Method and device for predicting national product oil demand quantity

Abstract

The invention discloses a method and a device for predicting national product oil demand quantity. The method comprises the following steps: acquiring N external variables related to product oil demand in the last year; converting the N variables into m principal component variables by adopting a principal component analysis method, wherein m is a positive integer smaller than N; and according to the determined m principal component variables, determining the national product oil demand quantity in the next year by adopting a pre-established national product oil demand prediction model.

Description

Technical Field

[0001] The present invention relates to the field of petroleum technology, and in particular to a method and device for predicting the national demand for refined oil. Background Art

[0002] Domestic refined oil consumption is the focus of oil companies. The changing trend of long-term demand involves the planning and layout of production capacity with a long construction period, short-term demand involves the utilization and deployment of production capacity, and regional demand involves the deployment of resources and the grasp of growth points. Accurately predicting changes in market demand is conducive to oil companies to formulate reasonable development plans, reasonably arrange the layout of production capacity and sales facilities, and formulate effective production and sales plans, which can help oil companies grasp growth and improve their market response capabilities.

[0003] There are many domestic refined oil consumption industries, involving more than ten industry fields such as civil automobiles, industry, construction, road transportation, railway transportation, water transportation, agricultural machinery, air transportation, and commercial applications. The influencing factors are numerous and complex, and are also affected by non-fundamental factors such as industry policies and upstream and downstream product prices. Therefore, accurately predicting changes in refined oil demand has always been an industry problem. Summary of the invention

[0004] The present application provides a method and device for predicting the national demand for refined oil products. The method can more accurately predict the national demand for refined oil products based on principal component analysis and a national refined oil demand prediction model.

[0005] The present application provides a method for predicting the national demand for refined oil products, the method comprising:

[0006] Obtain N external variables related to the demand for refined oil products in the previous year;

[0007] The principal component analysis method is used to transform N variables into m principal component variables, where m is a positive integer less than N;

[0008] Based on the determined m principal component variables, the national refined oil demand for the next year is determined using a pre-established national refined oil demand forecasting model.

[0009] In an exemplary embodiment, the external variables include one or more of the following: automobile data, transportation data, and industrial data.

[0010] In an exemplary embodiment, the method of converting N variables into m principal component variables by using a principal component analysis method includes:

[0011] The linear combination that maximizes the variance of N variables is determined by using the Lagrange multiplier method;

[0012] The m principal component variables are determined according to the determined linear combinations.

[0013] In an exemplary embodiment, after determining m principal component variables according to the determined linear combination, the method further includes:

[0014] The critical value of smoothing is determined by the scree plot method, and the value corresponding to the critical value is used as the number of principal components to be extracted.

[0015] In an exemplary embodiment, the

[0016] The pre-established national refined oil demand forecasting model is used to determine the national refined oil demand for the next year, including:

[0017] Cluster analysis was conducted on the 31 provinces in the country according to relevant indicators;

[0018] The 31 provinces were divided into multiple groups based on the principle of similar characteristics;

[0019] Each group corresponds to a component oil demand forecasting model;

[0020] Determine the demand for refined oil products in the next year for each province based on the corresponding refined oil demand forecast model;

[0021] The next year's refined oil demand of all provinces will be summed up to determine the next year's national refined oil demand.

[0022] In an exemplary embodiment, cluster analysis of 31 provinces in the country according to relevant indicators includes:

[0023] The 31 provinces in the country are clustered according to the proportion of each industry in GDP, per capita GDP and per capita disposable income of residents;

[0024] Cluster analysis of 31 provinces was performed using the k-means algorithm.

[0025] In an exemplary embodiment, the component oil demand forecasting model is established by the following process:

[0026] Extract the principal components of the provinces included in this group;

[0027] Regression modeling was performed for this group separately;

[0028] The regression model is:

[0029]

[0030] i=1,...,N;t=1,...,T i

[0031] In the above model, N is the number of provinces included in this group; T is the time length; K is the independent variable constant; u is the residual term; the extracted principal component is taken as the independent variable x, and the gasoline and diesel consumption data y of each province is taken as the dependent variable.

[0032] In an exemplary embodiment, the method of determining the linear combination that maximizes the variance of N variables by using the Lagrange multiplier method includes:

[0033] Establish the relationship between the principal components and the original variables;

[0034] Among them, the relationship is:

[0035] z=Xu

[0036] Among them, the original variable X, Principal component variable Z;

[0037] By using the Lagrange multiplier method, we can determine the A linear combination of .

