Building global collaborative awareness low-carbon control method

Abstract

The invention discloses a building global collaborative perception low-carbon control method. The method comprises the steps of 1, arranging a temperature and humidity sensor, an energy consumption monitoring sensor and an air quality monitoring sensor in internal and external environments of a building, and performing data acquisition on key control elements of temperature and humidity, energy consumption and air quality of the external environment of the building; step 2, carrying out preprocessing operation on the collected data; 3, constructing the data processed in the step 2 into a multi-view-angle data matrix, and analyzing the multi-view-angle data matrix by using a multi-view-angle K-Means clustering algorithm to obtain a comfort interval; 4, based on the comfort interval, an improved dynamic interval prediction and control algorithm is adopted to optimize the operation parameters of the building equipment; and 5, displaying data and analysis results of the whole process through a visual module. According to the invention, collaborative perception of information in multiple aspects such as internal and external environments and equipment operation states of the building is realized, intelligent control decision is made based on the information, and low-carbon and energy-saving effects are achieved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent building control technology, and in particular to a building global collaborative perception low-carbon control method. Background Art

[0002] At present, there are many intelligent control systems for buildings on the market, but the following problems are common:

[0003] 1. Lack of effective global collaborative perception capabilities, with information islands, low efficiency, slow response, poor user experience, and high maintenance costs. The decentralized control model ignores the comprehensive needs of the building as a whole, which may cause conflicts or duplication of work when each system performs tasks, and cannot make timely adjustments to respond to emergencies or environmental changes, affecting user satisfaction with smart buildings and increasing maintenance costs and complexity.

[0004] 2. Lack of refined energy consumption monitoring and management mechanisms. Due to the lack of or insufficient functions of the intelligent building energy management system, it is impossible to achieve real-time accurate monitoring and effective control of building energy consumption, which makes it impossible to adjust energy use in a timely manner, resulting in unnecessary waste of energy, which is not conducive to the sustainable operation and management of buildings and runs counter to the green energy-saving concept advocated by modern society.

[0005] In the research field of intelligent building control systems, an intelligent building control system for architectural design (patent number: 202311325080.3) integrates central management, intelligent control and data acquisition modules, and uses mathematical models to analyze historical data and predict future trends to improve the comfort of the indoor environment. However, the patent mainly relies on simple algorithms to independently control parameters such as temperature, humidity, and energy consumption. This control method lacks integrity and synergy, making it difficult to reduce energy consumption. In actual applications, since each algorithm runs independently, the mutual influence between different factors is not fully considered, which may cause energy waste.

[0006] Therefore, it is necessary to explore a global collaborative perception and low-carbon control method for smart buildings. Summary of the invention

[0007] In order to overcome the defects of the above existing technologies, the purpose of the present invention is to provide a low-carbon control method for the whole-area collaborative perception of buildings, which realizes the collaborative perception of various information such as the internal and external environment of the building, the operating status of the equipment, etc., and makes intelligent control decisions based on this to achieve the effect of low-carbon energy saving. It has the advantages of high control accuracy, strong adaptability, strong practicality, and good economic benefits.

[0008] In order to achieve the above object, the technical solution adopted by the present invention is:

[0009] A building global collaborative perception low-carbon control method, comprising the following steps;

[0010] Step 1: Arrange temperature and humidity sensors, energy consumption monitoring sensors and air quality monitoring sensors inside and outside the building to collect data on key control factors such as temperature and humidity, energy consumption and air quality in the building's internal and external environments;

[0011] Step 2: Preprocess the collected data to improve data quality and accuracy and lay the foundation for multi-perspective clustering analysis;

[0012] Step 3: construct the data processed in step 2 into a multi-view data matrix, and use the multi-view K-Means clustering algorithm to analyze the multi-view data matrix to obtain the comfort interval;

[0013] Step 4: Based on the comfort interval, an improved dynamic interval prediction and control (DIFC) algorithm is used to dynamically adjust the building equipment, so that the indoor environment is always in the best comfort state, reducing energy waste, reducing building energy consumption, and achieving low-carbon and environmental protection; through step 4, the equipment is controlled to achieve the purpose of energy saving and low carbon;

[0014] Step 5: Display the data and analysis results of the entire process through the visualization module.

[0015] The step 1 is specifically as follows:

[0016] Temperature and humidity sensors are used to accurately collect temperature and humidity data inside and outside the building, reflecting climate conditions and human comfort perception;

[0017] Energy consumption monitoring sensors are used to capture building electricity, water, gas and other energy consumption information in real time to help analyze energy consumption patterns;

[0018] Air quality detection sensors are used to monitor indoor and outdoor PM2.5, formaldehyde, and total volatile organic carbon (TVOC) concentrations, which are related to human health and ventilation needs.

[0019] Further:

[0020] Temperature and humidity sensors are located at the center of the wall of each room 1.5-2 meters above the ground, evenly spaced about 1.5-2 meters high along the wall of the corridor, about 1.5-2 meters high on each platform of the stairwell, 2-3 meters high on the outer wall of the outdoor building and in the open space on the roof;

[0021] Energy consumption monitoring sensors are installed on the power distribution boxes on each floor and the power supply lines of large power-consuming equipment. For water, they are installed on the main water inlet pipe and branch water pipes. For gas, they are installed on the main gas meter inlet pipe.

