An intelligent building curtain wall system

Abstract

The present invention proposes an intelligent building curtain wall system, which aims to achieve dynamic regulation and management of the building exterior wall environment by integrating advanced environmental response materials and intelligent control technology. The system includes a curtain wall module with a multi-layer structure, which integrates an external permeable layer with a microporous structure, a thermal insulation layer, a structural support layer, an internal decorative layer, and an intelligent material layer composed of electrochromic materials. The environmental monitoring module of the system is equipped with a variety of sensors, which is responsible for real-time monitoring of external and indoor environmental parameters. Based on these data, the intelligent processing module generates thermal control instructions and drainage instructions to adjust the transparency of the intelligent material layer and the opening and closing degree of the drainage valve. The drainage module further enhances the system's adaptability to extreme weather. In addition, the remote communication module allows the system to connect with a third-party system to achieve remote monitoring and control, which improves the user-friendliness of the system. The system provides a solution for buildings with high energy efficiency and strong environmental adaptability.

Description

Technical Field

[0001] The invention relates to the technical field of building curtain walls, and in particular to an intelligent building curtain wall system. Background Art

[0002] In the field of modern architectural design, curtain wall systems, as an important part of the building's external envelope, are responsible for protecting the interior space from the external environment. Traditional curtain wall systems mainly achieve functions such as heat insulation, sound insulation and waterproofing through static physical hierarchical structures, such as using double-layer glass and thermal insulation layers of specific materials. However, these traditional systems are often unable to dynamically adjust their performance according to environmental changes, such as actively adjusting the transmittance of light and heat under different weather conditions to maximize energy efficiency and indoor comfort.

[0003] The main problem with existing technologies is that they lack flexibility and intelligence, and cannot achieve real-time response to changes in indoor and outdoor environments. For example, it is impossible to reduce direct sunlight when the sun is strong, or enhance thermal insulation performance in low-temperature environments. In addition, the drainage design of traditional curtain wall systems is usually passive and cannot adjust the drainage rate according to changes in rainfall, which may cause rainwater accumulation or internal leakage, thus affecting the safety and comfort of the building.

[0004] Therefore, it is urgent to develop a new type of building curtain wall system. Summary of the invention

[0005] The present application provides an intelligent building curtain wall system to improve the adaptability of a building to environmental changes.

[0006] The present application provides an intelligent building curtain wall system, comprising:

[0007] The multi-layer curtain wall module includes an external water-permeable layer, a thermal insulation layer, a structural support layer, an internal decorative layer and an intelligent material layer; wherein the external water-permeable layer is made of a new material with a microporous structure, which can effectively prevent large particles from entering while allowing water to pass smoothly; the intelligent material layer is made of an electrochromic material, which is used to adjust the transparency of the intelligent material layer according to the thermal control instructions of the intelligent processing module, so as to realize active management of light and heat flow;

[0008] An environmental monitoring module is configured with a variety of sensors, which are distributed at key positions of the multi-layer curtain wall module and are used to detect environmental parameters in real time; the environmental parameters include external environmental parameters and indoor environmental parameters; the external environmental parameters include outdoor temperature, light intensity, wind speed, humidity and rainfall; the indoor environmental parameters include indoor temperature, light intensity and humidity;

[0009] The drainage module comprises a drainage valve and a drainage pipe; wherein the drainage pipe is used to discharge the water entering through the external water-permeable layer to the outside; the drainage valve is located at the entrance of the drainage pipe, and is used to adjust the opening and closing degree of the drainage valve according to the drainage instruction of the intelligent processing module;

[0010] The intelligent processing module is used to receive the environmental parameters provided by the environmental monitoring module; determine the target transparency of the intelligent material layer and the target opening and closing degree of the drainage valve according to the environmental parameters; generate a thermal control instruction and a drainage instruction according to the target transparency and the target opening and closing degree of the drainage valve; send the generated thermal control instruction to the curtain wall module of the multi-layer structure; and send the generated drainage instruction to the drainage module;

[0011] The remote communication module is connected with the intelligent processing module and the third-party system to realize the remote monitoring and adjustment control of the intelligent building curtain wall system.

[0012] Furthermore, the intelligent processing module uses a trained neural network model to determine the target transparency of the intelligent material layer and the target opening and closing degree of the drainage valve according to the environmental parameters; the neural network model includes a feature extraction layer, an environmental state recognition layer and a capsule network decision layer;

[0013] The feature extraction layer is implemented by a recurrent neural network, which processes the time series data of the input environmental parameters, captures and learns the dynamic characteristics of the external environmental parameters and the indoor environmental parameters changing over time, and obtains the time series characteristics of the environmental parameters;

[0014] The environmental state recognition layer is implemented by multiple fully connected layers, which are used to process the time series features provided by the feature extraction layer to obtain a high-level representation of the environmental state;

[0015] The capsule network decision layer is implemented using a capsule network, which is used to process the high-level representation of the environmental state provided by the environmental state recognition layer to obtain the target transparency of the smart material layer and the target opening and closing degree of the drainage valve.

[0016] Furthermore, the recurrent neural network used in the feature extraction layer is specifically a bidirectional long short-term memory network; the bidirectional long short-term memory network includes a first bidirectional LSTM unit layer and a second bidirectional LSTM unit layer; the first bidirectional LSTM unit layer includes a forward LSTM unit and a reverse LSTM unit, and the forward LSTM unit and the reverse LSTM unit each have 128 hidden neural units; the first bidirectional LSTM unit layer is used to receive input time series data of environmental parameters; the second bidirectional LSTM unit layer is connected to the first bidirectional LSTM unit layer and is implemented using the same structure as the first bidirectional LSTM unit layer; the second bidirectional LSTM unit layer receives the output of the first bidirectional LSTM unit layer as its input.

[0017] Furthermore, an adaptive time window mechanism is embedded in the bidirectional long short-term memory network, and the adaptive time window mechanism is used to automatically adjust the time window size of the forward LSTM unit and the reverse LSTM unit in the bidirectional long short-term memory network for processing data based on the change rate of the past and current environmental parameters before each processing cycle starts; the adaptive time window mechanism is implemented by the following steps:

[0018] Select a reference time window, calculate the average value within the reference time window for each environmental parameter; compare the calculated average value with the value at the current time point to obtain the difference between the calculated average value and the value at the current time point; and obtain the change rate index by dividing the difference by the time length of the reference time window;

[0019] Two preset thresholds are set, including a high threshold and a low threshold, which represent the upper and lower limits of the speed of change of environmental parameters respectively; if the change rate index exceeds the high threshold, indicating that the environmental parameters are changing rapidly, the time window size is reduced to 5 minutes to capture fine-grained change information; if the change rate index is lower than the low threshold, indicating that the environmental changes are stable, the time window is increased to 30 minutes to improve data processing efficiency.

[0020] Furthermore, the environment state recognition layer includes a first fully connected layer, a first fully connected layer, a third fully connected layer and an attention mechanism adjustment layer connected in sequence; wherein each fully connected layer is followed by an activation layer for enhancing the nonlinear processing capability of the network; the first fully connected layer has 256 neurons; the second fully connected layer has 128 neurons; and the third fully connected layer has 64 neurons;

[0021] The attention mechanism adjustment layer is directly connected to the third fully connected layer, and is used to automatically identify and weight the key features of the environmental state, thereby improving the transparency of the smart material layer and the accuracy and efficiency of the drainage valve opening and closing decision.

