Energy consumption performance prediction method applied to building cluster
By using cellular decomposition and physical constraint models, the problem of balancing cluster heterogeneity and coupling in building cluster energy consumption prediction is solved, achieving accurate energy consumption prediction and long-term adaptability, and improving prediction accuracy and reliability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-16
- Publication Date
- 2026-03-24
AI Technical Summary
Existing building cluster energy consumption prediction technologies struggle to balance cluster heterogeneity and coupling in complex scenarios, lack dynamic correlation capture capabilities, exhibit a disconnect between physical logic and data fitting, and suffer from poor long-term adaptability, leading to decreased prediction accuracy.
By using the cell decomposition method, building units are divided based on spatial correlation and functional similarity. Dynamic interactions are preserved by combining thermal environment and pipeline coupling matrix. Physical constraint models are configured to achieve deep integration of data fitting and mechanism constraints. Parameter migration is performed for cells with scarce data.
It enables precise quantification of energy consumption differences and their correlations within the cluster, improving the accuracy of predictions and the reliability of engineering applications, reducing prediction errors in data-scarce scenarios, and avoiding abnormal predictions.
Smart Images
Figure CN121723213A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of cluster energy consumption prediction, and in particular to a method for predicting the energy consumption performance of building clusters. Background Technology
[0002] With the acceleration of urbanization and the advancement of "dual carbon" goals, the demand for refined energy consumption management in building clusters (such as industrial parks, urban complexes, and university towns) is becoming increasingly urgent. Energy performance prediction, as a core aspect of management, has attracted much attention for its technological development. Currently, existing technologies in this field mainly fall into three categories, but their limitations in complex cluster scenarios are becoming increasingly apparent, prompting the development of innovative technological solutions.
[0003] Existing building cluster energy consumption prediction technologies can be divided into three categories: First, the overall modeling method, which treats the entire cluster as a single prediction object and inputs macroscopic data (such as total building area and average climate parameters) to construct a regression or machine learning model. The advantage is that the modeling is simple, but the disadvantage is that it completely ignores the heterogeneity of building functions, building envelopes, and energy consumption behavior within the cluster. Second, the single-building independent modeling method, which establishes an energy consumption model (such as the EnergyPlus mechanism model or LSTM data model) for each building in the cluster and directly sums the prediction results. Although it can reflect the differences between individual buildings, it severs the coupling effects of thermal radiation between buildings, shared pipeline losses, and peak energy consumption superposition. Third, the traditional fusion modeling method, which attempts to combine data-driven and physical mechanisms, mostly adopts a static weighted mode of "mechanism model output + data model correction", lacks a dynamic feedback mechanism, and has limited fusion depth.
[0004] As building clusters evolve towards "mixed-function, high-density, and complex energy systems," existing technologies have revealed four core problems: First, it is difficult to balance the heterogeneity and coupling of clusters. Overall modeling cannot distinguish the energy consumption differences between office, residential, and commercial buildings, while single-building modeling loses the cluster-level interactive influence, leading to prediction bias. Second, the ability to capture dynamic correlations is insufficient. Traditional data-driven methods rely on static spatiotemporal characteristics (such as historical averages and fixed time period proportions), which cannot accurately capture dynamic processes such as the transmission of peak energy consumption within the cluster and sudden changes in energy consumption under extreme weather conditions. Third, there is a disconnect between physical logic and data fitting. Pure data models are prone to "physical inaccuracies" due to data noise or distribution deviations (such as predicting heat transfer energy consumption in violation of the law of heat conduction), while pure mechanistic models are difficult to adapt to dynamic factors such as changes in user behavior. Fourth, long-term adaptability is poor. After deployment, most models operate statically and cannot cope with long-term dynamic scenarios such as building aging, equipment efficiency decline, and changes in energy consumption habits, resulting in a rapid decrease in prediction accuracy over time.
