Building hourly indoor temperature prediction method and system based on dual-drive mechanism

Abstract

The invention provides a building hourly indoor temperature prediction method and system based on a dual-drive mechanism, and belongs to the technical field of building energy consumption management and control. Comprising the steps that a heat balance differential equation with the building indoor temperature as an interface variable is selected as a reference, and meanwhile the heat transfer process of the enclosure structure is expressed in a space state so as to construct a guide module for a physical model; constructing a data driving module, and constructing and training a back propagation neural network by taking an indoor temperature change difference value caused by a non-principal component item as a target variable and taking an outdoor meteorological parameter as input; and combining the outputs of the physical model guiding module and the data driving module, and performing indoor temperature prediction based on a back propagation neural network. The physical authenticity of hourly indoor temperature prediction and the accuracy and rapidity of building hourly indoor temperature prediction can be considered at the same time.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of building energy consumption management and control, and particularly relates to a building hourly indoor temperature prediction method and system based on a double driving mechanism. BACKGROUND

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

[0003] The building indoor temperature is a core target variable for realizing building indoor thermal environmental control and is one of the key characteristic parameters for building energy consumption and load prediction. Therefore, predicting the building indoor temperature, especially hourly prediction, can not only enhance the stability of building indoor thermal environmental control and improve the building indoor environmental quality, but also guide the effective management of building energy consumption and load.

[0004] At present, the mainstream building indoor temperature prediction methods mainly include two types of physical models and data-driven models. The physical model has the advantages of higher result interpretability, stronger reliability and generalization ability, but the construction process is very complex and time-consuming; while the data-driven model has the advantages of fast, convenient and accurate modeling and prediction, but has the defect of non-interpretable results.

[0005] Therefore, related research has begun to focus on the modeling method of the physical and data double driving mechanism in the field of buildings, such as constructing a physical model for describing the heat change inside the building through thermodynamic principles, and constructing a data-driven model through a machine learning algorithm, and then combining the physical model with the data-driven model. However, this method has the following defects: (1) In the prior art, most fusion methods use highly simplified physical models, which are difficult to accurately depict the coupling effects of building envelope heat transfer, indoor and outdoor heat exchange, solar radiation, air permeation and other multi-physical processes, resulting in limited prediction accuracy under complex weather conditions or building structures.

[0006] (2) In the coupling of the two models, the existing double driving methods mostly use simple series or parallel structures, which cannot construct an explicit guide from the physical mechanism to the data compensation, resulting in shallow fusion of physical information and data information, weak interpretability, and unsatisfactory synergistic enhancement effect. SUMMARY

[0007] To overcome the above shortcomings of the prior art, the present application provides a building hourly indoor temperature prediction method and system based on a double driving mechanism, which can simultaneously consider the physical reality of the building hourly indoor temperature prediction and the accuracy and rapidity of the building hourly indoor temperature prediction.

[0008] To achieve the above object, one or more embodiments of the present application provide the following technical solutions: The first aspect of the present application provides a building hourly indoor temperature prediction method based on a double driving mechanism.

[0009] A building hourly indoor temperature prediction method based on a double driving mechanism, comprising: constructing a physical model of a building to be measured, determining an hourly building indoor temperature time series through building energy consumption simulation software according to basic information of the measured building and outdoor meteorological parameters; selecting a heat balance differential equation with building indoor temperature as an interface variable as a reference, and representing a heat transfer process of an envelope structure in a space state to construct a physical model guiding module, and obtaining a building indoor temperature time series after physical guidance; constructing a data-driven module, taking a difference value of indoor temperature change caused by a non-principal component item that is not explained by the physical model guiding module as a target variable, and taking outdoor meteorological parameters as input, constructing and training a back propagation neural network to obtain a residual prediction sequence; combining the building indoor temperature time series after physical guidance with the residual prediction sequence, obtaining a final building indoor temperature time series based on the back propagation neural network prediction, and performing error evaluation on the prediction result.

