Building energy consumption analysis method and system based on artificial intelligence

Abstract

The invention relates to the technical field of building energy consumption analysis, and discloses a building energy consumption analysis method and system based on artificial intelligence. The method comprises the following steps: collecting building environment data to form an energy consumption basic data set; processing the data set to generate an energy consumption feature vector; constructing a prediction model to obtain an energy consumption predictor; a predictor is used for comparing actual data to identify abnormity and generate a report; formulating an optimization scheme based on the report to generate a control instruction; and executing instruction record change data to update the feature library to complete a closed loop. According to the invention, closed-loop management of accurate prediction, anomaly detection, optimization control and effect evaluation of building energy consumption is realized, so that the building energy utilization efficiency is improved, and energy waste is reduced.

Description

Technical Field

[0001] This application relates to the technical field of building energy consumption analysis, and particularly to an artificial intelligence-based building energy consumption analysis method and system. Background Art

[0002] With the continuous improvement of energy consumption and environmental protection awareness, building energy conservation has become an important global issue. Building energy consumption accounts for a large proportion of total energy consumption. According to statistics, building energy consumption accounts for about 40% of global energy consumption. Traditional building energy consumption analysis methods mainly rely on manual experience and simple statistical analysis, such as regular meter reading and manual comparison of energy consumption data. With the development of Internet of Things technology, building energy management systems (BEMS) have begun to be applied to building energy consumption monitoring, and through the installation of various sensors and metering devices, automatic monitoring of building energy consumption has been achieved. In recent years, with the rapid development of artificial intelligence technology, some studies have begun to attempt to apply machine learning algorithms to building energy consumption prediction and anomaly detection, such as using algorithms like support vector machines and random forests to build energy consumption prediction models, or using clustering algorithms to identify abnormal energy consumption patterns.

[0003] However, the existing building energy consumption analysis methods still have many deficiencies. First, most methods fail to fully consider the spatio-temporal correlation of building energy consumption, ignoring the mutual influence relationship between energy consumption in different regions and different time periods, resulting in limited prediction accuracy. Second, existing methods usually use static thresholds for anomaly detection, unable to adapt to the dynamic changes of building energy consumption with seasons, weather, and usage patterns, and are prone to false alarms and missed alarms. Third, there is a lack of adaptive learning ability, unable to automatically adjust model parameters according to newly acquired data and feedback results, and it is difficult to adapt to the long-term evolution of building energy consumption characteristics. Fourth, existing methods often treat energy consumption prediction, anomaly detection, and energy conservation optimization as independent links, lacking an integrated closed-loop management mechanism, resulting in the implementation effect of energy conservation measures being difficult to accurately evaluate and continuously improve. Finally, the existing methods have limited ability to fuse and process multi-source heterogeneous data, and it is difficult to fully utilize the rich information provided by multi-dimensional data such as building temperature, humidity, and occupancy density. Summary of the Invention

[0004] This application provides an artificial intelligence-based building energy consumption analysis method and system, which is used to achieve closed-loop management of accurate prediction, anomaly detection, optimization control, and effect evaluation of building energy consumption, thereby improving building energy use efficiency and reducing energy waste.

[0005] In a first aspect, the present application provides an artificial intelligence-based building energy consumption analysis method. The artificial intelligence-based building energy consumption analysis method includes: collecting building temperature, humidity, energy consumption, and occupancy density data through multi-dimensional sensors to obtain a building energy consumption basic data set; performing data missing value filling and outlier processing on the basis of the building energy consumption basic data set, extracting building usage time periods, environmental parameters, and energy load characteristics, and generating a building energy consumption feature vector; establishing a building energy consumption prediction model considering time factors and spatial relationships based on the building energy consumption feature vector, and obtaining a building energy consumption predictor through training with historical data; calculating an expected energy consumption value using the building energy consumption predictor, comparing the actual energy consumption data with the expected energy consumption value, identifying energy consumption abnormal intervals, and generating a building energy consumption abnormal report; formulating an energy optimization parameter adjustment plan based on the building energy consumption abnormal report, in combination with the building usage plan and environmental conditions, generating a building energy conservation control instruction; by executing the building energy conservation control instruction, recording the energy consumption change data before and after adjustment, calculating an energy efficiency improvement index, updating the building energy consumption feature library, and completing the closed-loop of building energy consumption analysis.

[0006] In a second aspect, the present application provides an artificial intelligence-based building energy consumption analysis system. The artificial intelligence-based building energy consumption analysis system includes:

[0007] A collection module for collecting building temperature, humidity, energy consumption, and occupancy density data through multi-dimensional sensors to obtain a building energy consumption basic data set;

[0008] An extraction module for performing data missing value filling and outlier processing on the basis of the building energy consumption basic data set, extracting building usage time periods, environmental parameters, and energy load characteristics, and generating a building energy consumption feature vector;

[0009] A building module for establishing a building energy consumption prediction model considering time factors and spatial relationships based on the building energy consumption feature vector, and obtaining a building energy consumption predictor through training with historical data;

[0010] A comparison module for calculating an expected energy consumption value using the building energy consumption predictor, comparing the actual energy consumption data with the expected energy consumption value, identifying energy consumption abnormal intervals, and generating a building energy consumption abnormal report;

[0011] A generation module for formulating an energy optimization parameter adjustment plan based on the building energy consumption abnormal report, in combination with the building usage plan and environmental conditions, generating a building energy conservation control instruction;

[0012] An update module for recording the energy consumption change data before and after adjustment by executing the building energy conservation control instruction, calculating an energy efficiency improvement index, updating the building energy consumption feature library, and completing the closed-loop of building energy consumption analysis.

[0013] The third aspect of the present invention provides a computer device, comprising: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor invokes the instructions in the memory to cause the computer device to execute the above-mentioned artificial intelligence-based building energy consumption analysis method.

[0014] The fourth aspect of the present invention provides a computer-readable storage medium, in which instructions are stored, and when it runs on a computer, it causes the computer to execute the above-mentioned artificial intelligence-based building energy consumption analysis method.

[0015] In the technical solution provided by this application, by collecting building temperature, humidity, energy consumption, and occupancy density data through multi-dimensional sensors, a comprehensive building energy consumption basic data set is obtained, providing a rich data basis for subsequent analysis and solving the problem of single data source in traditional methods. By performing data missing value filling and outlier processing on the building energy consumption basic data set, extracting building usage time periods, environmental parameters, and energy load characteristics, and generating building energy consumption feature vectors, the data quality and feature expression ability are significantly improved, providing high-quality input for model training. Based on the building energy consumption feature vectors, a building energy consumption prediction model considering time factors and spatial relationships is established, fully capturing the spatio-temporal correlation of building energy consumption, overcoming the limitation of traditional models that ignore the mutual influence of energy consumption between regions, and improving the prediction accuracy. Using the building energy consumption predictor to calculate the expected energy consumption value, comparing the actual energy consumption data with the expected energy consumption value, identifying the energy consumption abnormal interval, and generating a building energy consumption abnormal report, realizing the transformation from passive discovery to active warning, and greatly shortening the abnormal discovery time. Based on the building energy consumption abnormal report, combined with the building usage plan and environmental conditions, a highly targeted energy optimization parameter adjustment plan is formulated, generating building energy-saving control instructions, realizing precise energy saving, and avoiding the problem of reduced comfort caused by the "one-size-fits-all" approach in traditional methods. By executing the building energy-saving control instructions, recording the energy consumption change data before and after adjustment, calculating the energy efficiency improvement index, and updating the building energy consumption feature library, the closed-loop of building energy consumption analysis is completed, realizing self-learning and continuous optimization. This solution makes full use of the advantages of artificial intelligence algorithms in the field of specific building energy consumption management, especially in spatio-temporal data processing. By using bidirectional long short-term memory networks to capture the time-dependent relationship of energy consumption, graph convolutional networks to model the spatial correlation between building regions, and attention mechanisms to dynamically adjust the importance weights of different features, these algorithm features contribute to the ability to adaptively process the dynamic change characteristics of building energy consumption, identify complex energy consumption patterns, and thus achieve more accurate prediction and more refined control. In addition, by using genetic algorithms to solve multi-objective optimization problems, the three dimensions of energy saving, cost, and comfort are balanced, realizing a globally optimal energy management strategy. The incremental learning mechanism enables the model to continuously learn from new data, improve the prediction accuracy, and adapt to the long-term evolution of building energy consumption characteristics. Brief Description of the Drawings

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0017] Figure 1 It is a schematic diagram of an embodiment of the building energy consumption analysis method based on artificial intelligence in the embodiments of the present application;

[0018] Figure 2 It is a schematic diagram of an embodiment of the building energy consumption analysis system based on artificial intelligence in the embodiments of the present application;

[0019] Figure 3 It is a schematic block diagram of the structure of the computer device in the embodiments of the present invention. Detailed Embodiments

[0020] The embodiments of the present application provide a building energy consumption analysis method and system based on artificial intelligence. The terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and the above drawings of the present application are used to distinguish similar objects, and do not necessarily need to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the term "including" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily need to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0021] For ease of understanding, the following describes the specific process of the embodiments of the present application. Please refer to Figure 1 , an embodiment of the building energy consumption analysis method based on artificial intelligence in the embodiments of the present application includes:

[0022] Step S101: Collect building temperature, humidity, energy consumption and personnel density data through multi-dimensional sensors to obtain a building energy consumption basic data set;

[0023] Step S102: According to the building energy consumption basic data set, perform data missing filling and outlier processing, extract building usage time periods, environmental parameters and energy load characteristics, and generate a building energy consumption feature vector;

[0024] Step S103: Based on the building energy consumption characteristic vector, establish a building energy consumption prediction model considering time factors and spatial relationships, and obtain a building energy consumption predictor through training with historical data;

[0025] Step S104: Use the building energy consumption predictor to calculate the expected energy consumption value, compare the actual energy consumption data with the expected energy consumption value, identify the energy consumption abnormal interval, and generate a building energy consumption abnormal report;

[0026] Step S105: Based on the building energy consumption abnormal report, combined with the building usage plan and environmental conditions, formulate an energy optimization parameter adjustment plan and generate a building energy saving control instruction;

[0027] Step S106: By executing the building energy saving control instruction, record the energy consumption change data before and after adjustment, calculate the energy efficiency improvement index, update the building energy consumption characteristic library, and complete the closed-loop of building energy consumption analysis.

