Construction method of constant-temperature energy efficiency test model
The method constructs a constant temperature energy efficiency test model using advanced data processing and Bayesian optimization to dynamically evaluate thermostat performance, addressing the limitations of traditional testing methods and improving energy efficiency assessment accuracy.
Patent Information
- Application Number
- CN202411268237.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-11
- Publication Date
- 2024-12-27
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional energy efficiency testing methods for thermostats lack adaptability to environmental changes and cannot accurately capture the energy efficiency performance under different working conditions, limiting their accuracy and applicability in practical applications.
A method for constructing a constant temperature energy efficiency test model that utilizes advanced data processing technology and optimization algorithms to dynamically evaluate energy efficiency by obtaining and fusing load, environmental, and log data, and optimizing the model using a Bayesian optimization algorithm.
The method provides accurate and reliable energy efficiency evaluation under varying conditions, enhancing the accuracy and reliability of energy utilization efficiency assessments.
Smart Images
Figure CN119203054A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of energy efficiency testing, and in particular to a method for constructing a constant temperature energy efficiency testing model. Background Art
[0002] In the current context of energy consumption, the energy efficiency evaluation of thermostats has become a key link in energy conservation and emission reduction. Traditional energy efficiency testing methods often rely on static test conditions and simplified models, which limits their accuracy and applicability in practical applications. With the development of technology, higher requirements are placed on the energy efficiency testing of thermostats, which requires the ability to dynamically reflect the energy utilization efficiency of the equipment under different working conditions. However, existing test models often lack adaptability to environmental changes and cannot accurately capture the energy efficiency performance of the equipment in actual operation. Therefore, developing a test model that can comprehensively consider multiple influencing factors and dynamically evaluate the energy efficiency of thermostats has important practical significance for improving energy utilization efficiency and reducing energy consumption. It is in this context that the present invention proposes an innovative method for constructing a thermostat energy efficiency test model, which aims to achieve accurate evaluation of the energy efficiency of thermostats through advanced data processing technology and optimization algorithms. Summary of the Invention
[0003] The purpose of the present invention is to provide a method for constructing a constant temperature energy efficiency test model.
[0004] To achieve the above object, the present invention is implemented according to the following technical solutions:
[0005] The present invention comprises the following steps:
[0006] Obtaining load data, environmental data, log data, and power consumption data of the thermostat, and calculating energy efficiency data based on the power consumption data;
[0007] Extracting impact features from the log data, and removing multicollinearity of the impact features to obtain impact factors;
[0008] Obtaining a loss factor according to the correlation between the load data and the environmental data and the energy efficiency data, and fusing the impact factor and the loss factor to obtain a dynamic impact factor;
[0009] A constant temperature energy efficiency test model is established based on the dynamic influencing factors, the constant temperature energy efficiency test model is optimized by using a Bayesian optimization algorithm, and the data to be tested is input into the constant temperature energy efficiency test model to output the test results.
[0010] Furthermore, a method for obtaining load data, environmental data, log data, and power consumption data of a thermostat and calculating energy efficiency data based on the power consumption data includes:
[0011] The load data is the mass and volume of the load in the thermostat. A resistive heating element with adjustable input power is placed at the center of the thermostat. After the thermostat temperature stabilizes at the set temperature, the input power of the resistive heating element is adjusted to reduce the thermostat's power consumption to zero and stabilize the temperature at the set temperature to obtain the baseline power consumption. Environmental data includes ambient temperature, ambient humidity, and atmospheric pressure. Log data is the thermostat's temperature data, and power consumption data is the baseline power consumption and the thermostat's power consumption.
[0012] The log data is sliced equidistantly according to the time sequence, and the energy efficiency data of each segment in the log data is calculated by the baseline power consumption and power consumption.
[0013] Furthermore, a method for extracting influencing features from the log data and removing multicollinearity of the influencing features to obtain influencing factors includes:
[0014] The temperature change rate, fluctuation frequency, standard deviation and relative deviation from the set temperature of the log data are extracted as influencing features.
