Intelligent building environment control method based on multi-modal perception

Abstract

This invention relates to an intelligent building environment control method based on multimodal perception, belonging to the field of environmental control system technology. This invention extracts the characteristics of personnel distribution and activity intensity from visual images and accurately converts them into dynamic metabolic rate parameters at the data layer. This enables the system to adapt to sudden changes in the state of personnel within the building in real time and proactively, greatly improving the adaptability and agility of environmental regulation. Simultaneously, it introduces physiological rhythm model parameters based on system time series to weight and correct the real-time predicted average vote index, effectively compensating for the shortcomings of traditional thermodynamic evaluations that neglect non-visual biological effects of the human body, thus improving the overall comfort of personnel within the space. Guided by the output comprehensive comfort target value, the algorithm coordinates the HVAC, lighting, and shading systems, completely changing the situation where each subsystem operates independently. Under the premise of strictly ensuring dynamic comfort, it achieves a deep reduction in the overall operating energy consumption of the building control system.

Description

Technical Field

[0001] This invention relates to the field of environmental control system technology, and in particular to an intelligent building environment control method based on multimodal sensing. Background Technology

[0002] Existing intelligent building environment control systems primarily rely on single environmental sensors (such as temperature and humidity sensors) and employ simple threshold triggering mechanisms (such as starting cooling when the temperature exceeds 26°C) or timed control (such as starting and stopping equipment according to get off work hours) to achieve regulation. This existing technology has the following significant drawbacks: Traditional systems often rely on single temperature and humidity sensors. However, physical comfort (PMV) is affected by various factors such as wind speed, mean radiant temperature, human activity level, and clothing thermal resistance. Existing sensor deployments struggle to cover the complex indoor physical field. Furthermore, many low-power IoT sensors have low sampling frequencies and are affected by their installation location (e.g., near air vents or in direct sunlight), leading to significant biases and lags in the collected feedback data. Existing systems typically set a constant overall temperature, but in the same environment, the comfort level varies greatly among individuals with different physical conditions and at different workstations. Many lighting environment control systems only focus on illuminance (brightness) and ignore the influence of the spectrum on the human circadian rhythm. Long-term exposure to such an environment can easily lead to visual fatigue or poor sleep quality for office workers. To address the aforementioned technical deficiencies, a solution is proposed. Summary of the Invention

[0003] The purpose of this invention is to extract the characteristics of personnel distribution and activity intensity from visual images and accurately convert them into dynamic metabolic rate parameters at the data layer. This enables the system to adapt to sudden changes in the state of people in the building in real time and proactively, greatly improving the adaptability and agility of environmental regulation. At the same time, it introduces physiological rhythm model parameters based on system time series to perform weighted correction on the real-time predicted average vote index, effectively making up for the shortcomings of traditional thermodynamic evaluation that ignores non-visual biological effects of the human body, and improving the overall comfort of people in the space.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent building environment control method based on multimodal perception, comprising the following steps: S1. Delineate target areas based on the spatial usage range of each area within the intelligent building, and provide a multimodal sensor group to acquire environmental physical data, human visual image sequences, and system time data of the target areas. The environmental physical data includes indoor temperature and humidity, illuminance, and carbon dioxide concentration. The system time data is used to extract human physiological rhythm model parameters at the corresponding time. S2. Perform target detection and behavior recognition on the visual image sequence of the personnel, extract personnel density features and personnel activity intensity features, and then calculate dynamic metabolic rate parameters based on the personnel activity intensity features to characterize the actual heat production and comfort needs of the personnel in the current target area. S3. Construct a dynamic comfort evaluation model. Input the dynamic metabolic rate parameter and environmental physical data into the preset comfort evaluation model, calculate the real-time predicted average vote index, and combine it with the physiological rhythm model parameters extracted from the system time data to perform weighted correction on the real-time predicted average vote index, and output the comprehensive comfort target value of the target area. S4. Based on the comprehensive comfort target value, calculate the current cooling / heating load deviation value and light demand deviation value of the target area. Minimize the deviation value as the first optimization objective and minimize the total energy consumption of the building control system as the second optimization objective. Generate coordinated control instructions for the HVAC system, lighting system and shading system through a multi-objective optimization algorithm, execute the coordinated control instructions, and return to S1 after a preset time period.