[0038] The present application also provides a device for predicting the national demand for refined oil, comprising: a memory and a processor;

[0039] The memory is used to store a program for predicting the national demand for refined oil products;

[0040] The processor is used to read and execute the program for predicting the national demand for refined oil, and perform the method for predicting the national demand for refined oil as described in any one of the above embodiments.

[0041] The present application also provides a computer storage medium, in which a program for predicting national refined oil demand is stored, and the program is configured to execute any of the methods in the above embodiments when running.

[0042] Compared with the related art, the embodiment of the present application provides a method and device for predicting the national demand for refined oil, the method comprising: obtaining N external variables related to the demand for refined oil in the previous year; converting the N variables into m principal component variables using the principal component analysis method, where m is a positive integer less than N; and determining the national demand for refined oil in the next year using a pre-established national demand for refined oil forecasting model based on the determined m principal component variables. Through the technical solution of the present invention, the national demand for refined oil can be more accurately predicted based on the principal component analysis and the national demand for refined oil forecasting model.

[0043] Other features and advantages of the present application will be described in the following description, and partly become apparent from the description, or be understood by implementing the present application. Other advantages of the present application can be realized and obtained by the schemes described in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings are used to provide an understanding of the technical solution of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the technical solution of the present application and do not constitute a limitation on the technical solution of the present application.

[0045] Figure 1 A flow chart of a method for predicting national refined oil demand according to an embodiment of the present application;

[0046] Figure 2 This is a schematic diagram of a device for predicting national refined oil demand according to an embodiment of the present application;

[0047] Figure 3 A schematic diagram of trend prediction modeling in some exemplary embodiments;

[0048] Figure 4 A schematic diagram of scree plot results in some exemplary embodiments;

[0049] Figure 5 A flow chart of a method for predicting national demand for refined oil products in some exemplary embodiments. DETAILED DESCRIPTION

[0050] The present application describes multiple embodiments, but the description is exemplary rather than restrictive, and it is obvious to those skilled in the art that there may be more embodiments and implementations within the scope of the embodiments described in the present application. Although many possible feature combinations are shown in the drawings and discussed in the specific embodiments, many other combinations of the disclosed features are also possible. Unless specifically limited, any feature or element of any embodiment may be used in combination with any other feature or element in any other embodiment, or may replace any other feature or element in any other embodiment.

[0051] The present application includes and contemplates combinations of features and elements known to those of ordinary skill in the art. The embodiments, features and elements disclosed in the present application may also be combined with any conventional features or elements to form a unique invention scheme defined by the claims. Any features or elements of any embodiment may also be combined with features or elements from other invention schemes to form another unique invention scheme defined by the claims. Therefore, it should be understood that any feature shown and / or discussed in the present application may be implemented individually or in any appropriate combination. Therefore, except for the limitations made according to the attached claims and their equivalents, the embodiments are not subject to other restrictions. In addition, various modifications and changes may be made within the scope of protection of the attached claims.

[0052] In addition, when describing representative embodiments, the specification may have presented the method and / or process as a specific sequence of steps. However, to the extent that the method or process does not rely on the specific order of the steps described herein, the method or process should not be limited to the steps of the specific order described. As will be understood by those of ordinary skill in the art, other sequences of steps are also possible. Therefore, the specific sequence of the steps set forth in the specification should not be interpreted as a limitation to the claims. In addition, the claims for the method and / or process should not be limited to the steps of performing them in the order written, and those skilled in the art can easily understand that these sequences can be changed and still remain within the spirit and scope of the embodiments of the present application.

[0053] Some technical forecasting methods for refined oil consumption mainly include:

[0054] 1) Elasticity coefficient method. This method is based on the ratio of the historical growth rate of refined oil consumption to the growth rate of GDP, and uses the GDP growth rate as the input condition to predict the growth rate of demand. The advantage is that it is simple and easy to use, and the calculation process only involves multiplication and division. The disadvantage is that the prediction results lack interpretability and the prediction accuracy is poor.

[0055] 2) Trend prediction model: According to the trend characteristics of the historical data of refined oil consumption changing with time or the absolute value of GDP or the number of cars, single variable regression models such as linear, quadratic curve, Gomperz curve, "S" curve, exponential growth, etc. are used to predict the future consumption value. The above models are all trend fitting models that come with Excel, that is, trend line fitting function. The advantage of this method is that it is simple and easy to use, and a small amount of data can be used to build a model. The disadvantage is that only a single variable is used, there is a lack of understanding of the mechanism of oil consumption, and only external predictions can be made based on existing trends, which cannot reflect the impact of changes in the external environment.