[0022] Air quality monitoring sensors are installed indoors in the corners of the room, 1-1.5 meters above the ground, near the vents, and outdoors on the roof. These installation locations are designed to accurately collect data, thereby achieving effective monitoring and control of indoor and outdoor environmental conditions and energy consumption.

[0023] The specific implementation steps of step 2 are:

[0024] Step 2.1: Filter the noise from the data collected by the sensor

[0025] The median filter method is used to process the data, making the processed data more stable and reducing data fluctuations. Taking temperature data as an example, humidity data, energy consumption data and other related data can all be processed in the same way.

[0026] The data sequence of the temperature sensor collected is set to be expressed as T = [t 1 , t 2 , t 3 , ..., t n , where t i represents the temperature value collected at the i-th moment, and n is the number of collected data points;

[0027] Median filtering is to select a window with a fixed length of odd number m for each data point in the sequence. i , the data in the window is Then sort the data in this window and take the middle value after sorting to replace the original data point t i .

[0028] Step 2.2: Fill in the missing data points of the sensor data

[0029] For data series with missing data, linear interpolation is used to fill in the missing data points, so that the entire data series is complete in the time dimension, ensuring that subsequent analysis will not be biased or impossible to perform for multiple data points with missing values ​​due to missing data. Taking energy consumption data as an example, humidity data, temperature data and other related data can refer to this example and use the same processing method.

[0030] In the collected energy consumption data series E(E=[e 1 , e 2 , e 3 , ..., e n ]), if the data has some missing values, such as e j For missing values ​​(1 <j<n);

[0031] Set known e j-1 and e j+1The value of , based on the linear relationship to calculate the missing e j , the specific linear interpolation calculation formula is formula (2-1):

[0032]

[0033] For multiple data points with missing values, this method is used to complete them one by one, so that the entire energy consumption data series is complete and available;

[0034] Step 2.3: Time-align and synchronize data from different sensors

[0035] The time alignment operation aims to build a unified data time base. Specifically, the earliest data collection time of each sensor is used as the starting point, and the data is reorganized and integrated according to fixed time intervals.

[0036] The data collected by different sensors are extracted and processed according to this unified time interval to ensure that at each unified time node after integration, data information on temperature, humidity, energy consumption, air quality, etc. can be simultaneously covered. This effectively guarantees the synchronization of data and lays a solid foundation for subsequent comprehensive analysis of the internal correlation between different control elements and other operations.

[0037] Step 2.4: Standardize the data

[0038] The data is processed using Z-score standardization to convert each indicator into a dimensionless form. In subsequent operations such as multi-perspective cluster analysis, different indicators can participate in calculation and analysis on the basis of equal importance, avoiding the problem of unbalanced weights of certain indicators in the analysis due to different dimensions.

[0039] Taking temperature data as an example, humidity data, temperature data and other related data can be processed in the same way as in this example.

[0040] For temperature T, the standardized value The calculation formula is (2-2), and the data sequence of the collected temperature sensor is expressed as T = [t 1 , t 2 , t 3 , ..., t n ], where t i represents the temperature value collected at the i-th moment, n is the number of collected data points, is the mean of the temperature data series, calculated as (2-3), σ t is the standard deviation of the temperature data series, and the calculation formula is (2-4):

[0041]

[0042] The specific implementation steps of step 3 are:

[0043] Step 3.1: Construct a multi-view data matrix

[0044] Extract multidimensional key control factors from the preprocessed data obtained in step 2, and construct a multi-perspective data matrix with time as the horizontal axis and indicators as the vertical axis. The perspectives include time perspective, space perspective, and functional perspective.

[0045] Time perspective: defined as in Represents the multidimensional indicator vector at the i-th time point;

[0046] Spatial perspective: defined as in Represents a multidimensional index vector at the spatial position s;

[0047] Functional perspective: defined as in A vector representing a specific functional indicator (such as energy consumption or air quality);

[0048] The matrix is ​​stored in [n,m] dimensions, where n is the number of time points or spatial locations, and m is the dimension of the control element (such as temperature, humidity, etc.);

[0049] Step 3.2: Divide the initial clusters

[0050] K-Means algorithm is applied to the local data subset of each perspective for preliminary clustering analysis. In each perspective, the similarity between data points is measured by weighted Euclidean distance, and the appropriate number of clusters is set to divide the multidimensional data into several clusters. Each cluster represents a typical environmental pattern under this perspective, and the clustering results of each perspective are output, including the cluster center point and the data distribution within the cluster;

[0051] For each viewpoint, the data matrix X v (v∈{t,s,f}) uses K-Means algorithm for preliminary clustering. The goal is to v Divide into k clusters The center point of each cluster The calculation formula is (3-1),

[0052]

[0053] The weighted Euclidean distance calculation formula from the data point within the cluster to the center point is as follows:

[0054]

[0055] where w l is the weight of the lth indicator, which is used to highlight the importance of specific control factors. The clustering objective function formula is formula (3-3):

[0056]

[0057] Finally, the preliminary clustering results of each perspective are output, including the center point of each cluster. And the distribution of data within the family

[0058] Step 3.3: Fusion of multi-view clustering

[0059] The clustering information of each perspective outputted in step 3.2 is integrated by a joint optimization method. Specifically, a global weighted optimization objective function is established based on the cluster center points of all perspectives. The cluster division result that enables the optimal classification of global data is found by iteratively adjusting the center positions of each cluster.

