CN111861002A - Building cold and hot load prediction method based on data-driven Gaussian learning technology - Google Patents

Building cold and hot load prediction method based on data-driven Gaussian learning technology Download PDF

Info

Publication number
CN111861002A
CN111861002A CN202010711865.4A CN202010711865A CN111861002A CN 111861002 A CN111861002 A CN 111861002A CN 202010711865 A CN202010711865 A CN 202010711865A CN 111861002 A CN111861002 A CN 111861002A
Authority
CN
China
Prior art keywords
data
cold
gaussian
heat load
building
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202010711865.4A
Other languages
Chinese (zh)
Inventor
冯波
艾春美
张翼
董昕昕
李实�
孙立
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Minghua Power Technology Co ltd
Original Assignee
Shanghai Minghua Power Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shanghai Minghua Power Technology Co ltd filed Critical Shanghai Minghua Power Technology Co ltd
Priority to CN202010711865.4A priority Critical patent/CN111861002A/en
Publication of CN111861002A publication Critical patent/CN111861002A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/08—Construction

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Mathematical Physics (AREA)
  • Economics (AREA)
  • Software Systems (AREA)
  • Human Resources & Organizations (AREA)
  • Data Mining & Analysis (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • General Business, Economics & Management (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Primary Health Care (AREA)
  • Algebra (AREA)
  • Databases & Information Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to a building cold and hot load prediction method based on a data-driven Gaussian learning technology, which comprises the following steps of: s1, acquiring a data set of the building cold and heat load and a plurality of characteristic data sets influencing the cold and heat load; s2, processing the data of the training data set by using a principal component analysis method, and extracting a set number of principal components; s3, constructing a Gaussian regression process prediction model of the building cold and heat load, training the constructed Gaussian regression process prediction model based on training set data so as to achieve the purpose of optimizing the model, and giving a prediction estimation interval in a final prediction model result; and S4, predicting the cold and heat load of the building and giving a prediction estimation interval based on the data set after principal component analysis and the optimized Gaussian process model. Compared with the prior art, the method has the advantages of being capable of achieving accurate prediction and the like.

