Dynamic building indoor temperature solving method, system, medium and equipment

Through the method of mechanism data fusion, a trend simulation model and a nonlinear model are built, combined with sliding window technology, the problem that traditional temperature control methods cannot respond to dynamic environmental changes is solved, and more accurate indoor temperature control and energy efficiency improvement is achieved.

CN118940356APending Publication Date: 2024-11-12ELECTRIC POWER RES INST OF STATE GRID ZHEJIANG ELECTRIC POWER COMAPNY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410926499.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-11
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

Traditional building indoor temperature control methods cannot accurately predict and respond to dynamic environmental changes, resulting in inefficient utilization of energy resources, and a single model is difficult to apply to various scenarios.

Method used

The dynamic building indoor temperature solution method is adopted with a dynamic building indoor temperature solution method with mechanism data fusion. By constructing a trend simulation model, a mechanism nonlinear model and a temperature solution model, combined with sliding window technology, considering multiple load change trends and mechanism correlations, to achieve accurate indoor temperature control.

Benefits of technology

It realizes more accurate indoor temperature control, improves the efficiency of building energy utilization, is suitable for various scenarios, reduces misjudgment of the real situation, and saves energy consumption.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940356A_ABST
    Figure CN118940356A_ABST
Patent Text Reader

Abstract

The invention discloses a dynamic building indoor temperature solving method and system, a medium and equipment, and belongs to the technical field of air conditioner power load control. The invention discloses a dynamic building indoor temperature solving method based on mechanism data fusion. A trend simulation model, a mechanism nonlinear model and a temperature solving model are constructed; according to the method, fresh air load and dehumidification load change rules are described through a trend simulation model, multiple load change trends and mechanism association are considered in combination with a mechanism nonlinear model, and the building indoor temperature is dynamically solved through a temperature solving model by applying a sliding window, so that instant and accurate calculation of the indoor average temperature is realized; therefore, temperature setting can be adjusted in a self-adaptive mode according to different scenes, energy consumption is effectively saved, the indoor environment comfort is improved, the precision and efficiency of building indoor temperature control are greatly improved, building energy utilization is more efficient, and a brand-new thought and technical support are brought to building temperature control.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a method, system, medium and equipment for solving the indoor temperature of a dynamic building, and belongs to the technical field of air-conditioning power load control. Background Art

[0002] As the world's requirements for sustainable development and energy efficiency continue to increase, energy management in the building sector has become one of the focuses of increasing attention. In buildings, indoor temperature control is an important factor in determining energy consumption and the comfort of the living / working environment. However, traditional temperature control methods are often oversimplified and cannot accurately predict and respond to dynamic environmental changes, resulting in inefficient use of energy resources. Currently, common building indoor temperature control methods often use linear models or control schemes based on static data. These methods usually ignore the complexity of time changes, seasonal changes, human activities or other influencing factors, resulting in misjudgment and delayed response to the actual indoor conditions. At the same time, in practical applications, due to regional differences, building structure characteristics and other reasons, a single model is difficult to apply to all scenarios.

[0003] The information disclosed in this Background Art is only for understanding the background of the inventive concept and therefore it may include information that does not constitute the prior art. Summary of the invention

[0004] In response to the above problem or one of the above problems, an object of the present invention is to provide a dynamic building indoor temperature solution method with mechanism data fusion, which accurately solves the building indoor temperature by constructing a trend simulation model, a mechanism nonlinear model, and a temperature solution model, thereby effectively considering the relationship between various load change trends and mechanisms, and being able to flexibly respond to different actual environmental scenarios, thereby achieving more precise indoor temperature control and making building energy utilization more efficient.

[0005] In response to the above problem or one of the above problems, the second object of the present invention is to provide a dynamic building indoor temperature solution system with mechanism data fusion, which can accurately solve the building indoor temperature by setting a trend simulation module, a mechanism nonlinear module, and a temperature solution module, so as to effectively consider the relationship between various load change trends and mechanisms, and can flexibly respond to different actual environmental scenarios, thereby achieving more precise indoor temperature control and making building energy utilization more efficient.

[0006] In response to the above problem or one of the above problems, the third object of the present invention is to provide a dynamic building indoor temperature solution method, system, medium and equipment. Through in-depth analysis of load changes, mechanism modeling and dynamic solution, the temperature setting can be adaptively adjusted for different scenarios, and energy consumption can be effectively saved, the indoor environmental comfort can be improved, the accuracy and efficiency of building indoor temperature control can be greatly improved, and the misjudgment of the actual indoor conditions can be reduced, so that it is suitable for various scenarios.

[0007] To achieve one of the above purposes, the first technical solution of the present invention is:

[0008] A dynamic building indoor temperature solution method based on mechanism data fusion includes the following contents:

[0009] Obtain data on conditions that affect changes in air conditioning load;

[0010] Use the pre-built trend simulation model to process the conditional data and obtain the load change prediction results;

[0011] Through the pre-built mechanism nonlinear model, the indoor temperature setting data is calculated according to the load change prediction results;

[0012] The pre-built temperature solution model is used, based on the indoor temperature setting data, and a sliding window is used for dynamic solution to obtain the indoor temperature of the building, realizing dynamic building indoor temperature solution based on mechanism data fusion.

[0013] After continuous exploration and experimentation, the present invention accurately solves the indoor temperature of the building by constructing a trend simulation model, a mechanism nonlinear model, and a temperature solution model, thereby being able to effectively consider the association between various load change trends and mechanisms, and being able to flexibly respond to different actual environmental scenarios, thereby achieving more precise indoor temperature control and making building energy utilization more efficient.

[0014] Furthermore, the present invention describes the changing rules of fresh air load and dehumidification load through trend simulation model, and considers the correlation between various load change trends and mechanisms in combination with mechanism nonlinear model, and then uses sliding window to dynamically solve the indoor temperature of the building through temperature solution model, thereby realizing real-time and accurate calculation of indoor average temperature, thus bringing new ideas and technical support for building temperature control.

[0015] Furthermore, the present invention can fully consider the complexity of time changes, seasonal changes, human activities or other influencing factors through in-depth analysis of load changes, mechanism modeling and dynamic solution, and can adaptively adjust temperature settings for different scenarios, effectively save energy consumption, improve indoor environmental comfort, greatly improve the accuracy and efficiency of indoor temperature control in buildings, reduce misjudgment of real indoor conditions, and make the multi-model structure of the present invention applicable to various scenarios.

[0016] As the preferred technical measures:

[0017] The condition data includes environmental conditions and condition coefficients;

[0018] The environmental conditions include outdoor temperature, humidity and solar radiation intensity; the condition coefficients include the influence coefficient of outdoor temperature on load, the influence coefficient of humidity on load and the influence coefficient of solar radiation intensity on load;

[0019] Or, the environmental conditions include outdoor temperature, humidity, solar radiation intensity, thermal properties of building materials, size and number of windows and indoor heat sources; the condition coefficients include the influence coefficient of outdoor temperature on the load, the influence coefficient of humidity on the load, the influence coefficient of solar radiation intensity on the load, the influence coefficient of thermal properties of building materials on the load, the influence coefficient of size and number of windows on the load and the influence coefficient of indoor heat source on the load.

[0020] The method of using the pre-built trend simulation model to process the conditional data and obtain the load change prediction result is as follows:

[0021] According to environmental conditions and condition coefficients, a regression analysis equation is constructed to characterize the load change trend;

[0022] The expression of the regression analysis equation is as follows:

[0023] Q load (t)=α+β1·Condition1(t)+β2·Condition2(t)+...+β n Condition n (t)+∈

[0024] In the formula, Q load (t) represents the load at a specific time, which includes the fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t), α is a constant term; β1,β2,...,β nis the relationship coefficient between load and outdoor temperature, humidity, ... solar radiation intensity; the coefficient β reflects the influence of each condition on the load; Condition1(t), Condition2(t)..., Condition n (t) represent the outdoor temperature, humidity, solar radiation intensity at time t respectively;

[0025] The least square method is used to adjust the relationship coefficient and constant term in the regression analysis equation to obtain new relationship coefficient and constant term;

[0026] Substitute the new relationship coefficient and constant term into the regression analysis equation and solve it to obtain a more accurate load change prediction result under static environmental conditions;

[0027] Alternatively or additionally, a method of processing conditional data using a pre-built trend simulation model to obtain load change prediction results is as follows:

[0028] According to environmental conditions and condition coefficients, considering the dynamic relationship of load over time, a time series formula is constructed, and its calculation formula is as follows:

[0029] Y t =f(Y t-1 ,Y t-2 ,…,Y t-k +e t )

[0030] Where Y t represents the load measurement value at time t; k is the order of historical data considered by the model; f(·) represents the function used to describe the dynamic relationship of the time series; e t represents the error term of the model;

[0031] The least square method is used to estimate the error term in the time series formula to obtain a new error term;

[0032] Substitute the new error term into the time series formula and solve it to obtain more accurate load change prediction results under time dynamic environmental conditions.