[0060] Based on the preliminary clustering results of all perspectives, a global weighted optimization objective function is established:

[0061]

[0062] Among them, α v is the weight for each view, μ j The global cluster center is the global cluster center. Through the selection and update of multi-view cluster center points, the characteristics of different perspectives are integrated to finally output the global family division {C 1 ,C 2 ,...C k} and the center point μ j ;

[0063] Step 3.4: Generate comprehensive comfort zone

[0064] Combined with the multi-view clustering results generated in step 3.3, the comprehensive comfort range of the building environment is extracted. The range of the comfort range is determined by the center point and distribution characteristics of each cluster, including the upper and lower limits of parameters such as temperature, humidity, and air quality (such as CO2 concentration and PM2.5 concentration). The output comfort range is expressed in a multi-dimensional range to provide a benchmark for subsequent control algorithms.

[0065] The comprehensive comfort interval is extracted based on the final clustering results, which is defined as the center point of each cluster and its data distribution range:

[0066] S={(min(x l ),max(x l ))|x l ∈C j ,l=1,2,...,m}

[0067] The comfort zone includes the upper and lower limits of parameters such as temperature, humidity, and air quality (such as CO2 concentration and PM2.5 concentration), providing a benchmark for subsequent control algorithms;

[0068] The step 4 is specifically as follows:

[0069] The dynamic interval prediction and control algorithm takes the current environmental state as the initial condition and sets the comfort interval as the target constraint. It builds a prediction model by integrating physical laws and data-driven methods, dynamically predicts the evolution trend of future environmental parameters, and then outputs the optimal operating parameters of each building equipment to achieve accurate and efficient control of the building environment.

[0070] Step 1): The input parameters of the prediction model include the current environmental parameters, the historical operating status of the building equipment, and the current setting values. The recursive optimization strategy is used to calculate the predicted values ​​of the environmental parameters at future moments.

[0071] Step 2): During the prediction process, the improved DIFC algorithm ensures that the predicted environmental parameter values ​​are always within the comfort range obtained by cluster analysis. At the same time, the operating parameters of each building equipment are dynamically adjusted with the goal of optimizing the equipment operating efficiency.

[0072] Step 3): The DIFC algorithm outputs key operating parameters such as fan speed, cooling capacity, and air circulation frequency for a variety of building equipment including air conditioners, ventilation equipment, humidifiers, etc., and converts these parameters into specific control instructions and sends them to the corresponding equipment to achieve real-time and dynamic adjustment of the building environment. Under the premise of ensuring the comfort of the indoor environment, it maximizes energy utilization efficiency and reduces building energy consumption.

[0073] The step 4 is specifically as follows:

[0074] Step 4.1: Build a state-space model

[0075] The state-space model of the built environment is defined as follows:

[0076] x t+1 =Ax t +Bu t +w t (4-1)

[0077] y t =Cx t +v t

[0078] in:

[0079] x t is the environmental state vector at time t (including temperature, humidity, CO2 concentration, etc.);

[0080] ut To control vectors (including air conditioning cooling capacity, ventilation equipment wind speed, etc.);

[0081] y t is the output observation vector;

[0082] A, B, and C are the dynamic matrices of the system, which are jointly determined by physical modeling and data fitting;

[0083] w t and v t are process noise and observation noise respectively, which satisfy Gaussian distribution.

[0084] Step 4.2: Initialize model input

[0085] The input data includes:

[0086] Real-time environment status x t (such as current temperature, humidity, air quality parameters);

[0087] History control records t-1 ,u t-2 , ...}

[0088] Comfort interval constraint S = {(x min , x max )};

[0089] Step 4.3: Predict future states

[0090] Through recursive optimization, the DIFC algorithm predicts the state evolution within the next T steps:

[0091]

[0092] in Represents the predicted future state. The goal is to ensure that the predicted value meets the comfort interval;

[0093]

[0094] Step 4.4: Optimize the control vector

[0095] To optimize equipment energy consumption, a performance index J is set to minimize the control cost:

[0096]

[0097] Where γ is the target value of the comfort interval:

[0098] Q is the state error weight matrix;

[0099] R is the control cost weight matrix;

[0100] Through the gradient descent of the constrained optimization algorithm, the optimal control vector that minimizes is solved:

[0101] {u t ,u t+1 ,...,u t+T}

[0102] Step 4.5: Output control instructions

[0103] According to the optimization results, extract the first control vector u t As the equipment operation parameters at the current moment, they specifically include:

[0104] Air conditioning cooling capacity (unit: kW);

[0105] Ventilation equipment fan speed (unit: RPM);

[0106] Humidifier working power (unit: W)

[0107] These instructions are sent to the corresponding equipment through the control interface to adjust the building environment in real time;

[0108] Step 4.6: Dynamic Adjustment and Feedback

[0109] The system updates the state x at regular intervals t and control vector u t , and re-execute steps 4.3 to 4.5 to form a closed-loop control to ensure that the environmental parameters are dynamically stable within the comfort range and the energy consumption is minimized.

