Cross-building energy consumption collaborative prediction method and device based on improved KAN network

Abstract

This application is applicable to the field of building energy consumption and provides a method and device for collaborative cross-building energy consumption prediction based on an improved KAN network. The method comprises: obtaining current building information and a target energy consumption prediction model for the building to be predicted, wherein the target energy consumption prediction model is determined by training an improved KAN network model based on knowledge sharing of historical building information of multiple original buildings, and the activation function of the improved KAN network model is a Mexican hat wavelet function; inputting the current building information into the target energy consumption prediction model to obtain the target energy consumption of the building to be predicted as output by the target energy consumption prediction model. This method can enhance the target energy consumption prediction model's ability to capture nonlinear relationships and multi-scale features in building energy consumption data while achieving cross-building knowledge sharing, thereby improving the prediction accuracy of building energy consumption.

Description

Technical Field

[0001] The present application belongs to the field of building energy consumption technology, and in particular relates to a method and device for collaborative prediction of cross-building energy consumption based on an improved KAN network. Background Art

[0002] With the rapid development of IoT technology and intelligent building management systems, a vast amount of building operational data is continuously being collected, including multi-dimensional information such as energy consumption, user behavior patterns, and indoor and outdoor environmental parameters. This rich data resource provides a solid foundation for building energy consumption forecasting. Data-driven deep learning methods, in particular, show promising application prospects in this field.

[0003] However, the building energy consumption prediction methods in existing technologies are mainly based on modeling based on data from a single building, which is often limited by objective factors such as sensor deployment time and equipment failure, resulting in insufficient data volume or incomplete data coverage of working conditions. At the same time, the deep learning models used in existing technologies are mainly traditional neural network structures, which have limited expression capabilities in capturing the complex nonlinear relationships and multi-scale characteristics of building energy consumption data, seriously reducing the prediction accuracy of building energy consumption. Summary of the Invention

[0004] The embodiments of the present application provide a method and device for collaborative prediction of cross-building energy consumption based on an improved KAN network, which can achieve cross-building knowledge sharing while enhancing the model's ability to capture complex nonlinear relationships and multi-scale features in building energy consumption data, thereby improving the prediction accuracy of building energy consumption.

[0005] In a first aspect, an embodiment of the present application provides a method for collaboratively predicting cross-building energy consumption based on an improved KAN network, comprising:

[0006] The current building information and target energy consumption prediction model of the building to be predicted are obtained, wherein the target energy consumption prediction model is determined by training an improved Kolmogorov-Arnold KAN network model based on knowledge sharing of historical building information of multiple original buildings, and the activation function of the improved KAN network model is the Mexican hat wavelet function; the current building information is input into the target energy consumption prediction model to obtain the target energy consumption of the building to be predicted output by the target energy consumption prediction model.

[0007] In a second aspect, an embodiment of the present application provides a cross-building energy consumption collaborative prediction device based on an improved KAN network, comprising:

[0008] An acquisition module is used to obtain the current building information and target energy consumption prediction model of the building to be predicted, wherein the target energy consumption prediction model is determined by training an improved Kolmogorov-Arnold KAN network model based on knowledge sharing of historical building information of multiple original buildings, and the activation function of the improved KAN network model is a Mexican hat wavelet function;

[0009] The input module is used to input the current building information into the target energy consumption prediction model to obtain the target energy consumption of the building to be predicted output by the target energy consumption prediction model.

[0010] In a third aspect, an embodiment of the present application provides a terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any of the methods of the first aspect when executing the computer program.

[0011] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the method of any one of the first aspects.

[0012] In a fifth aspect, an embodiment of the present application provides a computer program product, which, when executed on a terminal device, enables the terminal device to execute any one of the methods in the first aspect above.

[0013] The embodiments of the present application provide a method and device for collaborative prediction of cross-building energy consumption based on an improved KAN network, the method comprising: obtaining current building information and a target energy consumption prediction model of a building to be predicted, wherein the target energy consumption prediction model is determined by training an improved Kolmogorov-Arnold KAN network model based on knowledge sharing of historical building information of multiple original buildings, and the activation function of the improved KAN network model is a Mexican hat wavelet function; inputting the current building information into the target energy consumption prediction model to obtain the target energy consumption of the building to be predicted output by the target energy consumption prediction model. Utilizing the above technical solution, the target energy consumption prediction model is determined by training an improved KAN network model based on cross-building building information, and the activation function of the improved KAN network model is a Mexican hat wavelet function. This can enhance the target energy consumption prediction model's ability to capture nonlinear relationships and multi-scale features in building energy consumption data while achieving cross-building knowledge sharing, thereby improving the prediction accuracy of building energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0015] Figure 1 This is a flow chart of a cross-building energy consumption collaborative prediction method based on an improved KAN network provided in one embodiment of the present application;

[0016] Figure 2 This is a schematic diagram of the structure of an improved KAN network model provided in one embodiment of the present application;

[0017] Figure 3 This is a flow chart of a cross-building energy consumption collaborative prediction method based on an improved KAN network provided in another embodiment of the present application;

[0018] Figure 4 This is a schematic diagram of a cluster analysis process provided by an embodiment of the present application;

[0019] Figure 5 This is a flow chart of a cross-building energy consumption collaborative prediction method based on an improved KAN network provided in another embodiment of the present application;

[0020] Figure 6 This is a schematic diagram of a federated learning process provided by an embodiment of the present application;

[0021] Figure 7 1 is a schematic diagram of the architecture of a cross-building energy consumption collaborative prediction method based on an improved KAN network provided in another embodiment of the present application;

[0022] Figure 8 This is a structural block diagram of a cross-building energy consumption collaborative prediction device based on an improved KAN network provided by an embodiment of the present application;

[0023] Figure 9 This is a structural diagram of a terminal device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0024] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0025] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0026] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0027] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0028] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0029] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0030] It should be noted that the information collection process (such as the facial image collection process, the building information collection process, etc.) / feature extraction process involved in this application is performed with the user's knowledge and permission, that is, the information collection process / feature extraction process complies with the requirements of laws and regulations and does not constitute an act that harms the public interest.

[0031] It can be considered that building energy consumption prediction, as a core component of the building energy conservation management system, is of great strategic significance for achieving optimal energy utilization, reducing operating costs and reducing carbon emissions.

