Explanatable building energy consumption prediction method and system based on large language model

Abstract

The invention discloses an interpretable building energy consumption prediction method and system based on a large language model, and belongs to the technical field of building electric power energy consumption prediction. Comprising the following steps: acquiring and processing prior information of building energy consumption prediction, and generating prefix vector representation; energy consumption time series data of the building are obtained, reversible instance normalization processing and patch embedding operation are sequentially carried out on the energy consumption time series data, reediting is carried out in combination with the text prototype vector representation, and embedded vector representation is generated; splicing the prefix vector representation and the embedded vector representation, and performing feature processing on a splicing result through a large language model to obtain a multi-dimensional vector representation; and performing mode conversion on the multi-dimensional vector representation and mapping the multi-dimensional vector representation to an output space to obtain an energy consumption prediction result. The interpretability and the prediction efficiency of building energy consumption prediction are improved, and the problems that existing energy consumption prediction is limited by complex influence factors, the energy consumption prediction time is long, and the result lacks interpretability are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of building power consumption prediction, and particularly to an interpretable building energy consumption prediction method and system based on a large language model. Background Art

[0002] The statements in this section merely mention the background art related to the present invention and do not necessarily constitute prior art.

[0003] Building energy consumption accounts for a considerable proportion of energy consumption, and energy conservation has gradually entered people's lives. With the continuous improvement of data technology and computing power, building energy consumption prediction has gradually emerged. Building energy consumption prediction is an important process for analyzing the energy usage patterns of buildings and predicting future energy consumption, involving a variety of equipment, with a wide range of applications and large scale; accurate energy consumption prediction helps to formulate energy conservation strategies, which can help building managers identify energy consumption patterns, thereby formulating effective energy conservation strategies and reducing energy consumption.

[0004] Currently, statistical models, machine learning models, deep learning models, etc. have all been applied in building energy consumption prediction, but the existing building energy consumption prediction methods still have the following problems:

[0005] (1) Due to the fact that building energy consumption prediction involves many uncertain factors such as seasonal changes and weather changes, the performance of the prediction model is not good. Based on a variety of complex elements, deep learning models can provide more accurate predictions, but the training may take a long time, and the data training operation places high requirements on the hardware.

[0006] (2) For the building energy consumption time series prediction based on deep learning models, due to the continuous change of its internal parameters and structure, it is very difficult to explain its internal decision-making process, making the prediction results lack interpretability, which may lead to users or stakeholders lacking trust in the results of the model. Summary of the Invention

[0007] In order to solve the deficiencies of the prior art, the present invention provides an interpretable building energy consumption prediction method, system, electronic device, computer-readable storage medium and computer program product based on a large language model. By combining multi-modal information such as prior information, actual energy consumption data and text prototypes, building energy consumption prediction is carried out through a large language model, improving the prediction efficiency and ensuring the interpretability of the prediction results.

[0008] In the first aspect, the present invention provides an interpretable building energy consumption prediction method based on a large language model;

[0009] An interpretable building energy consumption prediction method based on a large language model includes:

[0010] Obtain the prior information for building energy consumption prediction and process it to generate a prefix vector representation;

[0011] Obtain the energy consumption time series data of a building, perform reversible instance normalization processing and patch embedding operations on the energy consumption time series data in sequence, and perform re-editing in combination with the text prototype vector representation to generate an embedded vector representation;

[0012] Concatenate the prefix vector representation and the embedded vector representation, and perform feature processing on the concatenated result through a large language model to obtain a multi-dimensional vector representation;

[0013] Perform modal conversion on the multi-dimensional vector representation and map it to the output space to obtain the energy consumption prediction result.

[0014] In some embodiments, processing the prior information to generate the prefix vector representation includes:

[0015] Perform word segmentation processing on the prior information to obtain multiple unit blocks;

[0016] Convert all unit blocks into an embedded vector matrix through a vector embedder, and perform required information retrieval based on the embedded vector matrix to generate a prefix vector representation.

[0017] In some embodiments, the sequentially performing reversible instance normalization processing and patch embedding operations on the energy consumption time series data includes:

[0018] Perform individual normalization on the energy consumption time series data and divide it into multiple consecutive patches to generate time series data patches;

[0019] Process the time series data patches through a patch embedder to generate a patch feature vector representation.

[0020] In some embodiments, the specifically performing re-editing in combination with the text prototype vector representation to generate the embedded vector representation is:

[0021] Process the patch feature vector representation and the text prototype vector representation through a multi-head attention mechanism, and through scaled dot product attention, utilize the matching between the patch feature vector representation and the text prototype vector representation to generate an embedded vector representation.

