A building energy management system based on intelligent gateway edge computing

Abstract

The present invention belongs to the technical field of energy management, and discloses a building energy management system based on intelligent gateway edge computing; the method comprises: collecting energy usage data and environmental data inside a building; transmitting the collected energy usage data and environmental data to an edge computing node through an intelligent gateway; in the edge computing node, preprocessing the collected energy usage data and environmental data to obtain preprocessed data; based on the preprocessed data, constructing an energy usage pattern diagram, optimizing the structure of the energy usage pattern diagram, and identifying key nodes and key associations; according to the identified key nodes and key associations, formulating an energy optimization strategy using an optimization algorithm, sending the energy optimization strategy to the intelligent gateway, and controlling the energy usage equipment inside the building through the intelligent gateway, so as to realize efficient utilization of energy, improve the intelligence and automation level of building energy management, and reduce management difficulty and labor cost.

Description

Technical Field

[0001] The present invention relates to the field of energy management technology, and more specifically, to a building energy management system based on intelligent gateway edge computing. Background Art

[0002] The patent application with application publication number CN118192325A discloses an intelligent building energy management system and method, including a client and a control end, the client and the control end are communicatively connected, the control end includes a sensor network, a data analysis module, a database module, and a control module, the sensor network is communicatively connected to the control module through the data analysis module, and the control module is also communicatively connected to the database module; by utilizing technologies such as sensors, data analysis and control modules, real-time energy monitoring and energy adjustment of various systems in the building are achieved, thereby reducing energy consumption and improving building energy utilization; it has the characteristics of high efficiency and reliability, and can predict and adjust energy demand in advance based on big data analysis and machine learning algorithms to further reduce energy consumption; at the same time, it can also ensure the comfort of the building environment and provide a good use environment.

[0003] However, there are still many problems in the existing building energy management system, which seriously restricts the effect of building energy conservation and emission reduction. First, traditional systems often adopt simple fixed strategies or rules, which cannot adapt to the complex and changeable energy consumption situation inside the building, resulting in widespread energy waste. Second, the existing system lacks in-depth analysis of the mutual influence between different areas and equipment inside the building, making it difficult to achieve global collaborative optimization. Third, most systems cannot effectively process and utilize multi-source heterogeneous data, such as energy consumption data and environmental data, resulting in insufficient decision-making basis and limited optimization effect. In addition, the traditional centralized architecture faces problems such as poor real-time performance and high network bandwidth pressure, which is particularly evident in large-scale building complexes. For example, when a rapid response to emergencies (such as a sudden increase in traffic) is required, the system response is often delayed. Finally, existing systems usually find it difficult to take into account multiple goals at the same time, often losing sight of one thing while focusing on another, making it difficult to achieve true comprehensive optimization. These problems lead to inefficient building energy management, which not only increases operating costs, but also hinders the realization of energy conservation and emission reduction goals.

[0004] In view of this, the present invention proposes a building energy management system based on intelligent gateway edge computing to solve the above problems. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solutions: A building energy management system based on intelligent gateway edge computing, comprising: a data acquisition module for collecting energy usage data and environmental data inside the building;

[0006] A data transmission module is used to transmit the collected energy usage data and environmental data to the edge computing node through the intelligent gateway; in the edge computing node, the collected energy usage data and environmental data are preprocessed to obtain preprocessed data;

[0007] The comprehensive identification module constructs an energy usage pattern diagram based on the preprocessed data, optimizes the structure of the energy usage pattern diagram, and identifies key nodes and key relationships;

[0008] The strategy formulation transmission module formulates energy optimization strategies based on the identified key nodes and key relationships using optimization algorithms, sends the energy optimization strategies to the smart gateway, and controls the energy-using devices inside the building through the smart gateway; each module is connected via wired and / or wireless means.

[0009] Furthermore, the energy usage data includes energy type, energy usage, energy usage power / flow, energy usage equipment / area ID and energy usage timestamp; the environmental data includes temperature, humidity, light intensity, carbon dioxide concentration, presence of personnel, area / room ID and environmental data collection timestamp.

[0010] Furthermore, the method of preprocessing the collected energy usage data and environmental data includes: creating an M×N two-dimensional array or matrix to represent a grid point matrix, and initializing the values ​​of all grid points in the grid point matrix to be empty or 0;

[0011] Extract key features from the collected energy usage data and environmental data; the key features include energy type, regional location and time; represent each data point as a vector composed of key features, recorded as a feature vector;

[0012] Map data points of the same energy type and regional location to adjacent grid points, map data points of different energy types or regional locations to areas consisting of non-adjacent grid points, and map data points at different times to different grid points; traverse all data points, and record the feature vector and original data of the data point in the grid lattice; and superimpose regional boundary information and time axis information on the grid lattice; cluster the data points in the grid lattice; obtain n data clusters; for each data cluster, extract the feature vector corresponding to the data point, map the feature vector to a low-dimensional space, that is, perform dimensionality reduction; and obtain preprocessed data.

[0013] Furthermore, the method of clustering data points in the grid lattice includes:

[0014] Randomly select K grid points in the grid lattice as the initial positions of the gas molecules, and record the coordinates of the grid point where each gas molecule is currently located;

[0015] Define the neighborhood radius r and calculate the diffusion probability of gas molecules within the neighborhood radius r ;in, , and All are control parameters; is the total mass of all gas molecules within the neighborhood radius r, is the average distance of a gas molecule from other gas molecules within the neighborhood radius r;

[0016] A diffusion threshold is preset. For each gas molecule, if the corresponding calculated diffusion probability is greater than the diffusion threshold, the corresponding gas molecule diffuses to the adjacent grid point; if the corresponding calculated diffusion probability is less than or equal to the diffusion threshold, the corresponding gas molecule keeps the original position unchanged; the coordinates of the new position of each gas molecule are updated;

[0017] Traverse each grid point in the grid matrix. If the grid point is the location of the data point, calculate the grid point The degree of aggregation of nearby gas molecules ; Take this as the degree of clustering of the data points;

[0018] ;in, are nearby gas molecules and lattice points The distance between is the base aggregation radius; is a positive parameter; is a positive exponential parameter; Grid point The quality of the corresponding data points; is the minimum value of the quality of all data points, is the maximum value of the mass of all data points;

[0019] Define a clustering threshold, traverse all data points, if the degree of clustering of data points is greater than the clustering threshold, they are classified into a new cluster; if the degree of clustering of data points is less than or equal to the clustering threshold, they are temporarily marked as noise points; for noise points, according to their distance from the existing clusters, they are classified into the nearest cluster to obtain n data clusters.