[0015] Using the influencing features as the feature matrix and the power consumption as the target variable, a regression model is established. The loss function of the regression model is:
[0016]
[0017] Where q is the parameter vector of the regression model, n is the number of samples in the feature matrix, V is the feature matrix, p is the target variable, λ1, λ2 and λ3 are regularization parameters, ρ is the smoothing coefficient, ρ∈[0,1], F represents the norm,
[0018] Optimize the loss function:
[0019]
[0020] Where sign is the sign function, and the update formula of the parameter vector is:
[0021]
[0022] where q t+1 is the parameter vector of the t+1th iteration, q t represents the parameter vector of the tth iteration, μ represents the learning rate, and its value range is [0.001, 0.1]. The parameter vector is selected according to the optimization loss function until the preset maximum number of iterations is reached.
[0023] The parameter vector corresponding to the influencing feature is taken as the influencing factor.
[0024] Furthermore, the method for obtaining the loss factor according to the correlation between the load data and the environmental data and the energy efficiency data includes:
[0025] Using set temperature, environmental data, load data and baseline power consumption as data sets, a prediction model is established.
[0026] The loss function of the prediction model is:
[0027]
[0028] Where Y is the predicted value of the baseline power consumption, is the input feature, is the set temperature, environmental data and load data, y is the target value, is the benchmark power consumption in the data set, ε is the allowable error, and the objective function of the prediction model is:
[0029]
[0030] Where ω is the weight vector, b is the bias, Φ(x) is the nonlinear mapping function, ω T Φ(x i )+b represents the separating hyperplane, ω T represents the transpose of ω, x i is the i-th input feature, is the number of input features, and x i The corresponding upper and lower bounds of the slack variables, y i is the i-th target value, C is the penalty factor, st represents the constraint condition, and the Lagrangian function is introduced to convert the objective function into a dual form:
[0031]
[0032] where α and are the Lagrange multipliers for the upper and lower boundaries, α i and x i The corresponding Lagrange multipliers for the upper and lower boundaries are, is the radial basis kernel function, x j is the jth input feature,
[0033] The weights and biases are obtained according to the dual form of the objective function:
[0034]
[0035] Where N nsv is the number of support vectors, α j and They are The corresponding Lagrange multipliers for the upper and lower boundaries are,
[0036] The decision function of the obtained prediction model is:
[0037]
[0038] Calculate the relevance of input features to energy efficiency data:
[0039]
[0040] in and z2 are the indirect impact weights of input features on energy efficiency data and the direct impact weights of input features on energy efficiency data, respectively. is the symbol of partial derivative, For the Input features and the covariance of energy efficiency data v, c(v) is the standard deviation of v,
[0041] The input features with a correlation degree less than a preset correlation threshold are removed, and the input features with a correlation degree greater than the preset correlation threshold are output as loss factors.
[0042] Furthermore, the method of fusing the impact factor and the loss factor to obtain a dynamic impact factor includes:
[0043] Dynamic influencing factors affecting characteristics:
[0044]
[0045] Among them D i is the dynamic impact factor of the i-th impact feature, q i is the impact factor of the i-th influencing feature, is the jth loss factor, is the decision function of the prediction model, is the number of types of loss factors, g(u j ,v) is the jth influencing feature u j and the covariance of energy efficiency data v, c(u j ) is u j The standard deviation of , c(v) is the standard deviation of v.
[0046] Furthermore, a method for establishing a constant temperature energy efficiency test model based on the dynamic impact factor and the energy efficiency data includes:
[0047] Through random sampling, the influencing characteristics, dynamic influencing factors and energy efficiency data are divided into training sets and test sets. The random forest algorithm is used to fit the prediction model of influencing characteristics and dynamic influencing factors for energy efficiency data through the training set, and a constant temperature energy efficiency test model is established. The hyperparameters of the constant temperature energy efficiency test model are optimized according to the test set.
