhoon cases, and has the ability to learn the short-term wind pressure change trend; after the wind pressure dynamic prediction model is run, it generates the wind pressure fluctuation trend prediction value Fyz for the next monitoring time period.

[0115] Furthermore, the wind pressure dynamic prediction model is trained offline based on historical typhoon path information and corresponding wind pressure monitoring data. Typhoon latitude and longitude, wind speed, air pressure and trajectory data are collected, and combined with high-frequency wind pressure records in the region to construct a multi-dimensional feature data set such as typhoon center distance, azimuth, terrain correction, etc.; the wind pressure change rate and fluctuation amplitude within every m seconds are extracted through a sliding window, the wind pressure fluctuation trend value Fyz is defined, and normalization and feature encoding processing are performed; a long short-term memory network enhanced by the spatiotemporal attention mechanism is used, and the actual wind pressure fluctuation trend value Fyz is used as the supervision label for training to optimize the mean square error and local trend loss; after the model is completed, the wind pressure fluctuation trend within the next m seconds is predicted based on the historical sequence, providing support for track safety monitoring and extreme weather warning.

[0116] The intelligent prediction module also includes an error evaluation unit;

[0117] After extracting the real-time wind pressure value Fa and ambient relative humidity Fb from the joint dataset and performing dimensionless processing, the ambient corrected wind pressure value Fre is calculated. The specific calculation formula is as follows:

[0118] ;

[0119] Furthermore, in the formula for the environmentally corrected wind pressure value Fre, the real-time wind pressure value Fa reflects the instantaneous aerodynamic force. However, in actual engineering environments, air humidity affects air density, which in turn affects the actual pressure change per unit area. When humidity is high, the air density becomes lower because the mass of water vapor molecules is smaller than that of dry air molecules. Therefore, at the same wind speed, the actual effect of wind pressure is slightly weakened. At low humidity, the wind pressure effect is relatively enhanced. Using the square root of the product rather than direct linear multiplication or weighting is to make the correction smoother and not introduce linear offset. It is used to take into account the changes in the proportion of the two, keep the dimensions reasonable, and prevent the values ​​from being too large or distorted.

[0120] Based on the environmentally corrected wind pressure value Fre, a real-time error evaluation is performed on the wind pressure fluctuation trend forecast value Fyz, and the current forecast error parameter Epr is obtained using the following formula:

[0121] ;

[0122] Where, e represents a small correction to prevent the denominator from being zero under zero wind pressure conditions;

[0123] Furthermore, the current formula for the prediction error parameter Epr is actually a relative error formula. The numerator, |Fre−Fyz|, represents the absolute error between the predicted value and the actual value; the denominator, Fre, is the actual value, which is equivalent to normalization, converting the absolute error into a proportional error to the actual wind pressure. In addition, regardless of how the wind pressure baseline changes, the error level can be uniformly evaluated.

[0124] The preset prediction error threshold E is compared with the current prediction error parameter Epr. The specific contents are as follows:

[0125] If the current prediction error parameter Epr ≤ the prediction error threshold E, the wind pressure fluctuation trend prediction result is judged to be normal, and the control instruction is continued to be executed based on the output of the current wind pressure dynamic prediction model;

[0126] If the current prediction error parameter Epr> the prediction error threshold E, then the wind pressure fluctuation trend prediction result is judged to be abnormal, and the self-check mechanism of the wind pressure dynamic prediction model is triggered.

[0127] Furthermore, the self-checking mechanism first identifies the source of input anomalies or model parameter drift by retrospectively analyzing the characteristic differences between historical wind pressure data and the current period; then, it constructs local characteristic disturbance samples based on the section with the largest error, combines them with recent measured wind pressure data for rapid incremental training, and adjusts key prediction parameters in the model, including trend learning weights and time window length; finally, it outputs the corrected wind pressure fluctuation trend prediction value Fyz′, which is used for dynamic response and strategy formulation of subsequent collaborative control modules.

[0128] The value of the prediction error threshold E is based on:

[0129] Without affecting the system's control accuracy and response speed, simulation analysis is conducted on the predicted stability and structural safety impacts under different error levels. Combining the load tolerance of the curtain wall structure with historical control performance, an empirically optimal safety critical error range is comprehensively determined.