[0032] Existing building energy consumption prediction methods are primarily based on modeling based on data from a single building. This approach has several significant limitations. First, data collection from individual buildings is often limited by objective factors such as sensor deployment time and equipment failures, resulting in insufficient data or incomplete data coverage of operating conditions, making it difficult to support high-precision model training. Second, models trained solely on data from a single building are unable to cope with the addition of new buildings or changes in operating modes, and their generalization capabilities are severely limited. Furthermore, this approach fails to fully utilize the common knowledge that may exist between different buildings, resulting in a waste of valuable data resources and low overall data utilization efficiency.

[0033] To address these issues, two new cross-building machine learning paradigms, transfer learning and federated learning, have emerged. However, both approaches currently have shortcomings. For one thing, existing transfer learning mechanisms for selecting source buildings have limitations. Source selection currently lacks quantitative criteria and relies excessively on subjective experience, making it difficult to ensure optimal similarity of source data. Furthermore, existing standard federated learning frameworks assume homogeneous data distribution across all participants. However, in real-world scenarios, significant differences in building function, scale, and climate conditions lead to highly heterogeneous data distribution.

[0034] On the other hand, the deep learning models currently used in building energy consumption prediction tasks are primarily based on traditional neural network structures, which present multiple limitations. For example, these deep learning models have limited expressive power in capturing the complex nonlinear relationships and multi-scale features found in building energy consumption data. Secondly, with limited data, the performance of traditional deep learning models significantly degrades, making effective learning difficult. Furthermore, traditional black-box models lack interpretability, making it difficult to understand the underlying basis for model decisions—a significant drawback in energy management decision support. Finally, traditional deep learning models exhibit unstable performance when migrated across buildings, making it difficult to flexibly adapt to the data characteristics and distribution patterns of different building types.

[0035] In summary, the existing cross-building machine learning building energy consumption prediction method is not perfect, and the deep learning model architecture has certain limitations. Based on this, in order to address the key technical bottlenecks in the existing technology, such as data island problems, imperfect cross-domain machine learning paradigms, and model architecture limitations, the embodiment of the present application provides a cross-building energy consumption collaborative prediction method based on an improved KAN (Kolmogorov Arnold Network) network model. By introducing the improved KAN model as the basic model architecture and integrating the two cross-domain machine learning paradigms of transfer learning and federated learning, a complete cross-building energy consumption prediction technology system considering building similarities is constructed, thereby improving the accuracy of energy consumption prediction for individual buildings.

[0036] Figure 1This is a flow chart of a cross-building energy consumption collaborative prediction method based on an improved KAN network provided by an embodiment of the present application. As an example and not a limitation, the method can be applied to terminal devices. The embodiment of the present application does not impose any restrictions on the specific type of terminal devices. For example, the terminal device can be a terminal computer, or a server, etc. Figure 1 As shown, the method includes:

[0037] S101. Obtain current building information and a target energy consumption prediction model of a building to be predicted.

[0038] Among them, the target energy consumption prediction model is determined by training an improved Kolmogorov-Arnold KAN network model based on the knowledge sharing of historical architectural information of multiple original buildings. The activation function of the improved KAN network model is the Mexican hat wavelet function.

[0039] The building to be predicted may refer to a building for which energy consumption prediction is to be performed. The current building information may be considered as the building information currently related to the building to be predicted. The type of building information is not limited and may involve indoor and outdoor environmental parameters, time variables and other related data.

[0040] Optionally, the current building information includes at least current time data, current environmental data, and building area data. The current time data may include, for example, the current month, the current day type (i.e., Monday to Sunday), and the current hour (i.e., 0 to 23); the current environmental data may include, for example, dry-bulb temperature, wet-bulb temperature, wind speed, and other data; and the building area data represents the total building area of ​​the building to be predicted.

[0041] The target energy consumption prediction model can be understood as the energy consumption prediction model corresponding to the building to be predicted, which can be used to predict the energy consumption of the building to be predicted; the target energy consumption prediction model can be considered to be determined by training an improved Kolmogorov-Arnold KAN network model based on the knowledge sharing of historical building information of multiple original buildings, wherein the multiple original buildings may include the building to be predicted, or may include other buildings different from the building to be predicted, or the building to be predicted may not be one of the multiple original buildings, as long as there are similar buildings in the multiple original buildings that have a similar category relationship with the building to be predicted. This embodiment does not limit this. Historical building information is building information related to the history of the original building, for example, it may include at least historical time data, historical environmental data, building area data and historical energy consumption data.

[0042] Furthermore, this embodiment employs an improved KAN network model for prediction. The KAN network model can be considered a novel neural network architecture based on the Kolmogorov-Arnold representation theorem, with learnable activation functions and efficient function approximation capabilities. However, because the KAN network model's inherent network structure utilizes different activation functions, it is difficult to directly use in transfer learning and federated learning. This embodiment innovatively selects the Mexican hat wavelet function as the baseline activation function for the KAN network model, and the improved network model is referred to as the Improved-KAN model (IKAN for short). This improves the network model's ability to capture multi-level data features, simplifies the model structure, and enhances prediction accuracy.

[0043] Figure 2 This is a schematic diagram of the structure of an improved KAN network model provided by an embodiment of the present application. Figure 2 As shown in the figure, the improved KAN network model can be composed of multiple activation functions. The benchmark of each activation function can be the Mexican hat wavelet function, and the parameters of the Mexican hat wavelet function corresponding to different activation functions are different. The Mexican hat wavelet function as an activation function can be used to control the learnable hyperparameters of the function height and width. At the same time, the Mexican hat wavelet function enables the IKAN model to analyze data at different scales, thereby effectively capturing and representing the multi-level features of the data. The simple expression of the IKAN model can be the weighted sum of multiple activation functions, such as , are multiple learnable nonlinear activation functions (i.e., Mexican hat wavelet functions), where the symbol " " can be understood as the weighted sum of multiple learnable nonlinear activation functions, that is, It can be further expressed as , They are Each nonlinear activation function can be used Indicates that, where p is the number of input variables and q is the number of output variables, multiple learnable nonlinear activation functions can be represented as multiple , each in, is an independent variable, representing time or position, which determines the evaluation point of the wavelet on the time axis; is the scale parameter, which controls the width of the wavelet. Make the wavelet wider, suitable for capturing low-frequency components; smaller Make the wavelet narrower and suitable for high-frequency analysis; w are learnable hyperparameters that control the height and width of the Mexican hat wavelet function.

