Electric vehicle ownership long-term prediction method considering multi-factor dynamic weight

Abstract

The invention discloses an electric vehicle ownership long-term prediction method considering a multi-factor dynamic weight, and belongs to the technical field of vehicle ownership prediction. Comprising the following steps: acquiring data of electric vehicle ownership, per capita GDP, policy subsidy amount and technical research and development investment; establishing a Logistic prediction model according to the electric vehicle inventory and the technical research and development investment; establishing a grey prediction model GM (1, 1) according to the policy subsidy amount; a Gompertz model is established according to the electric vehicle inventory and the per capita GDP; and according to the established Logistic prediction model, gray prediction model GM (1, 1) and Gompertz model, introducing correction factors, and establishing a comprehensive prediction model to perform long-term prediction on the electric vehicle inventory. According to the method, the prediction precision is improved, the limitation of the fixed coefficient of the combined model is solved, dominant factors of each stage are clarified, the interpretability is enhanced, and a scientific basis is provided for policy making and power grid planning.

Description

Technical Field

[0001] The present invention relates to the technical field of automobile ownership prediction, and in particular to a long-term prediction method for electric vehicle ownership considering the dynamic weights of multiple factors. Background Art

[0002] With the continued rise in global automotive carbon emissions and energy prices, more and more people are choosing environmentally friendly and energy-efficient electric vehicles as their new means of transportation. Driven by the "dual carbon" goals, my country's electric vehicle industry is booming, with sales reaching 6.887 million units in 2022. The large-scale integration of electric vehicles into the grid will significantly alter the current urban electricity load curve, potentially impacting power system stability and posing challenges to load planning. Accurately predicting the number of electric vehicles in use is key to optimizing grid dispatch and future grid planning.

[0003] Existing forecasting methods include regression analysis, elasticity coefficient method, gray model method, logistic model, GDP growth scenario analysis method, Bass model and combination forecasting method. Single model: such as logistic model (technology driven), gray forecasting (policy driven), Gompertz model (economic driven), cannot reflect the interactive impact of multiple factors. Static combination model: a combination of multiple models with fixed weights, without considering differences in development stages. Others predict short-term holdings based on the improved GM-Markov model, but do not solve the problem of long-term error accumulation. Per capita GDP is introduced into the logistic model, but the weight of economic factors is not dynamically adjusted. Multiple linear regression is used, but the stage-by-stage impact of policy and technical factors is ignored.

[0004] In summary, the above electric vehicle ownership prediction methods have the following problems:

[0005] 1. Reliance on a single factor: Existing technologies often use a single model (such as the Logistic, Grey Forecasting, or Gompertz models), ignoring the combined impact of multiple factors such as policy, technology, and economy, resulting in large long-term forecast errors.

[0006] 2. Static weight restriction: The existing combination forecasting model fixes the weight of each factor and does not consider the dynamic changes of the dominant factors in the electric vehicle industry at different development stages (introduction stage, growth stage, and maturity stage).

[0007] 3. Insufficient forecast accuracy: A single model or static combination model fails to adapt to the increased weight of the economic environment in the later stages of forecasting (such as the mature stage), causing the forecast results to deviate from the actual value.

[0008] Therefore, a long-term prediction method for electric vehicle ownership that considers the dynamic weights of multiple factors is needed to solve the above problems. Summary of the Invention

[0009] The purpose of the present invention is to propose a long-term prediction method for electric vehicle ownership considering the dynamic weights of multiple factors, comprising the following steps:

[0010] Obtain data on electric vehicle ownership, per capita GDP, policy subsidy amounts, and technology R&D investment;

[0011] Establish a logistic forecasting model based on the number of electric vehicles in use and technology R&D investment;

[0012] According to the policy subsidy amount, a grey prediction model GM(1,1) is established;

[0013] A Gompertz model is established based on the number of electric vehicles and GDP per capita;

[0014] According to the established Logistic prediction model, grey prediction model GM(1,1) and Gompertz model, correction factors are introduced to establish a comprehensive prediction model for long-term prediction of electric vehicle ownership.

[0015] Furthermore, the Logistic prediction model is:

[0016]

[0017] Among them, P (t) represents the number of electric vehicles at time t, C is the maximum number of electric vehicles, z is the slope of the growth rate, and t0 is the time point when the growth rate reaches half.

[0018] Furthermore, the specific process of establishing the grey prediction model GM(1,1) is as follows:

[0019] Get the original time series dataset

[0020] X=(x1,x2,…x n )

[0021] Among them, x1, x2, … x n The number of electric vehicles in the first to n years;

[0022] Accumulate the original data once to generate a new sequence:

[0023] Y(1)=(y1,y2,…y n )

[0024] in:

[0025] y1=x1

[0026]

[0027] Establish the grey difference equation through the new sequence:

[0028] y′ k +ay k =b(k=2,3,…n)

[0029] Where a and b are constants.

[0030] Furthermore, the Gompertz model is:

[0031] N (x) =N0·exp(-h·exp(-r·x))

[0032] Where: N (x) represents the number of electric vehicles under the independent variable x; N0 is the initial number of electric vehicles; h and r are the parameters of the model, h controls the height of the growth curve, r controls the growth rate, and x is the per capita GDP variable.

[0033] Furthermore, the correction factor is:

[0034]

[0035] Where t is time, t0 is the initial time, and θ and d are formula parameters.

