Building energy consumption optimization method based on artificial intelligence

Abstract

The invention relates to the technical field of energy consumption simulation, in particular to a building energy consumption optimization method based on artificial intelligence. The method comprises the following steps: acquiring multi-modal sensing data, and carrying out equipment physical space semantic mapping modeling to obtain a sensing time-space fusion feature set; extracting equipment energy transfer influence factors according to the sensing time-space fusion feature set to obtain a dynamic energy consumption influence factor matrix; constructing an energy consumption probability cloud picture according to the dynamic energy consumption influence factor matrix; performing adversarial strategy optimization on the energy consumption probability nephogram to obtain a multi-target energy-saving control strategy manifold of multi-target energy-saving control; generating an energy-saving optimization instruction set based on the multi-target energy-saving control strategy manifold; and acquiring real-time building energy consumption data, and implementing equipment control local comfort level instruction optimization on the real-time building energy consumption data according to the energy-saving optimization instruction set to obtain a personalized control instruction set. According to the invention, intelligence, refinement and individuation of building energy consumption management are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of energy consumption simulation, and in particular, to an artificial intelligence-based building energy consumption optimization method. Background Art

[0002] Existing building energy-saving systems usually rely on preset rules or static control strategies (such as fixed temperature thresholds, equipment start-stop control based on time periods). Although they reduce energy waste to a certain extent, they are difficult to adapt to complex environmental dynamic changes. For example, weather conditions (temperature, humidity, light intensity) may fluctuate rapidly over time, and the occupancy density and equipment usage inside the building also show non-linear changes. A single fixed strategy often cannot adjust control parameters in a timely manner, resulting in lagging energy management and poor energy efficiency optimization effects. Currently, most energy-saving management systems still focus on single-device optimization. For example, the air-conditioning system adjusts the cooling capacity according to the room temperature, and the lighting system controls the switch based on preset time or light sensors, while devices such as elevators, fresh air systems, and hot water supplies operate independently. This isolated optimization method ignores the coupling relationship between multiple devices inside the building. For example, the energy consumption of the air-conditioning system and the fresh air system affects each other. Changes in the fresh air volume of the fresh air system may cause drastic fluctuations in the air-conditioning load, thereby affecting the overall energy consumption balance. Similarly, the operation mode of the elevator and the lighting load distribution will also indirectly affect the heat load of the air-conditioning system. Without cross-system collaborative optimization, the overall energy scheduling efficiency will be low. In addition, although the application of Internet of Things (IoT) technology has improved the building energy consumption monitoring ability and can collect multi-dimensional data such as temperature, humidity, light intensity, and equipment operation status in real time, the utilization rate of these data is relatively low. Existing systems are mainly used for status monitoring and early warning, lacking in-depth data analysis and intelligent decision-making capabilities. In most scenarios, energy-saving strategies still rely on manual adjustment or simple rule setting, and it is difficult to achieve autonomous learning and dynamic optimization. For example, when the occupancy density changes or the external environment fluctuates, traditional systems usually cannot respond quickly, resulting in unreasonable energy scheduling, affecting user comfort and energy use efficiency. Summary of the Invention

[0003] Based on this, it is necessary for the present invention to provide an artificial intelligence-based building energy consumption optimization method to solve at least one of the above technical problems.

[0004] To achieve the above object, an artificial intelligence-based building energy consumption optimization method includes the following steps:

[0005] Step S1: Obtain multi-modal sensing data, and construct a heterogeneous sensing dynamic spatio-temporal adjacency matrix; perform device physical space semantic mapping modeling according to the heterogeneous sensing dynamic spatio-temporal adjacency matrix to obtain a sensing spatio-temporal fusion feature set;

[0006] Step S2: Perform cross - domain energy consumption impact mapping topology modeling on the sensor spatio - temporal fusion feature set to obtain a cross - domain energy consumption correlation topology graph; extract the device energy transfer impact factors according to the cross - domain energy consumption correlation topology graph to obtain a dynamic energy consumption impact factor matrix;

[0007] Step S3: Conduct multi - scale energy consumption evolution deduction based on the dynamic energy consumption impact factor matrix to generate a building energy consumption evolution map; construct an energy consumption probability cloud map according to the building energy consumption evolution map;

[0008] Step S4: Optimize the adversarial strategy for the energy consumption probability cloud map to obtain a multi - objective energy - saving control multi - objective energy - saving control strategy manifold; generate an energy - saving optimization instruction set based on the multi - objective energy - saving control strategy manifold and upload it to the building energy consumption control platform to control the devices;

[0009] Step S5: Obtain real - time building energy consumption data, track the deviation of predicted energy consumption of devices from the blood relationship for the real - time building energy consumption data according to the energy - saving optimization instruction set, and implement the optimization of the local comfort instruction of device control to obtain a personalized control instruction set and upload it to the building energy consumption control platform to control the devices.

[0010] Optionally, step S1 is specifically as follows:

[0011] Step S11: Obtain multi - modal sensing data through a multi - modal sensor group, where the multi - modal sensing data includes temperature and humidity sensing data, light sensing data, infrared sensing data of the population density, and device status sensing data;

[0012] Step S12: Perform time - step time - series alignment on the multi - modal sensing data to obtain a time - series aligned environmental data set;

[0013] Step S13: Perform non - Euclidean topology mapping on the time - series aligned environmental data set and construct an initial spatial semantic topology structure diagram based on the topology mapping result;

[0014] Step S14: Perform neighborhood - adaptive topology optimization based on the initial spatial semantic topology structure diagram to obtain a device spatial semantic graph, and fuse the device operation mode and environmental data based on the device spatial semantic graph to obtain a heterogeneous sensing dynamic spatio - temporal adjacency matrix;

[0015] Step S15: Establish a device physical space semantic mapping model according to the heterogeneous sensing dynamic spatio - temporal adjacency matrix, perform feature screening and correlation optimization on the model establishment result to obtain a sensor spatio - temporal fusion feature set.

[0016] Optionally, step S13 is specifically as follows:

[0017] Step S131: Obtain the operation data of building area devices, and perform synchronous time - series alignment on the operation data of building area devices and the time - series aligned environmental data set to obtain time - series aligned device operation data;

[0018] Step S132: Perform non-Euclidean topological mapping on the time-series aligned device operation data and the time-series aligned environment data set to construct a spatio-temporal relationship map between the devices and environment variables within the building area, obtaining the building area spatio-temporal relationship map;

[0019] Step S133: Based on the building area spatio-temporal relationship map, perform physical space interaction modeling, and map the physical space interaction modeling result to a preset building physical space coordinate system, convert the nodes in the spatio-temporal relationship map into actual physical space coordinate nodes, and construct device-environment physical space interaction data;

[0020] Step S134: Perform node spatial layout analysis on the device-environment physical space interaction data, and map the interaction relationship between the device and the environment according to the node spatial layout analysis result, obtaining the environment physical space interaction diagram;

[0021] Step S135: Based on the environment physical space interaction diagram, perform graph optimization processing, and construct an initial spatial semantic topological structure diagram.

[0022] Optionally, step S14 is specifically as follows:

[0023] Step S141: Calculate the local density of topological nodes according to the initial spatial semantic topological structure diagram, set the calculation neighborhood radius to 2, set the local density threshold to 0.1, and consider the nodes below the local density threshold as sparse regions, thereby obtaining the local density distribution of the topological structure;

[0024] Step S142: Set the minimum sample number to 5, the density threshold to 0.3, perform node neighborhood adaptive clustering on the initial spatial semantic topological structure diagram, and set the weight range to [0,1] to calculate the neighborhood adaptability weight based on the clustering result, generating the initial parameters for neighborhood adaptive topological optimization;

[0025] Step S143: Set the weight adjustment coefficient to 0.5, according to the initial parameters for neighborhood adaptive topological optimization, perform neighborhood topological weight adjustment on the initial spatial semantic topological structure diagram, and set the optimization scale to [1,3] to perform multi-scale topological structure optimization, obtaining the optimized device spatial semantic map;

[0026] Step S144: Extract device operation mode features with a feature dimension of 5 based on the optimized device spatial semantic map, and fuse the time-series aligned environment data set, perform normalization processing on the device operation mode features, generating standardized device operation mode data;

[0027] Step S145: Embed the standardized device operation mode data into the topological structure of the optimized device spatial semantic map, perform interaction modeling between the device operation mode and the environment data, obtaining a preliminary heterogeneous sensing dynamic spatio-temporal adjacency matrix;

[0028] Step S146: Perform feature screening and topological correlation optimization on the preliminary heterogeneous sensing dynamic spatio-temporal adjacency matrix, strengthen key interaction relationships, and remove low-correlation edge weights, so as to generate a heterogeneous sensing dynamic spatio-temporal adjacency matrix.

[0029] Optionally, step S2 is specifically as follows:

[0030] Step S21: Screen the cross-domain energy consumption impact factors for the sensing spatio-temporal fusion feature set, set the correlation threshold to [0.7, 0.8] to retain high-correlation features, calculate the energy consumption impact weights of each device, and construct an initial energy consumption impact feature matrix based on the energy consumption impact weights;

[0031] Step S22: Perform non-Euclidean graph topological mapping on the initial energy consumption impact feature matrix, set the maximum connection threshold of the adjacency matrix to 0.5, construct an energy consumption correlation topological structure across devices and spatial regions, and generate an initial cross-domain energy consumption correlation topological graph;

[0032] Step S23: Perform spectral clustering analysis on the initial cross-domain energy consumption correlation topological graph, select the first 20 eigenvectors for Laplacian eigen-decomposition, and set the feature truncation threshold to 0.1 to extract the main energy consumption impact patterns, generating a cross-domain energy consumption correlation topological graph;

[0033] Step S24: Model the energy consumption transfer relationship between devices based on the cross-domain energy consumption correlation topological graph, set the energy consumption transfer rate to [0.05, 0.2], and set the dynamic weight change interval to [0, 1] to allocate dynamic weights, generating an initial dynamic energy consumption impact factor matrix;

[0034] Step S25: Perform normalization processing on the initial dynamic energy consumption impact factor matrix, set the impact factor confidence interval to [0.95, 1.05] to correct the robustness of the impact factor, and optimize the energy consumption impact parameters, outputting a dynamic energy consumption impact factor matrix.

[0035] Optionally, the multi-scale energy consumption evolution deduction described in step S3 is specifically as follows:

[0036] Perform time series decomposition on the dynamic energy consumption impact factor matrix, set the multi-scale time step, and perform trend decomposition and periodic decomposition to generate an energy consumption time series decomposition data set;

[0037] Perform causal inference on the energy consumption propagation path for the energy consumption time series decomposition data set, calculate the energy consumption correlation strength between devices, and construct an energy consumption propagation network, generating an initial energy consumption propagation structure;

[0038] Perform energy consumption propagation mode clustering based on the initial energy consumption propagation structure, screen the energy consumption propagation paths, and remove low-contribution energy consumption paths, generating an optimized energy consumption propagation network;

[0039] Extract time - series features from the optimized energy - consumption propagation network, divide the energy - consumption time series into segments, and perform empirical mode trend decomposition on the energy - consumption time - series segments to obtain energy - consumption time - series feature data;

[0040] Based on the energy - consumption time - series feature data, conduct autoregressive integrated short - term energy - consumption trend modeling to generate short - term energy - consumption prediction data;

[0041] Conduct residual analysis on the short - term energy - consumption prediction data, input the results of the residual analysis into a preset long short - term memory network for long - term energy - consumption evolution modeling, and obtain long - term energy - consumption prediction data;

[0042] Fuse the long - term energy - consumption prediction data and the short - term energy - consumption prediction data to construct a comprehensive energy - consumption prediction sequence; based on the comprehensive energy - consumption prediction sequence, perform multi - resolution wavelet transform to extract energy - consumption change patterns at different time scales and generate a multi - scale energy - consumption evolution prediction matrix;

[0043] Based on the multi - scale energy - consumption evolution prediction matrix, conduct state probability modeling, calculate the energy - consumption state transition matrix, and combine the energy - consumption state transition matrix with the building energy - consumption evolution atlas to obtain the building energy - consumption evolution atlas.

