CN-120634102-B - Intelligent building energy consumption data monitoring and management method and system

CN120634102BCN 120634102 BCN120634102 BCN 120634102BCN-120634102-B

Abstract

The invention discloses a smart building energy consumption data monitoring management method and system, which relate to the technical field of smart building and energy management and comprise the steps of constructing a dynamic graph structure, extracting joint characteristics by using a graph injection force mechanism, weighting the mahalanobis distance of nodes by using a batch average attention coefficient to obtain abnormal scene characteristic vectors, defining a classification objective function, optimizing by using an EPC-PSO algorithm, classifying the abnormal scene characteristic vectors by using a Softmax classifier, defining an energy consumption efficiency objective function, decomposing the energy consumption efficiency objective function into sub-problems of each device by using a Lagrange method, outputting a global initial strategy vector by using a gradient descent method, defining a smart building task, constructing a matrix of comprehensive benefit weight, converting the devices and tasks into bipartite graph problems, solving by using a Hungary KM algorithm, weighting the mahalanobis distance by using the batch average attention coefficient, enhancing the robustness of abnormal detection, and improving the comprehensive benefit of resources by using the Lagrange method, the gradient descent method and the Hungary KM algorithm.

Inventors

LIU ZHONGSHUAI
Qiao Bingfeng
PENG KE
WANG MIAOYU
LIU LEISHENG
GUO TAO
LI JINGFEI
WANG BIN

Assignees

朗高科技有限公司

Dates

Publication Date: 20260512
Application Date: 20250527

Claims (9)