[0028] It can be understood that the execution subject of this application can be an artificial intelligence-based building energy consumption analysis system, or a terminal or a server. Specifically, it is not limited here. This application embodiment is described by taking the server as the execution subject as an example.

[0029] Specifically, the building temperature, humidity, energy consumption, and occupancy density data are collected through multi-dimensional sensors to obtain the basic building energy consumption dataset. Specifically, a temperature sensor network is arranged in each functional area of the building to collect temperature change data in each area, forming a temperature spatio-temporal data matrix; at the same time, the humidity distribution data inside the building is collected through multi-point humidity sensors, and the humidity environment feature set is generated in combination with the ventilation state parameters; the sub-item energy consumption data of the main energy-consuming equipment in the building is collected through an energy metering device, and the equipment operation power curve is recorded to construct an equipment energy consumption database; the personnel flow and density data inside the building are collected through a personnel detection device to establish a personnel activity pattern feature table. These data are processed through time synchronization, invalid data points are removed, and data for missing time periods are supplemented, finally forming a basic building energy consumption dataset containing multi-dimensional information. Based on the basic building energy consumption dataset, data missing filling and outlier processing are carried out, and the building usage period, environmental parameters, and energy load characteristics are extracted to generate a building energy consumption feature vector. The specific processing process includes filling the missing data through the sliding time window average filling method, which selects the valid data within a certain time window before and after the missing point to calculate the average value for filling; then the improved Z-score method is used to identify outliers, calculating the deviation degree of each data point from the surrounding data, and marking the points with too large deviation as outliers and replacing them with the median of the front and back data points; subsequently, the smoothed data is segmented in time series and divided into working periods, transitional periods, and idle periods according to the energy consumption load change characteristics; noise reduction processing is carried out through the wavelet transform method, decomposing the energy consumption data into a trend term, a seasonal term, and a residual term; environmental parameter characteristics such as temperature difference, thermal inertia index, sunshine influence coefficient, humidity change rate, and occupancy density distribution are extracted from the processed data; finally, the period division table, energy consumption decomposition characteristics, and environmental impact factor set are integrated through a feature fusion matrix to form a complete building energy consumption feature vector.

[0030] According to the building energy consumption characteristic vector, a building energy consumption prediction model considering time factors and spatial relationships is established. Through training with historical data, a building energy consumption predictor is obtained. This process first divides the building energy consumption characteristic vector into a time feature sub-vector and a spatial feature sub-vector; performs bidirectional long short-term memory network processing on the time feature sub-vector. This network calculates hidden states through forward and backward time steps and can capture long-term dependencies; performs graph convolution operations on the spatial feature sub-vector, represents each area of the building as a node in the graph, and represents the physical connection relationship between areas as edges, and captures the mutual influence of energy consumption between different areas through a message passing mechanism; fuses the processed time relationship feature matrix and the spatial association feature graph through an attention mechanism, calculates the weight coefficients of each feature channel for weighted combination; then pairs the fused features with historical building energy consumption data to form training samples, and optimizes the model parameters through the mini-batch gradient descent algorithm; finally, evaluates the prediction stability and accuracy of the model under different data partitions through the cross-validation method, and selects the model with the best performance as the building energy consumption predictor.

[0031] Use the building energy consumption predictor to calculate the expected energy consumption value, compare the actual energy consumption data with the expected energy consumption value, identify the energy consumption abnormal interval, and generate a building energy consumption anomaly report. The specific process is to input the current building parameter data into the predictor for forward calculation to obtain the expected energy consumption value at each time point; collect the actual building energy consumption data in the current period and align it with the expected energy consumption sequence in time; calculate the difference between the actual energy consumption sequence and the expected energy consumption sequence to generate an energy consumption residual sequence; analyze the residual sequence through a dynamic threshold detection algorithm, calculate the threshold at each time point, and determine it as an abnormal point when the absolute value of the residual exceeds the corresponding threshold; perform continuity analysis on the abnormal points, and mark the interval where three or more abnormal points appear continuously as the energy consumption abnormal interval; classify according to the characteristic parameters of the energy consumption abnormal interval combined with the decision tree algorithm to identify the abnormal type, and form a building energy consumption anomaly report including the abnormal location, type, severity, and recommended handling methods.

[0032] Based on the building energy consumption anomaly report, combined with the building usage plan and environmental conditions, formulate an energy optimization parameter adjustment plan and generate a building energy-saving control instruction. This step first extracts the abnormal type and energy consumption deviation data from the anomaly report to construct an optimization target list; collects building usage plan data, including work period arrangements, personnel occupancy plans, and special event times; obtains future environmental condition forecast data; constructs a multi-objective optimization function, including energy consumption indicators, economic cost indicators, and comfort indicators, sets the weight coefficients of each indicator, and solves the optimal parameter combination through a genetic algorithm; decomposes the obtained equipment operation parameter adjustment plan into equipment-level control parameters; finally, converts the control parameter table into an instruction format that conforms to the building control system communication protocol to form a building energy-saving control instruction.

[0033] By executing the building energy-saving control instructions, recording the energy consumption change data before and after adjustment, calculating the energy efficiency improvement indicators, and updating the building energy consumption feature library, the closed-loop of building energy consumption analysis is completed. The specific process is to send the energy-saving control instructions to the building control system and record the execution timestamp; use multi-dimensional sensors to monitor the adjusted operating status in real time and collect various adjusted data; align the adjusted data with the historical data of the same period before adjustment in terms of time and calculate the energy consumption difference value; conduct statistical analysis on the energy consumption change data and calculate indicators such as the average energy consumption reduction rate, peak energy consumption reduction amount, and load balance degree; fuse the key parameters with the building energy consumption feature vector and update the building energy consumption feature library; perform incremental training on the predictor through the updated feature library to form a closed-loop optimized analysis system.

[0034] For example, after collecting the basic energy consumption data for one month through multi-dimensional sensors, it is found that there are 10% missing points in the air-conditioning energy consumption data during the working hours from Monday to Friday. The missing points are filled by the sliding time window average filling method, and the window size is set to 3 hours before and after. The processed data set shows that the energy consumption peaks are from 9 to 11 o'clock and from 14 to 16 o'clock on weekdays. The daily cycle pattern and the weekly cycle pattern are separated through wavelet transform. The building energy consumption feature vector contains key information such as time features (weekday / non-weekday, day / night), environmental features (indoor-outdoor temperature difference, sunlight intensity), and usage features (personnel density distribution). The predictor trained based on these features shows a prediction accuracy of over 90% on the test set. In a practical application, the predictor finds that the actual energy consumption on a Tuesday afternoon is 30% higher than expected and lasts for 4 hours. The anomaly report identifies it as excessive cooling caused by improper air-conditioning parameter settings. Based on this report, the system generates control instructions to adjust the air-conditioning set temperature and air flow direction. After executing the instructions, the energy consumption in this area is reduced to the normal level, and the optimal operating parameters of this area are updated in the feature library.

[0035] In the embodiments of the present application, by collecting building temperature, humidity, energy consumption, and occupancy density data through multi-dimensional sensors, a comprehensive basic dataset of building energy consumption is obtained, providing a rich data foundation for subsequent analysis and solving the problem of single data source in traditional methods. By filling in missing data and processing outliers in the basic dataset of building energy consumption, extracting building usage periods, environmental parameters, and energy load characteristics, and generating building energy consumption feature vectors, the data quality and feature expression ability are significantly improved, providing high-quality input for model training. Based on the building energy consumption feature vectors, a building energy consumption prediction model considering time factors and spatial relationships is established, fully capturing the spatio-temporal correlation of building energy consumption, overcoming the limitation of traditional models that ignore the mutual influence of energy consumption between regions, and improving the prediction accuracy. Using the building energy consumption predictor to calculate the expected energy consumption value, comparing the actual energy consumption data with the expected energy consumption value, identifying the energy consumption abnormal interval, and generating a building energy consumption abnormal report, realizing the transformation from passive discovery to active early warning, and greatly shortening the abnormal discovery time. Based on the building energy consumption abnormal report, combined with the building usage plan and environmental conditions, a highly targeted energy optimization parameter adjustment plan is formulated, generating building energy-saving control instructions, realizing precise energy saving, and avoiding the problem of reduced comfort caused by the "one-size-fits-all" approach in traditional methods. By executing the building energy-saving control instructions, recording the energy consumption change data before and after adjustment, calculating the energy efficiency improvement index, and updating the building energy consumption feature library, the closed-loop of building energy consumption analysis is completed, realizing self-learning and continuous optimization. This solution makes full use of the advantages of artificial intelligence algorithms in the field of specific building energy consumption management, especially in spatio-temporal data processing. By using bidirectional long short-term memory networks to capture the time-dependent relationship of energy consumption, graph convolutional networks to model the spatial correlation between building regions, and attention mechanisms to dynamically adjust the importance weights of different features, these algorithm features contribute to the ability to adaptively process the dynamic change characteristics of building energy consumption, identify complex energy consumption patterns, and thus achieve more accurate prediction and more refined control. In addition, by using genetic algorithms to solve multi-objective optimization problems, the three dimensions of energy saving, cost, and comfort are balanced, realizing a globally optimal energy management strategy. The incremental learning mechanism enables the model to continuously learn from new data, improve the prediction accuracy, and adapt to the long-term evolution of building energy consumption characteristics.

[0036] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0037] Collecting the temperature change data of each region through a temperature sensor network arranged according to the building functional zones to form a temperature spatio-temporal data matrix;

[0038] Collecting the humidity distribution data inside the building through multi-point humidity sensors and generating a humidity environment feature set in combination with the ventilation state parameters;

[0039] Collect sub-item energy consumption data of the main energy-consuming equipment in the building through an energy metering device, record the equipment operation power curve, and construct an equipment energy consumption database;

[0040] Collect data on the personnel flow and density in the building through a personnel detection device, and establish a personnel activity pattern feature table;

[0041] Perform time synchronization processing on the temperature spatio-temporal data matrix, humidity environment feature set, equipment energy consumption database, and personnel activity pattern feature table, eliminate invalid data points, supplement missing time period data, and generate a building energy consumption basic data set;

[0042] Perform normalization processing on various types of data in the building energy consumption basic data set to eliminate the dimension difference and obtain the building energy consumption basic data set.