[0005] To address these issues, there is an urgent need for a method to predict the energy consumption performance of building clusters. Summary of the Invention
[0006] To address the aforementioned issues, this application proposes a method for predicting energy consumption performance in building clusters, the details of which are as follows: S1. Obtain the historical basic physical data and historical operation data of the building cluster, and preprocess the historical basic physical data and historical operation data to obtain building data, environmental data, energy consumption data, equipment operation data, and user behavior data; S2. Create a data curve graph with time as the horizontal axis and environmental data, energy consumption data, and equipment operation data as the vertical axis. Extract features from the data curve graph to obtain cluster features. S3. Based on the building data, each data point is clustered to obtain energy consumption performance cells; S4. Physical constraints are configured on the energy consumption performance cells to obtain the energy consumption performance prediction model. S5. Obtain the current operating data and input the current operating data into the energy consumption performance prediction model to obtain the energy consumption performance prediction result.
[0007] Preferably, in S2, a data curve is established with time as the horizontal axis and environmental data, energy consumption data, and equipment operation data as the vertical axis. The specific content of the cluster features obtained by feature extraction from the data curve includes: Create data curves with time on the horizontal axis and environmental data, energy consumption data, and equipment operation data on the vertical axis. Align the data curves to obtain a unified time series curve set; Standard curve sets are obtained by annotating building data and user behavior data on corresponding curves; Cluster features are obtained by combining standard curves with feature extraction. These cluster features include environmental features, energy consumption features, behavioral features, and facility features.
[0008] Preferably, the specific content of the energy consumption performance cell obtained by clustering each data point based on building data in S3 includes: Building data is processed to obtain building location information, building function information, and building sharing information; Based on building location information, building function information, and building sharing information, the spatial distance correlation, functional type similarity, and energy system coupling degree between different buildings are calculated respectively. Spatial distance correlation, functional type similarity, and energy system coupling degree are used as feature vectors; The elbow rule is used to determine the cell K value. The feature vector is input into the K-means clustering algorithm to obtain the cell affiliation label of each building, and then the initial energy consumption performance cell is obtained. The energy consumption performance cell is obtained by configuring the coupling relationship for the initial energy performance cell.
[0009] Preferably, the specific content of configuring the coupling relationship for the initial energy performance cell to obtain the energy consumption performance cell is as follows: Based on building geometric parameters and environmental data, the heat exchange between cells is calculated to obtain the thermal coupling strength coefficient and form a thermal coupling matrix. Based on the pipeline topology diagram, the pressure loss and heat transfer loss of the shared pipeline between different cells are calculated to obtain the pipeline coupling loss coefficient and form the pipeline coupling matrix.
[0010] Calculate the correlation coefficients of energy consumption time series data of different cells to form a time-period coupling matrix; By configuring weight conversion coefficients for the thermal coupling matrix, pipeline coupling matrix, and time-period coupling matrix, the comprehensive coupling strength matrix between cells is obtained.
[0011] Preferably, the energy consumption performance cell is configured with a migration mechanism for data-scarce cells, the details of which are as follows: Calculate the feature similarity between data-scarce cells and data-sufficient cells; The feature similarity includes functional type, building scale, and coupling mode, and the source cell with the highest similarity is selected as the migration object. Transfer the parameters trained in the source cells to the scarce cells.
[0012] Preferably, the specific content of the energy consumption performance prediction model obtained by configuring physical constraints on energy consumption performance cells in S4 is as follows: Based on the aging of the structural facilities, loss factors are configured for the theoretical structural heat transfer equation, theoretical functional balance equation and theoretical equipment efficiency to obtain the improved structural heat transfer equation, improved functional balance equation and improved equipment efficiency threshold. Obtain the threshold for improving equipment efficiency, and define the equipment efficiency constraint as an activation function; To improve the structural heat transfer equation and the functional balance equation, constraint weights are configured to obtain the loss function.
[0013] Preferably, the specific content of obtaining the current operating data in S5 and substituting the current operating data into the energy consumption performance prediction model to obtain the energy consumption performance prediction result includes: Cell weights are calculated based on the proportion of cell building area and the contribution of cell energy consumption. The cell energy consumption contribution rate is the proportion of historical energy consumption to the total energy consumption of the cluster. The energy consumption performance prediction results are calculated based on the predicted value and weight of the current cell, the predicted values of the cells associated with the current cell, and the comprehensive coupling strength between the current cell and the current cell.