[0020] Furthermore, the diffusion method includes:

[0021] Define the mass of each gas molecule; for gas molecule j, calculate the gravitational force between it and other gas molecules i ;in, is the gravitational constant, is the mass of gas molecule i, is the mass of gas molecule j, is the Euclidean distance between gas molecule j and other gas molecules i, To adjust the parameters;

[0022] For gas molecule j, calculate the repulsive force between it and other gas molecules i ;in, is the repulsive force constant, is a characteristic length, is a positive number, and To adjust the power index;

[0023] Based on the calculated attraction and repulsion of gas molecules, define the comprehensive influence field ;

[0024] ;in, is a random number between 0 and 1. is the random perturbation intensity constant;

[0025] For the current position of the gas molecule, calculate the value of the comprehensive influence field of its adjacent positions, and take the adjacent position corresponding to the value of the largest comprehensive influence field as the new position. If there are several adjacent positions with the same maximum comprehensive influence field value, randomly select one adjacent position as the new position; that is, the diffusion is completed.

[0026] Furthermore, the method of mapping the feature vector to the low-dimensional space includes:

[0027] For every two data points in the same data cluster, the corresponding eigenvectors and ; Calculate the comprehensive similarity between them ; The comprehensive similarity Construct the feature vector similarity matrix P as elements;

[0028] ;in, is the feature vector The k nearest neighbor set of ; is the feature vector The k nearest neighbor set of ; is the exponential parameter, and is the intra-cluster distance adjustment parameter; is the feature vector The distance from the corresponding data point to the cluster center of the data cluster to which it belongs; is the feature vector The distance from the corresponding data point to the cluster center of the data cluster to which it belongs;

[0029] Initialize N1 low-dimensional data points randomly in the preset low-dimensional space, and for every two low-dimensional data points X and Y in the low-dimensional space, calculate the comprehensive similarity p_XY between them; construct the low-dimensional space similarity matrix Q using the comprehensive similarity p_XY as an element;

[0030] The KL divergence between the low-dimensional space similarity matrix Q and the eigenvector similarity matrix P is calculated, and the position of the low-dimensional data points is iteratively optimized using the gradient descent method to minimize the KL divergence; after the iterative optimization converges, the final low-dimensional data points are obtained, corresponding to the data points in the data cluster, and the dimensionality reduction is completed.

[0031] Furthermore, the method of constructing the energy usage pattern diagram includes:

[0032] For each data cluster, a node is created and the preprocessed data in the data cluster is mapped to the attributes of the node. , calculate its relationship with other nodes The strength of the association between ;

[0033] ;

[0034] in, and Node and The corresponding time series data; and Node and The corresponding spatial coordinates; and Node and The corresponding activity time set; For Node Any pre-processed data in For Node Any pre-processed data in ; , , and are the weight coefficients of the corresponding items, and satisfy , , and The sum is 1;

[0035] A similarity threshold is set. No edge is connected between two nodes whose association strength is less than the similarity threshold, and an edge is connected between two nodes whose association strength is greater than or equal to the similarity threshold, and the weight of the edge is the corresponding association strength; then weighted edges are obtained; and an energy usage pattern graph G'=(V, E) is constructed, where V is a set of nodes and E is a set of weighted edges.

[0036] Furthermore, the method of optimizing the structure of the energy usage pattern diagram includes:

[0037] Assign a unique initial label to each node and record the initial labels of all nodes; set the maximum number of iterations mx and the convergence threshold co, initialize a node label list, and record the weight of each label, defining the weight of the label as the sum of the degrees of the corresponding nodes;

[0038] Copy the current node label list to get a new node label list, which is recorded as the reference node label list; traverse each node , get the node All connected nodes and its labels, calculate the comprehensive weight of each label, that is, the label corresponds to the node The sum of the degrees of the nodes The sum of the degrees of ; find the label ma with the largest comprehensive weight; if the comprehensive weight of ma is greater than the node The weight of the current label; then compare the node in the node label list The current label is updated to ma; otherwise the node The current label remains unchanged; calculate the ratio of nodes whose reference node label list is different from the node label list. If the ratio is less than the convergence threshold co, stop; otherwise, copy the reference node label list back to the node label list and continue the next round of iteration until the maximum number of iterations mx is reached, and then terminate the iteration;

[0039] Nodes with the same label are divided into a subgraph, and these subgraphs constitute the structure of the optimized energy usage pattern graph.

[0040] Furthermore, the method of identifying key nodes and key associations includes:

[0041] Calculate the degree centrality and betweenness centrality of each node; preset corresponding thresholds, and identify nodes whose degree centrality and betweenness centrality are higher than the corresponding thresholds as key nodes;

[0042] For each edge in the structure of the optimized energy usage pattern graph, the number of times it is between the shortest paths of other node pairs is calculated, and an edge betweenness threshold is set. The edges with a number higher than the edge betweenness threshold are identified as key associations.

[0043] Furthermore, the method of formulating the energy optimization strategy includes:

[0044] Each key node corresponds to an energy-using device, and the comprehensive attributes of the node are analyzed. The comprehensive attributes include the energy type, time series data, spatial coordinates, and activity time set;

[0045] According to the comprehensive attributes of each key node, its optimizable parameters are determined as decision variables. The value range of the decision variable is the reasonable variation range of the parameter.