[0033] As the preferred technical measures:

[0034] The coefficient of relationship is determined in regression analysis by minimizing the sum of squared errors (SSE) between the predicted load value and the actual load value, which is done using the least squares method, including the following:

[0035] Get a load data set, which includes multiple load time points, each of which has a corresponding load amount Q load(t) and a series of environmental conditions Condition1(t), Condition2(t)..., Condition n (t);

[0036] And the relationship coefficients β1,β2,...,β n Initialize with constant term α to obtain the initial values ​​of the relationship coefficient and constant term;

[0037] Substitute the current relationship coefficient value and constant term into the regression analysis equation to calculate the predicted load at each load time point. The corresponding calculation formula is as follows:

[0038] (Q load (t)predicted=α+β1·Condition1(t)+β2·Condition2(t)+...+β n Condition n (t));

[0039] Calculate the error between the predicted load value and the actual load value at each load time point to obtain the error data. The calculation formula is as follows:

[0040] e(t)=Q load (t) observed -Q load (t) predicted ;

[0041] According to the error data, a load coefficient linear equation group is established. The linear equation group is composed of the condition that the partial derivative of the error data to the relationship coefficient is equal to zero; the expression of the load coefficient linear equation group is as follows:

[0042]

[0043] Based on the principle of minimizing the sum of squared errors (SSE), the load factor linear equations are solved to obtain new relationship coefficients and constant terms.

[0044] Repeat steps 232) to 235 until a certain stopping criterion is met, and output the final relationship coefficient and constant term; the stopping criterion includes reaching the maximum number of iterations or / and the coefficient change is less than a certain threshold;

[0045] Or / and, the method to construct the time series formula is as follows:

[0046] Through the autoregressive model or the moving average model or the autoregressive moving average model, based on the data characteristics, a function f(·) is constructed to describe the dynamic relationship of the time series, and the time series formula is obtained;

[0047] In the autoregressive model, the future load value Yt is the past load value Y t-1 ,Y t-2 ,...,Y t-k A linear combination of t ; The time series formula constructed is as follows:

[0048] Y t =φ1Y t-1 +φ2Y t-2 +…+φ k Y t-k +e t

[0049] Where φ1, φ2, ..., φ k is the autoregressive coefficient;

[0050] In the moving average model, the future load value Y t is the past error term e t-1 ,e t-2 ,...,e t-k The linear combination of ; the time series formula constructed is as follows:

[0051] Y t =e t +θ1e t-1 +θ2e t-2 +…+θ k e t-k

[0052] Where θ1,θ2,...,θ k is the moving average coefficient;

[0053] The autoregressive moving average model combines the characteristics of autoregression and moving average; the time series formula it constructs is as follows:

[0054] Y t =φ1Y t-1 +φ2Y t-2 +…+φ k Y t-k +e t +θ1e t-1 +θ2e t-2 +…+θ q e t-q

[0055] Here, k is the order of autoregression and q is the order of moving average.

[0056] As the preferred technical measures:

[0057] The method of calculating the indoor temperature setting data based on the load change prediction results through the pre-built mechanism nonlinear model is as follows:

[0058] Obtain load change prediction results, including fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t);

[0059] Based on fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t) Establish a nonlinear mechanism formula to characterize the thermodynamic principles and building physics involved in the process of indoor temperature change in buildings;

[0060] The calculation formula of the nonlinear mechanism formula is as follows:

[0061]

[0062] Where, T set (t) represents the indoor temperature setting data; T out (t) represents the external temperature; Q fresh (t) represents the fresh air load; Q dehum (t) represents the dehumidification load; Q ac (t) represents the air conditioning power load; γ is the parameter of air conditioning system efficiency, which reflects the air conditioning power load Q ac (t) for indoor temperature T set (t); δ is the linear influence coefficient of fresh air and dehumidification load on indoor temperature setting; ζ is the reference temperature bias, which reflects the influence of the remaining factors not directly included in the model on the indoor temperature;

[0063] The nonlinear mechanism formula is solved to obtain the indoor temperature setting data.

[0064] As the preferred technical measures:

[0065] The parameter γ of the air conditioning system efficiency reflects the air conditioning power load Q ac (t) Set the indoor temperature data T set (t) Parameters of influence, which are related to air conditioning system efficiency, building thermal characteristics, and indoor and outdoor temperature difference;

[0066] The value of parameter γ is calculated by regression analysis and optimized by minimizing the prediction error. The calculation formula is as follows:

[0067] T set (t) = α + β·T out (t)+γ·Q oc (t)+χ

[0068] In the formula, α and β are constant terms, representing the relevant parameters, which are estimated by regression analysis, and α is used to adjust the baseline level of the model;

[0069] The linear influence coefficient δ of fresh air and dehumidification load on indoor temperature setting data is calculated by regression analysis method and optimized by minimizing the prediction error. The calculation formula is as follows:

[0070] T set (t) = α + β·T out (t)+γ·Q oc (t)+δ·(Q fresh (t)+Q dehum (t))+χ * .

[0071] As the preferred technical measures:

[0072] Using the pre-built temperature solution model, based on the indoor temperature setting data, and using the sliding window for dynamic solution, the method for obtaining the indoor temperature of the building is as follows:

[0073] Get indoor temperature setting data;

[0074] Based on the indoor temperature setting data, combined with the sliding window with time series data as the training data source, a dynamic prediction formula for indoor temperature is constructed. The calculation formula is as follows:

[0075] T avg (t n )=Φ(T set (t n-1 ),T set (t n-2 ),...,T set (t n-k ))

[0076] Where, T avg (t n ) represents the time t n Average indoor temperature; T set (t n-k ) represents the indoor temperature setting data of the previous k moments; Φ(·) represents a nonlinear function, which is used to capture the nonlinear relationship between temperature setting values ​​and predict future indoor temperatures; k represents the size of the sliding window, which determines the range of historical data considered by the model;

[0077] A multi-layer perceptron MLP neural network is used as a nonlinear function to solve the indoor temperature of the building; the method for solving the nonlinear function is as follows:

[0078] Build a multilayer perceptron with two hidden layers, each with h neurons and using ReLU as the activation function, and the output layer uses a linear activation function; the input layer has k neurons, corresponding to the k historical set temperature values ​​in the sliding window;

[0079] The input layer passes the historical set temperature values ​​in the sliding window to the hidden layer;

[0080] For the i-th neuron in the first hidden layer, i ranges from 1 to h, its output The calculation formula is as follows:

[0081]

[0082] In the formula, is the weight from the jth neuron in the input layer to the ith neuron in the first hidden layer, is the bias term of the i-th neuron in the first hidden layer;

[0083] For the i-th neuron in the second hidden layer, i ranges from 1 to h, its output The calculation formula is as follows:

[0084]

[0085]

[0086] In the formula, is the weight from the jth neuron in the first hidden layer to the ith neuron in the second hidden layer, is the bias term of the i-th neuron in the second hidden layer;

[0087] The output layer outputs the predicted average indoor temperature T avg (t n ), and the average indoor temperature T avg (t n ) is used as the indoor temperature of the building, and its calculation formula is as follows:

[0088]

[0089] T avg (t n )=z avg

[0090] In the formula, is the weight from the jth neuron in the second hidden layer to the output layer, b out is the bias term of the output layer.

[0091] Therefore, the present invention has intelligent learning capabilities and can adjust temperature settings in real time according to historical temperature data, thereby achieving more accurate indoor temperature control and making building energy utilization more efficient.

[0092] To achieve one of the above purposes, the second technical solution of the present invention is:

[0093] A dynamic building indoor temperature solution system with mechanism data fusion, including a trend simulation module, a mechanism nonlinear module and a temperature solution module;

[0094] Trend simulation module, used to process conditional data and obtain load change prediction results;

[0095] The mechanism nonlinear module is used to calculate the indoor temperature setting data based on the load change prediction results;

[0096] The temperature solving module is used to obtain the indoor temperature of the building based on the indoor temperature setting data and dynamically solve using a sliding window.

[0097] After continuous exploration and experimentation, the present invention accurately solves the indoor temperature of the building by setting a trend simulation module, a mechanism nonlinear module, and a temperature solving module, so as to effectively consider the association between various load change trends and mechanisms, and can flexibly respond to different actual environmental scenarios, thereby achieving more precise indoor temperature control and making building energy utilization more efficient.

[0098] Furthermore, the present invention achieves real-time and accurate calculation of the average indoor temperature by detailed research on the changing patterns of fresh air load and dehumidification load in the building air-conditioning system, combining the mechanism nonlinear model and using the sliding window dynamic solution technology, thus bringing new ideas and technical support to building temperature control.