1. A method for monitoring and managing intelligent building energy consumption data is characterized by comprising the following steps of, S1, constructing a distributed architecture, collecting energy consumption and environmental data for preprocessing, constructing a dynamic graph structure, extracting joint characteristics of the energy consumption and the environmental data by using a graph annotation mechanism GAT, weighting Mahalanobis distances of nodes by using a batch average attention coefficient, normalizing the abnormal scores of all the nodes as abnormal scores, setting an abnormal threshold by using an empirical method, screening out nodes with the abnormal score larger than the abnormal threshold by using a fixed threshold screening method, generating an abnormal node set, and transmitting the abnormal node set to a main control node by using an edge node; The main control node weights the anomaly score to the received feature vector in the anomaly node set by using a multiplication method to obtain an anomaly scene feature vector; S2, defining a classification objective function, performing iterative optimization by using an EPC-PSO algorithm, classifying the abnormal scene feature vectors by using a Softmax classifier, and calculating the severity of each abnormal type, wherein the method comprises the following steps: Initializing EPC-PSO particle groups on a main control node, and extracting collected historical energy consumption data with labels and abnormal data in historical environment data to be used as a historical abnormal data set; Defining a classification objective function, inputting each initialized particle position as a classification boundary parameter into the classification objective function, calculating an objective function value, selecting a global optimal particle position of the current iteration, updating each particle speed and particle position by using an EPC movement rule, outputting the global optimal position after the maximum iteration times are reached, substituting the global optimal position into the classification objective function to obtain an initialized individual optimal position, using PSO (particle swarm optimization) as the initialized individual optimal position to update the speed and position, calculating a classification score based on an abnormal scene feature vector, generating a classification score vector, calculating the probability of the classification score vector by using a Softmax function, setting an abnormal type weight, calculating the initial severity score of the abnormal score in the classification score vector by using an average severity calculation method, and calculating the adjusted severity score by using a multiplication method; Setting a high-risk threshold and a medium-risk threshold respectively based on a rule of thumb, and setting an alarm level grading rule by using the adjusted severity score; the update of each particle velocity using EPC movement rules, formula: , Wherein, the Is particle p at the first The speed of the number of iterations is such that, For the temperature factor, the control explore amplitude, the general value ensures global searching capability, For the random value, the search diversity is enhanced, As a distance factor, based on the euclidean distance between the particles and the global optimum, the universal value balances the convergence speed, To measure the euclidean distance of the proximity of a particle to the optimal solution, For the global optimal particle position for the c-th iteration, the best classification boundary parameter is represented, The position of particle p for the c-th iteration; S3, defining an energy consumption efficiency objective function, decomposing the energy consumption efficiency objective function into sub-problems of each device by using Lagrange, outputting a global initial strategy vector by using a gradient descent method, defining an intelligent building task, constructing a matrix of comprehensive benefit weights, converting the devices and the tasks into bipartite graph problems, and solving the bipartite graph problems by using a Hungary KM algorithm; the defining of the energy consumption efficiency objective function comprises integrating the abnormal score, the predicted abnormal type, the adjusted severity score, the node characteristics of the dynamic graph structure, the edge weight, the preprocessed energy consumption data and the preprocessed environment data, and carrying out normalization processing to generate a state vector; The master control node distributes the state vector to the edge node through a BACnet protocol, an energy consumption efficiency target is constructed at the edge node, and the ratio of the performance to the energy consumption is maximized; S4, collecting real-time data to perform intelligent building energy consumption management, and constructing data generated by visual interface display analysis.
2. The intelligent building energy consumption data monitoring and management method according to claim 1, wherein S1 comprises the following steps: the distributed architecture comprises an edge node and a main control node; the collected energy consumption and environment data comprise current, power, energy consumption rate, temperature, humidity and smoke concentration data; Constructing a time slice data matrix, mapping each data type in the time slice data matrix into a node, and adding edges of a dynamic graph structure by using pearson correlation coefficients; Calculating the comprehensive volatility of the edges among the nodes added with the edges, converting the comprehensive volatility into a time attenuation factor by using a linear mapping function, calculating the dynamic weight of the edges, constructing a dynamic graph structure, and calculating an adjacency matrix; Based on the adjacency matrix and the dynamic graph structure, extracting joint characteristics of energy consumption and environmental variables at edge nodes by using a graph annotation mechanism GAT; Calculating the weighted attention score of each pair of nodes by using weighted multi-head attention, calculating the average attention coefficient of each node, aggregating neighbor features for each head by using a ReLU activation function, performing multi-head splicing on feature vectors generated by each head to obtain a feature matrix H, and dividing the features in the feature matrix H and the average attention coefficient of each node into batches to obtain a batch feature matrix and a batch attention coefficient; Calculating the mean value and covariance matrix of each batch of feature matrixes, calculating the Mahalanobis distance of each node, weighting the Mahalanobis distance of each node by using a batch attention coefficient as an anomaly score, carrying out normalization processing, screening out nodes with anomaly scores larger than an anomaly threshold by using a fixed threshold screening method, generating an anomaly node set, and transmitting the anomaly node set to a main control node through an edge node; The main control node weights the anomaly score to the received feature vectors in the anomaly node set by using a multiplication method to obtain anomaly scene feature vectors, and all the anomaly scene feature vectors are used for generating the anomaly scene set.