[0046] Analyze the association attributes of key associations, including association strength and betweenness centrality; preset a constraint threshold, and subject two nodes corresponding to key associations whose association strength is greater than the constraint threshold to variable constraints, where the variable constraint is the absolute value of the sum of the decision variables corresponding to the two nodes, which is less than the preset variable threshold;

[0047] A path threshold is preset, and two nodes corresponding to the key association whose betweenness centrality is greater than the path threshold are subjected to path constraints. The path constraint is that the decision variables of the two nodes corresponding to the key association satisfy the temporal or quantitative relationship.

[0048] According to the number of decision variables n1, an initial group of explosive bodies consisting of N2 explosive bodies is generated. Each explosive body corresponds to an n1-dimensional vector, representing a candidate solution. The position coordinates of the explosive body are randomly initialized within the value range of the decision variables.

[0049] Define the objective function ;in, , and is the weight coefficient of the objective function; is a vector of decision variables, For the The vector of decision variables consisting of energy-using equipment The power under For the The operating time of each energy-using device, For the Unit rates for each energy-using device; is the vector of decision variables Next, The total consumption of energy-using equipment, For the CO2 emission coefficient of each energy-using equipment;

[0050] Substitute the candidate solution corresponding to each explosive body into the objective function, and calculate the function value, which is recorded as the explosion value; in the current group of explosive bodies, select the explosive body with the smallest explosion value as the optimal explosive body; the other explosive bodies are non-optimal explosive bodies; for each non-optimal explosive body, update its position, and the position update method is: calculate the distance between the explosive body and the optimal explosive body ; Generate a random number between 0 and 1 ,like , then the explosive body moves towards the optimal explosive body, and the step length of the movement is ; Otherwise, the explosion body moves away from the optimal explosion body, and the step length is ;

[0051] During the movement, check whether the new position of the explosion body exceeds the value range of the decision variable. If it exceeds, the position of the corresponding dimension is processed at the boundary.

[0052] If the preset maximum number of iterations is reached, the candidate solution corresponding to the current optimal explosive body is output as the final energy optimization strategy; otherwise, the position is continued to be updated until the preset maximum number of iterations is reached.

[0053] Technical effects and advantages of a building energy management system based on intelligent gateway edge computing in the present invention:

[0054] The present invention can achieve refined management and global optimization of building energy use, greatly improve energy utilization efficiency, and effectively reduce energy consumption and operating costs. The system deeply explores the complex energy use patterns and correlations within the building to formulate a more comprehensive and coordinated optimization strategy, avoiding the one-sidedness and limitations of traditional methods; at the same time, it can balance multiple goals such as energy consumption, cost and environmental protection, while improving economic benefits, it also reduces the carbon emissions of buildings and promotes sustainable development; the real-time and adaptability of the system are significantly enhanced, and it can quickly respond to changes in the internal environment of the building and energy demand, dynamically adjust the optimization strategy, and ensure that it always maintains the best operating state; in addition, the system also improves the intelligence and automation level of building energy management, reduces human intervention, and reduces management difficulty and labor costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A schematic diagram of a building energy management system based on intelligent gateway edge computing according to the present invention;

[0056] Figure 2 This is a schematic diagram of a building energy management method based on intelligent gateway edge computing of the present invention. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] Example 1

[0059] See also Figure 1 As shown, the building energy management system based on intelligent gateway edge computing described in this embodiment includes:

[0060] Data collection module, used to collect energy usage data and environmental data inside the building;

[0061] A data transmission module is used to transmit the collected energy usage data and environmental data to the edge computing node through the intelligent gateway; in the edge computing node, the collected energy usage data and environmental data are preprocessed to obtain preprocessed data;

[0062] The comprehensive identification module constructs an energy usage pattern diagram based on the preprocessed data, optimizes the structure of the energy usage pattern diagram, and identifies key nodes and key relationships;

[0063] The strategy formulation and transmission module formulates energy optimization strategies based on the identified key nodes and key relationships using optimization algorithms, sends the energy optimization strategies to the smart gateway, and controls the energy-using equipment inside the building through the smart gateway to achieve efficient energy utilization; each module is connected through wired and / or wireless means to achieve data transmission between modules.

[0064] The energy usage data includes energy type, energy usage, energy usage power / flow, energy usage equipment / area ID and energy usage timestamp; energy types include electricity, water, natural gas, fuel oil, etc.; energy usage, such as electricity, water, gas, etc.; energy usage power / flow such as electric power, water flow, etc.; environmental data includes temperature, humidity, light intensity, carbon dioxide concentration, presence of personnel, area / room ID and environmental data collection timestamp.

[0065] These data are collected through various sensors and metering devices and associated with corresponding IDs and timestamps to form a spatiotemporal data set; energy usage data reflects the energy consumption of different areas and equipment in the building at different times; environmental data reflects the environmental status of the corresponding area and can be used to analyze environmental factors that affect energy use.

[0066] In the process of collecting energy usage data and environmental data inside the building, a unique ID is assigned to each energy-using device or area, and the ID is recorded as a dimension of the energy usage data and environmental data.

[0067] The methods for preprocessing the collected energy usage data and environmental data include:

[0068] Create an M×N two-dimensional array or matrix to represent the grid point matrix, and initialize the values ​​of all grid points in the grid point matrix to be empty or 0; extract key features from the collected energy usage data and environmental data; the key features include energy type (electricity, water, gas, etc.), regional location and time; represent each data point as a vector composed of key features, recorded as a feature vector; map data points of the same energy type and regional location to adjacent grid points, map data points of different energy types or regional locations to areas composed of non-adjacent grid points, and map data points at different times to different grid points.