[0014] An energy consumption performance prediction system for building clusters, comprising: Data processing unit: acquires the building cluster's past basic physical data and past operational data, and preprocesses the past basic physical data and past operational data to obtain building data, environmental data, energy consumption data, equipment operation data, and user behavior data; Feature extraction unit: A data curve is created with time as the horizontal axis and environmental data, energy consumption data, and equipment operation data as the vertical axis. Feature extraction is performed on the data curve to obtain cluster features. Model building unit: Based on building data, each data point is clustered to obtain energy consumption performance cells, and physical constraints are configured on the energy consumption performance cells to obtain the energy consumption performance prediction model; Energy consumption performance prediction unit: acquires current operating data, inputs the current operating data into the energy consumption performance prediction model to obtain the energy consumption performance prediction result.
[0015] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor invokes the computer program in the memory to implement the content of the energy consumption performance prediction method applied to building clusters.
[0016] A storage medium storing computer-executable instructions, which, when loaded and executed by a processor, implement the content of the energy consumption performance prediction method applied to building clusters.
[0017] In summary, the energy consumption performance prediction method for building clusters proposed in this invention has the following advantages compared to traditional technologies: 1. This application uses cell decomposition to divide the cluster into building units according to indicators such as spatial correlation and functional similarity, thereby achieving dimensionality reduction. Furthermore, it retains the dynamic interaction between units through coupling matrices such as thermal environment and pipe network, thus accurately quantifying the energy consumption differences and related impacts of different buildings within the cluster, making the prediction basis more realistic. 2. This application targets data-scarce cells and quickly adapts them through parameter migration of similar cells, solving the pain point of traditional methods relying on a large amount of historical data. It can also reduce prediction errors in data-scarce scenarios. 3. This application transforms physical rules such as structural heat transfer and energy balance into hard constraints and soft loss terms, achieving deep integration of data fitting and mechanism constraints, thereby making the predicted values conform to physical logic, avoiding abnormal predictions caused by overfitting in traditional models, and improving the reliability of engineering applications.
[0018] The technical method of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0019] Figure 1 This is a flowchart illustrating the steps of an energy consumption performance prediction method for building clusters according to the present invention. Figure 2 This is a unit diagram of an energy consumption performance prediction system for building clusters according to the present invention. Detailed Implementation
[0020] The technical method of the present invention will be further described below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values of the components and steps described in these embodiments do not limit the scope of this application.
[0021] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the scope of this application and its application or use.
[0022] Techniques, systems, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, they should be considered part of the instruction manual.
[0023] In all the examples shown and discussed herein, any specific values should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.
[0024] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.
[0025] Example 1 This invention provides a method for predicting energy consumption performance in building clusters, such as... Figure 1 As shown, the specific steps are as follows: S1. Obtain the historical basic physical data and historical operational data of the building cluster, and preprocess the historical basic physical data and historical operational data to obtain building data, environmental data, energy consumption data, equipment operation data, and user behavior data.
[0026] Understandably, data is the foundation of predictive models and needs to cover all dimensions of information, including building physics, environment, operation, and behavior, to ensure data quality and standardization.
[0027] Building data, or basic building physical data, is extracted from design institute drawings and BIM models. It includes the building envelope parameters, building dimensions, number of floors, functional zoning, and shared energy system data for each building in the cluster.
[0028] Real-time environmental and operational data can be obtained by configuring and deploying Internet of Things (IoT) sensing networks.
[0029] User behavior data is obtained through smart sockets, user surveys, and property energy consumption logs, including the distribution of energy consumption time periods for different building types (e.g., office buildings have peak energy consumption from 9:00 to 18:00, and residential buildings have peak energy consumption from 18:00 to 22:00), equipment start-up and shutdown frequency (e.g., air conditioning set temperature distribution, lighting on-time), and special behaviors (e.g., number of overtime days in office buildings, and holiday travel rate in residential buildings).