[0130] In practical applications, in order to facilitate quantification and standardized management of the prediction error threshold E, the following formula can be used to set its numerical range, as a specific embodiment:

[0131] ;

[0132] in: is the mean of historical wind pressure prediction errors; is the standard deviation of historical wind pressure prediction errors, which is used to reflect the error fluctuation range;

[0133] λ is the empirical adjustment coefficient, which ranges from 1.0 to 2.0 and is used to control the tolerance range. A larger value indicates a stronger fault tolerance but a looser precision requirement.

[0134] The collaborative control module includes an instruction generation unit;

[0135] The instruction generation unit is used to receive the wind pressure fluctuation trend prediction value Fyz and the input of the multi-objective optimization algorithm;

[0136] Furthermore, the multi-objective optimization algorithm dynamically optimizes and adjusts the vent opening, damper stiffness, and shading angle, taking wind pressure safety threshold, energy efficiency, and comfort index as optimization targets;

[0137] Based on the wind pressure fluctuation trend prediction value Fyz, the optimal control objective function including the wind pressure safety threshold Zf, energy efficiency Xn and comfort Sn is constructed through a multi-objective optimization algorithm. The specific function is expressed as follows:

[0138] ;

[0139] In the formula, α, β, and γ are the weighted coefficients of wind pressure safety threshold, energy efficiency, and comfort, respectively, and are adjusted according to actual needs;

[0140] Based on the optimal control objective function, a control instruction set including vent opening, damper stiffness and shading angle is generated.

[0141] Furthermore, the wind pressure safety threshold Zf is used to ensure the safety of the curtain wall structure, the energy efficiency Xn optimizes the building's energy utilization by reducing system energy consumption, and the comfort Sn ensures that the indoor environment remains comfortable.

[0142] Furthermore, the wind pressure safety threshold Zf, energy efficiency Xn and comfort Sn represent the key performance indicators of the control system in terms of safety, energy consumption and human experience, respectively. The importance of the three is weighed and adjusted by setting weighting coefficients α, β and γ. The setting basis of the weighting coefficients includes the current environmental risk level, energy-saving priority and dynamic changes in user comfort needs. When the weight of α is high, the system pays more attention to safety. When the weight of β is high, energy consumption control is given priority. When the weight of γ is high, the goal is to improve comfort. In the final generated control instruction set, the vent opening directly affects the indoor and outdoor wind pressure exchange, and mainly acts to meet the control target of the wind pressure safety threshold Zf. The damper stiffness adjustment affects the response of the building structure to wind loads and optimizes the energy consumption level of the structure, which corresponds to the energy efficiency Xn. The shading angle adjustment affects sunlight exposure and indoor air flow distribution, thereby adjusting the perceived comfort, and therefore corresponds to the comfort Sn.

[0143] The collaborative control module also includes a real-time response unit;

[0144] The real-time response unit is used to deploy edge computing nodes at key locations on the building's curtain wall and connect directly to the sensor array. The edge computing node receives the control instruction set and performs control operations in real time, including adjusting the vent opening, damper stiffness, and shading angle.

[0145] When the current prediction error parameter Epr is greater than the prediction error threshold E, the emergency response mechanism is triggered and millisecond-level wind pressure control is performed.

[0146] Furthermore, when the current prediction error parameter Epr is greater than the prediction error threshold E, the system immediately triggers the emergency response mechanism and executes the millisecond-level wind pressure control process; specifically, it includes: quickly calling the high-frequency sampling data of the wind pressure sensor in the recent period to perform short-term trend fitting correction, combining the historical extreme working condition case library with the current environmental characteristics for rapid comparison, automatically switching to the redundant prediction model to obtain the adjusted wind pressure fluctuation trend prediction value Fyz', and then inputting Fyz' into the collaborative control module to reconstruct the weighted weight parameters α, β, and γ of the wind pressure safety threshold Zf, energy efficiency Xn, and comfort Sn, and using the multi-objective optimization algorithm to instantly generate a new round of control instruction sets to achieve dynamic fine-tuning of the vent opening, damper stiffness, and shading angle, thereby achieving rapid response and risk suppression within milliseconds.