[0044] This embodiment does not limit the specific means of obtaining the target energy consumption prediction model. For example, the terminal device can pre-train the target energy consumption prediction model, and each time energy consumption prediction is needed, the target energy consumption prediction model can be obtained directly from the terminal device, or the terminal device can obtain the target energy consumption prediction model through interactive communication with other devices. Alternatively, this embodiment can also comprehensively consider multiple factors and realize the real-time construction of the target energy consumption prediction model of the building to be predicted by selecting different cross-building machine learning paradigms and sharing the knowledge of historical building information of multiple original buildings. On the one hand, the selection of a cross-building machine learning paradigm can focus on key factors such as privacy constraints, data heterogeneity, and communication costs of the target scenario. On the other hand, the selection of a cross-building machine learning paradigm can combine the characteristics of different cross-domain machine learning paradigms. For example, transfer learning ultimately trains the target energy consumption prediction model of the target domain, while federated learning ultimately trains a target energy consumption prediction model for each participating building. Based on the above factors, a reasonable cross-domain machine learning paradigm can be selected to construct the target energy consumption prediction model of the building to be predicted.

[0045] Optionally, cross-domain machine learning paradigms can include transfer learning and federated learning. Transfer learning aims to transfer knowledge learned from a source domain to improve the performance of data-driven models in a target domain, with model parameter adjustments focused on the target domain. Transfer learning is suitable for situations where target domain data is limited but a large amount of relevant source domain data is available. The main steps in implementing transfer learning are as follows: First, source domain data relevant to the target domain must be selected; then, a model must be trained on the source domain data or a pre-trained model must be obtained that has learned the feature representations and patterns of the source domain; then, adjustments to the pre-trained model must be made based on the characteristics of the target domain; then, knowledge transfer is performed, applying the adjusted model to the target domain, leveraging the knowledge learned from the source domain to aid the target domain model's learning; finally, performance evaluation and optimization are performed to assess the model's performance on target domain data. Compared to traditional machine learning paradigms, transfer learning can fully leverage cross-domain data resources to build predictive models, significantly reducing the data requirements and computational costs required to develop data-driven models for new tasks.

[0046] Federated learning can be a cross-domain distributed learning method that allows multiple participants to collaborate on training models while maintaining data privacy. This method does not require centralized data storage. Instead, it optimizes the model through a cyclic iteration of client-side local training and server parameter aggregation, allowing the model to be jointly trained while protecting privacy. Federated learning can be described as a three-step learning process, namely (1) initializing a global model on the central server and distributing it to all participating clients; (2) the client performs model training based on local data; (3) uploading local model information to the central server for global model update using privacy-preserving methods, such as using gradients or the average value of model parameters. This process ensures data privacy through a privacy-preserving protocol and achieves model convergence through multiple rounds of parameter interaction. Compared with traditional deep learning, federated learning has significant advantages in solving data silo problems, reducing the risk of privacy leakage, and improving cross-domain collaboration efficiency. It is particularly suitable for multi-party collaborative modeling scenarios such as building energy consumption prediction.

[0047] S102: Input the current building information into the target energy consumption prediction model to obtain the target energy consumption of the building to be predicted output by the target energy consumption prediction model.

[0048] This embodiment provides a cross-building energy consumption collaborative prediction method based on an improved KAN network. The target energy consumption prediction model is determined by training an improved KAN network model based on building information across buildings. The activation function of the improved KAN network model is a Mexican hat wavelet function. While achieving cross-building knowledge sharing, it can enhance the target energy consumption prediction model's ability to capture nonlinear relationships and multi-scale features in building energy consumption data, thereby improving the prediction accuracy of building energy consumption.

[0049] Figure 3 This is a flow chart of a cross-building energy consumption collaborative prediction method based on an improved KAN network provided by another embodiment of the present application. This embodiment further optimizes the target energy consumption prediction model of the building to be predicted to: obtain historical building information of multiple original buildings, and the multiple original buildings at least include the building to be predicted; filter the source domain building information corresponding to the building to be predicted from the historical building information of the multiple original buildings, and the source domain building information is the historical building information with the highest similarity to the historical building information of the building to be predicted among the historical building information of the multiple original buildings; and adopt the transfer learning method to determine the target energy consumption prediction model of the building to be predicted based on the source domain building information. Figure 3 As shown, the method includes:

[0050] S201: Acquire current building information of a building to be predicted and historical building information of multiple original buildings.

[0051] The multiple original buildings at least include the building to be predicted.

[0052] S202: Filter source domain building information corresponding to the building to be predicted from historical building information of multiple original buildings.

[0053] The source domain building information is the historical building information with the highest similarity to the historical building information of the building to be predicted among the historical building information of multiple original buildings.

[0054] It can be considered that traditional transfer learning generally uses collected historical building data directly as source domain data. However, this embodiment filters the historical building data to select historical building data with high similarity to the target domain data as source domain data. That is, the historical building information with the highest similarity to the historical building information of the building to be predicted from the multiple original buildings is used as the source domain building information. The method for determining the source domain building information in this embodiment is not limited. For example, a pre-set neural network model can be used to directly output the source domain building information corresponding to the building to be predicted by inputting the historical building information of the multiple original buildings into the neural network model. Alternatively, cluster analysis can be performed on the historical building information of the multiple original buildings to obtain similar buildings in the multiple original buildings that have a similar category relationship with the building to be predicted; the historical building information of the similar buildings is then determined as the source domain building information corresponding to the building to be predicted. The cluster analysis process is not limited. For example, an autoencoder or other screening method can be used as the cluster analysis processing method to filter out other data with high similarity to the historical building information of the building to be predicted and determine it as the source domain building information.

[0055] As a feasible implementation method, cluster analysis is performed on the historical building information of multiple original buildings to obtain similar buildings in the multiple original buildings that have similar category relationships with the building to be predicted, including:

[0056] For each original building's historical building information, data compression processing is performed on the historical building information of each original building to obtain target building features of the historical building information of each original building;

[0057] Calculate the Euclidean distance between each target building feature to obtain the distance matrix corresponding to multiple original buildings;

[0058] The distance matrix is ​​categorized and similar buildings with similar category relationships to the building to be predicted are obtained from multiple original buildings.