[0043] Specifically, when collecting the temperature change data of each area through a temperature sensor network arranged according to the building function zones, install temperature sensors in different functional areas of the building, such as office areas, meeting rooms, corridors, equipment rooms, etc. These sensors are connected to the data acquisition gateway by RS-485 bus or wireless communication method. Multiple temperature measurement points are arranged in each area. Usually, sensors are installed at the four corners and the center of each room, and the acquisition frequency is set to once every 5 minutes. The collected temperature data is organized in two dimensions of time and space to form a temperature spatio-temporal data matrix. The rows of this matrix represent different time points, the columns represent different spatial positions, and the matrix element values are the temperature values corresponding to the time points and positions. When collecting the humidity distribution data in the building through multi-point humidity sensors, the humidity sensors are usually installed at the same positions as the temperature sensors, with the same acquisition frequency and an accuracy of ±1%RH. In addition to the humidity data, it is also necessary to collect the operating status of the building ventilation system, including parameters such as fresh air volume, return air ratio, and fan speed. The humidity data is combined with the ventilation status parameters for processing to calculate the air humidity change rate, absolute humidity value, and the correlation coefficient between humidity and ventilation status, thereby generating a humidity environment feature set. This feature set not only contains the original humidity data but also information on the humidity change trend and humidity adjustment ability.

[0044] When collecting the sub-item energy consumption data of the main energy-consuming equipment in the building through energy metering devices, it is necessary to install power monitoring devices, such as smart meters and sub-item metering instruments, on the main circuits and main electrical equipment of the building. These devices monitor the power consumption of each device in real time with an accuracy of ±0.5%. The monitoring data includes parameters such as the real-time power, cumulative power consumption, and power factor of the equipment. The power change curve is recorded according to the operating status of the equipment, and the start-stop characteristics, steady-state operating power, peak power and other characteristics of the equipment are analyzed. The equipment is classified and sorted according to multiple dimensions such as equipment type, functional area, and use period to build an equipment energy consumption database. Each record in the database contains fields such as equipment ID, timestamp, power value, cumulative power consumption, and operating status identification. When collecting personnel flow and density data in the building through personnel detection devices, various technical means such as infrared sensors, camera population counting systems, and WIFI probes are mainly used. These devices are installed at the entrances and exits of the building and the main passages to record the entry and exit of personnel and the density of personnel in the area in real time. The collected data is initially processed to form original data records containing information such as time, location, number of people, and moving direction. By analyzing these records, we extract the characteristics of personnel activities in each area, such as peak hours, activity paths, length of stay, etc., and establish a personnel activity pattern feature table. This feature table describes the regularity of personnel activities in each area in different time periods.

[0045] When performing time synchronization processing on the temperature spatiotemporal data matrix, humidity environment feature set, equipment energy consumption database, and personnel activity pattern feature table, it is first necessary to ensure that all data have timestamps in a standard format. The first step in time synchronization processing is to align data from different sources according to timestamps and unify them to the same time granularity, such as 15 minutes as a time unit. In this process, for data with a collection frequency higher than the unified granularity, the average value in the time window is used as a representative; for data with a collection frequency lower than the unified granularity, the interpolation method is used to supplement the data at the intermediate time points. The second step is to identify and eliminate invalid data points, including outliers caused by sensor failures and missing values caused by communication interruptions. The criteria for determining invalid data include situations where the value exceeds a reasonable range, fluctuates too much in a short period of time, and remains completely unchanged for a long time. The third step is to supplement the missing period data. For short-term data missing, linear interpolation or the average value of the previous and next data is used to fill in; for long-term data missing, data from the same period of similar dates are used to replace them. After these processes, a basic building energy consumption data set with a continuous time dimension and corresponding data time of various types is obtained.

[0046] When normalizing various types of data in the building energy consumption basic dataset, it is to eliminate the dimensional differences between different types of data and facilitate subsequent comprehensive analysis. The min-max normalization method is adopted for normalization, mapping all data into the interval [0, 1]. For each type of data, first determine its historical maximum and minimum values, and then convert according to the formula: normalized value = (original value - minimum value) / (maximum value - minimum value). After such processing, different types of data are mapped into the same numerical range, ensuring the comparability of the data. At the same time, to cope with the situation where new data may exceed the historical range, for data beyond the original range, truncation processing is adopted. Data exceeding the maximum value is normalized to 1, and data below the minimum value is normalized to 0. After the normalization process, a building energy consumption basic dataset with unified dimensions is obtained, laying a foundation for subsequent feature extraction and model training.

[0047] Taking the data collection process of an office building as an example, the building is divided into four functional areas: administrative office area, technology R & D area, meeting area, and public area. A temperature sensor network is installed in each area, with a total of 86 temperature measurement points. The temperature data is recorded every 5 minutes, forming a temperature spatio-temporal data matrix with 86 columns. At the same time, the humidity sensor is connected to the ventilation system controller to record the humidity value and ventilation parameters, generating a humidity environment feature set including features such as humidity change rate and absolute humidity. The main energy-consuming equipment includes central air conditioning, lighting system, office equipment, and elevators. Power monitoring devices are installed in each system to record the energy consumption data every 15 minutes, constructing a sub-item equipment energy consumption database. Personnel activities are monitored by infrared counters installed at entrances, exits, and main passages, recording the change in personnel density in different areas and establishing a personnel activity pattern feature table. The four types of data are synchronized through timestamps to a 15-minute granularity. The missing data of 8 temperature points and 3 humidity points are filled using linear interpolation method, and the abnormal energy consumption data points are screened by setting thresholds. Finally, all data is mapped into the interval [0, 1] through the min-max normalization method, forming a building energy consumption basic dataset.

[0048] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0049] Fill the missing values in the building energy consumption basic dataset through the sliding time window average filling method to obtain a preliminary filled dataset;

[0050] Identify the outliers in the preliminary filled dataset through the improved Z-score method, and replace the identified outliers with the median of the front and back data points to generate a smoothed dataset;

[0051] Perform time series segmentation on the smoothed dataset, divide it into working periods, transitional periods, and idle periods according to the characteristics of building energy consumption load changes, and form a period division table;

[0052] Perform noise reduction processing on the smoothed dataset through wavelet transform method, separate the trend term, seasonal term, and residual term of building energy consumption, and construct energy consumption decomposition features;

[0053] Extract five environmental parameter features of temperature difference, thermal inertia index, sunshine influence coefficient, humidity change rate, and personnel density distribution from the smoothed dataset to generate an environmental impact factor set;

[0054] Integrate the period division table, energy consumption decomposition features, and environmental impact factor set through a feature fusion matrix to generate a building energy consumption feature vector.

[0055] Specifically, fill the missing values in the building energy consumption basic dataset through the sliding time window average filling method. This method selects the valid data within a certain time range before and after the missing point to calculate the average value for filling the missing point. When implementing specifically, set a window size W, usually 24 hours. For the missing point t, search for the valid data points within the time range from t - W / 2 to t + W / 2, and calculate the average value of these data points as the filling value for the missing point. When the continuous missing exceeds the set threshold, for example, more than 6 hours, the similar day pattern substitution method is adopted, that is, search for the historical data at the same time period of the same type of day (working day or rest day) for filling. This sliding window average filling method takes into account the time continuity and periodicity characteristics of the data, and the preliminary filled dataset retains the basic change trend of the original data. When identifying outliers in the preliminary filled dataset through the improved Z - score method, calculate the deviation degree of each data point from its surrounding data. The ordinary Z - score method divides the difference between the data point and the mean by the standard deviation, while the improved Z - score method considers the time characteristics of the data and uses the mean and standard deviation within the local time window to judge. The specific implementation is that for each data point x(t), calculate the mean μ and standard deviation σ of the data within its time window [t - W / 2, t + W / 2], and then calculate the Z - score Z(t) = (x(t) - μ) / σ. When |Z(t)| exceeds the preset threshold (usually set to 3.0), it is determined as an outlier and replaced with the median of the data within the same time window. This method can effectively identify sudden data anomalies and avoid misjudging normal seasonal fluctuations as anomalies. The replaced dataset is called the smoothed dataset, which eliminates the sudden anomalies in the original data and is more suitable for subsequent pattern analysis.

[0056] Time series segmentation of the smoothed dataset is to identify the energy consumption behavior patterns in different periods. According to the characteristics of building energy consumption load changes, a day is usually divided into working periods, transitional periods, and idle periods. The working period refers to the time period when the building is operating normally and there is frequent human activity, such as 9:00 - 18:00 in an office building; the transitional period refers to the time period when the building changes from the idle state to the working state or from the working state to the idle state, such as 7:00 - 9:00 and 18:00 - 20:00; the idle period refers to the time period when the building is basically unoccupied, such as 20:00 - 7:00. The segmentation process determines the start and end times of each period by analyzing the comprehensive change characteristics of energy consumption data, personnel density data, and equipment operation status. For each working day and non - working day, a period division table is established respectively, recording the start and end times of different periods and the statistical values of energy consumption characteristics in each period, such as average value, standard deviation, maximum value, etc.

[0057] Noise reduction processing of the smoothed dataset by wavelet transform method is an important step in time - series data analysis. Wavelet transform is a multi - resolution analysis method that can represent signals in both the time - frequency domain. In building energy consumption analysis, the Daubechies wavelet basis is used and the decomposition level is 5 layers. The energy consumption time - series data is decomposed into coefficients of multiple frequency bands through wavelet transform. The high - frequency coefficients correspond to short - term fluctuations (residual terms), the medium - frequency coefficients correspond to periodic changes (seasonal terms), and the low - frequency coefficients correspond to long - term trends (trend terms). Noise is removed by threshold processing (such as soft threshold or hard threshold) of the coefficients of each frequency band, and then the signal is reconstructed to obtain the noise - reduced data. Based on the results of wavelet decomposition, energy consumption decomposition features can be constructed, including trend features (reflecting the long - term energy consumption change trend), seasonal features (reflecting the periodic energy consumption pattern), and residual features (reflecting random perturbations). These features reveal the multi - scale change law of building energy consumption.