Description

Building cold and hot load prediction method based on data-driven Gaussian learning technology
Technical Field
The invention relates to the field of building energy conservation, in particular to a method for predicting cold and hot loads of a building based on a data-driven Gaussian learning technology.
Background
The building energy consumption is one of the major issues in the field of building energy conservation, and the scientific analysis and reasonable prediction of the building energy consumption can effectively implement the development concept of building energy conservation. The cold and heat load of the building occupies a large part of the building load, and the accurate prediction of the cold and heat load of the building has great significance for energy consumption regulation and control and implementation of an energy-saving scheme.
The building cold and heat load is influenced by a plurality of factors and is in a nonlinear relation, the existing prediction technology generally adopts algorithm models such as a neural network or a support vector machine, and the like, and the models are used for predicting the problems of complex realization, large calculation amount, low prediction precision and the like.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a building cold and heat load prediction method based on a data-driven Gaussian learning technology, which can realize accurate prediction.
The purpose of the invention can be realized by the following technical scheme:
a building cold and heat load prediction method based on a data-driven Gaussian learning technology comprises the following steps:
s1, acquiring a data set of the building cold and heat load and a plurality of characteristic data sets influencing the cold and heat load;
s2, processing the data of the training data set by using a principal component analysis method, and extracting a set number of principal components;
s3, constructing a Gaussian regression process prediction model of the building cold and heat load, training the constructed Gaussian regression process prediction model based on training set data so as to achieve the purpose of optimizing the model, and giving a prediction estimation interval in a final prediction model result;
and S4, predicting the cold and heat load of the building and giving a prediction estimation interval based on the data set after principal component analysis and the optimized Gaussian process model.
Preferably, the data set of the building cold and heat load in step S1 is: 768 different building cold and heat load data sets were simulated.
Preferably, the plurality of characteristic data sets influencing the cold and heat loads in step S1 include eight kinds, namely, relative compactness, surface area, wall area, roof area, overall height, orientation, glass area and glass area distribution, and the eight kinds of influence characteristic data are expressed as input variables xiWhere i is 1,2, 8, and the thermal and cold loads are respectively represented as y1And y2。
Preferably, the step S2 specifically includes the following steps:
s21: correspondingly processing the acquired feature data to eliminate dimension inconsistency among the features;
s22: writing an input variable matrix into a column matrix form, and calculating an average characteristic value;
s23: calculating the difference d between the characteristic value of each sample and the average characteristic valuei；
S24: constructing a covariance matrix C, simultaneously calculating an eigenvalue and an eigenvector of the matrix C, and constructing an eigenvector space;
s25: after the characteristic vector space is constructed, the characteristic values and the characteristic vectors are respectively sorted and screened in the characteristic space, so that a new input characteristic data set after principal component analysis is obtained.
Preferably, the dimensional inconsistency between the elimination features in S21 is specifically:
for the acquired input variable xiNormalization is performed separately to eliminate dimensional inconsistencies between features.
Preferably, the step S3 specifically includes the following steps:
s31: selecting a proper kernel function by adopting a Gaussian regression process model GPR in machine learning, and constructing a composite kernel function to obtain a corresponding function model;
s32: correspondingly obtaining prior distribution of the output vector y;
s33: according to the Bayes framework, the prior distribution of the training output vector y is used to obtain the sum f*In the joint prior distribution, and accordingly, obtaining a test set output vector f*Posterior distribution and 95% confidence interval estimation;
s34: a suitable kernel function is selected as the kernel of the model.
Preferably, the kernel function of S34 is specifically: and selecting a square exponential covariance function to calculate the elements of the kernel matrix.
Preferably, the step S4 specifically includes the following steps:
s41: selecting different kernel functions to carry out model training, and carrying out optimization processing on the hyper-parameters;
s42: and comparing the difference of the training results by using the evaluation indexes so as to determine a final optimization model.
Preferably, the method further comprises:
s5, the prediction model is evaluated by comparing the indices of the plurality of machine learning regression models.
Preferably, the S5 specifically includes the following steps:
s51: meanwhile, other machine learning prediction models are constructed, wherein the prediction models comprise a support vector machine regression prediction model, a BP neural network regression prediction model and a random forest regression prediction model;