[0022] Furthermore, the capsule network decision layer includes a primary capsule layer, a dynamic routing process, and an output capsule layer;

[0023] The primary capsule layer is used to receive the high-level representation from the environment state recognition layer as input; the primary capsule layer includes 16 capsule units, each of which is responsible for capturing a specific feature combination in the environment state;

[0024] The dynamic routing process sends the output of the primary capsule layer to the output capsule layer through a dynamic routing algorithm;

[0025] The output capsule layer includes two capsule units, which correspond to the target transparency of the smart material layer and the target opening and closing degree of the drainage valve respectively; each capsule unit outputs a vector, and the length of the output vector represents the confidence of the corresponding decision; the direction of the output vector is used to explain the decision details.

[0026] Furthermore, the new type of microporous structure material used in the external water-permeable layer is a porous material with high air permeability and moisture conductivity. This material can block rainwater and large external particle pollutants while allowing air and steam to penetrate to maintain air pressure balance and dryness inside and outside the wall.

[0027] Furthermore, the drainage valve of the drainage module adopts an electric control mechanism, which can quickly adjust the opening and closing degree according to the instructions of the intelligent processing module, thereby effectively draining the water accumulated in the wall and preventing water erosion and accumulation.

[0028] Furthermore, the remote communication module adopts wireless communication technology to support data exchange and command transmission with third-party systems including smart home systems, weather stations and building management centers, thereby realizing remote monitoring and intelligent control of the system.

[0029] The beneficial effects of the present application mainly include: (1) By using the electrochromic material layer, the system can dynamically adjust the transparency of the curtain wall, respond to external light and temperature changes in real time, and effectively control the flow of light and heat into the building. This active management capability significantly improves the energy efficiency of the building and reduces energy consumption by reducing dependence on air conditioning and lighting. (2) By analyzing the data collected by the environmental monitoring module through the intelligent processing module, the system can automatically adjust according to indoor and outdoor environmental parameters to maintain the indoor temperature and light at the most comfortable level, thereby improving the comfort of living and working spaces. (3) The integrated drainage module can adjust the drainage strategy according to real-time weather conditions, prevent water accumulation and leakage problems, protect the building structure from water damage, and also reduce the need for maintenance of the drainage system. (4) The addition of the remote communication module allows building managers to remotely monitor and adjust the status of the curtain wall system, whether it is adjusting the transparency to cope with sudden weather conditions or adjusting the drainage system to cope with expected rainfall, which provides unprecedented flexibility and convenience for building management. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic diagram of an intelligent building curtain wall system provided in the first embodiment of the present application. DETAILED DESCRIPTION

[0031] Many specific details are described in the following description to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present application, so the present application is not limited by the specific implementation disclosed below.

[0032] The first embodiment of the present application provides an intelligent building curtain wall system. Figure 1 , which is a schematic diagram of the first embodiment of the present application. Figure 1 A first embodiment of the present application provides an intelligent building curtain wall system for detailed description.

[0033] The intelligent building curtain wall system includes a curtain wall module 101 with a multi-layer structure, an environmental monitoring module 102, a drainage module 103, an intelligent processing module 104 and a remote communication module 105.

[0034] The multi-layer curtain wall module 101 includes an external water-permeable layer, a thermal insulation layer, a structural support layer, an internal decorative layer and an intelligent material layer; wherein the external water-permeable layer adopts a new material with a microporous structure, which can effectively prevent large particles from entering while allowing water to pass smoothly; the intelligent material layer is made of electrochromic material, which is used to adjust the transparency of the intelligent material layer according to the thermal control instructions of the intelligent processing module, thereby realizing active management of light and heat flow.

[0035] In this embodiment, the core part of the intelligent building curtain wall system is a multi-layered curtain wall module 101, which is designed with five main hierarchical structures, each of which plays a unique and critical role to ensure that the curtain wall system can effectively cope with various environmental conditions while providing efficient energy utilization and a comfortable indoor environment.

[0036] First, the outer permeable layer is located on the outermost side of the curtain wall module. This layer uses a new type of material with a microporous structure, which can block the entry of large particles such as dust and other large suspended particles, while allowing water to pass smoothly. This design not only protects the internal structure of the curtain wall from pollution and damage, but also effectively handles rainwater and moisture, maintaining the air permeability and dryness of the curtain wall. The material with a microporous structure can be a polycarbonate sheet, which is widely used in building curtain walls due to its excellent light transmittance and impact resistance. Through special processing technology, a microporous structure can be formed in the PC board, so that it can maintain light transmittance while also having certain air permeability and drainage functions.

[0037] Secondly, the thermal insulation layer is located inside the outer permeable layer. This layer uses high-efficiency thermal insulation materials to reduce heat loss caused by the temperature difference between the inside and outside of the building, thereby reducing the energy consumption of the building. By optimizing the material and structural design of this layer, the energy efficiency of the entire building can be significantly improved.

[0038] Next, the structural support layer is the skeleton part of the curtain wall system, providing the necessary structural strength and stability. This layer needs to be made of high-strength and durable materials to ensure that the entire curtain wall system can withstand various external pressures, such as wind pressure and earthquake force.

[0039] The interior decoration layer is mainly responsible for the beauty of the curtain wall and the interior visual effect. The materials and designs need to be selected according to the overall style and usage functions of the building to create a comfortable and harmonious indoor environment.

[0040] The smart material layer uses electrochromic material, which is unique in that it can adjust its transparency according to the thermal control instructions issued by the intelligent processing module 104, thereby realizing active management of the light and heat flow inside the building. This means that according to the outdoor light intensity and temperature changes, the smart material layer can dynamically adjust the light and heat passing through the window, which can reduce the rise in indoor temperature on hot summer days and keep the indoor temperature warm in winter, greatly enhancing the energy efficiency and living comfort of the building. Electrochromic materials have been used in the current construction field, and their working principles will not be repeated here.

[0041] Furthermore, the new type of microporous structure material used in the external water-permeable layer is a porous material with high air permeability and moisture conductivity. This material can block rainwater and large external particle pollutants while allowing air and steam to penetrate to maintain air pressure balance and dryness inside and outside the wall.

[0042] In the intelligent building curtain wall system, the design of the external permeable layer is crucial, which is directly related to the building's thermal insulation, moisture retention and air permeability. Especially under changing climate conditions, the material selection and design method of the external permeable layer will determine the durability and comfort of the building curtain wall. This embodiment proposes to use a porous material with high air permeability and moisture conductivity as a new material for the external permeable layer, aiming to provide a solution that can both prevent rain and stains and maintain air circulation inside and outside the wall.

[0043] The key to this porous material is its microporous structure. These tiny pores can effectively block the entry of rainwater and large particles of pollutants, protecting the internal structure of the building from moisture and pollutant damage. At the same time, these micropores also provide passages for air and water vapor, allowing air and moisture inside and outside the wall to pass through, thereby maintaining the air pressure balance inside and outside the wall, reducing the risk of structural damage caused by air pressure differences, and helping the inside of the wall stay dry, preventing structural problems or mildew caused by moisture accumulation.

