Long-term net energy consumption prediction method and system based on adaptive analysis model

Abstract

The invention discloses a long-term net energy consumption prediction method and system based on an adaptive analysis model, and the method comprises the steps: collecting energy consumption data, and forming an energy consumption input sequence; constructing a multiple regression model according to the energy consumption input sequence; based on the multiple regression models and the energy consumption input sequence, the multiple regression model with the minimum error between the predicted value and the actual value of the net energy consumption is screened out to serve as the optimal multiple regression model; based on the energy consumption input sequence, constructing an ARIMA-BO-LSSVM model, and predicting the net energy consumption according to the model; and comparing the obtained predicted value of the net energy consumption of the optimal multiple regression model with the predicted value of the net energy consumption of the ARIMA-BO-LSSVM model, and selecting a model closest to the actual net energy consumption as a prediction model of the long-term net energy consumption. According to the method, an ARIMA model is adopted to capture a linear part hidden in the power consumption data, and a BO-LSSVM model is applied to capture a nonlinear part existing on the data, so that the prediction accuracy is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of long-term net energy consumption prediction, and in particular relates to a long-term net energy consumption prediction method and system based on an adaptive analysis model. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] In order to accurately predict future net energy consumption and demand, some scholars have proposed prediction models using econometrics, artificial intelligence and hybrid methods, such as regression analysis, autoregressive iterative moving average (ARIMA), seasonal autoregressive iterative moving average (SARIMA), grey theory, artificial neural network, support vector machine (SVM) and least squares support vector machine (LSSVM), etc. However, the existing models rarely consider the impact of other factors on net energy consumption, resulting in low accuracy of the prediction models.

[0004] CN201911217443.5A discloses a method for constructing an industry electricity demand forecasting model and a forecasting method, and a method for constructing an industry electricity demand forecasting model: according to a set industry electricity correlation analysis program, the first dominant factor of industry electricity growth is obtained; wherein the industry electricity correlation analysis program is a combination of an association rule mining program and a principal component analysis program; according to a pre-established historical industry electricity correlation analysis data set, the second dominant factor of industry electricity growth is extracted; according to the first dominant factor and the second dominant factor, the ARIMA model and the multivariate regression model are respectively corrected and coupled to obtain an industry electricity demand forecasting model. This invention does not consider the impact of factors such as installed capacity, power generation, population, number of users, export volume, and import volume on energy consumption prediction.

[0005] Therefore, there is an urgent need for a long-term net energy consumption forecasting method that can take into account factors such as installed capacity, power generation, population, number of users, export volume, import volume, etc. when forecasting energy consumption, so as to improve the accuracy of energy consumption forecasting. Summary of the invention

[0006] In order to solve the deficiencies in the prior art, the present invention provides a long-term net energy consumption prediction method and system based on an adaptive analysis model, which takes into account influencing factors such as installed capacity, power generation, population, number of users, export volume, import volume, etc., and constructs an adaptive analysis model.

[0007] The present invention adopts the following technical solution.

[0008] On the one hand, the present invention provides a method for predicting long-term net energy consumption based on an adaptive analysis model, characterized in that the method comprises:

[0009] Step 1: Collect energy consumption data to form an energy consumption input sequence;

[0010] Step 2: Construct a multiple regression model based on the energy consumption input sequence constructed in step 1;

[0011] Step 3: Based on the multiple regression model constructed in step 2 and the energy consumption input sequence obtained in step 1, the predicted value of net energy consumption is obtained, and the multiple regression model with the smallest error between the predicted value and the actual value of net energy consumption is selected as the optimal multiple regression model;

[0012] Step 4: Based on the energy consumption input sequence constructed in step 1, an ARIMA-BO-LSSVM model is constructed, and the ARIMA-BO-LSSVM model is trained based on the training sample sequence, and the net energy consumption is predicted according to the trained ARIMA-BO-LSSVM model;

[0013] Step 5: Compare the predicted value of net energy consumption of the optimal multiple regression model obtained in step 3 with the predicted value of net energy consumption of the ARIMA-BO-LSSVM model, and select the model closest to the actual net energy consumption as the prediction model for long-term net energy consumption.

[0014] More preferably,

[0015] In step 1, the energy consumption input sequence includes installed capacity, total power generation, population data, electricity import and export value, and total number of users in the historical time period.