[0030] S2. Create a data curve graph with time as the horizontal axis and environmental data, energy consumption data, and equipment operation data as the vertical axis. Extract features from the data curve graph to obtain cluster features. Furthermore, in S2, a data curve is created with time as the horizontal axis and environmental data, energy consumption data, and equipment operation data as the vertical axes. Feature extraction is performed on the data curve to obtain the specific content of the cluster features, including: Create corresponding data curves with time as the horizontal axis and environmental data, energy consumption data, and equipment operation data as the vertical axes, respectively. By aligning the data curves with time, a unified time series curve set is obtained; Standard curve sets are obtained by annotating building data and user behavior data on corresponding curves; Cluster features are obtained by combining standard curves with feature extraction. These cluster features include environmental features, energy consumption features, behavioral features, and facility features.
[0031] Understandably, this is a key operational description of the feature extraction stage in building cluster energy consumption prediction technology. It involves extracting four core cluster features from multi-source data using a set of standard curves, providing high-quality input for subsequent modeling. A set of standard curves refers to a pre-constructed collection of time-series curves with reference benchmark significance. These are generated based on historical data or industry standard data from a large amount of similar building clusters. Examples include "standard curves for daily energy consumption changes in typical office clusters," "standard curves for air conditioning load under high-temperature conditions in summer," and "standard curves for the distribution of user energy consumption during specific time periods in residential clusters." These serve as benchmarks for data calibration and feature quantification, eliminating data incomparability caused by differences in scale and location between different clusters.
[0032] Feature extraction does not directly use the original monitoring data, but compares and analyzes the real-time / historical data of the cluster with the standard curve group. By calculating indicators such as "deviation value", "similarity" and "fit coefficient", feature parameters with engineering significance and modeling value are extracted.
[0033] Environmental characteristics are the temporal patterns of environmental parameters such as temperature, humidity, solar radiation intensity, and wind speed in the area where the cluster is located, such as the daily average temperature difference, the deviation of sunshine duration, and the frequency of extreme weather. Real-time environmental monitoring data is compared with standard climate curves, and the fitting degree between the actual environmental curve and the standard curve, peak temperature difference, and other characteristics are calculated and displayed.
[0034] Energy consumption characteristics are the energy consumption patterns of the cluster as a whole and its sub-units, such as energy consumption per unit area, peak-valley energy consumption ratio, energy consumption fluctuation rate, and shared pipeline loss rate. By comparing the actual energy consumption curve with the standard energy consumption curve, features such as peak energy consumption offset duration, energy consumption intensity deviation rate, and coupled energy consumption ratio are extracted and displayed.
[0035] Behavioral characteristics are the energy consumption patterns of users within the cluster, such as peak energy consumption periods, equipment start-up and shutdown frequency, and the proportion of energy consumption during overtime. The actual energy consumption behavior curve is matched with the standard behavior curve, and features such as behavioral pattern similarity and peak period overlap are calculated and displayed.
[0036] Facility characteristics refer to the operational status characteristics of energy facilities within the cluster, such as air conditioning COP value, boiler thermal efficiency, photovoltaic equipment power generation efficiency, and pipeline transmission efficiency. By comparing the facility operation parameter curves with standard efficiency curves, features such as equipment efficiency decay rate and operating load deviation are extracted and displayed.
[0037] S3. Based on building data, each data point is clustered to obtain energy consumption performance cells. Cell decomposition achieves cluster dimensionality reduction and coupling preservation.
[0038] Furthermore, the specific content of the energy consumption performance cells obtained by clustering each data point based on building data in S3 includes: Building data is processed to obtain building location information, building function information, and building sharing information; Based on building location information, building function information, and building sharing information, the spatial distance correlation, functional type similarity, and energy system coupling degree between different buildings are calculated respectively. Spatial distance correlation, functional type similarity, and energy system coupling degree are used as feature vectors; The elbow rule is used to determine the cell K value. The feature vector is input into the K-means clustering algorithm to obtain the cell affiliation label of each building, and then the initial energy consumption performance cell is obtained.
[0039] Understandably, the Elbow Method is a common method for determining the number of clusters K in K-means clustering: by calculating the sum of squared clustering errors (SSE, i.e., the sum of squared distances from each sample to its cluster center) under different K values, a curve is plotted showing how SSE changes as the K value increases; the K value corresponding to the inflection point (resembling an "elbow") where the rate of SSE decreases from fast to slow is the optimal number of clusters, which can balance the clustering effect and computational complexity.