s52: and comparing the predicted performance indexes, wherein the predicted performance indexes comprise a mean absolute error MAE, a mean absolute percentage error MAPE and a root mean square error RMSE.
Compared with the prior art, the invention has the following advantages:
1) a building cold and heat load prediction model based on a data-driven Gaussian learning technology optimizes the problems of numerous data sets and long calculation time caused by various parameters, and the concept of effective utilization and energy-saving development is ensured to be implemented through estimation of prediction intervals of the building cold and heat load
2) The prediction model based on Gaussian Process Regression (GPR) has the advantages of easiness in programming realization, self-adaptive acquisition of hyper-parameters, probability distribution of output and the like when complex regression problems such as high dimensionality, nonlinearity and the like are processed, and is widely applied to multiple fields such as time series analysis, dynamic system model identification, system control and the like.
3) Compared with a neural network and a support vector machine, the method has the advantages of easiness in implementation, self-adaptive acquisition of hyper-parameters, flexibility in non-parameter inference, probability significance in output and the like.
4) The change of the building cold and heat load required to be predicted is nonlinear, and when random variables show obvious nonlinear trends, the Gaussian process regression of the method can well perform prediction analysis.
Drawings
FIG. 1 is a flow diagram of a method for predicting cold and heat loads of a building based on a data-driven Gaussian learning technique, according to an embodiment;
FIG. 2(a) is a graph of the predicted effect of thermal load based on a data-driven Gaussian learning technique (100 data points selected), according to one embodiment;
FIG. 2(b) is a graph of the predicted effect of cold load based on a data-driven Gaussian learning technique (100 data points selected), according to one embodiment;
FIG. 3(a) is a graph of heat load interval estimation based on a data-driven Gaussian learning technique (with 50 data points selected), according to one embodiment;
FIG. 3(b) is a cold load interval estimation graph based on a data-driven Gaussian learning technique (with 50 data points selected), according to one embodiment;
FIG. 4(a) is a graph of thermal load prediction effect and interval estimation based on a data-driven Gaussian learning technique after PCA is performed (100 data points are selected), according to one embodiment;
FIG. 4(b) is a graph of cold load prediction effectiveness and interval estimation based on data-driven Gaussian learning technique after PCA is performed (100 data points are selected), according to one embodiment;
FIG. 5(a) is a graph of heat load interval estimation based on a data-driven Gaussian learning technique after PCA is performed (50 data points are selected), according to one embodiment;
FIG. 5(b) is a plot of cold load interval estimation based on a data-driven Gaussian learning technique after PCA is performed (50 data points are selected), according to one embodiment;
FIG. 6(a) is a graph of thermal load prediction evaluation indicators based on a data-driven Gaussian learning technique, according to one embodiment;
FIG. 6(b) is a cold load prediction evaluation index plot based on a data-driven Gaussian learning technique, according to one embodiment.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, shall fall within the scope of protection of the present invention.
The invention discloses a building cold and hot load prediction method based on a data-driven Gaussian learning technology, which comprises the following steps of:
s1: acquiring a data set of a building cold and heat load and a plurality of characteristic data sets influencing the cold and heat load;
s2: performing data processing on the trained data set by using a principal component analysis method and extracting a proper amount of principal components;
s3: constructing a Gaussian regression process prediction model of the cold and hot load of the building, training the constructed Gaussian regression process model based on training set data so as to achieve the purpose of optimizing the model, and giving a prediction estimation interval in a final prediction model result;
s4: predicting the cold and heat load of the building and giving a prediction estimation interval by adopting an optimized Gaussian process model based on the data set subjected to principal component analysis;
s5: and evaluating the prediction model by comparing indexes of various machine learning regression models.
Further, the step S1 specifically includes the following steps:
acquiring a data set of the cold and heat load of the building and a plurality of characteristic data sets influencing the cold and heat load, wherein the data sets comprise: 768 different building cold and heat load data sets were simulated; the influence characteristic data set comprises 8 types of data which are respectively as follows: relative compactness, surface area, wall area, roof area, overall height, orientation, glass area distribution, representing the 8 impact characteristic data as input variable xi( i 1, 2.., 8) and the heat and cold loads are respectively represented as y1And y2As shown in table 1.
Mathematical representation Input/output volume
x1 Relative degree of compaction
x2 Surface area
x3 Area of wall
x4 Area of roof
x5 Overall height
x6 Orientation of
x7 Area of glass
x8 Area distribution of glass
y1 Thermal load
y2 Cold load
TABLE 1
Further, the step S2 specifically includes the following steps:
s21: correspondingly processing the acquired feature data, and eliminating dimension inconsistency among the features comprises the following steps: for the acquired input variable xiRespectively carrying out normalization to eliminate dimensional inconsistency among the characteristics;