[0016] More preferably,

[0017] In step 2, the formula for constructing the multiple regression model is as follows:

[0018] Y=β 0 +β 1 X1+β 2 X2+β 3 X3+β 4 X4+β 5 X5+∈

[0019] Where Y represents the predicted value of net energy consumption, β 0 is the intercept term, X1-X5 are all standardized variables, X1 represents installed capacity, X2 represents total power generation, X3 represents population data, X4 represents power import and export, X5 represents total number of users, β 1 , β 2 , β3 , β 4 , β 5 represents the partial regression coefficient of each variable, ∈ represents the error term, which is a zero-mean random variable.

[0020] More preferably,

[0021] In step 4, the training sample sequence includes an energy consumption input sequence and a net energy consumption corresponding to the energy consumption input sequence.

[0022] More preferably,

[0023] In step 4, the training of the ARIMA-BO-LSSVMM model includes:

[0024] Step 401: using the ARIMA model to divide the training sample sequence into a trend component and a residual component of the net energy consumption, and using the ARIMA model to fit the trend component of the net energy consumption to obtain a corresponding trend component prediction result;

[0025] Step 402: fitting the residual component of the net energy consumption using the BO-LSSVM model to obtain the corresponding residual component prediction value;

[0026] Step 403: superimpose the trend component prediction result of the net energy consumption and the residual component prediction result to obtain the corresponding net energy consumption prediction value.

[0027] More preferably,

[0028] In step 401, the trend component of the net energy consumption is fitted using the ARIMA model, and the fitting formula of the trend component is as follows:

[0029]

[0030] Among them, Y t represents the data to be fitted, Y t-1 To Y t-p Indicates the value of the data to be fitted at p time points before the current time t. arrive Represents the autoregressive coefficient, which reflects the influence of past observations on the current value and is obtained through fitting; θ 1 to θ p Represents the moving average coefficient, reflecting the prediction error ε at the past p time points t-1 -ε t-p The degree of influence on the current predicted value is obtained through fitting, ε t t represents the difference or residual between the predicted value and the actual observed value, and c is a fixed value that represents the average level or benchmark value of the time series data.

[0031] More preferably,

[0032] In step 402, the residual component of the net energy consumption is fitted using the BO-LSSVM model to obtain the corresponding residual component prediction value, including: calculating the prediction value according to the following formula:

[0033]

[0034] Among them, x i represents the data input of the i-th sample, σ represents the “tightness” of the data mapping into the high-dimensional space, a i represents one of the solutions of the linear equation. In LSSVM, the value can be obtained by solving the linear equation system; b is another solution of the linear equation, which is a bias term of the model and is used to adjust the intercept of the prediction function; β is a parameter that needs to be adjusted, which is used to control the influence of different terms in the model, K represents the kernel function, μ represents the average function in the Gaussian process, and l represents the number of iterations. At the same time, the present invention provides a prediction system for a long-term net energy consumption prediction method based on an adaptive analysis model, and the system includes:

[0035] The net energy consumption prediction module is used to select the multiple regression model with the smallest error between the predicted value and the actual value of the net energy consumption as the optimal multiple regression model based on the preset multiple regression model and its corresponding energy consumption input sequence; based on the ARIMA-BO-LSSVM model and its corresponding energy consumption input sequence, obtain the net energy consumption prediction value of the ARIMA-BO-LSSVM model;

[0036] The net energy consumption prediction model screening module is used to compare the predicted value of net energy consumption of the optimal multiple regression model with the predicted value of net energy consumption of the ARIMA-BO-LSSVM model, and screen out the model closest to the actual net energy consumption for the prediction of long-term net energy consumption.

[0037] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in any one of the methods for predicting long-term net energy consumption based on an adaptive analysis model.

[0038] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in any one of the methods for predicting long-term net energy consumption based on an adaptive analysis model are implemented.

[0039] A terminal comprises a processor and a storage medium; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the steps of any one of the methods.

[0040] A computer-readable storage medium stores a computer program, which implements the steps of any one of the methods when executed by a processor.

[0041] The beneficial effects of the present invention are that, compared with the prior art, the present invention establishes an adaptive prediction model that takes into account the influencing factors of installed capacity, power generation, population, number of users, export volume, and import volume, thereby improving the model accuracy; adopts an ARIMA-BO-LSSVM hybrid model to predict energy consumption, applies the ARIMA model to capture the linear part hidden in the electricity consumption data, and applies the BO-LSSVM model to capture the nonlinear part existing in the data, thereby improving the prediction accuracy; adopts the BO model to select a hyperparameter combination, thereby improving the performance of the LSSVM model, and thus improving the accuracy of the prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0043] Figure 1 is a flow chart of a method for predicting long-term net energy consumption based on an adaptive analysis model according to an embodiment of the present invention;

[0044] Figure 2 is a schematic diagram of the structure of a long-term net energy consumption prediction system based on an adaptive analysis model according to an embodiment of the present invention;

[0045] Figure 3 is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in this application are only embodiments of a part of the present invention, rather than all embodiments. Based on the spirit of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the protection scope of the present invention.