[0044] Optionally, the long - term energy - consumption evolution modeling is specifically as follows:

[0045] Align the time steps of the short - term energy - consumption prediction data and the results of the residual analysis to obtain a time - series aligned residual data set;

[0046] Extract local spatio - temporal features from the time - series aligned residual data set to generate residual feature data;

[0047] Based on the residual feature data set, construct a long short - term memory network model;

[0048] Perform backpropagation on the long short - term memory network model, use the gradient descent method to optimize the network parameters, conduct long - term energy - consumption evolution modeling on the time - series aligned residual data set, and obtain a long - term energy - consumption evolution model;

[0049] Use the long - term energy - consumption evolution model to perform long - term energy - consumption prediction on the residual feature data to obtain long - term energy - consumption prediction data.

[0050] Optionally, the generation of the energy - saving optimization instruction set in step S4 is specifically as follows:

[0051] Conduct device - category strategy adaptability analysis based on the multi - objective energy - saving control strategy manifold, and conduct hierarchical clustering of device control requirements to generate a device control classification index;

[0052] According to the device control classification index, perform energy - saving strategy constraint mapping on the multi - objective energy - saving control strategy manifold to obtain a constraint - optimized strategy set;

[0053] Perform time - series scheduling optimization on the constraint optimization policy set, adjust the policy execution time, and generate a time - optimized policy set;

[0054] Based on the time - optimized policy set, allocate policy execution priorities, and adaptively adjust policy execution parameters in combination with the time - series alignment environment data set to generate an adaptive energy - saving control instruction set;

[0055] Perform consistency verification on the adaptive energy - saving control instruction set, verify the execution feasibility of the adaptive energy - saving control instruction set, obtain an energy - saving optimized instruction set, and upload it to the building energy consumption control platform to execute equipment control tasks.

[0056] Optionally, the specific process of equipment predicted energy consumption deviation blood - relationship tracing in step S5 is as follows:

[0057] Obtain the historical energy consumption data of building equipment, and establish a benchmark energy consumption prediction model based on the historical energy consumption data of building equipment and the energy - saving optimized instruction set, so as to obtain equipment benchmark energy consumption prediction data;

[0058] Obtain real - time building energy consumption data, compare the error between the equipment benchmark energy consumption prediction data and the real - time building energy consumption data, calculate the prediction energy consumption deviation metric, and conduct preliminary screening of abnormal energy consumption based on the preset deviation prediction to obtain an equipment energy consumption deviation metric matrix;

[0059] Obtain the equipment control log, construct a causal relationship diagram of equipment energy consumption change according to the equipment control log and the dynamic energy consumption impact factor matrix, perform causal test of energy consumption impact factors, and generate an equipment energy consumption blood - relationship tracing chain;

[0060] Combine the equipment blood - relationship tracing chain to trace the source of energy consumption anomalies in the equipment energy consumption deviation metric matrix, and classify the abnormal energy consumption patterns of equipment based on the results of energy consumption anomaly source tracing to obtain an energy consumption deviation retrospective analysis report.

[0061] Optionally, the specific process of establishing the benchmark energy consumption prediction model is as follows

[0062] Obtain the historical energy consumption data of building equipment, and perform time - series alignment on the historical energy consumption data of building equipment to obtain a standardized equipment historical energy consumption data set;

[0063] Quantify the energy - saving optimization impact on the historical energy consumption data of equipment based on the energy - saving optimized instruction set and the dynamic energy consumption impact factor matrix to generate an energy - saving optimization impact factor data set;

[0064] Based on the standardized equipment historical energy consumption data set and the energy - saving optimization impact factor data set, perform benchmark energy consumption prediction modeling through a preset regression algorithm to obtain an equipment benchmark energy consumption prediction model;

[0065] Cross-validate the equipment benchmark energy consumption prediction model, iteratively adjust the model parameters, and obtain the benchmark energy consumption prediction model;

[0066] The energy-saving optimization instruction set is input into the baseline energy consumption prediction model to obtain the equipment baseline energy consumption prediction data.

[0067] The present invention can effectively improve the energy efficiency of buildings and optimize the energy consumption control strategy of equipment. First, the environmental data and equipment status information obtained by multimodal sensors are combined with non-Euclidean topological mapping methods to model the spatiotemporal relationship between equipment and environment, and generate a high-precision spatiotemporal fusion feature set. This provides an accurate physical and semantic mapping basis for subsequent energy consumption optimization, so that the energy consumption optimization process can better reflect the real interaction and change of equipment and environment. Then, through cross-domain energy consumption influencing factor screening and associated topological modeling, a dynamic energy consumption influencing factor matrix is ​​obtained, which can help accurately quantify the specific impact of various factors (such as temperature and humidity changes, equipment load, etc.) on building energy consumption, and then provide a scientific basis for the construction of building energy consumption evolution map. Through multi-scale energy consumption evolution deduction, not only can the short-term fluctuations of building energy consumption be predicted, but also scientific predictions can be provided for long-term energy consumption trends, thereby providing a basis for the formulation of energy-saving optimization strategies. Through the constructed energy consumption propagation network and time series analysis, the building energy consumption can be carefully decomposed and analyzed, thereby improving the understanding of the building energy consumption evolution process. Based on these analysis results, energy-saving optimization instruction sets can be generated through multi-objective energy-saving control strategy manifolds, and uploaded to the building energy consumption control platform for execution in real time, realizing personalized equipment energy consumption control. This process helps to make dynamic adjustments according to the needs of different equipment and environments, thereby improving building energy efficiency and ensuring that the operation of various types of equipment is more in line with energy-saving requirements. Through the technology of equipment predicted energy consumption deviation lineage tracking, when the real-time energy consumption data deviates from the predicted value, the anomaly can be quickly identified and traced back to the source analysis, so as to find out the root cause of the energy consumption anomaly and make timely adjustments. The combination of this energy consumption deviation measurement matrix and the energy consumption change causal relationship diagram can effectively enhance the system's control ability over equipment energy consumption fluctuations and further improve the accuracy and flexibility of energy consumption management. Finally, based on the generation of energy-saving optimization influencing factor data sets, it is possible to accurately quantify the impact of energy-saving optimization instruction sets, provide data support for the equipment benchmark energy consumption prediction model, ensure the accuracy of equipment energy consumption prediction, and improve the reliability of the system. In summary, the present invention, through the combination of multimodal data collection, dynamic energy consumption impact quantification, cross-domain energy consumption analysis and multi-objective energy-saving optimization strategy, can not only accurately predict and optimize building energy consumption, but also respond to changes in building energy efficiency demand in real time, and ultimately achieve maximization of building energy efficiency and optimization of equipment energy consumption control. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Other features, objectives, and advantages of the present invention will become more apparent from the following detailed description of non - restrictive embodiments read in conjunction with the accompanying drawings:

[0069] Figure 1 It is a schematic diagram of the step - by - step process of the artificial - intelligence - based building energy - consumption optimization method of the present invention;

[0070] Figure 2 It is a detailed step - by - step process schematic diagram of step S1 in the present invention;

[0071] The realization of the objectives of the present invention, its functional characteristics, and advantages will be further described with reference to the embodiments and the accompanying drawings. Detailed Embodiments

[0072] The following clearly and completely describes the technical method of the present invention with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0073] In addition, the accompanying drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and thus repeated descriptions of them will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0074] It should be understood that although terms such as "first", "second", etc. may be used here to describe each unit, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit can be called the second unit, and similarly, the second unit can be called the first unit. The term "and / or" used here includes any and all combinations of one or more of the listed related items.

[0075] To achieve the above - mentioned objective, please refer to Figures 1 to 2 , the present invention provides an artificial - intelligence - based building energy - consumption optimization method, and the method includes the following steps:

[0076] Step S1: Obtain multi - modal sensing data and construct a heterogeneous sensing dynamic spatio - temporal adjacency matrix; perform device physical - space semantic mapping modeling based on the heterogeneous sensing dynamic spatio - temporal adjacency matrix to obtain a sensing spatio - temporal fusion feature set;

[0077] In this embodiment, real-time data is collected in the building through multiple multimodal sensor devices (such as infrared sensors, ultrasonic sensors, light sensors, temperature and humidity sensors, etc.). Taking the collection time interval Δt = 0.1 second as an example, data on the operating states of building equipment (such as the states of fans, air conditioners, lighting equipment) and environmental parameters (such as temperature, humidity, carbon dioxide concentration, etc.) is obtained. These sensor data are first processed by noise removal, and the Kalman filtering algorithm is used to remove the noise components in the sensor readings. At the same time, missing data is interpolated, and the nearest neighbor interpolation or spline interpolation algorithm is used for filling to ensure data integrity. Then, time synchronization is performed to normalize all sensor data to a unified time step to ensure data consistency. Next, according to the spatio-temporal characteristics of the data, a heterogeneous sensing dynamic spatio-temporal adjacency matrix is constructed through a graph neural network (GNN). Each node in the matrix represents a device or an environmental parameter, the edges represent the degree of association between different devices or sensors, and the weights of the edges are calculated based on the spatial and temporal associations between the devices. Finally, based on this adjacency matrix, semantic mapping is performed in the three-dimensional building physical space, mapping the operating states of the devices and the environmental parameters to specific positions in the physical space, and extracting a set of sensing spatio-temporal fusion features as the input for the subsequent steps.

[0078] Step S2: Perform cross-domain energy consumption impact mapping topology modeling on the set of sensing spatio-temporal fusion features to obtain a cross-domain energy consumption correlation topology graph; extract the device energy transfer impact factors according to the cross-domain energy consumption correlation topology graph to obtain a dynamic energy consumption impact factor matrix;

[0079] In this embodiment, based on the constructed set of sensing spatio-temporal fusion features, cross-domain data collaborative modeling is performed on multiple devices in the building. At the local device end of each building area, the sensing data is first denoised, normalized, and feature extracted, and then learned and summarized through a weighted average model and a global model to ensure data privacy. According to the learning results of these device ends, a graph neural network (GNN) is used to map the energy consumption impact factors. Each node in the graph represents a building device, and the edges between the nodes represent the energy consumption transfer relationship between different devices. By calculating the spatio-temporal correlation degree between the devices and setting a maximum connection threshold T = 0.6, nodes with low correlation degrees are screened out, thereby constructing a cross-domain energy consumption correlation topology graph. Based on this topology graph, the energy transfer impact factor of each device is calculated, and the least squares method (LSE) is used to estimate the contribution degree of the device to the overall energy consumption. The dynamic energy consumption impact factor matrix is formed by the energy transfer rate and transfer relationship between the devices, and the specific rate is set between 0.05 and 0.2 to reflect the change of energy consumption between the devices.

[0080] Step S3: Perform multi-scale energy consumption evolution deduction based on the dynamic energy consumption impact factor matrix to generate a building energy consumption evolution map; construct an energy consumption probability cloud map based on the building energy consumption evolution map;

[0081] In this embodiment, based on the dynamic energy consumption impact factor matrix generated in the previous step, a long short-term memory network (LSTM) is used to model and predict historical building energy consumption data. To better capture the short-term energy consumption change trend, a 24-hour time sliding window is adopted for data analysis and short-term trend modeling. Empirical mode decomposition (EMD) is used to perform multi-scale decomposition on the building energy consumption data to extract the energy consumption patterns at different time scales. This step generates a building energy consumption evolution map, showing the long-term and short-term energy consumption evolution processes of each device. Based on the energy consumption evolution map, Gaussian kernel density estimation (KDE) is used to calculate the energy consumption probability distribution of the device, and then an energy consumption probability cloud map is generated to reflect the energy consumption fluctuations of each device in the building under different environmental and usage conditions.