[0110] The step 5 is specifically as follows:

[0111] The visualization interface includes the environmental data collected in real time in step 1 and the prediction results of the distributed control algorithm obtained in step 4; the interface supports real-time monitoring and adjustment, and users can tune system parameters according to actual needs.

[0112] Beneficial effects of the present invention:

[0113] The present invention has excellent control accuracy and powerful scene adaptation capabilities, and can implement fully automated, all-weather and multi-dimensional collaborative low-carbon control strategies for key control factors such as temperature and humidity, energy consumption and air quality. Through real-time feedback control and adjustment of specific circumstances, a complete closed-loop control system is constructed, and the control results are then visualized to achieve full-domain coordination and energy saving and consumption reduction effects. Compared with traditional methods, the present invention has good coordination control capabilities and practicality while meeting control requirements and accuracy, and can generate greater economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0114] Figure 1It is a flow chart of a building global collaborative perception low-carbon control method of the present invention.

[0115] Figure 2 This is a block diagram of the system composition of the present invention.

[0116] Figure 3 This is a block diagram of the global sensor system of the present invention. DETAILED DESCRIPTION

[0117] The present invention will be further described in detail below in conjunction with the accompanying drawings.

[0118] like Figure 1 , Figure 2 As shown, a building global collaborative perception low-carbon control method includes the following steps:

[0119] Step 1: Figure 3 As shown, a global sensor system is constructed. The global sensor system is used to obtain environmental data inside and outside the building. The system includes temperature and humidity sensors, energy consumption monitoring sensors, and air quality monitoring sensors.

[0120] Temperature and humidity sensors accurately collect temperature and humidity data inside and outside the building, reflecting climate conditions and human comfort perception; energy consumption monitoring sensors capture building electricity, water, gas and other energy consumption information in real time to help analyze energy consumption patterns; air quality detection sensors monitor indoor and outdoor PM2.5, formaldehyde, and total volatile organic carbon (TVOC) concentrations, which are related to personnel health and ventilation needs; among them,

[0121] Temperature and humidity sensors are located at the center of the wall of each room 1.5-2 meters above the ground, evenly spaced about 1.5-2 meters high along the wall of the corridor, about 1.5-2 meters high on each platform of the stairwell, 2-3 meters high on the outer wall of the outdoor building and in the open space on the roof;

[0122] Energy consumption monitoring sensors are installed on the power distribution boxes on each floor and the power supply lines of large power-consuming equipment. For water, they are installed on the main water inlet pipe and branch water pipes. For gas, they are installed on the main gas meter inlet pipe.

[0123] Air quality monitoring sensors are installed indoors in the corners of the room 1-1.5 meters from the ground and near the vents, and outdoors on the roof. These installation locations are designed to accurately collect data, thereby achieving effective monitoring and control of indoor and outdoor environmental conditions and energy consumption.

[0124] Step 2: Preprocess the collected data to improve data quality and accuracy, laying the foundation for multi-perspective cluster analysis. First, filter the noise of the collected data, and use a median filter-based method to remove outliers in the data to ensure the stability of the data. Then, interpolate to fill in missing values, and time-align data from different sensor sources to ensure data synchronization. Next, standardize the aligned data and convert each indicator into a dimensionless form to eliminate the interference of dimensional differences on subsequent analysis;

[0125] Specifically,

[0126] Step 2.1: Filter the noise from the data collected by the sensor

[0127] The median filter method is used to process the data, making the processed data more stable and reducing data fluctuations. Taking temperature data as an example, humidity data, energy consumption data and other related data can all be processed in the same way.

[0128] Assume that the data sequence collected by the temperature sensor is expressed as T = [t 1 , t 2 , t 3 , ..., t n ], where t i It represents the temperature value collected at the i-th moment, and n is the number of collected data points.

[0129] Median filtering is to select a window with a fixed length of odd number m for each data point in the sequence. i , the data in the window is Then sort the data in this window and take the middle value after sorting to replace the original data point t i .

[0130] Step 2.2: Fill in the missing data points of the sensor data

[0131] For data series with missing data, linear interpolation is used to fill in the missing data points, so that the entire data series is complete in the time dimension, ensuring that subsequent analysis will not be biased or impossible to perform for multiple data points with missing values ​​due to missing data. Taking energy consumption data as an example, humidity data, temperature data and other related data can refer to this example and use the same processing method.

[0132] In the collected energy consumption data sequence (set as E = [e 1 , e 2 , e 3 , ..., e n ]), if the data has some missing values, such as e jFor missing values ​​(1 <j<n)。

[0133] Assume that e is known j-1 and e j+1 The value of , based on the linear relationship to calculate the missing e j The specific linear interpolation calculation formula is formula (2-1):

[0134]

[0135] For multiple data points with missing values, this method is used to complete them one by one, so that the entire energy consumption data series is complete and available.

[0136] Step 2.3: Time-align and synchronize data from different sensors

[0137] The time alignment operation aims to build a unified data time reference. Specifically, the earliest time of data collection among each sensor is used as the starting point, and the data is reorganized and integrated according to fixed time intervals.