[0059] Figure 4 This is a flow chart of a cluster analysis provided by an embodiment of the present application. Figure 4As shown, cluster analysis can be implemented using an autoencoder model. Its operating mechanism mainly uses a neural network to learn the latent representation of the data and extract key features while reducing the dimensionality, effectively screening out source domain data with high similarity to the target domain. The autoencoder model can be composed of an encoder and a decoder. The encoder is responsible for compressing the input data into a low-dimensional bottleneck layer, while the decoder is dedicated to reconstructing the original input data based on the latent features. During the training phase, the autoencoder model forces the bottleneck layer to retain key information in the data by continuously minimizing the reconstruction loss. After training is completed, the latent features output by the encoder are used to construct a distance matrix, and a hierarchical clustering strategy is used for category division.

[0060] Taking the historical building information of the original building including the local data of building #A, building #B and building #C as an example, for the historical building information of each original building, an encoder can be used to perform data compression processing on the historical building information of each original building. For example, when the local data includes time variables, indoor and outdoor environment and energy consumption data, the encoder can compress the high-dimensional local data into one-dimensional target building features. The target building features can be seasonal data, or time feature data that encodes working days and rest days. The decoder can calculate the Euclidean distance between each target building feature (such as output #1, output #2 and output #3) to obtain the distance matrix corresponding to multiple original buildings. Finally, the hierarchical clustering strategy is used to classify the distance matrix, and it can be obtained that building #A and building #B belong to category #1, and building #C belongs to category #2. Assuming that the building to be predicted is building #A, then through the autoencoder clustering analysis, it can be determined that the similar building with a similar category relationship with building #A is building #B.

[0061] S203: Using a transfer learning method, a target energy consumption prediction model of the building to be predicted is determined based on the source domain building information.

[0062] This step can use the transfer learning method to directly transfer all parameters of the pre-trained model trained on the source domain data to the target task as the initial state of the target domain data model training. For example, the improved KAN network model can be trained based on the source domain building information to obtain the source domain energy consumption prediction model; the weight parameters of all layers in the source domain energy consumption prediction model are completely retained, and further training is performed on the target domain data. By fine-tuning these parameters to adapt to the specific needs of the target task, the source domain energy consumption prediction model is processed based on the historical building information of the building to be predicted, and the target energy consumption prediction model of the building to be predicted is obtained. On this basis, the general feature representations already learned in the source domain model can be used to accelerate the training process of the target domain and improve model performance.

[0063] S204: Input the current building information into the target energy consumption prediction model to obtain the target energy consumption of the building to be predicted output by the target energy consumption prediction model.

[0064] This embodiment provides a cross-building energy consumption collaborative prediction method based on an improved KAN network. The method screens source domain building information corresponding to a building to be predicted from the historical building information of multiple original buildings. The source domain building information is the historical building information with the highest similarity to the historical building information of the building to be predicted among the multiple original buildings. The method then adopts a transfer learning method to determine a target energy consumption prediction model for the building to be predicted based on the source domain building information. This can improve the efficiency of source domain knowledge transfer, solve the problems of insufficient data coverage and poor generalization ability, and further improve the prediction accuracy of building energy consumption.

[0065] Figure 5 This is a flow chart of a cross-building energy consumption collaborative prediction method based on an improved KAN network provided by another embodiment of the present application. This embodiment further optimizes the target energy consumption prediction model of the building to be predicted as follows: the improved KAN network model is trained based on the historical building information of the building to be predicted to obtain the current energy consumption prediction model; in each learning round of federated learning, the current random data sent by the server is dynamically obtained, and the current random data is input into the current energy consumption prediction model to obtain the current energy consumption prediction value; the current energy consumption prediction value is dynamically sent to the server so that the server clusters the current energy consumption prediction values ​​corresponding to multiple original buildings to obtain similar buildings with similar category relationships with the building to be predicted; the target model parameters returned by the server based on the current model parameters of the current energy consumption prediction model of the target building are received, and the target building includes the building to be predicted and similar buildings; the current energy consumption prediction model of the building to be predicted is dynamically updated based on the target model parameters until the updated current energy consumption prediction model reaches convergence; the updated current energy consumption prediction model that has reached convergence is determined as the target energy consumption prediction model of the building to be predicted. Figure 5 As shown, the method includes:

[0066] S301: Acquire current building information of the building to be predicted, and train an improved KAN network model based on historical building information of the building to be predicted to obtain a current energy consumption prediction model.

[0067] S302. In each learning round of federated learning, dynamically obtain current random data sent by the server, and input the current random data into the current energy consumption prediction model to obtain a current energy consumption prediction value.

[0068] S303. Dynamically send the current energy consumption prediction value to the server, so that the server clusters the current energy consumption prediction values ​​corresponding to multiple original buildings to obtain similar buildings that have a similar category relationship with the building to be predicted.

[0069] S304: Receive target model parameters returned by the server based on current model parameters of the current energy consumption prediction model of the target building, where the target building includes the building to be predicted and similar buildings.

[0070] S305 : Dynamically update the current energy consumption prediction model of the building to be predicted based on the target model parameters until the updated current energy consumption prediction model reaches convergence.

[0071] S306: Determine the updated current energy consumption prediction model that has reached convergence as the target energy consumption prediction model of the building to be predicted.

[0072] S307: Input the current building information into the target energy consumption prediction model to obtain the target energy consumption of the building to be predicted output by the target energy consumption prediction model.

[0073] The current random data may be a set of data randomly generated by the server and used to perform clustering processing on the current energy consumption prediction values ​​corresponding to multiple original buildings. The current energy consumption prediction value is the energy consumption prediction value output after the current random data is input into the current energy consumption prediction model.

[0074] In the specific process of training the model, after obtaining the current energy consumption prediction model in each round of training, the model parameters of the current energy consumption prediction model can be dynamically updated to obtain the updated current energy consumption prediction model, and then the next round of local model training can be carried out until the updated current energy consumption prediction model reaches convergence. Finally, the updated current energy consumption prediction model that has reached convergence can be determined as the target energy consumption prediction model of the building to be predicted.