[0082] Step S4: Perform adversarial strategy optimization on the energy consumption probability cloud map to obtain a multi-objective energy-saving control multi-objective energy-saving control strategy manifold; generate an energy-saving optimization instruction set based on the multi-objective energy-saving control strategy manifold and upload it to the building energy consumption control platform to control the devices;

[0083] In this embodiment, based on the generated energy consumption probability cloud map, a generative adversarial network (GAN) is used to perform adversarial optimization on the energy-saving control strategy. To improve the robustness of the energy-saving strategy, the number of training rounds is set to 1000 rounds, and through adversarial training, the model can adaptively adjust the energy-saving control strategy of the building devices. During this optimization process, for HVAC devices, an energy-saving control objective function is set, including controlling the deviation of the indoor temperature to be less than ±0.5°C, the humidity to be controlled within ±2%, and reducing the device energy consumption by at least 10%. The optimized energy-saving strategy manifold is adjusted through a multi-objective optimization algorithm (such as NSGA-II) and confirmed through consistency verification. The verified energy-saving optimization instruction set is uploaded to the building energy consumption control platform through the MQTT protocol, and the instruction set will automatically execute the device control tasks, including adjusting the operating states of devices such as air conditioners and lighting.

[0084] Step S5: Obtain real-time building energy consumption data, perform device predicted energy consumption deviation blood relationship tracking on the real-time building energy consumption data according to the energy-saving optimization instruction set, and implement device control local comfort instruction optimization to obtain a personalized control instruction set and upload it to the building energy consumption control platform to control the devices.

[0085] In this embodiment, real-time energy consumption data of a building is obtained through a building energy consumption control platform, and the baseline energy consumption of equipment is predicted in combination with an energy-saving optimization instruction set. The predicted equipment energy consumption is compared with the real-time building energy consumption data through deviation measurement. The Markov hidden state model is used to analyze the deviation trend of equipment energy consumption, and the deviation threshold is set at 5%. Possible abnormal equipment is screened out. In combination with the dynamic energy consumption impact factor matrix, Granger causality analysis (GCA) is used to trace the origin of the energy consumption deviation, analyze which equipment operating states have caused abnormal fluctuations in energy consumption, and generate an energy consumption deviation retrospective analysis report. The report details the abnormal energy consumption patterns of the equipment and the corresponding impact factors. According to the analysis report, the energy-saving control strategy of the equipment will be further optimized. For the control of specific equipment such as air conditioners, it is adjusted differentially according to the personnel distribution, and the comfort range is automatically adjusted to 22°C to 26°C, and the brightness of lighting equipment is adjusted to keep it between 300 lux and 500 lux. The optimized personalized control instruction set will be uploaded to the building energy consumption control platform to automatically execute equipment control tasks and ensure that the building energy consumption is effectively controlled within a reasonable range.

[0086] Optionally, step S1 is specifically as follows:

[0087] Step S11: Multimodal sensing data is obtained through a multimodal sensor group, where the multimodal sensing data includes temperature and humidity sensing data, light sensing data, infrared sensing data of the crowd density, and equipment status sensing data;

[0088] In this embodiment, multiple multimodal sensor devices (such as infrared sensors, ultrasonic sensors, light sensors, temperature and humidity sensors, etc.) are responsible for collecting different types of data. Among them, the temperature and humidity sensors monitor the changes in indoor temperature and humidity in real time, the light sensors capture the changes in ambient light intensity, the infrared sensors monitor the crowd density, and the equipment status sensors obtain the working status of the equipment in the building (such as air conditioners, lights, fans, etc.). The collection frequency is set to once per second (the sampling frequency is 1 Hz) to ensure the real-time and integrity of the data. The collected data undergoes preliminary preprocessing to remove outliers and noise data, and is normalized, converting all data to the standard interval from 0 to 1 to ensure the compatibility and comparability of different sensor data.

[0089] Step S12: Perform time-step time series alignment on the multimodal sensing data to obtain a time series aligned environment data set;

[0090] In this embodiment, the multi-modal data obtained from each sensor is aligned according to the timestamp. Since there may be differences in the acquisition frequencies and time sequences of each sensor, it is necessary to adjust the time step. For this purpose, a method combining linear interpolation and nearest neighbor interpolation is adopted to ensure that the data time step is unified and no information is lost. For example, for the temperature and humidity and light sensor data, a unified time step of 0.1 second is set, that is, a data point is generated every 0.1 second, and the interpolation method is used to fill in the missing time step data. The aligned data set will become the basis for subsequent analysis and modeling, ensuring the consistency and timeliness of the data.

[0091] Step S13: Perform non-Euclidean topological mapping on the temporally aligned environmental data set, and construct an initial spatial semantic topological structure diagram based on the topological mapping result;

[0092] In this embodiment, based on the temporally aligned environmental data set, a non-Euclidean graph (Non-Euclidean Graph) is used for topological mapping, and a graph neural network (GNN) is used to model the relationships between the data. Each sensor serves as a node in the graph, and different sensors are connected by edges, and the weights of the edges represent the association strength between them. For the relationship between the device state and the environmental data, by calculating the Euclidean distance and correlation, a weight threshold of 0.5 is set, the associated edges with larger weights are retained, and the redundant connections are removed. Through this non-Euclidean topological mapping, an initial spatial semantic topological structure diagram of each device and environmental factors in the building space is generated for subsequent neighborhood optimization and feature fusion.

[0093] Step S14: Perform neighborhood adaptive topological optimization based on the initial spatial semantic topological structure diagram to obtain a device spatial semantic graph, and fuse the device operation mode and environmental data based on the device spatial semantic graph to obtain a heterogeneous sensing dynamic spatio-temporal adjacency matrix;

[0094] In this embodiment, based on the generated initial spatial semantic topological structure diagram, a graph adaptive algorithm is used to optimize the topology. The algorithm adaptively adjusts the neighborhood relationship between devices according to the actual operation conditions of the building devices and sensors, optimizes the topological structure, so that the spatial semantic graph can better reflect the physical spatial relationship between devices. For example, if a device is frequently used during a certain period of time and has a strong correlation with the energy consumption of adjacent devices, the connection strength between them will automatically increase. The optimized device spatial semantic graph further fuses the device operation mode and environmental data, and generates a more representative heterogeneous sensing dynamic spatio-temporal adjacency matrix through the weighted average method. This matrix not only contains the physical position relationship between devices, but also can reflect the dynamic energy consumption interaction between devices and the environment.

[0095] Step S15: Establish a device physical space semantic mapping model based on the heterogeneous sensing dynamic spatio-temporal adjacency matrix, perform feature screening and correlation optimization on the model establishment results, and obtain a sensing spatio-temporal fusion feature set.

[0096] In this embodiment, based on the heterogeneous sensing dynamic spatio-temporal adjacency matrix, a device physical space semantic mapping model is established. Through the deep learning ability of the graph neural network, this model can map device and environmental data to the physical space of the building, constructing a more accurate correlation graph of device location and energy consumption. To improve the accuracy and robustness of the model, feature screening and correlation optimization techniques are adopted. First, the principal component analysis (PCA) method is used to reduce the dimension of the sensor data and reduce redundant features; secondly, device features with a greater impact on energy consumption are screened out through correlation analysis (such as Pearson correlation coefficient). Finally, combined with the space semantic mapping model, a set of high-quality sensing spatio-temporal fusion feature sets are obtained. These features can better reflect the relationship between device energy consumption and environmental changes, providing important input data for the subsequent formulation and optimization of energy-saving strategies.

[0097] Optionally, step S13 is specifically as follows:

[0098] Step S131: Obtain the operation data of the building area equipment, and perform synchronous time series alignment on the operation data of the building area equipment and the time series aligned environmental data set to obtain time series aligned equipment operation data;

[0099] In this embodiment, the operation data of all equipment in the building area is obtained through the equipment monitoring system, including the real-time operation status of equipment such as air conditioners, lighting, heating and ventilation systems (HVAC), fans, and pumps. The operation data of all equipment is sampled at a frequency of once per second and collected through standard network protocols (such as Modbus or BACnet). After obtaining this equipment data, it is synchronously time series aligned with the aligned environmental data set. The environmental data set includes temperature, humidity, light, personnel density, etc., and the sampling time step is set to 5 seconds. For the collected equipment operation data and environmental data, missing data is supplemented through interpolation methods (such as linear interpolation or spline interpolation) to ensure that all data is strictly aligned in time, thereby obtaining complete time series aligned equipment operation data. The key to this step is to ensure the precise synchronization of all equipment and environmental data in time series, laying a foundation for subsequent analysis.

[0100] Step S132: Perform non-Euclidean topological mapping on the time series aligned equipment operation data and the time series aligned environmental data set, construct a spatio-temporal relationship map between the equipment and environmental variables in the building area, and obtain the building area spatio-temporal relationship map;

[0101] In this embodiment, a topological mapping method of non-Euclidean graphs is used to analyze the operation data and environmental data of the time series alignment device. Each device and environmental variable is regarded as a node in the graph, and the relationship between the device operation state and the environmental variable is represented by the edges of the graph. A graph neural network (GNN) model is used to capture the spatio-temporal dependence relationship between these devices and environmental variables, and the weights of the edges are dynamically adjusted based on the correlation and influence of the devices and environmental variables. For example, when there is a strong correlation between the operation state of the air conditioning device and the indoor temperature change, the weight between the two is large; otherwise, the weight is small. In this way, a spatio-temporal relationship map between the devices and environmental variables in the building area is constructed. This map can reveal the complex interaction relationship between the devices and the environment, providing data support for subsequent optimization decisions.

[0102] Step S133: Perform physical space interaction modeling based on the spatio-temporal relationship map of the building area, map the results of the physical space interaction modeling to a preset building physical space coordinate system, convert the nodes in the spatio-temporal relationship map into actual physical space coordinate nodes, and construct device-environment physical space interaction data;

[0103] In this embodiment, based on the spatio-temporal relationship map of the building area, the positions of the devices (such as the installation positions of HVAC devices and lighting devices) and their impacts on the surrounding environment are analyzed. For example, the air conditioning device is located in a certain room, and the temperature change in that room will directly affect the energy efficiency performance of the air conditioning device. A model is established for the devices and environmental variables (such as temperature and humidity) through physical space relationships and mapped to the actual physical coordinate system of the building. Through the physical space coordinate system, each node in the spatio-temporal relationship map can be converted into an actual position in the building floor plan, making the relationship between each device and environmental variable more intuitive. For example, the positions of the air conditioner and the temperature and humidity sensors will directly affect their energy efficiency performance, so their spatial coordinates are accurately mapped onto the building floor plan.

[0104] Step S134: Perform node spatial layout analysis on the device-environment physical space interaction data, and map the interaction relationship between the devices and the environment according to the results of the node spatial layout analysis to obtain an environmental physical space interaction diagram;

[0105] In this embodiment, the device-environment physical space interaction data is optimized through node space layout analysis. First, the Voronoi diagram or spatial clustering algorithm is used to analyze the spatial layout of devices in the building area. The interaction relationship between the device and the environment is mapped between nodes, and the energy efficiency impact of the device is analyzed through the spatial relationship of the nodes (such as the relative distance between the device and the temperature and humidity sensor, and the distribution of the air conditioner and the lighting control device). The energy efficiency of the device is usually reflected by its impact on the environment. For example, the air conditioner directly affects the environment by adjusting the indoor temperature and humidity. If the distance between the air conditioner and the environmental sensor (such as the temperature and humidity sensor) is relatively close, its control of the environment is more accurate and efficient. Therefore, the energy efficiency of the air conditioner has a greater impact on the environment, especially in energy conservation and load optimization. Assume that the spatial distance between the air conditioner device and the temperature and humidity sensor in a certain area is relatively close, then the interaction relationship between the two is strong, the impact on the environment is large, and the energy efficiency impact is large. Through spatial layout analysis, the strong interaction relationships between devices are identified, providing a basis for subsequent energy efficiency optimization. In this step, the algorithms used also include spatial weighted analysis to ensure that the interaction between the device and the environment is accurately reflected. Finally, an environmental physical space interaction diagram is generated, which clearly indicates which devices have a close relationship with the environment and which devices have a greater impact on the environment during the energy efficiency optimization process.

