Dynamic energy management method for active cooling window system

Abstract

The invention belongs to the technical field of doors and windows, and particularly relates to an energy dynamic management method of an active cooling window system, which is used for realizing automatic energy dynamic management of the active cooling window system. The method comprises the four steps of establishing a to-be-evaluated matrix and performing data preprocessing, establishing a random forest prediction model and establishing a characteristic value weight matrix, establishing a TOPSIS decision matrix and determining an evaluation score, and calculating energy required by each active cooling window unit based on the evaluation score, so that the energy dynamic management of the active cooling window system is realized. The effects of optimizing the system energy distribution strategy, reducing the total energy consumption of the system, improving the working efficiency of the system and improving the indoor thermal environment comfort of the building are achieved.

Description

Technical Field

[0001] The invention belongs to the technical field of doors and windows, and in particular relates to a method for dynamic energy management of an active cooling window system. Background Art

[0002] At present, in the field of domestic door and window technology, a variety of active cooling window systems have been developed to enhance the cooling and heat insulation performance of windows. Such systems, with their excellent heat insulation performance, show particularly significant energy-saving effects in large-area glass curtain walls, windows and other parts of the building with weak heat insulation performance, or temporary buildings with poor heat insulation performance of the outer protective structure of industrial plants. However, in a building equipped with an active cooling window system, different window units of the building receive different amounts of solar radiation at the same time, so the temperature of each active cooling window unit is also different, and the required active cooling requirements and energy distribution are also different. Therefore, if the total energy is simply distributed evenly to each active cooling window unit, it will definitely be detrimental to the overall energy conservation and environmental protection of the building. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide an energy dynamic management method for an active cooling window system, which can realize automatic acquisition of monitoring data, automatic machine learning based on monitoring data, automatic optimization, and automatic realization of energy dynamic management when controlling an active cooling window unit of the active cooling window system.

[0004] In order to solve the above technical problems, the present invention provides an energy dynamic management method for an active cooling window system, characterized in that the energy dynamic management method for an active cooling window system mainly comprises the following steps:

[0005] Step S1, acquiring monitoring data of the active cooling window unit in real time through an active cooling window unit monitoring device based on the Internet of Things, constructing a matrix to be evaluated based on the monitoring data, and performing data preprocessing on the matrix to be evaluated to obtain a dimensionless matrix;

[0006] Step S2: establishing a random forest prediction model based on the dimensionless matrix, and constructing an eigenvalue weight matrix based on the random forest prediction model;

[0007] Step S3: construct a TOPSIS decision matrix based on the eigenvalue weight matrix, determine the evaluation score, and construct an evaluation score matrix;

[0008] Step S4: based on the evaluation score matrix, determine the energy required for each active cooling window unit to achieve dynamic energy management of the active cooling window system.

[0009] Furthermore, in step S1, the matrix to be evaluated X is constructed based on the monitoring data, and data preprocessing is performed on the matrix to be evaluated X to obtain a dimensionless matrix Z, which is specifically:

[0010] Step S1-1, construct the matrix X to be evaluated:

[0011]

[0012] where x ij represents the value of the index to be evaluated with index i in row and index j in column, i=1,…,n, j=1,…,p, n represents the total number of active cooling window units, i.e. the total number of rows, and p represents the number of indicators to be evaluated, i.e. the total number of columns;

[0013] Step S1-2: perform data preprocessing:

[0014] The data preprocessing step includes missing value checking and z-score standardization. The specific method of missing value checking is to traverse the matrix X to be evaluated, determine the position index of the missing value (NaN), calculate the mean of the non-missing values ​​in the column where the missing value is located, and replace the missing value (NaN) with the mean of the non-missing values ​​in the column where the missing value is located;

[0015] The specific method of the z-score standardization is to perform the z-score standardization process on the matrix to be evaluated X to obtain a dimensionless matrix Z:

[0016] Z=[z ij ] n×p ;

[0017]

[0018] where z ij represents the value of the index to be evaluated with index i in row and index j in column after the z-score normalization process, μ j represents the mean value of the index to be evaluated in the j-th column of the matrix to be evaluated X, σ j It represents the standard deviation of the index value to be evaluated in the column indexed by j in the matrix to be evaluated X.

