All-equipment energy consumption optimization management system in building

Abstract

The invention discloses a full-equipment energy consumption optimization management system in a building, and relates to the technical field of intelligent control, and the system is provided with a multi-modal data collection module which collects building environment parameters and equipment operation data through a distributed sensor network, and generates a multi-modal sensing matrix; an environment state prediction module is set to construct a dynamic knowledge graph based on a multi-modal sensing matrix, a graph attention network is used, an energy flow matrix of equipment topology is output, micro-environment state prediction data is generated through prediction of a finite element solver, a scheduling scheme solving module is set, a mixed integer programming model is constructed based on the micro-environment state prediction data, and a scheduling scheme is set. And solving the mixed integer programming model to obtain a scheduling scheme of the energy consumption equipment, setting a control instruction optimization module, and driving the edge execution equipment to operate based on the scheduling scheme. The global scheduling optimization of the whole building energy consumption equipment is realized, the energy consumption is effectively reduced, and the environmental comfort is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, in particular to a full-equipment energy consumption optimization management system in a building. Background Art

[0002] Modern buildings are equipped with a large number of energy-consuming devices, such as air conditioning, lighting, ventilation systems, and other automated equipment. While these devices meet the needs of occupant comfort and safety, they often use static, pre-set control strategies, resulting in energy waste and environmental fluctuations. Traditional building energy management systems, which mostly rely on single sensor monitoring and empirical regulation, struggle to simultaneously account for real-time environmental changes and equipment operating status, and therefore cannot achieve dynamic and accurate energy optimization. Therefore, there is an urgent need for an intelligent energy optimization management solution that can dynamically integrate the operating data of various energy-consuming devices based on real-time environmental perception, achieving the goal of reducing energy consumption while ensuring the comfort and safety of the building environment.

[0003] Existing systems are usually based on fixed parameters and static models, which are unable to cope with the complex and dynamically changing environmental factors in buildings. They also lack the means to establish a unified model for the interaction of multiple physical fields, resulting in a lack of effective connection between energy consumption regulation and environmental status prediction.

[0004] To this end, the present invention proposes an energy consumption optimization management system for all equipment in a building. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes a building-wide energy consumption optimization management system that achieves global scheduling optimization of all energy-consuming equipment in the building, effectively reducing energy consumption and improving environmental comfort.

[0006] In order to achieve the above objectives, a building full equipment energy consumption optimization management system is proposed.

[0007] A building-wide energy consumption optimization management system includes a multimodal data collection module, an environmental state prediction module, a scheduling solution module, and a control instruction optimization module; wherein each module is electrically connected;

[0008] The multimodal data collection module collects building environmental parameters and equipment operation data through a distributed sensor network, generates a multimodal perception matrix after spatiotemporal alignment and anomaly detection, and sends the multimodal perception matrix to the environmental state prediction module;

[0009] An environment state prediction module constructs a dynamic knowledge graph based on the multi-modal perception matrix, quantifies the energy coupling strength between devices in the dynamic knowledge graph using a graph attention network, outputs an energy flow matrix of the device topology, and then combines the energy flow matrix with a multi-physics coupling model to generate micro-environment state prediction data through a finite element solver; and sends the micro-environment state prediction data to a scheduling scheme solving module;

[0010] A scheduling scheme solving module constructs a mixed integer programming model based on the micro-environment state prediction data, solves the mixed integer programming model, obtains a scheduling scheme for energy-consuming devices, and sends the scheduling scheme to a control instruction optimization module;

[0011] A control instruction optimization module drives edge execution devices to operate based on the scheduling scheme.

[0012] The collection of building environment parameters and device operation data through a distributed sensor network includes the following steps:

[0013] Step 111: According to the design drawings and BIM data of the building, a plurality of positions in the building are selected as key positions, and environment sensors and state sensors are installed at the key positions to form a distributed sensor network;

[0014] Step 112: The environment sensors and state sensors collect parameter values of corresponding environment parameters and parameter values of operation state parameters in real time to form an original data set.

[0015] The multi-modal perception matrix is generated after time-space alignment and anomaly detection processing in the following manner:

[0016] Step 121: The original data set is transmitted to the edge gateway through the network connected by the corresponding physical sensors and the edge gateway; the edge node verifies the consistency of the time stamps of all data using the NTP protocol or the PTP protocol, and then uses a filtering algorithm to denoise the data to obtain a smoothed data set;

[0017] Step 122: Based on the smoothed data set, time-space alignment and multi-dimensional matrix construction are performed to form a structured multi-modal perception matrix.

[0018] The multi-modal perception matrix is constructed in the following manner:

[0019] After obtaining the smoothed data set, the parameter values collected from physical sensors of different positions and types are time-space aligned according to a unified time stamp to ensure that all data are corresponded at the same time point;

[0020] For each key position, the types of environmental parameters and operating state parameters are integrated into a multi-dimensional data matrix, each row of the multi-dimensional data matrix representing a physical sensor, and each column corresponding to an environmental parameter or operating state parameter;

[0021] The three-dimensional matrix formed by the multi-dimensional data matrices of all key positions is a multi-modal perception matrix.