[0075] Specifically, the update of the model parameters can be based on the result of clustering the multiple original buildings by the server. For example, each client can dynamically obtain the current random data sent by the server, and dynamically send the current energy consumption prediction value output by its own current energy consumption prediction model based on the current random data to the server. After receiving the current energy consumption prediction values ​​corresponding to the multiple original buildings, the server performs clustering processing on the current energy consumption prediction values ​​corresponding to the multiple original buildings to obtain a group of similar buildings, and can obtain similar buildings that have a similar category relationship with the building to be predicted; the server can determine the target model parameters according to the current model parameters of each current energy consumption prediction model corresponding to the similar building group, and The target model parameters are returned to the clients of the similar building group, so that the clients of the similar building group can use the federated learning method to receive the target model parameters and dynamically update their own current energy consumption prediction models. Taking the client corresponding to the building to be predicted as an example, the client can dynamically update the current energy consumption prediction model of the building to be predicted based on the target model parameters. That is, after each time the client receives the current random data sent by the server, it can realize the dynamic update of the model parameters of the current energy consumption prediction model according to the above steps until the updated current energy consumption prediction model reaches convergence, and then the updated current energy consumption prediction model that has reached convergence can be determined as the target energy consumption prediction model of the building to be predicted.

[0076] Furthermore, there is no limit to the means by which the server calculates the target model parameters. For example, one of the current model parameters can be selected as the target model parameter simply by comparing the current model parameters of each current energy consumption prediction model corresponding to the similar building group. For example, the largest current model parameter can be selected as the target model parameter from the current model parameters corresponding to the building to be predicted and the current model parameters corresponding to similar buildings that have a similar category relationship with the building to be predicted. Alternatively, the target model parameter can be obtained by performing certain calculations on the current model parameters of each current energy consumption prediction model corresponding to the similar building group.

[0077] Optionally, the target model parameters are obtained by averaging the server's current model parameters for the building being predicted with those for similar buildings. Based on this, the averaging aggregation method balances the contributions of each client to the global model simply by arithmetic averaging, without relying on complex weighting strategies. This approach is not only easy to implement and understand, but also, to a certain extent, avoids global model parameter deviations caused by abnormal data from individual clients or significant differences in model training.

[0078] Figure 6 This is a flowchart of a federated learning process provided by an embodiment of the present application. Figure 6As shown in the figure, in the initial stage of federated learning, the server can initialize the global model and distribute it to all clients for local training. Taking the current existence of building #A, building #B and building #C as an example, each client can be responsible for the model training of one of the buildings, that is, each client can perform local training based on the local data of its own building to obtain the local model of the current round (that is, the improved KAN network model is trained based on the historical building information of the building to be predicted to obtain the current energy consumption prediction model); then, the server can generate a set of random but fixed input data (that is, the current random data) and send it to each client, so that each client can generate a model prediction value (that is, the current energy consumption prediction value) based on its own local model.

[0079] Using these model predictions as the basis for cluster analysis, clustering results can be generated for personalized federated learning. The core principle is that models with similar parameters produce similar outputs given the same input. This allows the classification results for similar buildings to be dynamically adjusted in each round of federated learning, achieving more accurate and adaptable personalized federated learning. For example, during the current round of cluster analysis, a distance matrix can be generated based on the model predictions (e.g., output #1, output #2, and output #3). This distance matrix is ​​then used to generate hierarchical clustering results, indicating that buildings #A and #B belong to category #1, and building #C belongs to category #2. The server can then average the current model parameters for buildings in the same category to obtain the target model parameters, thereby updating the current energy consumption prediction model for buildings in the same category. In the next round of cluster analysis, the hierarchical clustering results for similar buildings can be updated based on the new model predictions. This allows the dynamic adjustment of the current energy consumption prediction model for similar buildings in each round of federated learning based on the different hierarchical clustering results.

[0080] It can be argued that, within the framework of federated learning, the parameters aggregated by the IKAN model proposed in this embodiment are not simple weights, but rather parameter changes of the Mexican hat wavelet function, namely, the current model parameters. Furthermore, this embodiment employs an average aggregation method for IKAN model parameter aggregation. Specifically, this method operates as follows: at the end of each training round, each client sends the Mexican hat wavelet function parameters of the IKAN model, obtained through local training, to a central server. After receiving the current model parameters from each client, the central server calculates the average of the current model parameters corresponding to the hierarchical clustering results of similar buildings and uses this average as the aggregated global model parameter (i.e., the target model parameter). This new global model parameter is then distributed to all clients corresponding to similar buildings. Each client can then update the current model parameters of the current energy consumption prediction model based on the target model parameters and then proceed to the next round of local training until the updated current energy consumption prediction model reaches convergence.

[0081] More specifically, the number of central servers can be one or more. For example, in each round of clustering analysis, different central servers can be divided according to different hierarchical clustering results. For example, the average value between the current model parameters of buildings of the same category can be calculated by the same central server, and the central server distributes the calculated target model parameters to the clients corresponding to the buildings of this category; while the average values ​​between the current model parameters of buildings of different categories are calculated by different central servers.

[0082] It's important to note that standard federated learning fixes participating buildings during distributed training and cannot dynamically adjust the grouping of similar buildings, making it difficult to effectively leverage similarities between buildings to improve model performance. This embodiment, by employing a federated learning strategy based on dynamic model parameters, dynamically identifies and leverages similarities between buildings while protecting data privacy. This not only improves the model's adaptability to the data distribution of various clients, but also enhances the flexibility and accuracy of the federated learning framework, enabling it to better cope with the dynamic changes and heterogeneity of building data, thus achieving dynamic and personalized federated learning.

[0083] Figure 7 This is a schematic diagram of the architecture of a cross-building energy consumption collaborative prediction method based on an improved KAN network provided by another embodiment of the present application, with reference to Figure 7 In the data acquisition phase, step S1 can acquire multi-source building operation data. In this embodiment, five building communities with no more than 20 buildings can be selected. Table 1 summarizes the characteristics of different building communities. The number of buildings in each building community ranges from 5 to 19, and the area of ​​a single building ranges from 943.9 m2 to 70,999.9 m2. The main building types are educational and office purposes.

[0084] Operational data for each building is collected hourly over a period of approximately two years. This data covers key information such as building type and floor area. It also includes time variables, energy consumption data, and indoor and outdoor environmental parameters such as dry-bulb temperature, wet-bulb temperature, and wind speed. Related data such as time variables are also included. Table 2 and Table 2 summarize several key numerical variables included in different building communities. Indoor and outdoor environmental parameters can include dry-bulb temperature, wet-bulb temperature, and wind speed. It can be seen that the outdoor environmental conditions within each building community are highly similar.

[0085] Table 1 Summary of characteristics of different building communities

[0086]

[0087] Table 2 Key numerical variables included in different building communities

[0088]

[0089] Table 2 Key numerical variables included in different building communities

[0090]

[0091] Step S2 can select a strategy based on the application scenario requirements, that is, select an appropriate cross-domain machine learning paradigm, including a transfer learning strategy or a federated learning strategy. Because the application scenario of this embodiment meets the requirements of both cross-domain machine learning paradigms, both transfer learning and federated learning paradigms can be used to construct a cross-building energy consumption prediction model.