[0058] The mathematical expression of wavelet transform is as follows:

[0059]

[0060] Among them, W ψ f(s, ∈) is the wavelet coefficient of the function f(t) at scale s and translation ∈, ψ * is the complex conjugate of the wavelet function, and f(t) represents the energy consumption time series. Through decomposition at different scales, a multi - scale representation of energy consumption data can be obtained:

[0061]

[0062] Among them, A J (t) is the approximation part (trend term) of the J - layer wavelet decomposition, D j (t) is the detail part of the j - th layer, and J is the total decomposition level. The seasonal term usually corresponds to D of the middle several layersj (t), and the residual term is the detail part of the highest frequency.

[0063] Extracting environmental parameter features from the smoothed dataset is the key to establishing the relationship between energy consumption and environmental factors. The temperature difference feature calculates the difference between indoor and outdoor temperatures, reflecting the intensity of heat conduction; the thermal inertia index represents the response speed of the building to temperature changes, calculated through the time delay between indoor temperature changes and external temperature changes; the solar radiation influence coefficient is obtained by analyzing the correlation between energy consumption and solar radiation intensity; the humidity change rate calculates the change in humidity per unit time; the personnel density distribution calculates the personnel density and its time-varying characteristics in each area based on personnel detection data. These five environmental parameter features together constitute the environmental impact factor set, comprehensively characterizing the impact of environmental factors on building energy consumption.

[0064] The calculation of the environmental impact factor set involves multiple formulas. Taking the thermal inertia index H_I as an example, its calculation formula is:

[0065]

[0066] where, ΔT in is the change in indoor temperature, ΔT out is the change in external temperature, and Δt delay is the time when the indoor temperature change lags behind the external temperature change. The calculation formula for the solar radiation influence coefficient S_C is:

[0067]

[0068] where, cov(E, I s ) is the covariance between energy consumption E and solar radiation intensity I s , and σ E and are the standard deviations of energy consumption and solar radiation intensity respectively. The time period division table, energy consumption decomposition features, and environmental impact factor set are integrated through the feature fusion matrix to generate the building energy consumption feature vector. The feature fusion matrix is a weight matrix used to adjust the importance of different types of features in the final feature vector. The fusion process first standardizes each type of feature and then generates the final feature vector through a weighted combination method. The formula for feature fusion is:

[0069] F = [W T ·T, W D ·D, W E ·E]

[0070] where, F is the final feature vector, T is the time period feature, D is the decomposition feature, E is the environmental feature, and W T , W D , W EThey are the corresponding weight coefficient matrices respectively. The finally generated building energy consumption feature vector is a high-dimensional vector, which contains the time characteristics, decomposition characteristics and environmental correlation characteristics of building energy consumption, providing comprehensive feature inputs for subsequent prediction models.

[0071] Taking a commercial building as an example, in the basic data set of weekly energy consumption obtained by multi-dimensional sensors, it is found that there are some missing data in the air-conditioning system data, specifically, the 4-hour data from 10:00 to 14:00 on Tuesday. The sliding time window average filling method is adopted, with the window size set to 24 hours, and the average values of the same time periods from 10:00 to 14:00 on Monday and from 10:00 to 14:00 on Wednesday are calculated for filling. An abnormal energy consumption peak in the afternoon of Thursday is identified in the filled data, and the Z-score calculation result is 4.2, exceeding the set threshold of 3.0. This value is replaced with the median within the same time window. After time series analysis of the processed smoothed data set, it is determined that the working period of this building is from 8:00 to 19:00, the transition periods are from 7:00 to 8:00 and from 19:00 to 21:00, and the idle period is from 21:00 to 7:00. The energy consumption data is decomposed by wavelet transform to extract the trend term reflecting the long-term energy consumption growth, the seasonal term reflecting the difference between weekdays / weekends, and the residual term reflecting random disturbances. At the same time, the average indoor-outdoor temperature difference is calculated to be 6.2 °C from the environmental data, the thermal inertia index is 1.8 hour·°C / °C, the sunshine influence coefficient is 0.72, the average humidity change rate is 1.4% / hour, and the peak period of personnel density is from 10:00 to 15:00. These features are integrated through the feature fusion matrix to generate a complete building energy consumption feature vector, which characterizes the time variation law of building energy consumption and environmental influencing factors.

[0072] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0073] Perform vector segmentation on the building energy consumption feature vector, dividing it into a time feature sub-vector representing time characteristics and a spatial feature sub-vector representing spatial distribution;

[0074] Perform bidirectional long short-term memory network processing on the time feature sub-vector, calculate the hidden state through forward and backward time steps, capture long-term dependencies, extract the temporal variation law of building energy consumption, and form a time relationship feature matrix;

[0075] Perform graph convolution operation on the spatial feature sub-vector, represent each area of the building as a node in the graph, represent the physical connection relationship between areas as edges, calculate the node feature update through the message passing mechanism, capture the energy consumption interaction relationship between different areas of the building, and generate a spatial association feature graph;

[0076] Fuse the time-relation feature matrix and the spatial-correlation feature map through an attention mechanism, calculate the weight coefficients of each feature channel, perform weighted combination, and construct a spatio-temporal fusion feature representation;

[0077] Pair the spatio-temporal fusion feature representation with the historical building energy consumption data to form a training sample set, iteratively optimize the parameters through the mini-batch gradient descent algorithm, calculate the gradient using a loss function that combines the weighted mean absolute percentage error and the root mean square error, update the connection weights, and at the same time introduce L2 regularization to prevent overfitting, forming an energy consumption prediction calculation framework;

[0078] Divide the data set into K subsets through the K-fold cross-validation method, alternately use K - 1 subsets for training and 1 subset for validation, evaluate the prediction stability and accuracy of the energy consumption prediction calculation framework under different data partitions, and select the model parameter configuration with the optimal validation performance as the building energy consumption predictor.

[0079] Specifically, the building energy consumption feature vector is vector-segmented into a time feature sub-vector representing time characteristics and a spatial feature sub-vector representing spatial distribution. This segmentation is based on the nature of the features. The time feature sub-vector contains the temporal variation characteristics of energy consumption, such as the time series of historical energy consumption values, working / non-working day markers, time-of-day markers, seasonal cycle characteristics, etc.; the spatial feature sub-vector contains the spatial layout of the building and the characteristics of each area, such as the area of each area, function type, relative position relationship, equipment distribution, etc. The vector segmentation uses the feature attribute classification method, which classifies each feature into the corresponding sub-vector according to its physical meaning to ensure the integrity and independence of the information. The bidirectional long short-term memory network processing of the time feature sub-vector is to capture the temporal variation law of building energy consumption. The bidirectional long short-term memory network (BiLSTM) is a special recurrent neural network that can learn sequence information from both the past and the future directions. During the processing, the time feature sub-vector is used as the input sequence, and the network consists of a forward LSTM layer and a backward LSTM layer. The forward LSTM layer processes the sequence from the beginning to the end, capturing the information flow from the past to the current; the backward LSTM layer processes the sequence from the end to the beginning, capturing the information flow from the future to the current. The LSTM layers in both directions calculate their hidden states respectively, and then these hidden states are concatenated to form a complete representation. Specifically, for each time point, the forward LSTM calculates the forward hidden state, and the backward LSTM calculates the backward hidden state. Finally, the output at this time point is the concatenation of the two. Through this bidirectional processing, the network can comprehensively capture long-term dependencies, extract the periodic, trend, and sudden change characteristics of building energy consumption, and form a time relationship feature matrix containing rich temporal information. The graph convolution operation on the spatial feature sub-vector is to capture the energy consumption interaction relationship between different regions of the building. The graph convolutional network (GCN) is a neural network specifically designed for processing graph-structured data. In building energy consumption analysis, each functional area of the building is represented as a node in the graph, and the node feature is the spatial feature sub-vector of this area; the physical connection relationship between regions (such as adjacent, shared equipment, etc.) is represented as an edge in the graph, and an adjacency matrix is constructed. The graph convolution operation updates the node features through a message passing mechanism. Each node not only considers its own features but also the information of its neighbor nodes. During the specific calculation process, for each node, its feature update is achieved by aggregating its own features and neighbor features. The stacking of multiple layers of graph convolution enables each node to perceive information in a wider range, expanding from directly adjacent nodes to the global relationship of the entire building. Through the graph convolution operation, a spatial association feature map is generated. This feature map not only contains the characteristics of each area itself but also encodes the energy consumption influence propagation relationship between regions.

[0080] Feature fusion of the time relationship feature matrix and the spatial association feature map through the attention mechanism is to comprehensively consider the information in both the time and space dimensions. The core idea of the attention mechanism is to calculate the importance weights of different features and then perform weighted combination according to the weights. In the specific implementation, first, the time relationship feature matrix and the spatial association feature map are converted into a unified form, and then their attention scores are calculated. The calculation process includes: calculating the query vector, key vector, and value vector for each feature channel; calculating the similarity between the query vector and the key vector to obtain the attention score; normalizing the attention score to obtain the weight coefficient; and finally, performing weighted summation on the value vector using the weight coefficient. In this way, the importance of time features and spatial features in different situations can be adaptively adjusted, and the fused features form a spatio-temporal fusion feature representation, which not only retains the dynamic change information of the time series but also contains the structured knowledge of the spatial layout.

[0081] Pairing the spatio-temporal fusion feature representation with the historical building energy consumption data to form a training sample set is the preparatory work for model training. Each training sample consists of an input feature (spatio-temporal fusion feature representation) and a target value (the actual energy consumption value at the corresponding time point). The training process uses the mini-batch gradient descent algorithm to iteratively optimize the parameters. Each time, a small batch of samples (usually 64) is randomly selected from the training sample set, the error between the model prediction value and the true value is calculated, and then the gradient is calculated through the backpropagation algorithm and the model parameters are updated. The loss function adopts a combined form of weighted mean absolute percentage error and root mean square error. The former focuses on the relative error and is suitable for energy consumption predictions of different magnitudes; the latter focuses on the absolute error and is more sensitive to outliers. At the same time, an L2 regularization term is introduced by adding a penalty term of the sum of squares of the parameters to the loss function to prevent the model from overfitting. During the training process, the parameters are gradually adjusted according to the set learning rate until the loss function converges or reaches the maximum number of training epochs, forming an energy consumption prediction calculation framework.