[0106] Step S135: Perform graph optimization processing based on the environmental physical space interaction diagram, and construct an initial spatial semantic topology structure diagram.

[0107] In this embodiment, based on the environmental physical space interaction diagram, an optimization algorithm in graph theory is used to optimize the graph, aiming to improve the expression ability of the graph and reduce the computational complexity. The key to graph optimization is to screen out the device-environment interaction relationships with the greatest impact on energy efficiency through algorithms (such as the minimum spanning tree and the maximum flow algorithm), and delete redundant low-impact nodes. For example, if the interaction relationship between certain devices and the environment is determined to have a small impact during the optimization process (such as a long distance and a weak impact), these nodes can be deleted in the optimized graph to simplify the calculation. The optimized graph will be able to more accurately display the main interaction relationships between the device and the environment, providing a basis for subsequent energy efficiency optimization decisions. Finally, an initial spatial semantic topology structure diagram is generated, which can clearly express the core interaction relationships between the devices and the environment in the building area and provide support for subsequent energy conservation optimization.

[0108] Optionally, step S14 is specifically as follows:

[0109] Step S141: Calculate the local density of the topological nodes according to the initial spatial semantic topology structure diagram, set the calculation neighborhood radius to 2, set the local density threshold to 0.1, and the nodes below the local density threshold are regarded as sparse areas, so as to obtain the local density distribution of the topological structure;

[0110] In this embodiment, when calculating the local density of topological nodes according to the initial spatial semantic topology structure diagram, the calculation neighborhood radius is set to 2, which means that each node in the initial spatial semantic topology structure diagram will consider the neighborhood nodes within a distance of no more than 2 nodes from it. The local density of each node is calculated through this radius, and the local density threshold is set to 0.1. Nodes below this threshold are regarded as sparse regions. The purpose of this step is to identify the sparse regions in the topological structure, and these sparse regions represent the regions with less impact on energy efficiency or lower environmental impact in the system. By calculating the local density of nodes, the regions that need to be optimized with emphasis can be effectively screened out, and the regions with lower density in the system can be determined, which is of great significance for energy efficiency improvement. In actual operation, a graph analysis library such as NetworkX in Python can be used to implement the calculation of local density, considering the neighborhood information of nodes and generating the corresponding local density distribution diagram. Assuming that the initial topology graph has 1000 nodes, the local_density() function will check 2 nodes around each node, calculate its density, and mark the nodes with local density lower than 0.1 as sparse regions. For example, in the ventilation system of a building, there are only a small number of devices in some areas, and the calculated density values are lower than the set threshold of 0.1. These areas are the sparse regions with low energy efficiency and are worthy of further optimization.

[0111] Step S142: Set the minimum sample number to 5, the density threshold to 0.3, perform node neighborhood adaptive clustering on the initial spatial semantic topology structure diagram, and set the weight range to [0,1]. Calculate the neighborhood adaptability weight based on the clustering result to generate the initial parameters for neighborhood adaptive topology optimization;

[0112] In this embodiment, setting the minimum sample number to 5 means that when performing clustering, the minimum number of nodes in each cluster cannot be less than 5. Setting the density threshold to 0.3 means that in the clustering process, only the regions with density higher than 0.3 in the neighborhood will be regarded as a cluster. According to these settings, the topological graph will be adaptively divided by using the density peak clustering algorithm to find the regions with similar characteristics and divide them into the same category. The calculated similarity between nodes (such as Euclidean distance or Manhattan distance) is used to determine the neighborhood adaptability weight, with a range of [0,1]. The higher the weight of a node, the closer its relationship with other neighborhood nodes, and it has a higher priority for optimizing energy efficiency and environmental impact. Through this step, the initial parameters for neighborhood adaptive topology optimization can be obtained, providing a basis for subsequent multi-scale optimization. The clustering result can be visualized on the physical space layout of the building.

[0113] Step S143: Set the weight adjustment coefficient to 0.5, adjust the neighborhood topology weights of the initial spatial semantic topology structure diagram according to the neighborhood adaptive topology optimization initial parameters, and set the optimization scale [1, 3] to perform multi-scale topology structure optimization to obtain an optimized device spatial semantic map;

[0114] In this embodiment, in the neighborhood topology weight adjustment stage, setting the weight adjustment coefficient to 0.5 means that the neighborhood weights of each node will be appropriately adjusted according to the influence of adjacent nodes during the optimization process. The process of neighborhood topology weight adjustment is to perform multi-scale topology structure optimization within the scale range of 1 to 3 by setting the optimization scale to [1, 3], so that the optimized graph can more accurately reflect the relationship between nodes and its impact on energy efficiency. This optimization process can be carried out algorithmically through the Weighted Shortest Path Algorithm, which uses the distance and correlation between nodes to adjust the weights, thereby reducing redundant calculations and optimizing the nodes with greater impact on energy efficiency. The weighted shortest path algorithm can calculate the "distance" between each node, and here the distance can be defined by the energy consumption, workload or other relevant metrics between devices. Nodes with higher correlation (such as the relationship between an air conditioner and a temperature and humidity sensor) will be assigned higher weights to optimize the energy efficiency of these nodes. This optimization method can efficiently integrate the interaction relationships between various devices and the environment in the building to ensure that the optimized graph more accurately reflects the key areas for energy efficiency improvement.

[0115] Step S144: Extract device operation mode features with a feature dimension of 5 based on the optimized device spatial semantic map, fuse the time-series aligned environmental data set, and perform normalization processing on the device operation mode features to generate standardized device operation mode data;

[0116] In this embodiment, the device operation mode data is standardized. The feature dimension is set to 5, that is, 5 important device operation features are selected. For example: the power consumption of air-conditioning equipment, indoor temperature, readings of humidity sensors, status of lighting equipment, and personnel density, etc. Through these features, the interaction relationship between the device operation status and environmental variables can be understood in detail. The device operation mode features are fused with the time-series aligned environmental data. This fusion process is carried out by a weighted average method based on data correlation. The operation mode features of each device (such as the power consumption of the air conditioner, the on / off status of the lighting, etc.) will be assigned different weights according to their performance under specific environmental conditions (such as temperature and humidity changes). For example, when the indoor temperature is relatively high, the energy consumption feature of the air conditioner will be higher. Therefore, the correlation between the air conditioner and temperature is stronger, and the corresponding weight is also larger. In this way, the fusion of the device operation mode and environmental data can accurately reflect the energy efficiency status of the device under different environmental conditions. During the standardization process, the Z-score standardization method is adopted, so that the data values of each feature are converted into a distribution with a mean of 0 and a standard deviation of 1, which can avoid the deviation of features with different dimensions from affecting the model. Suppose the power consumption of the air-conditioning equipment is 0.5 after standardization, indicating that the operation status of the device is relatively normal, while the value of the device with large temperature and humidity changes is -1.2 after standardization, indicating that the energy efficiency of the device is unstable. Through these standardized feature data, more accurate and consistent data can be provided for the next step of modeling.

[0117] Step S145: Embed the standardized device operation mode data into the topological structure of the optimized device space semantic graph, perform interactive modeling of the device operation mode and environmental data, and obtain a preliminary heterogeneous sensing dynamic spatio-temporal adjacency matrix;

[0118] In this embodiment, the standardized device operation mode data is embedded into the optimized device space semantic graph to perform interactive modeling between the device and the environment. To achieve this, the device nodes and environmental variable nodes are combined, and a graph neural network (GNN) model is used to learn the interaction relationship between the device and the environment. Suppose the air-conditioning equipment, temperature and humidity sensors, and lighting equipment are selected as the input nodes of the model. During the model training process, the edge weights between the air-conditioning equipment node and the temperature and humidity sensor nodes will be adjusted according to their interaction intensity. For example, if the air-conditioning equipment operates in a low-temperature environment and its temperature regulation effect is small, the edge weight between the air-conditioning node and the temperature and humidity sensor node will be small. In this way, an interaction model between the device operation mode and environmental data can be constructed, and a preliminary heterogeneous sensing dynamic spatio-temporal adjacency matrix is generated.

[0119] Step S146: Perform feature screening and topological correlation optimization on the preliminary heterogeneous sensing dynamic spatio-temporal adjacency matrix, strengthen the key interaction relationships, and remove the edge weights with low correlation degrees, so as to generate a heterogeneous sensing dynamic spatio-temporal adjacency matrix.

[0120] In this embodiment, in the feature screening and topological correlation optimization stage, a graph analysis method is used to optimize the preliminary heterogeneous sensing dynamic spatio-temporal adjacency matrix. By removing the edge weights with low correlation degrees and strengthening the key interaction relationships between nodes, a final heterogeneous sensing dynamic spatio-temporal adjacency matrix is generated. In this step, a feature selection algorithm (such as the L1 regularization method or the principal component analysis PCA) is used to perform feature screening on the matrix, removing the nodes and edges with less influence on energy efficiency optimization, thereby reducing the computational complexity. The finally generated heterogeneous sensing dynamic spatio-temporal adjacency matrix will more accurately reflect the key interaction relationships between the devices and the environment in the building, providing more effective data support for subsequent energy efficiency optimization. This process helps to extract the most influential relationships between the devices and the environment, ensuring that the optimization work focuses on the most critical interaction areas.

[0121] Optionally, step S2 is specifically as follows:

[0122] Step S21: Screen the cross-domain energy consumption impact factors for the sensing spatio-temporal fusion feature set, set the correlation threshold to [0.7, 0.8] to retain the features with high correlation, calculate the energy consumption impact weights of each device, and construct an initial energy consumption impact feature matrix based on the energy consumption impact weights;

[0123] In this embodiment, screening the energy consumption impact factors for the sensor spatio-temporal fusion feature set aims to select the parts that have the most influence on energy consumption from a large number of features. For this purpose, a correlation threshold range [0.7, 0.8] is set, and the correlation analysis is performed on the spatio-temporal feature data of all devices. Only when the feature correlation between two devices exceeds 0.7 and is less than 0.8, these features are retained. For example, there may be a strong correlation between the power consumption of air conditioning equipment and the indoor temperature and humidity, and such features with high correlation are selected and retained. Next, the energy consumption impact weights of each device are calculated according to the screened features. The higher the weight, the greater the impact of the device on the overall energy consumption. Specifically, the weight calculation uses weighted regression analysis based on the energy efficiency performance of each device in historical operation and its correlation with environmental factors. After completing the feature screening and weight calculation, an initial energy consumption impact feature matrix is constructed based on the weights. This matrix includes the energy efficiency impact degrees of each device and their mutual relationships.

