CN116663744A - Energy consumption prediction method and system for near-zero energy consumption building - Google Patents

Abstract

The invention provides an energy consumption prediction method and device for a near-zero energy consumption building. The method comprises the following steps: determining as many factors affecting building energy consumption as possible based on the specific structure of the building to be predicted; screening the factors based on the correlation to obtain N factors with the most obvious influence on the building energy consumption; and establishing a prediction model which takes the N factors as input variables and the building energy consumption as output variables, and predicting the building energy consumption by using the trained prediction model. According to the invention, the factors influencing the energy consumption of the building are screened based on the correlation, N factors which have the most obvious influence on the energy consumption of the building are used as input variables of the prediction model, and the prediction precision of the prediction model is greatly improved.

Description

A method and system for predicting energy consumption in near-zero-energy buildings

technical field

The invention belongs to the technical field of building energy consumption measurement and control, and in particular relates to an energy consumption prediction method and system for near-zero energy consumption buildings.

Background technique

The concept of building energy consumption includes two connotations of building and energy consumption. The so-called building energy consumption generally has a broad sense and a narrow sense. In a broad sense, building energy consumption refers to the energy consumption in the whole process from building material manufacturing, building construction, to building use; in a narrow sense, building energy consumption refers to the operating energy consumption of buildings, which refers to people’s daily energy consumption, such as heating, air conditioning, lighting, Energy consumption for cooking, laundry, etc. In order to reduce building energy consumption, people put forward the concept of near-zero energy building.

Nearly zero-energy buildings are undoubtedly to achieve "low energy consumption" and "high energy efficiency". Accurately predicting building energy consumption and taking corresponding control measures is the key to reducing building energy consumption and improving energy saving rate. For this reason, the present invention proposes an energy consumption prediction method and system for near-zero energy consumption buildings.

Contents of the invention

In order to solve the above-mentioned problems in the prior art, the present invention provides an energy consumption prediction method and system for near-zero energy consumption buildings.

In order to achieve the above object, the present invention adopts the following technical solutions.

In a first aspect, the present invention provides a method for predicting energy consumption for a nearly zero-energy building, comprising the following steps:

Determine as many factors as possible that affect building energy consumption based on the specific structure of the building to be predicted;

The factors are screened based on correlation to obtain N factors that have the most significant impact on building energy consumption;

A prediction model is established with the N factors as input variables and building energy consumption as output variable, and the trained prediction model is used to predict building energy consumption.

Further, the factors affecting building energy consumption include at least: heat transfer coefficient of exterior walls, heat transfer coefficient of interior walls, heat transfer coefficient of exterior windows, solar heat absorption coefficient of exterior windows, area ratio of south-facing windows and walls, area ratio of north-facing windows and walls, East-west window-to-wall area ratio.

Further, the screening of the factors based on the correlation includes:

The sample data sets X _j (x _ji ) and Y (y _i ) corresponding to each factor and building energy consumption are obtained based on historical data, where x _ji is the i-th sample of the sample data set X _j of the j-th factor Data, y _i is the i-th sample data of the sample data group Y of building energy consumption, j=1,2,...,m, m is the number of factors, i=1,2,...,n, n is the sample quantity;

Calculate the correlation coefficient between X _j (x _ji ) and Y(y _i ), and sort the m factors in descending order of the corresponding correlation coefficients;

Filter out the top N factors.

Further, the screening of the factors based on the correlation also includes:

Calculate the correlation coefficient between the sample data set X _j (x _ji ) and X _k (x _ki ) of any two factors in the N factors, for the X _j (x _ji ) whose correlation coefficient is greater than the set threshold , X _k (x _ki ), delete a factor corresponding to a sample data group ranked lower, where j≠k.

Furthermore, the formula for calculating the correlation coefficient between X _j (x _ji ) and Y(y _i ) is:

In the formula, R _j is the correlation coefficient between X _j (x _ji ) and Y(y _i ).

Further, the prediction model is a multiple linear regression model or an artificial neural network model.