[0147] The collaborative control module also includes an effect feedback unit;

[0148] The effect feedback unit is used to feed back the evaluation results of the current prediction error parameter Epr to the management terminal in a graphical form, providing abnormal prompts and control suggestions;

[0149] When the evaluation result shows that the current forecast error parameter Epr is greater than the forecast error threshold E, the forecast deviation abnormality is prompted through interface warnings, data curve abnormality marks, etc., and control optimization suggestions are submitted, including the choice of model update or error correction instructions based on the self-check mechanism;

[0150] When the evaluation result shows that the current prediction error parameter Epr is less than or equal to the prediction error threshold E, the current system operation status is synchronously displayed as stable, and trend visualization analysis is provided for operation and maintenance personnel to assist in decision-making and strategy correction.

[0151] Furthermore, when the current prediction error parameter Epr is greater than the prediction error threshold E, the effect feedback unit will feedback to the management terminal through the graphical interface with an abnormal mark and warning prompt, and submit a control optimization suggestion. The error correction process includes: calling historical error sample data and the latest high-frequency sampling values, combining false alarm and missed alarm cases to build an error distribution model, and using the model residual learning method to dynamically adjust the parameter weights of the wind pressure dynamic prediction model, including the normalized range of input features, the hidden state parameters of the LSTM network and the feature allocation weights of the attention mechanism, and then correct the prediction deviation in real time. If the residual offset exceeds the preset threshold, it will automatically switch to an error-sensitive model structure to enhance the prediction stability, thereby improving the accuracy of subsequent prediction values ​​and the reliability of the control response.

[0152] This embodiment, unlike existing technologies, achieves advanced prediction and dynamic response to wind pressure fluctuations on building curtain walls by introducing an intelligent prediction module and a coordinated control module. The wind pressure modeling unit, through access to a meteorological data platform, collects real-time typhoon path information and combines it with wind pressure data from each node. After normalization and time series processing, the data is input into a wind pressure dynamic prediction model based on a long-short-term memory neural network. This enables the system to have high-precision prediction capabilities for future wind pressure fluctuation trends within a short time window of m seconds.

[0153] The error assessment unit extracts the environmentally corrected wind pressure value Fre and compares it with the wind pressure fluctuation trend prediction value Fyz to calculate the prediction error parameter Epr. It dynamically determines the accuracy of the prediction result. When Epr exceeds the preset prediction error threshold E, the model self-check mechanism is triggered in a timely manner to ensure the reliability and robustness of the prediction output. The instruction generation unit in the collaborative control module uses the wind pressure fluctuation trend prediction value Fyz and a multi-objective optimization algorithm to comprehensively balance the wind pressure safety threshold Zf, energy efficiency Xn, and comfort Sn. It dynamically generates the optimal control instruction set for the vent opening, damper stiffness, and shading angle, thereby taking into account the curtain wall structure safety, energy efficiency, and indoor environmental comfort.

[0154] The real-time response unit quickly executes control instructions through edge computing nodes to ensure that the control response has a millisecond-level timeliness, especially when the prediction error parameter Epr is greater than the prediction error threshold E, the emergency mechanism is triggered, further improving the system resilience; the effect feedback unit feeds back the prediction error evaluation results through a graphical interface, realizing abnormal prompts, control suggestion push and visual display of stable operation status, which greatly facilitates the decision-making and system management of operation and maintenance personnel; the collection and calculation of various key parameters such as typhoon path information, wind pressure data, environmentally corrected wind pressure value Fre, prediction error parameter Epr, wind pressure safety threshold Zf, energy efficiency Xn, and comfort Sn correspond to the system's different functional requirements of external environment perception, wind pressure change trend prediction, prediction accuracy verification, dynamic control optimization and operation effect evaluation, and jointly construct an intelligent, efficient and highly adaptive curtain wall wind pressure fluctuation control system.