[0022] In some embodiments, the performing feature processing on the concatenated result through a large language model to obtain a multi-dimensional vector representation includes:

[0023] Process the concatenated result sequentially through a multi-head attention mechanism and a feed-forward network, and combine residual connections to obtain an output embedded representation;

[0024] Remove the prefix embedding in the output embedded representation to obtain a multi-dimensional vector representation.

[0025] In some embodiments, the step of performing modality conversion on the multi-dimensional vector representation and mapping it to the output space to obtain the energy consumption prediction result is specifically as follows: converting the multi-dimensional vector representation into a one-dimensional vector representation and flattening it, and performing a linear transformation on the flattened result through a fully connected linear layer to obtain the energy consumption prediction result.

[0026] In a second aspect, the present invention provides an interpretable building energy consumption prediction system based on a large language model;

[0027] An interpretable building energy consumption prediction system based on a large language model includes:

[0028] A prefix generation module, configured to: obtain and process the prior information of building energy consumption prediction to generate a prefix vector representation;

[0029] A re-encoding module, configured to: obtain the energy consumption time series data of a building, perform reversible instance normalization processing and patch embedding operations on the energy consumption time series data in sequence, and perform re-editing in combination with a pretext prototype vector representation to generate an embedded vector representation;

[0030] A feature processing module, configured to: splice the prefix vector representation and the embedded vector representation, and perform feature processing on the spliced result through a large language model to obtain a multi-dimensional vector representation;

[0031] A modality conversion module, configured to: perform modality conversion on the multi-dimensional vector representation and map it to the output space to obtain the energy consumption prediction result.

[0032] In a third aspect, the present invention provides an electronic device;

[0033] An electronic device includes a memory, a processor, and a computer program stored on the memory. The processor executes the computer program to implement the steps of the above-mentioned interpretable building energy consumption prediction method based on a large language model.

[0034] In a fourth aspect, the present invention provides a computer-readable storage medium;

[0035] A computer-readable storage medium, on which a computer program / instructions are stored. When the computer program / instructions are executed by a processor, the steps of the above-mentioned interpretable building energy consumption prediction method based on a large language model are implemented.

[0036] In a fifth aspect, the present invention provides a computer program product;

[0037] A computer program product includes computer program / instructions. When the computer program / instructions are executed by a processor, the steps of the above-mentioned interpretable building energy consumption prediction method based on a large language model are implemented.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0039] 1. For the technical solution provided by the present invention, prior information is used as the prefix of the building energy consumption time series data. The feature vectors can be directly processed under the knowledge of the building energy consumption field without fine-tuning the large language model, without changing its internal parameters and structure, making the internal decision-making process transparent and the prediction results interpretable.

[0040] 2. For the technical solution provided by the present invention, the large language model is used to process the building energy consumption time series data. While retaining the high prediction performance of deep learning, it also solves the limitation of a large amount of complex data on the prediction time, further shortening the prediction time.

[0041] 3. For the technical solution provided by the present invention, the time series patches are aligned and fused with the text prototype to form new feature vectors. On the one hand, the dimension of the feature vectors is reduced, reducing the computational redundancy. On the other hand, the operation information of the building energy consumption and the text prototype information are fully fused, enabling the model to have a more comprehensive understanding during energy consumption prediction and improving the prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The accompanying drawings forming a part of this invention are used to provide a further understanding of the present invention. The schematic embodiments and descriptions thereof of the present invention are used to explain the present invention and do not constitute an improper limitation of the present invention.

[0043] Figure 1 It is a schematic flow chart of an interpretable building energy consumption prediction method based on a large language model provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0045] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless otherwise clearly specified in the context, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those clearly listed steps or units, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0046] Without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0047] Embodiment 1

[0048] The influencing factors of building energy consumption prediction are complex, resulting in the need to improve the prediction timeliness of existing prediction methods based on deep learning models, and the lack of interpretability of their prediction results. Therefore, the present invention provides an interpretable building energy consumption prediction method based on a large language model, which combines prior information embedding and the processing of time series data by the large language model to achieve interpretable prediction of building energy consumption.

[0049] Next, in combination with Figure 1 , a detailed description will be given of an interpretable building energy consumption prediction method based on a large language model disclosed in this embodiment. The interpretable building energy consumption prediction method based on a large language model includes the following steps:

[0050] S1. Obtain the prior information of building energy consumption prediction and process it to generate a prefix vector representation.