[0048] Furthermore, the method of optimizing the constant temperature energy efficiency test model using a Bayesian optimization algorithm includes:
[0049] The Gaussian process used in Bayesian optimization is:
[0050]
[0051] in is the Gaussian process function used in Bayesian optimization, gp is the probability distribution of the Gaussian process, is the kernel function, and For different hyperparameter search spaces, the Gaussian process has a mean of 0, and the kernel function uses the Mattern kernel function:
[0052]
[0053] Among them, σ 2 is the variance parameter of the kernel function, w is the smoothing parameter of the Mattern kernel, and r is and The Euclidean distance between them, l is the length scale parameter of the kernel function, Γ(·) is the gamma function, K w (·) is the Bessel function,
[0054] Collection function:
[0055]
[0056] in is the mean of the Gaussian process, is the covariance function of the Gaussian process, θ is the hyperparameter of the constant temperature energy efficiency test model, β is the coefficient used to control the acquisition strategy, and the acquisition function is used to determine the next hyperparameter point to be evaluated.
[0057] The update rule for the prediction distribution is:
[0058]
[0059] Among them D 1:t is a set of data points from time point 1 to t, t is the current time point, f 1:t D 1:t The corresponding Gaussian process function Given a new input at time t The mean value predicted at time t is the expected value predicted using the Gaussian process, and K is the number of data points from time 1 to t. The kernel matrix, K -1 is the inverse matrix of K, where k is the kernel vector of the data point at time t+1;
[0060] Iterative optimization is used to reduce the mean square error of the predicted distribution and energy efficiency data.
[0061] A system for constructing a constant temperature energy efficiency test model, comprising:
[0062] The acquisition module obtains the load data, environmental data, log data and power consumption data of the thermostat;
[0063] a calculation module, calculating energy efficiency data based on the power consumption data, extracting influencing features from the log data, removing multicollinearity of the influencing features to obtain influencing factors, obtaining loss factors based on the correlation between the load data and the environmental data and the energy efficiency data, fusing the influencing factors and the loss factors to obtain a dynamic influencing factor, and establishing a constant-temperature energy efficiency test model based on the dynamic influencing factors;
[0064] An optimization module, which optimizes the constant temperature energy efficiency test model using a Bayesian optimization algorithm;
[0065] The output module inputs the test data into the constant temperature energy efficiency test model and outputs the test results.
[0066] The beneficial effects of the present invention are:
[0067] The proposed constant temperature energy efficiency test model construction method can dynamically evaluate the energy efficiency performance of a constant temperature device under different environmental conditions, improving the accuracy and reliability of the test. By comprehensively considering load, environmental, and log data, this method can more comprehensively reflect the actual energy utilization efficiency of the device, providing strong technical support for energy conservation and emission reduction. BRIEF DESCRIPTION OF THE DRAWINGS
[0068] Figure 1 The present invention is a flowchart of the steps of a method for constructing a constant temperature energy efficiency test model. DETAILED DESCRIPTION
[0069] The present invention will be further described below through specific examples. The illustrative examples and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.
[0070] The present invention provides a method for constructing a constant temperature energy efficiency test model, comprising the following steps:
[0071] like Figure 1 As shown, in this embodiment, the following steps are included:
[0072] Obtaining load data, environmental data, log data, and power consumption data of the thermostat, and calculating energy efficiency data based on the power consumption data;
[0073] In the actual evaluation, a resistive heating element with adjustable input power is placed in the center of the test chamber. After the test chamber reaches the set temperature, it is allowed to run stably for 2 hours. The power consumption and indicated temperature of the test chamber are monitored. The resistive heating element is connected and its input power is adjusted until the power consumption of the test chamber reaches zero and the indicated temperature stabilizes within ±0.5°C of the set temperature. The input power value of the resistive heating element is tested to obtain the baseline power consumption. The ratio of the baseline power consumption to the power consumption of the constant temperature device is used as the energy efficiency data. The slicing time of each segment of the log data is set.
[0074] Extracting impact features from the log data, and removing multicollinearity of the impact features to obtain impact factors;
[0075] In the actual evaluation, we conducted statistical analysis on the log data and extracted the temperature change rate, fluctuation frequency, standard deviation, and relative deviation from the set temperature for each segment of the log data as influencing features. We used the influencing features to fit the power consumption, and removed the cross-influence within the influencing features to obtain the influencing factors of the influencing features on power consumption.