s22: writing an input variable matrix into a column matrix form, and calculating an average characteristic value by using the following formula;
Figure BDA0002596849310000051
wherein the content of the first and second substances,
Figure BDA0002596849310000052
is an average characteristic value, xiThe number of the ith sample value is N;
s23: calculating the difference d between the characteristic value of each sample and the average characteristic valuei；
S24: constructing a covariance matrix C by using the following formula, simultaneously calculating the eigenvalue and the eigenvector of the matrix C, and constructing an eigenvector space;
Figure BDA0002596849310000053
s25: after the characteristic vector space is constructed, the characteristic values and the characteristic vectors are respectively sorted and screened in the characteristic space, so that a new input characteristic data set after principal component analysis is obtained.
Further, the step S3 specifically includes the following steps:
s31: selecting a proper kernel function by adopting a Gaussian regression process model (GPR) in machine learning, constructing a composite kernel function, and obtaining a corresponding function model:
for the regression problem, the statistical model is:
y＝f(X)+ (3)
in the formula: record as
Figure BDA0002596849310000061
Is a mean of 0 and a variance of
Figure BDA0002596849310000062
The white gaussian noise, I is a geometric identity matrix;
s32: the prior distribution of the corresponding obtained output vector y is:
Figure BDA0002596849310000063
in the formula:
Figure BDA0002596849310000064
is an n-order covariance matrix; k (X, X) ═ Kij) The order of n is the kernel matrix,
matrix element kij＝k(xi,xj)；
S33: according to the Bayes framework, the prior distribution of the training output vector y is used to obtain the sum f*Joint prior distribution of (c):
Figure BDA0002596849310000065
accordingly, test set output vector f*The posterior distribution of：
Figure BDA0002596849310000066
Figure BDA0002596849310000067
Figure BDA0002596849310000068
In the formula (f)*Obey a standard normal distribution; e (-) is the expectation function;
Figure BDA0002596849310000069
as expected, a deterministic prediction result and a 95% confidence interval estimate as test set output values;
s34: a suitable kernel function is selected as the kernel of the model. The method selects a square exponential covariance function (SE) to calculate an element of a kernel matrix, and the formula is
Figure BDA00025968493100000610
In the formula: unknown superparameter M ═ diag (l)-2) L is the variance scale;
Figure BDA00025968493100000611
is the kernel function signal variance;
Figure BDA00025968493100000612
is the noise variance.
Further, the step S4 specifically includes the following steps:
s41: selecting different kernel functions to carry out model training, and carrying out optimization processing on the hyper-parameters;
s42: and comparing the difference of the training results by using the evaluation indexes so as to determine a final optimization model.
Further, the step S5 specifically includes the following steps:
s51: in order to show the efficiency of the prediction model, other machine learning prediction models are constructed simultaneously, wherein the prediction models comprise a support vector machine regression prediction model, a BP neural network regression prediction model and a random forest regression prediction model.
S52: and comparing the predicted performance indexes, wherein the predicted performance indexes comprise Mean Absolute Error (MAE), Mean Absolute Percent Error (MAPE) and Root Mean Square Error (RMSE).
Figure BDA0002596849310000071
Figure BDA0002596849310000072
Figure BDA0002596849310000073
Wherein, observedtPredicted for the observed valuetIs a predicted value.
The specific embodiment discloses a method for predicting cold and hot loads of a building based on a data-driven Gaussian learning technology, which comprises the following steps:
s1: receiving related data about the building cold and heat load in a database, and performing simple processing including normalization on the data;
s2: performing principal component analysis on input variables aiming at building cold and hot load example data to generate new principal component variables;
s3: aiming at the building cold and hot load example data, constructing a prediction model based on Gaussian process regression;
s4: training a part of data sets according to a Gaussian regression prediction model, and optimizing parameters of the Gaussian regression prediction model;
s5: a plurality of machine learning tools are adopted to construct a prediction model which is used as a comparison item for comparing the evaluation index with the evaluation index of the invention.
Step S1 specifically includes the following steps:
the cold and heat load data of 768 different buildings are simulated; the influence characteristic data comprises 8 types of data which are respectively: relative compactness, surface area, wall area, roof area, overall height, orientation, glass area distribution, input and output quantities are represented by the following mathematical symbols, as shown in table 1.
The following transformation functions were used for the normalization of the data:
Figure BDA0002596849310000074
wherein, X*Is normalized data value, X is original value, XminIs the minimum value, X, in the data setmaxIs the maximum value in the data set.
Step S2 specifically includes the following steps:
the matrix is written in the form of a column matrix and the average eigenvalue is calculated using the following formula S21.
Figure BDA0002596849310000081
Wherein the content of the first and second substances,