[0124] Step S22: Perform a non-Euclidean graph topological mapping on the initial energy consumption impact feature matrix, set the maximum connection threshold of the adjacency matrix to 0.5, construct an energy consumption correlation topological structure across devices and spatial regions, and generate an initial cross-domain energy consumption correlation topological graph;

[0125] In this embodiment, the initial energy consumption impact feature matrix is mapped into a non-Euclidean graph topology. Here, "non-Euclidean graph topology" means that instead of using the traditional Euclidean space, devices are connected based on the energy consumption correlation and spatial layout between them. For example, if the correlation degree of energy efficiency between an air conditioner device and a temperature and humidity sensor device is greater than 0.5, a connection will be created between these two devices in the topology graph. To describe the complex energy consumption relationships between devices, the maximum connection threshold of the adjacency matrix is set to 0.5. When the energy consumption impact degree of two devices is greater than 0.5, they are considered to have a strong correlation, and an edge is created in the graph. For example, assume that there is a strong correlation in energy consumption between an air conditioner device and a temperature and humidity sensor (correlation coefficient is 0.6), while the correlation degree between the air conditioner and a light sensor is weak (correlation coefficient is 0.4), then only a connection will be formed between the air conditioner and the temperature and humidity sensor, and the connection with the light sensor will be ignored. In this way, the energy consumption data of different devices and spatial regions is transformed into a graphical structure, with devices as nodes and energy consumption transfer relationships as edges, constructing a preliminary cross-device and cross-spatial region energy consumption correlation topology graph. This provides a basis for the next spectral clustering analysis.

[0126] Step S23: Perform spectral clustering analysis on the initial cross-domain energy consumption correlation topology graph, select the first 20 eigenvectors for Laplacian eigen-decomposition, and set the eigen-truncation threshold to 0.1 to extract the main energy consumption impact patterns, generating a cross-domain energy consumption correlation topology graph;

[0127] In this embodiment, when performing spectral clustering analysis on the initial cross-domain energy consumption correlation topology graph, the first 20 eigenvectors are selected for Laplacian eigen-decomposition. The purpose of this process is to extract the most representative energy consumption impact patterns. The selection of eigenvectors is based on the significance of energy consumption patterns, and the first 20 eigenvectors represent the regions and devices with the greatest energy consumption impact. During this process, the eigen-truncation threshold is set to 0.1, which means that when the contribution degree of some features is lower than 0.1, these features will be excluded. In this way, noise and irrelevant features can be removed, and only the energy consumption impact patterns that are most helpful for energy efficiency optimization are retained. The generated cross-domain energy consumption correlation topology graph can more accurately reflect the energy efficiency interdependence relationships between various devices in the building and provide accurate data support for subsequent optimization steps.

[0128] Step S24: Based on the cross-domain energy consumption correlation topology graph, model the energy consumption transfer relationship between devices, set the energy consumption transfer rate to [0.05, 0.2], and set the dynamic weight change interval to [0, 1] to allocate dynamic weights, generating an initial dynamic energy consumption impact factor matrix;

[0129] In this embodiment, a cross-domain energy consumption correlation topology map is used to model the energy consumption transfer relationship between devices. By setting the energy consumption transfer rate to [0.05, 0.2], the transfer efficiency of energy consumption between different devices can be simulated. For example, the air conditioning equipment will directly affect the energy consumption of the temperature and humidity sensors, while the temperature and humidity sensors will affect the operating load of the air conditioner. By setting the range of the energy consumption transfer rate from 0.05 to 0.2, different situations of this transfer process can be simulated. Then, a dynamic weight change interval of [0, 1] is set, which enables the energy consumption transfer relationship to be dynamically adjusted under different working conditions. For example, when the external temperature rises sharply, the weight of the energy consumption impact of the air conditioner will increase, while the impact of other devices such as lighting will decrease. Through this dynamic weight adjustment mechanism, a more accurate dynamic energy consumption impact factor matrix can be obtained, further reflecting the energy consumption transfer relationship between devices.

[0130] Step S25: Normalize the initial dynamic energy consumption impact factor matrix, set the confidence interval of the impact factor to [0.95, 1.05], correct the robustness of the impact factor, and optimize the energy consumption impact parameters, and output the dynamic energy consumption impact factor matrix.

[0131] In this embodiment, the initial dynamic energy consumption impact factor matrix is normalized to ensure that all impact factors are within a reasonable range. For example, the confidence interval is set to [0.95, 1.05] to ensure that the fluctuation of the energy consumption impact factor within this interval does not exceed the predetermined value. If the energy efficiency factor of a certain device exceeds this range, such as an abnormal change in energy efficiency (such as a sharp increase), the energy efficiency of the device is considered abnormal and further adjustment is carried out (including data smoothing, recalibration, and abnormal data rejection or correction). For example, sudden changes in environmental factors such as external temperature and humidity lead to a sharp change in the energy efficiency of the device. In this case, compensation for environmental parameters is required to ensure the stability and accuracy of the model. The sensor itself may malfunction or there may be a deviation in data input, resulting in inaccurate energy consumption data collected. In this case, the energy consumption impact factor will lose credibility and needs to be adjusted by data cleaning or recalibrating the sensor. For example, if the energy consumption factor of a certain device exceeds 1.05 at a certain moment, it means that the energy consumption performance of the device is much higher than the normal range, which may be due to device failure, external environmental changes, or a decrease in device operating efficiency. If it is lower than 0.95, it may indicate that the energy consumption performance of the device is abnormally low, which may be due to abnormal device operation or some sensors not working properly, and early warning work needs to be carried out to report the faulty device. By normalizing and adjusting the impact factor, the robustness of the energy consumption matrix can be enhanced, making it more stable and accurate, and providing reliable data support for optimization decisions. Finally, through this optimized matrix, the system can allocate resources more efficiently, improve the overall energy efficiency performance, and ensure the stability and energy-saving effect of the system operation under different environmental conditions and device loads.

[0132] Optionally, the multi-scale energy consumption evolution deduction described in step S3 is specifically as follows:

[0133] Perform time series decomposition on the dynamic energy consumption impact factor matrix, set multi-scale time steps, and perform trend decomposition and periodic decomposition to generate an energy consumption time series decomposition data set;

[0134] In this embodiment, performing time series decomposition on the dynamic energy consumption impact factor matrix aims to extract energy consumption impact patterns at different time scales. For this purpose, multi-scale time steps (such as 5 minutes, 1 hour, 12 hours, and 24 hours) are selected for decomposition. These time scales can adapt to energy consumption change requirements at different granularities. For example, a 5-minute time step is suitable for capturing minute fluctuations in device operation, and a 24-hour time step is used to analyze daily energy consumption change patterns. The TrendDecomposition method is used to remove the long-term trend and extract the periodic energy consumption changes. SeasonalDecomposition helps identify daily or seasonal energy consumption patterns and extract the periodic fluctuation part. Finally, the obtained decomposition data set is used for subsequent analysis, such as causal inference of the energy consumption propagation path.

[0135] Perform causal inference on the energy consumption propagation path of the energy consumption time series decomposition data set, calculate the energy consumption correlation strength between devices, and construct an energy consumption propagation network to generate an initial energy consumption propagation structure;

[0136] In this embodiment, causal inference analysis is performed on the energy consumption time series decomposition data set. By using the Granger Causality Test, it is possible to detect whether there is a causal propagation relationship of energy between devices. For example, if the energy consumption fluctuation of device A causes the energy consumption fluctuation of device B, it can be considered that device A has a causal impact on the energy consumption of device B. Based on this causal inference result, calculate the energy consumption correlation strength between each pair of devices (such as measured by the Pearson correlation coefficient). Then, convert these correlation strengths into an energy consumption propagation network. By setting a threshold (such as 0.7), edges with a strength lower than this value are removed, thereby constructing a preliminary energy consumption propagation network structure.

[0137] Perform energy consumption propagation pattern clustering based on the initial energy consumption propagation structure, screen the energy consumption propagation paths, remove low-contribution energy consumption paths, and generate an optimized energy consumption propagation network;

[0138] In this embodiment, pattern clustering analysis is performed based on the initial energy consumption propagation structure. The spectral clustering algorithm is used to cluster the energy consumption propagation paths to identify device groups with similar propagation characteristics. By analyzing the clustering results, high-contribution propagation paths are screened out. For example, if the energy consumption fluctuations of certain paths can significantly affect the overall energy efficiency level, they are retained; while those paths with small contributions and weak propagation effects are removed. After this optimization step, a simplified and efficient energy consumption propagation network is obtained, which can more accurately reflect the energy consumption flow pattern between devices.

[0139] Time series feature extraction is performed on the optimized energy consumption propagation network, the energy consumption time series segments are divided, and empirical mode trend decomposition is performed on the energy consumption time series segments to obtain energy consumption time series feature data;

[0140] In this embodiment, the optimized energy consumption propagation network enters the time series feature extraction stage. First, the energy consumption time series segments are divided according to time periods, such as daily, weekly, and monthly data. Then, the empirical mode decomposition (EMD) method is used to decompose the energy consumption data of each time period to extract different feature patterns (such as trends, periodic components, noise, etc.) of the time series data. This method can adaptively extract the key features in the signal without relying on traditional statistical models. Each time series segment is decomposed into multiple intrinsic mode functions (IMFs), thus providing rich time series feature data for subsequent short-term and long-term energy consumption prediction.

[0141] An autoregressive integrated short-term energy consumption trend model is built based on the energy consumption time series feature data to generate short-term energy consumption prediction data;

[0142] In this embodiment, based on the energy consumption time series feature data obtained by empirical mode decomposition, an autoregressive integrated moving average (ARIMA) model is used to build a short-term energy consumption trend model. This model can build a model according to the characteristics of the trend, seasonality, and randomness of historical data, so as to generate short-term energy consumption predictions. The set prediction time range is 7 days, and further by adjusting the parameters in the ARIMA model (such as the AR order, MA order, etc.), the prediction results can be made more accurate. The generated short-term energy consumption prediction data will help decision-makers in daily energy management and scheduling.

[0143] Residual analysis is performed on the short-term energy consumption prediction data, and the residual analysis results are input into a preset long short-term memory network for long-term energy consumption evolution modeling to obtain long-term energy consumption prediction data;

[0144] In this embodiment, residual analysis is performed to compare the difference between the actual energy consumption and the predicted energy consumption. Residual analysis can help identify potential biases or error sources in the prediction model. If the residuals are large, it indicates that the model needs further adjustment. Then, the residual data is input into a preset long short-term memory network (LSTM) for long-term energy consumption evolution modeling. The LSTM network has memory and can effectively capture long-term dependencies to generate long-term energy consumption prediction data. This data can be used for monthly, quarterly, or annual energy budgeting and scheduling.

[0145] The long-term energy consumption prediction data and the short-term energy consumption prediction data are fused to construct a comprehensive energy consumption prediction sequence; based on the comprehensive energy consumption prediction sequence, multi-resolution wavelet transform is performed to extract the energy consumption change patterns at different time scales, generating a multi-scale energy consumption evolution prediction matrix;

[0146] In this embodiment, the long-term energy consumption prediction data and the short-term energy consumption prediction data are fused to construct a comprehensive energy consumption prediction sequence. The weighted average method or other fusion algorithms are used to assign different weights to the short-term and long-term data according to the prediction time length. For example, the weight of the short-term prediction can be set to 0.7, and the weight of the long-term prediction is 0.3. The specific weights can be adjusted according to actual application requirements. Finally, the fused energy consumption prediction sequence can provide a more accurate and comprehensive energy consumption prediction, providing a basis for energy efficiency management. Multi-resolution wavelet transform (Wavelet Transform) is performed on the comprehensive energy consumption prediction sequence to extract the energy consumption change patterns at different time scales. Wavelet transform can perform multi-level and multi-frequency analysis on the data to help identify rapid fluctuations in the short term and slow change patterns in the long term. After wavelet transform, a multi-scale energy consumption evolution prediction matrix is generated, which contains the energy consumption patterns at different time scales. This multi-scale analysis method can provide a more accurate basis for energy consumption scheduling, load balancing, and resource optimization.

[0147] Based on the multi-scale energy consumption evolution prediction matrix, state probability modeling is performed, the energy consumption state transition matrix is calculated, and the energy consumption state transition matrix is combined with the building energy consumption evolution map to obtain the building energy consumption evolution map.