[0047] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0048] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0049] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0050] Figure 1 : is a flow chart of a method for predicting long-term net energy consumption based on an adaptive analysis model in an embodiment of the present invention. The method for predicting long-term net energy consumption based on an adaptive analysis model in this embodiment includes:

[0051] Step 1: Collect energy consumption data and form an energy consumption input sequence.

[0052] In one embodiment of the present invention, in step 1, the energy consumption input sequence includes installed capacity, total power generation, population data, power import and export amount and total number of users in a historical period.

[0053] Step 2: Construct a multiple regression model based on the energy consumption input sequence constructed in step 1.

[0054] In one embodiment of the present invention, the construction of the multiple regression model includes defining dependent variables and independent variables, wherein the dependent variable includes net energy consumption Y, and the independent variables include installed capacity X1, total power generation X2, population data X3, power import and export amount X4, and total number of users X5; in this embodiment, it is assumed that there is a linear relationship between net energy consumption and each influencing factor, and the multiple regression model can be expressed as:

[0055] Y=β 0 +β 1 X1+β 2 X2+β 3 X3+β 4 X4+β 5 X5+∈ (1)

[0056] Among them, Y represents the predicted value of net energy consumption, X1-X5 are all standardized variables, X1 represents installed capacity, X2 represents total power generation, X3 represents population data, X4 represents electricity import and export, X5 represents the total number of users, β 0 is the intercept term, β 1 , β 2 , β 3 , β 4 , β 5is the coefficient of each variable, and ∈ is the error term. In the embodiment of the present invention, all independent variables and dependent variables are standardized, that is, the variable is subtracted from its mean and then divided by the standard deviation. The standardization formula is: Z = (X-μ) / σ, where Z is the standardized variable, X is the original variable, μ is the mean of the variable, and σ is the standard deviation of the variable. Step 3: Based on the multiple regression model constructed in step 2 and the energy consumption input sequence obtained in step 1, the predicted value of net energy consumption is obtained, and the multiple regression model with the smallest error between the predicted value and the actual value of net energy consumption is selected as the optimal multiple regression model.

[0057] In one embodiment of the present invention, the installed capacity X1, the total power generation X2, the population data X3, the power import and export amount X4 and the total number of users X5 are obtained, and the predicted value of the net energy consumption based on the multivariate regression model is obtained according to formula (1).

[0058] Input step 4: Based on the energy consumption input sequence constructed in step 1, an ARIMA-BO-LSSVM model is constructed, and the ARIMA-BO-LSSVM model is trained based on the training sample sequence. The net energy consumption is predicted based on the trained ARIMA-BO-LSSVM model.

[0059] In one embodiment of the present invention, based on the ARIMA-BO-LSSVM model and its corresponding energy consumption input sequence, a predicted value of net energy consumption of the ARIMA-BO-LSSVM model is obtained.

[0060] In this embodiment, the ARIMA model is a model that can fit a time series in order to better understand or predict the future values ​​of the series; in addition, this is a simple and popular linear method; the ARIMA model consists of two parts: autoregression (AR) and moving average (MA); the formula of the AR part is as follows:

[0061] f t =u t +a 1 f t-1 +...+a p f t-p +c (2)

[0062] Among them, f t represents the predicted value of the trend component at time t, p is the order of AR, u t represents the white noise (error) at time t, c is a constant, a 1 ...a p is the AR parameter.

[0063] The MA part formula is as follows:

[0064] ft =u t +m 1 u t-1 +...+m q u t-q +μ (3)

[0065] Among them, u t ,u t-1 ,...,u t-q is the error term, μ is the actual value (y t ) expectations, m 1 ,...,m q is the MA parameter, and q is the order of MA. These models are integrated with the same training data to create an ARIMA (p, q) model as follows:

[0066] f t =u t +a 1 f t-1 +...+a p f t-p +m 1 u t-1 +...+m q u t-q +c (4)

[0067] Where, is the order of the autoregressive (AR) term, and q is the order of the moving average term (MA). In this embodiment, μ represents the expected value (mean) of the time series. If the time series has been stabilized by steps such as difference, μ can be regarded as zero, that is, it no longer needs to be explicitly included in the model. In the ARIMA model, the mean of the time series has been eliminated due to the difference operation.