[0022] The manner of constructing a dynamic knowledge graph based on the multi-modal perception matrix is:

[0023] Step 211: By analyzing and extracting features of the building environmental parameters and equipment operating data collected in the multi-modal perception matrix, an initial graph node reflecting each key position in the building is constructed;

[0024] Step 212: Based on the spatio-temporal correlation of the data in the multi-modal perception matrix and the physical topology of the building, the relationship between the nodes is constructed, and a quantitative weight is assigned to each edge to reflect the energy coupling or operating state correlation strength between the nodes;

[0025] Step 213: Based on the extracted graph nodes and graph edges, a dynamic knowledge graph is constructed.

[0026] The manner of using a graph attention network to quantify the energy coupling strength between devices in the dynamic knowledge graph and outputting an energy flow matrix of the device topology is:

[0027] Step 221: Using a graph attention network, the initially constructed static knowledge graph is dynamically updated, the node representation and edge weight are adjusted according to the real-time collected multi-modal perception matrix, and a dynamic knowledge graph reflecting the real-time state change in the building is generated;

[0028] Step 222: Using a graph attention network, the updated node attributes of the dynamic knowledge graph are calculated for neighborhood attention coefficients, and the neighborhood information is aggregated to update the node hidden layer representation;

[0029] Step 223: The edge weight is dynamically adjusted according to the attention coefficients between the graph nodes and the initial edge weight, thereby constructing an energy flow matrix reflecting the real-time coupling strength;

[0030] Step 224: The attributes of the graph nodes and the edge weights of the edges in the dynamic knowledge graph are dynamically updated based on the updated representation of the updated graph nodes and the updated edge weights between the graph nodes.

[0031] The manner of combining the energy flow matrix with a multi-physical field coupling model to predict and generate micro-environment state prediction data through a finite element solver includes the following steps:

[0032] Step 231: Based on the building's geometric information and material properties, perform finite element meshing on the building's interior, discretize the continuous space into multiple finite element regions, and use the heat conduction equation and Navier-Stokes equations to construct a discretized description of the heat conduction and fluid flow physical fields.

[0033] Step 232: coupling the heat conduction equation and the Navier-Stokes equation within the finite element region, and introducing real-time coupling information provided by the energy flow matrix into the multi-physics coupling model to form a multi-physics coupling model that describes the interaction between temperature change and air flow;

[0034] Step 233: After constructing the coupled multi-physics coupling model, a finite element solver is used to numerically solve the discretized multi-physics coupling model to generate microenvironment state prediction data.

[0035] The constructing of the mixed integer programming model based on the microenvironment state prediction data comprises the following steps:

[0036] Step 311: using the microenvironment state prediction data generated by the finite element solver, extracting key physical field parameters in the building, and mapping the key physical field parameters into input parameters of the mixed integer programming model;

[0037] Step 312: Set control variables and continuous variables for each device, and based on the set control variables, continuous variables, and mapped microenvironment status indicators, construct the objective function and constraints of a mixed integer programming model that optimizes energy consumption while meeting the requirements of building interior environmental comfort.

[0038] The objective function and constraint conditions of the mixed integer programming model are constructed in the following ways:

[0039] The goal of the MIP model is to minimize the overall energy consumption of the equipment;

[0040] To ensure that the microenvironment in the building meets the predetermined comfort and safety requirements, an ambient temperature constraint is constructed for any area, which requires the temperature to be kept within the safety range, a wind speed comfort constraint requires the air flow rate to be kept within the wind speed comfort range, and a pressure comfort constraint requires the equipment pressure to be kept within the pressure comfort range.

[0041] The method of solving the mixed integer programming model to obtain the scheduling plan for energy-consuming equipment is:

[0042] The objective function and all constraint conditions of the mixed integer programming model are input into an integer programming model solver, which solves the mixed integer programming model for any discrete time t, generates an approximate optimal solution of the control variable and continuous variable for each device; thereby outputting a scheduling scheme to determine the operating state and output power or wind speed adjustment value of each device.

[0043] The driving edge execution device operation based on the scheduling scheme is as follows:

[0044] Based on the scheduling scheme, the edge control system takes the approximate optimal solution of the control variable and continuous variable of each graph node in the scheduling scheme as the direct operation basis, and automatically adjusts the operating state and output power or wind speed adjustment value of the device by the local controller of the device to the approximate optimal solution of the control variable and continuous variable.