[0005] Furthermore, the specific process of extracting the human physiological rhythm model parameters at the corresponding time based on the system time data is as follows: S101. Obtain the current timestamp t and input it into the preset circadian rhythm periodic function. Based on the periodic changes in human physiological characteristics over time, define the phase angle. for: in, The average wake-up time of people within the preset target area is based on the phase angle. Determine whether the person in question is currently in a period of high alertness, a period of slow metabolism, or a period of melatonin secretion. S102. Construct a light environment parameter model based on non-visual biological effects, and retrieve the corresponding target associated color temperature and target color rendering requirement from the preset physiological rhythm database according to the extracted phase angle. S103. Extract thermal comfort correction coefficient from system time data. : Where A is the amplitude term, The thermal comfort correction coefficient is for phase deviation. This reflects the physiological characteristic that core body temperature is lowest in the early morning and highest in the afternoon; S104. Extract the date attribute from the system time data, combine it with the current timestamp t, and automatically call the corresponding scene weight.

[0006] Furthermore, the specific process for calculating dynamic metabolic rate parameters is as follows: S201. Use the deep learning object detection algorithm to lock the bounding boxes of the people in the target area, and use the skeleton key point detection algorithm to extract the coordinate sequences of several core key points of the human body. Define the activity intensity operator K, and its calculation formula is as follows: , where n is the total number of people in the target area, is the centroid displacement vector of the i-th individual within the preset sampling time window, is the average motion rate of the limb key points of the i-th individual, and are the preset weight coefficients; S203. Compare the calculated activity intensity operator K with the preset behavior pattern interval (Kmin, Kmax), and divide the people's activity status into the following levels: When K < Kmin, it is divided into the first level, that is, the resting state; When Kmin ≤ K < Kmax, it is divided into the second level, that is, slight activity; When Kmax ≤ K, it is divided into the third level, that is, medium-high intensity activity; S204. Use the weighted average method to calculate the global dynamic metabolic rate parameter of the target area : , where is the number of people in the j-th activity level, is the standard metabolic rate value corresponding to the level, j = 1, 2, 3,..., n, and n is the total number of people; S205. Obtain the indoor temperature and humidity to perform a second-order correction on the global dynamic metabolic rate parameter. At the same time, if the carbon dioxide concentration in the target area surges, it indicates that the people's metabolic level is in the high-level rising period, and the metabolic rate weight will be predictively increased.

[0007] Furthermore, the specific process of calculating the real-time predicted average vote count index is as follows: S301. Obtain the environmental physical data of the target area and the dynamic metabolic rate parameter of the target area at the current moment. Through the time window interpolation algorithm, align the indoor temperature and humidity, illuminance, and carbon dioxide concentration to the same time stamp node; S302. Input the aligned environmental physical data and dynamic metabolic rate parameter into the preset comfort evaluation model. The comfort evaluation model is constructed based on the human body heat balance equation and is used to calculate the real-time predicted average vote count index , and the specific calculation formula is as follows:

[0008] Wherein, L is the human body heat load factor, which is calculated by combining input environmental physical data and dynamic metabolic rate parameters. The comfort evaluation model can respond in real time to sudden changes in the activity status of people in the target area, thereby outputting a real-time predicted average vote index that accurately represents the difference between the current physical heat production and the environmental heat absorption.

[0009] Furthermore, the specific process for outputting the overall comfort target value for the target area is as follows: Extract parameters of the human physiological rhythm model corresponding to the current system timestamp. The overall comfort prediction value is obtained by adjusting the weights and performing a weighted calculation on the index of the real-time predicted average vote count. The specific formula is as follows: ,in The illuminance-color temperature coupling factor for the current target area; Obtain the preset extreme value safety boundary. When the overall comfort prediction value is within the extreme value safety boundary, output the overall comfort prediction value as the overall comfort target value.