[0082] Evaluating the model performance through the K-fold cross-validation method is to ensure the generalization ability of the model. The K-fold cross-validation randomly divides the complete dataset into K subsets of similar sizes, usually with K set to 5 or 10. During the validation process, each time one subset is selected as the validation set, and the remaining K - 1 subsets are combined as the training set to train the model and evaluate its performance on the validation set. This process is repeated K times to ensure that each subset is used as the validation set once. Finally, the average performance metric of the K validations is calculated to obtain the comprehensive performance of the model under different data partitions. By comparing the cross-validation results under different parameter configurations, the model parameter configuration with the optimal validation performance is selected as the final building energy consumption predictor. This validation method can comprehensively evaluate the stability and adaptability of the model and avoid performance biases caused by specific data partitions.

[0083] Taking a commercial complex as an example, the building is divided into four functional areas: a retail area, an office area, a dining area, and a public area. The time feature sub-vector (including the 24-hour energy consumption change curve, weekday markers, season indicators, etc.) and the spatial feature sub-vector (including the area, orientation, equipment configuration, etc. of each area) are segmented from the building energy consumption feature vector. The time feature sub-vector is input into a three-layer BiLSTM network, with the number of units in each layer set to 128, 256, and 128 respectively, capturing the energy consumption change patterns at the hourly, daily, and weekly levels, and forming a feature matrix reflecting the time-dependent relationship. At the same time, the four functional areas of the building are constructed into a graph structure, and the weights of the edges between nodes are determined according to the physical distance between areas and the connection relationship of the energy system. Through processing by a three-layer graph convolutional network, the energy consumption influence propagation mechanism between areas is captured, such as the influence of air conditioner use in the office area on the temperature of adjacent public areas. The time and space features are fused through an attention layer, which automatically learns the pattern that the spatial factor weight is higher during the daytime on weekdays, while the time factor weight is higher at night and on weekends. The fused features are paired with the actual energy consumption data and trained using the mini-batch gradient descent algorithm with a batch size of 64. The weights of the root mean square error and the mean absolute percentage error in the loss function are set to 0.3 and 0.7 respectively. Finally, the performance of different network configurations is evaluated through 10-fold cross-validation, and the parameter combination with the smallest validation error is selected as the final model, which can accurately predict the energy consumption changes at different times and areas.

[0084] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0085] Input the current building parameter data into the building energy consumption predictor, and obtain the expected energy consumption values at each time point through forward calculation to form an expected energy consumption sequence;

[0086] Collect the actual building energy consumption data in the current period, align it according to the same time granularity as the expected energy consumption sequence to obtain the actual energy consumption sequence;

[0087] Calculate the difference between the actual energy consumption sequence and the expected energy consumption sequence to generate an energy consumption residual sequence;

[0088] Analyze the energy consumption residual sequence through a dynamic threshold detection algorithm, calculate the threshold τ value at each time point, and identify the data points with the absolute value of the residual exceeding the corresponding threshold as abnormal points;

[0089] Conduct a continuity analysis on the abnormal points, mark the intervals where three or more abnormal points continuously appear in time as energy consumption abnormal intervals, and record the start time, duration, maximum deviation value, and cumulative deviation amount of the abnormality;

[0090] Based on the characteristic parameters of the abnormal energy consumption interval, combined with the decision tree algorithm, classify the abnormal interval, identify the abnormal type, and generate a building energy consumption anomaly report according to the abnormal location, abnormal type, severity, and recommended handling methods.

[0091] Specifically, input the current building parameter data into the building energy consumption predictor, and obtain the expected energy consumption values at each time point through forward calculation to form an expected energy consumption sequence. Forward calculation means that the input data is passed layer by layer from the input layer to the output layer through the trained neural network model to obtain the final prediction result. Specifically, the current building parameter data includes the predicted values of temperature, humidity, personnel density, and building usage plan, etc., for the current and a period of time in the future. These data are sorted in the same format as the training data and then input into the predictor. The predictor calculates through each layer of the network and finally outputs the expected energy consumption values at each time point within a period of time in the future (such as 24 hours). These values are arranged in chronological order to form an expected energy consumption sequence. Collect the actual building energy consumption data for the current period, align it according to the same time granularity as the expected energy consumption sequence, and obtain the actual energy consumption sequence. The collection process records the energy consumption data such as actual electricity consumption, water consumption, and gas consumption in real time through the metering equipment of the building energy management system. Since different energy metering devices may have different sampling frequencies, time alignment processing is required to unify all data to the same time granularity (such as 15 minutes, 30 minutes, or 1 hour). The alignment processing includes downsampling the high-frequency data (such as taking the average value) or interpolating the low-frequency data to ensure that the actual energy consumption sequence corresponds one by one with the expected energy consumption sequence at each time point.

[0092] Calculate the difference between the actual energy consumption sequence and the expected energy consumption sequence to generate an energy consumption residual sequence. The residual calculation formula is: Residual = Actual value - Expected value. The residual value at the corresponding time point reflects the deviation degree between the actual energy consumption and the expected energy consumption. A positive residual indicates that the actual energy consumption is higher than expected, and there may be energy waste; a negative residual indicates that the actual energy consumption is lower than expected, and it may be that the equipment is not operating normally or the utilization rate is lower than expected. The residual sequence retains the time attributes of the original sequence, facilitating subsequent time series analysis.

[0093] Analyze the energy consumption residual sequence through the dynamic threshold detection algorithm, and calculate the threshold τ value at each time point. The threshold calculation formula is:

[0094] τ(p) = μ p +k p ·σ p

[0095] where τ(p) represents the dynamic threshold at time point p, μ p represents the mean value of the residual sequence within the time window near time point p, σ p represents the standard deviation within this window, and k pis an adaptive adjustment coefficient, which is automatically adjusted according to different time periods and different building functional areas. Compared with the traditional fixed threshold, the dynamic threshold can adapt to the time-varying characteristics of energy consumption data, setting a looser threshold during high-fluctuation periods and a stricter threshold during stable periods. When the absolute value of the residual |R(p)| exceeds the corresponding threshold k p , this data point is identified as an outlier. Continuity analysis is performed on the identified outliers, and the intervals where three or more outliers appear continuously in time are marked as energy consumption abnormal intervals. Continuity judgment is to filter out occasional random fluctuations and only focus on continuous abnormal states. For each abnormal interval, multiple characteristic parameters are recorded: the start time of the anomaly (the timestamp of the first outlier), the duration (the time length from the first outlier to the last outlier), the maximum deviation value (the maximum value of the absolute value of the residuals within the interval), and the cumulative deviation amount (the sum of the absolute values of all residuals within the interval). These characteristic parameters comprehensively describe the characteristics of the abnormal interval and provide a basis for subsequent anomaly classification.

[0096] Based on the characteristic parameters of the energy consumption abnormal interval and combined with the decision tree algorithm, the abnormal interval is classified to identify the type of anomaly. The decision tree algorithm is a typical classification method that divides the abnormal interval into different categories by constructing a tree structure. The construction of the decision tree is based on expert knowledge and historical anomaly cases. Each internal node of the tree represents a test on a certain characteristic (such as "whether the duration exceeds 2 hours"), each branch represents the possible result of the test (yes / no), and each leaf node represents the classification result (type of anomaly). The decision tree continuously divides the characteristics to classify the abnormal interval into different types such as equipment failure, control parameter deviation, abnormal energy consumption behavior, or external environment impact. Each type can be further divided into multiple sub-categories. The classification result, together with information such as the location, severity, and recommended handling method of the abnormal interval, forms a complete building energy consumption anomaly report to guide subsequent energy consumption optimization.

[0097] Taking the air conditioning system of an office building as an example, on a working day in summer, parameters such as the temperature forecast data of the day, the expected occupancy rate of personnel, and the meeting arrangements are input into the building energy consumption predictor to obtain the hourly expected air conditioning energy consumption values, forming an expected energy consumption sequence. At the same time, the actual air conditioning power consumption per hour is recorded in real time through the building energy management system to form an actual energy consumption sequence. Calculate the difference between the two sequences and find that during the period from 10 am to 3 pm, the residual values are continuously positive and large, indicating that the actual energy consumption is significantly higher than the expected value. Apply the dynamic threshold detection algorithm to the residual sequence. First, calculate the mean and standard deviation of different time periods. For example, the residual mean of the 10-11 am period is 2.1 kWh, and the standard deviation is 0.8 kWh. Set the adjustment coefficient to 2.5 according to the empirical value to obtain the dynamic threshold of this period as 4.1 kWh. The time points where the residual values exceed the threshold are marked as abnormal points. Continuous analysis finds that 18 abnormal points continuously appear from 10:15 to 14:45, constituting an energy consumption abnormal interval. Record the characteristic parameters such as the start time (10:15), duration (4.5 hours), maximum deviation value (6.2 kWh), and cumulative deviation amount (82.3 kWh) of this interval. Analyze these characteristics through the decision tree algorithm. This algorithm first judges the abnormal duration (>4 hours, go to the left branch), then judges whether it coincides with the outdoor temperature peak (yes, go to the right branch), and then judges the building area (central area, go to the middle branch). Finally, classify this anomaly as the type of "excessive air conditioning cooling caused by too low temperature set point", generate an anomaly report including the anomaly location, type, severity, and recommended treatment method (adjust the temperature set point up by 2°C), and provide accurate energy consumption anomaly diagnosis and treatment suggestions for building managers.