[0148] In this embodiment, based on the multi-scale energy consumption evolution prediction matrix, probability modeling of the energy consumption state is performed. By methods such as Markov chain, the transition probability of the energy consumption state is calculated to obtain the energy consumption state transition matrix. This matrix can describe the transition rules and probabilities between different energy consumption states. By combining the energy consumption state transition matrix with the building energy consumption evolution map, a map that comprehensively reflects the building energy efficiency evolution is obtained, which can play a key role in building energy efficiency management and optimization and support energy use prediction and decision-making.

[0149] Optionally, the long-term energy consumption evolution modeling is specifically as follows:

[0150] Align the short-term energy consumption prediction data with the results of residual analysis at the time step to obtain a time-series aligned residual data set;

[0151] In this embodiment, aligning the short-term energy consumption prediction data with the results of residual analysis requires ensuring the consistency of the two types of data in the time dimension. Usually, interpolation or resampling techniques can be used to complete the alignment. Select an appropriate time step (such as 1 hour or 30 minutes) for data alignment to ensure that the time intervals of the data are consistent and that the time step does not have too much impact on the prediction results.

[0152] Extract local spatio-temporal features from the time-series aligned residual data set to generate residual feature data;

[0153] In this embodiment, the time-series aligned residual data set will be processed by a local spatio-temporal feature extraction method. A convolutional neural network (CNN) is used to extract local features from the data. The extracted features include the energy consumption change trend in the short term and the time dependence in the long and short terms. In this process, a sliding window with a window size of 30 time steps (such as 30 hours) can be used for local feature extraction to capture the patterns of energy consumption fluctuations.

[0154] Construct a long short-term memory network model based on the residual feature data set;

[0155] In this embodiment, based on the extracted residual feature data set, a long short-term memory (LSTM) network model is constructed to learn the time series dependence of the data. The number of layers of this LSTM network is 3, the number of neurons in the hidden layer is 256, the optimization algorithm uses the Adam optimizer, and the learning rate is set to 0.001. Through training on historical data, the LSTM model can capture the time series law of energy consumption data, thus providing an accurate model for long-term energy consumption evolution.

[0156] Perform backpropagation on the long short-term memory network model, use the gradient descent method to optimize the network parameters, and model the long-term energy consumption evolution of the time-series aligned residual data set to obtain a long-term energy consumption evolution model;

[0157] In this embodiment, in the backpropagation stage, by using the gradient descent method to optimize the model parameters, a batch training method can be adopted, set the batch size to 64, and update the network parameters through the backpropagation algorithm to make the model fit the training data more accurately and reduce the error. In this process, 50 iterations are used for model training to ensure the accuracy of long-term energy consumption evolution.

[0158] Use the long-term energy consumption evolution model to predict the long-term energy consumption of the residual feature data to obtain long-term energy consumption prediction data.

[0159] In this embodiment, the trained long-term energy consumption evolution model will be used to perform long-term energy consumption prediction on the residual feature data. By inputting the residual feature data after time series alignment, the model can output the energy consumption prediction results for several future time steps (such as 72 hours or 168 hours). The prediction process not only depends on the learning of the training data, but also provides the dynamic evolution trend of the long-term energy consumption by dynamically adjusting the prediction window and the parameters of the model, so as to provide data support for the energy-saving optimization of building equipment.

[0160] Optionally, the generation of the energy-saving optimization instruction set described in step S4 is specifically as follows:

[0161] Perform device category strategy adaptability analysis based on the multi-objective energy-saving control strategy manifold, and perform hierarchical clustering of device control requirements to generate a device control classification index;

[0162] In this embodiment, by analyzing in detail the functional characteristics of various devices in the building (such as HVAC systems, lighting, elevators, smart sockets, etc.), important parameters such as the operating mode, energy consumption characteristics, and load response capabilities of the devices are extracted. Specifically, the characteristics of air-conditioning equipment include its temperature adjustment range, power consumption, response speed, etc.; the characteristics of lighting equipment include the brightness adjustment range, power consumption, and the ability to respond to changes in light intensity; while elevators have different operating modes, and electric curtains and fans have control requirements for automatic adjustment according to time or sensor data. Through data collection and analysis, a feature vector of each device is formed. Taking the air conditioner as an example, its feature vector may include information such as "maximum temperature adjustment range: ±2°C", "maximum power consumption: 3kW", "adjustment response time: 10 minutes", etc. The policy manifold can be regarded as a high-dimensional embedding space, where each policy point represents an energy-saving control strategy. To evaluate the suitability of each device for the energy-saving strategy, the feature vector of the device is compared with the policy points in the manifold, and the similarity between the device features and the policy points is calculated. Commonly used similarity evaluation methods include cosine similarity, Euclidean distance, and KL divergence, etc. For example, the temperature control characteristics of the air-conditioning equipment can be calculated for similarity with the policy points representing the temperature control optimization strategy in the policy manifold, and it is judged whether the device is suitable for this strategy through high similarity. For lighting equipment, the matching degree of its brightness adjustment ability with the lighting management strategy in the manifold is calculated. After evaluating the suitability of the device for the energy-saving strategy, the next step is to group all the devices in the building and classify them according to their adaptability to the energy-saving strategy. For this purpose, hierarchical clustering algorithms (such as K-means or DBSCAN) are used to group the devices. The devices will be divided into three categories according to their energy efficiency impact, control requirements, and operating mode: high-priority, medium-priority, and low-priority devices. For example, air-conditioning, lighting, and electric heater devices usually belong to the high-priority device group because they have a greater impact on the building's energy efficiency and require precise control; while electric curtains and fans belong to the low-priority device group because their impact on energy efficiency is relatively small. When classifying the devices, not only the load response capabilities and control requirements of the devices are considered, but also the energy efficiency boundary parameters of the devices are introduced to ensure that the control strategy of the device matches its actual capabilities. With the changes in the building environment and the operating state of the devices, attributes such as the load capacity, control time window, and environmental sensitivity of the devices may change. Therefore, in the device control classification, it is necessary to dynamically adjust the device classification. For example, the adjustment ability of the air-conditioning equipment within the temperature control range may be affected by external temperature changes, so in some cases it may need to be adjusted to a high-priority device for real-time energy-saving optimization. To ensure the accuracy of the analysis, energy efficiency boundary parameters of the devices are set, such as the temperature adjustment range (±1°C) of the air-conditioning equipment and the maximum brightness adjustment range (±20%) of the lighting equipment, to ensure that the classification of each device can be consistent with its actual control ability. Through the above steps, a device control classification index will be obtained.This index contains key information such as the priority, energy efficiency impact, and control requirements of each device, and can be dynamically adjusted as needed. The device control classification index will serve as the basis for the subsequent mapping of energy-saving strategy constraints to ensure that devices can be precisely matched with appropriate strategies when implementing energy-saving strategies. When generating the device control classification index, in addition to considering the control characteristics and energy efficiency impact of the devices, the actual operating status of the devices during the building operation process is also taken into account to improve the execution accuracy and effectiveness of the energy-saving strategies.

[0163] Map the energy-saving strategy constraints to the multi-objective energy-saving control strategy manifold based on the device control classification index to obtain a set of constrained optimization strategies;

[0164] In this embodiment, the specific control requirements of the devices in the device control classification index are mapped to the energy-saving control strategies. First, through the device category index, the energy efficiency requirements of the devices are mapped to appropriate energy-saving strategy manifolds to ensure that each device can minimize energy consumption to the greatest extent under specific strategies. For example, air-conditioning devices will adopt the "temperature rise limit strategy", that is, set the indoor temperature within a relatively stable range to avoid frequent on-off caused by excessive temperature fluctuations, thereby reducing energy consumption; for lighting devices, the "natural light induction dimming strategy" is adopted, and the indoor lighting brightness is adjusted in real time based on the external light intensity. During this process, the strategies are refined according to the environmental requirements, load distribution, and comfort requirements of the building. For example, the temperature control range of air-conditioning devices is set to 23°C - 26°C in summer and 18°C - 22°C in winter, and the indoor brightness of lighting devices is adjusted between 200 and 500 lux according to the external light intensity. In addition, other important constraint conditions are also set, such as the maximum delay of the control time does not exceed 5 minutes to ensure the real-time execution of the strategies.

[0165] Perform time series scheduling optimization on the set of constrained optimization strategies, adjust the strategy execution time, and generate a set of time-optimized strategies;

[0166] In this embodiment, time series scheduling optimization is carried out according to the device type and its control constraints. For example, air conditioning equipment usually has a high load during peak hours in the daytime, while the demand is less during off-peak hours at night. The running time of the equipment is optimized and adjusted by setting a time window. During the optimization process, based on historical data, environmental data (such as temperature changes and personnel activities), and prediction models (such as ARIMA models), the start time and running cycle of air conditioners and lighting equipment will be dynamically adjusted. For example, it is decided to lower the power of the air conditioning equipment during idle hours at night (such as adjusting from 20°C to 24°C), and increase the power during high-temperature hours in the daytime to keep the temperature stable. In addition, the lighting equipment will also be adjusted in real time according to the change of natural light, and will not waste energy when the light is sufficient. Through this time scheduling optimization, the peak and off-peak loads of building energy consumption can be effectively allocated, and the start and stop times of the equipment can be dynamically adjusted according to actual needs, so as to optimize the overall energy efficiency.

[0167] Allocate the execution priority based on the time optimization strategy set, and adaptively adjust the strategy execution parameters in combination with the time-series aligned environmental data set to generate an adaptive energy-saving control instruction set;

[0168] In this embodiment, based on the time optimization strategy set obtained from the previous optimization, the execution priority of each device in each time period is further analyzed. First, considering the energy efficiency impact of the device, the device priority and operation mode already allocated in the device control classification index. For example, due to its large energy efficiency impact and importance to indoor comfort, air conditioning equipment is usually assigned as a high-priority device, followed by lighting equipment, heaters, etc., while devices such as fans and smart sockets are usually assigned lower priorities. The operation mode of the device also affects its priority. For example, during high-temperature hours in summer, the air conditioning equipment needs to adjust the temperature frequently, so it will be temporarily adjusted to a high priority, while during low-temperature hours, it can be adjusted to a medium priority to save energy. During specific execution, according to the time-series aligned environmental data set (including indoor and outdoor temperature and humidity, personnel activities, and external climate changes), the execution parameters of the device will be adaptively adjusted. For example, if the system detects that the indoor temperature is high and there are many personnel activities, the operating power of the air conditioner will be immediately increased; at the same time, the system will combine the natural light sensing function of the lighting equipment, and if the external light intensity is sufficient, the lighting brightness will be reduced to reduce unnecessary energy consumption waste. In addition, based on the dynamic device control priority, the system will give priority to adjusting those devices with greater energy efficiency impact, such as air conditioners and lighting equipment, while ensuring comfort, to maximize energy efficiency.

[0169] Conduct consistency verification on the adaptive energy-saving control instruction set, verify the execution feasibility of the adaptive energy-saving control instruction set, obtain an energy-saving optimization instruction set, and upload it to the building energy consumption control platform to execute device control tasks.

[0170] In this embodiment, the consistency verification of the adaptive energy-saving control instruction set is carried out to ensure the feasibility and reliability of the execution of each instruction. First of all, all generated energy-saving control instructions will be verified through preset rules to ensure that each instruction conforms to the operating range of the equipment and environmental constraints. For example, the temperature adjustment instruction of the air-conditioning equipment will not exceed the set working temperature range (such as set between 18°C and 26°C), and the brightness adjustment of the lighting equipment will also be adjusted accordingly according to the real-time change of the external light intensity. The system will detect and verify whether the control instructions can be effectively executed in real time based on historical operation data, the current building energy efficiency status, and environmental feedback, so as to avoid problems such as overwork or underwork of the equipment. For any abnormal situation in the execution of instructions, the system will automatically make adjustments and correct the strategy parameters according to the feedback to ensure that the executed energy-saving instructions meet the actual requirements. Finally, the energy-saving optimization instruction set with verified consistency will be uploaded to the building energy consumption control platform to start executing specific equipment control tasks, ensuring the comfort of the environment while saving energy in the building.

