A building load forecasting method considering time-varying feature correlation degree

Abstract

The application discloses a building load prediction method considering time-varying characteristic correlation degree, and belongs to the technical field of calculation, estimation or counting. First, historical building load data is collected, and characteristic factor parameters related to the building load are extracted; a characteristic correlation degree matrix is established to describe the change of the correlation degree between each characteristic factor and the building load with time; then, a grey prediction method is used to predict the correlation degree between each characteristic factor and the building load in a future period; next, the regression coefficient of the next period is calculated according to the average value of the predicted characteristic correlation coefficient and the historical data characteristic correlation coefficient, so that a time-varying characteristic correlation degree prediction model about the building load is obtained. Compared with a traditional multiple linear regression model, the time-varying characteristic correlation degree prediction model improved based on the characteristic correlation degree matrix and the grey prediction method has higher fitting degree and stronger prediction adaptability.

Description

Technical Field

[0001] This invention relates to intelligent buildings, and specifically to grey relational analysis, grey prediction, and multiple linear regression models. It discloses a building load prediction method that considers time-varying relational characteristics, belonging to the technical field of calculation, estimation, or counting. Background Technology

[0002] Building load forecasting is a crucial aspect of building energy conservation control, essential for rationally regulating the energy consumption of various building equipment and optimizing building system performance to improve energy efficiency. Multiple linear regression and grey prediction models are two commonly used forecasting algorithms, suitable for real-time forecasting and rapid response scenarios. Therefore, they are often combined for short-term forecasting. Current short-term load forecasting algorithms first use a grey prediction model to predict the parameter values ​​of relevant data in the next time period based on extracted data. Then, these parameter values ​​are substituted into a multiple linear regression model to obtain the predicted values. However, the multiple linear regression model for predicting relevant data suffers from several instabilities, such as weak data correlation and low model fit, which affect the quality and accuracy of the final prediction model. Therefore, traditional correlation data prediction algorithms often produce significant errors.

[0003] The challenge of building load forecasting lies in the inability to obtain accurate prediction results using existing short-term load forecasting algorithms. Current short-term load forecasting techniques do not consider the time-varying nature of the correlation coefficients between characteristic factors and predictor variables in real-world building load forecasting. Furthermore, the regression coefficients of multiple linear regression models based on data of characteristic factors and predictor variable values ​​over a period of time cannot accurately reflect the parametric relationship between building load and characteristic factors. Currently, there is a lack of effective algorithms and models to solve this problem, and the inherent limitations of existing short-term load forecasting techniques also restrict their application in building load forecasting. Therefore, there is an urgent need to propose a model suitable for building load forecasting.

[0004] In summary, the present invention aims to propose a building load forecasting method that considers the correlation degree of time-varying characteristics to overcome the above-mentioned defects. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a building load forecasting method that considers time-varying characteristic correlations. This method establishes a characteristic correlation coefficient matrix to describe the change in the correlation between each characteristic factor and the building load over time. Then, a grey prediction method is used to predict the characteristic correlation coefficients between each characteristic factor and the building load in a future time period to calculate the regression coefficients for the next time period. This approach allows for a more refined representation of the relationships between variables and addresses the problem of unbalanced data distribution during building load forecasting. The invention aims to improve the accuracy and reliability of building load forecasting and solves the technical problem that traditional multiple linear regression models cannot accurately capture time-varying characteristic correlations when processing characteristic parameters.

[0006] To achieve the above-mentioned objectives, the present invention employs the following technical solution:

[0007] A building load forecasting method considering time-varying characteristic correlation includes the following steps:

[0008] Step 1: Collect historical building load data and extract characteristic factor parameters;

[0009] Step 2: Establish the characteristic factor parameter matrix and building load vector, and normalize the characteristic factor parameter matrix and building load vector;

[0010] Step 3: Calculate the time-varying characteristic correlation coefficient matrix between building load and characteristic factors in the current time period;

[0011] Step 4: Use the grey prediction method to predict the characteristic correlation coefficients between each characteristic factor and the building load in the future period;

[0012] Step 5: Establish the multiple linear regression equation for building load, and calculate the multiple linear regression coefficients and random error;

[0013] Step 6: Adjust the multiple linear regression coefficients based on the time-varying characteristic correlation coefficient matrix calculated in Step 3 and the characteristic correlation coefficients between each characteristic factor and the building load in the future time period predicted in Step 4, and obtain the time-varying characteristic correlation prediction model of the building load and the predicted value of the building load at future time.