[0069] Traverse all data points, and record the feature vector and original data of the data point in the grid matrix; and superimpose the regional boundary information and time axis information on the grid matrix; specifically, traverse all grid points, and mark the grid points belonging to the same area with the same color or pattern according to the ID corresponding to each grid point, draw regional boundary lines at the edge of the grid matrix, separate the grid points of different areas, and use different colors or line types to distinguish different areas for easy identification; draw the time axis on one side (above or left) of the grid matrix, and the scale of the time axis is set according to the time range and granularity of the data, which can be hours, days, months, etc.; traverse all grid points, and map each grid point to the corresponding time axis scale position according to the timestamp corresponding to the grid point; use different colors or marks to distinguish different time points or time periods for easy analysis of time patterns.

[0070] Simulate the diffusion and aggregation behavior of gas molecules in the grid lattice, and cluster the data points according to the degree of aggregation of gas molecules; randomly select K grid points in the grid lattice as the initial positions of gas molecules, and record the coordinates of the grid point where each gas molecule is currently located;

[0071] Define the neighborhood radius r and calculate the diffusion probability of gas molecules within the neighborhood radius r ;in, , and are control parameters, The baseline value that controls the diffusion probability is usually in the range of (0, 1]; Controlling the diffusion probability as the number of gas molecules within the neighborhood radius r The rate of change, The larger it is, the faster the diffusion probability decreases; Control diffusion probability The rate of change, The larger it is, the faster the diffusion probability decreases; is the total mass of all gas molecules within the neighborhood radius r, is the average distance (Euclidean distance) of a gas molecule to other gas molecules within the neighborhood radius r.

[0072] A diffusion threshold is preset. For each gas molecule, if the corresponding calculated diffusion probability is greater than the diffusion threshold, the corresponding gas molecule diffuses to the adjacent grid point; if the corresponding calculated diffusion probability is less than or equal to the diffusion threshold, the corresponding gas molecule maintains its original position unchanged; and the coordinates of the new position of each gas molecule are updated.

[0073] The ways of spreading include:

[0074] Define the mass of each gas molecule (data point). The larger the mass, the stronger the gravitational force. It should be noted that the mass value of a gas molecule is a weighted combination of multiple factors, such as data value size, importance weight, cluster size, time and space position, etc. By adjusting the weights of different factors, the calculation method of the mass value can be flexibly controlled. For gas molecule j, calculate the gravitational force between it and other gas molecules i. ;in, is the gravitational constant, is the mass of gas molecule i, is the mass of gas molecule j, is the Euclidean distance between gas molecule j and other gas molecules i, By adjusting the parameters and controlling the decay rate, we can better simulate the rapid decay of gravity over long distances in actual situations.

[0075] For gas molecule j, calculate the repulsive force between it and other gas molecules i ;in, is the repulsive force constant, is a characteristic length that controls the range of the repulsive force. is a small positive number used to avoid the denominator being zero. and To adjust the power exponent, the nonlinear degree of the repulsive force changing with distance is controlled and the attenuation rate of the repulsive force changing with distance is further adjusted.

[0076] Based on the calculated attraction and repulsion of gas molecules, define the comprehensive influence field ;

[0077] ;in is a random number between 0 and 1, simulating the random motion of gas molecules. is the random perturbation intensity constant;

[0078] For the current position of the gas molecule, calculate the value of the comprehensive influence field of its adjacent positions, and take the adjacent position corresponding to the value of the largest comprehensive influence field as the new position. If there are several adjacent positions with the same maximum comprehensive influence field value, randomly select one adjacent position as the new position; that is, the diffusion is completed.

[0079] Traverse each grid point in the grid matrix. If the grid point is the location of the data point, calculate the grid point The degree of aggregation of nearby gas molecules ; Take this as the degree of clustering of the data points;

[0080] ;in, are nearby gas molecules and lattice points The distance between (Euclidean distance), The base aggregation radius is used to determine the basic scale of the aggregation range; is a positive parameter that controls the degree of influence of quality or importance on the cluster radius; It is a positive exponential parameter that controls the nonlinearity of the change of the aggregation radius with the quality or importance; Grid point The quality of the corresponding data points; is the minimum value of the quality of all data points, is the maximum value of the mass of all data points.

[0081] Define a clustering threshold, traverse all data points, if the degree of clustering of data points is greater than the clustering threshold, classify them into a new cluster; if the degree of clustering of data points is less than or equal to the clustering threshold, temporarily mark them as noise points; for noise points, classify them into the nearest cluster according to their distance from the existing clusters; get n data clusters; n is an integer greater than 1; each corresponding data cluster corresponds to an energy-using device or area.

[0082] For each data cluster, extract the eigenvectors corresponding to the data points, map the eigenvectors to a low-dimensional space, and perform dimensionality reduction to reduce the dimension of the data; obtain the preprocessed data (vector); for every two data points in the same data cluster, the eigenvectors corresponding to the data points are and ; Calculate the comprehensive similarity between them ; The comprehensive similarity Construct the feature vector similarity matrix P as elements;

[0083] ;in, is the feature vector The k nearest neighbor set of ; is the feature vector The k nearest neighbor set of ; is an exponential parameter that can adjust the impact of the number of shared nearest neighbors on similarity. and Adjust the parameters for intra-cluster distance to flexibly control the influence of intra-cluster position on similarity; is the feature vector The distance from the corresponding data point to the cluster center of the data cluster to which it belongs; is the feature vector The distance from the corresponding data point to the cluster center of the data cluster to which it belongs.

[0084] Randomly initialize N1 low-dimensional data points in the preset low-dimensional space. For every two low-dimensional data points X and Y in the low-dimensional space, calculate the comprehensive similarity p_XY between them (the calculation method is the same as The same); the comprehensive similarity p_XY is used as an element to construct a low-dimensional space similarity matrix Q; the KL divergence between the low-dimensional space similarity matrix Q and the eigenvector similarity matrix P is calculated, and the gradient descent method is used to iteratively optimize the position of the low-dimensional data points to minimize the KL divergence, so that the similarity matrix Q in the low-dimensional space is as close as possible to the similarity matrix P in the high-dimensional space; after the iterative optimization converges, the final low-dimensional data points are obtained, which correspond to the data points in the data cluster, and the dimensionality reduction is completed.