[0051] Among them, the prior information is provided manually and includes information such as domain knowledge in building energy consumption prediction, input data format, and basic data features. Through the processing of the above information, a structured framework can be provided for subsequent building energy consumption prediction, that is, without modifying the internal parameters of the large language model, the application field of the data can be told to the large language model.

[0052] Specifically, the prior information is input into the prefix generation module of the large language model in the form of natural language for processing to obtain a prefix vector representation; specifically including:

[0053] S101. Preprocess the input prior information text T to obtain a tokenized text T'; use regular expressions to perform tokenization processing on the tokenized text T' to obtain multiple smaller unit blocks Tokens; among them, the tokenized text is expressed as:

[0054] T' = Preprocess(T);

[0055] In the formula, Preprocess represents the preprocessing operation. In this embodiment, the preprocessing operation refers to removing special characters in the prior information text and regularizing the input text format.

[0056] The unit block is expressed as:

[0057] Tokens = re.findall(pattern, T');

[0058] In the formula, re.findall represents the re.findall function, and pattern represents the regular expression pattern.

[0059] S102: Process all unit blocks through a vector embedder to convert all unit blocks into embedding vector matrices. The specific process is as follows:

[0060] (1) After processing in S101, a unit block set Tokens={t 1 ,t 2 ,...,t n}, t i Represents the i-th unit block.

[0061] (2) Set t i Mapped to a corresponding embedding vector e i To express the meaning of the unit block in a specific context, after all unit blocks are mapped, an embedded vector matrix is ​​formed.

[0062] Specifically, when searching for a unit block, the index function index(t i ) Find the embedding vector e corresponding to the required unit block i , where index(t i ) represents the unit block t i Index position in the vocabulary; for each input unit block t i , extract the corresponding vector representation e from the embedding matrix according to its index in the vocabulary i , the embedding vector matrix composed of all the required embedding vectors retrieved is used as the output vector representation to form a prefix vector representation. The embedding vector matrix is ​​expressed as:

[0063]

[0064] Where |V| represents the size of the vocabulary, e i Represents the embedding vector. Here, the index function is the index formula of the vector embedder in the large language model.

[0065] S2. Obtain the energy consumption time series data of the building, perform reversible instance normalization and patch embedding operations on the energy consumption time series data in sequence, re-edit the data in combination with the text prototype vector representation, and generate an embedded vector representation.

[0066] In this embodiment, S1 and S2 can be executed in parallel.

[0067] In this step, reversible instance normalization is performed on the energy consumption time series data to alleviate the offset of the time series distribution, and the time series data patch is aligned with the text prototype generated by the large language model pre-training to form a new feature vector, which is conducive to capturing key information and reducing computational complexity. As an implementation method, S2 specifically includes:

[0068] S201. Individually normalize the energy consumption time series data and divide it into multiple consecutive patches to generate time series data patches.

[0069] Specifically, first, the energy consumption time series data X (i) is individually normalized to have a mean of zero and a standard deviation of unit standard deviation; then, the normalized energy consumption time series data is subjected to sliding decomposition and divided into several consecutive non - overlapping patches to generate time series data patches.

[0070] Here, the total number of input patches is expressed as:

[0071]

[0072] In the formula, T represents the total length of the time series data, L p represents the patch length, which means subtracting the length of one patch from the total data to obtain the length available for dividing patches; S represents the horizontal sliding step size, and adding 2 ensures a reasonable number of patches.

[0073] S202. Input the time series data patches into a patch embedder to extract the features of each part and generate a patch feature vector representation.

[0074] Specifically, first, perform feature extraction on the time series data patches, and combine the extracted features into a feature vector, expressed as:

[0075] FeatureVector i = [mean(P i ), std(P i ), max(P i ), min(P i )];

[0076] Traverse all the patches in the energy consumption time series data, repeat the above - mentioned feature extraction steps, and generate a list of patch feature vectors, that is, the patch feature vector representation, expressed as:

[0077] FeatureVectors = [FeatureVector 0 , FeatureVector 1 ,..., FeatureVector T-S ;

[0078] S203. Process the word embeddings through a linear layer and output a text prototype vector representation.

[0079] Here, the word embeddings are the word vector representations related to the construction field obtained by pre-training the large language model with information such as web knowledge, that is, the responses of the large language model itself configured for construction field problems. These word vector representations can capture the semantic relationships and context information between words, making words with similar meanings closer in the vector space.

[0080] The text prototype vector representation contains information such as basic building information, usage, timestamps (including dates and times) for recording building energy consumption, and historical building energy consumption data. Fusing it with the patch feature vector representation can enhance the understanding and reasoning ability of energy consumption time series data, making the final building energy consumption prediction result closer to the real building energy consumption scenario and having higher prediction accuracy.