[0068] The basic assumption of the ARIMA model is that the time series data includes statistical stationarity, that is, the statistical properties of the measurement, such as autocorrelation, variance, and mean remain unchanged; if the training data is not stationary, then different data are needed to convert the ARIMA model to stationary; in this case, the model is ARIMA(p, d, q), where d is the degree of difference.

[0069] The LSSVM model was first proposed by Vapnik. SVM is an important machine learning technology used to solve the problems of nonlinear classification and function estimation. This technology maps data to a high-dimensional space (future space) and constructs an optimal separation hyperplane in this space. The quality of the SVM solution does not directly depend on the dimension of the input space. SVM is quite powerful in nonlinear regression prediction problems. LSSVM is an improvement on SVM. The performance of LSSVM is superior to traditional SVM in terms of fast convergence, simple calculation and high accuracy. On the other hand, in many applications, it is very difficult to select appropriate LSSVM parameters, and the selection of LSSVM parameters will affect the regression accuracy.

[0070] LSSVM simplifies the SVM procedure. LSSVM uses an organized least squares function and equality constraints to construct a linear system that satisfies the Karush-Kuhn-Tucker (KKT) conditions to obtain an optimal solution so that the regression problem can be solved by a linear system of equations. LSSVM is often used in control theory, classification, and regression problems. This model briefly introduces LSSVM for regression. The least squares support vector regression (LSSVR) method approximates an unknown function by using samples of a given training data series. The formula of the linear SVM regression algorithm is as follows:

[0071] y=f(x)=w T φ(x)+b (5)

[0072] in, φ(.): n → nh It is a mapping of the high-dimensional feature space. w represents the normal vector of the hyperplane C, which is perpendicular to the vector of the hyperplane and determines the direction of the hyperplane. b represents the intercept, which is used to determine the position of the hyperplane. The LSSVM method regards the regression problem as the following optimization problem:

[0073]

[0074] Wherein, formula (6) is constrained by:

[0075] y i =w T φ(x i )+b+e i ,i=1,2,...l (7)

[0076] where is the prediction error term for data point i, x i represents the data input of the i-th sample, γ is the regularization constant, φ(x i ) is an infinite-dimensional feature map, e iRepresents the prediction error term of data point i, and introduces the Lagrangian function:

[0077]

[0078] in is the Langer-Langer multiplier, which can be positive or negative in the LSSVM formula, where the optimality condition for the linear system of the Karush-Kuhn-Tucker (KKT) condition is as follows:

[0079]

[0080] For i = 1, 2, 3...lk(x i ,x j ) specifies a kernel function as follows:

[0081] k(x i ,x j )=φ(x i ) T φ(x j ) (10)

[0082] Among them, x i ,x j Respectively represent the feature vectors of the i-th and j-th data points, and the kernel function k(x i ,x j ) is equal to φ(x i ) and φ(x j ) i and x j The inner product of a vector, φ(x i ) and φ(x j ) represents an infinite-dimensional feature map.

[0083] The basic characteristics of the kernel function are derived from Mercer's theorem. Applicable kernel functions must satisfy Mercer's conditions, as shown in the following formula. Linear and polynomial functions are used as kernel functions in many applications:

[0084] k(x i ,x j ) = exp(-||x i -x j || 2 / σ 2 )(RBF) (11)

[0085] Where RBF represents radial basis function, σ represents the “tightness” of data mapping into high-dimensional space, and the Mercer condition is valid for all values ​​in the case of radial basis function and for positive t values ​​in the case of polynomial function. Therefore, K = (k ij )|x| ,k ij = k(x i ,x j ),V=diag(1 / γ,1 / γ,...,1 / γ), therefore, the LSSVM regression model can be described as:

[0086]

[0087] Among them, A = K + V, by solving formula (15), we can find the regression model in formula (7). Using the Lagrangian method, the LSSVM model can be expressed as:

[0088]

[0089] Among them, a i and b is the solution of the linear equation. ARIMA-BO-LSSVM prediction model:

[0090] Factors such as economic growth have a significant impact on the changing trend of net energy consumption. Therefore, it is necessary to explore the nonlinear component of the net energy consumption time series to improve the prediction accuracy. This embodiment uses LSSVM to fit the nonlinear part of the long-term net energy consumption series.