[0010] Further, the basic information of the measured building includes the geographical position, appearance size, orientation, internal structure, envelope structure type, thermal parameter and room function of the building.

[0011] Further, the heat balance differential equation with building indoor temperature as an interface variable is represented as: ; wherein, Cp represents an air sensible heat capacity, , , and represent surface temperatures of indoor air, outdoor air, wall inner surface and air conditioning coil respectively; represents a sum of convective heat transfer of all internal heat sources, represents a sum of convective heat transfer of all wall inner surfaces, represents heat brought in by air permeation.

[0012] Further, the heat transfer process of the envelope structure is represented in a space state, that is, a space state equation with two internal nodes is used to represent the heat transfer process of the envelope structure.

[0013] Further, the construction of the physical model guiding module comprises: discretizing the reference formula corresponding to the physical model and the spatial state equation used for representing the heat transfer process of the envelope in time in a differential manner, and screening a plurality of physical variables, so that the remaining physical variables after screening are used as principal components to construct the physical model guiding module.

[0014] Further, the outdoor meteorological parameters comprise: outdoor air temperature, direct solar radiation intensity, scattered radiation intensity, sky temperature and wind speed.

[0015] Further, the prediction result is error-evaluated by simultaneously using MSE error, normalized average deviation error and root mean square error variation coefficient.

[0016] The second aspect of the present application provides a building hourly indoor temperature prediction system based on a double driving mechanism.

[0017] A building hourly indoor temperature prediction system based on a double driving mechanism comprises: A physical model construction unit is configured to: construct a physical model of a building to be measured, determine an hourly building indoor temperature time series by using building energy consumption simulation software according to basic information of the measured building and outdoor meteorological parameters; A physical guiding calculation unit is configured to: select a heat balance differential equation taking the building indoor temperature as an interface variable as a reference, and represent the heat transfer process of the envelope in a spatial state to construct a physical model guiding module, and obtain a building indoor temperature time series after physical guidance; A data-driven calculation unit is configured to: construct a data-driven module, taking the difference value of the indoor temperature change caused by the non-principal component term not explained by the physical model guiding module as a target variable, and taking outdoor meteorological parameters as input, constructing and training a back propagation neural network to obtain a residual prediction sequence; A temperature prediction evaluation unit is configured to: combine the building indoor temperature time series after physical guidance with the residual prediction sequence, obtain a final building indoor temperature time series based on the back propagation neural network prediction, and error-evaluate the prediction result. The third aspect of the present application provides a computer readable storage medium having a program stored thereon, the program being executed by a processor to implement the steps of the building hourly indoor temperature prediction method based on a double driving mechanism according to the first aspect of the present application.

[0018] The fourth aspect of the present application provides an electronic device comprising a memory, a processor and a program stored on the memory and executable on the processor, wherein the processor executes the program to implement the steps of the building hourly indoor temperature prediction method based on a double driving mechanism according to the first aspect of the present application.

[0019] The above one or more technical solutions have the following beneficial effects: (1) According to the basic information of the measured building and the outdoor meteorological parameters, the physical model of the building to be measured is constructed through the building energy consumption simulation software EnergyPlus, and the hourly building indoor temperature time sequence is determined. Compared with the prior art, multiple factors such as building geographical location, building envelope thermal characteristics, indoor load and complex time-varying meteorological parameters can be integrated, and the coupling physical processes of building envelope heat transfer, indoor and outdoor air heat exchange, solar radiation heat gain, air permeation and internal heat source influence can be simulated completely and accurately. Therefore, the present application fundamentally overcomes the defect that the simplified physical model is difficult to describe the complex thermal process, significantly improves the model-based precision and adaptability under variable weather conditions or special-shaped building structures, and can provide more reliable and more physically realistic benchmark and data basis for subsequent fusion prediction.