[0099] As the preferred technical measures:

[0100] The trend simulation module is used to process the conditional data to obtain the load change prediction results, which include the following:

[0101] According to environmental conditions and condition coefficients, a regression analysis equation is constructed to characterize the load change trend;

[0102] The expression of the regression analysis equation is as follows:

[0103] Q load (t)=α+β1·Condition1(t)+β2·Condition2(t)+...+β n Condition n (t)+∈

[0104] In the formula, Qload (t) represents the load at a specific time, which includes the fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t), α is a constant term; β1,β2,...,β n is the relationship coefficient between load and outdoor temperature, humidity, ... solar radiation intensity; the coefficient β reflects the influence of each condition on the load; Condition1(t), Condition2(t)..., Condition n (t) represent the outdoor temperature, humidity, solar radiation intensity at time t respectively;

[0105] Considering the dynamic relationship of load over time, a time series formula is constructed, and its calculation formula is as follows:

[0106] Y t =f(Y t-1 ,Y t-2 ,…,Y t-k +e t )

[0107] Where Y t represents the load measurement value at time t; k is the order of historical data considered by the model; f(·) represents the function used to describe the dynamic relationship of the time series; e t represents the error term of the model;

[0108] The least square method is used to adjust the relationship coefficient and constant term in the regression analysis equation to obtain new relationship coefficient and constant term;

[0109] The least square method is used to estimate the error term in the time series formula to obtain a new error term;

[0110] Substitute the new relationship coefficient and constant term into the regression analysis equation and solve it to obtain a more accurate load change prediction result under static environmental conditions;

[0111] Substitute the new error term into the time series formula and solve it to obtain more accurate load change prediction results under time dynamic environmental conditions;

[0112] The load change prediction result under static environmental conditions is coupled with the load change prediction result under dynamic environmental conditions to obtain the final load change prediction result;

[0113] Or / and, a mechanism nonlinear module is used to calculate the indoor temperature setting data according to the load change prediction results, which includes the following contents:

[0114] Obtain load change prediction results, including fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t);

[0115] Based on fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t) Establish a nonlinear mechanism formula to characterize the thermodynamic principles and building physics involved in the process of indoor temperature change in buildings;

[0116] The calculation formula of the nonlinear mechanism formula is as follows:

[0117]

[0118] Where, T set (t) represents the indoor temperature setting data; T out (t) represents the external temperature; Q fresh (t) represents the fresh air load; Q dehum (t) represents the dehumidification load; Q ac (t) represents the air conditioning power load; γ is the parameter of air conditioning system efficiency, which reflects the air conditioning power load Q ac (t) for indoor temperature T set (t); δ is the linear influence coefficient of fresh air and dehumidification load on indoor temperature setting; ζ is the reference temperature bias, which reflects the influence of the remaining factors not directly included in the model on the indoor temperature;

[0119] Solve the nonlinear mechanism formula to obtain the indoor temperature setting data;

[0120] Or / and, a temperature solving module is used to dynamically solve the indoor temperature of the building based on the indoor temperature setting data and using a sliding window, which includes the following contents:

[0121] Get indoor temperature setting data;

[0122] Based on the indoor temperature setting data, combined with the sliding window with time series data as the training data source, a dynamic prediction formula for indoor temperature is constructed. The calculation formula is as follows:

[0123] T avg (t n )=Φ(T set (t n-1 ),T set (t n-2 ),...,T set (t n-k ))

[0124] Where, T avg (t n ) represents the time t n Average indoor temperature; T set (t n-k ) represents the indoor temperature setting data of the previous k moments; Φ(·) represents a nonlinear function, which is used to capture the nonlinear relationship between temperature setting values ​​and predict future indoor temperatures; k represents the size of the sliding window, which determines the range of historical data considered by the model;

[0125] A multi-layer perceptron (MLP) neural network is used as a nonlinear function to solve and obtain the indoor temperature of the building.

[0126] To achieve one of the above purposes, the third technical solution of the present invention is:

[0127] An electronic device comprising:

[0128] one or more processors;

[0129] A storage device for storing one or more programs;

[0130] When the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned method for solving the dynamic building indoor temperature by mechanism data fusion.

[0131] To achieve one of the above purposes, the fourth technical solution of the present invention is:

[0132] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned dynamic building indoor temperature solution method using mechanism data fusion.

[0133] Compared with the prior art solutions, the present invention has the following beneficial effects:

[0134] After continuous exploration and experimentation, the present invention accurately solves the indoor temperature of the building by constructing a trend simulation model, a mechanism nonlinear model, and a temperature solution model, thereby being able to effectively consider the association between various load change trends and mechanisms, and being able to flexibly respond to different actual environmental scenarios, thereby achieving more precise indoor temperature control and making building energy utilization more efficient.

[0135] Furthermore, the present invention describes the changing rules of fresh air load and dehumidification load through trend simulation model, and considers the correlation between various load change trends and mechanisms in combination with mechanism nonlinear model, and then uses sliding window to dynamically solve the indoor temperature of the building through temperature solution model, thereby realizing real-time and accurate calculation of indoor average temperature, thus bringing new ideas and technical support for building temperature control.

[0136] Furthermore, the present invention can adaptively adjust temperature settings for different scenarios through in-depth analysis of load changes, mechanism modeling and dynamic solution, effectively save energy consumption, improve indoor environmental comfort, and greatly improve the accuracy and efficiency of indoor temperature control in buildings.

[0137] Furthermore, after continuous exploration and experimentation, the present invention accurately solves the indoor temperature of the building by setting a trend simulation module, a mechanism nonlinear module, and a temperature solving module, so that it can effectively consider the relationship between various load change trends and mechanisms, and can flexibly respond to different actual environmental scenarios, thereby achieving more precise indoor temperature control and making building energy utilization more efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0138] Figure 1 It is a first flow chart of the method for solving the dynamic building indoor temperature by fusion of mechanism data of the present invention;

[0139] Figure 2 This is a second flow chart of the method for solving the dynamic building indoor temperature by fusion of mechanism data of the present invention. DETAILED DESCRIPTION

[0140] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0141] On the contrary, the present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention as defined by the claims. Further, in order to make the public have a better understanding of the present invention, some specific details are described in detail in the detailed description of the present invention below. Those skilled in the art can fully understand the present invention without the description of these details.

[0142] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention pertains. The terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "or / and" as used herein includes any and all combinations of one or more of the associated listed items.

[0143] The first specific embodiment of the dynamic building indoor temperature solution method of the mechanism data fusion of the present invention is:

[0144] A dynamic building indoor temperature solution method based on mechanism data fusion includes the following contents:

[0145] Obtain data on conditions that affect changes in air conditioning load;

[0146] Use the pre-built trend simulation model to process the conditional data and obtain the load change prediction results;

[0147] Through the pre-built mechanism nonlinear model, the indoor temperature setting data is calculated according to the load change prediction results;

[0148] The pre-built temperature solution model is used, based on the indoor temperature setting data, and a sliding window is used for dynamic solution to obtain the indoor temperature of the building, thereby realizing dynamic building indoor temperature solution based on mechanism data fusion.

[0149] like Figure 1 As shown, the second specific embodiment of the dynamic building indoor temperature solution method of the mechanism data fusion of the present invention is:

[0150] A dynamic building indoor temperature solution method based on mechanism data fusion includes the following steps:

[0151] The first step is to obtain the condition data that affects the change of air conditioning load;

[0152] In the second step, the conditional data is processed using the pre-built trend simulation model to obtain the load change prediction results;

[0153] The third step is to calculate the indoor temperature setting data based on the load change prediction results through the pre-built mechanism nonlinear model;

[0154] The fourth step is to use the pre-built temperature solution model, based on the indoor temperature setting data, and use the sliding window for dynamic solution to obtain the indoor temperature of the building, thereby realizing dynamic building indoor temperature solution based on mechanism data fusion.

[0155] In this embodiment:

[0156] In the first step, the condition data includes environmental conditions and condition coefficients;

[0157] The environmental conditions include outdoor temperature, humidity and solar radiation intensity; the condition coefficients include the influence coefficient of outdoor temperature on load, the influence coefficient of humidity on load and the influence coefficient of solar radiation intensity on load.