[0076] Obtaining a loss factor according to the correlation between the load data and the environmental data and the energy efficiency data, and fusing the impact factor and the loss factor to obtain a dynamic impact factor;
[0077] In the actual evaluation, the set temperature, load data, and environmental data are used as input features to fit the baseline power consumption. The correlation between the predicted value of the baseline power consumption and the partial derivative of the input features and the input features and energy efficiency data is used to obtain the correlation degree. The input features are screened based on the correlation degree to obtain the loss factor. The loss factor and the correlation degree between the influencing features and energy efficiency data are combined to obtain the dynamic impact factor of the influencing features.
[0078] Establishing a constant temperature energy efficiency test model based on the dynamic influencing factors, optimizing the constant temperature energy efficiency test model using a Bayesian optimization algorithm, inputting the test data into the constant temperature energy efficiency test model to output the test results;
[0079] In the actual evaluation, in the Bayesian optimization algorithm, as the time points increase, more data points are continuously collected, the estimate of the objective function is updated through the Gaussian process, and the next hyperparameter point to be evaluated is selected. The changes in the predicted distribution before and after optimization are compared. The mean square error between the optimized predicted distribution and the actual energy efficiency data is reduced, indicating that the accuracy and stability of the constant temperature energy efficiency test model have been improved.
[0080] In this embodiment, the method of obtaining load data, environmental data, log data, and power consumption data of a thermostat and calculating energy efficiency data based on the power consumption data includes:
[0081] The load data is the mass and volume of the load in the thermostat. A resistive heating element with adjustable input power is placed at the center of the thermostat. After the thermostat temperature stabilizes at the set temperature, the input power of the resistive heating element is adjusted to reduce the thermostat's power consumption to zero and stabilize the temperature at the set temperature to obtain the baseline power consumption. Environmental data includes ambient temperature, ambient humidity, and atmospheric pressure. Log data is the thermostat's temperature data, and power consumption data is the baseline power consumption and the thermostat's power consumption.
[0082] The log data is sliced equidistantly according to the time sequence, and the energy efficiency data of each segment in the log data is calculated by the baseline power consumption and power consumption.
[0083] In this embodiment, the method of extracting influencing features from the log data and removing multicollinearity of the influencing features to obtain influencing factors includes:
[0084] The temperature change rate, fluctuation frequency, standard deviation and relative deviation from the set temperature of the log data are extracted as influencing features.
[0085] Using the influencing features as the feature matrix and the power consumption as the target variable, a regression model is established. The loss function of the regression model is:
[0086]
[0087] Where q is the parameter vector of the regression model, n is the number of samples in the feature matrix, V is the feature matrix, p is the target variable, λ1, λ2 and λ3 are regularization parameters, ρ is the smoothing coefficient, ρ∈[0,1], F represents the norm,
[0088] Optimize the loss function:
[0089]
[0090] Where sign is the sign function, and the update formula of the parameter vector is:
[0091]
[0092] where q t+1 is the parameter vector of the t+1th iteration, q t represents the parameter vector of the tth iteration, μ represents the learning rate, and its value range is [0.001, 0.1]. The parameter vector is selected according to the optimization loss function until the preset maximum number of iterations is reached.
[0093] The parameter vector corresponding to the influencing feature is taken as the influencing factor.
[0094] In this embodiment, the method for obtaining the loss factor according to the correlation between the load data and the environmental data and the energy efficiency data includes:
[0095] Using set temperature, environmental data, load data and baseline power consumption as data sets, a prediction model is established.