Figure BDA0002596849310000082
is an average characteristic value, xiIs the ith sample value, and N is the number of samples.
S22, calculating the difference d between the characteristic value of each sample and the average characteristic valuei。
And S23, constructing a covariance matrix C by using the following formula, simultaneously calculating the eigenvalue and the eigenvector of the matrix C, and constructing an eigenvector space.
Figure BDA0002596849310000083
And S24, after the construction of the feature vector space is completed, sorting and screening the feature values and the feature vectors in the feature space respectively to obtain a new input feature data set after principal component analysis.
The composition matrix in which the SPSS 26 is used to analyze data and the principal component analysis can be used to obtain the interpretation of each index is shown in the following table.
Figure BDA0002596849310000084
TABLE 2
The values in the component matrix are divided by the square root of the eigenvalue of the component to obtain the eigenvector of the eigenvalue, and the result is shown in table 3.
Figure BDA0002596849310000091
TABLE 3
Therefore, the principal component has the following relationship with the original input variable:
Z1＝0.508X1-0.511X2+0.130X3-0.512X4+0.508X5
Z2＝-0.469X1+0.456X2+0.896X3-0.430X4+0.435X5
Z3＝0.801X7+0.801X8
Z4＝X6
Z5＝0.893X7-0.893X8(16)
step S3 specifically includes the following steps:
s31: and selecting a proper kernel function by adopting a Gaussian regression process model (GPR) in machine learning, and constructing a composite kernel function to obtain a corresponding function model.
For the regression problem, the statistical model is:
y＝f(X)+ (17)
in the formula: record as
Figure BDA0002596849310000092
Is a mean of 0 and a variance of
Figure BDA0002596849310000093
I is a geometric identity matrix.
S32: the prior distribution of the corresponding obtained output vector y is:
Figure BDA0002596849310000094
in the formula:
Figure BDA0002596849310000095
is an n-order covariance matrix; k (X, X) ═ Kij) Is an n-th order kernel matrix, matrix element kij＝k(xi,xj)。
S33: according to the Bayes framework, the prior distribution of the training output vector y is used to obtain the sum f*Joint prior distribution of (c):
Figure BDA0002596849310000101
accordingly, test set output vector f*The posterior distribution of (A) is:
Figure BDA0002596849310000102
Figure BDA0002596849310000103
Figure BDA0002596849310000104
in the formula (f)*Obey a standard normal distribution; e (-) is the expectation function;
Figure BDA0002596849310000105
as expected, the results of the deterministic predictions as well as the 95% confidence interval estimates are output as test sets.
Step S4 specifically includes the following steps:
and S41, optimizing the kernel function of the Gaussian regression model, wherein the kernel function of the Gaussian regression process has the main form: ardexponentaial, ardsquaredeplonential, ardmatern32, ardmatern52, ardratinalquadratic. By calculating evaluation indexes including Mean Absolute Error (MAE), mean percent absolute error (MAPE) and Root Mean Square Error (RMSE), the expression is as follows:
Figure BDA0002596849310000106
Figure BDA0002596849310000107
Figure BDA0002596849310000108
wherein, observedtPredicted for the observed valuetIs a predicted value.
Finally, an ardsquaredexpointial function, namely a square exponential covariance function, is selected as a Gaussian regression process to calculate the kernel matrix element, and the formula is
Figure BDA0002596849310000109
In the formula: unknown superparameter M ═ diag (l)-2) L is the variance scale;
Figure BDA00025968493100001010
is the kernel function signal variance;
Figure BDA00025968493100001011
is the noise variance.
Step S5 specifically includes the following steps:
according to the method, while a Gaussian regression model is built, in order to compare the advantages and benefits of the Gaussian regression model, other various machine learning tool comparison models are built, the final prediction result is evaluated by adopting a unified evaluation index, and the machine learning work comprises support vector machine regression (SVR), BP neural network and random forest.
The evaluation indexes of the heat load are shown in the following table:
thermal load MAE MAPE RMSE
GPR 0.3077 1.3783 0.4349
GPR via PCA 0.3121 1.4166 0.4307
SVR 1.2810 6.3988 1.7374
BP random network 0.9919 5.3466 1.2357
Random forest 0.5947 2.9956 0.8864
TABLE 4
The evaluation indexes of the cooling load are shown in the following table:
cold load MAE MAPE RMSE
GPR 0.1602 0.7193 0.2293
GPR via PCA 0.1896 0.8381 0.2736
SVR 1.3964 5.9792 1.8456
BP random network 1.1049 4.0280 1.6339
Random forest 1.2302 4.4010 1.8162
TABLE 5
As can be seen from tables 4 and 5, the error of the prediction result of the gaussian process regression is the lowest no matter the prediction result of the cold load or the heat load is fit, which means that the gaussian regression model of the present invention has significant advantages in the prediction model of the building load, and the analysis and the result prediction can be performed more accurately. The difference between the prediction result of the Gaussian process regression model after PCA and the prediction result of the original data is very small, which means that the model after PCA can reduce the calculation amount required by prediction, so that the prediction model is more efficient, and the result is not adversely affected.
While the invention has been described with reference to specific embodiments, the invention is not limited thereto, and various equivalent modifications and substitutions can be easily made by those skilled in the art within the technical scope of the invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (10)