[0158] Or, the environmental conditions include outdoor temperature, humidity, solar radiation intensity, thermal properties of building materials, size and number of windows and indoor heat sources; the condition coefficients include the influence coefficient of outdoor temperature on load, the influence coefficient of humidity on load, the influence coefficient of solar radiation intensity on load, the influence coefficient of thermal properties of building materials on load, the influence coefficient of size and number of windows on load and the influence coefficient of indoor heat source on load;

[0159] Or / and, in the second step, the method of obtaining the load change prediction result by using the trend simulation model is as follows:

[0160] Step 21, constructing a regression analysis equation based on environmental conditions and condition coefficients to characterize the load change trend;

[0161] The expression of the regression analysis equation is as follows:

[0162] Q load (t)=α+β1·Condition1(t)+β2·Condition2(t)+...+β n Condition n (t)+∈

[0163] In the formula, Q load (t) represents the load at a specific time, which includes the fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t), α is a constant term; β1,β2,...,β n is the relationship coefficient between load and outdoor temperature, humidity, ... solar radiation intensity; the coefficient β reflects the influence of each condition on the load; Condition1(t), Condition2(t)..., Condition n (t) represent the outdoor temperature, humidity, solar radiation intensity at time t respectively;

[0164] Step 22, using the least square method to adjust the relationship coefficient and constant term in the regression analysis equation to obtain new relationship coefficient and constant term;

[0165] Step 23, substituting the new relationship coefficient and constant term into the regression analysis equation and solving it to obtain a more accurate load change prediction result under static environmental conditions;

[0166] Or / and, in the second step, the method of obtaining the load change prediction result by using the trend simulation model is as follows:

[0167] Step 21, based on environmental conditions and condition coefficients, considering the dynamic relationship of load over time, construct a time series formula, the calculation formula is as follows:

[0168] Y t =f(Y t-1 ,Y t-2 ,…,Y t-k +e t )

[0169] Where Y trepresents the load measurement value at time t; k is the order of historical data considered by the model; f(·) represents the function used to describe the dynamic relationship of the time series; e t represents the error term of the model;

[0170] Step 22, using the least square method to estimate the error term in the time series formula to obtain a new error term;

[0171] Step 23, substituting the new error term into the time series formula and solving it to obtain a more accurate load change prediction result under time dynamic environmental conditions.

[0172] In this embodiment:

[0173] The relationship coefficient is determined in the regression analysis by minimizing the sum of squared errors (SSE) between the predicted load value and the actual load value, which is done using the least squares method, including the following steps:

[0174] Step 231, obtain a load data set, which includes multiple load time points, each of which has a corresponding load value Q load (t) and a series of environmental conditions Condition1(t), Condition2(t)..., Condition n (t);

[0175] And the relationship coefficients β1,β2,...,β n Initialize with constant term α to obtain the initial values ​​of the relationship coefficient and constant term;

[0176] Step 232, substitute the current relationship coefficient value and constant term into the regression analysis equation to calculate the predicted load at each load time point. The corresponding calculation formula is as follows:

[0177] (Q load (t)predicted=α+β1·Condition1(t)+β2·Condition2(t)+...+β n Condition n (t));

[0178] Step 233, calculate the error between the predicted load value and the actual load value at each load time point to obtain error data, and the calculation formula is as follows:

[0179] e(t)=Q load (t) observed -Q load (t) predicted ;

[0180] Step 234, based on the error data, a load coefficient linear equation group is established, the linear equation group is composed of the condition that the partial derivative of the error data with respect to the relationship coefficient is equal to zero; the expression of the load coefficient linear equation group is as follows:

[0181]

[0182] Step 235, solving the load factor linear equations based on the principle of minimizing the sum of squared errors SSE to obtain new relationship coefficients and constant terms;

[0183] Step 236, repeating steps 232) to 235 until a certain stopping criterion is met, and outputting the final relationship coefficient and constant term; the stopping criterion includes reaching a maximum number of iterations or / and a coefficient change being less than a certain threshold;

[0184] Or / and, the method to construct the time series formula is as follows:

[0185] Through the autoregressive model or the moving average model or the autoregressive moving average model, based on the data characteristics, a function f(·) is constructed to describe the dynamic relationship of the time series, and the time series formula is obtained;

[0186] In the autoregressive model, the future load value Y t is the past load value Y t-1 ,Y t-2 ,...,Y t-k A linear combination of t ; The constructed time series formula is as follows:

[0187] Y t =φ1Y t-1 +φ2Y t-2 +…+φ k Y t-k +e t

[0188] Where φ1, φ2, ..., φ k is the autoregressive coefficient;

[0189] In the moving average model, the future load value Y t is the past error term e t-1 ,e t-2 ,...,e t-k The linear combination of ; the time series formula constructed is as follows:

[0190] Y t =e t +θ1e t-1 +θ2e t-2 +…+θ k e t-k

[0191] Where θ1,θ2,...,θ k is the moving average coefficient;

[0192] The autoregressive moving average model combines the characteristics of autoregression and moving average; the time series formula it constructs is as follows:

[0193] Y t =φ1Y t-1 +φ2Y t-2 +…+φ k Y t-k +e t +θ1e t-1 +θ2e t-2 +…+θ q e t-q

[0194] Here, k is the order of autoregression and q is the order of moving average.

[0195] In this embodiment:

[0196] In the third step, the method for obtaining the indoor temperature setting data through the mechanism nonlinear model is as follows:

[0197] Step 31, obtain load change prediction results, which include fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t);

[0198] Step 32: Based on the fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t) Establish a nonlinear mechanism formula to characterize the thermodynamic principles and building physics involved in the process of indoor temperature change in buildings;

[0199] The calculation formula of the nonlinear mechanism formula is as follows:

[0200]

[0201] Where, T set (t) represents the indoor temperature setting data; T out (t) represents the external temperature; Q fresh (t) represents the fresh air load; Q dehum (t) represents the dehumidification load; Q ac (t) represents the air conditioning power load; γ is the parameter of air conditioning system efficiency, which reflects the air conditioning power load Q ac (t) for indoor temperature T set(t); δ is the linear influence coefficient of fresh air and dehumidification load on indoor temperature setting; ζ is the reference temperature bias, which reflects the influence of the remaining factors not directly included in the model on the indoor temperature;

[0202] Step 33, solving the nonlinear mechanism formula to obtain indoor temperature setting data.

[0203] In this embodiment:

[0204] The parameter γ of the air conditioning system efficiency reflects the air conditioning power load Q ac (t) Set the indoor temperature data T set (t) Parameters of influence, which are related to air conditioning system efficiency, building thermal characteristics, and indoor and outdoor temperature difference;

[0205] The value of parameter γ is calculated by regression analysis and optimized by minimizing the prediction error. The calculation formula is as follows:

[0206] T set (t) = α + β·T out (t)+γ·Q oc (t)+χ

[0207] In the formula, α and β are constant terms, representing the relevant parameters, which are estimated by regression analysis, and α is used to adjust the baseline level of the model;

[0208] The linear influence coefficient δ of fresh air and dehumidification load on indoor temperature setting data is calculated by regression analysis method and optimized by minimizing the prediction error. The calculation formula is as follows:

[0209] T set (t) = α + β·T out (t)+γ·Q oc (t)+δ·(Q fresh (t)+Q dehum (t))+χ * .

[0210] In this embodiment:

[0211] In the fourth step, the temperature solution model is used to obtain the indoor temperature of the building as follows:

[0212] Step 41, obtaining indoor temperature setting data;

[0213] Step 42, based on the indoor temperature setting data, combined with the sliding window with time series data as the training data source, construct a dynamic prediction formula for indoor temperature, the calculation formula is as follows:

[0214] T avg (t n )=Φ(Tset (t n-1 ),T set (t n-2 ),...,T set (t n-k ))

[0215] Where, T avg (t n ) represents the time t n Average indoor temperature; T set (t n-k ) represents the indoor temperature setting data of the previous k moments; Φ(·) represents a nonlinear function, which is used to capture the nonlinear relationship between temperature setting values ​​and predict future indoor temperatures; k represents the size of the sliding window, which determines the range of historical data considered by the model;

[0216] Step 43, using a multi-layer perceptron MLP neural network as a nonlinear function to solve and obtain the indoor temperature of the building; the method for solving the nonlinear function is as follows:

[0217] Step 431, establish a multilayer perceptron including two hidden layers, each hidden layer has h neurons, and uses ReLU as an activation function, and the output layer uses a linear activation function; the input layer has k neurons, corresponding to k historical set temperature values ​​in the sliding window;

[0218] Step 432, the input layer transmits the historical set temperature value in the sliding window to the hidden layer;

[0219] Step 433, for the i-th neuron in the first hidden layer, i ranges from 1 to h, its output is a i 1 The calculation formula is as follows:

[0220]

[0221] In the formula, is the weight from the jth neuron in the input layer to the ith neuron in the first hidden layer, is the bias term of the i-th neuron in the first hidden layer;

[0222] Step 434, for the i-th neuron in the second hidden layer, i ranges from 1 to h, its output The calculation formula is as follows:

[0223]

[0224]

[0225] In the formula, is the weight from the jth neuron in the first hidden layer to the ith neuron in the second hidden layer, is the bias term of the i-th neuron in the second hidden layer;

[0226] Step 435: the output layer outputs the predicted indoor average temperature T avg (t n ), and the average indoor temperature T avg (t n ) is used as the indoor temperature of the building, and its calculation formula is as follows:

[0227]

[0228] T avg (t n )=z avg

[0229] like Figure 2 As shown, the third specific embodiment of the dynamic building indoor temperature solution method of the mechanism data fusion of the present invention is:

[0230] A dynamic building indoor temperature solution method based on mechanism data fusion studies the changing trends of air conditioning fresh air load and dehumidification load, constructs a nonlinear relationship between set temperature and air conditioning power load under different conditions, and combines the sliding window dynamic solution method to achieve accurate calculation of indoor average temperature; it includes the following steps:

[0231] S1, study the changing trends of fresh air load and dehumidification load of typical building air conditioning under different conditions;

[0232] S2, establish a mechanism nonlinear model including the set temperature of air conditioning fresh air, dehumidification load and air conditioning power load;

[0233] S3, dynamically solves the sliding window with time series data as the training data source to achieve accurate calculation of the average indoor temperature.