[0138] The data collected by different sensors are extracted and processed separately according to this unified time interval to ensure that at each unified time node after integration, data information on temperature and humidity, energy consumption, air quality, etc. can be covered at the same time. This effectively guarantees the synchronization of data and lays a solid foundation for subsequent comprehensive analysis of the internal correlation between different control elements, so that it can be smoothly promoted and more accurate and valuable analysis results can be obtained.

[0139] Step 2.4: Standardize the data

[0140] The data is processed using Z-score standardization to convert each indicator into a dimensionless form. In subsequent operations such as multi-perspective cluster analysis, different indicators can participate in calculation and analysis on the basis of equal importance, avoiding the problem of unbalanced weights of certain indicators in the analysis due to different dimensions.

[0141] Taking temperature data as an example, humidity data, temperature data and other related data can be processed in the same way. The calculation formula is (2-2), is the mean of the temperature data series, calculated as (2-3), σ t is the standard deviation of the temperature data series, and the calculation formula is (2-4).

[0142]

[0143] Step 3: Build a multi-view data matrix for the data processed in step 2, and input the multi-view data matrix into the multi-view K-Means clustering algorithm for analysis. The algorithm takes the building environment comfort and low carbon as the goal, comprehensively analyzes the correlation between key control factors, mines the potential environmental patterns in the data, and determines the comprehensive comfort range;

[0144] The algorithm takes the comfort and low carbon of the building environment as its goal, comprehensively analyzes the correlation between key control factors, and mines the potential environmental patterns in the data. The multi-perspective K-Means clustering algorithm first performs a preliminary division of the multi-dimensional data based on the multi-perspective data matrix, and uses the weighted Euclidean distance as a metric in each perspective to cluster the data points into different clusters; each cluster represents an environmental pattern, and the multi-perspective K-Means clustering algorithm further integrates the clustering results of all perspectives, and uses a joint optimization method to determine a comprehensive comfort interval. The comprehensive comfort interval is expressed as a multi-dimensional parameter range, including the upper and lower limits of temperature, humidity, and air quality indicators (PM2.5, formaldehyde, and total volatile organic carbon (TVOC)).

[0145] The specific implementation steps of step 3 are:

[0146] 3.1 Constructing a multi-view data matrix

[0147] Extract multidimensional key control factors from the preprocessed data obtained in step 2, and construct a multi-perspective data matrix with time as the horizontal axis and indicators as the vertical axis. The perspectives include time perspective, space perspective, and functional perspective.

[0148] Time perspective: defined as in Represents the multidimensional indicator vector at the i-th time point;

[0149] Spatial perspective: defined as in Represents a multidimensional index vector at the spatial position s;

[0150] Functional perspective: defined as in A vector representing a specific functional metric, such as energy consumption or air quality.

[0151] The matrix is ​​stored in [n,m] dimensions, where n is the number of time points or spatial locations, and m is the dimension of the control factor (such as temperature, humidity, etc.)

[0152] 3.2 Preliminary clustering

[0153] K-Means algorithm is applied to the local data subset of each perspective for preliminary clustering analysis. In each perspective, the similarity between data points is measured by weighted Euclidean distance, and the appropriate number of clusters is set to divide the multidimensional data into several clusters. Each cluster represents a typical environmental pattern under this perspective, and the clustering results of each perspective are output, including the cluster center point and the data distribution within the cluster.

[0154] For each viewpoint, the data matrix X v (v∈{t,s,f}) uses K-Means algorithm for preliminary clustering. The goal is to v Divide into k clusters The center point of each cluster Defined as, the calculation formula is formula (3-1):

[0155]

[0156] The weighted Euclidean distance calculation formula from the data point within the cluster to the center point is formula (3-2),

[0157]

[0158] where w l is the weight of the lth indicator, which is used to highlight the importance of specific control factors. The clustering objective function calculation formula is formula (3-3),

[0159]

[0160] Output the preliminary clustering results for each viewpoint, including the center point of each cluster And the distribution of data within the family

[0161] 3.3 Fusion of multi-view results

[0162] In order to integrate the clustering results of multiple perspectives, a joint optimization method is used to fuse the clustering information of each perspective. Specifically, based on the cluster center points of all perspectives, a global weighted optimization objective function is established. By iteratively adjusting the center positions of each cluster, the cluster division result that optimally classifies the global data is found. This process not only retains the characteristics of each perspective, but also comprehensively reflects the relevance of the global pattern.

[0163] Based on the preliminary clustering results of all perspectives, a global weighted optimization objective function is established. The objective function calculation formula is (3-4):

[0164]

[0165] Among them, α v is the weight for each view, μ jis the global cluster center point, and the objective function updates the multi-view cluster center point through selection, integrates the characteristics of different viewpoints, and finally outputs the global family division {C 1 ,C 2 ,...C k} and the center point μ j .

[0166] 3.4 Generate comprehensive comfort zone

[0167] Combined with the final global clustering results, the comprehensive comfort range of the building environment is extracted. The range of the comfort range is determined by the center point and distribution characteristics of each cluster, including the upper and lower limits of parameters such as temperature, humidity, and air quality (such as CO2 concentration and PM2.5 concentration). The output comfort range is expressed in a multidimensional range to provide a benchmark for subsequent control algorithms.