1. A method for predicting cold and hot loads of a building based on a data-driven Gaussian learning technology is characterized by comprising the following steps:
s1, acquiring a data set of the building cold and heat load and a plurality of characteristic data sets influencing the cold and heat load;
s2, processing the data of the training data set by using a principal component analysis method, and extracting a set number of principal components;
s3, constructing a Gaussian regression process prediction model of the building cold and heat load, training the constructed Gaussian regression process prediction model based on training set data so as to achieve the purpose of optimizing the model, and giving a prediction estimation interval in a final prediction model result;
and S4, predicting the cold and heat load of the building and giving a prediction estimation interval based on the data set after principal component analysis and the optimized Gaussian process model.
2. The method for predicting the cold and heat load of the building based on the data-driven Gaussian learning technique as claimed in claim 1, wherein the data set of the cold and heat load of the building in step S1 is as follows: 768 different building cold and heat load data sets were simulated.
3. The method according to claim 1, wherein the plurality of feature data sets influencing cold and heat loads in step S1 include eight kinds, namely, relative compactness, surface area, wall area, roof area, overall height, orientation, glass area and glass area distribution, and the eight kinds of influence feature data are expressed as input variable xiWhere i is 1,2, 8, and the thermal and cold loads are respectively represented as y1And y2。
4. The method for predicting the cold and heat load of the building based on the data-driven Gaussian learning technology as claimed in claim 1, wherein the step S2 specifically comprises the following steps:
s21: correspondingly processing the acquired feature data to eliminate dimension inconsistency among the features;
s22: writing an input variable matrix into a column matrix form, and calculating an average characteristic value;
s23: calculating the difference d between the characteristic value of each sample and the average characteristic valuei；
S24: constructing a covariance matrix C, simultaneously calculating an eigenvalue and an eigenvector of the matrix C, and constructing an eigenvector space;
s25: after the characteristic vector space is constructed, the characteristic values and the characteristic vectors are respectively sorted and screened in the characteristic space, so that a new input characteristic data set after principal component analysis is obtained.
5. The method for predicting the cold and heat load of the building based on the data-driven Gaussian learning technology as claimed in claim 4, wherein the dimensional inconsistency between the elimination features in S21 is specifically:
for the acquired input variable xiNormalization is performed separately to eliminate dimensional inconsistencies between features.
6. The method for predicting the cold and heat load of the building based on the data-driven Gaussian learning technology as claimed in claim 1, wherein the step S3 specifically comprises the following steps:
s31: selecting a proper kernel function by adopting a Gaussian regression process model GPR in machine learning, and constructing a composite kernel function to obtain a corresponding function model;
s32: correspondingly obtaining prior distribution of the output vector y;
s33: according to the Bayes framework, the prior distribution of the training output vector y is used to obtain the sum f*In the joint prior distribution, and accordingly, obtaining a test set output vector f*Posterior distribution and 95% confidence interval estimation;
s34: a suitable kernel function is selected as the kernel of the model.
7. The method for predicting the cold and heat load of the building based on the data-driven Gaussian learning technology as claimed in claim 6, wherein the kernel function of S34 is specifically as follows: and selecting a square exponential covariance function to calculate the elements of the kernel matrix.
8. The method for predicting the cold and heat load of the building based on the data-driven Gaussian learning technology as claimed in claim 1, wherein the step S4 specifically comprises the following steps:
s41: selecting different kernel functions to carry out model training, and carrying out optimization processing on the hyper-parameters;
s42: and comparing the difference of the training results by using the evaluation indexes so as to determine a final optimization model.
9. The method for predicting the cold and heat load of the building based on the data-driven Gaussian learning technology as claimed in claim 1, further comprising:
s5, the prediction model is evaluated by comparing the indices of the plurality of machine learning regression models.
10. The method for predicting the cold and heat load of the building based on the data-driven Gaussian learning technology as claimed in claim 9, wherein the step S5 specifically comprises the following steps:
s51: meanwhile, other machine learning prediction models are constructed, wherein the prediction models comprise a support vector machine regression prediction model, a BP neural network regression prediction model and a random forest regression prediction model;
s52: and comparing the predicted performance indexes, wherein the predicted performance indexes comprise a mean absolute error MAE, a mean absolute percentage error MAPE and a root mean square error RMSE.
CN202010711865.4A 2020-07-22 2020-07-22 Building cold and hot load prediction method based on data-driven Gaussian learning technology Pending CN111861002A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010711865.4A CN111861002A (en) 2020-07-22 2020-07-22 Building cold and hot load prediction method based on data-driven Gaussian learning technology