[0234] In this embodiment: step S1 includes the following steps:

[0235] S1.1, load change trend analysis using regression analysis method:

[0236] Q load (t)=α+β1·Condition1(t)+β2·Condition2(t)+...+β n Condition n (t)+∈

[0237] In the formula, Q load (t) represents the load at a specific time; α is a constant term; β1,β2,...,β nIt is the relationship coefficient between the load and various conditions, including environmental factors such as outdoor temperature, humidity, and solar radiation intensity; the β coefficient reflects the influence of each condition on the load; Condition1(t), Condition2(t)..., Condition n (t) represents different environmental conditions at time t; ∈ is the error term.

[0238] S1.2, if the dynamic relationship of data over time needs to be considered, the time series model can be expressed as:

[0239] Y t =f(Y t-1 ,Y t-2 ,…,Y t-k +e t )

[0240] Where Y t represents the measured value at time t; k is the order of historical data considered by the model; f(·) represents the function used to describe the dynamic relationship of the time series; e t represents the error term of the model.

[0241] S1.3, the parameters in the two models are estimated using statistical methods such as the least squares method to make the models fit the actual measured values ​​as closely as possible. After adjusting the parameters, more accurate load prediction results under time dynamic or static environmental conditions can be obtained.

[0242] In this embodiment, step S2 includes the following steps:

[0243] The nonlinear mechanism model is constructed to represent the thermodynamic principles and building physics involved in solving the indoor temperature change process of the building:

[0244]

[0245] Where: T set (t) represents the set indoor temperature; T out (t) represents the external temperature; Q fresh (t) represents the fresh air load; Q dehum (t) represents the dehumidification load; Q ac (t) represents the air conditioning power load; γ is the parameter of air conditioning system efficiency, which reflects the air conditioning power load Q ac (t) for indoor temperature T set (t); δ is the linear influence coefficient of fresh air and dehumidification load on indoor temperature setting; ζ is the reference temperature bias, which reflects the influence of other factors not directly included in the model on indoor temperature.

[0246] In this embodiment: step S3 includes the following steps:

[0247] Dynamically solve the sliding window with time series data as the training data source to achieve accurate calculation of the average indoor temperature.

[0248] Based on sliding window technology, combined with machine learning methods, dynamic prediction of indoor temperature:

[0249] T avg (t n )=Φ(T set (t n-1 ),T set (t n-2 ),...,T set (t n-k ))

[0250] Where: T avg (t n ) represents the time t n Average indoor temperature; T set (t n-k ) represents the set temperature of the previous k moments; Φ(·) represents a nonlinear function, including but not limited to machine learning models such as neural networks and support vector machines, which are used to capture the nonlinear relationship between temperature set values ​​and predict future indoor temperatures; k represents the size of the sliding window, which determines the scope of historical data considered by the model.

[0251] Through in-depth analysis of load changes, mechanism modeling and dynamic solution, the present invention can adaptively adjust temperature settings for different scenarios, effectively save energy consumption, improve indoor environmental comfort, and greatly improve the accuracy and efficiency of indoor temperature control in buildings.

[0252] Furthermore, the present invention first analyzes the changing trends of the fresh air load and dehumidification load of air conditioning in typical buildings under different environmental conditions, uses regression analysis or time series analysis methods to obtain a relationship model between the load and different environmental factors, and further introduces a nonlinear mechanism model based on the principles of thermodynamics and building physics. By considering factors such as external temperature, fresh air load, dehumidification load and air conditioning power load, a complex nonlinear relationship between the set temperature and the actual indoor temperature is established. On this basis, using time series data as the training data source and utilizing machine learning methods, a sliding window dynamic solution model is designed to dynamically predict the indoor temperature and continuously adjust the temperature setting to achieve high-precision calculation of the indoor average temperature. The solution is scientific, reasonable and feasible.

[0253] The fourth specific embodiment of the dynamic building indoor temperature solution method of the mechanism data fusion of the present invention is:

[0254] A dynamic building indoor temperature solution method based on mechanism data fusion includes the following steps:

[0255] S1, study the changing trends of the fresh air load and dehumidification load of typical building air conditioning under different conditions, including the following steps:

[0256] S1.1, load change trend analysis using regression analysis method:

[0257] Q load (t)=α+β1·Condition1(t)+β2·Condition2(t)+...+β n Condition n (t)+∈

[0258] In the formula, Q load (t) represents the load at a specific moment, respectively Q fresh (t) Fresh air load, Q dehum (t) Dehumidification load, Q ac (t) represents the load such as air conditioning power load, α is a constant term; β1,β2,...,β n It is the relationship coefficient between the load and various conditions, including environmental factors such as outdoor temperature, humidity, and solar radiation intensity; the β coefficient reflects the influence of each condition on the load; Condition1(t), Condition2(t)..., Condition n (t) represents different environmental conditions at time t, including outdoor temperature, humidity, solar radiation intensity and other environmental factors; ∈ is the error term.

[0259] S1.2, if the dynamic relationship of data over time needs to be considered, the time series model can be expressed as:

[0260] Y t =f(Y t-1 ,Y t-2 ,…,Y t-k +e t )

[0261] Where Y t represents the measured value at time t; k is the order of historical data considered by the model; f(·) represents the function used to describe the dynamic relationship of the time series; e t represents the error term of the model.

[0262] The function f(·) used to describe the dynamic relationship of the time series is related to the data characteristics, and the autoregressive model, moving average model, and autoregressive moving average model are taken as examples.

[0263] 1) Autoregressive model:

[0264] In an autoregressive model, the future value Y t is the past value Yt-1 ,Y t-2 ,...,Y t-k A linear combination of t .

[0265] Y t =φ1Y t-1 +φ2Y t-2 +…+φ k Y t-k +e t

[0266] Where φ1, φ2, ..., φ k is the autoregressive coefficient.

[0267] 2) Moving average model:

[0268] In the moving average model, the future value Y t is the past error term e t-1 ,e t-2 ,...,e t-k A linear combination of .

[0269] Y t =e t +θ1e t-1 +θ2e t-2 +…+θ k e t-k

[0270] Where θ1,θ2,...,θ k is the moving average coefficient.

[0271] 3) Autoregressive Moving Average Model:

[0272] The autoregressive moving average model combines the characteristics of the above autoregressive and moving average.

[0273] Y t =φ1Y t-1 +φ2Y t-2 +…+φ k Y t-k +e t +θ1e t-1 +θ2e t-2 +…+θ q e t-q

[0274] Here, k is the order of autoregression and q is the order of moving average.

[0275] S1.3, the parameters in the two models are estimated by using statistical methods such as the least squares method to make the model fit the actual measured value as much as possible. After adjusting the parameters, more accurate load change prediction results under time dynamic or static environmental conditions can be obtained. The load change prediction results refer to the load mentioned in S1, which refer to Q fresh (t) Fresh air load, Q dehum (t) Dehumidification load, Q ac (t) represents loads such as air conditioning power load.

[0276] In this embodiment, the method for determining the relationship coefficient is as follows:

[0277] Relation coefficients β1, β2, ..., β n In regression analysis, it is determined by minimizing the sum of squared errors (SSE) between the predicted value and the actual value. This is usually done using the Ordinary Least Squares (OLS) method.

[0278] Assume that there is a data set that includes multiple load time points (time points), each of which has a corresponding load amount Q load (t) and a series of environmental conditions Condition1(t), Condition2(t)..., Condition n (t). The goal is to find a set of relationship coefficients β1,β2,...,β n and the constant term α, so that the sum of squared errors between the predicted value and the actual value is minimized.

[0279] The specific processing steps are as follows:

[0280] Input data Q load (t) and Condition1(t),Condition2(t)...,Condition n (t).

[0281] 1) Initialization: relationship coefficients β1, β2, ..., β n and the constant term α sets the initial value (usually 0 or a random value).

[0282] 2) Calculate the predicted value: Use the current coefficient value according to the formula:

[0283] (Q load (t)predicted=α+β1·Condition1(t)+β2·Condition2(t)+...+β n Condition n (t))

[0284] Calculate the predicted load at each load time point.