[0098] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0099] Extract the anomaly type and energy consumption deviation data from the building energy consumption anomaly report to construct an energy consumption optimization target list;

[0100] Collect building usage plan data, including work period arrangements, personnel occupancy plans, and special event times, to form a building usage time sequence table;

[0101] Obtain future environmental condition forecasts, including temperature change trends, humidity distributions, and light intensities, to generate an environmental impact factor prediction sequence;

[0102] Construct a multi-objective optimization function, including three dimensions of energy consumption index, economic cost index, and comfort index, set the weight coefficients of each index, and solve the optimal parameter combination through the genetic algorithm to obtain the equipment operation parameter adjustment plan;

[0103] Decompose the equipment operation parameter adjustment plan into equipment-level control parameters, including air-conditioning temperature set values, fresh air ratio adjustment parameters, lighting intensity parameters, and equipment start / stop times, and generate an equipment control parameter table;

[0104] Convert the equipment control parameter table into an instruction format compliant with the building control system communication protocol to form building energy-saving control instructions.

[0105] Specifically, extract the abnormal types and energy consumption deviation data from the building energy consumption anomaly report to construct an energy consumption optimization target list. The abnormal types generally include four categories: equipment failures, control parameter offsets, abnormal energy consumption behaviors, and external environmental impacts. Each type is further divided into multiple sub-categories. For example, under the control parameter offset category, there are sub-categories such as too low air-conditioning temperature set point and too high lighting intensity. The energy consumption deviation data is the difference value between the actual energy consumption and the expected energy consumption within the abnormal interval, including indicators such as deviation amount, deviation rate, and duration. The energy consumption optimization target list is prioritized according to factors such as abnormal type, energy consumption deviation amount, duration, and occurrence frequency. Abnormalities with large energy consumption deviations, long durations, and frequent occurrences are processed first. Collecting building usage plan data to form a building usage time sequence table is to ensure that energy-saving optimization does not affect the normal usage functions of the building. The building usage plan data is sourced from the scheduling information of the building management system, including work period arrangements (such as office hours from 9:00 to 18:00), personnel occupancy plans (such as the expected number and distribution in each area), and special event times (such as meetings, exhibitions, and other temporary events). These data are sorted in chronological order to form a building usage time sequence table for a future period (usually 7 days). This time sequence table marks the usage status of the building at different times, providing constraint conditions for energy optimization and ensuring energy-saving adjustments are made while meeting usage requirements.

[0106] Obtaining future environmental condition forecasts and generating a prediction sequence of environmental impact factors is a key step in considering the impact of the external environment on building energy consumption. Future environmental condition forecasts mainly come from weather forecast data of meteorological departments, including temperature change trends (such as hourly temperature forecasts for the next few days), humidity distributions (such as relative humidity change curves), and lighting intensities (such as sunshine duration and intensity). These original meteorological data are processed and converted into environmental impact factors directly related to building energy consumption, such as temperature difference factors (indoor-outdoor temperature difference), humidity factors (indoor-outdoor humidity difference), and lighting factors (availability of natural daylight). The prediction sequence of environmental impact factors is arranged with the same time granularity as the building usage time sequence table, providing environmental background information for the energy optimization algorithm. Constructing a multi-objective optimization function is the core link of energy optimization. The multi-objective optimization function includes three dimensions: energy consumption index, economic cost index, and comfort index, and comprehensively considers these three objectives through weighted summation. The mathematical expression of this optimization function is:

[0107] F = γ1·Econs +γ2·C econ +γ3·D comf

[0108] Among them, F is the comprehensive optimization objective value that needs to be minimized; E cons is the energy consumption index, representing the energy usage per unit time; C econ is the economic cost index, considering the electricity price differences at different times and the equipment operation costs; D comf is the comfort index, quantifying the deviation degree of indoor environmental parameters (temperature, humidity, lighting, etc.) from the comfort standard; γ1, γ2, γ3 are the weight coefficients of the three indexes, satisfying γ1 + γ2 + γ3 = 1, and are dynamically set according to the building type and management requirements. The multi-objective optimization problem is solved by the genetic algorithm to obtain the equipment operation parameter adjustment scheme. The genetic algorithm is an optimization algorithm inspired by biological evolution, searching for the optimal solution by simulating natural selection and genetic mechanisms. The algorithm process includes: initializing the population (randomly generating multiple groups of equipment parameter combinations); evaluating the fitness (calculating the scores of each group of parameters under the optimization function); selection operation (retaining the individuals with high fitness); crossover operation (two individuals exchange part of the parameters to generate new individuals); mutation operation (randomly changing some parameters of the individuals to increase diversity); termination condition judgment (stop when reaching the maximum number of generations or meeting the convergence condition, otherwise return to the step of evaluating the fitness). After multiple generations of evolution, the algorithm finally converges to a solution close to the global optimum, forming the equipment operation parameter adjustment scheme. The equipment operation parameter adjustment scheme is decomposed into equipment-level control parameters to generate the equipment control parameter table. The equipment operation parameter adjustment scheme is an overall optimization result and needs to be converted into specific equipment parameters that can be directly controlled. The control parameters include the air-conditioning temperature setting value (such as the return air temperature of the central air-conditioning is set to 24°C), the fresh air ratio adjustment parameter (such as the opening degree of the fresh air valve is set to 30%), the lighting intensity parameter (such as the brightness of the dimming system is set to 80%), and the equipment start-stop time (such as the pre-cooling start time of the air-conditioning system is advanced to 8:00), etc. Each control parameter has a corresponding value range and adjustment step, and the overall optimization objective needs to be achieved on the premise of meeting the equipment operation constraints. The equipment control parameters are sorted according to information such as equipment type, control point ID, parameter name, parameter value, and execution time to form the equipment control parameter table.

[0109] Convert the device control parameter table into an instruction format compliant with the communication protocol of the building control system to form building energy-saving control instructions. Different building control systems may use different communication protocols, such as BACnet, Modbus, or LonWorks, etc. The conversion process needs to map each control parameter to the corresponding protocol command according to the protocol specifications of the target system, including fields such as device address, function code, data area, and check code. The building energy-saving control instructions are a set of finally executable commands, which are sent to each controller and execution device through the communication network of the building automation system to achieve optimized control of building energy consumption.

[0110] Taking a commercial office building as an example, an anomaly of too low temperature setpoint of the air-conditioning system was extracted from the energy consumption anomaly report. During the working hours from 9:00 to 18:00, the actual energy consumption was 15% higher than expected. An optimization list with the main goal of adjusting the air-conditioning temperature setpoint was constructed. By collecting the building usage plan, it was found that there were two important meetings scheduled from 10:00 to 12:00 the next morning, and the personnel density would be 30% higher than usual. The period from 14:00 to 17:00 in the afternoon was in the normal office state. At the same time, the weather forecast showed that the temperature would gradually rise the next day, from 22°C in the morning to 30°C in the afternoon, and the relative humidity was stable at about 65%. Based on this information, a multi-objective optimization function was constructed, with the weight of the energy consumption index set to 0.5, the weight of the economic cost index set to 0.3, and the weight of the comfort index set to 0.2. Through iterative calculation using the genetic algorithm (population size 100, number of iterations 50), a differentiated air-conditioning temperature setting scheme was obtained: maintaining at 24°C during the meeting period, adjusting to 26°C during the normal office period, setting to 28°C during the non-office period, and at the same time increasing the fresh air ratio to 40% during the meeting period and keeping it at 25% during other periods. These parameters were decomposed into specific control point parameters of the air-conditioning system, including the temperature set value of each area, the opening degree of the air supply volume regulating valve, and the position of the fresh air valve, etc. Finally, these parameters were converted into control instructions according to the BACnet protocol format used in the building. The instructions contained information such as device ID, attribute ID, value type, adjustment value, and execution time, forming a building energy-saving control instruction set, waiting to be executed in the planned time sequence to achieve refined control that meets the usage requirements and saves energy.

[0111] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0112] Send the building energy-saving control instructions to the building control system, start the device parameter adjustment process, and record the instruction execution timestamp;

[0113] Real-time monitor the operating state of the building after adjustment through multi-dimensional sensors, collect data on the temperature, humidity, energy consumption, and personnel density of the building after adjustment, and form a post-adjustment energy consumption monitoring data set;

[0114] Perform time alignment comparison between the adjusted energy consumption monitoring data set and the historical data in the same period before adjustment, calculate the energy consumption difference value under the same conditions, and generate an energy consumption change data table;

[0115] Conduct statistical analysis on the data in the energy consumption change data table, calculate three energy efficiency improvement indicators: the average energy consumption reduction rate, the peak energy consumption reduction amount, and the load balance degree, and construct an energy-saving effect evaluation report;

[0116] Fuse the key parameters in the energy-saving effect evaluation report with the building energy consumption feature vector, supplement new time series data points, and update the building energy consumption feature library;

[0117] Perform incremental training on the building energy consumption predictor through the updated building energy consumption feature library to improve the prediction accuracy and form a closed-loop optimized building energy consumption analysis system.

[0118] Specifically, send the building energy-saving control instruction to the building control system, start the equipment parameter adjustment process, and record the instruction execution timestamp. This process transmits the previously generated control instructions to each level of controller through the building automation network, such as the air-conditioning host controller, lighting control module, fan frequency converter, etc. Each instruction contains information such as the control object identifier, parameter type, parameter value, and execution time. When the control instruction is received and executed by the equipment, the system records the execution timestamp, forms an instruction execution log, and records the instruction issuance time, receipt confirmation time, and actual parameter effective time. These timestamp information provides an accurate time reference point for subsequent effect evaluation. Real-time monitor the operating state of the building after adjustment through multi-dimensional sensors, collect the adjusted building temperature, humidity, energy consumption, and personnel density data, and form an adjusted energy consumption monitoring data set. These multi-dimensional sensors are the same as those used in the previous data collection stage, including temperature sensors, humidity sensors, power monitoring devices, and personnel density sensors distributed in each functional area of the building. The monitoring process continuously records the real-time changes of each parameter, and the data collection frequency is set according to the parameter characteristics, generally 5 minutes / time for temperature and humidity, 15 minutes / time for energy consumption, and 10 minutes / time for personnel density. All the collected data is accompanied by accurate timestamps and stored in a unified data format to form an adjusted energy consumption monitoring data set. This data set not only contains energy consumption data, but also contains environmental parameters and usage parameters that affect energy consumption, providing a data basis for comprehensively evaluating the energy-saving effect.