[0091] However, the performance of the LSSVM algorithm depends largely on the selection of its hyperparameters, such as regularization parameters and kernel function parameters. Therefore, the present invention introduces the BO algorithm based on LSSVM; BO is an optimization method based on a probability model, which can effectively search for the optimal solution in the hyperparameter space. By balancing exploration and utilization, BO can find the hyperparameter combination that improves the model performance within a small number of iterations, thereby improving the accuracy of the LSSVM model; wherein, the BO algorithm is mainly used to find the best hyperparameter combination in the parameter space, thereby improving the performance of the LSSVM model; the BO algorithm determines the search range of the LSSVM model hyperparameters; the BO algorithm constructs a probability model to describe the relationship between hyperparameters and model performance. The BO algorithm predicts the performance of unknown points in the parameter space through this probability model, and finds the most suitable hyperparameter combination to optimize the LSSVM model.

[0092] In this embodiment, the energy consumption input sequence of the ARIMA-BO-LSSVM model consists of installed capacity, total power generation, population data, power import and export amounts, and total number of users in a historical period.

[0093] The ARIMA-BO-LSSVM model is trained by a set of training sample sequences; the training sample sequence consists of an energy consumption input sequence and its corresponding net energy consumption.

[0094] In one embodiment of the present invention, the training of the ARIMA-BO-LSSVM model includes:

[0095] Step 401: Use the ARIMA model to divide the training sample sequence into a trend component and a residual component of the net energy consumption, and use the ARIMA model to fit the trend component of the net energy consumption to obtain the corresponding trend component prediction result; for example, use the ARIMA method to fit the trend component of the training sample sequence, and create a series of trend component prediction values ​​(f t ), by taking the actual value of net energy consumption (y t ) minus the predicted value of the trend component to obtain a series of irregular component values ​​(e r ):

[0096] e r =y t -f t (14)

[0097] Step 402: Fitting the residual component of the net energy consumption using the BO-LSSVM model to obtain the corresponding residual component prediction result;

[0098] Step 403: Add the trend component prediction result of the net energy consumption and the residual component prediction result to obtain the corresponding net energy consumption prediction value.

[0099] In one embodiment of the present invention, in step 401, fitting the trend component of net energy consumption using the ARIMA model includes:

[0100] The fitting formula for the trend component is as follows:

[0101]

[0102] Among them, Y t represents the data to be fitted, Y t-1 To Y t-p Indicates the value of the data to be fitted at p time points before the current time t. arrive Represents the autoregressive coefficient, which reflects the influence of past observations on the current value and is obtained through fitting; θ 1 to θ p Represents the moving average coefficient, reflecting the prediction error ε at the past p time points t-1 -ε t-p The degree of influence on the current predicted value is obtained through fitting, ε trepresents the difference or residual between the predicted value and the actual observed value, and c is a fixed value, which represents the average level or benchmark value of the time series data. In one embodiment of the present invention, fitting the residual component of the net energy consumption using the BO-LSSVM model includes:

[0103] Fitting irregular component values ​​using BO-LSSVM r , and generate a series of predicted values ​​f(x) according to the following formula:

[0104]

[0105] Among them, x i represents the data input of the i-th sample, σ represents the “tightness” of the data mapping into the high-dimensional space, a i and b represent solutions to linear equations, β is a constant, K represents the kernel function, and μ represents the average function in the Gaussian process;

[0106] The final predicted value of net energy consumption (y') is calculated as shown in equation (17):

[0107] y'=f t +f(x i ) (17)

[0108] In this embodiment, the predicted value of net energy consumption of the optimal multiple regression model is compared with the predicted value of net energy consumption (y') of the ARIMA-BO-LSSVM model, and the model closest to the actual net energy consumption is selected for the prediction of long-term net energy consumption.

[0109] In one embodiment of the present invention, three different models are constructed in the multivariate linear regression analysis; these models respectively consider the total power generation (GEG), population (P), installed capacity (IC), import (I), export (E) and the total number of users (TS), as shown in Table 1.

[0110] In this embodiment, in order to establish an adaptive analysis model taking the above-mentioned influencing factors into account, six types of data were collected; including historical population data from 2000 to 2020, electricity import and export amounts in the same period, installed capacity and total power generation data, net energy consumption and number of users; the data used in the model were divided into three groups, and in all prediction models, 85% of the input data was used for training, 11% of the input data was used for testing, and 4% of the input data was used for verification.

[0111] In order to evaluate the effectiveness of the proposed model in work, the present invention calculates the mean square error (MSE), root mean square error (RMSE), and mean absolute percentage error (MAPE); in addition, when selecting the optimal ARIMA model, the Akaike information criterion (AIC), Bayesian information criterion (BIC, or Schwartz criterion), and Hannan-Quinn criterion (HQC) values ​​are used for comparison.