[0014] As a further optimization of the building load forecasting method that considers the correlation of time-varying characteristics, the feature factors extracted in step 1 include, but are not limited to, air temperature, dew point temperature, solar radiation intensity, air pressure, wind speed, and building occupancy rate.

[0015] As a further optimization scheme for the building load forecasting method that considers the correlation of time-varying characteristics, the feature factor parameter matrix established in step 2 is as follows: The building load vector is Y = [y0(1)...y0(t)], where X0 is the feature factor parameter matrix, x0(1,1) is the first feature factor parameter in the first time period, x0(1,t) is the first feature factor parameter in the t time period, x0(n,1) is the nth feature factor parameter in the first time period, x0(n,t) is the nth feature factor parameter in the t time period, n is the number of feature factors, t is the number of time periods selected, t≥n+1, n is a positive integer, and t is a positive integer greater than or equal to 2.

[0016] As a further optimization of the building load forecasting method that considers the correlation of time-varying characteristics, step 2 is achieved through the expression... The feature factor parameter matrix is ​​normalized using the expression. The building load vector is normalized, where x′0(i,j) is the normalized result of the i-th feature factor parameter in the j-th time period, x0(i,j) is the parameter of the i-th feature factor in the j-th time period, and max j x0(i,j), min j x0(i,j) represents the maximum and minimum values ​​of the i-th characteristic factor parameter on the date of the j-th time period, respectively; y′0(j) is the normalized data of the building load data for the j-th time period; y0(j) represents the building load data for the j-th time period; max j y0(j), min j y0(j) represents the maximum and minimum values ​​of the building load data for the j-th time period on the date, i≤n, j≤t.

[0017] As a further optimization of the building load forecasting method that considers time-varying characteristic correlation, step 3 calculates the expression for the characteristic correlation coefficient matrix between building load and characteristic factors in the current time period as follows: Where δ(i,j) is the characteristic correlation coefficient between building load and the ith characteristic factor in the j-th time period, r i Let be the Pearson correlation coefficient between the i-th characteristic factor and the building load. avg j x′0(i,j) is the average value of the normalized result of the i-th feature factor parameter over t time periods. avg j y′0(j) is the average value of the normalized results of the building load data over t time periods. γ is a fixed term, 0 < γ < 1, min i min j |x′0(i,j)-y′0(j)| is the minimum difference between x′0(i,j) and y′0(j) in the normalized matrix. The normalized matrix consists of the normalized result of the feature factor parameter matrix and the normalized result of the building load vector.i max j |x′0(i,j)-y′0(j)| is the maximum difference between x′0(i,j) and y′0(j) in the normalized matrix, and ρ is the resolution coefficient, ρ=0.5.

[0018] As a further optimization of the building load forecasting method that considers the correlation of time-varying characteristics, the multiple linear regression equation for building load established in step 5 is: in, Let k be the predicted building load for time period j. i Let ε be the multiple linear regression coefficient of the building load and the i-th characteristic factor, and let ε be the random error. The multiple linear regression coefficient and random error are calculated using the least squares method.

[0019] As a further optimization of the building load forecasting method that considers time-varying characteristic correlation, step 6 adjusts the expression of the multiple linear regression coefficients based on the time-varying characteristic correlation coefficient matrix calculated in step 3 and the characteristic correlation coefficients between each characteristic factor and the building load predicted in step 4 for a future time period. in, Let be the multiple linear regression coefficient of building load and the i-th characteristic factor at time t+1. The characteristic correlation coefficient between the building load and the i-th characteristic factor during the t+1-th time period predicted in step 4 is given.

[0020] As a further optimization of the building load forecasting method that considers the correlation of time-varying characteristics, the building load time-varying characteristic correlation prediction model obtained in step 6 is as follows: in, Let x0(i, t+1) be the building load forecast value for time period t+1, and let x0(i, t+1) be the parameter of the i-th characteristic factor for time period t+1.

[0021] An apparatus comprising:

[0022] One or more processors;

[0023] Memory, used to store one or more programs;

[0024] When one or more of the programs are executed by one or more processors, the one or more processors implement the building load forecasting method described above.