[0155] Example 4

[0156] The fault-tolerance guarantee module includes a signal reconstruction unit;

[0157] The signal reconstruction unit is used to reconstruct the signal when the kth node's credibility is lower than the threshold Cth k When it is identified as a failed node, the nearest valid nodes k1 on both sides of the failed node are used. <k errThe real-time wind pressure data Dk1(t) and Dk2(t) of <k2>, combined with the wind pressure dynamic prediction model, extract the wind pressure time series characteristics. The interpolation algorithm is used to obtain the missing data at the current time t for data reconstruction, and a prediction compensation value Drec(t) is generated;

[0158] Among them, k1 represents the number of the effective node adjacent to the previous one of the failure node kerr and with a current credibility higher than the threshold Cth k ; kerr represents the node number in the failure state, that is, the k-th node, whose current credibility is lower than the set Cth k threshold; k2 represents the number of the effective node adjacent to the next one of the failure node kerr and with a current credibility higher than the threshold Cth k ;

[0159] Furthermore, the effective nodes k1 and k2 are respectively the nearest normal nodes spatially close to the failure node kerr, satisfying the relationship k1 < kerr < k2, which is used to support signal reconstruction;

[0160] Furthermore, Dk1(t) and Dk2(t) represent the instantaneous wind pressure measurement values respectively collected by the sensors of nodes k1 and k2 at time t; they are obtained by synchronous sampling at a fixed time interval through a wind pressure sensor array distributed on the track or the surface of the structure; the acquisition process undergoes preliminary filtering and denoising processing to ensure the stability and reliability of the data; the difference between Dk1(t), Dk2(t) and the wind pressure value Fa is that the wind pressure value Fa is the original wind pressure value measured by a single-point sensor and is used for preliminary wind pressure analysis;

[0161] Furthermore, the prediction compensation value Drec(t) is obtained by using the real-time wind pressure data Dk1(t) and Dk2(t) of the effective nodes adjacent to the failure node, combining the time series characteristics extracted by the wind pressure dynamic prediction model, and applying the interpolation algorithm for estimation at the current reconstruction time t; the specific calculation formula for the prediction compensation value Drec(t) is:

[0162] ;

[0163] Based on the maximum tolerable error boundary between the reconstruction deviation range of historical node failure data and the regulation accuracy requirement, a prediction compensation threshold D is preset, and it is compared and evaluated with the prediction compensation value Drec(t) to evaluate the effectiveness of the reconstructed data. The specific content is as follows:

[0164] If the prediction compensation value Drec(t) ≤ the prediction compensation threshold D, it is determined that the current reconstructed data is valid. At this time, the current reconstructed data replaces the failure node and an instruction adjustment is made;

[0165] If the predicted compensation value Drec(t) is greater than the predicted compensation threshold D, the current reconstructed data is considered invalid and the backup actuator is triggered;

[0166] Furthermore, the backup actuator specifically refers to a redundant control component that is automatically enabled when the main wind pressure control execution unit fails or the predicted compensation value Drec(t) exceeds the predicted compensation threshold D, including an independently controlled wind pressure regulator, emergency ventilation equipment or bypass airflow adjustment module, which is used to maintain the ambient wind pressure within a safe range when the main system data is unavailable, thereby ensuring the stable operation and safety of the system.

[0167] The fault-tolerance guarantee module also includes an instruction fine-tuning unit;

[0168] The instruction fine-tuning unit evaluates the error between the predicted compensation value Drec(t) and the actual feedback state of the system, including the vent opening, damper stiffness, and shading angle, and the predicted control target value to obtain the current control deviation value Ectrl. The specific calculation formula is as follows:

[0169]

[0170] Among them, Ureal represents the actual control state value, and Uexp represents the predicted control target value;

[0171] Furthermore, based on key performance indicators including wind pressure response time and structural response stability, a simulation analysis is conducted on the maximum allowable error range under the most unfavorable climatic conditions, and a control deviation threshold Ec is preset;

[0172] The control deviation threshold Ec is compared with the current control deviation value Ectrl for evaluation. The specific evaluation contents are as follows:

[0173] If the current control deviation value Ectrl ≤ the control deviation threshold Ec, the current control state is determined to be normal and the original control instruction is maintained;

[0174] If the current control deviation value Ectrl>the control deviation threshold Ec, the current control state is determined to be abnormal, indicating that there is a deviation in the current control. At this time, the instruction correction mechanism is activated to fine-tune the control parameters according to the degree of deviation and historical trends.