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010711865.4A CN111861002A (en) 2020-07-22 2020-07-22 Building cold and hot load prediction method based on data-driven Gaussian learning technology

Publications (1)

Publication Number Publication Date
CN111861002A true CN111861002A (en) 2020-10-30

Family

ID=72950664

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010711865.4A Pending CN111861002A (en) 2020-07-22 2020-07-22 Building cold and hot load prediction method based on data-driven Gaussian learning technology

Country Status (1)

Country Link
CN (1) CN111861002A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112699601A (en) * 2020-12-28 2021-04-23 电子科技大学 Space-time reconstruction method for sensor network data
CN113762534A (en) * 2021-09-10 2021-12-07 广东电网有限责任公司 Building cold and heat load prediction method, device, equipment and storage medium

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106971240A (en) * 2017-03-16 2017-07-21 河海大学 The short-term load forecasting method that a kind of variables choice is returned with Gaussian process
CN108022001A (en) * 2017-09-20 2018-05-11 河海大学 Short term probability density Forecasting Methodology based on PCA and quantile estimate forest
CN109543203A (en) * 2017-09-22 2019-03-29 山东建筑大学 A kind of Building Cooling load forecasting method based on random forest
CN109978201A (en) * 2017-12-27 2019-07-05 深圳市景程信息科技有限公司 Probability load prediction system and method based on Gaussian process quantile estimate model
CN111062517A (en) * 2019-11-21 2020-04-24 上海航天智慧能源技术有限公司 GBDT-based LightGBM model cold and heat load prediction method
CN111245024A (en) * 2020-01-14 2020-06-05 山东大学 Comprehensive energy system robust optimization operation method based on model predictive control