[0056] 3) Location quotient method. This method assumes that gasoline consumption and per capita gasoline consumption are similar between regions. Based on the relative level of gasoline consumption per unit of output value and per capita gasoline consumption in the region compared with the national level, given the future growth rate of gasoline consumption in each region, the demand in each region is calculated, and adjusted proportionally based on the national total forecast results. The advantage of this model is that it is simple and easy to use, and the model only contains division operations. The disadvantage is that it can only calculate the relative proportion of each province in the country, and the specific value depends on the forecast value of the national model.

[0057]

[0058] 4) Multiple regression method. Multiple linear regression is an improvement over the previous two methods. It selects as many explanatory factors as possible and uses regression methods to determine the impact of each factor on demand. However, due to the strong collinearity between a large number of independent variables, most variables will be filtered out, leaving only two or three independent variables in the final equation. In this case, the prediction of the dependent variable is almost equivalent to the prediction of the independent variable, which is a very difficult task, and the prediction accuracy is not guaranteed at all. More importantly, there is no systematic explanatory power.

[0059] In the field of refined oil demand forecasting, traditional methods have the following problems:

[0060] 1) Insufficient data feature mining. Methods such as elasticity coefficient and location quotient seek the direct relationship between economic growth and oil consumption growth, but do not analyze the impact of economic structure and industrial development on the demand for refined oil. Therefore, they are similar to black boxes; there are problems such as information overflow and poor accuracy.

[0061] 2) Information loss problem. Multiple linear regression is an improvement over the previous two methods. It selects as many explanatory factors as possible and determines the impact of each factor on demand through regression methods. However, this method cannot solve the problem of information overlap. For example, the destination of fixed asset investment includes the purchase of industrial products, transportation of industrial products, and personnel expenses. Among them, the purchase cost of industrial products will be converted into industrial added value. This causes information overlap between the two variables. This cannot be solved by the multiple regression method. Only one variable can be eliminated, resulting in information loss.

[0062] 3) The problem of predicting demand by province.

[0063] In view of the problems existing in the above-mentioned traditional technologies and the fact that there is currently no technical method at home and abroad that can systematically predict the demand for refined oil in all provinces, the inventors proposed a method that can more accurately predict the demand for refined oil in all provinces and the demand for refined oil in each province.

[0064] The disclosed embodiment provides a method for predicting the national demand for refined oil products. Figure 1 As shown, the method includes steps S100-S120, which are specifically as follows:

[0065] S100. Obtain N external variables related to the demand for refined oil products in the previous year;

[0066] S110. Using principal component analysis method to transform N variables into m principal component variables, m is a positive integer less than N;

[0067] S120. Based on the determined m principal component variables, the national refined oil demand for the next year is determined using a pre-established national refined oil demand forecasting model.

[0068] In an exemplary embodiment, the external variables include one or more of the following: automobile data, transportation data, and industrial data.

[0069] In an exemplary embodiment, a principal component analysis method is used to convert N variables into m principal component variables, including:

[0070] Step 1, determine the linear combination that maximizes the variance of N variables by using the Lagrange multiplier method;

[0071] Step 2: Determine m principal component variables based on the determined linear combinations.

[0072] In this embodiment, the specific implementation process of the principal component analysis method is as follows:

[0073] Combined with market research experience, 155 independent variables in three categories were included in the modeling indicator pool, and principal component analysis was performed on each group of independent variables. The process is as follows:

[0074] Assume that a given set of variables is, That is, there are p variables, and each variable has n samples.

[0075] For the convenience of analysis, it is assumed that the above variables are standardized variables. Principal component analysis is to find a linear combination that maximizes the variance of the original variables, that is,

[0076] z=Xu

[0077] The u vector is such that, Since X is a standardized variable, That is the correlation coefficient. So we have,

[0078] var(z)=u′Ru

[0079] A condition u′u=1 should be added here, that is, the unit vector, otherwise the variance will become infinite.

[0080] In summary, principal component analysis is to select u so that u′Ru is maximized and constrained to u′u=1.

[0081] In an exemplary embodiment, a linear combination that maximizes the variance of N variables is determined by using a Lagrange multiplier method, including:

[0082] Step 1. Establish the relationship between the principal component and the original variable;

[0083] Among them, the relationship is:

[0084] z=Xu

[0085] Among them, the original variable X, Principal component variable Z;

[0086] Step 2. Determine the Lagrange multiplier method to satisfy u′u=1 and A linear combination of .