[0085] By optimizing the position of data points in low-dimensional space, the similarity in low-dimensional space is made to preserve the local and global structures in high-dimensional space as much as possible.

[0086] Ways to construct a map of energy usage patterns include:

[0087] For each data cluster, a node is created and the preprocessed data in the data cluster is mapped to the attributes of the node (several). , calculate its relationship with other nodes The strength of the association between ;

[0088] ;

[0089] in, and Node and Corresponding time series data (the time series corresponding to the data points); and Node and The corresponding spatial coordinates (the centroid of the spatial locations of all data points in the data cluster or the central geographic coordinates of the energy-using device / area); and Node and Corresponding activity time set (analyze the time series data of the data cluster and extract the time period with non-zero values, which is the activity time set corresponding to the data cluster); For Node Any pre-processed data in For Node Any pre-processed data in ; , , and are the weight coefficients of the corresponding items, and satisfy , , and The sum is 1.

[0090] A similarity threshold is set. No edge is connected between two nodes whose association strength is less than the similarity threshold, and an edge is connected between two nodes whose association strength is greater than or equal to the similarity threshold, and the weight of the edge is the corresponding association strength; then weighted edges are obtained; and an energy usage pattern graph G'=(V, E) is constructed, where V is a set of nodes and E is a set of weighted edges.

[0091] Ways to optimize the structure of the energy usage pattern map include:

[0092] Assign a unique initial label to each node, such as a node ID, and record the initial labels of all nodes; set the maximum number of iterations mx and the convergence threshold co, initialize a node label list, and record the weight of each label, defining the weight of the label as the sum of the degrees of the corresponding nodes (the number of edges directly connected to the node); copy the current node label list to obtain a new node label list, which is recorded as the control node label list.

[0093] Traverse each node , get the node All connected nodes and its labels, calculate the comprehensive weight of each label, that is, the label corresponds to the node The sum of the degrees of the nodes The sum of the degrees of ; find the label ma with the largest comprehensive weight; if the comprehensive weight of ma is greater than the node The weight of the current label; then compare the node in the node label list The current label is updated to ma; otherwise the node The current label remains unchanged.

[0094] Calculate the ratio of nodes whose control node label list is different from the node label list. If the ratio is less than the convergence threshold co, stop. Otherwise, copy the control node label list back to the node label list and continue the next round of iteration until the maximum number of iterations mx is reached, then terminate the iteration. Divide the nodes with the same label into a subgraph, and these subgraphs constitute the structure of the optimized energy usage pattern graph.

[0095] Ways to identify key nodes and key relationships include:

[0096] Calculate the degree centrality and betweenness centrality of each node; preset corresponding thresholds, and identify nodes whose degree centrality and betweenness centrality are higher than the corresponding thresholds as key nodes; for each edge in the structure of the optimized energy usage pattern graph, calculate the number of shortest paths between other node pairs, set the edge betweenness threshold, and identify the edges whose number is higher than the edge betweenness threshold as key associations; these key associations may be bridging edges or edge edges in the network.

[0097] At the same time, based on the centrality index of the nodes and the betweenness centrality of the edges, the nodes and associations that play a key role in the optimized graph structure can be identified; these key nodes and associations are crucial for understanding energy usage patterns and formulating optimization strategies.

[0098] Ways to develop an energy optimization strategy include:

[0099] Each key node corresponds to an energy-using device. The comprehensive attributes of the node are analyzed, including the energy type, time series data, spatial coordinates and activity time set. According to the comprehensive attributes of each key node, its optimizable parameters are determined as decision variables. For example, for power type nodes, the decision variables can be power supply time, power size, etc.; for heating type nodes, the decision variables can be heating time, temperature setting value, etc.; for water supply and drainage type nodes, the decision variables can be water supply / drainage time, flow size, etc. The time series data reflects the energy usage of the node at different times; peak and trough periods are extracted from it, and the decision variables are set as the operation / optimization strategies in different periods; nodes with adjacent spatial positions may have mutual influence and constraints; the decision variables are set so that the sum of the decision variables of the nodes in the same area does not exceed a certain threshold; the activity time set reflects the time period of non-zero energy usage of the node; the constraint condition of the decision variable is set to be 0 (energy-saving state) during the inactive time; the value range of the decision variable is the reasonable variation range of the parameter.

[0100] Key associations represent important associations and constraints between nodes (devices / areas). The association properties of key associations are analyzed. The association properties include association strength and betweenness centrality (the number of shortest paths between other node pairs).

[0101] A constraint threshold is preset, and variable constraints are applied to the two nodes corresponding to the key association whose association strength is greater than the constraint threshold. The variable constraint is the absolute value of the sum of the decision variables corresponding to the two nodes, which is less than the preset variable threshold. A path threshold is preset, and path constraints are applied to the two nodes corresponding to the key association whose betweenness centrality is greater than the path threshold. The path constraint is that the decision variables of the two nodes corresponding to the key association satisfy a temporal or quantitative relationship.

[0102] Specifically, for some key associations, the operation of two nodes may need to follow a specific timing order, that is, the operation of one node must be before or after another node; for example, the water supply equipment must be operated before the water use equipment to ensure water supply, the air conditioning refrigeration unit must be operated after the blower to ensure the cooling effect, and the boiler must be operated before the heating system to ensure heating.

[0103] In this case, timing constraints are imposed on the decision variables (such as startup time) of the two nodes, that is, if node A1 must run before node B1, then t_A1+δ_A1≤t_B1; where t_A1 and t_B1 are the startup times of nodes A1 and B1 respectively, and δ_A1 is the running time of node A1.