[0045] Compared with the prior art, the beneficial effects of the present application are:

[0046] The present application collects environmental parameters and device operating state data in each area of the building through a distributed sensor network, and pre-processes and structures the data for storage, forming a multi-modal perception matrix. Subsequently, a dynamic knowledge graph is constructed using a graph attention network, quantifying the energy coupling relationship between devices and outputting an energy flow matrix. Then, the energy flow matrix is combined with a multi-physical field coupling model based on building geometric information and material properties, and the micro-environment state data in each area of the building, such as temperature field, air flow field and pressure field, is predicted by a finite element solver. On this basis, a mixed integer programming model is constructed, taking the micro-environment state prediction data as a key constraint parameter to determine the operating state and output parameters of each device to achieve minimum energy consumption while meeting environmental comfort requirements. The global scheduling optimization of all building energy-consuming devices is realized, effectively reducing energy consumption and improving environmental comfort. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 A module connection relationship diagram of a building-wide device energy optimization management system in an embodiment of the present application;

[0048] Figure 2 An example diagram of sensor deployment in a building in an embodiment of the present application. DETAILED DESCRIPTION

[0049] The technical solutions of the present application will be described in detail below with reference to the embodiments, obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0050] like Figure 1 As shown, a building full-equipment energy consumption optimization management system includes a multimodal data collection module, an environmental state prediction module, a scheduling solution module, and a control instruction optimization module; wherein each module is electrically connected;

[0051] The multimodal data collection module collects building environmental parameters and equipment operation data through a distributed sensor network, generates a multimodal perception matrix after spatiotemporal alignment and anomaly detection, and sends the multimodal perception matrix to the environmental state prediction module;

[0052] The environmental state prediction module builds a dynamic knowledge graph based on the multimodal perception matrix, uses a graph attention network to quantify the energy coupling strength between devices in the dynamic knowledge graph, outputs an energy flow matrix of the device topology, combines the energy flow matrix with the multi-physics field coupling model, and generates microenvironmental state prediction data through a finite element solver; and sends the microenvironmental state prediction data to the scheduling solution solution module;

[0053] a scheduling solution module, which constructs a mixed integer programming model based on the microenvironment state prediction data, solves the mixed integer programming model to obtain a scheduling solution for the energy-consuming equipment, and sends the scheduling solution to the control instruction optimization module;

[0054] The control instruction optimization module drives the edge execution device to run based on the scheduling scheme.

[0055] In an embodiment of the present invention, collecting building environment parameters and equipment operation data through a distributed sensor network includes the following steps:

[0056] Step 111: Based on the building design drawings and BIM data, multiple locations within the building are selected as key locations, and environmental sensors and status sensors are installed at the key locations to form a distributed sensor network.

[0057] As an example, the key locations in an office building may include but are not limited to key areas such as office areas, conference rooms, computer rooms, as well as areas with intensive human activity, equipment concentrations, and corner locations. The specific selection of key locations is determined based on the building's design drawings and the area's functions.

[0058] The environmental sensor refers to a physical sensor used to collect various environmental parameters, including but not limited to temperature, humidity, illumination, CO2 concentration and other physical parameters used to characterize the building environment;

[0059] The state sensor refers to a physical sensor used to collect operating state parameters of each energy-consuming device in a building. The operating state parameters include but are not limited to physical parameters that characterize the operating state of the device, such as current, voltage, power, and vibration.

[0060] During actual deployment, all physical sensors can use industrial-grade high-precision equipment to ensure a sampling frequency of one second or faster, and assign a unique identifier to each physical sensor to ensure that the raw data collected by each sensor contains detailed monitoring values ​​and location information to facilitate subsequent data processing.

[0061] An example of a sensor deployed in a building is shown in the figure below. Figure 2 As shown, in Figure 2 In the distributed sensor network, a number of appropriate locations are pre-selected to deploy corresponding sensors, thereby forming a distributed sensor network.

[0062] Step 112: The parameter values ​​of the corresponding environmental parameters and the parameter values ​​of the operating status parameters are collected in real time through the environmental sensors and the status sensors to form an original data set.

[0063] It can be understood that the original data set contains parameters and data of the building environment and equipment operation.

[0064] Furthermore, the method of generating a multimodal perception matrix after the spatiotemporal alignment and anomaly detection processing is as follows:

[0065] Step 121: The original data set is transmitted to the edge gateway through a network connected to the corresponding physical sensor and the edge gateway; the edge node verifies the consistency of the timestamps of all data using the NTP protocol or the PTP protocol, and then uses a filtering algorithm to denoise the data to obtain a smoothed data set.

[0066] Specifically, the filtering algorithm includes any one of a Kalman filtering algorithm and a low-pass filtering algorithm.

[0067] Step 122: Based on the smoothed data set, a structured multimodal perception matrix is ​​formed through spatiotemporal alignment and multidimensional matrix construction.