[0040] The initial energy performance cells are coupled to obtain energy consumption performance cells. The preliminary results are verified. If situations such as "building spacing within the cell > 50m" or "excessive functional differences" occur, the assignment is manually adjusted to ensure that the cell division conforms to the principle of "homogeneous cohesion and heterogeneous separation".
[0041] Furthermore, the specific content of the energy consumption performance cell obtained by configuring the coupling relationship for the initial energy performance cell is as follows: Based on building geometric parameters and environmental data, the heat exchange between cells is calculated to obtain the thermal coupling strength coefficient and form a thermal coupling matrix.
[0042] Understandably, a cluster 3D flow field model is constructed based on CFD software (such as Fluent), and building geometric parameters and environmental data (wind speed, wind direction, temperature) are input. The heat exchange between cells is simulated and calculated to obtain the "thermal coupling strength coefficient" (unit: W / ℃, which represents the heat exchange power between cells when the temperature difference changes by 1℃), forming a thermal coupling matrix.
[0043] Based on the pipeline topology diagram, the pressure loss and heat transfer loss of the shared pipeline between different cells are calculated to obtain the pipeline coupling loss coefficient and form the pipeline coupling matrix.
[0044] Understandably, based on the pipeline network topology diagram, hydraulic calculation software (such as PipeFlow) is used to calculate the pressure loss and heat transfer loss of the shared pipeline network between different cells, and obtain the "pipeline network coupling loss coefficient" (unit: % / km, representing the proportion of energy loss per kilometer of pipeline network), forming the pipeline network coupling matrix.
[0045] The correlation coefficients of energy consumption time series data of different cells are calculated to form a time period coupling matrix. The larger the absolute value of the correlation coefficient, the stronger the time period coupling (e.g., the energy consumption correlation coefficient between office building cells and shopping mall cells is 0.7, indicating that the peak energy consumption periods of the two overlap significantly).
[0046] By configuring weight conversion coefficients for the thermal coupling matrix, pipeline coupling matrix, and time-period coupling matrix, the comprehensive coupling strength matrix between cells is obtained.
[0047] Furthermore, the energy consumption performance cell is configured with a migration mechanism for data-scarce cells. Spatiotemporal attention migration captures dynamic associations and adapts to data scarcity, as detailed below: Calculate the feature similarity between data-scarce cells and data-sufficient cells; The feature similarity includes functional type, building scale, and coupling mode, and the source cell with the highest similarity is selected as the migration object. Transfer the parameters trained in the source cells to the scarce cells.
[0048] Understandably, when calculating the feature similarity between data-scarce cells and data-sufficient cells, it is necessary to quantify the similarity for three core features: function type, building scale, and coupling mode, and then obtain the comprehensive similarity through weighted summation.
[0049] Functional type is the core influencing factor of cell energy consumption pattern. The category matching quantification method is adopted to establish a hierarchical table of building functional types, which is divided into major categories and subcategories (e.g., major categories: office, residential, commercial, industrial; subcategories: office buildings, R&D offices, ordinary residential buildings, commercial complexes, etc.). Then, the matching weight is defined. If the cell is a mixed function (e.g., containing 2 office buildings + 1 commercial building), it is calculated by weighting the proportion of functional building area.
[0050] Building scale determines the energy consumption base of a cell. The relative deviation quantification method is adopted to eliminate the influence of absolute scale differences. Core scale indicators are selected: total building area of the cell, average number of floors, and building density (building area / land area). The normalized deviation is calculated for each indicator, and the scale similarity is obtained by weighting according to the indicator weight.
[0051] The coupling mode reflects the interaction characteristics between a cell and its surrounding units. The cosine similarity method of the coupling matrix is used: the comprehensive coupling strength matrix of the cell is used as the feature carrier (the matrix elements are the coupling strength values between the cell and all other cells), and the coupling matrix vector of the scarce cell is denoted as... The coupling matrix vector of a sufficient cell is denoted as The cosine similarity is calculated with a value range of [0,1]. The closer the value is to 1, the more consistent the coupling objects and coupling strength distribution of the cells are.