[0044] In addition, the selection of this new porous material also takes into account its durability and environmental adaptability, ensuring that it can maintain good performance under different environmental conditions (such as high temperature, low temperature, high humidity or dry environment), which not only improves the energy efficiency of the building, but also improves the comfort of the indoor and outdoor environment. This material can be expanded polytetrafluoroethylene (ePTFE), which is known for its unique microporous structure that can effectively block moisture and pollutants while allowing air and water vapor to pass through. ePTFE materials are widely used in outdoor clothing, building coverings and other occasions that require waterproof and breathable.

[0045] By adopting this new porous material with high air permeability and moisture conductivity, the intelligent building curtain wall system can effectively protect the building from the adverse effects of the external environment while maintaining the comfort and health of the internal environment, providing a new solution for modern building design.

[0046] The environmental monitoring module is equipped with a variety of sensors and distributed at key positions of the multi-layer curtain wall module for real-time detection of environmental parameters; the environmental parameters include external environmental parameters and indoor environmental parameters; the external environmental parameters include outdoor temperature, light intensity, wind speed, humidity and rainfall; the indoor environmental parameters include indoor temperature, light intensity and humidity.

[0047] In the intelligent building curtain wall system, the environmental monitoring module 102 plays a vital role. By precisely monitoring the surrounding environment and the conditions inside the building, it provides a data basis for the intelligent response of the entire system. The module uses a series of sensors distributed at key locations of the curtain wall module to capture environmental parameters including temperature, light intensity, wind speed, humidity and rainfall in real time, as well as monitor the corresponding parameters of the indoor environment.

[0048] To ensure the effective operation of the environmental monitoring module, the system uses a variety of sensors. Temperature sensors are used to measure outdoor and indoor temperatures, providing decision data for the transparency adjustment of the smart material layer; light intensity sensors monitor the intensity of daylight to ensure that light management can increase or decrease indoor light as needed; wind speed sensors measure external wind conditions to evaluate and adjust the stability and safety of the curtain wall against wind impacts; humidity sensors monitor the humidity of the air, aiming to adjust the comfort of the indoor air through the smart processing module; and rainfall sensors detect rainfall conditions to provide real-time data support for the operation of the drainage module.

[0049] These sensors are not only placed on the outside of the curtain wall to monitor environmental conditions, but also arranged inside to collect indoor environmental data, so as to ensure that the system can fully understand and adapt to the changes in the indoor and outdoor environments. By collecting this data in real time, the environmental monitoring module transmits the information to the intelligent processing module 104, which analyzes the data and generates corresponding control instructions, such as adjusting the transparency of the smart material layer to control the light and heat flow, or adjusting the opening and closing of the drainage valve to manage rainwater discharge.

[0050] The design and implementation of the environmental monitoring module needs to take into account the accuracy, reliability and long-term stability of the sensors, as well as how they are integrated into the curtain wall system to ensure that they can work seamlessly with other modules in the system. Through precise environmental monitoring and data analysis, the intelligent building curtain wall system can achieve highly automated environmental adaptability, optimize energy utilization, improve the comfort and safety of indoor and outdoor environments, and meet the needs of modern buildings for intelligence and energy saving.

[0051] The drainage module 103 includes a drainage valve and a drainage pipe; wherein the drainage pipe is used to discharge the moisture entering through the external permeable layer to the outside; the drainage valve is located at the entrance of the drainage pipe, and is used to adjust the opening and closing degree of the drainage valve according to the drainage instruction of the intelligent processing module.

[0052] In the present intelligent building curtain wall system, the drainage module 103 is designed and implemented to efficiently manage and guide the moisture captured by the external permeable layer to ensure the protection of the building and the comfort of the indoor environment. The drainage module includes a drainage pipe and a drainage valve, which work together to dynamically adjust the response of the drainage system according to the instructions of the intelligent processing module 104, thereby optimizing the drainage process of moisture.

[0053] The main function of the drain pipe is to safely and effectively guide the captured moisture from the building curtain wall to the designated drainage area, avoid moisture accumulation and penetration, and protect the building structure from water damage. The design of the drain pipe takes into account the optimal path of water flow and drainage efficiency, ensuring that moisture can be quickly discharged under heavy rain or other wet conditions to avoid adverse effects on the curtain wall structure and indoor environment.

[0054] The drain valve is located at the entrance of the drain pipe. Its core function is to adjust the opening and closing degree according to the drainage instructions of the intelligent processing module. This design allows the system to automatically adjust the drainage speed and amount according to real-time environmental monitoring data and changes in indoor and outdoor environmental conditions, such as rainfall and humidity levels, to optimize drainage efficiency. By precisely controlling the opening and closing degree of the drain valve, the system can flexibly respond to different weather conditions, ensure the efficient operation of the drainage system, and reduce energy consumption.

[0055] The intelligent processing module 104 plays a central role in this process. It receives and analyzes data from the environmental monitoring module 102, such as rainfall and humidity. Based on this information, the intelligent processing module calculates the optimal opening and closing degree of the drainage valve and generates corresponding drainage instructions. These instructions are then sent to the drainage module to guide the drainage valve to adjust to the appropriate position to cope with current and predicted environmental conditions.

[0056] Furthermore, the drainage valve of the drainage module adopts an electric control mechanism, which can quickly adjust the opening and closing degree according to the instructions of the intelligent processing module, thereby effectively draining the water accumulated in the wall and preventing water erosion and accumulation.

[0057] The drainage module in the intelligent building curtain wall system is a key component designed to protect the structural safety of the building and the comfort of the internal environment. In this system, the drainage module contains a key element - an electrically controlled drainage valve. The design of this valve enables it to quickly and accurately adjust its opening and closing state according to specific instructions from the intelligent processing module, thereby effectively controlling the discharge of water from the wall and preventing water erosion and accumulation from damaging the building structure.

[0058] The core of the electric control mechanism is its ability to receive and execute electronic signals from the intelligent processing module, which is based on real-time monitoring of environmental parameters outside and inside the building. For example, when the environmental monitoring module detects an increase in rainfall and the wall may be flooded, the intelligent processing module will calculate an appropriate drain valve opening and closing degree and send this instruction to the electric control system of the drain valve.

[0059] The electric control mechanism usually includes a motor, a transmission device (such as a gear box), and an associated circuit board, which work together to accurately control the position of the drain valve. The motor adjusts the speed and direction according to the instructions received, which are converted into the opening and closing action of the valve through the transmission device, thereby controlling the drainage rate and drainage volume to cope with different environmental conditions.

[0060] In addition, the design of the electrically controlled drainage valve also takes into account rapid response in emergency situations. In extreme weather conditions or when the drainage system is detected to be blocked, the intelligent processing module can quickly adjust the drainage valve to a fully open state to maximize the removal of accumulated water and ensure building safety.

[0061] The intelligent processing module 104 is used to receive the environmental parameters provided by the environmental monitoring module; determine the target transparency of the intelligent material layer and the target opening and closing degree of the drainage valve according to the environmental parameters; generate thermal control instructions and drainage instructions according to the target transparency and the target opening and closing degree of the drainage valve respectively; send the generated thermal control instructions to the curtain wall module of the multi-layer structure; and send the generated drainage instructions to the drainage module.