[0112] Since it is a big problem to determine the correlation and the degree of correlation between many variables, regression is one of the most commonly used techniques to study the correlation between variables. In the first stage of analysis, multiple linear regression analysis was used to model the net energy consumption. Three models considering different influencing factors are as follows:

[0113] Model 1: y = b 1 x 2 +c 1 x 3 +f 1

[0114] Model 2: y = a 1 x 1 +b 2 x 2 +c 2 x 3 +d 1 x 4 +e 1 x 5 +g 1 x 6 +f 2

[0115] Model 3: y = a 2 x 1 +b 3 x 2 +c 3 x 3 +d 2 x 4 +f 3

[0116] Where y represents the predicted net energy consumption, a 1 -a 2 , b 1 -b 3 、c 1 -c 3 d 1 -d 2 、f 1 -f 3 、e 1 , g 1is the regression coefficient, x represents the six independent variables that are predictors of y (Table 1), and the coefficients of adjusted R2 for models 1, 2, and 3 are 0.988, 0.994, and 0.996, respectively.

[0117] Table 1 Summary of variables

[0118] variable describe <![CDATA[x 1 ]]> Installed capacity <![CDATA[x 1 ]]> Total power generation <![CDATA[x 3 ]]> population <![CDATA[x 4 ]]> Total number of users <![CDATA[x 5 ]]> exit <![CDATA[x 6 ]]> import <![CDATA[x 6 ]]> Net energy consumption

[0119] Table 2 Statistical values ​​of different regression models

[0120]

[0121] As can be seen from Table 2, the training errors of Model 1 and Model 2 are low, but the test errors of Model 1 and Model 2 are high. On the other hand, the training error of Model 3 is high, but the test error rate is low, with an RMSE of 2.54; in addition, the higher the adjusted R2 coefficient of Model 3, the more appropriate Model 3 is than the other models; therefore, the regression equation of Model 3 was considered when comparing the results.

[0122] The net energy consumption has a linear distribution from year to year, except for some recessionary periods; no considerable fluctuations and seasonality were observed in the historical consumption data investigated. To fit the ARIMA model to a stationary time series, the annual net energy consumption data needs to be periodically differencing to become stationary. The order of the ARIMA model is identified using the autocorrelation function (ACF) and partial autocorrelation function (PACF). The parameters of the ARIMA model are estimated by a maximum likelihood function, and the goodness of fit of the model is attempted on the model residuals. The MLR analysis predicts the net energy consumption based on historical data through a combination of multiple factors. The models take into account different influencing factors and calculate the regression coefficients, which are analyzed by the adjusted R 2 The coefficients are used to measure the goodness of fit of the model. When evaluating which provisional model best fits the data, AIC and BIC are used to select the best fitting model. The goal is to select the model sequence that produces the minimum values ​​of AIC and BIC. The ARIMA model is applied using the MATLAB 2010b programming language, which can be used to create econometric models.

[0123] Table 3 summarizes the statistical measures of the most successful ARIMA models; it can be concluded that the best fitting model for the time series data is ARIMA(1,1,2) due to its lower values ​​of AIC, BIC, and HQC; in addition, the R2 (simple and adjusted) statistical criteria used to evaluate the fit of the regression model have higher values ​​than the other criteria; the ARIMA(1,1,2) model represents one lag of the dependent variable, the variables used are first difference stationary and two lags of the error term.

[0124] Table 3 Statistics of ARIMA model

[0125] ARIMA Model R2 Adjusted R2 p-value AIC BIC HQC ARIMA(1,1,2) 0.852 0.85 0.00 -3.3705 -3.2501 -3.3256 ARIMA(1,1,0) 0.842 0.84 0.00 -3.3529 -3.2325 -3.3080 ARIMA(1,1,1) 0.839 0.821 0.00 -3.3314 -3.1708 -3.2715

[0126] Among them, R2: Coefficient of Determination, used to measure the goodness of model fit, represents the ratio of the variability explained by the model to the total variability, R 2 The value of is between 0 and 1, and the closer the value is to 1, the stronger the explanatory power of the model.

[0127] p-value: p-value is used to determine whether the statistical results are significant. If the p-value of a coefficient is less than a certain significance level, the coefficient is considered to be statistically significant, and it makes sense to include the coefficient in the model.

[0128] AIC: Akaike Information Criterion, is a criterion for measuring the goodness of model fit. It takes into account both the fit and complexity (i.e., the number of parameters) of the model. The smaller the AIC value, the simpler the model is in fitting the data.

[0129] BIC: Bayesian Information Criterion, a criterion for measuring the goodness of model fit. The smaller the BIC value, the better the model.

[0130] HQC: Hannan-Quinn Information Criterion, a criterion for measuring the goodness of model fit. The smaller the HQC value, the better the model.