[0168] The comprehensive comfort interval is extracted based on the final clustering results, which is defined as the center point of each cluster and its data distribution range:

[0169] S={(min(x l ),max(x l ))|x l ∈C j ,l=1,2,...,m}

[0170] The comfort zone includes the upper and lower limits of parameters such as temperature, humidity, and air quality (such as CO2 concentration and PM2.5 concentration), providing a benchmark for subsequent control algorithms.

[0171] Step 4: After obtaining the comfort interval in step 3, the DIFC algorithm uses the current environmental state as the initial condition and the comfort interval as the target constraint, and uses the DIFC algorithm to dynamically predict the evolution trend of future environmental parameters, and finally outputs the optimal operating parameters of each building equipment.

[0172] The step 4 is specifically as follows:

[0173] The dynamic interval prediction and control algorithm takes the current environmental state as the initial condition and sets the comfort interval as the target constraint. It builds a prediction model by integrating physical laws and data-driven methods, dynamically predicts the evolution trend of future environmental parameters, and then outputs the optimal operating parameters of each building equipment to achieve accurate and efficient control of the building environment.

[0174] Step 1): The input parameters of the prediction model include the current environmental parameters, the historical operating status of the building equipment, and the current setting values. The recursive optimization strategy is used to calculate the predicted values ​​of the environmental parameters at future moments.

[0175] Step 2): During the prediction process, the improved DIFC algorithm ensures that the predicted environmental parameter values ​​are always within the comfort range obtained by cluster analysis. At the same time, the operating parameters of each building equipment are dynamically adjusted with the goal of optimizing the equipment operating efficiency.

[0176] Step 3): The DIFC algorithm outputs key operating parameters such as fan speed, cooling capacity, and air circulation frequency for a variety of building equipment including air conditioners, ventilation equipment, humidifiers, etc., and converts these parameters into specific control instructions and sends them to the corresponding equipment to achieve real-time and dynamic adjustment of the building environment. Under the premise of ensuring the comfort of the indoor environment, it maximizes energy utilization efficiency and reduces building energy consumption.

[0177] The specific implementation steps of step 4 are:

[0178] Step 4.1: Build a state-space model

[0179] The state-space model of the built environment is defined as follows:

[0180] x t+1 =Ax t +Bu t +w t (4-1)

[0181] y t =Cx t +v t

[0182] Among them, x t is the environmental state vector at time t (including temperature, humidity, CO2 concentration, etc.); u t is the control vector (including air conditioning cooling capacity, ventilation equipment wind speed, etc.); t is the output observation vector; A, B, C are the dynamic matrices of the system, which are determined jointly by physical modeling and data fitting; w t and v t are process noise and observation noise respectively, which satisfy Gaussian distribution.

[0183] Step 4.2: Initialize model input

[0184] The input data includes:

[0185] Real-time environment status x t (such as current temperature, humidity, air quality parameters);

[0186] History control records t-1 ,u t-2 ,...}

[0187] Comfort interval constraint S = {(x min ,xmax )}.

[0188] Step 4.3: Predict future environmental states

[0189] Through recursive optimization, the DIFC algorithm predicts the state evolution within the next T steps:

[0190]

[0191] in Represents the predicted future state. The goal is to ensure that the predicted value meets the comfort interval;

[0192]

[0193] Step 4.4: Optimize the control vector

[0194] To optimize equipment energy consumption, a performance index J is set to minimize the control cost:

[0195]

[0196] Where γ is the target value of the comfort interval:

[0197] Q is the state error weight matrix;

[0198] R is the control cost weight matrix.

[0199] Through the gradient descent of the constrained optimization algorithm, the optimal control vector that minimizes

[0200] {u t ,u t+1 ,...,u t+T}

[0201] Step 4.5: Output control instructions

[0202] According to the optimization results, extract the first control vector u t As the equipment operation parameters at the current moment, they specifically include:

[0203] Air conditioning cooling capacity (unit: kW);

[0204] Ventilation equipment fan speed (unit: RPM);

[0205] Humidifier working power (unit: W)

[0206] These instructions are sent to the corresponding equipment through the control interface to adjust the building environment in real time;

[0207] Step 4.6: Dynamic Adjustment and Feedback

[0208] The system updates the state x at regular intervalst and control vector u t , and re-execute steps 4.3 to 4.5 to form a closed-loop control to ensure that the environmental parameters are dynamically stable within the comfort range and the energy consumption is minimized.

[0209] Step 5: Display the data and analysis results of the entire process through the visualization module.

[0210] The visualization interface includes real-time collected environmental data curves, prediction results of distributed control algorithms, and equipment operating status, etc. The interface supports real-time monitoring and adjustment, and users can tune system parameters according to actual needs.

[0211] Through the above method, efficient management and optimization of the building environment can be finally achieved. This method collects key data by arranging sensors, determines the comfort zone through preprocessing and multi-view clustering analysis, and then uses the improved DIFC algorithm to dynamically adjust the building equipment to keep the indoor environment in the best comfortable state, effectively reduce building energy consumption, reduce energy waste, achieve the goal of low-carbon and environmental protection, and intuitively display data through visualization modules to provide strong support for the sustainable development of buildings.