[0104] For other key associations, the operation of the two nodes satisfies a certain quantitative relationship. For example, the heating supply of the heating system needs to match the heat dissipation of the room, the fresh air volume of the fresh air system needs to match the ventilation needs of indoor personnel, and the output power of the generator needs to match the total load of electrical equipment.

[0105] In this case, quantitative constraints are imposed on the decision variables of the two nodes (such as power, flow, etc.), that is, if the output of node A1 needs to match the demand of node B1, then Q_A1=k2×Q_B1; where Q_A1 and Q_B1 are the decision variables corresponding to nodes A1 and B1 respectively, and k2 is the matching coefficient.

[0106] These timing or quantitative constraints are directly expressed in the constraint conditions, or indirectly constrained by penalty functions. By imposing these constraints, we can ensure that the nodes corresponding to the key associations maintain good coordination and avoid conflicts or waste, thereby improving the operating efficiency and energy utilization of the overall system.

[0107] According to the number of decision variables n1, an initial group of explosive bodies consisting of N2 explosive bodies is generated. Each explosive body corresponds to an n1-dimensional vector, representing a candidate solution. The position coordinates of the explosive body are randomly initialized within the value range of the decision variables.

[0108] Define the objective function ;in, , and is the weight coefficient of the objective function, which is used to balance the importance of each sub-goal; is a vector of decision variables, containing all the parameters that need to be optimized. For the The vector of decision variables consisting of energy-using equipment The power under For the The operating time of each energy-using device, For the Unit rates for individual energy-using equipment (such as electricity and water charges); is the vector of decision variables Next, The total consumption of energy-using equipment, For the The CO2 emission coefficient of each energy-using device.

[0109] Substitute the candidate solution corresponding to each explosive body into the objective function, and calculate the function value, which is recorded as the explosion value; in the current group of explosive bodies, select the explosive body with the smallest explosion value as the optimal explosive body; the other explosive bodies are non-optimal explosive bodies; for each non-optimal explosive body, update its position, and the position update method is: calculate the distance between the explosive body and the optimal explosive body ; Generate a random number between 0 and 1 ,like , then the explosive body moves towards the optimal explosive body, and the step length of the movement is ; Otherwise, the explosion body moves away from the optimal explosion body, and the step length is During the movement, check whether the new position of the explosive body exceeds the value range of the decision variable. If so, the position of the corresponding dimension is processed by boundary processing; that is, if the explosive body exceeds the upper boundary of the corresponding parameter in a certain dimension, the coordinate value of the dimension is mapped between the lower boundary and the upper boundary; if it is lower than the lower boundary of the corresponding parameter, it is mapped between the upper boundary and the lower boundary; if the preset maximum number of iterations is reached, the candidate solution corresponding to the current optimal explosive body is output as the final energy optimization strategy; otherwise, the position is updated until the preset maximum number of iterations is reached.

[0110] The formulated energy optimization strategy is converted into an executable control instruction set. According to different equipment types and optimization strategies, corresponding control instructions are generated. The control instruction set is packaged into a data packet that can be sent down, and a secure communication connection with the intelligent gateway is established. The packaged optimization strategy data packet is transmitted to the intelligent gateway through a predefined communication protocol. The intelligent gateway receives and parses the optimization strategy data packet; the intelligent gateway extracts the corresponding control instruction set according to the optimization strategy data packet, and distributes the control instruction set to different subsystem modules, such as the power control module, the heating control module, etc. Each subsystem module parses and executes the corresponding control instructions. The intelligent gateway establishes a communication connection with various energy-using devices inside the building through wired or wireless means, and sends control commands to the corresponding devices according to the control instructions of the optimization strategy. The devices receive and execute control commands, such as adjusting the power supply time, changing the temperature setting, and controlling the water flow. The intelligent gateway monitors the operating status and energy usage data of each device in real time, and uploads the monitored data to the edge computing node for analysis and adjustment of the optimization strategy; according to the feedback data, the optimization strategy is dynamically adjusted to form a closed-loop control.

[0111] Throughout the entire process, the smart gateway plays a key role as a hub and controller. It is a bridge and executor between the optimization strategy and the actual energy-using equipment. Through the coordination of the smart gateway, the energy optimization strategy can be efficiently applied to every corner of the building to achieve comprehensive energy-saving management and control. At the same time, the smart gateway will also collect real-time operation data on site to provide a basis for the dynamic adjustment of the optimization strategy, so that the entire system can be continuously optimized and improved.

[0112] This embodiment can achieve refined management and global optimization of building energy use, greatly improve energy utilization efficiency, and effectively reduce energy consumption and operating costs. The system deeply explores the complex energy use patterns and correlations within the building to develop a more comprehensive and coordinated optimization strategy, avoiding the one-sidedness and limitations of traditional methods; at the same time, it can balance multiple goals such as energy consumption, cost and environmental protection, while improving economic benefits, it also reduces the carbon emissions of buildings and promotes sustainable development; the real-time and adaptability of the system are significantly enhanced, and it can quickly respond to changes in the internal environment of the building and energy demand, dynamically adjust the optimization strategy, and ensure that it always maintains the best operating state; in addition, the system also improves the intelligence and automation level of building energy management, reduces human intervention, and reduces management difficulty and labor costs.

[0113] Example 2

[0114] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, which provides a building energy management method based on intelligent gateway edge computing, including:

[0115] S1. Collect energy usage data and environmental data inside the building;

[0116] S2. Transmit the collected energy usage data and environmental data to the edge computing node through the intelligent gateway; in the edge computing node, pre-process the collected energy usage data and environmental data to obtain pre-processed data;

[0117] S3. Based on the preprocessed data, construct an energy usage pattern diagram, optimize the structure of the energy usage pattern diagram, and identify key nodes and key relationships;

[0118] S4. Based on the identified key nodes and key relationships, an energy optimization strategy is formulated using an optimization algorithm, and the energy optimization strategy is sent to the smart gateway. The energy-using equipment inside the building is controlled through the smart gateway to achieve efficient use of energy.