[0285] 3) Calculation error: Calculate the error between the predicted value and the actual value at each load time point, that is:

[0286] e(t)=Q load (t) observed -Q load (t) predicted .

[0287] 4) Update coefficients: Update the relationship coefficients and constant term using the least squares method to minimize the sum of squared errors (SSE). This usually involves solving a system of linear equations consisting of conditions for the partial derivatives of the error with respect to the coefficients to be equal to zero.

[0288] 5) Iteration: Repeat steps 2)-4) until a certain stopping criterion is met (such as reaching the maximum number of iterations, the coefficient change is less than a certain threshold, etc.).

[0289] Output data: α and β1, β2, ..., β n .

[0290] In the least squares method, the relationship coefficients β1,β2,...,β n The solution of the constant term α can be obtained by solving the following normal equations:

[0291]

[0292] S2, establish a mechanism nonlinear model including the set temperature of air conditioning fresh air, dehumidification load and air conditioning power load, which specifically includes the following steps:

[0293] Calculate Q based on step S1 fresh (t) Fresh air load, Q dehum (t) Dehumidification load, Q ac (t) represents loads such as air conditioning power load, and then further constructs a nonlinear mechanism model to represent the thermodynamic principles and building physics involved in solving the process of indoor temperature change in the building:

[0294]

[0295] Where, T set (t) represents the set indoor temperature; T out (t) represents the external temperature; Q fresh (t) represents the fresh air load; Q dehum (t) represents the dehumidification load; Q ac (t) represents the air conditioning power load; γ is the parameter of air conditioning system efficiency, which reflects the air conditioning power load Q ac(t) for indoor temperature T set (t); δ is the linear influence coefficient of fresh air and dehumidification load on indoor temperature setting; ζ is the reference temperature bias, which reflects the influence of other factors not directly included in the model on indoor temperature. The reference temperature bias is usually initialized to a small value, such as 0 or a random number close to 0, or a more appropriate initial value can be selected based on experience.

[0296] Among them, the parameter γ of the air conditioning system efficiency reflects the air conditioning power load Q ac (t) for indoor temperature T set (t) Parameters of influence. It is usually related to the energy efficiency ratio (EER) or coefficient of performance (COP) of the air conditioning system. In this application, it represents a comprehensive parameter, including multiple factors such as system efficiency, building thermal characteristics, indoor and outdoor temperature difference, etc. It needs to be determined through experimental data or engineering experience, involving measuring the changes in indoor temperature under different conditions (such as different indoor and outdoor temperature differences, different power loads, etc.), and estimating the value of γ through regression analysis and other methods. The calculation model is as follows:

[0297] T set (t) = α + β·T out (t)+γ·Q oc (t)+χ

[0298] In the formula, α and β are constant terms, which represent other parameters and can be estimated through regression analysis. The value of γ is optimized by minimizing the prediction error (such as mean square error). α is mainly used to adjust the baseline level of the model; β is a set of coefficients used to represent other linear factors in the model that may affect the dependent variable. This application has included fresh air load, dehumidification load, air conditioning power load and outdoor temperature as independent variables, but there may be other factors that have not been considered (such as thermal properties of building materials, size and number of windows, indoor heat sources, etc.) that will also affect indoor temperature. These factors that are not explicitly included in the model can be approximately captured by the β coefficient; if other factors need to be considered, the χ term can be added to represent other factors that need to be considered.

[0299] The linear influence coefficient δ of fresh air and dehumidification load on indoor temperature setting also needs to be determined through experimental data or engineering experience. Usually, this involves measuring the change of indoor temperature under different fresh air and dehumidification loads, and estimating the value of δ through regression analysis and other methods. The calculation model is as follows:

[0300] T set (t) = α + β·T out (t)+γ·Q oc (t)+δ·(Q fresh (t)+Q dehum (t))+χ *

[0301] The calculation is the same as above, and the value of δ is optimized by minimizing the prediction error.

[0302] S3, dynamically solves the sliding window with time series data as the training data source to achieve accurate calculation of the average indoor temperature, which specifically includes the following steps:

[0303] Dynamically solve the sliding window with time series data as the training data source to achieve accurate calculation of the average indoor temperature.

[0304] Based on sliding window technology, combined with machine learning methods, dynamic prediction of indoor temperature:

[0305] T avg (t n )=Φ(T set (t n-1 ),T set (t n-2 ),...,T set (t n-k ))

[0306] Where, T avg (t n ) represents the time t n Average indoor temperature; T set (t n-k ) represents the set temperature of the previous k moments; Φ(·) represents a nonlinear function, including but not limited to machine learning models such as neural networks and support vector machines, which are used to capture the nonlinear relationship between temperature set values ​​and predict future indoor temperatures; k represents the size of the sliding window, which determines the scope of historical data considered by the model.

[0307] The first specific embodiment of the present invention for solving the nonlinear function Φ(·):

[0308] This example uses a multi-layer perceptron (MLP) neural network as the nonlinear function Φ(·) to solve. Assume that there is a multi-layer perceptron with two hidden layers, each hidden layer has h neurons, and uses ReLU (Rectified Linear Unit) as the activation function, and the output layer uses a linear activation function (because we want to predict the temperature value, no nonlinear transformation is required). In addition, assume that the input layer has k neurons, corresponding to the k historical set temperature values ​​in the sliding window. The steps are as follows:

[0309] 1) Input layer:

[0310] The input layer directly passes the historical set temperature values ​​in the sliding window to the hidden layer.

[0311] 2) The first hidden layer:

[0312] For the i-th neuron in the first hidden layer, i ranges from 1 to h, its output It can be calculated as:

[0313]

[0314] In the formula, is the weight from the jth neuron in the input layer to the ith neuron in the first hidden layer, is the bias term of the i-th neuron in the first hidden layer.

[0315] 3) Second hidden layer:

[0316] For the i-th neuron in the second hidden layer, i ranges from 1 to h, its output It can be calculated as:

[0317]

[0318]

[0319] In the formula, is the weight from the jth neuron in the first hidden layer to the ith neuron in the second hidden layer, is the bias term of the i-th neuron in the second hidden layer.

[0320] 4) Output layer:

[0321] The output of the output layer is T avg (t n ) (i.e. the predicted average indoor temperature) can be calculated as:

[0322]

[0323] T avg (t n )=z avg

[0324] In the formula, is the weight from the jth neuron in the second hidden layer to the output layer, b out is the bias term of the output layer.

[0325] The weights and biases of the neural network (i.e. b out ) needs to be learned and optimized through training data to minimize the error between the predicted temperature and the actual temperature. In addition, the architecture of the neural network (such as the number of layers, the number of neurons in each layer, the activation function, etc.) can also be adjusted according to the specific task and data set.

[0326] The second specific embodiment of the present invention for solving the nonlinear function Φ(·):

[0327] The example uses a neural network as the nonlinear function Φ(·) to solve the problem and selects a long short-term memory network (LSTM). The steps are as follows:

[0328] Data preparation: Historical set temperature data is organized to construct features and labels. For each time step t n , the first k set temperatures are used as input features, and the current average indoor temperature is used as the output label.

[0329] Training model: The selected deep learning model is trained using the training set, during which the model parameters are optimized to minimize the gap between the actual value and the predicted value.

[0330] Evaluation and prediction: The trained model is evaluated using the validation set and the average indoor temperature at future moments is predicted. This involves calculating the average indoor temperature based on the previous k set temperatures.

[0331] The model is expressed as:

[0332] T avg (t n )=Φ(T set (t n-1 ),T set (t n-2 ),…,T set (t n-k ))

[0333] Where, T avg (t n ) represents the time t n Average indoor temperature; T set (t n-k ) represents the set temperature at the previous k moments; Φ(·) is a neural network model, which in this example is a function composed of LSTM hidden layers, used to extract and learn patterns and regularities in time series data from the input sequence.

[0334] In neural networks, T set (t n-1 ),T set (t n-2 ),…,T set (t n-k ) as input, and T avg (t n ), thereby completing the prediction of the average indoor temperature.

[0335] A system embodiment using the method of the present invention:

[0336] A dynamic building indoor temperature solution system with mechanism data fusion, including a trend simulation module, a mechanism nonlinear module and a temperature solution module;

[0337] Trend simulation module, used to process conditional data and obtain load change prediction results;

[0338] The mechanism nonlinear module is used to calculate the indoor temperature setting data based on the load change prediction results;

[0339] The temperature solving module is used to obtain the indoor temperature of the building based on the indoor temperature setting data and dynamically solve using a sliding window.