[0131] The residual components in the estimation obtained by the ARIMA model are separated and BO-LSSVM analysis is performed. In BO-LSSVM analysis, data normalization is performed before the training process begins, and each input space is normalized to [-1, +1] using the following formula:

[0132]

[0133] Among them, x n , x represent normalized data and original data respectively, x min and x max They respectively represent the minimum and maximum values ​​in the original data. For the BO-LSSVM model, the MATLAB platform is used to implement and analyze the algorithm. The radial basis function (RBF) is used as the kernel function. The RBF parameters selected in the BO-LSSVM analysis should also reflect the distribution of the input values ​​of the training data, otherwise, it may cause serious changes in the optimal regression line. Subsequently, the prediction of the ARIMA (1,1,2) model is combined with the residual value estimated by the BO-LSSVM model to obtain the final result of the proposed model.

[0134] The prediction results using the ARIMA-BO-LSSVM hybrid method are compared with other methods; it is concluded that this model is better than artificial neural networks, support vector machines, and optimized gray models. The results for 2021 and 2022 obtained by the model of the present invention are closer to the actual values, with a smaller error value and lower than previous models, as shown in Table 4.

[0135] Table 4 Comparison with similar studies of the present invention

[0136]

[0137] Figure 2 is a schematic diagram of the structure of a long-term net energy consumption prediction system based on an adaptive analysis model in an embodiment of the present invention, such as Figure 2 As shown, the long-term net energy consumption prediction system based on the adaptive analysis model in this embodiment includes:

[0138] The net energy consumption prediction module 201 is used to select the multiple regression model with the smallest error between the predicted value and the actual value of the net energy consumption as the optimal multiple regression model based on the preset multiple regression model and its corresponding energy consumption input sequence; based on the ARIMA-BO-LSSVM model and its corresponding energy consumption input sequence, obtain the net energy consumption prediction value of the ARIMA-BO-LSSVM model;

[0139] The net energy consumption prediction model screening module 202 is used to compare the predicted value of the net energy consumption of the optimal multiple regression model with the predicted value of the net energy consumption of the ARIMA-BO-LSSVM model, and screen out the model closest to the actual net energy consumption for long-term net energy consumption prediction.

[0140] It should be noted here that the long-term net energy consumption prediction system based on the adaptive analysis model in this embodiment corresponds one to one with each step in the long-term net energy consumption prediction method based on the adaptive analysis model, and the specific implementation process is the same, which will not be described in detail here.

[0141] Reference Figure 3 , a schematic diagram of an electronic device is given. It should be noted that, Figure 3 The electronic device 300 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0142] like Figure 3As shown, the electronic device 300 includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage part 308 into the random access memory (RAM) 303. In the RAM 303, various programs and data required for system operation are also stored. The central processing unit 301, the ROM 302 and the RAM 303 are connected to each other through a bus 304. An input / output (I / O) interface 305 is also connected to the bus 104.

[0143] The following components are connected to the I / O interface 305: an input section 306 including a keyboard, a mouse, etc.; an output section 307 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a local area network (LAN) card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed so that a computer program read therefrom is installed into the storage section 308 as needed.

[0144] When the central processing unit 301 in the electronic device of this embodiment executes the program, the following is achieved: Figure 1 The steps in the method for predicting long-term net energy consumption based on an adaptive analytical model are shown.

[0145] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer readable medium, the computer program including a computer program for executing Figure 1 In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 309, and / or installed from the removable medium 311. When the computer program is executed by the central processing unit 301, various functions defined in the apparatus of the present application are executed.

[0146] in, Figure 1 The computer program instructions corresponding to the method shown may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0147] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0148] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention. The present disclosure may be a system, method and / or computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for enabling a processor to implement various aspects of the present disclosure.

[0149] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0150] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0151] The computer program instructions for performing the operation of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be customized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for predicting long-term net energy consumption based on an adaptive analytical model, characterized in that: The method comprises: Step 1: Collect energy consumption data to form an energy consumption input sequence; Step 2: Construct a multiple regression model based on the energy consumption input sequence constructed in step 1; Step 3: Based on the multiple regression model constructed in step 2 and the energy consumption input sequence obtained in step 1, the predicted value of net energy consumption is obtained, and the multiple regression model with the smallest error between the predicted value and the actual value of net energy consumption is selected as the optimal multiple regression model; Step 4: Based on the energy consumption input sequence constructed in step 1, an ARIMA-BO-LSSVM model is constructed, and the ARIMA-BO-LSSVM model is trained based on the training sample sequence, and the net energy consumption is predicted according to the trained ARIMA-BO-LSSVM model; Step 5: Compare the predicted value of net energy consumption of the optimal multiple regression model obtained in step 3 with the predicted value of net energy consumption of the ARIMA-BO-LSSVM model, and select the model closest to the actual net energy consumption as the prediction model for long-term net energy consumption.