[0052] The final similarity value is obtained by weighted summation of the three types of feature similarities, with the weights set according to the degree of influence of the features on energy consumption prediction.
[0053] S4. Physical constraints are configured on the energy consumption performance cells to obtain the energy consumption performance prediction model. The physical constraint self-encoding ensures the physical consistency of the prediction.
[0054] Furthermore, the specific content of the energy consumption performance prediction model obtained by configuring physical constraints on energy consumption performance cells in S4 is as follows: Based on the aging of structural facilities, the theoretical structural heat transfer equation, theoretical functional balance equation, and theoretical equipment efficiency are configured with loss factors to obtain improved structural heat transfer equation, improved functional balance equation, and improved equipment efficiency threshold. The quantitative impact of structural facility aging is integrated into the physical mechanism model of energy consumption prediction, and the ideal assumptions of theoretical equations and efficiency thresholds are corrected to make them fit the actual operating scenario.
[0055] The core aging indicators of building cluster structures and facilities are quantitatively selected, including: Aging indicators of building envelope: aging attenuation coefficient of external wall / roof insulation layer (obtained through ultrasonic or thermal imaging detection; the higher the degree of aging, the greater the attenuation coefficient), and the rate of decline in door and window sealing performance (obtained through airtightness testing); Aging indicators of energy system facilities: damage rate of pipeline insulation layer, rate of decline in the operating efficiency of water pumps / fans, and rate of decline in the energy efficiency of boilers / air conditioning compressors (obtained through equipment operation records and periodic testing data). Based on these indicators, a comprehensive aging coefficient is calculated. (Value range [1.05, 1.5], the larger the value, the higher the degree of aging), the formula is: =0.4 w +0.3 p +0.3 e ,in, w The aging and degradation coefficient of the building envelope. p The aging coefficient of the pipeline network facilities. e The aging factor of energy-consuming equipment.
[0056] Improved structural heat transfer equation: Q0 Where K is the theoretical heat transfer coefficient and A is the heat transfer area. T represents the indoor / outdoor temperature difference, f k =1+0.2 ; Improved functional equilibrium equation: Q1=Q 2+ f l Where Q1 is the theoretical total energy supply, Q2 is the theoretical end-user energy consumption, and Q3 is the theoretical pipeline loss), f l =1+0.3 .
[0057] The theoretical equipment efficiency threshold is the efficiency range under ideal operating conditions, based on the equipment aging factor. e Configure efficiency decay factor f e f e =1+0.2 e ; For example, the theoretical COP threshold for air conditioners is [1.5, 4.5], and the improved COP threshold for air conditioners is: [1.5]. f e 4.5 fe ].
[0058] Obtain the threshold for improving equipment efficiency, and define the constraint on improving equipment efficiency as an activation function; The improved equipment efficiency threshold is transformed into the activation function of the model output layer, which forces the predicted equipment efficiency parameters to fall within the reasonable range of reality, thus avoiding physical inaccuracies.
[0059] Construct a piecewise constrained activation function g(x), taking the COP value of air conditioning as an example: ; in, , , which are the upper and lower limits of the improved air conditioner COP threshold, respectively, and x is the original COP value predicted by the model; The activation function is embedded into the output layer of the prediction model to impose real-time constraints on the prediction results related to equipment efficiency, ensuring that the output values match the actual operating efficiency of the aging facility. Constraint weights are configured for the improved structural heat transfer equation and the improved functional balance equation to obtain the loss function.
[0060] The improved structural heat transfer equation and the improved functional balance equation are transformed into loss function constraints of the model, and the physical mechanism and data fitting are deeply integrated by configuring weights.
[0061] Definition of equations, constraint loss terms, heat transfer equations, constraint loss, Loss k : Measures the heat transfer energy consumption Q predicted by the model pred Deviation from the calculated value Q0 of the improved heat transfer equation: Loss k =|Q pred -Q0|.
[0062] Balance equation constraint loss b : Measures the total energy supply Q predicted by the model supply,pred Deviation from the calculated value Q1 of the improved equilibrium equation: Loss b =|Q supply,pred -Q1|.