[0020] (2) From the output of the high-fidelity physical model, the principal component physical variables and their relationships that dominate the indoor temperature change are explicitly identified and extracted through partial derivative analysis and other methods to form an interpretable physical guided calculation core; and the remaining part that cannot be completely explained by the physical model is defined as the target residual, which is learned and predicted by the data-driven model. This architecture makes the learning goal of the data-driven module very clear, and realizes the deep fusion and clear division of physical information and data information at different levels. This not only enhances the physical interpretability of the final prediction result, but also accurately compensates for the complex residual through data-driven.

[0021] The advantages of the additional aspects of the application will be partially given in the following description, partially will become obvious from the following description, or will be understood through the practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0022] The drawings accompanying the specification of this application form a part thereof, serve to provide further understanding of the application, and together with the description of the exemplary embodiments of the application and the explanation thereof serve to explain the application, and do not constitute an improper limitation of the application.

[0023] Figure 1 A flowchart of a building hourly indoor temperature prediction method based on a double driving mechanism in embodiment one of the present application.

[0024] Figure 2 A space state diagram with two internal nodes in embodiment one of the present application.

[0025] Figure 3 A schematic diagram of the actual value time sequence of the building indoor temperature simulated by the EnergyPlus software in embodiment one of the present application.

[0026] Figure 4A schematic diagram of the time series of the building indoor temperature prediction value calculated by the physical model guiding module in the first embodiment of the present application.

[0027] Figure 5 A comparison chart of the building indoor temperature prediction results in the first embodiment of the present application. DETAILED DESCRIPTION

[0028] It should be noted that the following detailed description is exemplary in nature and is intended to provide further description of the present application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0029] It should be noted that the terms used herein are only intended to describe specific embodiments and are not intended to limit the exemplary embodiments according to the present application.

[0030] In the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

[0031] Embodiment One The present embodiment discloses a building hourly indoor temperature prediction method based on a double driving mechanism.

[0032] As shown in Figure 1 A building hourly indoor temperature prediction method based on a double driving mechanism comprises the following steps: Step S1, constructing a physical model of the building to be measured, determining the hourly building indoor temperature time series through building energy consumption simulation software according to the basic information and outdoor meteorological parameters of the measured building; Step S2, selecting a heat balance differential equation with building indoor temperature as an interface variable as a reference, and representing the heat transfer process of the envelope structure in the spatial state to construct a physical model guiding module, and obtaining the building indoor temperature time series after physical guidance; Step S3, constructing a data driven module, taking the indoor temperature change difference caused by the unexplained non-principal component term of the physical model guiding module as the target variable, and taking the outdoor meteorological parameters as the input, constructing and training a back propagation neural network to obtain a residual prediction sequence; Step S4, combining the building indoor temperature time series after physical guidance with the residual prediction sequence, obtaining the final building indoor temperature time series based on the back propagation neural network prediction, and performing error evaluation on the prediction results.

[0033] The first object of the present application is to give full play to the role of the physical model, form simultaneous efficient use of its internal physical principles and output data, to ensure the physical authenticity and explainability of the building hourly indoor temperature prediction; the second object is to give full play to the ability of data-driven model to handle nonlinear problems, to fit the nonlinear part of the physical model output data which is difficult to explain with simple and explicit physical principles, to ensure the accuracy and rapidity of the building hourly indoor temperature prediction. Based on the above process, the present application can simultaneously consider the physical authenticity of the hourly indoor temperature prediction and the accuracy and rapidity of the building hourly indoor temperature prediction. For the convenience of understanding the technical scheme of the present application, the specific implementation method in the technical scheme of the present application is further explained and described below.

[0034] In step S1, the physical model of the building to be tested is constructed, and the hourly building indoor temperature time series is determined by building energy consumption simulation software according to the basic information and outdoor meteorological parameters of the tested building.

[0035] The physical model of the building is established using EnergyPlus simulation software. The inputs of modeling include basic information of the building such as geographical location, appearance size, orientation, internal structure, envelope type, thermal parameters, room function, and hourly time series data of outdoor meteorological parameters; the output is set as the hourly building indoor temperature time series ; wherein, represents the indoor temperature at the th hour of the output, represents the total number of the output indoor temperature.