3. The intelligent building energy consumption data monitoring and management method according to claim 2, wherein the defining the energy consumption efficiency objective function, decomposing the energy consumption efficiency objective function into sub-problems of each device by using Lagrangian, and outputting a global initial strategy vector by using a gradient descent method comprises: Constructing an energy consumption efficiency target at an edge node, maximizing the ratio of performance to energy consumption, setting an edge node power and performance constraint condition, decomposing a local optimization target and the constraint condition into sub-problems of each device by using Lagrange, iteratively updating power by using a gradient descent method for each device, and calculating the environmental quality by the updated power; Calculating updated equipment efficiency by using a ratio method, calculating updated comprehensive efficiency of each equipment by using a weighted summation method, setting an equipment efficiency threshold by using a fixed value method, comparing the equipment efficiency threshold with the updated comprehensive efficiency to judge the equipment switching state, and obtaining a quantized value of a mode by using a mapping method based on updated power; Updating the Lagrangian multiplier based on the updated power, outputting the optimal power, switching state and running mode of the equipment after the maximum iteration times are reached, and generating a global initial strategy vector at a main control node.
4. The intelligent building energy consumption data monitoring and management method according to claim 3, wherein the defining intelligent building tasks, constructing a matrix of comprehensive benefit weights, converting equipment and tasks into bipartite graph problems, and solving by using Hungary KM algorithm comprises: defining standard forms of four tasks of the intelligent building and candidate action rules of a normal scene and an abnormal scene, setting comprehensive benefit weights of each device under the tasks and the candidate actions, constructing the comprehensive benefit weights of each device into a weight matrix, and generating candidate action sets by the candidate actions of the normal scene and the abnormal scene; the four tasks comprise a ventilation task, a refrigeration task, an indoor temperature reduction task, an illumination task, an indoor illumination task and an entertainment task, wherein the ventilation task is used for adjusting air quality, the refrigeration task is used for reducing indoor temperature, the illumination task is used for providing indoor illumination, and the entertainment task is used for supporting functions of entertainment equipment; Each edge node transmits a weight matrix and a candidate action set to a main control node, the main control node aggregates the weight matrix of all the edge nodes through a mean aggregation method to obtain initial global weights, action constraint weight correction is carried out on the initial global weights based on the candidate action set to obtain final global weights, and the main control node distributes the global weights and the candidate action set to each edge node; Based on global weight, constructing a bipartite graph, removing unreasonable equipment-task edges by using an equipment-task suitability verification method, supplementing nodes with equipment numbers unequal to task numbers by using virtual nodes, adding the virtual nodes into the bipartite graph by using a graph expansion method, and obtaining an updated bipartite graph; Setting single allocation constraint by taking maximized global weight and task matching as objective functions; based on the updated bipartite graph, setting an empty matching set, performing equipment-task edge matching by using a KM Hungary algorithm, and outputting a final matching set; and each edge node transmits the final matching set to the main control node, and the main control node converts the final matching set into equipment control instructions and distributes the equipment control instructions to the edge nodes to execute task allocation.
5. The intelligent building energy consumption data monitoring and management method according to claim 4, wherein the collecting real-time data performs intelligent building energy consumption management, and comprises: The edge node collects real-time data, the main control node analyzes normal and abnormal conditions of the data, the main control node aims at minimizing energy consumption based on different scenes, tasks are sequentially executed at an alarm level, and a final matching set is used for controlling equipment sending instructions controlled by the edge node.
6. The intelligent building energy consumption data monitoring and management method according to claim 5, wherein the constructing the visual interface displays the data generated by analysis, and the method comprises the following steps: building a visual interface by using BIM, and displaying four areas; An interactive interface is provided that supports a manager to manually adjust device status and confirm automated advice.
7. An intelligent building energy consumption data monitoring and management system based on the intelligent building energy consumption data monitoring and management method of any one of claims 1-6 is characterized by comprising the following steps of, The collection construction module is used for constructing a distributed architecture, collecting energy consumption and environment data for preprocessing, constructing a dynamic graph structure, extracting joint characteristics of the energy consumption and the environment data by using a graph annotation force mechanism GAT, and weighting the Mahalanobis distance of the nodes by using a batch average attention coefficient to obtain an abnormal scene feature vector; The definition classification module is used for defining a classification objective function, performing iterative optimization by using an EPC-PSO algorithm, classifying the abnormal scene feature vectors by using a Softmax classifier, and calculating the severity of each abnormal type; The decomposition solving module is used for defining an energy consumption efficiency objective function, decomposing the energy consumption efficiency objective function into sub-problems of each device by using Lagrange, outputting a global initial strategy vector by using a gradient descent method, defining an intelligent building task, constructing a matrix of comprehensive benefit weights, converting the devices and the tasks into bipartite graph problems, and solving the bipartite graph problems by using a Hungary KM algorithm; and the execution visualization module is used for collecting real-time data, executing intelligent building energy consumption management and constructing data generated by visual interface display analysis.
8. The computer equipment comprises a memory and a processor, wherein the memory stores a computer program, and the computer program is characterized in that the processor realizes the steps of the intelligent building energy consumption data monitoring and management method according to any one of claims 1-6 when executing the computer program.
9. A computer readable storage medium having a computer program stored thereon, wherein the computer program when executed by a processor implements the steps of the intelligent building energy consumption data monitoring and management method of any one of claims 1 to 6.