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106971240A (en) * 2017-03-16 2017-07-21 河海大学 The short-term load forecasting method that a kind of variables choice is returned with Gaussian process
CN108022001A (en) * 2017-09-20 2018-05-11 河海大学 Short term probability density Forecasting Methodology based on PCA and quantile estimate forest
CN109543203A (en) * 2017-09-22 2019-03-29 山东建筑大学 A kind of Building Cooling load forecasting method based on random forest
CN109978201A (en) * 2017-12-27 2019-07-05 深圳市景程信息科技有限公司 Probability load prediction system and method based on Gaussian process quantile estimate model
CN111062517A (en) * 2019-11-21 2020-04-24 上海航天智慧能源技术有限公司 GBDT-based LightGBM model cold and heat load prediction method
CN111245024A (en) * 2020-01-14 2020-06-05 山东大学 Comprehensive energy system robust optimization operation method based on model predictive control

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
魏来: ""基于机器学习方法的建筑能耗性能研究"", 《中国优秀博硕士学位论文全文数据库（硕士） 工程科技II辑》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112699601A (en) * 2020-12-28 2021-04-23 电子科技大学 Space-time reconstruction method for sensor network data
CN113762534A (en) * 2021-09-10 2021-12-07 广东电网有限责任公司 Building cold and heat load prediction method, device, equipment and storage medium

Similar Documents

Publication Publication Date Title
CN109659933B (en) Electric energy quality prediction method for power distribution network with distributed power supply based on deep learning model
CN111950854B (en) Coke quality index prediction method based on multilayer neural network
CN107463993B (en) Medium-and-long-term runoff forecasting method based on mutual information-kernel principal component analysis-Elman network
Gaur Neural networks in data mining
CN111861002A (en) Building cold and hot load prediction method based on data-driven Gaussian learning technology
CN113822499B (en) Train spare part loss prediction method based on model fusion
CN106296434B (en) Grain yield prediction method based on PSO-LSSVM algorithm
CN111931983A (en) Precipitation prediction method and system
CN115587666A (en) Load prediction method and system based on seasonal trend decomposition and hybrid neural network
CN115423594A (en) Enterprise financial risk assessment method, device, equipment and storage medium
CN116187835A (en) Data-driven-based method and system for estimating theoretical line loss interval of transformer area
CN114357870A (en) Metering equipment operation performance prediction analysis method based on local weighted partial least squares
CN110738363A (en) photovoltaic power generation power prediction model and construction method and application thereof
CN114330815A (en) Ultra-short-term wind power prediction method and system based on improved GOA (generic object oriented architecture) optimized LSTM (least Square TM)
CN112241832B (en) Product quality grading evaluation standard design method and system
CN117875481A (en) Carbon emission prediction method, electronic device, and computer-readable medium
CN117497038A (en) Method for rapidly optimizing culture medium formula based on nuclear method
CN116341929A (en) Prediction method based on clustering and adaptive gradient lifting decision tree
Yu et al. A hybrid learning-based model for on-line monitoring and diagnosis of out-of-control signals in multivariate manufacturing processes
CN110288724B (en) Batch process monitoring method based on wavelet function principal component analysis
Sallehuddin et al. Forecasting small data set using hybrid cooperative feature selection
CN115204535B (en) Purchasing business volume prediction method based on dynamic multivariate time sequence and electronic equipment
El-Sheikh et al. Proposed two variable selection methods for big data: simulation and application to air quality data in Italy
CN118097435B (en) Supergraph neural network-based corn lodging classification method and device
Han et al. A modified fast recursive hidden nodes selection algorithm for ELM

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20201030

RJ01 Rejection of invention patent application after publication