[0096] The loss function of the prediction model is:
[0097]
[0098] Where Y is the predicted value of the baseline power consumption, is the input feature, is the set temperature, environmental data and load data, y is the target value, is the benchmark power consumption in the data set, ε is the allowable error, and the objective function of the prediction model is:
[0099]
[0100] Where ω is the weight vector, b is the bias, Φ(x) is the nonlinear mapping function, ω T Φ(x i )+b represents the separating hyperplane, ω T represents the transpose of ω, x i is the i-th input feature, is the number of input features, and x i The corresponding upper and lower bounds of the slack variables, y i is the i-th target value, C is the penalty factor, st represents the constraint condition, and the Lagrangian function is introduced to convert the objective function into a dual form:
[0101]
[0102]
[0103] where α and are the Lagrange multipliers for the upper and lower boundaries, α i and x i The corresponding Lagrange multipliers for the upper and lower boundaries are, is the radial basis kernel function, x j is the jth input feature,
[0104] The weights and biases are obtained according to the dual form of the objective function:
[0105]
[0106] Where N nsv is the number of support vectors, α j and They are The corresponding Lagrange multipliers for the upper and lower boundaries are,
[0107] The decision function of the obtained prediction model is:
[0108]
[0109] Calculate the relevance of input features to energy efficiency data:
[0110]
[0111] in z1 and z2 are the indirect influence weights of input features on energy efficiency data and the direct influence weights of input features on energy efficiency data, with values of 0.5 and 0.5 respectively. is the symbol of partial derivative, For the Input features and the covariance of energy efficiency data v, c(v) is the standard deviation of v,
[0112] The input features with a correlation degree less than a preset correlation degree threshold are removed. The preset correlation degree threshold is 0.47, and the input features with a correlation degree greater than the preset correlation degree threshold are output as loss factors.
[0113] In this embodiment, the method of fusing the impact factor and the loss factor to obtain a dynamic impact factor includes:
[0114] Dynamic influencing factors affecting characteristics:
[0115]
[0116] Among them D i is the dynamic impact factor of the i-th impact feature, q i is the impact factor of the i-th influencing feature, is the jth loss factor, is the decision function of the prediction model, is the number of types of loss factors, g(u j ,v) is the jth influencing feature u j and the covariance of energy efficiency data v, c(u j ) is u j The standard deviation of , c(v) is the standard deviation of v.
[0117] In this embodiment, the method for establishing a constant temperature energy efficiency test model based on the dynamic influencing factor and the energy efficiency data includes:
[0118] Through random sampling, the influencing characteristics, dynamic influencing factors and energy efficiency data are divided into training sets and test sets. The random forest algorithm is used to fit the prediction model of influencing characteristics and dynamic influencing factors for energy efficiency data through the training set, and a constant temperature energy efficiency test model is established. The hyperparameters of the constant temperature energy efficiency test model are optimized according to the test set.
[0119] In this embodiment, the method of optimizing the constant temperature energy efficiency test model using the Bayesian optimization algorithm includes:
[0120] The Gaussian process used in Bayesian optimization is:
[0121]
[0122] in is the Gaussian process function used in Bayesian optimization, gp is the probability distribution of the Gaussian process, is the kernel function, and For different hyperparameter search spaces, the Gaussian process has a mean of 0, and the kernel function uses the Mattern kernel function:
[0123]
[0124] Among them, σ 2 is the variance parameter of the kernel function, w is the smoothing parameter of the Mattern kernel, and r is and The Euclidean distance between them, l is the length scale parameter of the kernel function, Γ(·) is the gamma function, K w (·) is the Bessel function,
[0125] Collection function:
[0126]
[0127] in is the mean of the Gaussian process, is the covariance function of the Gaussian process, θ is the hyperparameter of the constant temperature energy efficiency test model, β is the coefficient used to control the acquisition strategy, and the acquisition function is used to determine the next hyperparameter point to be evaluated.
[0128] The update rule for the prediction distribution is:
[0129]
[0130] Among them D 1:t is a set of data points from time point 1 to t, t is the current time point, f 1:t D 1:t The corresponding Gaussian process function Given a new input at time t The mean value predicted at time t is the expected value predicted using the Gaussian process, and K is the number of data points from time 1 to t. The kernel matrix, K -1 is the inverse matrix of K, where k is the kernel vector of the data point at time t+1;
[0131] Iterative optimization is used to reduce the mean square error of the predicted distribution and energy efficiency data.