[0010] Furthermore, the specific process of generating coordinated control instructions through a multi-objective optimization algorithm is as follows: S401. Obtain the target value of comprehensive comfort in the target area, and obtain the current real-time environmental physical data to calculate the environmental state deviation of the target area in order to construct a deviation vector. : ,in, The cooling / heating load deviation value is calculated based on the difference between the current environmental physics data and the overall comfort target value. The light demand deviation value is calculated based on the difference between the current illuminance and the target illuminance required by the physiological rhythm model. S402. Define a set of control variables X = [x1, x2, x3], where x1 is the operating frequency of the HVAC system, x2 is the power of the lighting system, and x3 is the opening angle of the louvers of the shading system. Construct the first optimization objective function. as follows: ,in and The preset weighting coefficients, To achieve the optimal illuminance, the first optimization objective function adjusts the set of control variables X so that the expected environmental state after adjustment approaches the overall comfort target value infinitely. S403. Construct the second optimization objective function. as follows: ,in This is the rated energy consumption of the air conditioner. Given the rated lighting energy consumption, the second optimization objective function is used to minimize the total energy consumption of the building control system; S404. The first and second optimization objective functions are combined into a multi-objective optimization problem. The non-dominated sorting genetic algorithm with an elite strategy is used to solve iteratively. The solution space of the set of control variables X is searched globally, and finally a set of non-dominated solutions is output. According to the preset decision preference matrix, an optimal combination of control parameters with the highest overall benefit is automatically selected from the non-dominated solution set. S405. The selected optimal control parameter combination is parsed into the corresponding underlying device communication message. According to the preset collaborative execution logic, the opening and closing angle command is sent to the shading system in the intelligent building first for energy-free adjustment. After evaluating and predicting the delay, the frequency conversion command is sent to the HVAC system and the dimming and color temperature adjustment command is sent to the lighting system simultaneously.

[0011] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: This intelligent building environment control method based on multimodal perception extracts the characteristics of personnel distribution and activity intensity from visual images and accurately converts them into dynamic metabolic rate parameters at the data layer. This enables the system to adapt to sudden changes in the state of personnel within the building in real time and proactively, greatly improving the adaptability and agility of environmental regulation. At the same time, it introduces physiological rhythm model parameters based on system time series to weight and correct the real-time prediction average vote index, effectively making up for the shortcomings of traditional thermodynamic evaluation that ignores non-visual biological effects of the human body, and improving the overall comfort of personnel in the space. Guided by the output comprehensive comfort target value, the algorithm coordinates the HVAC, lighting and shading systems, completely changing the situation where each subsystem operates independently. Under the premise of strictly ensuring dynamic comfort, it achieves a deep reduction in the overall operating energy consumption of the building control system. Attached Figure Description

[0012] Figure 1 A schematic diagram of the overall method flow of the present invention is shown. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] Example: like Figure 1As shown, the intelligent building environment control method based on multimodal perception includes the following steps: S1. Delineate target areas based on the spatial usage range of each area within the intelligent building, and provide a multimodal sensor group to acquire environmental physical data, human visual image sequences, and system time data of the target areas. The environmental physical data includes indoor temperature and humidity, illuminance, and carbon dioxide concentration. The system time data is used to extract human physiological rhythm model parameters at the corresponding time. The specific process of extracting the parameters of the human physiological rhythm model at the corresponding moment based on the system time data is as follows: S101. Obtain the current timestamp t and input it into the preset circadian rhythm periodic function. Based on the periodic changes in human physiological characteristics over time, define the phase angle. for: in, The average wake-up time of people within the preset target area is based on the phase angle. Determine whether the person in question is currently in a period of high alertness, a period of slow metabolism, or a period of melatonin secretion. S102. Construct a light environment parameter model based on non-visual biological effects, and retrieve the corresponding target associated color temperature and target color rendering requirement from the preset physiological rhythm database according to the extracted phase angle. During the morning hours (e.g., 08:00-10:00): the light environment parameter model is automatically adjusted to a high color temperature (e.g., 5000K-6500K) and high illuminance feedback to suppress melatonin secretion and improve work efficiency. During lunch breaks and evenings: the light environment parameter model automatically and smoothly switches to a low color temperature (such as 2700K-3000K) and reduces light intensity to meet the physiological relaxation needs of the human body; S103. Extract thermal comfort correction coefficient from system time data. : Where A is the amplitude term, The thermal comfort correction coefficient is for phase deviation. This reflects the physiological characteristic that core body temperature is lowest in the early morning and highest in the afternoon; S104. Extract the date attribute (such as weekdays and holidays) from the system time data, and automatically call the corresponding scene weight based on the current timestamp t.