[0087] In this embodiment, in order to determine the linear combination that maximizes the variance of N variables, the Lagrange multiplier method can be used, that is, first construct a Lagrange function, then set its first-order derivative equal to zero, and then solve the equation. That is,

[0088] L=u′Ru-λ(u′u-1)

[0089] Taking the first-order derivative,

[0090]

[0091] Let it be equal to zero, and we can solve it to get,

[0092] Ru=λu or, (R-λI)u=0

[0093] The above equation is actually the problem of solving the eigenvector u; where vector u is the eigenvector and the value of λ is the eigenvalue.

[0094] Since there are p variables in X, we can find a total of p such vectors, which can form a matrix, u = [u 1 ,u 2 ,…,u p ]′ is an element in .

[0095] Therefore, the principal component variable Z can be obtained, where Z = [z 1 ,z 2 ,…,z p ]

[0096] On the surface, it seems meaningless to transform p variables to get p variables, just that the Z vectors are orthogonal. But in fact, this transformation is very important, because in actual use, we will never use all the variables in Z, but only a few, that is, the principal component variables that are retained.

[0097] In an exemplary embodiment, after determining m principal component variables according to the determined linear combination, it is also necessary to determine the number of principal component variables to be retained, that is, to determine the smoothing critical value by the scree plot method, and use the value corresponding to the critical value as the number of principal components to be extracted. Figure 3As shown in the figure, the scree plot actually displays the eigenvalues ​​of the principal components in a graphical way, making it more intuitive to determine the number of principal component variables. Generally speaking, the number of principal component variables should be selected at the position where the scree plot begins to become smooth. The figure shows that the number of principal components becomes very smooth starting from 4, so it can be determined that 3 is a more appropriate number of principal component variables.

[0098] As shown in Figure 1 below, the number of principal components determined by the scree plot method improves the efficiency of the analysis. The number of principal components extracted is set to 3 in the model. By analyzing the variation range of each independent variable in three directions, the variance proportion explained by each principal component can be calculated. From the variance analysis table, it can be seen that the first principal component can explain 60.2% of the variance of all 15 independent variables, the first two principal components can explain 89.8% of the variance of all 15 independent variables, and the first three principal components can explain 97.5% of the variance of all 15 independent variables. The efficiency of principal component analysis is relatively high.

[0099] Table 1 Principal component variance analysis

[0100] Total Variance Explained

[0101]

[0102] Extraction Method: Principal Component Analysis

[0103] In addition, the model can further calculate the principal component load. The principal component load is the correlation coefficient between the principal component variable Z and the original variable X. From the principal component load table of the diesel model, it can be seen that the principal component variable FAC1 has the strongest correlation with lhzgy, 1hgdpb and lzgk, reaching more than 90%. The principal component variable FAC2 has the strongest correlation with lccdt and hpen; the principal component variable FAC3 has the strongest correlation with pbcgdpa.

[0104] Table 2 Principal component loading and table

[0105] Rotated Component Matrix a

[0106]

[0107] Extraction Method: Principal ComponentAnalysis.

[0108] Rotation Method: Varimax with Kaiser Normalization.

[0109] a.Rotation converged in 6 iterations.

[0110] In addition, by calculating the principal component coefficient matrix, it can be clearly seen that the principal component variables are obtained through the linear combination of the original variables, see Table 3.

[0111] Table 3 Principal component coefficients

[0112] Component Score Coefficient Matrix

[0113]

[0114] Extraction Method: Principal Component Analysis

[0115] Rotation Method: Varimax with Kaiser Norrnalization.

[0116] Cormponent Scores.

[0117] In an exemplary embodiment, the national refined oil demand forecasting model established in advance is used to determine the national refined oil demand for the next year, including:

[0118] The first step is to cluster the 31 provinces in the country according to relevant indicators;

[0119] The second step is to divide the 31 provinces into multiple groups according to the principle of similar characteristics;

[0120] Step 3: Each group corresponds to a component oil demand forecasting model;

[0121] Step 4: Determine the next year's refined oil demand for each province based on the corresponding refined oil demand forecast model;

[0122] Step 5: Sum up the demand for refined oil in all provinces in the next year to determine the national demand for refined oil in the next year.

[0123] In an exemplary embodiment, cluster analysis of 31 provinces in the country according to relevant indicators includes:

[0124] The 31 provinces in the country are clustered according to the proportion of each industry in GDP, per capita GDP and per capita disposable income of residents;

[0125] Cluster analysis of 31 provinces was performed using the k-means algorithm.