[0063] The constraint weights are configured based on the degree of influence of the equations on energy consumption prediction, and the constraint weights w of the heat transfer equations are set accordingly. k =0.4, equilibrium equation constraint weight w b =0.3, data fitting basic loss weight w m =0.3.
[0064] The total loss function is constructed by weighting the data fitting loss, the heat transfer equation constraint loss, and the equilibrium equation constraint loss: Loss total =w m Loss mse +w k Loss k +w b Loss b .
[0065] Among them, Loss mse This represents the mean square error loss between the model's predicted values and the actual energy consumption values.
[0066] S5. Obtain the current operating data and input the current operating data into the energy consumption performance prediction model to obtain the energy consumption performance prediction result.
[0067] Furthermore, the specific details of obtaining current operating data in S5 and substituting this data into the energy consumption performance prediction model to obtain the energy consumption performance prediction results include: Cell weights are calculated based on the proportion of cell building area and the contribution of cell energy consumption. The cell energy consumption contribution rate is the proportion of historical energy consumption to the total energy consumption of the cluster. The energy consumption performance prediction results are calculated based on the predicted value and weight of the current cell, the predicted values of the cells associated with the current cell, and the comprehensive coupling strength between the current cell and the current cell.
[0068] Example 2 An energy consumption performance prediction system for building clusters, used to implement an energy consumption performance prediction method for building clusters, such as... Figure 2 As shown, it includes: Data processing unit: acquires the building cluster's past basic physical data and past operational data, and preprocesses the past basic physical data and past operational data to obtain building data, environmental data, energy consumption data, equipment operation data, and user behavior data; Feature extraction unit: A data curve is created with time as the horizontal axis and environmental data, energy consumption data, and equipment operation data as the vertical axis. Feature extraction is performed on the data curve to obtain cluster features. Model building unit: Based on building data, each data point is clustered to obtain energy consumption performance cells, and physical constraints are configured on the energy consumption performance cells to obtain the energy consumption performance prediction model; Energy consumption performance prediction unit: acquires current operating data, inputs the current operating data into the energy consumption performance prediction model to obtain the energy consumption performance prediction result.
[0069] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor invokes the computer program in the memory to implement the content of the energy consumption performance prediction method applied to building clusters.
[0070] A storage medium storing computer-executable instructions, which, when loaded and executed by a processor, implement the content of the energy consumption performance prediction method applied to building clusters.
[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical methods of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical methods of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical methods to deviate from the spirit and scope of the technical methods of the present invention.
Claims
1. A method for predicting energy consumption performance in building clusters, characterized in that, Includes the following steps: S1. Obtain the historical basic physical data and historical operation data of the building cluster, and preprocess the historical basic physical data and historical operation data to obtain building data, environmental data, energy consumption data, equipment operation data, and user behavior data; S2. Create a data curve graph with time as the horizontal axis and environmental data, energy consumption data, and equipment operation data as the vertical axis. Extract features from the data curve graph to obtain cluster features. S3. Based on the building data, each data point is clustered to obtain energy consumption performance cells; S4. Physical constraints are configured on the energy consumption performance cells to obtain the energy consumption performance prediction model. S5. Obtain the current operating data and input the current operating data into the energy consumption performance prediction model to obtain the energy consumption performance prediction result.
2. The energy consumption performance prediction method for building clusters according to claim 1, characterized in that, In S2, a data curve is created with time as the horizontal axis and environmental data, energy consumption data, and equipment operation data as the vertical axis. Feature extraction from this data curve yields the following specific cluster features: Create data curves with time on the horizontal axis and environmental data, energy consumption data, and equipment operation data on the vertical axis. Align the data curves to obtain a unified time series curve set; Standard curve sets are obtained by annotating building data and user behavior data on corresponding curves; Cluster features are obtained by combining standard curves with feature extraction. These cluster features include environmental features, energy consumption features, behavioral features, and facility features.