[0015] For example, if S2 detects that there are still a few people active late at night, the system will adjust the circadian rhythm parameters to the sleep-deprivation compensation mode, increase the blue light component and slightly increase the air conditioner's air outlet temperature to maintain the basic comfort of people and compensate for the cold feeling caused by the decrease in metabolism.

[0016] S2. Perform object detection and behavior recognition on the visual image sequence of the personnel, extract the personnel density feature and the personnel activity intensity feature, and then calculate the dynamic metabolic rate parameter according to the personnel activity intensity feature to characterize the actual heat production and comfort demand of the personnel in the current target area; The specific process of calculating the dynamic metabolic rate parameter is as follows: S201. Use the deep learning object detection algorithm to lock the bounding boxes of the personnel in the target area, and use the skeleton key point detection algorithm to extract the coordinate sequences of several core key points of the human body. Define the activity intensity operator K, and its calculation formula is as follows: , where n is the total number of personnel in the target area, is the centroid displacement vector of the i-th individual within the preset sampling time window, is the average motion rate of the limb key points of the i-th individual, and are the preset weight coefficients; S203. Compare the calculated activity intensity operator K with the preset behavior pattern interval (Kmin, Kmax), and divide the personnel activity state into the following levels: When K < Kmin, it is divided into the first level, that is, the resting state; When Kmin ≤ K < Kmax, it is divided into the second level, that is, the slight activity; When Kmax ≤ K, it is divided into the third level, that is, the medium-high intensity activity; S204. Use the weighted average method to calculate the global dynamic metabolic rate parameter of the target area : , where is the number of personnel in the j-th activity level, is the standard metabolic rate value corresponding to the level, j = 1, 2, 3,..., n, and n is the total number of people; S205. Obtain the indoor temperature and humidity to perform second-order correction on the global dynamic metabolic rate parameter. At the same time, if the carbon dioxide concentration in the target area surges, it indicates that the personnel metabolic level is in the high-rise period, and the metabolic rate weight will be predictively increased.

[0017] S3. Construct a dynamic comfort evaluation model, input the dynamic metabolic rate parameter and the environmental physical data into the preset comfort evaluation model, calculate the real-time predicted mean vote count index, and combine the physiological rhythm model parameters extracted from the system time data to perform weighted correction on the real-time predicted mean vote count index, and output the comprehensive comfort target value of the target area; The specific process of calculating the real-time predicted mean vote count index is as follows: S301. Obtain the environmental physical data of the target area and the dynamic metabolic rate parameters of the target area at the current moment, and align the indoor temperature, humidity, illuminance and carbon dioxide concentration to the same timestamp node through a time window interpolation algorithm. S302. Input the aligned environmental physical data and dynamic metabolic rate parameters into the preset comfort evaluation model. The comfort evaluation model is constructed based on the human body heat balance equation and is used to calculate the real-time predicted average vote index. The specific calculation formula is as follows:

[0018] Wherein, L is the human body heat load factor, which is calculated by combining input environmental physical data and dynamic metabolic rate parameters. The comfort evaluation model can respond in real time to sudden changes in the activity status of people in the target area, thereby outputting a real-time predicted average vote index that accurately represents the difference between the current physical heat production and the environmental heat absorption.