[0036] In this embodiment, an EnergyPlus software is used to establish a building model, a typical meteorological year is selected for the meteorological parameter file, the simulation time is set from July 1st to 31st, the simulation time step is 1 hour, and other specific settings are shown in Table 1. It should be noted that the parameters in the envelope setting and indoor setting in Table 1 are set from the built-in parameter setting options of the EnergyPlus simulation software, therefore, the parameters in Table 1 are not explained in detail in this embodiment.

[0037] Table 1: Modeling settings of EnergyPlus simulation software

[0038] According to the above modeling settings, the simulated building indoor temperature true value time series is shown in Figure 3 , each value in the indoor temperature time series shown in Figure 3 is calculated according to specific physical principles and formulas, and has complete physical explainability, therefore, as the indoor temperature true value.

[0039] In step S2, the thermal balance differential equation with the building's indoor temperature as the interface variable is selected as the benchmark, and the heat transfer process of the building envelope is represented in the spatial state to construct a physical model guidance module and obtain the time series of the building's indoor temperature after physical guidance.

[0040] First, based on the principles of EnergyPlus simulation software, the thermal equilibrium differential equation with indoor building temperature as the interface variable is selected as the benchmark formula for the physical model guiding module: ; in, This indicates the sensible heat capacity of air. , , and These represent the indoor and outdoor air temperatures, the inner surface of the wall, and the surface temperature of the air conditioning coil, respectively. This represents the sum of convective heat transfer from all internal heat sources. This represents the sum of all convective heat transfers within the walls. This indicates the heat brought in by air infiltration.

[0041] Subsequently, for the most critical heat transfer process of the building envelope, the following methods were adopted: Figure 2 The state equation for a space with two internal nodes is shown below: ; ; in, Indicates the temperature outside the wall. and These represent the heat flux flowing through the outside and inside of the wall, respectively. and These represent the equivalent thermal resistance and equivalent heat capacity of the building envelope, respectively. and These represent the convective heat transfer coefficient and heat transfer area of ​​the building envelope, respectively.

[0042] Then, using the finite difference method, the baseline formula and the state-space equation are written in time-discrete form and further derived to obtain the following: Time and The value of the physical variable at time represents Indoor temperature at any time The expression, and the calculation of the physical variables in it with respect to... The partial derivatives are used to select physical variables whose partial derivatives are not zero. Terms containing only these physical variables are then used as principal components to form the calculation formula for the physical module. ; in, This indicates the sensible heat capacity of air.

[0043] Finally, the output of the EnergyPlus software... Each Moment and Moment Substituting into the above formula, we can calculate... Indoor temperature at any time Among them, time 0 Other parameters in this calculation formula can be obtained through the settings and output of the EnergyPlus software. Completing the calculations for all time points yields... time series This is the output of the physical module.

[0044] The time series of predicted building indoor temperatures obtained through calculations guided by the physical model are as follows: Figure 4 As shown, the calculation results of the physics model-guided module have already shown a trend consistent with... Figure 3 The results are largely consistent with the actual values, indicating that the calculation results of the physical model-guided module have included the main physical principles in the process of indoor temperature change, and this part has physical interpretability. However, it can also be seen that there are large errors between the calculation results of the physical model-guided module and the actual values. These errors are mainly caused by nonlinear non-principal component terms. Therefore, it is necessary to build a data-driven module to fit the non-principal components to reduce the final prediction error.

[0045] In step S3, a data-driven module is constructed, using the difference in indoor temperature change caused by non-principal component terms not explained by the physical model guidance module as the target variable and outdoor meteorological parameters as input, to construct and train a backpropagation neural network to obtain the residual prediction sequence.

[0046] First, the non-principal component terms other than the principal components will cause... Time and Difference in indoor temperature over time As the target variable of the data-driven module Output using EnergyPlus software Indoor temperature at all times With the output of the physics model module Subtracting the two gives us: ; in, express Chinese correspondence The elements of time, namely Difference in indoor temperature at any given time .