[0062] In this intelligent building curtain wall system, the intelligent processing module 104 plays the role of the central nervous system, responsible for integrating and analyzing the data from the environmental monitoring module 102, and formulating and issuing execution strategies based on these data to ensure the optimal performance of the curtain wall system. The data received by the intelligent processing module include but are not limited to external environmental parameters such as temperature, light intensity, wind speed, humidity and rainfall, as well as corresponding indoor environmental parameters. These data provide the module with the necessary information to accurately control the response of the curtain wall system, such as adjusting the transparency of the intelligent material layer or the opening and closing of the drainage system to adapt to the changing environmental conditions.

[0063] The intelligent processing module has highly advanced data processing capabilities and can use algorithms to analyze the collected environmental data to identify trends and patterns of environmental changes. Based on these analysis results, the module will determine the target transparency of the smart material layer and the specific opening and closing degree of the drainage valve. In this process, the intelligent processing module takes into account a variety of factors, including but not limited to current weather conditions, expected weather changes, indoor and outdoor temperature differences, and light requirements during specific time periods.

[0064] Once the target transparency and the opening and closing degree of the drainage valve are determined, the intelligent processing module will generate corresponding thermal control instructions and drainage instructions. These instructions are sent to the corresponding components through the communication network within the system: the thermal control instructions are sent to the curtain wall modules of the multi-layer structure, especially the smart material layer, to adjust its transparency; the drainage instructions are sent to the drainage module, especially the drainage valve, to control the speed and amount of water discharge.

[0065] In order to achieve remote monitoring and adjustment control, the intelligent processing module is also connected to the remote communication module 105. This connection allows the system to not only operate autonomously and make adjustments based on real-time data, but also receive remote commands from building managers or maintenance personnel, allowing them to manually adjust the system when necessary. This design improves the flexibility and controllability of the system, ensuring that the building curtain wall system can maintain optimal performance under various circumstances.

[0066] The intelligent processing module can determine the target transparency of the smart material layer and the target opening degree of the drainage valve according to preset logical rules or thresholds, and the logical rules or thresholds include: when the external light intensity exceeds a preset high threshold, reducing the transparency of the smart material layer to reduce the indoor light and heat entering; when the external light intensity is lower than the preset low threshold, increasing the transparency to allow more natural light to enter; when the temperature difference between indoor and outdoor is greater than a preset value, adjusting the transparency to control the heat flow according to whether insulation or heat dissipation is desired; when the rainfall detected by the environmental monitoring module exceeds a preset threshold, fully opening the drainage valve to quickly drain the accumulated water; when the rainfall is less than the preset threshold, maintaining the drainage valve in a slightly open state; and when the outdoor humidity is significantly higher than the indoor humidity and there is a risk of water accumulation, fully opening the drainage valve to prevent water seepage.

[0067] Determining the target transparency of the smart material layer and the target opening and closing degree of the drainage valve can be achieved through a series of simple logic rules or threshold judgments. This method is based on preset parameter ranges and conditional responses, allowing the system to make predefined responses under specific conditions. The following describes how to determine these two target states based on external environmental parameters and indoor environmental parameters.

[0068] The target transparency of the smart material layer may be determined by at least one of the following high and low thresholds of light intensity and preset values ​​of indoor and outdoor temperature differences:

[0069] 1. Light intensity: If the external light intensity exceeds the preset high threshold, it indicates that the light is sufficient or too strong. At this time, the transparency of the smart material layer can be reduced to reduce the indoor light and heat entering, maintain indoor comfort and reduce cooling requirements. If the light intensity is lower than the preset low threshold, it indicates that the external light is insufficient. At this time, the transparency can be increased to allow more natural light to enter, improve indoor lighting and save energy.

[0070] 2. Indoor and outdoor temperature difference: If the indoor and outdoor temperature difference is greater than the preset value, it means that the external temperature conditions are far different from the desired indoor temperature. At this time, according to whether you want to keep warm or dissipate heat, adjust the transparency to control the heat flow. For example, if you want to keep the room warm in winter, increase the transparency during the day to absorb more sunlight.

[0071] The target opening degree of the drain valve may be determined by at least one of the following rainfall and humidity thresholds:

[0072] 1. Rainfall: If the rainfall detected by the environmental monitoring module exceeds the preset threshold, it indicates that there is a large amount of rainwater that needs to be discharged. At this time, the drainage valve should be fully opened to quickly discharge the accumulated water. If the rainfall is small, maintaining the valve in a slightly open state is sufficient to handle less rainwater discharge needs.

[0073] 2. Humidity: The difference in humidity between indoors and outdoors can also be a basis for adjusting the opening and closing of the drain valve. If the outdoor humidity is significantly higher than the indoor humidity and there is a risk of water accumulation, the drain valve may need to be fully opened to prevent water seepage.

[0074] By setting appropriate thresholds and adjusting the transparency of the smart material layer and the opening and closing of the drainage valve according to these simple logical rules, the environmental parameters can also be effectively used to perform basic intelligent control of the curtain wall system. The advantages of this method are simple implementation, easy understanding and operation, and suitable for the initial realization of intelligent control of building curtain walls.

[0075] Furthermore, the intelligent processing module uses a trained neural network model to determine the target transparency of the intelligent material layer and the target opening and closing degree of the drainage valve according to the environmental parameters; the neural network model includes a feature extraction layer, an environmental state recognition layer and a capsule network decision layer;

[0076] The feature extraction layer is implemented by a recurrent neural network, which processes the time series data of the input environmental parameters, captures and learns the dynamic characteristics of the external environmental parameters and the indoor environmental parameters changing over time, and obtains the time series characteristics of the environmental parameters;

[0077] The environmental state recognition layer is implemented by multiple fully connected layers, which are used to process the time series features provided by the feature extraction layer to obtain a high-level representation of the environmental state;

[0078] The capsule network decision layer is implemented using a capsule network, which is used to process the high-level representation of the environmental state provided by the environmental state recognition layer to obtain the target transparency of the smart material layer and the target opening and closing degree of the drainage valve.

[0079] The feature extraction layer plays a vital role, and its main responsibility is to extract useful information from the input environmental parameter time series data. To achieve this goal, a recurrent neural network (RNN) is used as the core technology. RNN is particularly suitable for processing time series data because it can remember the output of the previous moment and affect the output of the current moment, which makes RNN very suitable for capturing and learning the dynamic change characteristics in the time series.

[0080] Specifically in this system, the RNN of the feature extraction layer receives environmental parameter data from the outside and inside of the building, such as temperature, humidity, light intensity, etc. These data change over time and exist in the form of time series. Through its internal structure, RNN gradually processes these time series data and effectively captures the patterns and trends of these environmental parameters changing over time.

[0081] To ensure efficient feature extraction, RNN uses a series of hierarchical structures when processing data, and each layer further analyzes and refines the data. This hierarchical structure not only increases the depth of the network, but also enhances the network's ability to extract deep features. In this way, the feature extraction layer ultimately generates a set of high-quality feature vectors that represent the dynamic changes in the environment state.

[0082] These feature vectors are then passed to the environment state recognition layer, which further analyzes these features and further processes and simplifies them through multiple fully connected layers to obtain a high-level representation of the environment state. This high-level representation captures the core features of the environment state and provides an important information basis for the final decision.