Further, the method further includes: performing a significance test on the prediction accuracy of the prediction model, and if the inspection is qualified, it indicates that the prediction accuracy meets the requirements; otherwise, the prediction model does not meet the requirements, and optimizes the prediction model.

Further, the method for optimizing the prediction model includes:

Combining any two or more input variables of the prediction model to obtain new input variables; the combination includes linear combination and nonlinear combination;

All input variables are screened based on correlation to obtain M input variables that have the most significant impact on building energy consumption;

The input variables of the prediction model are updated to the M input variables, and the updated prediction model is retrained.

Furthermore, methods for combining any two input variables include:

Combining the input variables X _i and X _j is: X=k*X _i +(1-k)*X _j , wherein, i≠j, 0<k<1;

Calculate the correlation coefficient R _X between the combined variable X and the output Y when k takes different values, and calculate k _max when R _X takes the maximum value R _X-max ;

Substituting k _max into X ₁ and X ₂ to obtain the best combined variable: X=k _max *X ₁ +(1-k _max )*X ₂ .

In the second aspect, the present invention provides an energy consumption prediction system for a near-zero-energy building, including:

The factor determination module is used to determine as many factors as possible that affect building energy consumption based on the specific structure of the building to be predicted;

A factor screening module, configured to screen the factors based on correlation to obtain the N factors that have the most significant impact on building energy consumption;

The energy consumption prediction module is used to establish a prediction model with the N factors as input variables and building energy consumption as an output variable, and use the trained prediction model to predict building energy consumption.

Compared with the prior art, the present invention has the following beneficial effects.

The present invention determines as many factors affecting building energy consumption as possible based on the specific structure of the building to be predicted, and screens the factors based on correlation to obtain N factors that have the most significant impact on building energy consumption. A prediction model with three factors as the input variable and building energy consumption as the output variable, using the trained prediction model to predict the building energy consumption, realizes the automatic prediction of the near-zero energy consumption building energy consumption. The present invention screens the factors affecting building energy consumption based on correlation, and takes the N most significant influencing factors on building energy consumption as input variables of the prediction model, thereby greatly improving the prediction accuracy of the prediction model.

Description of drawings

Fig. 1 is a flowchart of an energy consumption prediction method for a near-zero energy building according to an embodiment of the present invention.

Fig. 2 is a block diagram of an energy consumption prediction system for a near-zero energy building according to an embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solution and advantages of the present invention clearer and clearer, the present invention will be further described below in conjunction with the accompanying drawings and specific embodiments. Apparently, the described embodiments are only some of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

Fig. 1 is a flowchart of an energy consumption prediction method for a near-zero energy building according to an embodiment of the present invention, including the following steps:

Step 101, determining as many factors as possible that affect building energy consumption based on the specific structure of the building to be predicted;

In step 102, the factors are screened based on the correlation to obtain N factors that have the most significant impact on building energy consumption;

Step 103, establishing a prediction model with the N factors as input variables and building energy consumption as output variable, and using the trained prediction model to predict building energy consumption.

In this embodiment, step 101 is mainly used to determine various factors affecting building energy consumption. This embodiment predicts building energy consumption by establishing a predictive model, and the key to establishing an effective predictive model is to correctly select the input variables of the predictive model, that is, to select the most significant impact on the output of the predictive model, that is, building energy consumption. influencing factors as its input variables. Therefore, in order to correctly select input variables, it is necessary to list as comprehensively and as many factors that have an impact on building energy consumption as possible, so as not to omit some factors that seem to have little impact but actually have significant impact. That is to say, in the preliminary selection stage of influencing factors in step 101, based on the principle of "multiple and comprehensive", "it is better to choose one thousand by mistake than to let one go", because the following step 102 will also strictly control all the listed factors. Screening to filter out factors that have no significant impact.