[0119] Time-align and compare the adjusted energy consumption monitoring data set with the historical data of the same period before adjustment, calculate the energy consumption difference value under the same conditions, and generate an energy consumption change data table. Time alignment and comparison refer to selecting the historical data of the same period before adjustment (such as the same working day last week) as the benchmark and comparing it with the adjusted data. To ensure the accuracy of the comparison, external factors affecting energy consumption, such as weather conditions and building usage status, need to be considered. When processing data, the similar-day screening method is used. According to indicators such as temperature similarity, humidity similarity, and occupancy density similarity, the data closest to the current conditions are selected from the historical data as the comparison benchmark. When calculating the energy consumption difference, the formula is used: difference value = adjusted energy consumption - pre-adjusted energy consumption, and the difference rate = (adjusted energy consumption - pre-adjusted energy consumption) / pre-adjusted energy consumption. The calculation results are sorted according to dimensions such as equipment type, functional area, and time period to generate an energy consumption change data table, visually showing the actual effect of the energy-saving measures. Statistical analysis is performed on the data in the energy consumption change data table, and three energy efficiency improvement indicators, namely the average energy consumption reduction rate, peak energy consumption reduction amount, and load balance degree, are calculated to construct an energy-saving effect evaluation report. The average energy consumption reduction rate refers to the average percentage of energy consumption reduction during the entire evaluation period, and the calculation method is to weight-average the energy consumption reduction rates of each time period (the weight is the time period length). The peak energy consumption reduction amount refers to the reduction in energy consumption during the peak energy consumption period, and the energy-saving effect during the peak energy consumption period (such as the peak air-conditioning electricity consumption period in summer) is mainly concerned. The load balance degree is an indicator to measure the uniformity of energy consumption distribution, and the calculation method is the ratio of the energy consumption standard deviation to the average value. The smaller this value is, the more uniform the energy consumption distribution is, which is beneficial to reducing the equipment capacity configuration and improving the energy utilization efficiency. These three indicators evaluate the energy-saving effect from different perspectives and comprehensively form an energy-saving effect evaluation report, including the overall energy-saving rate, sub-item energy-saving effect, energy-saving situation in each area, and economic benefit analysis, etc.

[0120] Fuse the key parameters in the energy-saving effect evaluation report with the building energy consumption feature vector, supplement new time series data points, and update the building energy consumption feature library. The building energy consumption feature library is a structured storage of building energy consumption data, including historical energy consumption data, environmental parameter data, usage parameter data, and the association relationships between them. The update process first extracts the key parameters from the energy-saving effect evaluation report, such as the actual energy consumption values of each area under different control parameters, the energy response characteristics (such as the relationship between the temperature adjustment range and the energy consumption change), etc. Then these parameters are fused with the original building energy consumption feature vector to generate a new feature vector, which is used as a new data point to supplement the feature library. The feature library is updated in an incremental update manner. While retaining the original data, new data points are added to enrich the diversity of the data samples and improve the representativeness and timeliness of the feature library.

[0121] Incrementally train the building energy consumption predictor using the updated building energy consumption feature library to improve the prediction accuracy and form a closed-loop optimized building energy consumption analysis system. Incremental training means that, based on the original model, new data is used to fine-tune the parameters without retraining the entire model. During the incremental training process, first extract the new data from the updated feature library to construct a training set; then, with the current predictor model as the initial state, set a small learning rate (usually 1 / 10 of the original learning rate), and use the new training set to update the parameters; finally, evaluate the prediction accuracy of the updated model. If it is better than the original model, replace it; otherwise, keep the original model. Incremental training enables the model to gradually adapt to the dynamic changes in building energy consumption characteristics, such as seasonal changes, equipment aging, and usage pattern adjustments, continuously improving the prediction accuracy. In this way, the building energy consumption analysis forms a complete closed-loop optimization system: from data collection to anomaly detection, from optimization strategies to execution evaluation, and then to model update. Each link is interconnected and continuously optimized to achieve the intelligent and refined management of building energy consumption.

[0122] Taking the optimization of the air conditioning system in an educational building as an example, first, the generated energy-saving control instructions are sent to the air conditioning automatic control system through the BACnet protocol, adjusting the temperature set points of different classrooms (from the original unified 24°C to differential settings according to actual usage, such as 25°C in large classrooms and 26°C in small classrooms), and at the same time modifying the start-stop time strategy (starting 30 minutes in advance according to the class schedule instead of a fixed time). The execution situation of all instructions is recorded with accurate timestamps. The temperature sensors record the actual temperature change curves of each classroom after adjustment, the power monitoring system collects the electricity consumption of the air conditioning system, and the occupancy sensors record the actual usage of each classroom. These data together form an energy consumption monitoring data set for the week after adjustment. Comparing these data with the historical data of the same class schedule last week, considering that the average temperature this week is 2°C higher than last week, the reference energy consumption is adjusted through a temperature correction formula, and finally the energy consumption differences of the air conditioning system in each classroom at each time period are calculated. The statistical analysis results show that the differential temperature setting strategy reduces the average energy consumption by 8.2 kWh / day, cuts the peak load during peak hours (12:00 - 14:00) by 12 kW, and improves the system load balance from 0.42 to 0.31, indicating a more uniform energy distribution. After fusing these key parameters with the original building energy consumption feature vectors, they are supplemented into the building energy consumption feature library, especially recording the actual energy consumption performance under different temperature setting strategies. Finally, the building energy consumption predictor is incrementally trained with these newly added sample data, enabling the predictor to more accurately predict the energy consumption performance of the differential temperature strategy.

[0123] The above described the method for analyzing building energy consumption based on artificial intelligence in the embodiments of the present application. Next, the system for analyzing building energy consumption based on artificial intelligence in the embodiments of the present application will be described. Please refer to Figure 2 , an embodiment of the system for analyzing building energy consumption based on artificial intelligence in the embodiments of the present application includes:

[0124] A collection module, configured to collect building temperature, humidity, energy consumption, and occupancy density data through multi-dimensional sensors, and obtain a basic building energy consumption dataset;

[0125] An extraction module, configured to perform data missing value filling and outlier processing according to the basic building energy consumption dataset, extract building usage time periods, environmental parameters, and energy load characteristics, and generate a building energy consumption feature vector;

[0126] A building module, configured to establish a building energy consumption prediction model considering time factors and spatial relationships according to the building energy consumption feature vector, and obtain a building energy consumption predictor through training with historical data;

[0127] A comparison module, configured to calculate an expected energy consumption value using the building energy consumption predictor, compare the actual energy consumption data with the expected energy consumption value, identify energy consumption abnormal intervals, and generate a building energy consumption abnormal report;

[0128] A generation module, configured to formulate an energy optimization parameter adjustment plan based on the building energy consumption abnormal report, combine the building usage plan and environmental conditions, and generate a building energy saving control instruction;

[0129] An update module, configured to record the energy consumption change data before and after adjustment by executing the building energy saving control instruction, calculate the energy efficiency improvement index, update the building energy consumption feature library, and complete the closed-loop of building energy consumption analysis.

[0130] Through the collaborative cooperation of the above-mentioned various components, data on building temperature, humidity, energy consumption, and occupancy density are collected through multi-dimensional sensors, obtaining a comprehensive basic dataset of building energy consumption, providing a rich data foundation for subsequent analysis, and solving the problem of single data source in traditional methods. By filling in missing data and dealing with outliers in the basic dataset of building energy consumption, the building usage period, environmental parameters, and energy load characteristics are extracted to generate building energy consumption feature vectors, significantly improving the data quality and feature expression ability, and providing high-quality input for model training. Based on the building energy consumption feature vectors, a building energy consumption prediction model considering time factors and spatial relationships is established, fully capturing the spatio-temporal correlation of building energy consumption, overcoming the limitation of traditional models that ignore the mutual influence of energy consumption between regions, and improving the prediction accuracy. The building energy consumption predictor is used to calculate the expected energy consumption value, compare the actual energy consumption data with the expected energy consumption value, identify the abnormal energy consumption interval, and generate a building energy consumption anomaly report, realizing the transformation from passive discovery to active warning, and greatly shortening the anomaly discovery time. Based on the building energy consumption anomaly report, combined with the building usage plan and environmental conditions, a highly targeted energy optimization parameter adjustment plan is formulated to generate building energy-saving control instructions, achieving precise energy saving and avoiding the problem of reduced comfort caused by the "one-size-fits-all" approach in traditional methods. By executing the building energy-saving control instructions, recording the energy consumption change data before and after adjustment, calculating the energy efficiency improvement index, and updating the building energy consumption feature library, the closed-loop of building energy consumption analysis is completed, realizing self-learning and continuous optimization. This solution makes full use of the advantages of artificial intelligence algorithms in the field of specific building energy consumption management, especially in spatio-temporal data processing. The time-dependent relationship of energy consumption is captured through a bidirectional long short-term memory network, the spatial correlation between building regions is modeled through a graph convolutional network, and the importance weights of different features are dynamically adjusted through an attention mechanism. These algorithm features contribute to the solution by being able to adaptively process the dynamic change characteristics of building energy consumption, identify complex energy consumption patterns, thereby achieving more accurate prediction and more refined control. In addition, a genetic algorithm is used to solve the multi-objective optimization problem, balancing the three dimensions of energy saving, cost, and comfort, and realizing a globally optimal energy management strategy. The incremental learning mechanism enables the model to continuously learn from new data, improve the prediction accuracy, and adapt to the long-term evolution of building energy consumption characteristics.

[0131] Referring to Figure 3 , an embodiment of the present invention also provides a computer device, which may be a server, and its internal structure may be as Figure 3As shown in the figure. The computer device includes a processor, a memory, a display screen, an input device, a network interface, and a database connected through a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0132] Those skilled in the art can understand that Figure 3 the structure shown in the figure is only a block diagram of some structures related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0133] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0134] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium provided by the present invention and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or an external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0135] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0136] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0137] The above is described. The above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. An artificial intelligence-based building energy consumption analysis method, characterized in that, The artificial intelligence-based building energy consumption analysis method includes: Collecting building temperature, humidity, energy consumption, and occupancy density data through multi-dimensional sensors to obtain a building energy consumption basic data set; Based on the building energy consumption basic data set, performing data missing value filling and outlier processing, extracting building usage time periods, environmental parameters, and energy load characteristics, and generating a building energy consumption feature vector; Based on the building energy consumption feature vector, establishing a building energy consumption prediction model considering time factors and spatial relationships, and obtaining a building energy consumption predictor through training with historical data; Using the building energy consumption predictor to calculate the expected energy consumption value, comparing the actual energy consumption data with the expected energy consumption value, identifying energy consumption abnormal intervals, and generating a building energy consumption abnormal report; Based on the building energy consumption abnormal report, combining the building usage plan and environmental conditions, formulating an energy optimization parameter adjustment plan, and generating a building energy conservation control instruction; By executing the building energy conservation control instruction, recording the energy consumption change data before and after adjustment, calculating the energy efficiency improvement index, updating the building energy consumption feature library, and completing the building energy consumption analysis closed loop.