[0019] Furthermore, in step S2, a random forest prediction model is established based on the dimensionless matrix Z, and a eigenvalue weight matrix W is constructed based on the random forest prediction model, which is specifically:

[0020] Step S2-1, dividing the dimensionless matrix Z into a random forest model training set and a random forest model prediction set according to a certain ratio, wherein the random forest model training set includes: a training set prediction index input matrix and a training set prediction target output matrix, and the random forest model prediction set includes: a prediction set prediction index input matrix and a prediction set prediction target output matrix;

[0021] Step S2-2: Establish the random forest prediction model based on the random forest model training set, set the random forest prediction model parameters and run the random forest algorithm to obtain the eigenvalue weight matrix W:

[0022] W=[w j ] 1×p ;

[0023] W=[w 1 w 2 w 3 … w j … w p ];

[0024] w j represents the eigenvalue weight of the indicator to be evaluated with index j in the dimensionless matrix Z;

[0025] Based on the established random forest prediction model, the data in the prediction set prediction target output matrix is ​​predicted, and the random forest prediction accuracy A is output. RF , set the random forest prediction accuracy threshold A t , where A t ∈[0,1], compare A RF With A t The value is large. If A RF t , then return to step S2-1, if A RF ≥A t , then proceed to the next step.

[0026] Furthermore, the TOPSIS decision matrix is ​​constructed based on the eigenvalue weight matrix W in step S3 to determine the evaluation score k i , construct the evaluation score matrix K, specifically:

[0027] Step S3-1: Perform weighting on the dimensionless matrix to obtain a weighted matrix E = [e ij ] n×p :

[0028] E=[e ij ] n×p =[w j z ij ] n×p ​;

[0029] Step S3-2: Select the maximum value of each column in the weighted matrix to form the optimal solution matrix E + , the minimum value constitutes the worst solution matrix E - :

[0030]

[0031] in represents the maximum value in the j-th column of the weight matrix E, represents the minimum value in the column of the weighted matrix E;

[0032] Step S3-3: Determine the Euclidean distance between the optimal solution and the worst solution in each round

[0033]

[0034] Step S3-4: Calculate the score k of each active cooling window unit i , construct the evaluation score matrix K:

[0035]

[0036] K=[k 1 ;k 2 ;k 3 ;…;k i ;…;k n ].

[0037] Furthermore, the energy required for each active cooling window unit is determined based on the evaluation score matrix K in step S4. i , specifically:

[0038]

[0039] Where p represents the total energy, p i Represents the energy required by the active cooling window unit with index i.

[0040] The benefits of the present invention are as follows: (1) the present invention can realize dynamic, intelligent and precise management of the energy of the active cooling window system through dynamic collection and dynamic analysis of monitoring data, optimize the system energy allocation strategy, reduce the total energy consumption of the system while improving the system working efficiency and enhancing the indoor thermal environment comfort of the building; (2) the present invention is based on a dynamic random forest prediction model and a machine learning algorithm of the TOPSIS decision matrix, and has the characteristics of machine self-learning and self-optimization. Therefore, the longer the present invention is used in actual project applications, the better its optimization performance for the energy management of the active cooling window system. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is the overall flow chart of the present invention;

[0042] Figure 2 This is a data preprocessing flow chart of the present invention;

[0043] Figure 3 A random forest prediction model flow chart is established for the present invention;

[0044] Figure 4 Constructing an evaluation score matrix flow chart for TOPSIS of the present invention;

[0045] Figure 5 This is a graph showing the error variation of the random forest prediction model of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is described in detail below in conjunction with the accompanying drawings and implementation cases. It should be understood that the specific implementation cases described herein are only used to explain the present invention and are not used to limit the invention.

[0047] See also Figure 1 The present invention provides the following technical solution: a method for dynamic energy management of an active cooling window system, characterized in that the method for dynamic energy management of an active cooling window system mainly comprises the following steps:

[0048] Step S1, acquiring monitoring data of the active cooling window unit in real time through an active cooling window unit monitoring device based on the Internet of Things, constructing a matrix X to be evaluated based on the monitoring data, and performing data preprocessing on the matrix X to be evaluated to obtain a dimensionless matrix Z;

[0049] Step S2: establishing a random forest prediction model based on the dimensionless matrix Z, and constructing an eigenvalue weight matrix W based on the random forest prediction model;

[0050] Step S3: Based on the eigenvalue weight matrix W, a TOPSIS decision matrix is ​​constructed to determine the evaluation score k i , construct the evaluation score matrix K;

[0051] Step S4: Based on the evaluation score matrix K, determine the energy p required for each active cooling window unit. i , realizing dynamic energy management of active cooling window system.