[0132] A system for constructing a constant temperature energy efficiency test model, comprising:
[0133] The acquisition module obtains the load data, environmental data, log data and power consumption data of the thermostat;
[0134] a calculation module, calculating energy efficiency data based on the power consumption data, extracting influencing features from the log data, removing multicollinearity of the influencing features to obtain influencing factors, obtaining loss factors based on the correlation between the load data and the environmental data and the energy efficiency data, fusing the influencing factors and the loss factors to obtain a dynamic influencing factor, and establishing a constant-temperature energy efficiency test model based on the dynamic influencing factors;
[0135] An optimization module, which optimizes the constant temperature energy efficiency test model using a Bayesian optimization algorithm;
[0136] The output module inputs the test data into the constant temperature energy efficiency test model and outputs the test results.
[0137] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A method for constructing a constant temperature energy efficiency test model, characterized in that: The following steps are involved: Obtaining load data, environmental data, log data, and power consumption data of the thermostat, and calculating energy efficiency data based on the power consumption data; Extracting influencing features from the log data, and removing multicollinearity of the influencing features to obtain influencing factors; Obtaining a loss factor according to the correlation between the load data and the environmental data and the energy efficiency data, and fusing the impact factor and the loss factor to obtain a dynamic impact factor; A constant temperature energy efficiency test model is established based on the dynamic influencing factors, the constant temperature energy efficiency test model is optimized by using a Bayesian optimization algorithm, and the data to be tested is input into the constant temperature energy efficiency test model to output the test results.
2. The method for constructing a constant temperature energy efficiency test model according to claim 1, characterized in that: The method of obtaining load data, environmental data, log data and power consumption data of a thermostat and calculating energy efficiency data according to the power consumption data includes: The load data is the mass and volume of the load in the thermostat. A resistive electric heating element with adjustable input power is placed at the center of the thermostat. After the temperature of the thermostat is stabilized at the set temperature, the input power of the resistive electric heating element is adjusted to make the power consumption of the thermostat zero and the temperature stabilized at the set temperature to obtain the benchmark power consumption. The environmental data includes the ambient temperature, ambient humidity and atmospheric pressure. The log data is the temperature data of the thermostat. The power consumption data is the benchmark power consumption and the power consumption of the thermostat. The log data is sliced equidistantly according to the time sequence, and the energy efficiency data of each segment in the log data is calculated by the baseline power consumption and power consumption.
3. The method for constructing a constant temperature energy efficiency test model according to claim 1, characterized in that: The method of extracting influencing features from the log data and removing the multicollinearity of the influencing features to obtain influencing factors includes: The temperature change rate, fluctuation frequency, standard deviation and relative deviation from the set temperature of the log data are extracted as influencing features. The influencing features are used as the feature matrix, the power consumption is used as the target variable, and a regression model is established. The loss function of the regression model is: Where q is the parameter vector of the regression model, n is the number of samples in the feature matrix, V is the feature matrix, p is the target variable, λ1, λ2 and λ3 are regularization parameters, ρ is the smoothing coefficient, ρ∈[0,1], F represents the norm, Optimize the loss function: Where sign is the sign function, and the update formula of the parameter vector is: where q t+1 is the parameter vector for the t+1th iteration, q t represents the parameter vector of the tth iteration, μ represents the learning rate, and its value range is [0.001, 0.1]. The parameter vector is selected according to the optimization loss function until the preset maximum number of iterations is reached. The parameter vector corresponding to the influencing feature is taken as the influencing factor.
4. The method for constructing a constant temperature energy efficiency test model according to claim 1, characterized in that: The method for obtaining a loss factor according to the correlation between the load data and the environmental data and the energy efficiency data includes: Using set temperature, environmental data, load data and baseline power consumption as data sets, a prediction model is established. The loss function of the prediction model is: Where Y is the predicted value of the baseline power consumption, is the input feature, is the set temperature, environmental data and load data, y is the target value, is the benchmark power consumption in the data set, ε is the allowable error, and the objective function of the prediction model is: In the formula, ω is the weight vector, b is the bias, Φ(x) is the nonlinear mapping function, ω T Φ(x i )+b represents the separating hyperplane, ω T represents the transpose of ω, x i is the i-th input feature, is the number of input features, and x i The corresponding upper and lower bounds of the slack variables, y i is the i-th target value, C is the penalty factor, st represents the constraint condition, and the Lagrangian function is introduced to convert the objective function into a dual form: where α and are the Lagrange multipliers for the upper and lower boundaries, α i and x i The corresponding Lagrange multipliers for the upper and lower boundaries are, is the radial basis kernel function, x j is the jth input feature, The weights and biases are obtained according to the dual objective function: Where N nsv is the number of support vectors, α j and They are The corresponding Lagrange multipliers for the upper and lower boundaries are, The decision function of the obtained prediction model is: Calculate the relevance of input features to energy efficiency data: in z1 and z2 are the indirect impact weights of the input features on the energy efficiency data and the direct impact weights of the input features on the energy efficiency data, respectively. is the symbol of partial derivative, For the Input features and the covariance of energy efficiency data v, c(v) is the standard deviation of v, The input features whose correlation is less than the preset correlation threshold are removed, and the input features whose correlation is greater than the preset correlation threshold are output as loss factors.