[0175] The fine-tuning mainly includes the following three aspects:

[0176] Fine-tune the vent opening: If the wind pressure is too high, adjust the vent angle appropriately to increase the airflow damping; otherwise, fine-tune and increase the opening;

[0177] Fine-tuning of damper stiffness: Adjust the damper response sensitivity according to the wind vibration response trend to reduce system hysteresis;

[0178] Fine-tuning the shading angle: Taking into account wind pressure and indoor temperature and humidity comfort, adjust the shading system angle to assist pressure control;

[0179] The entire adjustment process is driven in real time by a closed-loop feedback system, ensuring that each command correction quickly converges between the predicted model and actual feedback, thereby improving the accuracy and robustness of the system response. The advantage of this mechanism is that the fine-tuning process is flexible and controllable, avoiding over-adjustment that may lead to system instability.

[0180] In this embodiment, the robustness and emergency handling capability of the system are significantly improved by introducing the fault-tolerant protection module; when the signal reconstruction unit detects that the credibility of a node is lower than the threshold Cth k When the wind pressure is regulated, the system can use the data of the adjacent valid nodes and the wind pressure dynamic prediction model through interpolation algorithm to generate the predicted compensation value Drec(t) and perform effectiveness evaluation to ensure that the system can quickly restore data in the event of node failure, thereby ensuring the continuity and accuracy of wind pressure control; if the reconstructed data is invalid, the system will automatically start the backup actuator to ensure the stable operation of the system;

[0181] The command fine-tuning unit evaluates the error between the predicted control target value and the evaluation result of the predicted compensation value Drec(t), combined with actual feedback data such as vent opening, damper stiffness and shading angle. It determines whether the command correction mechanism needs to be activated by calculating the current control deviation value Ectrl, ensuring that each fine-tuning adjustment is within the optimal control range and avoiding excessive adjustment that may cause system instability. Through real-time monitoring and error evaluation of key parameters such as wind pressure, temperature and humidity, the system can flexibly respond to different environmental changes, ensure the accuracy of control and the timeliness of response, and at the same time protect the safety and structural stability of the building curtain wall under adverse climatic conditions.

[0182] It should be noted that: All calculation formulas in this application document use regression analysis including but not limited to machine learning algorithms to deeply analyze the relevant parameters collected and identify their natural trends and relationships. Use professional software, such as Python's Scikit-learn library or R language, to automatically generate mathematical models that match the data. Then, objectively evaluate the performance of the model through methods such as cross-validation, and combine continuous feedback and optimization to ensure that the created formula truly reflects the inherent laws of the data, thereby ensuring its effectiveness and accuracy. In all calculation formulas in this application, the parameters in each formula are dimensionlessly processed in a consistent range to ensure that different physical quantities are compared on the same scale; dimensionless technical means include but are not limited to Min-Max Normalization and Z-Score normalization;

[0183] The technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0184] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0185] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0186] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. Intelligent pressure fluctuation detection and control system for building curtain walls, characterized by: include: Data acquisition module: used to collect wind pressure data, ambient temperature and humidity data, and structural strain data of each node on the curtain wall surface in real time during the current monitoring period to generate a pre-processed joint data set; Credibility calculation module: used to receive the pre-processed joint data set and perform credibility Ckx calculation on the monitoring data of each node in the joint data set through the pre-built multi-dimensional data quality assessment model; Intelligent prediction module: This module is used for the dynamic wind pressure prediction model based on the LSTM neural network. It uses the historical wind pressure data and climate information of each node on the curtain wall surface as the model input, and the wind pressure fluctuation trend as the model output to perform dynamic wind pressure prediction for the next monitoring period. Collaborative control module: This module receives the wind pressure fluctuation trend forecast results for the next monitoring period and generates control instructions for curtain wall control parameters using a multi-objective optimization algorithm. Curtain wall control parameters include vent opening, damper stiffness, and shading angle. Fault-tolerant guarantee module: It is used to obtain the adjustment margin index of the curtain wall control parameters of each node on the curtain wall surface and the monitoring data credibility Ckx of each monitoring node at the current moment, so as to build a fault-tolerant adjustment model and reconstruct the signal, and at the same time execute the fine-tuning calibration strategy for the control instructions.