[0047] Subsequently, the input variables for the data-driven module are determined, namely: Theoretically, it is a function of non-principal component terms, and the physical variables involved in the non-principal component terms are used as input variables for the data-driven module. Specifically includes outdoor air temperature at any time Intensity of direct solar radiation Scattered radiation intensity Sky temperature and wind speed The time series, and the variables mentioned above A time series composed of the values ​​at each moment, i.e. .

[0048] Then, Dividing the dataset according to a 70%:30% ratio, the first 70% is used as the training set, denoted as... ,in Round to the nearest integer; the last 30% is used as the test set, denoted as... . Similarly, it is divided into training sets. and test set ;in, Indicates training set Chinese correspondence Elements of time Indicates test set Chinese correspondence The element of time.

[0049] Based on this, a backpropagation neural network (BPNN) is built using the "newff" function built into Matlab software. The input of BPNN is a vector. The output is a vector. The predicted value is denoted as .use and The BPNN is trained using the following loss function: and Mean square error (MSE): ; in, Indicates the training set Chinese correspondence Elements of time The predicted value.

[0050] In step S4, the physical-guided building indoor temperature time series is combined with the residual prediction sequence, and the final building indoor temperature time series is predicted based on the backpropagation neural network. The prediction results are then evaluated for error.

[0051] A well-trained BPNN can be used based on the input test set variable values. To obtain the corresponding predicted value This is the output of the data-driven module. Moment In fact The predicted value is compared with the output of the physics model guidance module. By adding them together, we can obtain the result. Predicted indoor temperature at any time ,Right now Therefore, within the test set coverage From the moment to the end At any given moment, through hourly calculations, a time series of the final predicted indoor building temperature values ​​can be generated. .

[0052] To ensure the accuracy of the prediction, three error evaluation indicators were used simultaneously to analyze the time series of the predicted indoor temperature values. and the corresponding EnergyPlus output of the building's indoor temperature time series for that time period. The errors between the values ​​are evaluated. In addition to MSE, the three error evaluation indicators include the Normalized Mean Bias Error (NMBE) and the Root Mean Square Error Coefficient of Variation (CVRMSE). Generally, the smaller the absolute values ​​of MSE, NMBE, and CVRMSE, the higher the prediction accuracy. Furthermore, if the absolute value of NMBE is less than 10% or the absolute value of CVRMSE is less than 30%, the prediction can be considered reliable. The three error evaluation indicators are expressed as follows: ; ; ; In this embodiment, the BPNN is built using the "newff" function in MATLAB software, the model is trained using the "train" function, and the model's prediction simulation is performed using the "sim" function. The hyperparameters and configuration parameters are set as shown in Table 2.

[0053] Table 2. Main hyperparameters and corresponding settings of the backpropagation neural network

[0054] The final predicted indoor building temperature calculated after coupling the physical model-guided module and the data-driven module, and a comparison with the results obtained from the simplified physical and data-driven dual-mechanism model 2R2C and the pure data-driven BPNN, are shown below. Figure 5 As shown. By Figure 5As can be seen, in general, the fluctuation trends of the method of this invention and the existing methods are relatively consistent with the actual results, indicating that their prediction results are relatively reasonable. However, the physical interpretability of the method of this invention and the 2R2C model is not possessed by the purely data-driven BPNN. In addition, although the results of the method of this invention and the purely data-driven BPNN model both show a certain degree of oscillation, they are more consistent with the actual results set in this embodiment. The results of the 2R2C model show a certain degree of deviation. Therefore, considering the above two points, in this embodiment, the method of this invention has certain advantages in terms of practicality, as it is more consistent with the actual results and can provide physical interpretation of the results.