[0126] In an exemplary embodiment, the component oil demand forecasting model is established by the following process:

[0127] Extract the principal components of the provinces included in this group;

[0128] Regression modeling was performed for this group separately;

[0129] The regression model is:

[0130]

[0131] i=1,...,N;t=1,...,T i

[0132] In the above model, N is the number of provinces included in this group; T is the time length; K is the independent variable constant; u is the residual term; the extracted principal component is taken as the independent variable x, and the gasoline and diesel consumption data y of each province is taken as the dependent variable.

[0133] In this embodiment, the k-means algorithm is used to perform cluster analysis on 31 provinces. The results show that the 31 provinces in China can be divided into the following 7 groups.

[0134] Table 4 Group names of 31 provinces

[0135] Province name Group Province name Group Liaoning 1 Shanxi 5 Fujian 1 Heilongjiang 5 Shandong 1 Hunan 5 Guangdong 1 Shaanxi 5 Beijing 2 Ningxia 5 Tianjin 2 Xinjiang 5 Shanghai 2 Anhui 6 Guizhou 3 Jiangxi 6 Yunnan 3 Henan 6 Gansu 3 Guangxi 6 Tibet 3 Hainan 6 Inner Mongolia 4 Sichuan 6 Jiangsu 4 Qinghai 6 Zhejiang 4 Hebei 7 Jilin 7 Hubei 7 Chongqing 7

[0136] In this embodiment, after extracting the principal components of each province, matrix data regression modeling is performed on each group. The general pattern of matrix data regression is

[0137]

[0138] i=1,...,N;t=1,...,T i

[0139] N is the number of cases, here it is the number of provinces included in each group; T is the length of time, if all T i =T, then it is balanced data, otherwise it is unbalanced data; K is the number of independent variables, which is a constant in the matrix data model, and K will be different for different models; u is the residual term, and its difference directly leads to different matrix data models. The principal components extracted previously are used as independent variables, and the gasoline and diesel consumption data of each province are used as dependent variables, and the modeling is completed by regression method. Using the above method, the 7 groups are modeled one by one to complete the construction of the entire model.

[0140] When using the model for prediction, the matrix data models of the seven groups are run one by one to obtain the prediction data for 31 provinces, which are then added together to obtain the national prediction data.

[0141] The present application also provides a device for predicting the national demand for refined oil products, such as Figure 2 As shown, it includes: a memory 210 and a processor 220;

[0142] The memory is used to store a program for predicting the national demand for refined oil products;

[0143] The processor is used to read and execute the program for predicting the national demand for refined oil, and perform the method for predicting the national demand for refined oil as described in any one of the above embodiments.

[0144] The present application also provides a computer storage medium, in which a program for predicting national refined oil demand is stored, and the program is configured to execute any of the methods in the above embodiments when running.

[0145] Example 1

[0146] Adopt the method of forecasting national refined oil demand, such as Figure 5 As shown in the figure, the process of predicting the national demand for refined oil is further explained:

[0147] Step 500: Select a series of external variables related to the demand for refined oil, such as the number of cars in use, industrial added value, and more than 20 other variables, and use the principal component analysis method to convert the 20 variables into three to four principal component variables.

[0148] In this step, taking the first group of provinces as an example, in order to verify the necessity of the principal component analysis method to verify the process; the results of the diesel sphericity test for the first group of provinces are shown in Table 5. The sig value is significantly lower than 0.05, rejecting the original hypothesis that the data is spherically distributed, indicating that the principal component analysis method can be used to convert twenty variables into three to four principal component variables.

[0149] Table 5 Sphericity test results

[0150] KMO and Bartlett's Test

[0151]

[0152] Step 510: Regress the consumption of refined oil products using the principal components to construct a national prediction model.

[0153] Step 520: cluster the 31 provinces according to indicators such as GDP and population, and divide the 31 provinces into 7 groups according to the principle of similar characteristics to form matrix data.

[0154] Step 530: Perform matrix data regression on each group of finished oil consumption, establish a prediction model for each group, and then integrate the prediction values ​​of each province to form a national prediction value.

[0155] Using the above method, according to the forecast results released from April 2017 to December 2019, the forecast error rate of the absolute value of annual refined oil demand is 3.5%. Selecting the forecast results of the same period, in the field of annual refined oil demand forecast, the forecast error rate of the Planning Institute is 1.7 percentage points ahead of the domestic advanced level.