Different building structures have different factors affecting building energy consumption, and the same factors have different influence degrees. For example, the invention patent with the application number 202111604996.3 "A data-driven building heating load forecasting method, device and equipment". The prediction model takes the equivalent heat loss coefficient of the building and the area ratio of the building's transparent envelope as input variables; and Dou Baoyue's master's degree thesis "Research on Influencing Factors and Prediction Models of Energy Consumption of Near-Zero-Energy Residential Buildings" (2015) The prediction model established in , its input variables include the heat transfer coefficient of the exterior wall, the heat transfer coefficient of the interior wall, the heat transfer coefficient of the exterior window, etc. In addition, in addition to the parameters related to the building structure, the factors may also consider environmental factors and human factors (such as the aforementioned invention patents). Therefore, this embodiment does not limit specific influencing factors.

In this embodiment, step 102 is mainly used to screen various influencing factors obtained in step 101 . As mentioned above, the purpose of screening various influencing factors is to determine the effective input variables of the prediction model; in addition, another purpose of factor screening is to reduce the number of input variables, because when there are many input variables in the prediction model, not only will the The prediction model becomes complicated, and the calculation amount will increase rapidly, seriously affecting the running speed. In this embodiment, the degree of influence of various factors on building energy consumption is judged according to the correlation between various factors and building energy consumption. The stronger the correlation, the more significant the effect. The correlation coefficient can be used to represent the degree of correlation, and the larger the absolute value of the correlation coefficient, the higher the degree of correlation. When the correlation coefficient is positive, it is a positive correlation, that is, the larger the input, the larger the output; when the correlation coefficient is negative, it is a negative correlation, that is, the larger the input, the smaller the output. The relationship between the absolute value of the correlation coefficient and the degree of influence is: 0.8-1.0 is a very strong correlation, 0.6-0.8 is a strong correlation, 0.4-0.6 is a moderate degree of correlation, 0.2-0.4 is a weak correlation, 0.0-0.2 is a very weak correlation or no correlation relevant.

In this embodiment, step 103 is mainly used to predict building energy consumption. In this embodiment, the trained prediction model is used to predict building energy consumption. The input variables of the prediction model are the factors that have the most significant impact on building energy consumption that have been screened out above, and the output variable is building energy consumption. Building energy consumption can be total energy consumption or air conditioning energy consumption (easier to verify). Predictive model training requires building a training dataset based on historical data. If it is difficult to collect historical data structures, the existing modeling software (Dest software) can be used to simulate and calculate supplementary training data. After the prediction model is trained, the values of the input variables are input into the prediction model, and the output of the prediction model is the building energy consumption.

As an optional embodiment, factors affecting building energy consumption at least include: external wall heat transfer coefficient, internal wall heat transfer coefficient, external window heat transfer coefficient, external window solar heat absorption coefficient, south-facing window-to-wall area ratio, north-facing window Wall area ratio, east-west window-to-wall area ratio.

This example gives several factors that affect building energy consumption. As mentioned above, different building structures have different factors that affect building energy consumption; and the architectural styles in regions that cannot meet the climate conditions are also obviously different, for example, the difference between the northern and southern architectural structures is more obvious. Several influencing factors given in this embodiment are suitable for building construction in cold northern regions. The main structure of a near-zero-energy building generally adopts steel frame + cast-in-place polystyrene particle foam concrete wall, and adopts better thermal insulation technical measures for the outer protective structure. The high-performance envelope structure insulation system can control the heat transfer coefficient of the building envelope to be very low. The heat transfer performance of the building envelope directly affects the heating and air conditioning energy consumption of the building. Therefore, the thermal insulation performance of the building is a crucial factor to reduce the building energy consumption. The building envelope mainly refers to building exterior walls, roofs, doors and windows, etc., which influence and restrict each other. Combined with the thermal insulation technical measures of the outer enclosure structure, this example gives several factors that have a significant impact on building energy consumption, including the heat transfer coefficient of the inner and outer walls, the heat transfer coefficient of the outer window, the solar heat absorption coefficient of the outer window, and the Window-to-wall area ratios facing (south, north, and east-west).