[0025] A storage medium containing computer-executable instructions that, when executed by a computer processor, perform the aforementioned building load forecasting method.

[0026] The present invention, by adopting the above technical solution, has the following beneficial effects:

[0027] (1) The building load prediction method proposed in this invention is based on the feature correlation matrix and the grey prediction model. The feature correlation matrix is ​​used to describe the change of the correlation between each feature factor and the building load over time. It can accurately reflect the influence of feature factors on building load and capture the time-varying relationship between feature factors and building load. The feature correlation coefficient between each feature factor and building load in the future time period is used to improve the multivariate linear equation, thereby obtaining a new prediction model for building load. Compared with the traditional load prediction model, the prediction model of this invention has higher prediction accuracy and can more accurately predict the future building load.

[0028] (2) The building load prediction model proposed in this invention adjusts the regression coefficient based on the characteristic correlation coefficient of gray prediction. Compared with the traditional multiple linear regression model that regards the weight of characteristic factors as fixed parameters, this model has better stability and adaptability and can better cope with the impact of environmental changes and abnormal data.

[0029] (3) The prediction model provided by this invention can be widely applied in the field of building load prediction. It can be applied to situations with multiple characteristic factors related to building load, and can more accurately predict future building load, providing a more reliable reference for decision-making in related fields. At the same time, the prediction method can explain the correlation between the prediction results and the characteristic factors by adjusting and controlling the correlation between the characteristic factors and the prediction results through regression coefficient adjustment. Furthermore, by processing the grey relational coefficient through the weighted linear change of the Pearson coefficient and then calculating the characteristic relational coefficient, the influence of nonlinear relationship on the prediction results can be weakened. Therefore, the model has better robustness and interpretability, and can better meet the needs of practical applications. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a flowchart of the building load prediction method proposed in this invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] like Figure 1 As shown, a building load forecasting method that considers the correlation degree of time-varying characteristics includes the following 6 steps.

[0034] Step 1: Collect historical building load data and extract characteristic factor parameters.

[0035] The extracted features include, but are not limited to: air temperature, dew point temperature, solar radiation intensity, air pressure, wind speed, and building occupancy rate.

[0036] Step 2: Establish the feature factor parameter matrix and building load vector, and perform normalization processing.

[0037] Let the feature factor parameter matrix of the selected time period be denoted as . The building load vector for the selected time period is Y = [y0(1)…y0(t)], where n represents the number of characteristic factors and t represents the number of selected time periods. To ensure that the multiple linear regression equation based on the selected time periods has a solution, t ≥ n+1 should be satisfied. Let t = n+1, then the normalized result x′0(i,j) of the i-th characteristic factor parameter in the j-th time period is expressed as:

[0038]

[0039] The normalized result y′0(j) of the building load data for time period j is expressed as:

[0040]

[0041] In the formula, x0(i,j) represents the parameter of the i-th characteristic factor in the j-th time period, i≤n, j≤t; max j x0(i,j), min j x0(i,j) represent the maximum and minimum values ​​of the i-th characteristic factor parameter on the j-th time period, respectively; y0(j) represents the building load data for the j-th time period, max j y0(j), min j y0(j) represents the maximum and minimum values ​​of the building load data for the j-th time period on the date.

[0042] Therefore, the normalized feature factor parameter matrix is ​​obtained as follows: The normalized building load vector is Y′=[y′0(1) … y′0(t)], and the two are combined into a normalized matrix.

[0043] Step 3: Calculate the time-varying characteristic correlation coefficient matrix between building load and characteristic factors.

[0044] The feature correlation coefficient, by weighting the linear correlation between feature factors and the desired building load with the grey feature correlation coefficient, can weaken feature factors with weak correlation and insignificant linear correlation during grey prediction, thereby reducing their interference with the prediction results. First, based on the normalized data Z from step 2, the Pearson correlation coefficient between the time series of normalized values ​​of different feature factor parameters and the time series of normalized values ​​of building load is calculated.

[0045] The Pearson correlation coefficient r between the i-th characteristic factor and the building load i for:

[0046]

[0047] In the above formula, avg j x′0(i,j) represents the average value of the normalized result of the i-th feature parameter over t time periods, avg j y′0(j) represents the average value of the normalized results of the building load data over t time periods.