[0068] Specifically, the multimodal perception matrix is ​​constructed as follows:

[0069] After obtaining the smoothed data set, spatial and temporal alignment is performed on parameter values ​​collected from physical sensors of different locations and types according to a unified timestamp to ensure that all data correspond to the same time point;

[0070] For each key position, the types of environmental parameters and operating state parameters are integrated into a multi-dimensional data matrix, each row of the multi-dimensional data matrix representing a physical sensor, and each column corresponding to an environmental parameter or operating state parameter;

[0071] For example, the number of key positions is marked as s, and for the s-th key position, the corresponding multi-dimensional data matrix may be expressed as:

[0072]

[0073] wherein Z is the number of all physical sensors including environmental sensors and state sensors, Y is the number of sampling time points collected by the distributed sensor network, so that the parameter value in the z-th row and the y-th column of the multi-dimensional data matrix represents the parameter value of the z-th physical sensor at the s-th position at the y-th time point;

[0074] The three-dimensional matrix formed by the multi-dimensional data matrices of all key positions is a multi-modal perception matrix.

[0075] In an embodiment of the present application, the manner of constructing a dynamic knowledge graph based on the multi-modal perception matrix is:

[0076] Step 211: By analyzing and extracting features of the building environmental parameters and equipment operating data collected in the multi-modal perception matrix, an initial graph node reflecting each key position in the building is constructed.

[0077] Specifically, the manner of constructing an initial graph node reflecting each key position in the building is:

[0078] The multi-modal perception matrix formed by the physical sensors at each key position at multiple time nodes is normalized to ensure that the data of the environmental sensors and the state sensors are within a comparable scale.

[0079] Subsequently, using predefined rules and machine learning methods such as threshold segmentation or clustering algorithms, key features are automatically extracted, and representative monitoring objects in the building such as temperature monitoring points, humidity monitoring points, current monitoring points, and equipment state monitoring points are defined as graph nodes.

[0080] Each graph node includes not only the time series data collected by the corresponding physical sensor at the corresponding position in the multi-modal perception matrix, but also static attribute information such as sensor type, installation position, sampling frequency, and initial calibration parameters, thereby constructing an initial graph node set.

[0081] Step 212: based on the spatio-temporal correlation of data in the multi-modal perception matrix and the physical topology of the building, the relationship edges between nodes are constructed, and a quantitative weight is assigned to each edge to reflect the energy coupling or running state correlation strength between nodes;

[0082] Specifically, the way of constructing the relationship edges between nodes and assigning a quantitative weight to each edge is:

[0083] For the time series data of any two nodes in the multi-modal perception matrix, the statistical correlation and physical interaction possibility between any two graph nodes are determined by using time series correlation analysis and spatial proximity index, and the similarity and dependency relationship between any two nodes are converted into edge weight by weighted summation; the time series correlation analysis is, for example, Pearson correlation coefficient, mutual information metric, etc.

[0084] For any two nodes, if the edge weight between the two nodes is greater than a preset weight threshold, a graph edge is set between the two nodes, and the weight of the graph edge is set as the corresponding edge weight.

[0085] Step 213: based on the extracted graph nodes and graph edges, a dynamic knowledge graph is constructed.

[0086] It can be understood that, in the case of known knowledge graph nodes, node attributes, and connection relationships between edges, it is common knowledge in the art to construct a corresponding knowledge graph, for example, by using neo4j to construct a knowledge graph; the present application will not be repeated here.

[0087] Further, the way of using a graph attention network to quantify the energy coupling strength between devices in the dynamic knowledge graph and output the energy flow matrix of the device topology is:

[0088] Step 221: using a graph attention network, the initially constructed static knowledge graph is dynamically updated, the node representation and edge weight are adjusted according to the real-time collected multi-modal perception matrix, and a dynamic knowledge graph reflecting the real-time state change in the building is generated.

[0089] Specifically, the way of dynamically updating the initially constructed static knowledge graph is:

[0090] The latest received multi-modal perception matrix is input into the graph attention network, and the initial attribute vector of each node is iteratively updated.

[0091] The original data of each sensor is extracted and normalized according to a predefined mapping function, the extracted feature value is fused with the original static attribute of the graph node, and a recursive updating strategy is used, so that the new attribute vector of the graph node can reflect both static information and the latest dynamic change; the recursive updating strategy, for example, uses weighted average or a gated recurrent unit to update the state; thus, the attribute representation of each graph node not only includes prior information such as its installation location and sensor type, but also incorporates real-time environmental and device state extracted from the latest sampling data.

[0092] As an example, for example, in an office building, three types of key sensors are deployed: a temperature sensor in the conference room as node 1, a humidity sensor in the conference room as node 2, and a power monitoring sensor in the machine room as node 3, and the static attribute vector of each node is collected in advance through the static attribute information collection method in step 211;

[0093] The collected static attribute vector of each node includes:

[0094] The static attribute vector of node 1 is a1=[type: temperature, location: conference room, c1], the static attribute vector of node 2 is a2=[type: humidity, location: conference room, c2], and the static attribute vector of node 3 is a3=[type: power monitoring, location: machine room, c3]; c1, c2 and c3 are the initial calibration parameters of each node;

[0095] At time t, the real-time data obtained through the multi-modal perception matrix is: the temperature sensor collects a temperature of 25.3°C, the humidity sensor collects a humidity of 45%, and the power monitoring sensor collects a current of 10.5A; after feature extraction and normalization mapping function, the new node attribute vector is updated as:

[0096]

[0097] The mapping function is a function that fuses the original measurement value with the calibration information and outputs a vector with a dimension of 3, so that the state of each node not only retains static information, but also embeds the latest collected real-time data.