5. The method for constructing a constant temperature energy efficiency test model according to claim 1, characterized in that: The method of fusing the impact factor and the loss factor to obtain a dynamic impact factor comprises: Dynamic influencing factors affecting characteristics: Where D i is the dynamic impact factor of the i-th influencing feature, q i is the impact factor of the i-th influencing feature, is the jth loss factor, is the number of types of loss factors, g(u j ,v) is the jth influencing feature u j and the covariance of energy efficiency data v, c(u j ) is u j The standard deviation of , c(v) is the standard deviation of v.
6. The method for constructing a constant temperature energy efficiency test model according to claim 1, characterized in that: The method for establishing a constant temperature energy efficiency test model based on the dynamic influencing factor and the energy efficiency data includes: Through random sampling, the influencing characteristics, dynamic influencing factors and energy efficiency data are divided into training sets and test sets. The random forest algorithm is used to fit the prediction model of influencing characteristics and dynamic influencing factors for energy efficiency data through the training set, and a constant temperature energy efficiency test model is established. The hyperparameters of the constant temperature energy efficiency test model are optimized according to the test set.
7. The method for constructing a constant temperature energy efficiency test model according to claim 1, characterized in that: The method of optimizing the constant temperature energy efficiency test model by using a Bayesian optimization algorithm comprises: The Gaussian process used in Bayesian optimization is: in is the Gaussian process function used in Bayesian optimization, gp is the probability distribution of the Gaussian process, is the kernel function, and For different hyperparameter search spaces, the Gaussian process has a mean of 0, and the kernel function uses the Mattern kernel function: Among them, σ 2 is the variance parameter of the kernel function, w is the smoothing parameter of the Mattern kernel, and r is and The Euclidean distance between them, l is the length scale parameter of the kernel function, Γ(·) is the gamma function, K w (·) is the Bessel function, Collection function: in is the mean of the Gaussian process, is the covariance function of the Gaussian process, θ is the hyperparameter of the constant temperature energy efficiency test model, β is the coefficient used to control the acquisition strategy, and the acquisition function is used to determine the next hyperparameter point to be evaluated. The update rule for the prediction distribution is: Where D 1:t is a set of data points from time point 1 to t, t is the current time point, f 1:t D 1:t The corresponding Gaussian process function Given a new input at time t is the mean value predicted at time t, is the expected value predicted using the Gaussian process, and K is the t data points from time point 1 to t. The kernel matrix, K -1 is the inverse matrix of K, k is the kernel vector of the data point at time t+1; Iterative optimization to reduce the mean square error of the predicted distribution and energy efficiency data.
8. A system for constructing a constant temperature energy efficiency test model, characterized in that: include: The acquisition module obtains the load data, environmental data, log data and power consumption data of the thermostat; a calculation module, calculating energy efficiency data according to the power consumption data, extracting influencing features from the log data, removing multicollinearity of the influencing features to obtain influencing factors, obtaining loss factors according to the correlation between the load data and the environmental data and the energy efficiency data, fusing the influencing factors and the loss factors to obtain dynamic influencing factors, and establishing a constant temperature energy efficiency test model based on the dynamic influencing factors; An optimization module, which uses a Bayesian optimization algorithm to optimize the constant temperature energy efficiency test model; The output module inputs the data to be tested into the constant temperature energy efficiency test model to output the test results.