[0055] Furthermore, the prediction result error evaluation indexes of the method of the present invention and existing methods are shown in Table 3: Table 3 Comparison of Prediction Result Error Evaluation Indicators

[0056] As shown in Table 3, the errors of the method of this invention and the purely data-driven BPNN model are basically at the same level, and both are relatively low, while the error level of the 2R2C model is relatively high. This result further confirms that the method of this invention is relatively accurate in predicting hourly indoor temperatures in buildings.

[0057] Based on the hourly indoor temperature prediction method for buildings based on a dual-drive mechanism provided by this invention, the following technical effects can be achieved: 1) This invention provides a new idea and a new technical route for the research direction of building indoor temperature prediction, which is of certain significance for improving the quality of building indoor environment and helping to achieve the goal of building energy conservation and carbon reduction; 2) The proposed building indoor temperature prediction method forms a simultaneous application of physical model principles and processes and high-quality data, which can achieve relatively accurate prediction of building indoor temperature while making the prediction results have a certain degree of interpretability.

[0058] Example 2 This embodiment discloses a building hourly indoor temperature prediction system based on a dual-drive mechanism.

[0059] A building hourly indoor temperature prediction system based on a dual-drive mechanism includes: The physical model building unit is configured to: build a physical model of the building to be measured, and determine the hourly indoor temperature time series of the building based on the basic information of the building and outdoor meteorological parameters using building energy consumption simulation software; The physical guidance calculation unit is configured to: take a heat balance differential equation with the indoor temperature of the building as an interface variable as a reference, represent a heat transfer process of the envelope structure in a spatial state, construct a physical model guidance module, and obtain the indoor temperature time sequence of the building after physical guidance; The data-driven calculation unit is configured to: construct a data-driven module, take a difference value of the indoor temperature change caused by a non-principal component term that is not explained by the physical model guidance module as a target variable, take outdoor meteorological parameters as inputs, construct and train a back propagation neural network, and obtain a residual prediction sequence. The temperature prediction evaluation unit is configured to: combine the indoor temperature time sequence of the building after physical guidance and the residual prediction sequence, obtain a final indoor temperature time sequence of the building based on the prediction of the back propagation neural network, and evaluate the prediction result. Embodiment three An object of the embodiment is to provide a computer-readable storage medium.

[0060] The computer-readable storage medium has a computer program stored thereon, and the program, when executed by a processor, implements the steps in the building hourly indoor temperature prediction method based on a double driving mechanism according to Embodiment One of the present disclosure.

[0061] Embodiment four An object of the embodiment is to provide an electronic device.

[0062] The electronic device includes a memory, a processor, and a program stored on the memory and executable on the processor, and the processor implements the steps in the building hourly indoor temperature prediction method based on a double driving mechanism according to Embodiment One of the present disclosure when executing the program.

[0063] The steps and methods in the above embodiments two, three, and four correspond to Embodiment One, and the specific embodiments can be referred to the relevant description in Embodiment One. The term “computer-readable storage medium” should be understood to include a single medium or multiple media of one or more instruction sets; and should also be understood to include any medium that can store, encode, or carry instruction sets for execution by a processor and cause the processor to perform any of the methods in the present disclosure.

[0064] Those skilled in the art should understand that each module or step of the present disclosure described above can be implemented by a general computer device, and alternatively, they can be implemented by program code executable by a computing device, so that they can be stored in a storage device and executed by a computing device, or they can be respectively manufactured into individual integrated circuit modules, or a plurality of modules or steps among them can be manufactured into a single integrated circuit module. The present disclosure is not limited to any specific combination of hardware and software.

[0065] The above describes the specific embodiments of the present application in conjunction with the drawings, but is not a limitation on the scope of protection of the present application. Those skilled in the art should understand that various modifications or variations made by those skilled in the art on the basis of the technical solutions of the present application without creative labor are still within the scope of protection of the present application.