[0092] Step S3 can be divided into steps A3 and B3. Step A3 can include an autoencoder screening source domain data, that is, selecting a building as a local data set from the collected multi-source building data, that is, determining the target domain data; then, the autoencoder can be used as a clustering processing means to screen out other data with high similarity to the target domain data and determine it as the source domain data; finally, all model parameters trained on the source domain data can be migrated and fine-tuned on the target domain to obtain the target domain energy consumption prediction model (i.e., the target energy consumption prediction model). Step B3 can include a strategy based on dynamic model parameters. That is, in this embodiment, federated learning can adopt a strategy based on dynamic model parameters. This strategy can achieve dynamic personalization, allowing similar building groupings to change over time, better adapting to the dynamic and complex requirements of data characteristics, and can operate effectively under computing resource constraints.

[0093] Whether choosing transfer learning or federated learning as the cross-building machine learning paradigm, implementing building energy consumption prediction requires selecting input variables for the prediction model. In this embodiment, since building operation data typically exhibits significant seasonality, three temporal variables can be selected as input: month, day type (i.e., Monday to Sunday), and hour (i.e., 0 to 23). These variables serve as representative variables to reflect the influence of implicit information (such as unpredictable factors like the number of residents). Environmental variables, on the other hand, can describe seasonal variations in outdoor conditions and significantly impact building operation patterns and energy consumption. Furthermore, considering that differences in scale within the same type of building may lead to completely different operating patterns, the total building area is also used in the predictive modeling. During the data preprocessing stage, traditional normalization methods can be used to normalize the numerical input variables. This allows the construction of an improved KAN network prediction model (i.e., the target energy consumption prediction model) based on these preprocessed input variables.

[0094] Step S4 can implement the building energy consumption prediction task based on the constructed improved KAN network prediction model. For example, the improved KAN network prediction model can be used to predict the building's hourly electricity consumption based on data such as time, environmental variables, and building characteristics.

[0095] In addition, this embodiment can also use multi-layer perceptron (MLP) predictions as a control group to evaluate the performance advantages of the proposed transfer learning and federated learning methods. MLP can be a basic feedforward artificial neural network consisting of multiple layers, including an input layer, a hidden layer, and an output layer. Each layer consists of multiple neurons, connected by weights. MLP is trained using a supervised learning technique called backpropagation, optimizing weights using the chain rule to minimize prediction error.

[0096] Comparative analysis of implementation results revealed that the proposed transfer learning method performed significantly better in energy consumption forecasting tasks than the traditional localized MLP modeling and prediction method, reducing the mean energy consumption prediction error by 7.31%. The federated learning method also demonstrated significant advantages, reducing the mean energy consumption prediction error by 6.70%. Furthermore, the prediction technology based on the IKAN model further improved prediction accuracy by 8.06% and 10.07%, respectively, compared to the previous two methods. This demonstrates that both transfer learning and federated learning methods can effectively mine knowledge from building data and enable cross-building sharing, effectively resolving the problem of building data silos. They enable cross-building knowledge transfer without sharing original data, thereby improving the accuracy of energy consumption forecasts for individual buildings. Furthermore, the IKAN model, with its unique neural network architecture, is more efficient than traditional MLP models in both transfer and federated learning methods for building energy consumption forecasting. It can adapt to the diverse characteristics of buildings of different sizes and types, exhibiting strong generalization and practical value. It provides more accurate prediction support for building energy management systems, contributing to energy conservation, emission reduction, and the achievement of carbon neutrality.

[0097] From the above description, it can be found that the core ideas of the cross-building energy consumption collaborative prediction method based on the improved KAN network in this embodiment mainly include: on the one hand, a transfer learning strategy based on data similarity is designed. Unlike the existing transfer learning that relies on experience to select source domain buildings, this embodiment uses the clustering method of the autoencoder model to systematically screen source domain data with high similarity to the target building, thereby improving the efficiency of source domain knowledge transfer and solving the problems of insufficient data coverage and poor generalization ability.

[0098] On the other hand, in response to the heterogeneity of building data, a federated learning strategy based on dynamic model parameters was proposed. The grouping of similar buildings was dynamically adjusted through model prediction results, effectively solving the data heterogeneity problem caused by differences in building operation rules. The model can adapt to the data distribution differences of different types of buildings, improve the dynamic adjustment ability of the model, and provide an efficient and scalable solution for building energy conservation management.

[0099] At the same time, to address the problem of limited expressive power of traditional neural network models, an improved KAN model based on the Mexican hat wavelet function was designed, which enhanced the model's ability to capture multi-scale nonlinear features and its performance in small sample scenarios. At the same time, it simplified the network structure, making the model more interpretable and significantly improving the accuracy of energy consumption prediction in small sample scenarios.

[0100] Therefore, this embodiment, based on an improved KAN network, uses a collaborative cross-building energy consumption prediction method to address the data silo problem and technical bottlenecks of insufficient model generalization in the field of building energy consumption prediction. It achieves cross-building knowledge sharing and collaborative prediction through an innovative machine learning paradigm. It also uses an improved KAN model as the neural network architecture, significantly improving the accuracy of single-building energy consumption prediction while protecting data privacy. This provides effective technical support for building energy conservation, emission reduction, and achieving carbon neutrality. Therefore, this embodiment can be applied to various building systems requiring energy consumption prediction, and is particularly suitable for scenarios where data on a single building is limited but data on similar building types is abundant.

[0101] Corresponding to the cross-building energy consumption collaborative prediction method based on the improved KAN network in the above embodiment, Figure 8 This is a structural block diagram of a cross-building energy consumption collaborative prediction device based on an improved KAN network provided in an embodiment of the present application. For the sake of convenience, only the parts related to the embodiment of the present application are shown.

[0102] Reference Figure 8 , the device comprises:

[0103] Acquisition module 401 is used to obtain current building information and a target energy consumption prediction model of the building to be predicted, wherein the target energy consumption prediction model is determined by training an improved Kolmogorov-Arnold KAN network model based on knowledge sharing of historical building information of multiple original buildings, and the activation function of the improved KAN network model is a Mexican hat wavelet function;

[0104] The input module 402 is used to input the current building information into the target energy consumption prediction model to obtain the target energy consumption of the building to be predicted output by the target energy consumption prediction model.