2. The artificial intelligence-based building energy consumption analysis method according to claim 1, wherein The step of collecting building temperature, humidity, energy consumption, and occupancy density data through multi-dimensional sensors to obtain a building energy consumption basic data set includes: Collecting regional temperature change data through a temperature sensor network arranged according to building functional zones to form a temperature spatio-temporal data matrix; Collecting building interior humidity distribution data through multi-point humidity sensors and generating a humidity environment feature set in combination with ventilation state parameters; Collecting sub-item energy consumption data of major energy-consuming equipment in the building through an energy metering device, recording the equipment operation power curve, and constructing an equipment energy consumption database; Collecting building interior personnel flow and density data through a personnel detection device and establishing a personnel activity pattern feature table; Performing time synchronization processing on the temperature spatio-temporal data matrix, the humidity environment feature set, the equipment energy consumption database, and the personnel activity pattern feature table, removing invalid data points, and supplementing missing time period data to generate a building energy consumption basic data set; Performing normalization processing on various types of data in the building energy consumption basic data set to eliminate dimension differences and obtain a building energy consumption basic data set.

3. The method for analyzing building energy consumption based on artificial intelligence according to claim 1, characterized in that, The step of based on the building energy consumption basic data set, performing data missing value filling and outlier processing, extracting building usage time periods, environmental parameters, and energy load characteristics, and generating a building energy consumption feature vector includes: Performing filling processing on the missing values in the building energy consumption basic data set through a sliding time window average filling method to obtain a preliminary filling data set; Identifying outliers in the preliminary filling data set through an improved Z-score method, and replacing the identified outliers with the median of the front and back data points to generate a smoothed data set; Performing time series segmentation on the smoothed data set, and dividing it into working periods, transition periods, and idle periods according to the building energy consumption load change characteristics to form a time period division table; Performing noise reduction processing on the smoothed data set through a wavelet transform method, separating the trend term, seasonal term, and residual term of the building energy consumption, and constructing an energy consumption decomposition feature; Extract five environmental parameter features, namely temperature difference, thermal inertia index, sunshine influence coefficient, humidity change rate, and personnel density distribution, from the smoothed dataset to generate an environmental impact factor set; Integrate the time period division table, the energy consumption decomposition features, and the environmental impact factor set through a feature fusion matrix to generate a building energy consumption feature vector.

4. The artificial intelligence-based building energy consumption analysis method according to claim 1, wherein Based on the building energy consumption feature vector, establish a building energy consumption prediction model considering time factors and spatial relationships, and through training with historical data, obtain a building energy consumption predictor, including: Perform vector segmentation on the building energy consumption feature vector, dividing it into a time feature sub-vector representing time characteristics and a spatial feature sub-vector representing spatial distribution; Perform bidirectional long short-term memory network processing on the time feature sub-vector, calculate the hidden state through forward and backward time steps, capture long-term dependencies, extract the temporal variation law of building energy consumption, and form a time relationship feature matrix; Perform graph convolution operation on the spatial feature sub-vector, represent each area of the building as a node in the graph, and represent the physical connection relationship between areas as edges. Calculate the node feature update through the message passing mechanism, capture the energy consumption interaction relationship between different areas of the building, and generate a spatial association feature map; Perform feature fusion on the time relationship feature matrix and the spatial association feature map through an attention mechanism, calculate the weight coefficients of each feature channel, perform weighted combination, and construct a spatio-temporal fusion feature representation; Pair the spatio-temporal fusion feature representation with historical building energy consumption data to form a training sample set, iteratively optimize the parameters through the mini-batch gradient descent algorithm, calculate the gradient using a loss function combined with weighted mean absolute percentage error and root mean square error, update the connection weights, and at the same time introduce L2 regularization to prevent overfitting, forming an energy consumption prediction calculation framework; Divide the dataset into K subsets through the K-fold cross-validation method, alternately use K - 1 subsets for training and 1 subset for validation, evaluate the prediction stability and accuracy of the energy consumption prediction calculation framework under different data partitions, and select the model parameter configuration with the best validation performance as the building energy consumption predictor.

5. The artificial intelligence-based building energy consumption analysis method according to claim 1, wherein, Use the building energy consumption predictor to calculate the expected energy consumption value, compare the actual energy consumption data with the expected energy consumption value, identify the energy consumption abnormal interval, and generate a building energy consumption abnormal report, including: Input the current building parameter data into the building energy consumption predictor, and obtain the expected energy consumption value at each time point through forward calculation to form an expected energy consumption sequence; Collect the actual building energy consumption data of the current time period, align it according to the same time granularity as the expected energy consumption sequence, and obtain the actual energy consumption sequence; Calculate the difference between the actual energy consumption sequence and the expected energy consumption sequence to generate an energy consumption residual sequence; Analyze the energy consumption residual sequence through a dynamic threshold detection algorithm, calculate the threshold τ value at each time point, and identify the data points with the absolute value of the residual exceeding the corresponding threshold as abnormal points; Perform continuity analysis on the abnormal points, mark the interval where three or more abnormal points continuously appear in time as the energy consumption abnormal interval, and record the start time, duration, maximum deviation value, and cumulative deviation amount of the abnormality; Based on the characteristic parameters of the abnormal energy consumption interval, classify the abnormal interval by combining the decision tree algorithm, identify the abnormal types, and generate a building energy consumption anomaly report according to the abnormal location, abnormal type, severity, and recommended treatment methods.

6. The method for analyzing building energy consumption based on artificial intelligence according to claim 1, wherein Based on the building energy consumption anomaly report, combine the building usage plan and environmental conditions to formulate an energy optimization parameter adjustment plan and generate a building energy-saving control instruction, including: Extract the abnormal type and energy consumption deviation data from the building energy consumption anomaly report to construct an energy consumption optimization target list; Collect building usage plan data, including work period arrangements, personnel occupancy plans, and special event times, to form a building usage time series table; Obtain future environmental condition forecasts, including temperature change trends, humidity distributions, and light intensities, to generate an environmental impact factor prediction sequence; Construct a multi-objective optimization function, including three dimensions of energy consumption indicators, economic cost indicators, and comfort indicators, set the weight coefficients of each indicator, and solve the optimal parameter combination through the genetic algorithm to obtain the equipment operation parameter adjustment plan; Decompose the equipment operation parameter adjustment plan into equipment-level control parameters, including air-conditioning temperature set values, fresh air ratio adjustment parameters, lighting intensity parameters, and equipment start-stop times, to generate an equipment control parameter table; Convert the equipment control parameter table into an instruction format that conforms to the building control system communication protocol to form a building energy-saving control instruction.

7. The method for analyzing building energy consumption based on artificial intelligence according to claim 1, wherein By executing the building energy-saving control instruction, record the energy consumption change data before and after the adjustment, calculate the energy efficiency improvement indicators, and update the building energy consumption feature library to complete the building energy consumption analysis closed-loop, including: Send the building energy-saving control instruction to the building control system, start the equipment parameter adjustment process, and record the instruction execution timestamp; Real-time monitor the adjusted operation state of the building through multi-dimensional sensors, collect the adjusted building temperature, humidity, energy consumption, and personnel density data to form an adjusted energy consumption monitoring data set; Perform time alignment comparison on the adjusted energy consumption monitoring data set and the historical data of the same period before the adjustment, calculate the energy consumption difference value under the same conditions, and generate an energy consumption change data table; Statistically analyze the data in the energy consumption change data table, calculate three energy efficiency improvement indicators of the average energy consumption reduction rate, peak energy consumption reduction amount, and load balance degree, and construct an energy-saving effect evaluation report; Fuse the key parameters in the energy-saving effect evaluation report with the building energy consumption feature vector, supplement new time series data points, and update the building energy consumption feature library; Incrementally train the building energy consumption predictor through the updated building energy consumption feature library to improve the prediction accuracy and form a closed-loop optimized building energy consumption analysis system.

8. An artificial intelligence-based building energy consumption analysis system for implementing the artificial intelligence-based building energy consumption analysis method according to any one of claims 1-7, characterized in that, The artificial intelligence-based building energy consumption analysis system includes: A collection module for collecting building temperature, humidity, energy consumption, and personnel density data through multi-dimensional sensors to obtain a building energy consumption basic data set; An extraction module for filling in missing data and processing outliers according to the building energy consumption basic data set, extracting building usage periods, environmental parameters, and energy load characteristics, and generating a building energy consumption feature vector; A building energy consumption prediction model considering time factors and spatial relationships is established according to the building energy consumption feature vector, and through training with historical data, a building energy consumption predictor is obtained; A comparison module is used to calculate the expected energy consumption value by using the building energy consumption predictor, compare the actual energy consumption data with the expected energy consumption value, identify the energy consumption abnormal interval, and generate a building energy consumption abnormal report; A generation module is used to formulate an energy optimization parameter adjustment plan based on the building energy consumption abnormal report, combined with the building usage plan and environmental conditions, and generate a building energy-saving control instruction; An update module is used to record the energy consumption change data before and after adjustment by executing the building energy-saving control instruction, calculate the energy efficiency improvement index, update the building energy consumption feature library, and complete the closed-loop of building energy consumption analysis.

9. A computer device, characterized in that, It includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the artificial intelligence-based building energy consumption analysis method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is stored thereon. When the computer program is run by the processor, the processor is caused to execute the artificial intelligence-based building energy consumption analysis method according to any one of claims 1 to 7.

Images