[0171] Optionally, the specific process of tracking the deviation of the predicted energy consumption of the equipment described in step S5 is as follows:

[0172] Obtain the historical energy consumption data of the building equipment, and establish a benchmark energy consumption prediction model based on the historical energy consumption data of the building equipment and the energy-saving optimization instruction set, so as to obtain the equipment benchmark energy consumption prediction data;

[0173] In this embodiment, data collection is carried out through the building energy consumption control platform to obtain its historical energy consumption data. These historical data usually include the power consumption, operation cycle, external environment (such as temperature, humidity, light intensity) of each equipment in different time periods, and the load status of the equipment. The energy consumption data of the equipment is collected in real time through Internet of Things sensors and intelligent metering devices (such as electricity meters, temperature control sensors, etc.) and stored in the cloud database of the building energy consumption control platform. Then, combined with the historical energy consumption data of the equipment and the formulated energy-saving optimization instruction set (such as temperature control range, lighting brightness control, etc.), a benchmark energy consumption prediction model is constructed. This model can be trained using machine learning algorithms (such as support vector machine regression, decision tree regression, etc.), and the benchmark energy consumption of each equipment in the future time period is predicted through historical data. During the training process of the model, multi-dimensional features, such as environmental temperature and humidity, equipment load, etc., are used for data feature engineering to optimize the prediction accuracy of the model. Finally, the equipment benchmark energy consumption prediction data based on equipment characteristics, historical energy consumption data, and energy-saving instructions is obtained.

[0174] Obtain the real-time building energy consumption data, compare the error between the equipment benchmark energy consumption prediction data and the real-time building energy consumption data, calculate the deviation metric of the predicted energy consumption, and perform preliminary screening of abnormal energy consumption based on the preset deviation prediction to obtain the equipment energy consumption deviation metric matrix;

[0175] In this embodiment, energy consumption data is obtained in real time through a building energy consumption control platform, and this data is fed back to the cloud platform or the local control system in real time through an intelligent metering system (such as an intelligent electricity meter, a temperature and humidity sensor, etc.). The real-time energy consumption data usually includes the current power consumption value of each device, the load change, and the real-time data of the external environment (such as temperature, humidity, light, etc.). Then, the real-time energy consumption data is compared with the previously established device benchmark energy consumption prediction data, the error is calculated, and the analysis is carried out based on the deviation. Common methods for calculating errors include the mean square error (MSE), the mean absolute error (MAE), etc. Through the calculated deviation metric, the difference between the actual energy consumption of the device and the predicted value can be judged, and then the devices with large energy consumption deviation can be screened out to form a device energy consumption deviation metric matrix. This matrix contains the difference between the predicted energy consumption and the actual energy consumption of each device at each moment, the error size and its relative deviation degree. For devices with a deviation exceeding the set threshold (such as 10%), a preliminary screening of abnormal energy consumption is carried out as the basis for further diagnosis.

[0176] Obtain the device control log, and construct a causal relationship diagram of device energy consumption change based on the device control log and the dynamic energy consumption impact factor matrix, perform a causal test on the energy consumption impact factor, and generate a device energy consumption bloodline tracking chain;

[0177] In this embodiment, the control log of the device is obtained through the building energy consumption control platform, which records the running state of the device, operation instructions (such as power on / off, temperature adjustment, brightness adjustment, etc.) and the impact of environmental variables (such as external temperature, humidity, etc.) on the device. By analyzing the device control log, the control behavior of each device in different time periods can be understood, and then combined with the historical energy consumption data and the current environmental data of the device, a causal relationship diagram of device energy consumption change can be constructed. This diagram models the causal relationship between the control operations of the device and its energy consumption change, environmental factors, etc. To ensure the accuracy of the model, a dynamic energy consumption impact factor matrix is used, which calculates the impact degree of different control factors on the device energy consumption by analyzing the energy consumption fluctuations of the device under various environmental conditions. By testing these causal relationships and using causal inference methods (such as Granger causality test, Bayesian network, etc.), it is ensured that the causal relationships of each node in the relationship diagram are accurate. Finally, a device energy consumption bloodline tracking chain is generated, which records the energy consumption change path and influencing factors of each device. For example, the energy consumption change of an air conditioning device may be closely related to changes in various factors such as external temperature, the number of people in the room, and temperature control settings. The root cause of the energy consumption change can be traced through the bloodline tracking chain.

[0178] Combine the device bloodline tracking chain to trace the source of energy consumption anomalies in the device energy consumption deviation metric matrix, and classify the abnormal energy consumption patterns of the device based on the results of the energy consumption anomaly source tracing to obtain an energy consumption deviation retrospective analysis report.

[0179] In this embodiment, the energy consumption lineage tracking chain of the device is combined with the device energy consumption deviation measurement matrix to trace the origin of abnormal energy consumption. Through the lineage tracking chain, the root cause of the energy consumption deviation of each device can be determined, and the factors causing the deviation can be found, such as abnormal control instructions, changes in the external environment, aging of the device, etc. By tracing each device with a large deviation degree in the deviation measurement matrix one by one, the pattern of energy consumption change is analyzed to further determine whether there is abnormal energy consumption behavior. According to the tracing results, the devices are divided into different abnormal energy consumption modes. For example, the energy consumption of air conditioning equipment may deviate due to sudden changes in external temperature, while the energy consumption of lighting equipment may increase unnecessarily due to the failure of the automatic control system. Finally, an energy consumption deviation retrospective analysis report is generated, which includes the energy consumption deviation of each device, the analysis of abnormal reasons, the classification of abnormal energy consumption modes, and the recommended improvement measures, providing support for subsequent energy-saving optimization and device management.

[0180] Optionally, the establishment of the benchmark energy consumption prediction model is specifically

[0181] Obtain the historical energy consumption data of building equipment, and perform time series alignment on the historical energy consumption data of building equipment to obtain a standardized historical energy consumption dataset of the equipment;

[0182] In this embodiment, the historical energy consumption data of building equipment is obtained through the building energy consumption control platform. These data include the energy consumption records of the equipment in different time periods, such as the energy consumption values per hour or per minute. After obtaining the data, time series alignment is required, which is usually completed by resampling or interpolating data from different time sources. Specifically, a unified time step can be set, such as 1 hour, and the linear interpolation method is applied to align the data with unequal time intervals to a unified time point. After alignment, all data is standardized, and the standardization method of zero mean and unit variance is adopted to make the mean of the data 0 and the standard deviation 1, ensuring the comparability and stability of the data in subsequent modeling, and obtaining a standardized historical energy consumption dataset of the equipment.

[0183] Quantify the energy-saving optimization impact on the historical energy consumption data of the equipment based on the energy-saving optimization instruction set and the dynamic energy consumption impact factor matrix to generate an energy-saving optimization impact factor dataset;

[0184] In this embodiment, the energy-saving optimization instruction set includes a series of energy-saving measures during the operation of the equipment. These measures usually include adjusting the temperature setting value of the air conditioner, changing the brightness of the lights, adjusting the working mode of the heater, etc. Each energy-saving operation can affect the energy consumption of the equipment by changing the operating state of the equipment or the environmental conditions. Therefore, it is necessary to analyze according to the influencing factors of each energy-saving optimization instruction (such as environmental temperature, humidity, equipment load, working duration, etc.) and quantify the specific impact of these factors on the equipment energy consumption. Based on the dynamic energy consumption impact factor matrix, the weighted average method or the regression analysis method can be used for the quantitative analysis of the energy-saving optimization instructions. Specifically, through the regression model or the weighted average model, the influencing factors corresponding to each operation in the energy-saving optimization instruction set are combined with the factors in the dynamic energy consumption impact factor matrix to quantify the impact degree of each energy-saving operation on the equipment energy consumption. For example, if an energy-saving operation instruction adjusts the temperature setting value of the air conditioner, and according to historical data, the influence coefficient of temperature change on energy consumption is 0.15, then the energy-saving effect of this instruction can be calculated by multiplying this coefficient by the adjusted temperature change range, and finally the energy-saving optimization impact factor is obtained. Finally, through the above calculation process, the impact of each energy-saving optimization instruction on the equipment energy consumption is quantified into a value, generating an energy-saving optimization impact factor data set. This data set records the quantitative relationship between each energy-saving optimization instruction and the equipment energy consumption.

[0185] Based on the standardized equipment historical energy consumption data set and the energy-saving optimization impact factor data set, a baseline energy consumption prediction model is established through a preset regression algorithm to obtain the equipment baseline energy consumption prediction model;

[0186] In this embodiment, based on the standardized equipment historical energy consumption data set and the energy-saving optimization impact factor data set, a baseline energy consumption prediction model is established through a preset regression algorithm. The support vector regression (SVR) algorithm is used as the regression model. During specific operation, first select a suitable kernel function (such as the RBF kernel) and set the parameter C and ε values. During the training process, by using the standardized equipment historical energy consumption data and the energy-saving optimization impact factor data as inputs, the model is trained to predict the baseline energy consumption of the equipment without energy-saving optimization intervention. The SVR model generates the equipment baseline energy consumption prediction model by minimizing the error function, which can accurately reflect the normal energy consumption trend of the equipment.

[0187] Perform cross-validation on the equipment baseline energy consumption prediction model and iteratively adjust the model parameters to obtain the baseline energy consumption prediction model;

[0188] In this embodiment, the k-fold cross-validation method is adopted, and k is set to 5. The energy consumption data of the device is divided into 5 subsets. Each subset is used as the validation set in turn, and the remaining subsets are used as the training set for model training and validation. After each validation, the prediction error is calculated, and the parameters of the SVR model, especially the C value and the ε value, are adjusted according to the error to optimize the generalization ability and accuracy of the model. Through multiple rounds of cross-validation, the optimal device baseline energy consumption prediction model is finally obtained.

[0189] The energy-saving optimization instruction set is input into the baseline energy consumption prediction model to obtain the device baseline energy consumption prediction data.

[0190] In this embodiment, the energy-saving optimization instruction set is input into the baseline energy consumption prediction model to obtain the device baseline energy consumption prediction data. At this time, the energy-saving optimization instruction set contains operation instructions for adjusting the device operation mode. By inputting these instructions into the obtained baseline energy consumption prediction model, the energy consumption prediction data of the device after executing these energy-saving optimization instructions is generated. For example, if the energy-saving instruction set contains instructions to increase or decrease the air conditioner temperature, the model will calculate the energy consumption change of the device under this condition, and finally obtain the optimized device baseline energy consumption prediction data. These data will be used for subsequent energy efficiency evaluation and optimization decisions.

[0191] Therefore, from any point of view, the embodiment should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the application document are intended to be encompassed within the present invention.

[0192] The above are only the specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A building energy consumption optimization method based on artificial intelligence, characterized in that: The following steps are involved: Step S1: Acquire multimodal sensor data and construct a heterogeneous sensor dynamic spatiotemporal adjacency matrix; perform device physical space semantic mapping modeling based on the heterogeneous sensor dynamic spatiotemporal adjacency matrix to obtain a sensor spatiotemporal fusion feature set; Step S2: Perform cross-domain energy consumption impact mapping topological modeling on the sensing spatiotemporal fusion feature set to obtain a cross-domain energy consumption correlation topological map; extract the device energy transfer influencing factors according to the cross-domain energy consumption correlation topological map to obtain a dynamic energy consumption influencing factor matrix; Step S3: Perform multi-scale energy consumption evolution deduction according to the dynamic energy consumption influencing factor matrix to generate a building energy consumption evolution map; construct an energy consumption probability cloud map according to the building energy consumption evolution map; Step S4: performing adversarial strategy optimization on the energy consumption probability cloud map to obtain a multi-objective energy-saving control strategy manifold; generating an energy-saving optimization instruction set based on the multi-objective energy-saving control strategy manifold, and uploading it to the building energy consumption control platform to control the equipment; Step S5: Acquire real-time building energy consumption data, perform equipment predicted energy consumption deviation lineage tracking on the real-time building energy consumption data according to the energy-saving optimization instruction set, and implement equipment control local comfort instruction optimization to obtain a personalized control instruction set, and upload it to the building energy consumption control platform to control the equipment.