[0048]

[0049]

[0050] Calculate the characteristic correlation coefficient δ(i,j) between building load and the i-th characteristic factor during the j-th time period:

[0051]

[0052] In the formula, γ is a fixed term, generally 0 < γ < 1, and here γ is taken as 0.5; min i min j |x′0(i,j)-y′0(j)| represents the minimum difference between x′0(i,j) and y′0(j) in the normalized matrix Z; max i max j |x′0(i,j)-y′0(j)| represents the maximum difference between x′0(i,j) and y′0(j) in the normalized matrix Z; ρ is the resolution coefficient, usually taken as ρ=0.5.

[0053] Therefore, the time-varying characteristic correlation coefficient matrix between building load and characteristic factors is:

[0054]

[0055] The column vector represents the characteristic correlation coefficient between different characteristic factors and building load in the same period, and the row vector represents the characteristic correlation coefficient between the same characteristic factor and building in different periods.

[0056] Step 4: Use the grey prediction method to predict the characteristic correlation coefficients between various characteristic factors and building load in the future time period.

[0057] First let

[0058]

[0059] 1≤i≤n

[0060] 1≤j≤t

[0061] When i = 1, let

[0062]

[0063] Therefore, the original sequence of the characteristic correlation coefficients is:

[0064] x (0) =(x 0 (1),x 0 (2),…x 0 (t))

[0065]

[0066] x (1) =(x 1 (1),x 1 (2),…x 1 (t))

[0067] Define x (1) The gray derivative is

[0068] d(j)=x 0 (j)=x 1 (j)-x 1 (j-1), j = 2, 3, ..., t

[0069] Let z (1) (j) is a sequence x (1) The neighboring values ​​generate a sequence, that is...

[0070] z (1) (j)=αx (1) (j)+(1-α)x (1) (j-1)

[0071] Therefore, the grey differential equation model of GM(1,1) is defined as follows:

[0072] d(j)+αz (1) (j) = β or x 0 (j)+αz (1) (j)=β

[0073] Substituting the times q = 2, 3, ..., t into the above equation, we have:

[0074]

[0075] Introducing matrix-vector notation:

[0076]

[0077]

[0078]

[0079] Therefore, the GM(1,1) model can be represented as Y = Bu.

[0080]

[0081] After obtaining the estimated values ​​of α and β, the GM(1,1) model is as follows:

[0082] x 0 (j)+αz (1) (j)=β

[0083] Solution

[0084]

[0085] q = 2, 3, ..., t

[0086] Thus, the predicted value is obtained.

[0087]

[0088] j = 2, 3, ..., t

[0089] Thus, the predicted value is obtained accordingly:

[0090]

[0091] j = 2, 3, ..., t

[0092] therefore

[0093]

[0094] Similarly, we can obtain

[0095]

[0096] Step 5: Calculate the multiple linear regression coefficients k of the multiple linear regression equation for building load. i and random error σ

[0097] Let the linear regression equation for building load be:

[0098]

[0099] The multiple linear regression coefficient k of the building load was calculated using the least squares method. i And random error ε.

[0100] Step 6: Adjust the regression coefficients of the multiple linear equation based on the characteristic correlation coefficients between each characteristic factor and the building load in the future time period predicted in Step 4, as well as the average historical characteristic correlation coefficients, to obtain the time-varying characteristic correlation prediction model of the building load and the predicted building load value for future times.

[0101] The time-varying characteristic correlation coefficient matrix R between building load and characteristic factors obtained in step 3, and the characteristic correlation coefficient of time period t+1 predicted in step 4. The regression coefficient between building load and the i-th characteristic factor in time period t+1 can be obtained as follows:

[0102]

[0103] This represents the average value of the historical feature correlation coefficients, which are obtained from the time-varying feature correlation coefficient matrix calculated in step 3.

[0104] The predicted building load for time period t+1 is:

[0105]

[0106] Based on the same inventive concept, this invention also provides a computer device, comprising: one or more processors, and a memory for storing one or more computer programs; the programs include program instructions, and the processor executes the program instructions stored in the memory. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, used to implement one or more instructions, specifically for loading and executing one or more instructions stored in a computer storage medium to implement the above-described method.