[0156] Table 6

[0157] mechanism 2016 2017 2018 2019 average Planning Institute 1.8％ 4.2％ 1.0％ 0.1％ 1.8％ Economic and Technological Research Institute 0.6％ 2.4％ 4.5％ 6.6％ 3.5％

[0158] This disclosure example solves the problems existing in the prior art by predicting the national demand for refined oil products:

[0159] 1) Use the principal component method to overcome the problem of information loss. Information overlap is actually due to too many data dimensions. Introducing the principal component method to reduce the data dimension, transforming more than a hundred interrelated independent variables into two to four unrelated variables, can effectively solve the problem of information overlap.

[0160] 2) Solve the information overflow problem with a bottom-up model. National demand is composed of provincial demand. Accurately predicting the demand of each province can accurately predict national demand. There is a similar dependency pattern in the growth of finished products among provinces, which can clearly explain the reasons for the growth of refined oil demand and solve the information overflow problem of location quotient and other methods.

[0161] 3) Use clustering method to group and perform panel regression forecast. The development patterns of refined oil demand in each province are quite different. Using clustering method to classify each province for forecasting can better reflect different growth patterns. Using panel regression method to forecast provinces in the same group can reflect the internal connection between provinces in the same group.

[0162] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In hardware implementations, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all components may be implemented as software executed by a processor, such as a digital signal processor or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those of ordinary skill in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

Claims

1. A method for predicting national demand for refined oil, characterized in that: The method comprises: Obtain N external variables related to the demand for refined oil products in the previous year; The principal component analysis method is used to transform N variables into m principal component variables, where m is a positive integer less than N; Based on the determined m principal component variables, the national refined oil demand for the next year is determined using a pre-established national refined oil demand forecasting model.

2. The method for predicting national refined oil demand according to claim 1, characterized in that: The external variables include one or more of the following: automobile data, transportation data, and industrial data.

3. The method for predicting national refined oil demand according to claim 2, characterized in that: The principal component analysis method is used to convert N variables into m principal component variables, including: The linear combination that maximizes the variance of N variables is determined by using the Lagrange multiplier method; The m principal component variables are determined according to the determined linear combinations.

4. The method for predicting national refined oil demand according to claim 3, characterized in that: After determining m principal component variables according to the determined linear combination, the method further comprises: The critical value of smoothing is determined by the scree plot method, and the value corresponding to the critical value is used as the number of principal components to be extracted.

5. The method for predicting national refined oil demand according to claim 1, characterized in that: Said The pre-established national refined oil demand forecasting model is used to determine the national refined oil demand for the next year, including: Cluster analysis was conducted on the 31 provinces in the country according to relevant indicators; The 31 provinces were divided into multiple groups based on the principle of similar characteristics; Each group corresponds to a component oil demand forecasting model; Determine the demand for refined oil products in the next year for each province based on the corresponding refined oil demand forecast model; The next year's refined oil demand of all provinces will be summed up to determine the next year's national refined oil demand.

6. The method for predicting national refined oil demand according to claim 1, characterized in that: The 31 provinces in the country were clustered according to relevant indicators, including: The 31 provinces in the country are clustered according to the proportion of each industry in GDP, per capita GDP and per capita disposable income of residents; Cluster analysis of 31 provinces was performed using the k-means algorithm.

7. The method for predicting national refined oil demand according to claim 1, characterized in that: The component oil demand forecasting model is established through the following process: Extract the principal components of the provinces included in this group; Regression modeling was performed for this group separately; The regression model is: In the above model, N is the number of provinces included in this group; T is the time length; K is the independent variable constant; u is the residual term; the extracted principal component is taken as the independent variable x, and the gasoline and diesel consumption data y of each province is the dependent variable.

8. The method for predicting national refined oil demand according to claim 2, characterized in that: The method of determining the linear combination that maximizes the variance of N variables by using the Lagrange multiplier method includes: Establish the relationship between the principal components and the original variables; Among them, the relationship is: z=Xu Among them, the original variable X, The principal component variable Z is determined by using the Lagrange multiplier method to satisfy u′u=1 and A linear combination of .

9. A device for predicting the national demand for refined oil, characterized in that: include: Memory and processor; The memory is used to store a program for predicting the national demand for refined oil products; The processor is used to read and execute the program for predicting the national demand for refined oil products, and perform the method for predicting the national demand for refined oil products as described in any one of claims 1-8.

10. A computer storage medium, characterized in that: The storage medium stores a program for predicting national refined oil demand, and the program is configured to execute the method according to any one of claims 1 to 8 when running.

Images