Claims

1. A low-carbon control method for global collaborative perception of buildings, characterized in that: The steps include: Step 1: Arrange temperature and humidity sensors, energy consumption monitoring sensors and air quality monitoring sensors inside and outside the building to collect data on key control factors such as temperature and humidity, energy consumption and air quality in the building's internal and external environments; Step 2: Preprocess the collected data; Step 3: construct the data processed in step 2 into a multi-view data matrix, and use the multi-view K-Means clustering algorithm to analyze the multi-view data matrix to obtain the comfort interval; Step 4: Based on the comfort interval, an improved dynamic interval prediction and control algorithm is used to dynamically adjust the building equipment, so that the indoor environment is always in the best comfort state, reducing energy waste, reducing building energy consumption, and achieving low-carbon environmental protection; Step 5: Display the data and analysis results of the entire process through the visualization module.

2. A building global collaborative perception low-carbon control method according to claim 1, characterized in that: The step 1 is specifically as follows: Temperature and humidity sensors are used to accurately collect temperature and humidity data inside and outside the building, reflecting climate conditions and human comfort perception; Energy consumption monitoring sensors are used to capture building electricity, water, gas and other energy consumption information in real time to help analyze energy consumption patterns; Air quality detection sensors are used to monitor indoor and outdoor PM2.5, formaldehyde, and total volatile organic carbon (TVOC) concentrations, which are related to human health and ventilation needs.

3. A building global collaborative perception low-carbon control method according to claim 2, characterized in that: Temperature and humidity sensors are located at the center of the wall of each room 1.5-2 meters above the ground, evenly spaced about 1.5-2 meters high along the wall of the corridor, about 1.5-2 meters high on each platform of the stairwell, 2-3 meters high on the outer wall of the outdoor building and in the open space on the roof; Energy consumption monitoring sensors are installed on the power distribution boxes on each floor and the power supply lines of large power-consuming equipment. For water, they are installed on the main water inlet pipe and branch water pipes. For gas, they are installed on the main gas meter inlet pipe. The air quality monitoring sensor is placed indoors in the corner of the room 1-1.5 meters above the ground and near the vents, and outdoors on the roof.

4. A building global collaborative perception low-carbon control method according to claim 2, characterized in that: The specific implementation steps of step 2 are: Step 2.1: Filter the noise from the data collected by the sensor The data is processed using the median filtering method; The data sequence of the temperature sensor collected is assumed to be T = [t1, t2, t3, ..., t n ], where t i represents the temperature value collected at the i-th moment, and n is the number of collected data points; Median filtering is to select a window with a fixed length of odd number m for each data point in the sequence. i , the data in the window is Then sort the data in this window and take the middle value after sorting to replace the original data point t i . Step 2.2: Fill in the missing data points of the sensor data For data series with missing data, linear interpolation method is used to fill in the missing data points, so that the entire data series is complete in the time dimension, ensuring that subsequent analysis will not be biased or unable to be performed due to missing data for multiple data points with missing values; In the collected energy consumption data sequence E(E=[e1,e2,e3,...,e n ]), if there are some missing values ​​in the data, e j For missing values ​​(1 <j<n); Assume known j-1 and e j+1 The value of , based on the linear relationship to calculate the missing e j , the specific linear interpolation calculation formula is formula (2-1): For multiple data points with missing values, fill them in one by one according to this method; Step 2.3: Time-align and synchronize data from different sensors The earliest collection time of each sensor is used as the starting point, and the data is reorganized and integrated according to fixed time intervals; The data collected by different sensors are extracted and processed separately according to this unified time interval to ensure that at each unified time node after integration, data information on temperature, humidity, energy consumption, air quality, etc. can be simultaneously covered, so as to effectively ensure the synchronization of data; Step 2.4: Standardize the data The data were processed using Z-score standardization; For temperature T, the standardized value The calculation formula is (2-2), and the data sequence of the collected temperature sensor is expressed as T = [t1, t2, t3, ..., t n ], where t i represents the temperature value collected at the i-th moment, n is the number of collected data points, is the mean of the temperature data series, calculated as (2-3), σ t is the standard deviation of the temperature data series, and the calculation formula is (2-4):