[0098] Step 222: using a graph attention network to calculate the neighborhood attention coefficient of the updated node attribute of the dynamic knowledge graph, and aggregating neighborhood information to update the node hidden layer representation;

[0099] Specifically, the neighborhood attention coefficient calculation and neighborhood information aggregation of the updated node attribute of the dynamic knowledge graph to update the node hidden layer representation include:

[0100] The dynamic knowledge graph is marked as G=(V, E), wherein G is the dynamic knowledge graph, V is a set of graph nodes, and E is a set of graph edges;

[0101] For each graph node i and its adjacent nodes in the dynamic knowledge graph , a preliminary attention score is calculated , wherein N(i) is a set of adjacent nodes of the graph node i; the preliminary attention score is calculated in the following manner:

[0102] , wherein a is an attention vector, is a transpose of the attention vector, the symbol represents a vector connection operation, W is a shared linear transformation matrix, and are new attribute vectors of the updated node i and node j, respectively;

[0103] The attention scores of the adjacent nodes are normalized by a Softmax function to obtain an attention coefficient ;

[0104] The hidden layer features of the adjacent nodes of each graph node are weighted and summed by using the attention score and the attention coefficient to generate an updated representation of the graph node ;

[0105] The specific calculation manner is as follows: ; wherein represents a nonlinear activation function; the extraction process of the updated representation realizes adaptive information aggregation of each graph node in the time and space dimensions, and ensures that the graph node representation can sufficiently capture the dynamic change information of other graph nodes in the local network of the graph node.

[0106] As an example, on the basis of the example of step 221, the process of updating the hidden layer representation of the node by step 222 is as follows:

[0107] For node 1, the following is calculated by the formula of : , so that the attention coefficient is obtained by Softmax after normalization:

[0108] ;

[0109] Similarly, the following is obtained: ;

[0110] Similarly, the following is respectively calculated: , , , ;

[0111] Thus, the hidden layer representation of node 1 is:

[0112]

[0113] Similarly, the hidden layer representation of node 2 is calculated as: ≈[1.3875, 0.5375, 0.1875], and the hidden layer representation of node 3 is calculated as .

[0114] Step 223: dynamically adjusting the edge weight according to the attention coefficient between the nodes of the graph and the initial edge weight, so as to construct an energy flow matrix reflecting the real-time coupling strength;

[0115] Specifically, the dynamic adjustment is performed in the following manner:

[0116] For each pair of adjacent nodes of the graph, the edge weight of the pair of nodes is updated according to the product of the calculated edge weight and the attention coefficient;

[0117] Each row of the energy flow matrix corresponds to a node of the graph, and each column also corresponds to a node of the graph. Each element in the energy flow matrix is set to the updated edge weight between the pair of nodes corresponding to the row and the column of the element.

[0118] It can be understood that the initial edge weight is determined by the static data and the physical topology relationship, and reflects the basic coupling possibility between the two nodes of the graph. The attention coefficient reflects the interaction strength in the dynamic state according to the latest data. Therefore, by calculating the product of the initial edge weight and the attention coefficient as the new edge weight, the real-time coupling characteristics between the monitoring points in the building are embodied.

[0119] As an example, based on the example of step 222, the calculation process of the energy flow matrix obtained after step 223 is as follows:

[0120] The initial edge weight of each node is collected in advance, and is respectively: The edge weight between node 1 and node 2 is updated to ; Similarly, the edge weight between node 1 and node 3 is calculated as , , , , ;

[0121] Thus, the energy flow matrix F obtained is:

[0122]

[0123] Step 224: dynamically updating the attributes of the graph nodes and the edge weights of the edges in the dynamic knowledge graph with the updated representations of the graph nodes and the updated edge weights between the pairs of graph nodes.

[0124] Further, the combining the energy flow matrix with the multi-physics coupling model to predict the micro-environment state prediction data by the finite element solver comprises the following steps:

[0125] Step 231: discretizing the interior of the building into a plurality of finite element regions by finite element meshing according to the geometric information and material properties of the building, and constructing a discretized description for the heat conduction and fluid flow physical fields;

[0126] Specifically, the way of constructing the discretized description comprises:

[0127] The interior of the building is divided into a plurality of finite element units according to functional areas or spatial features by using the building BIM data and field measurement, and each finite element unit is assigned with unique physical properties such as density ρ, specific heat c and thermal conductivity k, etc.

[0128] The finite element region plays a role of discretizing the continuous physical field in the multi-physics coupling model, so that the complex partial differential equation can be converted into an algebraic equation solved on the discrete grid.