[0107] It should be further explained that, based on the same inventive concept, the present invention also provides a computer storage medium storing a computer program, which, when executed by a processor, performs the above-described method. This storage medium can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific, and not exhaustive, examples of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0108] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0109] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the present invention without departing from its spirit and scope, and all such changes and modifications fall within the protection scope claimed by the present invention.

Claims

1. A building load forecasting method considering time-varying characteristic correlation, characterized in that, Includes the following steps: Step 1: Collect historical building load data and extract characteristic factor parameters; Step 2: Establish the feature factor parameter matrix and the building load vector, and normalize the feature factor parameter matrix and the building load vector; Step 3: Calculate the time-varying characteristic correlation coefficient matrix between building load and characteristic factors in the current time period. The specific expression is as follows: ,in, For the first The characteristic correlation coefficient between building load and the i-th characteristic factor within a time period. Let be the Pearson correlation coefficient between the i-th characteristic factor and the building load. , For the i-th feature factor parameter in The average of the normalized results for each time period. , For building load data in The average of the normalized results for each time period. , As a fixed item, , In the normalized matrix and The minimum difference, the normalization matrix is ​​composed of the normalization result of the feature factor parameter matrix and the normalization result of the building load vector, In the normalized matrix and The maximum difference, The resolution coefficient, ; Step 4: Use the grey prediction method to predict the characteristic correlation coefficients between each characteristic factor and the building load in the future period; Step 5: Establish the multiple linear regression equation for building load, and calculate the multiple linear regression coefficients and random error; Step 6: Adjust the multiple linear regression coefficients based on the time-varying characteristic correlation coefficient matrix calculated in Step 3 and the characteristic correlation coefficients between each characteristic factor and building load in the future time period predicted in Step 4. The specific expression is as follows: ,in, For the building load and the i-th characteristic factor in The multiple linear regression coefficients, For the prediction of step 4 The characteristic correlation coefficient between building load and the i-th characteristic factor within a time period is used to obtain the time-varying characteristic correlation prediction model of building load and the predicted building load value at future times. The obtained time-varying characteristic correlation prediction model of building load is as follows: ,in, for Building load forecast for the time period For the i-th feature factor Parameters for the time period.

2. The building load forecasting method considering time-varying characteristic correlation as described in claim 1, characterized in that, The feature factors extracted in step 1 include, but are not limited to, air temperature, dew point temperature, solar radiation intensity, air pressure, wind speed, and building occupancy rate.

3. The building load forecasting method considering time-varying characteristic correlation as described in claim 2, characterized in that, The feature factor parameter matrix established in step 2 is as follows: The established building load vector is ,in, For the characteristic factor parameter matrix, For the first The first characteristic factor parameter of the time period, For the first The first characteristic factor parameter of the time period, For the first Time period Each characteristic factor parameter, For the first Time period Each characteristic factor parameter, t represents the number of characteristic factors, and t represents the number of time periods selected. , It is a positive integer. It is a positive integer greater than or equal to 2.

4. The building load forecasting method considering time-varying characteristic correlation as described in claim 3, characterized in that, Step 2 is achieved through an expression. The feature factor parameter matrix is ​​normalized using the expression. The building load vector is normalized, where, For the first The normalized result of the i-th feature factor parameter in the time period. For the i-th feature factor Time period parameters, , The first The i-th feature parameter of the time period falls within the maximum and minimum values. For the first Normalization of building load data Indicates the building load. Data for a specific time period The first The maximum and minimum values ​​of building load data for the given time period and date. , .

5. The building load forecasting method considering time-varying characteristic correlation as described in claim 1, characterized in that, The multiple linear regression equation for building load established in step 5 is as follows: ,in, Let be the predicted building load for time period j. Let be the multiple linear regression coefficient of building load and the i-th characteristic factor. The random error is calculated using the least squares method to determine the multiple linear regression coefficients and the random error.

6. A device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When one or more of the programs are executed by one or more of the processors, the one or more of the processors implement the building load forecasting method as described in any one of claims 1-5.

7. A storage medium containing computer-executable instructions, characterized in that, The computer-executable instructions, when executed by a computer processor, perform the building load forecasting method as described in any one of claims 1-5.

Images