[0340] In this embodiment, the load change prediction result is obtained by using the trend simulation module, which includes the following contents:

[0341] Step 21, constructing a regression analysis equation based on environmental conditions and condition coefficients to characterize the load change trend;

[0342] The expression of the regression analysis equation is as follows:

[0343] Q load (t)=α+β1·Condition1(t)+β2·Condition2(t)+...+β n Condition n (t)+∈

[0344] In the formula, Q load (t) represents the load at a specific time, which includes the fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t), α is a constant term; β1,β2,...,β n is the relationship coefficient between load and outdoor temperature, humidity, ... solar radiation intensity; the coefficient β reflects the influence of each condition on the load; Condition1(t), Condition2(t)..., Condition n (t) represent the outdoor temperature, humidity, solar radiation intensity at time t respectively;

[0345] Step 22, considering the dynamic relationship of load over time, construct a time series formula, the calculation formula is as follows:

[0346] Y t =f(Y t-1 ,Y t-2 ,…,Y t-k +e t )

[0347] Where Y t represents the load measurement value at time t; k is the order of historical data considered by the model; f(·) represents the function used to describe the dynamic relationship of the time series; e t represents the error term of the model;

[0348] Step 23, using the least square method to adjust the relationship coefficient and constant term in the regression analysis equation to obtain new relationship coefficient and constant term;

[0349] The least square method is used to estimate the error term in the time series formula to obtain a new error term;

[0350] Step 24, substituting the new relationship coefficient and constant term into the regression analysis equation and solving it to obtain a more accurate load change prediction result under static environmental conditions;

[0351] Substitute the new error term into the time series formula and solve it to obtain more accurate load change prediction results under time dynamic environmental conditions;

[0352] In this embodiment, the indoor temperature setting data is obtained through the mechanism nonlinear module, which includes the following contents:

[0353] Step 31, obtain load change prediction results, which include fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t);

[0354] Step 32: Based on the fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t) Establish a nonlinear mechanism formula to characterize the thermodynamic principles and building physics involved in the process of indoor temperature change in buildings;

[0355] The calculation formula of the nonlinear mechanism formula is as follows:

[0356]

[0357] Where, T set (t) represents the indoor temperature setting data; T out (t) represents the external temperature; Q fresh (t) represents the fresh air load; Q dehum (t) represents the dehumidification load; Q ac (t) represents the air conditioning power load; γ is the parameter of air conditioning system efficiency, which reflects the air conditioning power load Q ac (t) for indoor temperature T set(t); δ is the linear influence coefficient of fresh air and dehumidification load on indoor temperature setting; ζ is the reference temperature bias, which reflects the influence of the remaining factors not directly included in the model on the indoor temperature;

[0358] Step 33, solving the nonlinear mechanism formula to obtain indoor temperature setting data;

[0359] In this embodiment, a temperature solving module is used to obtain the indoor temperature of the building, which includes the following contents:

[0360] Step 41, obtaining indoor temperature setting data;

[0361] Step 42, based on the indoor temperature setting data, combined with the sliding window with time series data as the training data source, construct a dynamic prediction formula for indoor temperature, the calculation formula is as follows:

[0362] T avg (t n )=Φ(T set (t n-1 ),T set (t n-2 ),...,T set (t n-k ))

[0363] Where, T avg (t n ) represents the time t n Average indoor temperature; T set (t n-k ) represents the indoor temperature setting data of the previous k moments; Φ(·) represents a nonlinear function, which is used to capture the nonlinear relationship between temperature setting values ​​and predict future indoor temperatures; k represents the size of the sliding window, which determines the range of historical data considered by the model;

[0364] Step 43, using a multi-layer perceptron MLP neural network as a nonlinear function to solve and obtain the indoor temperature of the building.

[0365] An embodiment of a device applying the method of the present invention:

[0366] An electronic device comprising:

[0367] one or more processors;

[0368] A storage device for storing one or more programs;

[0369] When the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned method for solving the dynamic building indoor temperature by mechanism data fusion.

[0370] A computer medium embodiment using the method of the present invention:

[0371] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned dynamic building indoor temperature solution method using mechanism data fusion.

[0372] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, and computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0373] The present application is described by flowcharts or / and block diagrams of the methods, devices (systems), and computer program products of the embodiments of the present application. It should be understood that each process or / and block in the flowchart or / and block diagram and the combination of the processes or / and blocks in the flowchart or / and block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart or / and block diagram. Figure 1 Process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0374] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 Process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0375] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 Process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0376] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A dynamic building indoor temperature solution method based on mechanism data fusion, characterized by: It includes the following: Obtain data on conditions that affect changes in air conditioning load; Use the pre-built trend simulation model to process the conditional data and obtain the load change prediction results; Through the pre-built mechanism nonlinear model, the indoor temperature setting data is calculated according to the load change prediction results; The pre-built temperature solution model is used, based on the indoor temperature setting data, and a sliding window is used for dynamic solution to obtain the indoor temperature of the building, realizing dynamic building indoor temperature solution based on mechanism data fusion.

2. The method for solving the dynamic indoor temperature of a building by using mechanism data fusion as claimed in claim 1, characterized in that: The method of using the pre-built trend simulation model to process the conditional data and obtain the load change prediction result is as follows: According to environmental conditions and condition coefficients, a regression analysis equation is constructed to characterize the load change trend; The expression of the regression analysis equation is as follows: Q load (t)=α+β1·Condition1(t)+β2·Condition2(t)+...+β n ·Condition n (t)+∈ In the formula, Q load (t) represents the load at a specific time, which includes the fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t), α is a constant term; β1,β2,...,β n is the relationship coefficient between load and outdoor temperature, humidity, ... solar radiation intensity; the coefficient β reflects the influence of each condition on the load; Condition1(t), Condition2(t)..., Condition n (t) represent the outdoor temperature, humidity, solar radiation intensity at time t respectively; The least square method is used to adjust the relationship coefficient and constant term in the regression analysis equation to obtain new relationship coefficient and constant term; Substitute the new relationship coefficient and constant term into the regression analysis equation and solve it to obtain a more accurate load change prediction result under static environmental conditions; Alternatively or additionally, a method of processing conditional data using a pre-built trend simulation model to obtain load change prediction results is as follows: According to environmental conditions and condition coefficients, considering the dynamic relationship of load over time, a time series formula is constructed, and its calculation formula is as follows: AND t =f(Y t-1 ,AND t-2 ,…,AND t-k +e t ) Where Y t represents the load measurement value at time t; k is the order of historical data considered by the model; f(·) represents the function used to describe the dynamic relationship of the time series; e t represents the error term of the model; The least square method is used to estimate the error term in the time series formula to obtain a new error term; Substitute the new error term into the time series formula and solve it to obtain more accurate load change prediction results under time dynamic environmental conditions.

3. The method for solving the dynamic indoor temperature of a building by using mechanism data fusion as claimed in claim 2, characterized in that: The coefficient of relationship is determined in regression analysis by minimizing the sum of squared errors (SSE) between the predicted load value and the actual load value, which is done using the least squares method, including the following: Get a load data set, which includes multiple load time points, each of which has a corresponding load amount Q load (t) and a series of environmental conditions Condition1(t), Condition2(t)..., Condition n (t); And the relationship coefficients β1,β2,...,β n Initialize with constant term α to obtain the initial values ​​of the relationship coefficient and constant term; Substitute the current relationship coefficient value and constant term into the regression analysis equation to calculate the predicted load at each load time point. The corresponding calculation formula is as follows: (Q load (t)predicted=α+β1·Condition1(t)+β2·Condition2(t)+...+β n ·Condition n (t)); Calculate the error between the predicted load value and the actual load value at each load time point to obtain the error data. The calculation formula is as follows: e(t)=Q load (t) observed -Q load (t) predicted ; According to the error data, a load coefficient linear equation group is established. The linear equation group is composed of the condition that the partial derivative of the error data to the relationship coefficient is equal to zero; the expression of the load coefficient linear equation group is as follows: Based on the principle of minimizing the sum of squared errors (SSE), the load factor linear equations are solved to obtain new relationship coefficients and constant terms. Repeat steps 232) to 235 until a certain stopping criterion is met, and output the final relationship coefficient and constant term; the stopping criterion includes reaching the maximum number of iterations or / and the coefficient change is less than a certain threshold; Or / and, the method to construct the time series formula is as follows: Through the autoregressive model or the moving average model or the autoregressive moving average model, based on the data characteristics, a function f(·) is constructed to describe the dynamic relationship of the time series, and the time series formula is obtained; In the autoregressive model, the future load value Y t is the past load value Y t-1 ,Y t-2 ,...,Y t-k A linear combination of t ; The time series formula constructed is as follows: Y t =φ1Y t-1 +φ2Y t-2 +…+φ k Y t-k +e t Where φ1, φ2, ..., φ k is the autoregressive coefficient; In the moving average model, the future load value Y t is the past error term e t-1 ,e t-2 ,...,e t-k The linear combination of ; the time series formula constructed is as follows: Y t =e t +θ1e t-1 +θ2e t-2 +…+θ k e t-k Where θ1,θ2,...,θ k is the moving average coefficient; The autoregressive moving average model combines the characteristics of autoregression and moving average; the time series formula it constructs is as follows: Y t =φ1Y t-1 +φ2Y t-2 +…+φ k Y t-k +e t +θ1e t-1 +θ2e t-2 +…+θ q e t-q Here, k is the order of autoregression and q is the order of moving average.