[0105] This embodiment provides a cross-building energy consumption collaborative prediction device based on an improved KAN network. An acquisition module acquires the current building information and a target energy consumption prediction model of the building to be predicted. The target energy consumption prediction model is determined by training an improved Kolmogorov-Arnold KAN network model based on knowledge sharing of historical building information from multiple original buildings. The activation function of the improved KAN network model is a Mexican hat wavelet function. The current building information is input into the target energy consumption prediction model through an input module to obtain the target energy consumption of the building to be predicted as output by the target energy consumption prediction model. Using this device, the target energy consumption prediction model is determined by training an improved KAN network model based on cross-building building information. The activation function of the improved KAN network model is a Mexican hat wavelet function. This device can achieve cross-building knowledge sharing while enhancing the target energy consumption prediction model's ability to capture nonlinear relationships and multi-scale features in building energy consumption data, thereby improving the accuracy of building energy consumption prediction.

[0106] Optionally, the acquisition module includes:

[0107] an acquisition unit, configured to acquire historical building information of a plurality of original buildings, wherein the plurality of original buildings at least includes the building to be predicted;

[0108] An information screening unit is used to screen source domain building information corresponding to the building to be predicted from the historical building information of the plurality of original buildings, wherein the source domain building information is the historical building information having the highest similarity to the historical building information of the building to be predicted among the historical building information of the plurality of original buildings;

[0109] The model determination unit is used to determine the target energy consumption prediction model of the building to be predicted based on the source domain building information by adopting the transfer learning method.

[0110] Optionally, the information screening unit includes:

[0111] The cluster analysis subunit is used to perform cluster analysis on the historical building information of multiple original buildings to obtain similar buildings in the multiple original buildings that have a similar category relationship with the building to be predicted;

[0112] The determination subunit is used to determine the historical building information of similar buildings as the source domain building information corresponding to the building to be predicted.

[0113] Optionally, the cluster analysis subunit is specifically used to:

[0114] For each original building's historical building information, data compression processing is performed on the historical building information of each original building to obtain target building features of the historical building information of each original building;

[0115] Calculate the Euclidean distance between each target building feature to obtain the distance matrix corresponding to multiple original buildings;

[0116] The distance matrix is ​​categorized and similar buildings with similar category relationships to the building to be predicted are obtained from multiple original buildings.

[0117] Optionally, the model determination unit is specifically configured to:

[0118] The improved KAN network model is trained based on the source domain building information to obtain the source domain energy consumption prediction model;

[0119] Based on the historical building information of the building to be predicted, the source domain energy consumption prediction model is subjected to knowledge transfer processing to obtain the target energy consumption prediction model of the building to be predicted.

[0120] Optionally, the acquisition module is specifically used to:

[0121] The improved KAN network model is trained based on the historical building information of the building to be predicted to obtain the current energy consumption prediction model;

[0122] In each round of federated learning, the current random data sent by the server is dynamically obtained and input into the current energy consumption prediction model to obtain the current energy consumption prediction value;

[0123] Dynamically sending the current energy consumption prediction value to the server so that the server can cluster the current energy consumption prediction values ​​corresponding to multiple original buildings and obtain similar buildings that have a similar category relationship with the building to be predicted;

[0124] receiving target model parameters returned by the server based on current model parameters of a current energy consumption prediction model of a target building, where the target building includes a building to be predicted and similar buildings;

[0125] Dynamically update the current energy consumption prediction model of the building to be predicted based on the target model parameters until the updated current energy consumption prediction model reaches convergence;

[0126] The updated current energy consumption prediction model that has reached convergence is determined as the target energy consumption prediction model of the building to be predicted.

[0127] Optionally, the target model parameters are model parameters obtained by the server performing mean processing on the current model parameters corresponding to the building to be predicted and the current model parameters corresponding to similar buildings.

[0128] Optionally, the current building information includes at least current time data, current environment data and building area data.

[0129] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of this application. Their specific functions and technical effects can be found in the method embodiment section and will not be repeated here.

[0130] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0131] The embodiment of the present application also provides a terminal device, Figure 9 This is a schematic diagram of the structure of a terminal device provided by an embodiment of the present application. Figure 9 As shown, the terminal device includes: at least one processor 501, a memory 502, an input device 503, an output device 504, and a computer program stored in the memory 502 and executable on at least one processor 501. When the processor 501 executes the computer program, the steps in any of the above method embodiments are implemented.

[0132] The input device 503 may be used to receive input digital or character information and generate key signal input related to user settings and function control of the terminal device. The output device 504 may include a display device such as a display screen.

[0133] The embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by the processor 501, the steps in the above-mentioned method embodiments can be implemented.

[0134] An embodiment of the present application provides a computer program product. When the computer program product is run on a terminal device, the terminal device can implement the steps in the above-mentioned method embodiments when executing the computer program product.

[0135] If the integrated unit is implemented as a software functional unit and sold or used as a standalone product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the process steps in the above-mentioned method embodiments by using a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When executed by processor 501, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. A computer-readable storage medium can include at least: any entity or device capable of carrying computer program code to a device / terminal equipment, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signals, telecommunication signals, and software distribution media. Examples include USB flash drives, removable hard drives, magnetic disks, or optical disks. In some jurisdictions, based on legislation and patent practice, computer-readable storage media cannot be electric carrier signals or telecommunication signals.