As an optional embodiment, the screening of the factors based on the correlation includes:

The sample data sets X _j (x _ji ) and Y (y _i ) corresponding to each factor and building energy consumption are obtained based on historical data, where x _ji is the i-th sample of the sample data set X _j of the j-th factor Data, y _i is the i-th sample data of the sample data group Y of building energy consumption, j=1,2,...,m, m is the number of factors, i=1,2,...,n, n is the sample quantity;

Calculate the correlation coefficient between X _j (x _ji ) and Y(y _i ), and sort the m factors in descending order of the corresponding correlation coefficients;

Filter out the top N factors.

This embodiment provides a technical solution for screening influencing factors based on correlation. The technical principle of screening the influencing factors in this embodiment is: factors that are more closely related to building energy consumption have a more significant impact on building energy consumption. Therefore, it is only necessary to calculate the correlation coefficient between each influencing factor and building energy consumption, and then sort the influencing factors according to the order of the correlation coefficient from large to small, and select the top N influencing factors, that is, the unfinished correlation-based Screening of influencing factors. The size of N is determined empirically, neither too large nor too small.

As an optional embodiment, the screening of the factors based on the correlation also includes:

Calculate the correlation coefficient between the sample data set X _j (x _ji ) and X _k (x _ki ) of any two factors in the N factors, for the X _j (x _ji ) whose correlation coefficient is greater than the set threshold , X _k (x _ki ), delete a factor corresponding to a sample data group ranked lower, where j≠k.

This embodiment is a further screening on the basis of the previous embodiment. Although the influencing factors screened out in the previous embodiment are factors that have a relatively significant impact on building energy consumption, the correlation between the screened out influencing factors is not considered. If the correlation between several influencing factors is large, their influence effects are similar, and only one of them can be used to replace other influencing factors, that is, only one of them can be reserved as the input variable of the prediction model. This further simplifies the prediction model. Of course, one of the remaining factors should be the one with the strongest correlation with building energy consumption, that is, the top one. The specific screening method is as above, and will not be repeated here.

As an optional embodiment, the formula for calculating the correlation coefficient between X _j (x _ji ) and Y(y _i ) is:

In the formula, R _j is the correlation coefficient between X _j (x _ji ) and Y(y _i ).

This embodiment provides a technical solution for calculating the correlation coefficient. In this embodiment, the Pearson correlation coefficient (Pearson correlation coefficient) is used to calculate the correlation coefficient between X _j (x _ji ) and Y(y _i ). The Pearson correlation coefficient is widely used to measure the degree of correlation between two variables, and its value is between -1 and 1. This correlation coefficient is also called the Pearson product-moment correlation coefficient.

As an optional embodiment, the prediction model is a multiple linear regression model or an artificial neural network model.

This example presents two available predictive model structures. Regression analysis refers to a statistical analysis method that determines the interdependent quantitative relationship between the objective function and various influencing factors through a large amount of observation data. According to the number of influencing factors, it can be divided into unary regression and multiple regression; according to the relationship between the objective function and influencing factors, it can be divided into linear regression and nonlinear regression. Therefore, there can be multiple linear regression models and multiple nonlinear regression models. Regression analysis generally adopts the least squares method, which is to minimize the sum of squares of the deviations between the measured values and the regression values that contain random errors. Artificial neural network (ANN) abstracts the human brain neuron network from the perspective of information processing, establishes a simple model, and forms different networks according to different connection methods. In engineering and academia, it is often referred to directly as a neural network or a neural network. A neural network is an operational model consisting of a large number of interconnected nodes. Each node represents a specific output function, called the activation function. Each connection between two nodes represents a weighted value for the signal passing through the connection, called weight, which is equivalent to the memory of the artificial neural network. The output of the network varies according to the way the network is connected, the weight value and the activation function. The network itself is usually an approximation to a certain algorithm or function in nature, or it may be an expression of a logical strategy. Compared with the multiple linear (or nonlinear) regression model, the artificial neural network has a more complex structure and a larger amount of calculation, but its accuracy is also higher.