[0129] For heat conduction, the heat conduction equation is used for each finite element unit:

[0130]

[0131] where T1 represents the temperature field, and Q is the energy source term of the finite element unit.

[0132] In each finite element region, the Navier-Stokes equation describing fluid motion with convection effect is discretized according to the geometric boundary conditions and initial conditions, and these finite element regions are not only used to solve the temperature distribution, but also provide numerical support for the prediction of fluid flow.

[0133] Step 232: coupling the heat conduction equation and the Navier-Stokes equation in the finite element region, and introducing the real-time coupling information provided by the energy flow matrix into the multi-physics coupling model to form a multi-physics coupling model describing the interaction between temperature change and air flow;

[0134] Specifically, in the coupling construction, the energy source term Q in the heat conduction equation is obtained through the mapping of the energy flow matrix, for example, for a certain finite element unit, set Q where is the number of the finite element unit in the energy flow matrix, The element value of the first row and the first column in the energy flow matrix is obtained by multiplying the element value of the first row and the first column in the coupling matrix by the element value of the first

[0135] column in the energy flow matrix, which represents the coupling interaction strength between two atlas nodes, so that Q can reflect the dynamic energy exchange between regions.

[0136]

[0137] The modified formula can represent the multi-physical field coupling model, in which temperature changes affect local air density and flow rate, and air flow transfers heat to adjacent regions through convection. Finally, in each finite element region, the multi-physical field coupling model can describe both heat diffusion and convection heat transfer phenomena, so that the model can more realistically reflect the physical state of the microenvironment in the building.

[0138] Step 233: After constructing the coupled multi-physical field coupling model, a finite element solver is used to numerically solve the discretized multi-physical field coupling model to generate microenvironment state prediction data.

[0139] Specifically, the process of generating microenvironment state prediction data includes:

[0140] The discretized algebraic equation set of the coupled multi-physical field coupling model is input into a finite element solver, which includes but is not limited to ANSYS, COMSOL, etc.

[0141] The finite element solver uses an iterative algorithm to solve the physical quantity values of each element on the finite element grid at each time step;

[0142] Specifically, the physical quantity values at least include temperature T1(t), air flow rate v(t), and pressure p(t) of each finite element at each time point t, etc.

[0143] In a further preferred embodiment, the solved physical quantity values are stored in the form of a multi-dimensional array or a grid file, such as VTK or HDF5 format, etc., wherein each data point not only identifies its spatial coordinates x in the building, but also records the corresponding physical field value at the corresponding time.

[0144] It can be understood that through the solving process of the finite element solver, the microenvironment state prediction data in the building can be predicted and generated, fully reflecting the spatio-temporal variation characteristics of heat conduction, convection heat transfer, and air flow in the building.

[0145] Further, the constructing a mixed integer programming model based on the microenvironment state prediction data comprises the following steps:

[0146] Step 311: Using the microenvironment state prediction data generated by the finite element solver, extracting the key physical field parameters in the building, and mapping the key physical field parameters to the input parameters of the mixed integer programming model;

[0147] Specifically, the manner of extracting the key physical field parameters in the building comprises:

[0148] Extracting the key physical quantities such as temperature field T1(x, t), air flow velocity field v(x, t) and pressure field p(x, t) from the microenvironment state prediction data output by the finite element solver; wherein x represents the spatial coordinate, and t is the time node;

[0149] Then, by means of spatial averaging or regional segmentation data fusion, the distributed key physical quantities of each spatial coordinate are mapped to the index values representing each key region; for example, the temperature prediction value of a certain region is denoted as , the air flow velocity prediction value is denoted as , and the pressure prediction value is denoted as .

[0150] Step 312: Setting control variables and continuous variables for each device, and based on the set control variables, continuous variables and mapped microenvironment state indexes, constructing the objective function and constraint conditions of the mixed integer programming model for optimizing energy consumption while meeting the comfort requirements of the building environment.

[0151] Specifically, the constructing manner of the objective function and constraint conditions of the mixed integer programming model comprises:

[0152] Determining the objective of the MIP model to be the minimization of the total energy consumption of the device;

[0153] Setting the control variable to represent the running state of the graph node i at time t, wherein, is 0 or 1, respectively representing off or on, and setting the continuous variable to represent the output power or air speed adjustment value of the device; then the objective function can be expressed in the form:

[0154]

[0155] wherein, represents the energy consumption of the graph node i at the output , which is calculated based on the type of the device and the energy consumption curve; n is the number of all graph nodes; T is the total number of time points;

[0156] To ensure that the micro-environment in the building meets the predetermined comfort and safety requirements, for the region r, an environmental temperature constraint is required to keep the temperature within a safe interval, a wind speed comfort constraint is required to keep the air flow rate within a wind speed comfort interval, and a pressure comfort constraint is required to keep the device pressure within a pressure comfort interval;

[0157] Specifically, the environmental temperature constraint can be set as: Wherein, represents the temperature compensation or influence caused by the device operation control; Specifically, the compensation or influence can be preliminarily established according to the device operation principle and the building physical model to establish a physical response model, and then based on historical operation data or experimental data, a statistical regression method such as linear regression, nonlinear regression or machine learning algorithm is used to calibrate the parameters in the model to obtain; and are preset minimum temperature threshold and maximum temperature threshold, respectively;

[0158] For example, for air conditioning equipment, there is usually a certain linear or nonlinear relationship between the refrigerating capacity and the local temperature reduction, so, can be set as ; wherein R is a set of graph nodes affecting the temperature change in the region r, is a parameter to be calibrated.