2. The method for optimizing building energy consumption based on artificial intelligence according to claim 1 is characterized in that: Step S1 is specifically as follows: Step S11: acquiring multimodal sensing data through a multimodal sensor group, wherein the multimodal sensing data includes temperature and humidity sensing data, light sensing data, crowd density infrared sensing data, and equipment status sensing data; Step S12: performing time step timing alignment on the multimodal sensing data to obtain a time-series aligned environment data set; Step S13: performing non-Euclidean topological mapping on the time-series aligned environment dataset, and constructing an initial spatial semantic topological structure graph based on the topological mapping result; Step S14: performing neighborhood adaptive topology optimization based on the initial spatial semantic topology structure graph to obtain a device spatial semantic graph, and fusing the device operation mode and environmental data based on the device spatial semantic graph to obtain a heterogeneous sensor dynamic spatiotemporal adjacency matrix; Step S15: Establish a device physical space semantic mapping model based on the heterogeneous sensor dynamic spatiotemporal adjacency matrix, perform feature screening and association optimization on the model establishment results, and obtain a sensor spatiotemporal fusion feature set.

3. The method for optimizing building energy consumption based on artificial intelligence according to claim 2 is characterized in that: Step S13 is specifically as follows: Step S131: Acquire the equipment operation data of the building area, and synchronize the equipment operation data of the building area and the timing alignment environment data set to obtain the timing alignment equipment operation data; Step S132: Perform non-Euclidean topological mapping on the timing-aligned device operation data and the timing-aligned environment data set, construct a spatiotemporal relationship map between the devices and the environment variables in the building area, and obtain a spatiotemporal relationship map of the building area; Step S133: Perform physical space interaction modeling based on the building area spatiotemporal relationship map, map the physical space interaction modeling results to a preset building physical space coordinate system, convert nodes in the spatiotemporal relationship map into actual physical space coordinate nodes, and construct device-environment physical space interaction data; Step S134: performing node space layout analysis on the device-environment physical space interaction data, and mapping the interaction relationship between the device and the environment according to the node space layout analysis result to obtain an environment physical space interaction diagram; Step S135: Optimize the graph based on the environmental physical space interaction graph and construct an initial spatial semantic topology structure graph.

4. The method for optimizing building energy consumption based on artificial intelligence according to claim 2 is characterized in that: Step S14 is specifically as follows: Step S141: Calculate the local density of topological nodes according to the initial spatial semantic topological structure graph, set the calculation neighborhood radius to 2, set the local density threshold to 0.1, and regard nodes below the local density threshold as sparse areas, thereby obtaining the local density distribution of the topological structure; Step S142: Set the minimum number of samples to 5, the density threshold to 0.3, perform node neighborhood adaptive clustering on the initial spatial semantic topology structure graph, and set the weight range to [0, 1] to calculate the neighborhood adaptive weight based on the clustering result, and generate the initial parameters of neighborhood adaptive topology optimization; Step S143: setting the weight adjustment coefficient to 0.5, adjusting the neighborhood topology weight of the initial spatial semantic topology structure graph according to the initial parameters of the neighborhood adaptive topology optimization, and setting the optimization scale [1,3] to perform multi-scale topology structure optimization to obtain the optimized device spatial semantic graph; Step S144: extracting device operation mode features with a feature dimension of 5 based on the optimized device space semantic graph, fusing the time-series alignment environment data set, normalizing the device operation mode features, and generating standardized device operation mode data; Step S145: embedding the standardized device operation mode data into the topological structure of the optimized device space semantic graph, performing interactive modeling of the device operation mode and environmental data, and obtaining a preliminary heterogeneous sensor dynamic spatiotemporal adjacency matrix; Step S146: Perform feature screening and topological association optimization on the preliminary heterogeneous sensor dynamic spatiotemporal adjacency matrix, strengthen key interactive relationships, and remove low-correlation edge weights, thereby generating a heterogeneous sensor dynamic spatiotemporal adjacency matrix.

5. The method for optimizing building energy consumption based on artificial intelligence according to claim 1 is characterized in that: Step S2 is specifically as follows: Step S21: Screen the cross-domain energy consumption influencing factors of the sensing spatiotemporal fusion feature set, set the correlation threshold to [0.7, 0.8] to retain high correlation features, calculate the energy consumption impact weight of each device, and construct an initial energy consumption impact feature matrix based on the energy consumption impact weight; Step S22: Perform non-Euclidean graph topology mapping on the initial energy consumption impact feature matrix, set the maximum connection threshold of the adjacency matrix to 0.5, construct an energy consumption correlation topology structure across devices and across spatial regions, and generate an initial cross-domain energy consumption correlation topology map; Step S23: Perform spectral clustering analysis on the initial cross-domain energy consumption correlation topology map, select the first 20 eigenvectors for Laplace eigendecomposition, and set the feature cutoff threshold to 0.1 to extract the main energy consumption impact mode, and generate a cross-domain energy consumption correlation topology map; Step S24: Modeling the energy consumption transfer relationship between devices based on the cross-domain energy consumption association topology diagram, setting the energy consumption transfer rate to [0.05, 0.2], and setting the dynamic weight change interval to [0, 1] to assign dynamic weights, and generating an initial dynamic energy consumption impact factor matrix; Step S25: normalize the initial dynamic energy consumption influencing factor matrix, set the influencing factor confidence interval [0.95, 1.05] to correct the robustness of the influencing factors, optimize the energy consumption influencing parameters, and output the dynamic energy consumption influencing factor matrix.

6. The method for optimizing building energy consumption based on artificial intelligence according to claim 1, characterized in that: The multi-scale energy consumption evolution deduction described in step S3 is specifically as follows: Perform time series decomposition on the dynamic energy consumption influencing factor matrix, set multi-scale time steps, and perform trend decomposition and cycle decomposition to generate an energy consumption time series decomposition data set; Perform causal inference on the energy consumption propagation path of the energy consumption time series decomposition data set, calculate the energy consumption correlation strength between devices, and build an energy consumption propagation network to generate the initial structure of energy consumption propagation; Based on the initial structure of energy consumption propagation, the energy consumption propagation pattern is clustered, the energy consumption propagation path is screened, the low-contribution energy consumption path is removed, and the optimized energy consumption propagation network is generated; Extract time series features from the optimized energy consumption propagation network, divide the energy consumption time series segments, and perform empirical mode trend decomposition on the energy consumption time series segments to obtain energy consumption time series feature data; Based on the energy consumption time series characteristic data, autoregressive integral short-term energy consumption trend modeling is performed to generate short-term energy consumption forecast data; Perform residual analysis on short-term energy consumption forecast data, input the residual analysis results into the preset long short-term memory network to perform long-term energy consumption evolution modeling, and obtain long-term energy consumption forecast data; The long-term energy consumption forecast data is integrated with the short-term energy consumption forecast data to construct a comprehensive energy consumption forecast sequence; multi-resolution wavelet transform is performed based on the comprehensive energy consumption forecast sequence to extract the energy consumption change patterns at different time scales and generate a multi-scale energy consumption evolution prediction matrix; Based on the multi-scale energy consumption evolution prediction matrix, state probability modeling is performed, the energy consumption state transfer matrix is ​​calculated, and the energy consumption state transfer matrix is ​​combined with the building energy consumption evolution map to obtain the building energy consumption evolution map.

7. The method for optimizing building energy consumption based on artificial intelligence according to claim 1 is characterized in that: The long-term energy consumption evolution modeling is specifically as follows: The short-term energy consumption forecast data and the residual analysis results are aligned in time steps to obtain a time-series aligned residual data set; Extract local spatiotemporal features from the time-series aligned residual data set to generate residual feature data; Based on the residual feature dataset, a long short-term memory network model is constructed; The long short-term memory network model is back-propagated, the network parameters are optimized using the gradient descent method, and the long-term energy consumption evolution model is modeled for the time-series alignment residual data set to obtain the long-term energy consumption evolution model; The long-term energy consumption evolution model is used to perform long-term energy consumption prediction on the residual characteristic data to obtain long-term energy consumption prediction data.

8. The method for optimizing building energy consumption based on artificial intelligence according to claim 1, characterized in that: The generation of the energy-saving optimization instruction set described in step S4 is specifically as follows: Based on the multi-objective energy-saving control strategy manifold, the adaptability analysis of equipment category strategy is carried out, and the equipment control demand hierarchical clustering is performed to generate the equipment control classification index; According to the equipment control classification index, the energy-saving strategy constraints are mapped to the multi-objective energy-saving control strategy manifold to obtain the constrained optimization strategy set; Perform time series scheduling optimization on the constraint optimization strategy set, adjust the strategy execution time, and generate a time optimization strategy set; Assign policy execution priorities based on the time optimization policy set, and adaptively adjust policy execution parameters based on the timing alignment environment data set to generate an adaptive energy-saving control instruction set; The adaptive energy-saving control instruction set is verified for consistency, the feasibility of its execution is verified, the energy-saving optimization instruction set is obtained, and uploaded to the building energy consumption control platform to execute equipment control tasks.

9. The method for optimizing building energy consumption based on artificial intelligence according to claim 1, characterized in that: The device predicted energy consumption deviation lineage tracking described in step S5 is specifically as follows: Obtain historical energy consumption data of building equipment, and establish a baseline energy consumption prediction model based on the historical energy consumption data of building equipment and an energy-saving optimization instruction set, thereby obtaining equipment baseline energy consumption prediction data; Obtain real-time building energy consumption data, compare the equipment benchmark energy consumption forecast data and real-time building energy consumption data, calculate the forecast energy consumption deviation metric, and perform preliminary screening of abnormal energy consumption based on the preset deviation forecast to obtain the equipment energy consumption deviation metric matrix; Obtain the equipment control log, and build a causal relationship diagram of equipment energy consumption changes based on the equipment control log and the dynamic energy consumption influencing factor matrix, perform causal test of energy consumption influencing factors, and generate the equipment energy consumption lineage tracking chain; The energy consumption anomaly is traced to the equipment energy consumption deviation measurement matrix in combination with the equipment lineage tracing chain, and the abnormal energy consumption patterns of the equipment are classified based on the energy consumption anomaly tracing results to obtain an energy consumption deviation backtracking analysis report.

10. The method for optimizing building energy consumption based on artificial intelligence according to claim 9, characterized in that: The establishment of the benchmark energy consumption prediction model is specifically as follows: Obtain historical energy consumption data of building equipment, and perform time-series alignment on the historical energy consumption data of building equipment to obtain a standardized historical energy consumption data set of equipment; Based on the energy-saving optimization instruction set and the dynamic energy consumption impact factor matrix, the energy-saving optimization impact of the equipment's historical energy consumption data is quantified to generate an energy-saving optimization impact factor data set; Based on the standardized equipment historical energy consumption data set and the energy-saving optimization influencing factor data set, the benchmark energy consumption prediction model is built through the preset regression algorithm to obtain the equipment benchmark energy consumption prediction model; Cross-validate the equipment benchmark energy consumption prediction model, iteratively adjust the model parameters, and obtain the benchmark energy consumption prediction model; The energy-saving optimization instruction set is input into the baseline energy consumption prediction model to obtain the equipment baseline energy consumption prediction data.

Images