3. The energy consumption performance prediction method for building clusters according to claim 1, characterized in that, The specific content of the energy consumption performance cells obtained by clustering each data point based on building data in S3 includes: Building data is processed to obtain building location information, building function information, and building sharing information; Based on building location information, building function information, and building sharing information, the spatial distance correlation, functional type similarity, and energy system coupling degree between different buildings are calculated respectively. Spatial distance correlation, functional type similarity, and energy system coupling degree are used as feature vectors; The elbow rule is used to determine the cell K value. The feature vector is input into the K-means clustering algorithm to obtain the cell affiliation label of each building, and then the initial energy consumption performance cell is obtained. The energy consumption performance cell is obtained by configuring the coupling relationship for the initial energy performance cell.
4. The energy consumption performance prediction method for building clusters according to claim 1, characterized in that, The specific content of the energy consumption performance cell obtained by configuring the coupling relationship of the initial energy performance cell is as follows: Based on building geometric parameters and environmental data, the heat exchange between cells is calculated to obtain the thermal coupling strength coefficient and form a thermal coupling matrix. Based on the pipeline topology diagram, calculate the pressure loss and heat transfer loss of the shared pipeline between different cells to obtain the pipeline coupling loss coefficient and form the pipeline coupling matrix. Calculate the correlation coefficients of energy consumption time series data of different cells to form a time-period coupling matrix; By configuring weight conversion coefficients for the thermal coupling matrix, pipeline coupling matrix, and time-period coupling matrix, the comprehensive coupling strength matrix between cells is obtained.
5. The energy consumption performance prediction method for building clusters according to claim 1, characterized in that, The energy consumption performance cell is configured with a migration mechanism for data-scarce cells, as detailed below: Calculate the feature similarity between data-scarce cells and data-sufficient cells; The feature similarity includes functional type, building scale, and coupling mode, and the source cell with the highest similarity is selected as the migration object. Transfer the parameters trained in the source cells to the scarce cells.
6. The energy consumption performance prediction method for building clusters according to claim 1, characterized in that, The specific content of the energy consumption performance prediction model obtained by configuring physical constraints on energy consumption performance cells in S4 is as follows: Based on the aging of the structural facilities, loss factors are configured for the theoretical structural heat transfer equation, theoretical functional balance equation and theoretical equipment efficiency to obtain the improved structural heat transfer equation, improved functional balance equation and improved equipment efficiency threshold. Obtain the threshold for improving equipment efficiency, and define the equipment efficiency constraint as an activation function; To improve the structural heat transfer equation and the functional balance equation, constraint weights are configured to obtain the loss function.
7. The energy consumption performance prediction method for building clusters according to claim 1, characterized in that, The specific steps involved in obtaining current operating data in S5 and substituting that data into the energy consumption performance prediction model to obtain the energy consumption performance prediction results include: Cell weights are calculated based on the proportion of cell building area and the contribution of cell energy consumption. The cell energy consumption contribution rate is the proportion of historical energy consumption to the total energy consumption of the cluster. The energy consumption performance prediction results are calculated based on the predicted value and weight of the current cell, the predicted values of the cells associated with the current cell, and the comprehensive coupling strength between the current cell and the current cell.
8. An energy consumption performance prediction system for building clusters, used to implement the energy consumption performance prediction method for building clusters as described in any one of claims 1 to 7, characterized in that, include: Data processing unit: acquires the building cluster's past basic physical data and past operational data, and preprocesses the past basic physical data and past operational data to obtain building data, environmental data, energy consumption data, equipment operation data, and user behavior data; Feature extraction unit: A data curve is created with time as the horizontal axis and environmental data, energy consumption data, and equipment operation data as the vertical axis. Feature extraction is performed on the data curve to obtain cluster features. Model building unit: Based on building data, each data point is clustered to obtain energy consumption performance cells, and physical constraints are configured on the energy consumption performance cells to obtain the energy consumption performance prediction model; Energy consumption performance prediction unit: acquires current operating data, inputs the current operating data into the energy consumption performance prediction model to obtain the energy consumption performance prediction result.
9. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program, and the processor, when calling the computer program in the memory, implements the content of the energy consumption performance prediction method for building clusters as described in any one of claims 1 to 7.
10. A storage medium, characterized in that, The storage medium stores computer-executable instructions, which, when loaded and executed by a processor, implement the energy consumption performance prediction method for building clusters as described in any one of claims 1 to 7.