[0052] See also Figure 2 In this embodiment, the step S1 constructs the matrix X to be evaluated based on the monitoring data, and performs data preprocessing on the matrix X to be evaluated to obtain a dimensionless matrix Z, which is specifically:

[0053] Step S1-1, construct the matrix X to be evaluated:

[0054]

[0055] where x ij represents the value of the index to be evaluated with index i in row and index j in column, i=1,…,n, j=1,…,p, n represents the total number of active cooling window units, i.e. the total number of rows, and p represents the number of indicators to be evaluated, i.e. the total number of columns;

[0056] Step S1-2: perform data preprocessing:

[0057] The data preprocessing step includes missing value checking and z-score standardization. The specific method of missing value checking is to traverse the matrix X to be evaluated, determine the position index of the missing value (NaN), calculate the mean of the non-missing values ​​in the column where the missing value is located, and replace the missing value (NaN) with the mean of the non-missing values ​​in the column where the missing value is located;

[0058] The specific method of the z-score standardization is to perform the z-score standardization process on the matrix to be evaluated X to obtain a dimensionless matrix Z:

[0059] Z=[z ij ] n×p ;

[0060]

[0061] where z ij represents the value of the index to be evaluated with index i in row and index j in column after the z-score normalization process, μ j represents the mean value of the index to be evaluated in the j-th column of the matrix to be evaluated X, σ j It represents the standard deviation of the index value to be evaluated in the column indexed by j in the matrix to be evaluated X.

[0062] See also Figure 3 , Figure 5 In this embodiment, the random forest prediction model is established based on the dimensionless matrix Z in step S2, and the eigenvalue weight matrix W is constructed based on the random forest prediction model, which is specifically:

[0063] Step S2-1, dividing the dimensionless matrix Z into a random forest model training set and a random forest model prediction set according to a certain ratio, wherein the random forest model training set includes: a training set prediction index input matrix and a training set prediction target output matrix, and the random forest model prediction set includes: a prediction set prediction index input matrix and a prediction set prediction target output matrix;

[0064] Step S2-2: Establish the random forest prediction model based on the random forest model training set, set the random forest prediction model parameters and run the random forest algorithm to obtain the eigenvalue weight matrix W:

[0065] W=[w j ] 1×p ;

[0066] W=[w 1 w 2 w 3 … w j … w p ];

[0067] w j represents the eigenvalue weight of the indicator to be evaluated with index j in the dimensionless matrix Z;

[0068] Based on the established random forest prediction model, the data in the prediction set prediction target output matrix is ​​predicted, and the random forest prediction accuracy A is output. RF , set the random forest prediction accuracy threshold A t , where A t ∈[0, 1], compare A RF With A t The value is large. If A RF <A t , then return to step S2-1, if A RF ≥A t , then proceed to the next step. In this embodiment, the random forest prediction accuracy threshold is set to A t =0.80, see Figure 5 It can be seen that A RF =1-0.16=0.84, A RF >A t , then proceed to the next step.

[0069] See also Figure 4 In this embodiment: the step S3 constructs a TOPSIS decision matrix based on the eigenvalue weight matrix W to determine the evaluation score k i , construct the evaluation score matrix K, specifically:

[0070] Step S3-1: Perform weighted processing on the dimensionless matrix Z to obtain a weighted matrix E:

[0071] E=[e ij ] n×p =[w j z ij ] n×p ;

[0072] Step S3-2: Select the maximum value of each column in the weighted matrix to form the optimal solution matrix E + , the minimum value constitutes the worst solution matrix E - :

[0073]

[0074] in represents the maximum value in the j-th column of the weight matrix E, represents the minimum value in the jth column of the weight matrix E;

[0075] Step S3-3: Determine the Euclidean distance between each active cooling window unit and the optimal solution and the worst solution.

[0076]

[0077] Step S3-4: Calculate the score k of each active cooling window unit i , construct the evaluation score matrix K:

[0078]

[0079] K=[k 1 ;k 2 ;k 3 ;…;k i ;…;k n ].

[0080] See also Figure 1 In this embodiment: the energy required for each active cooling window unit is determined based on the evaluation score matrix K in step S4. i , specifically:

[0081]

[0082] Where p represents the total energy, p i Represents the energy required by the active cooling window unit with index i.

[0083] It should be understood that parts not elaborated in detail in this specification belong to the prior art.

[0084] For those skilled in the art, according to the teachings of the present invention, without departing from the principles and spirit of the present invention, changes, modifications, substitutions and variations made to the implementation methods are still within the protection scope of the present invention.