2. The building curtain wall intelligent pressure fluctuation detection and control system according to claim 1, characterized in that: The data acquisition module is used to perform preliminary filtering and formatting preprocessing on the collected wind pressure data, ambient temperature and humidity data, and structural strain data, including using the adaptive Kalman filter algorithm to denoise the original data and construct a joint data set.

3. The building curtain wall intelligent pressure fluctuation detection and control system according to claim 1, characterized in that: The credibility calculation module is used to build a multi-dimensional data quality assessment model. In each data collection cycle, the multi-dimensional data quality assessment model is called to calculate the credibility of the collected data of all nodes at the current time t. The specific logic includes: The node sequence set evenly distributed on the curtain wall surface is set to {1, 2, …, k, …, K}, where k is the index of the node and K is the total number of nodes on the curtain wall surface. Based on the multi-dimensional data quality assessment model, the physical fluctuation rationality index Ca, the adjacent node data consistency index Cb and the historical data trend matching degree Cc of various sensors used in the joint data set acquisition are analyzed and quantified in real time; After extracting the rationality of physical fluctuations, the consistency of adjacent node data, and the matching degree of historical data trends and performing dimensionless processing, the credibility Ckx of node k is calculated in real time. The specific calculation formula is as follows: ； Where C i It represents the i-th value after dimensionless processing, i=1 is used to represent the rationality of physical fluctuation Ca; i=2 is used to represent the consistency of adjacent node data Cb; i=3 is used to represent the trend matching degree of historical data Cc, is the credibility of node k Ckx, where credibility The value range of is [0,1]; The credibility of all nodes Arrange in descending order from high to low to generate a credibility sorted list, where Ckx1≥Ckx2≥...≥Ckx K ; Preset the credibility threshold Cth of node k k , and credibility Conduct comparative assessments; When credibility <Confidence threshold Cth k When , the current node data is deleted.

4. The building curtain wall intelligent pressure fluctuation detection and control system according to claim 3 is characterized by: The intelligent prediction module includes a wind pressure modeling unit; The wind pressure modeling unit is connected to the meteorological data platform or meteorological warning platform through the real-time typhoon monitoring and forecast data interface to collect typhoon path information in real time; The wind pressure data recorded by each node in the data acquisition module is extracted and used as the input data set along with the typhoon path information. After data normalization and time series processing, the data is input into the wind pressure dynamic prediction model built based on the long-short-term memory neural network. The wind pressure dynamic prediction model is trained offline through historical wind pressure events and extreme weather cases, and has the ability to learn the short-term wind pressure change trend; after the wind pressure dynamic prediction model is run, it generates the wind pressure fluctuation trend forecast value Fyz for the next monitoring time period.

5. The building curtain wall intelligent pressure fluctuation detection and control system according to claim 4, characterized in that: The intelligent prediction module also includes an error evaluation unit; After extracting the real-time wind pressure value Fa and ambient relative humidity Fb from the joint dataset and performing dimensionless processing, the ambient corrected wind pressure value Fre is calculated. The specific calculation formula is as follows: ； Based on the environmentally corrected wind pressure value Fre, a real-time error evaluation is performed on the wind pressure fluctuation trend forecast value Fyz, and the current forecast error parameter Epr is obtained using the following formula: ； Where, e represents a small correction to prevent the denominator from being zero under zero wind pressure conditions; The preset prediction error threshold E is compared with the current prediction error parameter Epr. The specific contents are as follows: If the current prediction error parameter Epr ≤ the prediction error threshold E, the wind pressure fluctuation trend prediction result is judged to be normal, and the control instruction is continued to be executed based on the output of the current wind pressure dynamic prediction model; If the current prediction error parameter Epr> the prediction error threshold E, then the wind pressure fluctuation trend prediction result is judged to be abnormal, and the self-check mechanism of the wind pressure dynamic prediction model is triggered.

6. The building curtain wall intelligent pressure fluctuation detection and control system according to claim 5, characterized in that: The collaborative control module includes an instruction generation unit; The instruction generation unit is used to receive the final wind pressure fluctuation trend prediction value Fyz and the input of the multi-objective optimization algorithm; based on the wind pressure fluctuation trend prediction value Fyz, the multi-objective optimization algorithm is used to construct the optimal control objective function including the wind pressure safety threshold Zf, energy efficiency Xn and comfort Sn. The specific function is expressed as follows: ； Where α, β, and γ are the weighted coefficients of wind pressure safety threshold, energy efficiency, and comfort, respectively; Based on the optimal control objective function, a control instruction set including vent opening, damper stiffness and shading angle is generated.