Description

Intelligent building energy consumption data monitoring and management method and system Technical Field The invention relates to the technical field of intelligent building and energy management, in particular to an intelligent building energy consumption data monitoring and management method and system. Background Along with the continuous promotion of ideas such as smart city, green building, etc., building energy consumption management becomes one of the key links of promoting city sustainable development ability, in recent years, the development of emerging technologies such as internet of things (IoT), edge calculation, artificial intelligence, etc., provides important support for acquisition, processing and analysis of energy consumption data in smart building, current smart building system generally relies on various sensor nodes deployed in the building for collecting multidimensional energy consumption data such as electric power, water, gas, etc. and environmental parameters, and gathers, models and analyzes through data centralized platform to realize functions such as energy consumption trend prediction, operation efficiency evaluation and abnormal early warning. Although the related technology achieves a certain achievement, many challenges and limitations are faced, most of the existing methods only analyze energy consumption data or environment data separately, lack deep mining of correlation features between the two, cause misjudgment or missed judgment in the anomaly detection process, often fail to capture complex dependency relationships between nodes effectively and lack robustness when the traditional anomaly detection algorithm processes high-dimensional time sequence data, and in the aspects of task scheduling and energy consumption optimization, most of the current common methods are based on heuristic algorithms or centralized solving frameworks, are difficult to operate efficiently in a distributed system, and fail to fully consider nonlinear coupling relationships between equipment and tasks and resource constraint conditions. Disclosure of Invention The present invention has been made in view of the above-described problems occurring in the prior art. Therefore, the invention provides a method and a system for monitoring and managing intelligent building energy consumption data, which solve the problems that most of the existing methods only analyze energy consumption data or environment data separately, lack of deep mining on correlation characteristics between the two, cause misjudgment or missed judgment easily to occur in an anomaly detection process, often cannot effectively capture complex dependency relationships between nodes and lack robustness when a traditional anomaly detection algorithm processes high-dimensional time sequence data, and most of the current common methods are based on heuristic algorithms or centralized solving frames in terms of task scheduling and energy consumption optimization, are difficult to operate efficiently in a distributed system, and fail to fully consider nonlinear coupling relationships between equipment and tasks and resource constraint conditions. In order to solve the technical problems, the invention provides the following technical scheme: in a first aspect, the present invention provides a method for monitoring and managing energy consumption data of a smart building, comprising, Constructing a distributed architecture, collecting energy consumption and environmental data for preprocessing, constructing a dynamic graph structure, extracting joint characteristics of the energy consumption and the environmental data by using a graph annotation force mechanism GAT, and weighting the Mahalanobis distance of the nodes by using a batch average attention coefficient to obtain an abnormal scene feature vector; Defining a classification objective function, performing iterative optimization by using an EPC-PSO algorithm, classifying the abnormal scene feature vectors by using a Softmax classifier, and calculating the severity of each abnormal type; Defining an energy consumption efficiency objective function, decomposing the energy consumption efficiency objective function into sub-problems of each device by using Lagrange, outputting a global initial strategy vector by using a gradient descent method, defining an intelligent building task, constructing a matrix of comprehensive benefit weights, converting the devices and the tasks into bipartite graph problems, and solving the bipartite graph problems by using a Hungary KM algorithm; And collecting real-time data, performing intelligent building energy consumption management, and constructing data generated by visual interface display analysis. The intelligent building energy consumption data monitoring management method is a preferable scheme, wherein the construction of a dynamic graph structure, the extraction of the combined characteristics of energy consumption and environmental data