[0019] The specific process for outputting the overall comfort target value for the target area is as follows: Extract parameters of the human physiological rhythm model corresponding to the current system timestamp. The overall comfort prediction value is obtained by adjusting the weights and performing a weighted calculation on the index of the real-time predicted average vote count. The specific formula is as follows: ,in The illuminance-color temperature coupling factor for the current target area; When it is determined that the person is currently in a period of physiological fatigue (such as in the afternoon), through The system automatically adjusts the overall comfort target value towards a slightly cooler or higher color temperature to compensate for the decline in physiological function through environmental stimulation.

[0020] Obtain the preset extreme value safety boundary. When the overall comfort prediction value is within the extreme value safety boundary, output the overall comfort prediction value as the overall comfort target value.

[0021] S4. Based on the comprehensive comfort target value, calculate the current cooling / heating load deviation value and light demand deviation value of the target area. Minimize the deviation value as the first optimization objective and minimize the total energy consumption of the building control system as the second optimization objective. Generate coordinated control instructions for the HVAC system, lighting system and shading system through a multi-objective optimization algorithm, execute the coordinated control instructions, and return to S1 after a preset time period.

[0022] The specific process of generating coordinated control instructions using a multi-objective optimization algorithm is as follows: S401. Obtain the target value of comprehensive comfort in the target area, and obtain the current real-time environmental physical data to calculate the environmental state deviation of the target area in order to construct a deviation vector. : ,in, The cooling / heating load deviation value is calculated based on the difference between the current environmental physics data and the overall comfort target value. The light demand deviation value is calculated based on the difference between the current illuminance and the target illuminance required by the physiological rhythm model. S402. Define a set of control variables X = [x1, x2, x3], where x1 is the operating frequency of the HVAC system, x2 is the power of the lighting system, and x3 is the opening angle of the louvers of the shading system. Construct the first optimization objective function. as follows: ,in and The preset weighting coefficients, To achieve the optimal illuminance, the first optimization objective function adjusts the set of control variables X so that the expected environmental state after adjustment approaches the overall comfort target value infinitely. S403. Construct the second optimization objective function. as follows: ,in This is the rated energy consumption of the air conditioner. Given the rated lighting energy consumption, the second optimization objective function is used to minimize the total energy consumption of the building control system, and the shading system louver opening angle x3 is cross-introduced into the air conditioning energy consumption. Energy consumption for lighting The calculations are designed to reflect the coupled effects of opening and closing the sunshade on the introduction of natural light (reducing lighting energy consumption) and the introduction of solar radiation heat (increasing / reducing air conditioning energy consumption). S404. The first and second optimization objective functions are combined into a multi-objective optimization problem. The non-dominated sorting genetic algorithm with an elite strategy is used to solve iteratively. The solution space of the set of control variables X is searched globally, and finally a set of non-dominated solutions is output. According to the preset decision preference matrix, an optimal combination of control parameters with the highest overall benefit is automatically selected from the non-dominated solution set. S405. The selected optimal control parameter combination is parsed into the corresponding underlying device communication message. According to the preset collaborative execution logic, the opening and closing angle command is sent to the shading system in the intelligent building first for energy-free adjustment. After evaluating and predicting the delay, the frequency conversion command is sent to the HVAC system and the dimming and color temperature adjustment command is sent to the lighting system. After each system executes the command, it returns to step S1 after a preset time period to reacquire multimodal data and form a real-time dynamic closed loop for environmental control.

[0023] This invention extracts the distribution and activity intensity features of people from visual images and accurately converts them into dynamic metabolic rate parameters at the data layer. This enables the system to adapt to sudden changes in the state of people within the building in real time and proactively, greatly improving the adaptability and agility of environmental control. At the same time, it introduces physiological rhythm model parameters based on system time series to weight and correct the real-time prediction average vote index, effectively making up for the shortcomings of traditional thermodynamic evaluation that ignores non-visual biological effects of the human body, and improving the overall comfort of people in the space. Guided by the output comprehensive comfort target value, the algorithm coordinates the HVAC, lighting and shading systems, completely changing the situation where each subsystem operates independently. Under the premise of strictly ensuring dynamic comfort, it achieves a deep reduction in the overall operating energy consumption of the building control system.