[0083] Furthermore, the recurrent neural network used in the feature extraction layer is specifically a bidirectional long short-term memory network; the bidirectional long short-term memory network includes a first bidirectional LSTM unit layer and a second bidirectional LSTM unit layer; the first bidirectional LSTM unit layer includes a forward LSTM unit and a reverse LSTM unit, and the forward LSTM unit and the reverse LSTM unit each have 128 hidden neural units; the first bidirectional LSTM unit layer is used to receive input time series data of environmental parameters; the second bidirectional LSTM unit layer is connected to the first bidirectional LSTM unit layer and is implemented using the same structure as the first bidirectional LSTM unit layer; the second bidirectional LSTM unit layer receives the output of the first bidirectional LSTM unit layer as its input.

[0084] An adaptive time window mechanism is embedded in the bidirectional long short-term memory network. The adaptive time window mechanism is used to automatically adjust the time window size of the forward LSTM unit and the reverse LSTM unit in the bidirectional long short-term memory network for processing data based on the change rate of the past and current environmental parameters before each processing cycle starts. The adaptive time window mechanism is implemented by the following steps:

[0085] Select a reference time window, calculate the average value within the reference time window for each environmental parameter; compare the calculated average value with the value at the current time point to obtain the difference between the calculated average value and the value at the current time point; and obtain the change rate index by dividing the difference by the time length of the reference time window;

[0086] Two preset thresholds are set, including a high threshold and a low threshold, which represent the upper and lower limits of the speed of change of environmental parameters respectively; if the change rate index exceeds the high threshold, indicating that the environmental parameters are changing rapidly, the time window size is reduced to 5 minutes to capture fine-grained change information; if the change rate index is lower than the low threshold, indicating that the environmental changes are stable, the time window is increased to 30 minutes to improve data processing efficiency.

[0087] In the intelligent building curtain wall system, one of the core components is a recurrent neural network specifically designed to process environmental parameter data, especially a bidirectional long short-term memory network (Bi-LSTM). This network design allows the system to not only understand the changes in environmental parameters over time, but also capture the dynamics of reverse time, ensuring a comprehensive understanding of environmental changes.

[0088] This network consists of two layers, each of which contains forward and reverse LSTM units, each with 128 hidden neurons. These units work together to receive and process input environmental parameter time series data, such as temperature, humidity, or light, which record the changes in environmental conditions over time in the form of time series. The bidirectional LSTM units in the first layer focus on initially capturing and analyzing the time-dependent characteristics in the data, and their output is then passed to the second layer, which further strengthens the extraction and analysis of these characteristics to generate more refined time series features.

[0089] In order to make this processing more sophisticated and adaptable, the system introduces an innovative technology - adaptive time window mechanism. This mechanism can automatically adjust the time window size of data processing according to the rate of change of past and current environmental parameters before each processing cycle. Specifically, the system will first select a baseline time window to calculate the average change value of the environmental parameter, and then compare this average value with the current environmental parameter value to calculate a change rate indicator through this comparison. Then, the system compares this change rate indicator with the preset high and low thresholds to decide whether to reduce the time window to 5 minutes to capture rapidly changing details, or increase the time window to 30 minutes to improve data processing efficiency and reduce resource consumption.

[0090] This adaptive time window mechanism not only improves the system's response speed and sensitivity to environmental changes, but also optimizes the data processing efficiency of the entire network by dynamically adjusting the processing strategy. This design ensures that the intelligent building curtain wall system can accurately adjust the transparency of the intelligent material layer and the opening and closing degree of the drainage valve under different environmental changes, thereby effectively managing the building's energy consumption and internal environmental comfort.

[0091] The following is the reference implementation code of the feature extraction layer:

[0092]

[0093]

[0094] In this example, a hypothetical `adjust_input_data_based_on_rate_change` function is used to directly adjust the sampling frequency of the network input data according to the rate of change. In this way, the adaptive time window mechanism directly affects how the data is processed by the network, making the model more flexible in responding to environmental changes. It should be noted that the `adjust_input_data_based_on_rate_change` function and data sampling strategy here are only to illustrate how to achieve the direct association between the adaptive time window mechanism and the network input, and are not the only or optimal implementation method. In actual applications, the adjustment strategy and implementation details may vary depending on the specific environmental parameters and requirements.

[0095] Furthermore, the environment state recognition layer includes a first fully connected layer, a first fully connected layer, a third fully connected layer and an attention mechanism adjustment layer connected in sequence; wherein each fully connected layer is followed by an activation layer for enhancing the nonlinear processing capability of the network; the first fully connected layer has 256 neurons; the second fully connected layer has 128 neurons; and the third fully connected layer has 64 neurons;

[0096] The attention mechanism adjustment layer is directly connected to the third fully connected layer, and is used to automatically identify and weight the key features of the environmental state, thereby improving the transparency of the smart material layer and the accuracy and efficiency of the drainage valve opening and closing decision.

[0097] The environmental state recognition layer plays a crucial role. The main task of this layer is to further process and refine the time series features from the feature extraction layer, and finally generate a high-level representation of the environmental state. This high-level representation not only captures the key information of environmental parameters, but also provides the necessary input for the decision-making layer, enabling the system to accurately determine the target transparency of the smart material layer and the target opening degree of the drainage valve.

[0098] Specifically, the environmental state recognition layer consists of a series of fully connected layers and an attention mechanism adjustment layer. These fully connected layers are connected in sequence, and each layer is connected to an activation layer to enhance the nonlinear processing capabilities of the network. The first fully connected layer is configured with 256 neurons and receives the time series features output by the feature extraction layer. Through this layer of processing, the data is converted into an intermediate representation that captures preliminary environmental state information. Next, the second fully connected layer further processes this data, with 128 neurons, which helps the system focus on more important features by reducing the number of neurons. Then, the third fully connected layer, with 64 neurons, continues to streamline and refine features to provide a more accurate representation of the environmental state for the final decision.

[0099] The most critical part is the attention mechanism adjustment layer, which is directly connected to the third fully connected layer. This layer uses the attention mechanism to automatically identify and weight the key features of the environment state. By learning which features are most important to the decision-making process, this layer can adjust the weights of the features to ensure that the decision-making process pays more attention to information that has a significant impact on the transparency of the smart material layer and the opening and closing of the drain valve. This not only improves the accuracy of decision-making, but also improves the efficiency of system processing.

[0100] Through such a structural design, the environmental state recognition layer can effectively process and transform the data provided by the feature extraction layer to generate a high-level representation of the environmental state. In addition, the introduced attention mechanism adjustment layer further enhances the system's ability to identify and utilize key information, providing high-quality input for the capsule network decision layer. This design not only ensures the accuracy and response speed of the intelligent building curtain wall system decision, but also provides an effective data processing process, ensuring the system's adaptability and flexibility in the face of complex environmental changes.

[0101] The following is the reference implementation code of the environment state recognition layer. The code uses Python and assumes that the TensorFlow and Keras frameworks are used to build the model. This code example provides a sequence connection of the fully connected layer and a basic attention mechanism model implementation.

[0102]

[0103]

[0104]

[0105] In this code, `AttentionLayer` is a custom attention mechanism layer that automatically identifies and weights the key features of the environment state by learning the weights of each feature. This simple attention mechanism layer is used to improve the accuracy and efficiency of the smart material layer transparency and drain valve opening and closing decision. By connecting this attention mechanism layer directly to the third fully connected layer, it ensures that the model pays more attention to information that has a significant impact on the final decision.