[0119] Example 3

[0120] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the building energy management method based on intelligent gateway edge computing provided above is implemented.

[0121] Since the electronic device introduced in this embodiment is an electronic device used to implement a building energy management method based on intelligent gateway edge computing in the embodiment of this application, based on the building energy management method based on intelligent gateway edge computing introduced in the embodiment of this application, the technical personnel of this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application is not introduced in detail here. As long as the technical personnel of this field implement the electronic device used in the building energy management method based on intelligent gateway edge computing in the embodiment of this application, it belongs to the scope of protection of this application.

[0122] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0123] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A building energy management system based on intelligent gateway edge computing, characterized in that: include: Data collection module, used to collect energy usage data and environmental data inside the building; Data transmission module, used to transmit the collected energy usage data and environmental data to the edge computing node through the intelligent gateway; In the edge computing node, the collected energy usage data and environmental data are preprocessed to obtain preprocessed data; The comprehensive identification module constructs an energy usage pattern diagram based on the preprocessed data, optimizes the structure of the energy usage pattern diagram, and identifies key nodes and key relationships; The strategy formulation transmission module formulates energy optimization strategies based on the identified key nodes and key relationships using optimization algorithms, sends the energy optimization strategies to the smart gateway, and controls the energy-using devices inside the building through the smart gateway; each module is connected by wired and / or wireless means; The method of preprocessing the collected energy usage data and environmental data includes: creating an M×N two-dimensional array or matrix to represent a grid point matrix, and initializing the values ​​of all grid points in the grid point matrix to be empty or 0; Extract key features from the collected energy usage data and environmental data; the key features include energy type, regional location and time; represent each data point as a vector composed of key features, recorded as a feature vector; Map data points of the same energy type and regional location to adjacent grid points, map data points of different energy types or regional locations to areas consisting of non-adjacent grid points, and map data points at different times to different grid points; traverse all data points, and record the feature vector and original data of the data point in the grid lattice; and superimpose regional boundary information and time axis information on the grid lattice; cluster the data points in the grid lattice; obtain n data clusters; for each data cluster, extract the feature vector corresponding to the data point, map the feature vector to a low-dimensional space, that is, perform dimensionality reduction; obtain preprocessed data; The method of constructing the energy usage pattern diagram includes: For each data cluster, a node is created and the preprocessed data in the data cluster is mapped to the attributes of the node. , calculate its relationship with other nodes The strength of the association between ; ； in, and Node and The corresponding time series data; and Node and The corresponding spatial coordinates; and Node and The corresponding activity time set; For Node Any pre-processed data in For Node Any pre-processed data in ; , , and are the weight coefficients of the corresponding items, and satisfy , , and The sum is 1; A similarity threshold is set. No edge is connected between two nodes whose association strength is less than the similarity threshold, and an edge is connected between two nodes whose association strength is greater than or equal to the similarity threshold, and the weight of the edge is the corresponding association strength; then weighted edges are obtained; and an energy usage pattern graph G'=(V, E) is constructed, where V is a set of nodes and E is a set of weighted edges.

2. A building energy management system based on intelligent gateway edge computing according to claim 1, characterized in that: The energy usage data includes energy type, energy usage, energy usage power / flow, energy usage equipment / area ID and energy usage timestamp; the environmental data includes temperature, humidity, light intensity, carbon dioxide concentration, presence of personnel, area / room ID and environmental data collection timestamp.

3. A building energy management system based on intelligent gateway edge computing according to claim 2, characterized in that: The method of clustering data points in the grid lattice includes: Randomly select K grid points in the grid lattice as the initial positions of the gas molecules, and record the coordinates of the grid point where each gas molecule is currently located; Define the neighborhood radius r and calculate the diffusion probability of gas molecules within the neighborhood radius r ;in, , and All are control parameters; is the total mass of all gas molecules within the neighborhood radius r, is the average distance of a gas molecule from other gas molecules within the neighborhood radius r; A diffusion threshold is preset. For each gas molecule, if the corresponding calculated diffusion probability is greater than the diffusion threshold, the corresponding gas molecule diffuses to the adjacent grid point; if the corresponding calculated diffusion probability is less than or equal to the diffusion threshold, the corresponding gas molecule keeps the original position unchanged; the coordinates of the new position of each gas molecule are updated; Traverse each grid point in the grid matrix. If the grid point is the location of the data point, calculate the grid point The degree of aggregation of nearby gas molecules ; Take this as the degree of clustering of the data points; ;in, are nearby gas molecules and lattice points The distance between is the base aggregation radius; is a positive parameter; is a positive exponential parameter; Grid point The quality of the corresponding data points; is the minimum value of the quality of all data points, is the maximum value of the mass of all data points; Define a clustering threshold, traverse all data points, if the degree of clustering of data points is greater than the clustering threshold, they are classified into a new cluster; if the degree of clustering of data points is less than or equal to the clustering threshold, they are temporarily marked as noise points; for noise points, according to their distance from the existing clusters, they are classified into the nearest cluster to obtain n data clusters.

4. A building energy management system based on intelligent gateway edge computing according to claim 3, characterized in that: The diffusion method includes: Define the mass of each gas molecule; for gas molecule j, calculate the gravitational force between it and other gas molecules i ;in, is the gravitational constant, is the mass of gas molecule i, is the mass of gas molecule j, is the Euclidean distance between gas molecule j and other gas molecules i, To adjust the parameters; For gas molecule j, calculate the repulsive force between it and other gas molecules i ;in, is the repulsive force constant, is a characteristic length, is a positive number, and To adjust the power index; Based on the calculated attraction and repulsion of gas molecules, define the comprehensive influence field ; ;in, is a random number between 0 and 1. is the random perturbation intensity constant; For the current position of the gas molecule, calculate the value of the comprehensive influence field of its adjacent positions, and take the adjacent position corresponding to the value of the largest comprehensive influence field as the new position. If there are several adjacent positions with the same maximum comprehensive influence field value, randomly select one adjacent position as the new position; that is, the diffusion is completed.