[0024] The size of the interval and threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by those skilled in the art for each set of sample data; as long as it does not affect the ratio between the parameter and the quantized value.

[0025] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation. In the two embodiments provided in this application, it should be understood that the disclosed methods can be implemented in other ways; for example, the division of modules is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed; another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be indirect coupling or communication connection through some interfaces, devices or modules, and can be electrical, mechanical or other forms. The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent building environment control based on multimodal perception, characterized in that, The steps include the following: S1. Define a target area according to the spatial usage range of each area in the intelligent building, and provide a multi-modal sensor group to obtain the environmental physical data, personnel visual image sequence and system time data of the target area. The environmental physical data includes indoor temperature and humidity, illuminance and carbon dioxide concentration. The system time data is used to extract the human physiological rhythm model parameters at the corresponding moment; S2. Conduct target detection and behavior recognition on the personnel visual image sequence, extract the personnel density feature and personnel activity intensity feature, and then calculate the dynamic metabolic rate parameter according to the personnel activity intensity feature to characterize the actual heat production and comfort demand of the personnel in the current target area; S3. Construct a dynamic comfort evaluation model, input the dynamic metabolic rate parameter and environmental physical data into a preset comfort evaluation model, calculate the real-time predicted mean vote number index, and combine the physiological rhythm model parameters extracted from the system time data to weight and correct the real-time predicted mean vote number index, and output the comprehensive comfort target value of the target area; S4. Based on the comprehensive comfort target value, calculate the current cold / hot load deviation value and lighting demand deviation value of the target area. Taking the minimum deviation value as the first optimization goal and the minimum total energy consumption of the building control system as the second optimization goal, generate the coordinated control instructions for the HVAC system, lighting system and shading system through a multi-objective optimization algorithm, execute the coordinated control instructions, and return to S1 after a preset time period.

2. The intelligent building environment control method based on multimodal perception according to claim 1, characterized in that, The specific process of extracting the human physiological rhythm model parameters at the corresponding moment according to the system time data is as follows: S101. Obtain the current timestamp t and input it into the preset circadian rhythm periodic function. Based on the periodic changes in human physiological characteristics over time, define the phase angle. for: in, This indicates the average wake-up time of people within a preset target area, based on the phase angle. Determine whether the person in question is currently in a period of high alertness, a period of slow metabolism, or a period of melatonin secretion. S102. Construct a light environment parameter model based on non-visual biological effects, and retrieve the corresponding target correlated color temperature and target color rendering index requirements from a preset physiological rhythm database according to the extracted phase angle; S103. Extracting parameters of the human physiological rhythm model from system time data. : Where A is the amplitude term, For phase deviation, the parameters of the human physiological rhythm model This reflects the physiological characteristic that core body temperature is lowest in the early morning and highest in the afternoon; S104. Extract the date attribute from the system time data, and automatically call the corresponding scene weight in combination with the current timestamp t.

3. The intelligent building environment control method based on multimodal perception according to claim 1, characterized in that, The specific process of calculating the dynamic metabolic rate parameter is as follows: S201. Use the deep learning target detection algorithm to lock the bounding box of the personnel in the target area, and use the skeleton key point detection algorithm to extract the coordinate sequence of several core key points of the human body. Define the activity intensity operator K, and its calculation formula is as follows: Where n is the total number of people in the target area, Let be the centroid displacement vector of the i-th individual within the preset sampling time window. Let be the average movement rate of the limb key points of the i-th individual. and These are preset weighting coefficients; S203. Compare the calculated activity intensity operator K with the preset behavior pattern interval (Kmin, Kmax), and divide the personnel activity state into the following levels: When K < Kmin, it is divided into the first level, that is, the resting state; When Kmin ≤ K < Kmax, it is divided into the second level, that is, the slight activity; When Kmax ≤ K, it is divided into the third level, that is, the medium-high intensity activity; S204. Calculate the global dynamic metabolic rate parameters of the target region using the weighted average method. : ,in Let j represent the number of people at the j-th activity level. The standard metabolic rate value corresponding to the level, j=1, 2, 3, ..., n, where n is the total number of people; S205. Obtain indoor temperature and humidity data to perform a second-order correction on the global dynamic metabolic rate parameters. Simultaneously, if the carbon dioxide concentration in the target area surges, indicating a high-level increase in human metabolic activity, the metabolic rate weight will be predictively increased to obtain the optimized global dynamic metabolic rate parameters. .