2. The method for predicting long-term net energy consumption based on an adaptive analytical model according to claim 1, characterized in that: In step 1, the energy consumption input sequence includes installed capacity, total power generation, population data, electricity import and export value, and total number of users in the historical time period.

3. The method for predicting long-term net energy consumption based on an adaptive analytical model according to claim 1, characterized in that: In step 2, the formula for constructing the multiple regression model is as follows: Y=β0+β1X1+β2X2+β3X3+β4X4+β5X5+∈ Among them, Y represents the predicted value of net energy consumption, β0 is the intercept term, X1-X5 are all standardized variables, X1 represents installed capacity, X2 represents total power generation, X3 represents population data, X4 represents electricity import and export amount, X5 represents the total number of users, β1, β2, β3, β4, β5 represent the partial regression coefficients of their respective variables, ∈ represents the error term, which is a zero-mean random variable.

4. The method for predicting long-term net energy consumption based on an adaptive analytical model according to claim 1, characterized in that: In step 4, the training sample sequence includes an energy consumption input sequence and a net energy consumption corresponding to the energy consumption input sequence.

5. The method for predicting long-term net energy consumption based on an adaptive analytical model according to claim 2, characterized in that: In step 4, the training of the ARIMA-BO-LSSVMM model includes: Step 401: using the ARIMA model to divide the training sample sequence into a trend component and a residual component of the net energy consumption, and using the ARIMA model to fit the trend component of the net energy consumption to obtain a corresponding trend component prediction result; Step 402: fitting the residual component of the net energy consumption using the BO-LSSVM model to obtain the corresponding residual component prediction value; Step 403: superimpose the trend component prediction result of the net energy consumption and the residual component prediction result to obtain the corresponding net energy consumption prediction value.

6. The method for predicting long-term net energy consumption based on an adaptive analytical model according to claim 5, characterized in that: In step 401, the trend component of the net energy consumption is fitted using the ARIMA model, and the fitting formula of the trend component is as follows: Among them, Y t represents the data to be fitted, Y t-1 To Y t-p Indicates the value of the data to be fitted at p time points before the current time t. arrive Represents the autoregressive coefficient, which reflects the influence of past observations on the current value and is obtained through fitting; θ1 to θ p Represents the moving average coefficient, reflecting the prediction error ε at the past p time points t-1 -ε t-p The degree of influence on the current predicted value is obtained through fitting, ε t t represents the difference or residual between the predicted value and the actual observed value, and c is a fixed value that represents the average level or benchmark value of the time series data.

7. The method for predicting long-term net energy consumption based on an adaptive analytical model according to claim 5, characterized in that: In step 402, the residual component of the net energy consumption is fitted using the BO-LSSVM model to obtain the corresponding residual component prediction value, including: calculating the prediction value according to the following formula: Among them, x i represents the data input of the i-th sample, σ represents the "tightness" of the data mapping into the high-dimensional space, a i It represents one of the solutions of the linear equation. In LSSVM, the value can be obtained by solving the linear equation system. b is another solution of the linear equation, which is a bias term of the model and is used to adjust the intercept of the prediction function. β is a parameter that needs to be adjusted. It is used to control the influence of different terms in the model. K represents the kernel function, μ represents the average function in the Gaussian process, and l represents the number of iterations.

8. A long-term net energy consumption prediction system according to the prediction method according to any one of claims 1 to 7, characterized in that: The system includes: a net energy consumption prediction module, which is used to screen out a multiple regression model with the smallest error between the predicted value and the actual value of the net energy consumption as the optimal multiple regression model based on a preset multiple regression model and its corresponding energy consumption input sequence; and obtain the net energy consumption prediction value of the ARIMA-BO-LSSVM model based on the ARIMA-BO-LSSVM model and its corresponding energy consumption input sequence; The net energy consumption prediction model screening module is used to compare the predicted value of net energy consumption of the optimal multiple regression model with the predicted value of net energy consumption of the ARIMA-BO-LSSVM model, and screen out the model closest to the actual net energy consumption for the prediction of long-term net energy consumption.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the long-term net energy consumption prediction method based on an adaptive analysis model as described in any one of claims 1 to 7 are implemented.

10. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the method for predicting long-term net energy consumption based on an adaptive analysis model as described in any one of claims 1 to 7 are implemented.

11. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1-7.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images