[0081] S204. Process the patch feature vector representation and the text prototype vector representation through the multi-head attention mechanism. By scaling the dot product attention and using the matching between the patch feature vector representation and the text prototype vector representation, generate the embedded vector representation. The specific process is as follows:

[0082] (1) Define the query matrix, key matrix, and value matrix for each attention head k = {1,..., K} in the multi-head attention mechanism. Among them, the query matrix is expressed as:

[0083]

[0084] The key matrix is expressed as:

[0085]

[0086] The value matrix is expressed as:

[0087]

[0088] In the formula, represents the weight, represents the patch feature vector representation, and E′ represents the text prototype vector representation.

[0089] (2) Each attention head determines the relevance of different parts in the input sequence through scaled dot product attention by matching the query with the key to obtain the attention weights, and weights the value vector to generate the attention output.

[0090] Here, the calculation process of the scaled dot product attention is expressed as:

[0091]

[0092] In the formula, d k represents the dimension of the key vector.

[0093] (3) Concatenate and linearly transform the results obtained by multiple independent attention heads to obtain an embedded vector representation.

[0094] S3. Concatenate the prefix vector representation and the embedded vector representation, and perform feature processing on the concatenated result through a large language model to obtain a multi-dimensional vector representation.

[0095] In this step, the embedding with the prefix vector is input into the large language model. Under the action of the prefix vector, the large language model can directly perform feature vector processing under the knowledge of the building energy consumption field limited by the prefix without fine-tuning parameters or changing the model structure, so that the result will not be affected by unknown factors due to changes in the internal structure, and the final building energy consumption prediction result is interpretable.

[0096] As an implementation, S3 includes:

[0097] S301. Concatenate the prefix vector representation and the embedded vector representation to obtain an initial vector representation; process the initial vector representation with the prefix vector through a multi-head attention mechanism to obtain a first intermediate vector representation.

[0098] Through multiple attention heads, the model can learn information from different subspaces. Each head focuses on different parts of the input, synthesizes information from multiple angles, improves the model's ability to understand context, and processes multiple feature dimensions to achieve parallel computing.

[0099] S302. Perform an addition operation on the first intermediate vector representation and the prefix vector representation, and normalize the added result to obtain a second intermediate vector representation.

[0100] Based on this, residual connection is achieved through the addition operation to avoid gradient disappearance in deep networks; the stability of the model is improved and the training process is accelerated through the normalization operation.

[0101] S303. Input the second intermediate vector representation into a feed-forward network for processing to obtain a third intermediate vector representation.

[0102] Specifically, the input layer corresponds each feature in the second intermediate vector representation to a node and transmits it to a feed-forward network with multiple hidden layers. The neurons in the hidden layer use the weighted sum activation function ReLU to process the input and pass it layer by layer to obtain the final output, that is, the third intermediate vector representation.

[0103] Here, the calculation process of each neuron output is expressed as

[0104]

[0105] Among them, represents the weight, Bias is denoted as, φ represents the weighted sum activation function, l represents a specific hidden layer, i represents the neuron number of the (l - 1)th layer, and j represents the neuron number of the lth layer.

[0106] S304. Add the second intermediate vector representation and the third intermediate vector representation to achieve residual connection and normalization, obtain the output embedding representation, remove the prefix embedding therein, and obtain the multi-dimensional vector representation.

[0107] S4. Perform modal transformation on the multi-dimensional vector representation and map it to the output space to obtain the energy consumption prediction result.

[0108] Specifically, convert the multi-dimensional vector representation into a one-dimensional vector representation and flatten it, perform a linear transformation on the flattened result through a trained fully connected linear layer, and obtain the energy consumption prediction result through the learned weights and bias.

[0109] Embodiment 2

[0110] This embodiment discloses an interpretable building energy consumption prediction system based on a large language model, including:

[0111] A prefix generation module, configured to: obtain and process the prior information of building energy consumption prediction, and generate a prefix vector representation;

[0112] A re-encoding module, configured to: obtain the energy consumption time series data of the building, perform reversible instance normalization processing and patch embedding operations on the energy consumption time series data in sequence, and perform re-editing in combination with the text prototype vector representation to generate an embedding vector representation;

[0113] A feature processing module, configured to: splice the prefix vector representation and the embedding vector representation, and perform feature processing on the splicing result through a large language model to obtain a multi-dimensional vector representation;

[0114] A modal transformation module, configured to: perform modal transformation on the multi-dimensional vector representation and map it to the output space to obtain the energy consumption prediction result.