Claims

1. A method for predicting hourly indoor temperature in buildings based on a dual-drive mechanism, characterized in that, include: A physical model of the building to be tested is constructed. Based on the basic information of the building and outdoor meteorological parameters, the hourly indoor temperature time series is determined using building energy consumption simulation software. The heat balance differential equation with the building's indoor temperature as the interface variable is selected as the benchmark. At the same time, the heat transfer process of the building envelope is represented in the spatial state to construct a physical model guidance module and obtain the time series of the building's indoor temperature after physical guidance. A data-driven module is constructed, with the difference in indoor temperature change caused by non-principal component terms not explained by the physical model guidance module as the target variable and outdoor meteorological parameters as input. A backpropagation neural network is constructed and trained to obtain the residual prediction sequence. The physical-guided building indoor temperature time series is combined with the residual prediction sequence, and the final building indoor temperature time series is obtained based on the backpropagation neural network. The prediction results are then evaluated for error.

2. The hourly indoor temperature prediction method for buildings based on a dual-drive mechanism as described in claim 1, characterized in that, The basic information of the building being measured includes its geographical location, external dimensions, orientation, internal structure, building envelope type, thermal parameters, and room functions.

3. The hourly indoor temperature prediction method for buildings based on a dual-drive mechanism as described in claim 1, characterized in that, The heat balance differential equation with the building's indoor temperature as the interface variable is expressed as: ； in, This indicates the sensible heat capacity of air. , , and These represent the indoor and outdoor air temperatures, the inner surface of the wall, and the surface temperature of the air conditioning coil, respectively. This represents the sum of convective heat transfer from all internal heat sources. This represents the sum of all convective heat transfers within the walls. This indicates the heat brought in by air infiltration.

4. The hourly indoor temperature prediction method for buildings based on a dual-drive mechanism as described in claim 1, characterized in that, The heat transfer process of the building envelope is represented in a spatial state, that is, the heat transfer process of the building envelope is represented by a spatial state equation with two internal nodes.

5. The hourly indoor temperature prediction method for buildings based on a dual-drive mechanism as described in claim 1, characterized in that, The construction of the physical model guidance module includes: using a difference method to discretize the benchmark formula corresponding to the physical model and the spatial state equation used to represent the heat transfer process of the building envelope in time, and filtering multiple physical variables, using the remaining physical variables after filtering as principal components to construct the physical model guidance module.

6. The hourly indoor temperature prediction method for buildings based on a dual-drive mechanism as described in claim 1, characterized in that, The outdoor meteorological parameters include: outdoor air temperature, direct solar radiation intensity, diffuse radiation intensity, sky temperature, and wind speed.

7. The hourly indoor temperature prediction method for buildings based on a dual-drive mechanism as described in claim 1, characterized in that, Meanwhile, the MSE error, normalized average deviation error, and root mean square error coefficient of variation are used to evaluate the prediction results.

8. A building hourly indoor temperature prediction system based on a dual-drive mechanism, characterized in that, include: The physical model building unit is configured to: build a physical model of the building to be measured, and determine the hourly indoor temperature time series of the building based on the basic information of the building and outdoor meteorological parameters using building energy consumption simulation software; The physical-guided calculation unit is configured to: select the thermal balance differential equation with the building's indoor temperature as the interface variable as the benchmark, and represent the heat transfer process of the building envelope in the spatial state, so as to construct a physical model guidance module and obtain the time series of the building's indoor temperature after physical guidance; The data-driven computing unit is configured to: construct a data-driven module, using the difference in indoor temperature change caused by non-principal component terms not explained by the physical model guidance module as the target variable and outdoor meteorological parameters as input, construct and train a backpropagation neural network to obtain a residual prediction sequence; The temperature prediction and evaluation unit is configured to combine the physically guided building indoor temperature time series with the residual prediction sequence, predict the final building indoor temperature time series based on the backpropagation neural network, and evaluate the error of the prediction results.

9. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the hourly indoor temperature prediction method for buildings based on a dual-drive mechanism as described in any one of claims 1-7.

10. An electronic device, comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the hourly indoor temperature prediction method for buildings based on a dual-drive mechanism as described in any one of claims 1-7.

Images