4. The method for solving the dynamic indoor temperature of a building by mechanism data fusion according to claim 1, characterized in that: The method of calculating the indoor temperature setting data based on the load change prediction results through the pre-built mechanism nonlinear model is as follows: Obtain load change prediction results, including fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t); Based on fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t) Establish a nonlinear mechanism formula to characterize the thermodynamic principles and building physics involved in the process of indoor temperature change in buildings; The calculation formula of the nonlinear mechanism formula is as follows: Where, T set (t) represents the indoor temperature setting data; T out (t) represents the external temperature; Q fresh (t) represents the fresh air load; Q dehum (t) represents the dehumidification load; Q ac (t) represents the air conditioning power load; γ is the parameter of air conditioning system efficiency, which reflects the air conditioning power load Q ac (t) for indoor temperature T set (t); δ is the linear influence coefficient of fresh air and dehumidification load on indoor temperature setting; ζ is the reference temperature bias, which reflects the influence of the remaining factors not directly included in the model on the indoor temperature; The nonlinear mechanism formula is solved to obtain the indoor temperature setting data.

5. The method for solving the dynamic indoor temperature of a building by fusion of mechanism data as claimed in claim 4, characterized in that: The parameter γ of the air conditioning system efficiency reflects the air conditioning power load Q ac (t) Set the indoor temperature data T set (t) Parameters of influence, which are related to air conditioning system efficiency, building thermal characteristics, and indoor and outdoor temperature difference; The value of parameter γ is calculated by regression analysis and optimized by minimizing the prediction error. The calculation formula is as follows: T set (t)=α+β·T out (t)+γ·Q oc (t)+x In the formula, α and β are constant terms, representing the relevant parameters, which are estimated by regression analysis, and α is used to adjust the baseline level of the model; The linear influence coefficient δ of fresh air and dehumidification load on indoor temperature setting data is calculated by regression analysis method and optimized by minimizing the prediction error. The calculation formula is as follows: T set (t)=α+β·T out (t)+γ·Q oc (t)+δ·(Q fresh (t)+Q dehum (t))+χ * 。 6. The method for solving the dynamic indoor temperature of a building by mechanism data fusion according to claim 1, characterized in that: Using the pre-built temperature solution model, based on the indoor temperature setting data, and using the sliding window for dynamic solution, the method for obtaining the indoor temperature of the building is as follows: Get indoor temperature setting data; Based on the indoor temperature setting data, combined with the sliding window with time series data as the training data source, a dynamic prediction formula for indoor temperature is constructed. The calculation formula is as follows: T avg (t n )=Φ(T set (t n-1 ),T set (t n-2 ),...,T set (t n-k )) Where, T avg (t n ) represents the time t n Average indoor temperature; T set (t n-k ) represents the indoor temperature setting data of the previous k moments; Φ(·) represents a nonlinear function, which is used to capture the nonlinear relationship between temperature setting values ​​and predict future indoor temperatures; k represents the size of the sliding window, which determines the range of historical data considered by the model; A multi-layer perceptron MLP neural network is used as a nonlinear function to solve the indoor temperature of the building; the method for solving the nonlinear function is as follows: Build a multilayer perceptron with two hidden layers, each with h neurons and using ReLU as the activation function, and the output layer uses a linear activation function; the input layer has k neurons, corresponding to the k historical set temperature values ​​in the sliding window; The input layer passes the historical set temperature values ​​in the sliding window to the hidden layer; For the i-th neuron in the first hidden layer, i ranges from 1 to h, its output The calculation formula is as follows: In the formula, is the weight from the jth neuron in the input layer to the ith neuron in the first hidden layer, is the bias term of the i-th neuron in the first hidden layer; For the i-th neuron in the second hidden layer, i ranges from 1 to h, its output The calculation formula is as follows: In the formula, is the weight from the jth neuron in the first hidden layer to the ith neuron in the second hidden layer, is the bias term of the i-th neuron in the second hidden layer; The output layer outputs the predicted average indoor temperature T avg (t n ), and the average indoor temperature T avg (t n ) is used as the indoor temperature of the building, and its calculation formula is as follows: T avg (t n )=z avg In the formula, is the weight from the jth neuron in the second hidden layer to the output layer, b out is the bias term of the output layer.

7. A dynamic building indoor temperature solution system with mechanism data fusion, characterized by: Including trend simulation module, mechanism nonlinear module and temperature solution module; Trend simulation module, used to process conditional data and obtain load change prediction results; The mechanism nonlinear module is used to calculate the indoor temperature setting data based on the load change prediction results; The temperature solving module is used to obtain the indoor temperature of the building based on the indoor temperature setting data and dynamically solve using a sliding window.

8. The dynamic building indoor temperature solution system with mechanism data fusion as claimed in claim 7, characterized in that: Trend simulation module is used to process conditional data and obtain load change prediction results. Includes the following: According to environmental conditions and condition coefficients, a regression analysis equation is constructed to characterize the load change trend; The expression of the regression analysis equation is as follows: Q load (t)=α+β1·Condition1(t)+β2·Condition2(t)+...+β n ·Condition n (t)+∈ In the formula, Q load (t) represents the load at a specific time, which includes the fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t), α is a constant term; β1,β2,...,β n is the relationship coefficient between load and outdoor temperature, humidity, ... solar radiation intensity; the coefficient β reflects the influence of each condition on the load; Condition1(t), Condition2(t)..., Condition n (t) represent the outdoor temperature, humidity, solar radiation intensity at time t respectively; Considering the dynamic relationship of load over time, a time series formula is constructed, and its calculation formula is as follows: AND t =f(Y t-1 ,AND t-2 ,…,AND t-k +e t ) Where Y t represents the load measurement value at time t; k is the order of historical data considered by the model; f(·) represents the function used to describe the dynamic relationship of the time series; e t represents the error term of the model; The least square method is used to adjust the relationship coefficient and constant term in the regression analysis equation to obtain new relationship coefficient and constant term; The least square method is used to estimate the error term in the time series formula to obtain a new error term; Substitute the new relationship coefficient and constant term into the regression analysis equation and solve it to obtain a more accurate load change prediction result under static environmental conditions; Substitute the new error term into the time series formula and solve it to obtain more accurate load change prediction results under time dynamic environmental conditions; The load change prediction result under static environmental conditions is coupled with the load change prediction result under dynamic environmental conditions to obtain the final load change prediction result; Or / and, a mechanism nonlinear module is used to calculate the indoor temperature setting data according to the load change prediction results, which includes the following contents: Obtain load change prediction results, including fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t); Based on fresh air load Q fresh (t), dehumidification load Q dehum (t), air conditioning power load Q ac (t) Establish a nonlinear mechanism formula to characterize the thermodynamic principles and building physics involved in the process of indoor temperature change in buildings; The calculation formula of the nonlinear mechanism formula is as follows: Where, T set (t) represents the indoor temperature setting data; T out (t) represents the external temperature; Q fresh (t) represents the fresh air load; Q dehum (t) represents the dehumidification load; Q ac (t) represents the air conditioning power load; γ is the parameter of air conditioning system efficiency, which reflects the air conditioning power load Q ac (t) for indoor temperature T set (t); δ is the linear influence coefficient of fresh air and dehumidification load on indoor temperature setting; ζ is the reference temperature bias, which reflects the influence of remaining factors not directly included in the model on indoor temperature; Solve the nonlinear mechanism formula to obtain the indoor temperature setting data; Or / and, a temperature solving module is used to dynamically solve the indoor temperature of the building based on the indoor temperature setting data and using a sliding window, which includes the following contents: Get indoor temperature setting data; Based on the indoor temperature setting data, combined with the sliding window with time series data as the training data source, a dynamic prediction formula for indoor temperature is constructed. The calculation formula is as follows: T avg (t n )=Φ(T set (t n-1 ),T set (t n-2 ),...,T set (t n-k )) Where, T avg (t n ) represents the time t n Average indoor temperature; T set (t n-k ) represents the indoor temperature setting data of the previous k moments; Φ(·) represents a nonlinear function, which is used to capture the nonlinear relationship between temperature setting values ​​and predict future indoor temperatures; k represents the size of the sliding window, which determines the range of historical data considered by the model; A multi-layer perceptron (MLP) neural network is used as a nonlinear function to solve and obtain the indoor temperature of the building.

9. An electronic device, characterized in that: It includes: one or more processors; A storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement a dynamic building indoor temperature solution method using mechanism data fusion as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the program is executed by a processor, a dynamic building indoor temperature solution method using mechanism data fusion as described in any one of claims 1 to 6 is implemented.