[0115] It should be noted here that the above prefix generation module, re-encoding module, feature processing module, and modal transformation module correspond to the steps in Embodiment 1. The examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1 above. It should be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer executable instructions.

[0116] Embodiment 3

[0117] Embodiment III of the present invention provides an electronic device, including a memory, a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the steps of the above-mentioned interpretable building energy consumption prediction method based on the large language model are completed.

[0118] Embodiment IV

[0119] Embodiment IV of the present invention provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps of the above-mentioned interpretable building energy consumption prediction method based on the large language model are completed.

[0120] Embodiment V

[0121] Embodiment V of the present invention provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the steps of the above-mentioned interpretable building energy consumption prediction method based on the large language model are implemented.

[0122] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0123] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured product including an instruction device, and the instruction device implements the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0124] These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operation steps are executed on the computer or other programmable device to generate computer-implemented processing. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0125] In the above embodiments, the descriptions of the respective embodiments have their own focuses. For the parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0126] The foregoing is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An interpretable building energy consumption prediction method based on a large language model, characterized in that: include: Obtain and process the prior information of building energy consumption prediction to generate prefix vector representation; Obtaining energy consumption time series data of the building, performing reversible instance normalization processing and patch embedding operations on the energy consumption time series data in sequence, re-editing the data in combination with text prototype vector representation, and generating an embedded vector representation; concatenating the prefix vector representation with the embedded vector representation, and performing feature processing on the concatenated result through a large language model to obtain a multi-dimensional vector representation; The multidimensional vector representation is modally converted and mapped to an output space to obtain an energy consumption prediction result.

2. The interpretable building energy consumption prediction method based on a large language model as claimed in claim 1, characterized in that: Processing the prior information to generate a prefix vector representation includes: Performing word segmentation processing on the prior information to obtain multiple unit blocks; All unit blocks are converted into embedding vector matrices through a vector embedder, and the required information is retrieved based on the embedding vector matrix to generate a prefix vector representation.

3. The interpretable building energy consumption prediction method based on a large language model as claimed in claim 1, characterized in that: The step of sequentially performing reversible instance normalization processing and patch embedding operations on the energy consumption time series data includes: Normalizing the energy consumption time series data individually and dividing it into a plurality of continuous patches to generate time series data patches; The time series data patches are processed by a patch embedder to generate patch feature vector representations.

4. The interpretable building energy consumption prediction method based on a large language model as claimed in claim 3 is characterized in that: The re-editing in combination with the text prototype vector representation to generate the embedded vector representation is specifically as follows: The patch feature vector representation and the text prototype vector representation are processed by a multi-head attention mechanism, and an embedded vector representation is generated by scaling click attention and utilizing the match between the patch feature vector representation and the text prototype vector representation.

5. The interpretable building energy consumption prediction method based on a large language model as claimed in claim 1, characterized in that: The process of performing feature processing on the concatenation result by using the large language model to obtain a multi-dimensional vector representation includes: The splicing results are processed sequentially through the multi-head attention mechanism and feed-forward network, combined with residual connection to obtain the output embedding representation; The prefix embedding in the output embedding representation is removed to obtain a multi-dimensional vector representation.

6. The interpretable building energy consumption prediction method based on a large language model as claimed in claim 1, characterized in that: The method of performing modal conversion on the multidimensional vector representation and mapping it to the output space to obtain the energy consumption prediction result is specifically as follows: converting the multidimensional vector representation into a one-dimensional vector representation and flattening it, performing a linear transformation on the flattened result through a fully connected linear layer, and obtaining the energy consumption prediction result.

7. An interpretable building energy consumption prediction system based on a large language model, characterized by: include: The prefix generation module is configured to: obtain and process the prior information of the building energy consumption prediction to generate a prefix vector representation; The re-encoding module is configured to: obtain energy consumption time series data of the building, perform reversible instance normalization processing and patch embedding operations on the energy consumption time series data in sequence, and re-edit the data in combination with the text prototype vector representation to generate an embedding vector representation; A feature processing module is configured to: concatenate the prefix vector representation with the embedded vector representation, perform feature processing on the concatenation result through a large language model, and obtain a multi-dimensional vector representation; The modal conversion module is configured to: perform modal conversion on the multidimensional vector representation and map it to an output space to obtain an energy consumption prediction result.

8. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the interpretable building energy consumption prediction method based on a large language model as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the steps of the interpretable building energy consumption prediction method based on a large language model described in any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the interpretable building energy consumption prediction method based on a large language model described in any one of claims 1 to 6 are implemented.

Images