[0106] The capsule network decision layer is implemented using a capsule network, which is used to process the high-level representation of the environmental state provided by the environmental state recognition layer to obtain the target transparency of the smart material layer and the target opening and closing degree of the drainage valve.

[0107] Furthermore, the capsule network decision layer includes a primary capsule layer, a dynamic routing process, and an output capsule layer;

[0108] The primary capsule layer is used to receive the high-level representation from the environment state recognition layer as input; the primary capsule layer includes 16 capsule units, each of which is responsible for capturing a specific feature combination in the environment state;

[0109] The dynamic routing process sends the output of the primary capsule layer to the output capsule layer through a dynamic routing algorithm;

[0110] The output capsule layer includes two capsule units, which correspond to the target transparency of the smart material layer and the target opening and closing degree of the drainage valve respectively; each capsule unit outputs a vector, and the length of the output vector represents the confidence of the corresponding decision; the direction of the output vector is used to explain the decision details.

[0111] The capsule network decision layer plays a vital role. It is responsible for the final decision output, that is, determining the target transparency of the smart material layer and the target opening and closing degree of the drainage valve. The capsule network decision layer is designed using capsule network technology, which can more effectively capture the spatial hierarchical relationship of entities and provide more accurate and detailed decision-making basis compared to traditional neural networks.

[0112] The capsule network decision layer mainly consists of three parts: the primary capsule layer, the dynamic routing process, and the output capsule layer. Each part plays a specific role and works together to generate the final decision output.

[0113] The primary capsule layer is the first stop in the capsule network decision layer, which receives high-level representations from the environment state recognition layer as input. This layer contains 16 capsule units, each of which is designed to capture a specific combination of features in the environment state. These capsule units are able to understand complex patterns in the input data and encode these patterns into a series of vector forms, providing a rich information basis for the subsequent decision-making process.

[0114] The dynamic routing process is the core of the capsule network decision layer, which is responsible for processing the output of the primary capsule layer and transmitting these outputs to the output capsule layer. Through the dynamic routing algorithm, the system can adaptively adjust the connection weights between capsules according to the characteristics of the input data. This process ensures that the network can efficiently focus on the features that are most important for the final decision, while ignoring irrelevant or minor information.

[0115] The output capsule layer is the final stage of the capsule network decision layer. It contains two capsule units, corresponding to the target transparency of the smart material layer and the target opening degree of the drain valve. Each capsule unit outputs a vector whose length represents the confidence of the corresponding decision, while the direction of the vector is used to explain the details of the decision. This output method not only provides the system with the specific value of the decision, but also provides interpretable information about the decision, making the decision process of the system more transparent and credible.

[0116] Through the above design, the capsule network decision layer can make full use of the high-level representation obtained from the environmental state recognition layer, conduct in-depth analysis and processing, and finally generate accurate and detailed decision outputs. This decision output is not only based on the understanding of the deep-level characteristics of the environmental state, but also takes into account the complex relationship between these characteristics, thereby ensuring that the intelligent building curtain wall system can make the most appropriate adjustments under various environmental conditions to achieve the purpose of energy saving and comfort improvement.

[0117] The following is the reference code of the capsule network decision layer. This code uses the TensorFlow framework to implement the basic structure of the capsule network decision layer, including the primary capsule layer, dynamic routing process, and output capsule layer.

[0118]

[0119]

[0120]

[0121] input_shape = (28,28,1) # Example input shape

[0122] n_class=2#Example: Two decisions on the transparency of the smart material layer and the opening and closing degree of the drainage valve

[0123] model=CapsuleNetwork(input_shape,n_class,num_routing=3)

[0124] model.summary()

[0125] In the above code, the `PrimaryCapsule` function constructs a primary capsule layer that uses convolution operations to capture the features of the input and converts it into a series of capsule vectors. The `CapsuleLayer` class implements a custom capsule layer that contains the framework of the dynamic routing algorithm (the actual dynamic routing algorithm details are omitted in this code). Finally, the `CapsuleNetwork` function integrates these components into a complete capsule network model for the final decision output.

[0126] To implement the details of the dynamic routing algorithm, it is necessary to define in detail how capsules interact with each other and how to iteratively update the connection weights to achieve the optimal routing effect. The following is an implementation detail of a dynamic routing algorithm based on TensorFlow, which is used in capsule networks. This implementation extends the capsule network structure code provided previously.

[0127] The key to the dynamic routing algorithm is to calculate the "agreement" between capsules and update the connection weights accordingly. The algorithm is performed iteratively, and each iteration updates the weights to better route the input capsule to the correct output capsule.

[0128]

[0129]

[0130] In this implementation, `CapsuleLayer` is a custom capsule layer that contains the core logic of the dynamic routing algorithm. During each iteration, the "soft assignment" weights `c_ij` are first calculated, which is achieved by applying the softmax function to `b_ij`. These weights determine the contribution of the input capsule to the output capsule. Next, the weighted sum `s_j` is calculated and the squash function is applied to obtain the output capsule `v_j`. Except for the last iteration, `b_ij` is updated according to the agreement between the current output capsule `v_j` and the input capsule's prediction `inputs_hat` to proceed to the next round of iteration.

[0131] In the above code, the implementation of the dynamic routing algorithm in the capsule network layer can be summarized as the following steps:

[0132] 1. Initialization: First, a capsule layer class `CapsuleLayer` is defined, which receives several key parameters, including the number of capsules `num_capsule`, the dimension of each capsule `dim_capsule`, and the number of routing iterations `num_routing`. This class creates a weight matrix `kernel` when constructed. This matrix is ​​the weight connecting the previous layer capsule and the current layer capsule. Its shape depends on the shape of the input data and the configuration of the capsule.

[0133] 2. Data processing: In the `call` method, the input data is first dimensionally expanded so that it can be matrix multiplied with the weight matrix `kernel`. In this way, the input data is converted into `inputs_hat`, the prediction vector, which represents the prediction of the activation vector of the next layer of capsules.

[0134] 3. Dynamic routing: Next is the core part of the dynamic routing algorithm. First, initialize a log probability of all zeros (b_ij) to represent the initial probability of connection between capsules. Then, the algorithm enters the iterative process. In each iteration:

[0135] - Calculate the probability of the current connection (c_ij) using the `softmax` function. This is achieved by applying `softmax` to `b_ij`, ensuring that the weights of all prediction vectors received by each lower layer capsule sum to 1.

[0136] - The total input `s_j` for each capsule is calculated by weighted summing (weights provided by c_ij) all input prediction vectors `inputs_hat`, and then applying the `squash` function to obtain the output vector `v_j`, which makes the vector length compressed between 0 and 1 while preserving the directional information.

[0137] -If it is not the last iteration, the algorithm updates `b_ij` to increase the weights of those prediction vectors that have a higher consistency (i.e. "agreement") with the current output vector `v_j`.

[0138] 4. Output: After a set number of iterations, the final output vector `v_j` is obtained by compressing `s_j`, which represents the activation state of the capsule in the current layer. These activation vectors contain the capsule's understanding of the input data, including the probability of the entity (the length of the vector) and the parameters of the entity (the direction of the vector).