As an optional embodiment, the method further includes: performing a significance test on the prediction accuracy of the prediction model, if the inspection is qualified, it indicates that the prediction accuracy meets the requirements; otherwise, the prediction accuracy does not meet the requirements, and the prediction model is optimized .

This embodiment provides a technical solution for testing the prediction model. After the prediction model is trained, a significance test is required to verify whether the prediction model meets the requirements. The significance level of this example is 0.05. If the test result does not meet the requirements, the prediction model needs to be further optimized.

As an optional embodiment, the method for optimizing the prediction model includes:

Combining any two or more input variables of the prediction model to obtain new input variables; the combination includes linear combination and nonlinear combination;

All input variables are screened based on correlation to obtain M input variables that have the most significant impact on building energy consumption;

The input variables of the prediction model are updated to the M input variables, and the updated prediction model is retrained.

This embodiment provides a technical solution for further optimizing the prediction model. Practice has shown that it is sometimes difficult to find factors that have a significant impact on the prediction results during the modeling process, resulting in the prediction accuracy of the prediction model not meeting the requirements. The inventor found through repeated experiments that it is difficult to obtain a high-precision prediction model by using some influencing factors alone as input variables of the prediction model; however, the new variables obtained by combining some influencing factors can produce unexpected effects—— The significance of the influence of the prediction results is significantly higher than that of the single factor before combination. This embodiment is based on this finding that the input variables of the predictive model to be optimized are combined (it can be in any combination) to obtain a large number of new variables, and then these new variables and the original input variables are screened based on the correlation to screen out The variables with the strongest correlation with building energy consumption are used as the input variables of the prediction model, and the model training is carried out again. Of course, this optimization method can also be applied in step 102 after combining various influencing factors and then screening based on correlation to obtain the most effective input variables for the prediction model.

As an optional embodiment, the method for combining any two input variables includes:

Combining the input variables X _i and X _j is: X=k*X _i +(1-k)*X _j , wherein, i≠j, 0<k<1;

Calculate the correlation coefficient R _X between the combined variable X and the output Y when k takes different values, and calculate k _max when R _X takes the maximum value R _X-max ;

Substituting k _max into X ₁ and X ₂ to obtain the best combined variable: X=k _max *X ₁ +(1-k _max )*X ₂ .

This embodiment provides a technical solution for linear combination of two input variables. First give the general expression of the linear combination X of two input variables, the weighting coefficients of the two input variables are k and 1-k respectively; then calculate the maximum value R _X-max of the correlation coefficient between X and output Y and the corresponding The weighting coefficient k _max of ; substituting k _max into the general expression of X can get the combination variable with the strongest correlation with the output Y. It is worth noting that although this embodiment only provides a combination scheme for two input variables, it can be used for the combination scheme of more than two input variables with a little modification, but it is more difficult to find the maximum value. There are mature optimization algorithms to solve the maximum value.

Fig. 2 is a schematic diagram of the composition of an energy consumption prediction system for a near-zero-energy building according to an embodiment of the present invention, and the system includes:

A factor determination module 11, configured to determine as many factors as possible that affect building energy consumption based on the specific structure of the building to be predicted;

The factor screening module 12 is used to screen the factors based on the correlation to obtain the N factors that have the most significant impact on building energy consumption;

The energy consumption prediction module 13 is configured to establish a prediction model with the N factors as input variables and building energy consumption as an output variable, and use the trained prediction model to predict building energy consumption.

The system of this embodiment can be used to execute the technical solution of the method embodiment shown in FIG. 1 , and its implementation principle and technical effect are similar, and will not be repeated here.

The above is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Anyone skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present invention. All should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims (10)

1. An energy consumption prediction method for a near zero energy consumption building is characterized by comprising the following steps:

determining as many factors affecting building energy consumption as possible based on the specific structure of the building to be predicted;

screening the factors based on the correlation to obtain N factors with the most obvious influence on the building energy consumption;

and establishing a prediction model which takes the N factors as input variables and the building energy consumption as output variables, and predicting the building energy consumption by using the trained prediction model.