[0159] Similarly, the wind speed comfort constraint and the pressure comfort constraint can be represented as:

[0160] and

[0161] Wherein, are the minimum speed threshold and the maximum speed threshold, respectively, and are the minimum pressure threshold and the maximum pressure threshold, respectively, and respectively represent the speed compensation or influence caused by the device operation control and the pressure compensation or influence caused by the device operation control, and Both can be obtained by fitting the compensation or influence of the temperature.

[0162] Further, the way to solve the mixed integer programming model to obtain the scheduling scheme of the energy-consuming device is:

[0163] The objective function and all constraints of the mixed integer programming model are input into an integer programming model solver, which solves the mixed integer programming model for any discrete time t, and generates the approximate optimal solution of the control variables and continuous variables for each device; thereby outputting a scheduling scheme that determines the operating state and output power or wind speed adjustment value of each device.

[0164] It can be understood that the scheduling scheme explicitly indicates the operating state and output parameter of each device at each time step, so that the temperature, flow rate, pressure, and other physical field indicators of each region in the building are maintained within the design requirement range, while the energy consumption is minimized.

[0165] Further, the manner in which the edge execution device is driven based on the scheduling scheme is as follows:

[0166] Based on the scheduling scheme, the edge control system takes the approximate optimal solution of the control variables and continuous variables of each graph node in the scheduling scheme as the direct operation basis, and automatically adjusts the operating state and output power or wind speed adjustment value of the device by the local controller of the device to the approximate optimal solution of the control variables and continuous variables.

[0167] In addition, the parts of the above technical solutions in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive repetition.

[0168] The specific embodiments described above further illustrate the objects, technical solutions, and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0169] The above preset parameters or preset thresholds are set by a person skilled in the art according to actual conditions or obtained by a large amount of data simulation.

[0170] The above embodiments are only used to illustrate the technical method of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. A building-wide energy consumption optimization management system, characterized in that: It includes a multimodal data collection module, an environmental state prediction module, a scheduling solution module, and a control instruction optimization module; each module is electrically connected; The multimodal data collection module collects building environmental parameters and equipment operation data through a distributed sensor network, generates a multimodal perception matrix after spatiotemporal alignment and anomaly detection, and sends the multimodal perception matrix to the environmental state prediction module; An environmental state prediction module builds a dynamic knowledge graph based on the multimodal perception matrix, uses a graph attention network to quantify the energy coupling strength between devices in the dynamic knowledge graph, outputs an energy flow matrix of the device topology, and then combines the energy flow matrix with the multi-physics field coupling model to generate microenvironmental state prediction data through a finite element solver. And send the micro-environment state prediction data to the scheduling solution module; a scheduling solution module, which constructs a mixed integer programming model based on the microenvironment state prediction data, solves the mixed integer programming model to obtain a scheduling solution for the energy-consuming equipment, and sends the scheduling solution to the control instruction optimization module; The control instruction optimization module drives the edge execution device to run based on the scheduling scheme.

2. The system for optimizing energy consumption of all equipment in a building according to claim 1, characterized in that: The method of collecting building environment parameters and equipment operation data through a distributed sensor network includes the following steps: Step 111: Based on the building design drawings and BIM data, multiple locations within the building are selected as key locations, and environmental sensors and status sensors are installed at the key locations to form a distributed sensor network. Step 112: The parameter values ​​of the corresponding environmental parameters and the parameter values ​​of the operating status parameters are collected in real time through the environmental sensors and the status sensors to form an original data set.

3. The system for optimizing energy consumption of all equipment in a building according to claim 2, characterized in that: The method of generating a multimodal perception matrix after the spatiotemporal alignment and anomaly detection processing is as follows: Step 121: The original data set is transmitted to the edge gateway through a network connected to the corresponding physical sensor and the edge gateway; the edge node verifies the consistency of the timestamps of all data using the NTP protocol or the PTP protocol, and then uses a filtering algorithm to denoise the data to obtain a smoothed data set; Step 122: Based on the smoothed data set, a multimodal perception matrix is ​​formed through spatiotemporal alignment and multidimensional matrix construction.