[0139] The dynamic routing process optimizes the connection weights through multiple iterations, allowing the network to dynamically determine which lower-level capsules contribute more to higher-level capsules, thereby preserving the spatial hierarchical relationship without using a maximum pooling layer. This is the key feature that distinguishes capsule networks from traditional convolutional neural networks.

[0140] For the neural network model provided in this embodiment, including a feature extraction layer, an environment state recognition layer and a capsule network decision layer, the training steps include:

[0141] Step 1: Data preparation and preprocessing

[0142] First, collect relevant environmental parameter data from the building environment, including but not limited to temperature, humidity, light intensity, rainfall, etc., which are time-varying sequence data. The data is cleaned, denoised, and normalized through preprocessing steps to ensure the quality of the input data. For time series data, it may also be necessary to split the time window and prepare it into a format suitable for recurrent neural network processing.

[0143] Step 2: Model training

[0144] 1. Feature extraction layer training: Taking the time series data of environmental parameters as input, a recurrent neural network (such as LSTM or GRU) is used to learn the dynamic characteristics of external environmental parameters and indoor environmental parameters changing over time. In this stage, the network focuses on extracting useful features from the original time series, such as patterns and trends of environmental changes.

[0145] 2. Training of the environment state recognition layer: Next, the time series features learned by the feature extraction layer are passed to the environment state recognition layer, which consists of multiple fully connected layers. The training goal of this stage is to further process and refine these features to generate a high-level representation of the environment state. The feature information can be gradually abstracted and compressed by reducing the number of neurons layer by layer, and finally a compact representation of the environment state is formed.

[0146] 3. Capsule network decision layer training: Finally, the high-level representation of the environment state is sent to the capsule network decision layer, which is processed by the capsule network's dynamic routing mechanism and ultimately outputs the target transparency of the smart material layer and the target opening and closing degree of the drainage valve. The training of this layer needs to focus on optimizing the dynamic routing process to ensure the accuracy and efficiency of decision-making.

[0147] Step 3: Model evaluation and tuning

[0148] After training is completed, the validation set and test set are used to evaluate the performance of the model, including key indicators such as decision accuracy and response speed. Tuning is performed based on the evaluation results, which may involve adjusting the network structure, optimizing the routing algorithm, and other steps to improve the overall performance of the model.

[0149] Step 4: Model deployment and monitoring

[0150] Deploy the trained and fully verified model to the actual building management system, process environmental parameter data in real time, and dynamically adjust the state of the building curtain wall according to the model's prediction results. At the same time, continuously monitor the operation and performance of the model, and collect new environmental data for continuous optimization and iterative updates of the model.

[0151] Through this series of training and implementation steps, we ensure that the neural network model can make decisions accurately and efficiently based on environmental changes, realize adaptive adjustment of the intelligent building curtain wall system, and improve the building's energy efficiency and living comfort.

[0152] The remote communication module 105 is connected with the intelligent processing module and the third-party system to realize the remote monitoring and adjustment control of the intelligent building curtain wall system.

[0153] The remote communication module 105 plays a bridge role in the intelligent building curtain wall system. It enables the system to not only automatically adjust its own state according to environmental data, but also receive instructions from remote operators to achieve a higher level of interactivity and control flexibility. This module allows building managers or maintenance teams to monitor and adjust the operating status of the curtain wall system from a remote location through connection with the intelligent processing module 104 and third-party systems, including but not limited to adjusting the transparency of the intelligent material layer, the operation of the drainage system, etc.

[0154] The remote communication module uses modern communication technologies, such as wireless networks, Bluetooth, the Internet or dedicated networks, to ensure the stability and security of data transmission. This means that no matter where the operator is, as long as he can connect to the network, he can receive real-time system status information, including environmental monitoring data and system performance indicators, and can also send control instructions to adjust system settings to cope with sudden environmental changes or special needs.

[0155] In order to achieve this function, the remote communication module 105 integrates necessary hardware and software components. The hardware components include but are not limited to communication interfaces, data encryption modules, signal converters, etc., which work together to ensure accurate information transmission and system security. The software components include communication protocols, data processing algorithms, user interfaces, etc., so that remote operators can easily monitor system status, perform data analysis, and issue instructions.

[0156] The design of the remote communication module also takes into account compatibility with third-party systems, ensuring that the intelligent curtain wall system can be seamlessly integrated into a wider range of building management systems, such as building automation systems, energy management systems, etc. This not only increases the application scope of the system, but also improves the overall intelligence level of the building.

[0157] Through the above design, the remote communication module 105 becomes an indispensable part of the intelligent building curtain wall system, which not only enhances the autonomy and intelligence of the system, but also provides unprecedented control convenience and flexibility for building managers.

[0158] Furthermore, the remote communication module adopts wireless communication technology to support data exchange and command transmission with third-party systems including smart home systems, weather stations and building management centers, thereby realizing remote monitoring and intelligent control of the system.

[0159] In the intelligent building curtain wall system, the remote communication module plays a vital role. It not only enables the system to exchange data and transmit instructions in real time, but also expands the system's functions and application scope. By adopting advanced wireless communication technology, the module can seamlessly connect to multiple third-party systems including smart home systems, weather stations and building management centers, thereby realizing remote monitoring and intelligent control of the system.

[0160] The selection of wireless communication technologies covers a variety of standards and protocols, such as Wi-Fi, Bluetooth, Zigbee, LTE, etc., to adapt to different communication needs and environmental conditions. These technologies not only ensure the high efficiency and stability of data transmission, but also ensure the security of the communication process, and prevent data from being unauthorized access or tampering through encryption and authentication mechanisms.

[0161] The remote communication module is designed with ease of use and compatibility in mind, enabling it to be interconnected with a variety of standard smart home devices, weather monitoring equipment, and building management software platforms. In this way, the system can not only receive real-time weather information from the weather station and automatically adjust the thermal control and drainage strategies of the curtain wall according to weather changes, but also adjust the transparency of the curtain wall to maximize natural light utilization and visual comfort according to user preferences and behavior patterns in the smart home system.

[0162] In addition, through the remote communication module, the building management center can remotely monitor the operating status of the curtain wall system, obtain real-time environmental parameter data and system performance reports, and promptly discover and deal with potential problems. Managers can also issue control instructions remotely, such as adjusting the curtain wall protection status in advance when extreme weather events are predicted, or remotely activating the drainage system to prevent water accumulation in certain circumstances.

[0163] In order to realize these functions, detailed planning and configuration are required during the system design and implementation phase, including selecting appropriate wireless communication technologies and equipment, setting up secure communication protocols and encryption mechanisms, and ensuring compatibility and interoperability with third-party systems. At the same time, the system should have a certain degree of fault tolerance and redundancy design to ensure that basic communication and control functions can be maintained when the network is unstable or the equipment fails.

[0164] Through this design, the intelligent building curtain wall system can effectively utilize the latest wireless communication technology, which not only realizes efficient management and intelligent regulation of the system, but also provides a higher level of intelligent and automated services for the building, improving the building's energy efficiency, safety and living comfort.

[0165] Although the present application is disclosed as above in the form of a preferred embodiment, it is not intended to limit the present application. Any technical personnel in this field may make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be based on the scope defined by the claims of the present application.