7. The building curtain wall intelligent pressure fluctuation detection and control system according to claim 6, characterized in that: The collaborative control module also includes a real-time response unit; The real-time response unit is used to deploy edge computing nodes at node locations on the building curtain wall and directly connect to the sensor array. The edge computing node receives the control instruction set and performs control operations in real time, including adjusting the vent opening, damper stiffness, and shading angle. When the current prediction error parameter Epr is greater than the prediction error threshold E, the emergency response mechanism is triggered and millisecond-level wind pressure control is performed.

8. The building curtain wall intelligent pressure fluctuation detection and control system according to claim 7, characterized in that: The collaborative control module also includes an effect feedback unit; The effect feedback unit is used to feed back the evaluation results of the current prediction error parameter Epr to the management terminal in a graphical form, providing abnormal prompts and control suggestions; When the evaluation result shows that the current forecast error parameter Epr is greater than the forecast error threshold E, the forecast deviation abnormality is prompted through interface warnings, data curve abnormality marks, etc., and control optimization suggestions are submitted, including the choice of model update or error correction instructions based on the self-check mechanism; When the evaluation result shows that the current prediction error parameter Epr is less than or equal to the prediction error threshold E, the current system operation status is synchronously displayed as stable, and trend visualization analysis is provided for operation and maintenance personnel to assist in decision-making and strategy correction.

9. The building curtain wall intelligent pressure fluctuation detection and control system according to claim 8, characterized in that: The fault-tolerance guarantee module includes a signal reconstruction unit; The signal reconstruction unit is used to identify it as a failed node when the credibility of the k-th node is lower than the threshold Cth k At this time, the real-time wind pressure data Dk1(t) and Dk2(t) of the two nearest effective nodes k1 < k err < k2 on both sides of the failed node are used. Combining with the wind pressure dynamic prediction model, the wind pressure time series characteristics are extracted. The interpolation algorithm is used to obtain the missing data at the current time t for data reconstruction, and a prediction compensation value Drec(t) is generated; Among them, k1 represents the previous neighbor of the failed node kerr, and its current credibility is higher than the threshold Cth k The valid node number; kerr represents the node number of the failed state, that is, the kth node, whose current credibility is lower than the set Cth k Threshold; k2 indicates the next adjacent node after the failed node kerr, and its current credibility is higher than the threshold Cth k A valid node number; Based on the maximum tolerable error boundary between the reconstruction deviation range of historical node failure data and the control accuracy requirement, a prediction compensation threshold D is preset and compared with the prediction compensation value Drec(t) to evaluate the effectiveness of the reconstructed data. The specific contents are as follows: If the predicted compensation value Drec(t) ≤ the predicted compensation threshold D, the current reconstructed data is considered valid. At this time, the current reconstructed data replaces the failed node and the control instruction is adjusted; If the predicted compensation value Drec(t) is greater than the predicted compensation threshold D, the current reconstructed data is considered invalid, and the backup actuator is triggered.

10. The building curtain wall intelligent pressure fluctuation detection and control system according to claim 9, characterized in that: The fault-tolerance guarantee module also includes an instruction fine-tuning unit; The command fine-tuning unit evaluates the predicted compensation value Drec(t) based on the actual feedback state of the system, including the vent opening, damper stiffness, and shading angle. It evaluates the error between the actual control state value and the predicted control target value to obtain the current control deviation value Ectrl. The specific calculation formula is as follows: ； Among them, Ureal represents the actual control state value, and Uexp represents the predicted control target value; The preset control deviation threshold Ec is compared with the current control deviation value Ectrl for evaluation. The specific evaluation contents are as follows: If the current control deviation value Ectrl ≤ the control deviation threshold Ec, the current control state is determined to be normal and the original control instruction is maintained; If the current control deviation value Ectrl>the control deviation threshold Ec, the current control state is determined to be abnormal, indicating that there is a deviation in the current control. At this time, the instruction correction mechanism is activated to fine-tune the control parameters according to the degree of deviation and historical trends.

Images