Claims

1. A method for dynamic energy management of an active cooling window system, characterized in that: The energy dynamic management method of the active cooling window system mainly includes the following steps: Step S1, acquiring monitoring data of the active cooling window unit in real time through an active cooling window unit monitoring device based on the Internet of Things, constructing a matrix X to be evaluated based on the monitoring data, and performing data preprocessing on the matrix X to be evaluated to obtain a dimensionless matrix Z; Step S2: establishing a random forest prediction model based on the dimensionless matrix Z, and constructing an eigenvalue weight matrix W based on the random forest prediction model; Step S3: Based on the eigenvalue weight matrix W, a TOPSIS decision matrix is ​​constructed to determine the evaluation score k i , construct the evaluation score matrix K; Step S4: Based on the evaluation score matrix K, determine the energy p required for each active cooling window unit. i , realizing dynamic energy management of active cooling window system.

2. The method for dynamic energy management of an active cooling window system according to claim 1, characterized in that: The step S1 constructs the matrix X to be evaluated based on the monitoring data, and performs data preprocessing on the matrix X to be evaluated to obtain a dimensionless matrix Z, which is specifically: Step S1-1, construct the matrix X to be evaluated: where x ij represents the value of the index to be evaluated with index i in row and index j in column, i=1,…,n, j=1,…,p, n represents the total number of active cooling window units, i.e. the total number of rows, and p represents the number of indicators to be evaluated, i.e. the total number of columns; Step S1-2: perform data preprocessing: The data preprocessing step includes missing value checking and z-score standardization. The specific method of missing value checking is to traverse the matrix X to be evaluated, determine the position index of the missing value (NaN), calculate the mean of the non-missing values ​​in the column where the missing value is located, and replace the missing value (NaN) with the mean of the non-missing values ​​in the column where the missing value is located; The specific method of the z-score standardization is to perform the z-score standardization process on the matrix to be evaluated X to obtain a dimensionless matrix Z: From=[from ij ] n×p ; where z ij represents the value of the index to be evaluated with index i in row and index j in column after the z-score normalization process, μ j represents the mean value of the index to be evaluated in the j-th column of the matrix to be evaluated X, σ j It represents the standard deviation of the index value to be evaluated in the column indexed by j in the matrix to be evaluated X.

3. The method for dynamic energy management of an active cooling window system according to claim 1, characterized in that: In step S2, a random forest prediction model is established based on the dimensionless matrix Z, and a eigenvalue weight matrix W is constructed based on the random forest prediction model, which is specifically: Step S2-1, dividing the dimensionless matrix Z into a random forest model training set and a random forest model prediction set according to a certain ratio, wherein the random forest model training set includes: a training set prediction index input matrix and a training set prediction target output matrix, and the random forest model prediction set includes: a prediction set prediction index input matrix and a prediction set prediction target output matrix; Step S2-2: Establish the random forest prediction model based on the random forest model training set, set the random forest prediction model parameters and run the random forest algorithm to obtain the eigenvalue weight matrix W: in=[in j ] 1×p ; <h2 style=";text-align:left;direction:ltr">W=[w1 w2 w3 … w<h2 style=";text-align:left;direction:ltr"> j <h2 style=";text-align:left;direction:ltr"> …w<h2 style=";text-align:left;direction:ltr"> p <h2 style=";text-align:left;direction:ltr"> ]; w j represents the eigenvalue weight of the indicator to be evaluated with index j in the dimensionless matrix Z; Based on the established random forest prediction model, the data in the prediction set prediction target output matrix is ​​predicted, and the random forest prediction accuracy A is output. RF , set the random forest prediction accuracy threshold A t , where A t ∈[0,1], compare A RF With A t The value is large. If A RF t , then return to step S2-1, if A RF ≥A t , then proceed to the next step.​ 4. The method for dynamic energy management of an active cooling window system according to claim 1, characterized in that: In step S3, based on the eigenvalue weight matrix W, a TOPSIS decision matrix is ​​constructed to determine the evaluation score k i , construct the evaluation score matrix K, specifically: Step S3-1: Perform weighted processing on the dimensionless matrix Z to obtain a weighted matrix E: E=[e ij ] n×p =[w j z ij ] n×p ; Step S3-2: Select the maximum value of each column in the weighted matrix to form the optimal solution matrix E + , the minimum value constitutes the worst solution matrix E - : in represents the maximum value in the j-th column of the weight matrix E, represents the minimum value in the column of the weighted matrix E; Step S3-3: Determine the Euclidean distance between each active cooling window unit and the optimal solution and the worst solution. Step S3-4: Calculate the score k of each active cooling window unit i , construct the evaluation score matrix K: K=[k1;k2;k3;…;k i ;…;k n ]。 5. The method for dynamic energy management of an active cooling window system according to claim 1, characterized in that: In step S4, based on the evaluation score matrix K, the energy p required for each active cooling window unit is determined. i , specifically: Where p represents the total energy, p i Represents the energy required by the active cooling window unit with index i.

Images