5. A building energy management system based on intelligent gateway edge computing according to claim 4, characterized in that: The method of mapping the feature vector to the low-dimensional space includes: For every two data points in the same data cluster, the corresponding eigenvectors and ; Calculate the comprehensive similarity between them ; The comprehensive similarity Construct the feature vector similarity matrix P as elements; ;in, is the feature vector The k nearest neighbor set of ; is the feature vector The k nearest neighbor set of ; is the exponential parameter, and is the intra-cluster distance adjustment parameter; is the feature vector The distance from the corresponding data point to the cluster center of the data cluster to which it belongs; is the feature vector The distance from the corresponding data point to the cluster center of the data cluster to which it belongs; Initialize N1 low-dimensional data points randomly in the preset low-dimensional space, and for every two low-dimensional data points X and Y in the low-dimensional space, calculate the comprehensive similarity p_XY between them; construct the low-dimensional space similarity matrix Q using the comprehensive similarity p_XY as an element; The KL divergence between the low-dimensional space similarity matrix Q and the eigenvector similarity matrix P is calculated, and the position of the low-dimensional data points is iteratively optimized using the gradient descent method to minimize the KL divergence; after the iterative optimization converges, the final low-dimensional data points are obtained, corresponding to the data points in the data cluster, and the dimensionality reduction is completed.

6. A building energy management system based on intelligent gateway edge computing according to claim 5, characterized in that: The method of optimizing the structure of the energy usage pattern diagram includes: Assign a unique initial label to each node and record the initial labels of all nodes; set the maximum number of iterations mx and the convergence threshold co, initialize a node label list, and record the weight of each label, defining the weight of the label as the sum of the degrees of the corresponding nodes; Copy the current node label list to get a new node label list, which is recorded as the reference node label list; traverse each node , get the node All connected nodes and its labels, calculate the comprehensive weight of each label, that is, the label corresponds to the node The sum of the degrees of the nodes The sum of the degrees of ; find the label ma with the largest comprehensive weight; if the comprehensive weight of ma is greater than the node The weight of the current label; then compare the node in the node label list The current label is updated to ma; otherwise the node The current label remains unchanged; calculate the ratio of nodes whose reference node label list is different from the node label list. If the ratio is less than the convergence threshold co, stop; otherwise, copy the reference node label list back to the node label list and continue the next round of iteration until the maximum number of iterations mx is reached, and then terminate the iteration; Nodes with the same label are divided into a subgraph, and these subgraphs constitute the structure of the optimized energy usage pattern graph.

7. A building energy management system based on intelligent gateway edge computing according to claim 6, characterized in that: The method of identifying key nodes and key associations includes: Calculate the degree centrality and betweenness centrality of each node; preset corresponding thresholds, and identify nodes whose degree centrality and betweenness centrality are higher than the corresponding thresholds as key nodes; For each edge in the structure of the optimized energy usage pattern graph, the number of times it is between the shortest paths of other node pairs is calculated, and an edge betweenness threshold is set. The edges with a number higher than the edge betweenness threshold are identified as key associations.

8. A building energy management system based on intelligent gateway edge computing according to claim 7, characterized in that: The methods for developing energy optimization strategies include: Each key node corresponds to an energy-using device, and the comprehensive attributes of the node are analyzed. The comprehensive attributes include the energy type, time series data, spatial coordinates, and activity time set; According to the comprehensive attributes of each key node, its optimizable parameters are determined as decision variables. The value range of the decision variable is the reasonable variation range of the parameter. Analyze the association attributes of key associations, including association strength and betweenness centrality; preset a constraint threshold, and subject two nodes corresponding to key associations whose association strength is greater than the constraint threshold to variable constraints, where the variable constraint is the absolute value of the sum of the decision variables corresponding to the two nodes, which is less than the preset variable threshold; A path threshold is preset, and two nodes corresponding to the key association whose betweenness centrality is greater than the path threshold are subjected to path constraints. The path constraint is that the decision variables of the two nodes corresponding to the key association satisfy the temporal or quantitative relationship. According to the number of decision variables n1, an initial group of explosive bodies consisting of N2 explosive bodies is generated. Each explosive body corresponds to an n1-dimensional vector, representing a candidate solution. The position coordinates of the explosive body are randomly initialized within the value range of the decision variables. Define the objective function ;in, , and is the weight coefficient of the objective function; is a vector of decision variables, For the The vector of decision variables consisting of energy-using equipment The power under For the The operating time of each energy-using device, For the Unit rates for each energy-using device; is the vector of decision variables Next, The total consumption of energy-using equipment, For the CO2 emission coefficient of each energy-using equipment; Substitute the candidate solution corresponding to each explosive body into the objective function, and calculate the function value, which is recorded as the explosion value; in the current group of explosive bodies, select the explosive body with the smallest explosion value as the optimal explosive body; the other explosive bodies are non-optimal explosive bodies; for each non-optimal explosive body, update its position, and the position update method is: calculate the distance between the explosive body and the optimal explosive body ; Generate a random number between 0 and 1 ,like , then the explosive body moves towards the optimal explosive body, and the step length of the movement is ; Otherwise, the explosion body moves away from the optimal explosion body, and the step length is ; During the movement, check whether the new position of the explosion body exceeds the value range of the decision variable. If it exceeds, the position of the corresponding dimension is processed at the boundary. If the preset maximum number of iterations is reached, the candidate solution corresponding to the current optimal explosive body is output as the final energy optimization strategy; otherwise, the position is continued to be updated until the preset maximum number of iterations is reached.

Images