[0136] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0137] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0138] In the embodiments provided in this application, it should be understood that the disclosed devices / terminal devices and methods can be implemented in other ways. For example, the device / terminal device embodiments described above are merely illustrative. For example, the division of modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0139] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0140] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A cross-building energy consumption collaborative prediction method based on an improved KAN network, characterized in that: include: Obtaining current building information and a target energy consumption prediction model of the building to be predicted, wherein the target energy consumption prediction model is determined by training an improved Kolmogorov-Arnold KAN network model based on knowledge sharing of historical building information of multiple original buildings, and the activation function of the improved KAN network model is a Mexican hat wavelet function; Inputting the current building information into the target energy consumption prediction model to obtain the target energy consumption of the building to be predicted output by the target energy consumption prediction model; The target energy consumption prediction model of the building to be predicted is obtained, including: Acquire historical building information of a plurality of original buildings, wherein the plurality of original buildings at least includes the building to be predicted; Filtering source domain building information corresponding to the building to be predicted from the historical building information of the multiple original buildings; Training the improved KAN network model based on the source domain building information to obtain a source domain energy consumption prediction model; Performing knowledge migration processing on the source domain energy consumption prediction model based on the historical building information of the building to be predicted, to obtain a target energy consumption prediction model of the building to be predicted; The step of screening source domain building information corresponding to the building to be predicted from the historical building information of the plurality of original buildings includes: Performing cluster analysis on the historical building information of the plurality of original buildings to obtain similar buildings in the plurality of original buildings that have a similar category relationship with the building to be predicted; Determining the historical building information of the similar building as the source domain building information corresponding to the building to be predicted; The cluster analysis of the historical building information of the plurality of original buildings to obtain similar buildings in the plurality of original buildings that have a similar category relationship with the building to be predicted includes: For each of the historical building information of the original building, data compression processing is performed on the historical building information of each of the original building to obtain target building features of the historical building information of each of the original building; Calculating the Euclidean distance between each of the target building features to obtain a distance matrix corresponding to the multiple original buildings; The distance matrix is ​​subjected to category processing to obtain similar buildings in the plurality of original buildings that have a similar category relationship with the building to be predicted.

2. A cross-building energy consumption collaborative prediction method based on an improved KAN network, characterized in that: include: Obtaining current building information and a target energy consumption prediction model of the building to be predicted, wherein the target energy consumption prediction model is determined by training an improved Kolmogorov-Arnold KAN network model based on knowledge sharing of historical building information of multiple original buildings, and the activation function of the improved KAN network model is a Mexican hat wavelet function; Inputting the current building information into the target energy consumption prediction model to obtain the target energy consumption of the building to be predicted output by the target energy consumption prediction model; The target energy consumption prediction model of the building to be predicted is obtained, including: Training the improved KAN network model based on the historical building information of the building to be predicted to obtain a current energy consumption prediction model; In each learning round of federated learning, dynamically obtain current random data sent by the server, and input the current random data into the current energy consumption prediction model to obtain a current energy consumption prediction value; Dynamically sending the current energy consumption prediction value to the server, so that the server clusters the current energy consumption prediction values ​​corresponding to multiple original buildings to obtain similar buildings that have a similar category relationship with the building to be predicted; receiving target model parameters returned by the server based on current model parameters of a current energy consumption prediction model of a target building, wherein the target building includes the building to be predicted and the similar buildings; Dynamically updating the current energy consumption prediction model of the building to be predicted based on the target model parameters until the updated current energy consumption prediction model reaches convergence; The updated current energy consumption prediction model that has reached convergence is determined as the target energy consumption prediction model of the building to be predicted.

3. The cross-building energy consumption collaborative prediction method based on the improved KAN network according to claim 2 is characterized in that: The target model parameters are model parameters obtained by the server after performing mean processing on the current model parameters corresponding to the building to be predicted and the current model parameters corresponding to the similar buildings.

4. The cross-building energy consumption collaborative prediction method based on the improved KAN network according to any one of claims 1 to 3, characterized in that: The current building information at least includes current time data, current environment data and building area data.

5. A cross-building energy consumption collaborative prediction device based on an improved KAN network, characterized in that: include: an acquisition module for acquiring current building information and a target energy consumption prediction model of the building to be predicted, wherein the target energy consumption prediction model is determined by training an improved Kolmogorov-Arnold KAN network model based on knowledge sharing of historical building information of multiple original buildings, and the activation function of the improved KAN network model is a Mexican hat wavelet function; An input module, configured to input the current building information into the target energy consumption prediction model to obtain the target energy consumption of the building to be predicted output by the target energy consumption prediction model; Wherein, the acquisition module includes: an acquiring unit, configured to acquire historical building information of a plurality of original buildings, wherein the plurality of original buildings at least includes the building to be predicted; An information screening unit, configured to screen source domain building information corresponding to the building to be predicted from the historical building information of the plurality of original buildings; A model determination unit is configured to train the improved KAN network model based on the source domain building information to obtain a source domain energy consumption prediction model; perform knowledge transfer processing on the source domain energy consumption prediction model based on the historical building information of the building to be predicted to obtain a target energy consumption prediction model of the building to be predicted; The information screening unit includes: A cluster analysis subunit, configured to perform cluster analysis on the historical building information of the plurality of original buildings to obtain similar buildings in the plurality of original buildings that have a similar category relationship with the building to be predicted; a determination subunit, configured to determine the historical building information of the similar building as the source domain building information corresponding to the building to be predicted; The cluster analysis subunit is specifically used for: For each of the historical building information of the original building, data compression processing is performed on the historical building information of each of the original building to obtain target building features of the historical building information of each of the original building; Calculating the Euclidean distance between each of the target building features to obtain a distance matrix corresponding to the multiple original buildings; The distance matrix is ​​subjected to category processing to obtain similar buildings in the plurality of original buildings that have a similar category relationship with the building to be predicted.

6. A cross-building energy consumption collaborative prediction device based on an improved KAN network, characterized in that: include: an acquisition module for acquiring current building information and a target energy consumption prediction model of the building to be predicted, wherein the target energy consumption prediction model is determined by training an improved Kolmogorov-Arnold KAN network model based on knowledge sharing of historical building information of multiple original buildings, and the activation function of the improved KAN network model is a Mexican hat wavelet function; An input module, configured to input the current building information into the target energy consumption prediction model to obtain the target energy consumption of the building to be predicted output by the target energy consumption prediction model; The acquisition module is specifically used for: Training the improved KAN network model based on the historical building information of the building to be predicted to obtain a current energy consumption prediction model; In each learning round of federated learning, dynamically obtain current random data sent by the server, and input the current random data into the current energy consumption prediction model to obtain a current energy consumption prediction value; Dynamically sending the current energy consumption prediction value to the server, so that the server clusters the current energy consumption prediction values ​​corresponding to multiple original buildings to obtain similar buildings that have a similar category relationship with the building to be predicted; receiving target model parameters returned by the server based on current model parameters of a current energy consumption prediction model of a target building, wherein the target building includes the building to be predicted and the similar buildings; Dynamically updating the current energy consumption prediction model of the building to be predicted based on the target model parameters until the updated current energy consumption prediction model reaches convergence; The updated current energy consumption prediction model that has reached convergence is determined as the target energy consumption prediction model of the building to be predicted.

7. A terminal device comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the terminal device implements the method according to any one of claims 1 to 4.

Images