5. A building global collaborative perception low-carbon control method according to claim 4, characterized in that: The specific implementation steps of step 3 are: Step 3.1: Construct a multi-view data matrix Extract multi-dimensional key control factors from the pre-processed data obtained in step 2, and construct a multi-perspective data matrix with time as the horizontal axis and indicators as the vertical axis. The perspectives include time perspective, space perspective, and functional perspective. Time perspective: defined as in Represents the multidimensional indicator vector at the i-th time point; Spatial perspective: defined as in Represents a multidimensional index vector at the spatial position s; Functional perspective: defined as in A vector representing a specific functional indicator (such as energy consumption or air quality); The matrix is ​​stored in [n,m] dimensions, where n is the number of time points or spatial locations, and m is the dimension of the control element (such as temperature, humidity, etc.); Step 3.2: Divide the initial clusters K-Means algorithm is applied to the local data subset of each perspective for preliminary clustering analysis. In each perspective, the similarity between data points is measured by weighted Euclidean distance, and the appropriate number of clusters is set to divide the multidimensional data into several clusters. Each cluster represents a typical environmental pattern under this perspective, and the clustering results of each perspective are output, including the cluster center point and the data distribution within the cluster; For each viewpoint, the data matrix X v (v∈{t,s,f}) uses the K-Means algorithm for preliminary clustering, the goal is to v Divide into k clusters The center point of each cluster The calculation formula is (3-1), The weighted Euclidean distance calculation formula from the data point within the cluster to the center point is as follows: where w l is the weight of the lth indicator, and the clustering objective function formula is (3-3): Finally, the preliminary clustering results of each perspective are output, including the center point of each cluster. And the distribution of data within the family Step 3.3: Fusion of multi-view clustering The clustering information of each perspective outputted in step 3.2 is integrated by a joint optimization method. Specifically, a global weighted optimization objective function is established based on the cluster center points of all perspectives. The cluster division result that enables the optimal classification of global data is found by iteratively adjusting the center positions of each cluster. Based on the preliminary clustering results of all perspectives, a global weighted optimization objective function is established: Among them, α v is the weight for each view, μ j The global cluster center is the center point of the cluster. Through the selection and update of the multi-view cluster center point, the characteristics of different viewpoints are integrated, and the global family partition {C1, C2, ...C k } and the center point μ j ; Step 3.4: Generate comprehensive comfort zone Combined with the multi-view clustering results generated in step 3.3, the comprehensive comfort range of the building environment is extracted. The range of the comfort range is determined by the center point and distribution characteristics of each cluster, including the upper and lower limits of parameters such as temperature, humidity, and air quality. The output comfort range is expressed in a multi-dimensional range to provide a benchmark for subsequent control algorithms. The comprehensive comfort interval is extracted based on the final clustering results, which is defined as the center point of each cluster and its data distribution range: S={(min(x l ),max(x l ))|x l ∈C j ,l=1,2,...,m} The comfort zone includes the upper and lower limits of parameters such as temperature, humidity, and air quality, providing a benchmark for subsequent control algorithms.

6. A building global collaborative perception low-carbon control method according to claim 5, characterized in that: The step 4 is specifically as follows: The dynamic interval prediction and control algorithm takes the current environmental state as the initial condition and sets the comfort interval as the target constraint. It builds a prediction model by integrating physical laws and data-driven methods, dynamically predicts the evolution trend of future environmental parameters, and then outputs the optimal operating parameters of each building equipment to achieve accurate and efficient control of the building environment. Step 1): The input parameters of the prediction model include the current environmental parameters, the historical operating status of the building equipment, and the current setting values. The recursive optimization strategy is used to calculate the predicted values ​​of the environmental parameters at future moments. Step 2): During the prediction process, the improved DIFC algorithm ensures that the predicted environmental parameter values ​​are always within the comfort range obtained by cluster analysis. At the same time, the operating parameters of each building equipment are dynamically adjusted with the goal of optimizing the equipment operating efficiency. Step 3): The DIFC algorithm outputs key operating parameters for building equipment, and converts the parameters into specific control instructions and sends them to the corresponding equipment to achieve real-time and dynamic adjustment of the building environment.

7. A building global collaborative perception low-carbon control method according to claim 6, characterized in that: The step 4 is specifically as follows: Step 4.1: Build a state-space model The state space model of the built environment is defined as follows: x t+1 =Ax t +Bu t +w t (4-1) y t =Cx t +v t in: x t is the environmental state vector at time t; u t is the control vector; y t is the output observation vector; A, B, and C are the dynamic matrices of the system, which are determined jointly by physical modeling and data fitting; w t and v t are process noise and observation noise respectively, satisfying Gaussian distribution; Step 4.2: Initialize model input The input data includes: Real-time environment status x t ; History control records t-1 ,u t-2 ,...} Comfort interval constraint S = {(x min ,x max )}; Step 4.3: Predict future states Through recursive optimization, the DIFC algorithm predicts the state evolution within the next T steps: in, Represents the predicted future state. The goal is to ensure that the predicted value meets the comfort interval; Step 4.4: Optimize the control vector To optimize equipment energy consumption, a performance index J is set to minimize the control cost: Where γ is the target value of the comfort interval: Q is the state error weight matrix; R is the control cost weight matrix; Through the gradient descent of the constrained optimization algorithm, the optimal control vector that minimizes is solved: {in t ,in t+1 ,...,in t+T } Step 4.5: Output control instructions According to the optimization results, extract the first control vector u t As the equipment operation parameters at the current moment, they specifically include: Air conditioning cooling capacity (unit: kW); Ventilation equipment fan speed (unit: RPM); Humidifier working power (unit: W) These instructions are sent to the corresponding equipment through the control interface to adjust the building environment in real time; Step 4.6: Dynamic Adjustment and Feedback The system updates the state x at regular intervals t and control vector u t , and re-execute steps 4.3 to 4.5 to form a closed-loop control to ensure that the environmental parameters are dynamically stable within the comfort range and the energy consumption is minimized.

8. The method for building global collaborative perception low-carbon control according to claim 1 is characterized in that: The step 5 is specifically as follows: The visualization interface includes the environmental data collected in real time in step 1 and the prediction results of the distributed control algorithm obtained in step 4; the interface supports real-time monitoring and adjustment, and users can tune system parameters according to actual needs.

Images