4. The system for optimizing energy consumption of all equipment in a building according to claim 3, characterized in that: The multimodal perception matrix is ​​constructed as follows: After obtaining the smoothed data set, spatial and temporal alignment is performed on parameter values ​​collected from physical sensors of different locations and types according to a unified timestamp; For each key location, the types of environmental parameters and operating status parameters are integrated into a multidimensional data matrix, where each row of the multidimensional data matrix represents a physical sensor and each column corresponds to an environmental parameter or an operating status parameter; The three-dimensional matrix composed of the multidimensional data matrices of all key positions is the multimodal perception matrix.

5. The system for optimizing energy consumption of all equipment in a building according to claim 4, characterized in that: The method of constructing a dynamic knowledge graph based on the multimodal perception matrix is: Step 211: By analyzing and extracting features from the building environment parameters and equipment operation data collected in the multimodal perception matrix, initial graph nodes reflecting key locations in the building are constructed; Step 212: Based on the spatiotemporal correlation of the data in the multimodal sensing matrix and the physical topology of the building, construct relationship edges between nodes and assign a quantitative weight to each edge to reflect the energy coupling or operating status correlation strength between the nodes; Step 213: Construct a dynamic knowledge graph based on the extracted graph nodes and graph edges.

6. The system for optimizing energy consumption of all equipment in a building according to claim 5, characterized in that: The graph attention network is used to quantify the energy coupling strength between devices in the dynamic knowledge graph and output the energy flow matrix of the device topology as follows: Step 221: Using the graph attention network, the static knowledge graph constructed initially is dynamically updated. The node representations and edge weights are adjusted according to the multimodal perception matrix collected in real time, thereby generating a dynamic knowledge graph that reflects the real-time state changes within the building. Step 222: Calculate the neighborhood attention coefficient for the updated node attributes of the dynamic knowledge graph using the graph attention network, and aggregate the neighborhood information to update the node hidden layer representation; Step 223: Dynamically adjust the edge weights based on the attention coefficients between the graph nodes and the initial edge weights to construct an energy flow matrix that reflects the real-time coupling strength; Step 224: Dynamically update the attributes of the graph nodes and the edge weights of the edges in the dynamic knowledge graph using the updated representation of the updated graph nodes and the updated edge weights between each pair of graph nodes.

7. The system for optimizing energy consumption of all equipment in a building according to claim 6, characterized in that: Combining the energy flow matrix with the multi-physics field coupling model and generating microenvironment state prediction data through finite element solver prediction includes the following steps: Step 231: Based on the building's geometric information and material properties, perform finite element meshing on the building's interior, discretize the continuous space into multiple finite element regions, and use the heat conduction equation and Navier-Stokes equations to construct a discretized description of the heat conduction and fluid flow physical fields. Step 232: coupling the heat conduction equation and the Navier-Stokes equation within the finite element region, and introducing real-time coupling information provided by the energy flow matrix into the multi-physics coupling model to form a multi-physics coupling model that describes the interaction between temperature change and air flow; Step 233: After constructing the coupled multi-physics coupling model, a finite element solver is used to numerically solve the discretized multi-physics coupling model to generate microenvironment state prediction data.

8. The system for optimizing energy consumption of all equipment in a building according to claim 7, characterized in that: The constructing of the mixed integer programming model based on the microenvironment state prediction data comprises the following steps: Step 311: using the microenvironment state prediction data generated by the finite element solver, extracting key physical field parameters in the building, and mapping the key physical field parameters into input parameters of the mixed integer programming model; Step 312: Set control variables and continuous variables for each device, and based on the set control variables, continuous variables, and mapped microenvironment status indicators, construct the objective function and constraints of a mixed integer programming model that optimizes energy consumption while meeting the requirements of building interior environmental comfort.

9. The system for optimizing energy consumption of all equipment in a building according to claim 8, characterized in that: The objective function and constraint conditions of the mixed integer programming model are constructed in the following ways: The goal of the MIP model is to minimize the overall energy consumption of the equipment; For any area, an ambient temperature constraint is constructed, which requires the temperature to be kept within the safe range, a wind speed comfort constraint is constructed, which requires the air flow rate to be kept within the wind speed comfort range, and a pressure comfort constraint is constructed, which requires the equipment pressure to be kept within the pressure comfort range.

10. The system for optimizing energy consumption of all equipment in a building according to claim 9, characterized in that: The method of solving the mixed integer programming model to obtain the scheduling plan for energy-consuming equipment is: The objective function and all constraints of the mixed integer programming model are input into an integer programming model solver. The integer programming model solver solves the mixed integer programming model for any discrete time t, generates an approximate optimal solution for the control variables and continuous variables for each device, and outputs a scheduling plan to determine the operating status and output power or wind speed adjustment value of each device.

11. A building-wide energy consumption optimization management system according to claim 10, characterized in that: The method of driving the edge execution device to run based on the scheduling scheme is: Based on the scheduling scheme, the edge control system uses the approximate optimal solutions of the control variables and continuous variables of each graph node in the scheduling scheme as the direct operation basis, and the local controller of the device automatically adjusts its operating status and output power or wind speed adjustment value to the approximate optimal solutions of the corresponding control variables and continuous variables.

Images