Claims

1. An intelligent building curtain wall system, characterized in that: include: The multi-layer curtain wall module includes an external water-permeable layer, a thermal insulation layer, a structural support layer, an internal decorative layer and an intelligent material layer; wherein the external water-permeable layer is made of a new material with a microporous structure, which can effectively prevent large particles from entering while allowing water to pass smoothly; the intelligent material layer is made of an electrochromic material, which is used to adjust the transparency of the intelligent material layer according to the thermal control instructions of the intelligent processing module, so as to realize active management of light and heat flow; An environmental monitoring module is configured with a variety of sensors, which are distributed at key positions of the multi-layer curtain wall module and are used to detect environmental parameters in real time; the environmental parameters include external environmental parameters and indoor environmental parameters; the external environmental parameters include outdoor temperature, light intensity, wind speed, humidity and rainfall; the indoor environmental parameters include indoor temperature, light intensity and humidity; The drainage module comprises a drainage valve and a drainage pipe; wherein the drainage pipe is used to discharge the water entering through the external water-permeable layer to the outside; the drainage valve is located at the entrance of the drainage pipe, and is used to adjust the opening and closing degree of the drainage valve according to the drainage instruction of the intelligent processing module; The intelligent processing module is used to receive the environmental parameters provided by the environmental monitoring module; determine the target transparency of the intelligent material layer and the target opening and closing degree of the drainage valve according to the environmental parameters; generate a thermal control instruction and a drainage instruction according to the target transparency and the target opening and closing degree of the drainage valve; send the generated thermal control instruction to the curtain wall module of the multi-layer structure; and send the generated drainage instruction to the drainage module; Remote communication module, connected with intelligent processing module and third-party system, to realize remote monitoring and adjustment control of intelligent building curtain wall system; The intelligent processing module uses a trained neural network model to determine the target transparency of the intelligent material layer and the target opening and closing degree of the drainage valve according to the environmental parameters; the neural network model includes a feature extraction layer, an environmental state recognition layer and a capsule network decision layer; The feature extraction layer is implemented by a recurrent neural network, which processes the time series data of the input environmental parameters, captures and learns the dynamic characteristics of the external environmental parameters and the indoor environmental parameters changing over time, and obtains the time series characteristics of the environmental parameters; The environmental state recognition layer is implemented by multiple fully connected layers, which are used to process the time series features provided by the feature extraction layer to obtain a high-level representation of the environmental state; The capsule network decision layer is implemented by a capsule network, and is used to process the high-level representation of the environmental state provided by the environmental state recognition layer to obtain the target transparency of the smart material layer and the target opening and closing degree of the drainage valve; Among them, the recurrent neural network used in the feature extraction layer is specifically a bidirectional long short-term memory network; the bidirectional long short-term memory network includes a first bidirectional LSTM unit layer and a second bidirectional LSTM unit layer; the first bidirectional LSTM unit layer includes a forward LSTM unit and a reverse LSTM unit, and the forward LSTM unit and the reverse LSTM unit each have 128 hidden neural units; the first bidirectional LSTM unit layer is used to receive input time series data of environmental parameters; the second bidirectional LSTM unit layer is connected to the first bidirectional LSTM unit layer and is implemented using the same structure as the first bidirectional LSTM unit layer; the second bidirectional LSTM unit layer receives the output of the first bidirectional LSTM unit layer as its input; Wherein, an adaptive time window mechanism is embedded in the bidirectional long short-term memory network, and the adaptive time window mechanism is used to automatically adjust the time window size of the forward LSTM unit and the reverse LSTM unit in the bidirectional long short-term memory network for processing data based on the change rate of the past and current environmental parameters before each processing cycle starts; the adaptive time window mechanism is implemented by the following steps: Select a reference time window, calculate the average value within the reference time window for each environmental parameter; compare the calculated average value with the value at the current time point to obtain the difference between the calculated average value and the value at the current time point; and obtain the change rate index by dividing the difference by the time length of the reference time window; Set two preset thresholds, including a high threshold and a low threshold, which represent the upper and lower limits of the environmental parameter change rate respectively; if the change rate index exceeds the high threshold, indicating that the environmental parameters are changing rapidly, the time window size is reduced to 5 minutes to capture fine-grained change information; if the change rate index is lower than the low threshold, indicating that the environmental changes are stable, the time window is increased to 30 minutes to improve data processing efficiency; The environment state recognition layer includes a first fully connected layer, a second fully connected layer, a third fully connected layer and an attention mechanism adjustment layer connected in sequence; each fully connected layer is followed by an activation layer to enhance the nonlinear processing capability of the network; the first fully connected layer has 256 neurons; the second fully connected layer has 128 neurons; and the third fully connected layer has 64 neurons; The attention mechanism adjustment layer is directly connected to the third fully connected layer, and is used to automatically identify and weight the key features of the environmental state, thereby improving the transparency of the smart material layer and the accuracy and efficiency of the drainage valve opening and closing decision; The capsule network decision layer includes a primary capsule layer, a dynamic routing process, and an output capsule layer; The primary capsule layer is used to receive the high-level representation from the environment state recognition layer as input; the primary capsule layer includes 16 capsule units, each of which is responsible for capturing a specific feature combination in the environment state; The dynamic routing process sends the output of the primary capsule layer to the output capsule layer through a dynamic routing algorithm; The output capsule layer includes two capsule units, which correspond to the target transparency of the smart material layer and the target opening and closing degree of the drainage valve respectively; each capsule unit outputs a vector, and the length of the output vector represents the confidence of the corresponding decision; the direction of the output vector is used to explain the decision details; The dynamic routing algorithm specifically includes the following steps: Expand the dimension of the data input to the capsule, generate a prediction vector, and initialize the log probability matrix of the connection weights between capsules to an all-zero matrix; During each iteration, the connection weights between the input capsule and the output capsule are calculated by applying the softmax function to the log probability matrix, where the sum of the weights is 1; Apply weights to the input prediction vectors and perform a weighted sum to calculate the sum of the input vectors for each output capsule; Apply the squash function to the sum of the input vectors to compress the vector of the output capsule to between 0 and 1 to ensure that its length can represent the probability information of the entity; At the end of each iteration except the last one, the agreement between the input prediction vector and the current output vector is calculated, and the log probability matrix is ​​updated to increase the weight of the prediction vector that is consistent with the current output; After a preset number of iterations, a final capsule vector is output, where the final capsule vector represents the final activation state of the capsule layer.

2. The intelligent building curtain wall system according to claim 1, characterized in that: The new type of microporous structure material used in the external permeable layer is a porous material with high air permeability and moisture conductivity. The material can block rainwater and large external particle pollutants while allowing air and steam to penetrate to maintain air pressure balance and dryness inside and outside the wall.

3. The intelligent building curtain wall system according to claim 1, characterized in that: The drainage valve of the drainage module adopts an electric control mechanism, which can quickly adjust the opening and closing degree according to the instructions of the intelligent processing module, thereby effectively draining the water accumulated in the wall and preventing water erosion and accumulation.

4. The intelligent building curtain wall system according to claim 1, characterized in that: The remote communication module adopts wireless communication technology to support data exchange and command transmission with third-party systems including smart home systems, weather stations and building management centers, and realizes remote monitoring and intelligent control of the system.

Images