2. The energy consumption prediction method for near zero energy consumption buildings according to claim 1, wherein the factors influencing building energy consumption include at least: the heat transfer coefficient of the outer wall, the heat transfer coefficient of the inner wall, the heat transfer coefficient of the outer window, the solar heat absorption coefficient of the outer window, the area ratio of the south window wall, the area ratio of the north window wall and the east-west window wall.

3. The energy consumption prediction method for near zero energy consumption buildings according to claim 1, wherein the filtering the factors based on correlation comprises:

obtaining a sample data set X corresponding to each factor and building energy consumption based on historical data _j (x _ji ) And Y (Y) _i ) Wherein x is _ji Sample data set X as the j-th factor _j Is the ith sample data, y _i The i-th sample data of the sample data set Y for building energy consumption, j=1, 2, …, m, m is the number of factors, i=1, 2, …, n,n is the number of samples;

calculate X _j (x _ji ) And Y (Y) _i ) And ordering m factors according to the sequence from the big to the small of the corresponding correlation coefficient;

the top N factors are screened out.

4. The energy consumption prediction method for near zero energy consumption buildings according to claim 3, wherein the filtering the factors based on correlation further comprises:

calculating a sample data set X of any two factors of the N factors _j (x _ji ) And X is _k (x _ki ) A correlation coefficient between the two, X for which the correlation coefficient is larger than a set threshold value _j (x _ji )、X _k (x _ki ) One factor corresponding to the last sample data set is deleted, where j+.k.

5. A method of energy consumption prediction for near zero energy consumption buildings according to claim 3, wherein X _j (x _ji ) And Y (Y) _i ) The calculation formula of the correlation coefficient of (2) is as follows:

wherein R is _j Is X _j (x _ji ) And Y (Y) _i ) Is used for the correlation coefficient of the (c).

6. The energy consumption prediction method for near zero energy consumption buildings according to claim 1, wherein the prediction model is a multiple linear regression model or an artificial neural network model.

7. The energy consumption prediction method for near zero energy consumption buildings of claim 1, further comprising: performing significance test on the prediction precision of the prediction model, and if the prediction precision is qualified, indicating that the prediction precision meets the requirement; otherwise, the prediction precision does not meet the requirement, and the prediction model is optimized.

8. The energy consumption prediction method for near zero energy consumption buildings according to claim 7, characterized in that the method of optimizing the prediction model comprises:

combining any two or more input variables of the prediction model to obtain a new input variable; the combination includes a linear combination and a nonlinear combination;

screening all input variables based on correlation to obtain M input variables with the most obvious influence on building energy consumption;

and updating the input variables of the prediction model into the M input variables, and retraining the updated prediction model.

9. The energy consumption prediction method for near zero energy consumption buildings of claim 8, wherein the method of combining any two input variables comprises:

will input variable X _i 、X _j The combination is as follows: x=k×x _i +(1-k)*X _j Wherein i+.j, 0<k<1；

Calculating the correlation coefficient R of the combined variable X and the output Y when k takes different values _X And calculate R _X Take the maximum value R _X-max K at time _max ；

Will k _max Substitution to obtain X ₁ 、X ₂ Is the best combined variable of (a): x=k _max *X ₁ +(1-k _max )*X ₂ 。

10. An energy consumption prediction system for a near zero energy consumption building, comprising:

the factor determining module is used for determining factors influencing the energy consumption of the building as much as possible based on the specific structure of the building to be predicted;

the factor screening module is used for screening the factors based on the correlation to obtain N factors with the most obvious influence on the building energy consumption;

and the energy consumption prediction module is used for establishing a prediction model which takes the N factors as input variables and takes the building energy consumption as output variables, and predicting the building energy consumption by using the trained prediction model.