4. The intelligent building environment control method based on multimodal perception according to claim 1, characterized in that, The specific process of calculating the real-time predicted mean vote number index is as follows: S301. Obtain the environmental physical data of the target area and the dynamic metabolic rate parameter of the target area at the current moment, and align the indoor temperature and humidity, illuminance and carbon dioxide concentration to the same timestamp node through the time window interpolation algorithm; S302. Input the aligned environmental physical data and dynamic metabolic rate parameters into the preset comfort evaluation model. The comfort evaluation model is constructed based on the human body heat balance equation and is used to calculate the real-time predicted average vote index. The specific calculation formula is as follows: Where L is the human body heat load factor, the calculation of which is determined by the input environmental physical data and dynamic metabolic rate parameters, by introducing... The comfort evaluation model can respond in real time to sudden changes in the activity status of people in the target area, thereby outputting a real-time predicted average vote index that accurately represents the difference between the current physical heat production and the environmental heat absorption.

5. The intelligent building environment control method based on multimodal perception according to claim 1, characterized in that, The specific process of outputting the comprehensive comfort target value of the target area is as follows: Extract parameters of the human physiological rhythm model corresponding to the current system timestamp. The overall comfort prediction value is obtained by adjusting the weights and performing a weighted calculation on the index of the real-time predicted average vote count. The specific formula is as follows: ,in The illuminance-color temperature coupling factor for the current target area; Obtain the preset extreme value safety boundary. When the overall comfort prediction value is within the extreme value safety boundary, output the overall comfort prediction value as the overall comfort target value.

6. The intelligent building environment control method based on multimodal perception according to claim 1, characterized in that, The specific process of generating coordinated control instructions using a multi-objective optimization algorithm is as follows: S401. Obtain the target value of comprehensive comfort in the target area, and obtain the current real-time environmental physical data to calculate the environmental state deviation of the target area in order to construct a deviation vector. : ,in, The cooling / heating load deviation value is calculated based on the difference between the current environmental physics data and the overall comfort target value. The light demand deviation value is calculated based on the difference between the current illuminance and the target illuminance required by the physiological rhythm model. S402. Define a set of control variables X = [x1, x2, x3], where x1 is the operating frequency of the HVAC system, x2 is the power of the lighting system, and x3 is the opening angle of the louvers of the shading system. Construct the first optimization objective function. as follows: ,in and The preset weighting coefficients, To achieve the optimal illuminance, the first optimization objective function adjusts the set of control variables X so that the expected environmental state after adjustment approaches the overall comfort target value infinitely. S403. Construct the second optimization objective function. as follows: ,in This is the rated energy consumption of the air conditioner. Given the rated lighting energy consumption, the second optimization objective function is used to minimize the total energy consumption of the building control system; S404. The first and second optimization objective functions are combined into a multi-objective optimization problem. The non-dominated sorting genetic algorithm with an elite strategy is used to solve iteratively. The solution space of the set of control variables X is searched globally, and finally a set of non-dominated solutions is output. According to the preset decision preference matrix, an optimal combination of control parameters with the highest overall benefit is automatically selected from the non-dominated solution set. S405. The selected optimal control parameter combination is parsed into the corresponding underlying device communication message. According to the preset collaborative execution logic, the opening and closing angle command is first sent to the shading system in the intelligent building for energy-free adjustment. After evaluating and predicting the delay, the frequency conversion command is sent to the HVAC system and the dimming and color temperature adjustment command is sent to the lighting system.

Images