[0036] Furthermore, the comprehensive prediction model is:

[0037] P(i)=C gm (i)·P gm (i)+C log (i)·P log (i)+C gom (i)·P gom (i)

[0038] C gom (i) = C 0gom ·Y(i)

[0039]

[0040] Among them, P(i) represents the comprehensive forecast value of year i, P gm (i) P log (i) P gom (i) are the predicted values ​​of the grey prediction model GM(1,1), the Logistic prediction model, and the Gompertz model in year i, respectively. gm (i) C log (i) C gom (i) are the weights of the grey prediction model GM(1,1), the logistic prediction model, and the Gompertz model in year i, respectively. 0gomrepresents the weight of the Gompertz model for the first year of prediction, Y(i) is the correction factor for the i-th year, ΔC gm , ΔC log is the change in weight value of the grey prediction model GM(1,1) and the Logistic prediction model at the beginning and end of the prediction.

[0041] The beneficial effects of the present invention are:

[0042] The present invention integrates the Logistic (technology), grey prediction (policy), and Gompertz (economic) models, which can cover the influence of multiple factors and improve prediction accuracy; allocates initial weights based on the entropy weight method, and introduces correction factors to simulate the dynamic adjustment of weights with the development stage; through the weight smooth transition mechanism, it solves the limitation of fixed coefficients of the combined model; the present invention clarifies the dominant factors in each stage, enhances interpretability, and provides a scientific basis for policy formulation and power grid planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a flow chart of the long-term prediction method for electric vehicle ownership that takes into account the dynamic weights of multiple factors of the present invention. DETAILED DESCRIPTION

[0044] The present invention proposes a long-term prediction method for electric vehicle ownership that considers the dynamic weights of multiple factors. The present invention is further described below with reference to the accompanying drawings and specific embodiments.

[0045] Figure 1 This is a flow chart of the long-term prediction method for electric vehicle ownership that considers the dynamic weights of multiple factors.

[0046] 1. Data collection: Obtain data on electric vehicle ownership, per capita GDP, policy subsidy amounts, and technology R&D investment.

[0047] 2. Single factor modeling:

[0048] Logistic Model: Electric vehicle technology primarily encompasses powertrains, charging and energy management, intelligence and automation, and body and structural technologies. Battery technology in the powertrain is crucial for vehicle range, convenience, cost, and service life. Therefore, battery technology is a key determinant of electric vehicle ownership growth. In the early stages of the forecast, the maturation of related technologies has significantly reduced electric vehicle costs, leading to explosive growth in electric vehicle ownership. To reflect this, a logistic forecasting model is introduced. This function exhibits a medium-term "rapid growth" pattern, reflecting the explosive growth in electric vehicle ownership driven by technological breakthroughs.

[0049] The form of the Logistic quantity forecasting model is:

[0050]

[0051] In this model, P (t) represents the number of electric vehicles in the country at time t, C is the maximum number of electric vehicles in the country, z is the slope of the growth rate, and t0 is the time point when the growth rate reaches half. Therefore, by setting the parameter t0, the model can reflect the growth of electric vehicle ownership under the influence of technological innovation.

[0052] In order to establish a Logistic electric vehicle ownership prediction model, it is necessary to collect relevant electric vehicle ownership data and time information, and then estimate the model parameters C and z by fitting the model.

[0053] Gray Forecasting Model: The government intervenes in industry development through subsidies, taxes, and mandatory regulations. For China's electric vehicle industry, policy regulation often involves quantitative planning. Therefore, when simulating the impact of policies on electric vehicle population, we must not simply analyze based on data but must consider the mandatory nature of the policies. Using the equal-dimensional new information gray coefficient method as the basis, we simulate population growth under different policy scenarios by setting different growth parameters.

[0054] The common grey prediction model is GM(1,1), and its basic construction process is as follows:

[0055] Original data: Given a time series dataset

[0056] X=(x1,x2,…x n ) (2)

[0057] Among them, x1, x2, … x n The number of electric vehicles in years 1-n; generate a cumulative sequence: accumulate the original data once to generate a new sequence:

[0058] Y(1)=(y1,y2,…y n ) (3)

[0059] in:

[0060] y1=x1 (4)

[0061]

[0062] Establish model: Establish grey difference equation through the cumulative sequence of data:

[0063] y′ k +ay k =b(k=2,3,…n) (6)

[0064] Where a and b are constants, obtained by least squares fitting.

[0065] In actual load forecasting using grey models, it has been found that when data dispersion is large, forecast accuracy decreases significantly. This is especially true for medium- and long-term load forecasting, where the forecast results at the end of the forecast period are suboptimal. Based on this, the equal-dimensional new information replenishment forecasting method is employed. New information obtained from the forecast model is fed into the original data sequence, and outdated data is removed, repeating this cycle until the target is achieved.

[0066] This improvement not only overcomes the drawback of the fixed mathematical model in the simple grey prediction method, but also takes advantage of the high short-term prediction accuracy of the grey prediction method, so that the prediction model can be effectively revised and its prediction accuracy is significantly improved.

[0067] Gompertz model: The economic environment influences the number of electric vehicles through multiple channels, including income levels, energy prices, government policies, and the financing environment. The economic environment not only increases the number of electric vehicles during the rapid growth phase of the electric vehicle industry, but also, after the market reaches saturation, economic growth can increase the maximum number of electric vehicles in the market, thus affecting the number of electric vehicles in the later stages of development.

[0068] Without the influence of the implementation of control policies, the per capita number of motor vehicles has a strong positive correlation with the per capita GDP. Therefore, based on the characteristics of China's economic growth, it is believed that the maximum number of vehicles in the simulation will slowly increase in the middle and late stages. This embodiment uses my country's per capita GDP as the independent variable data and uses the Gompertz model to describe the impact of the economic environment on the number of electric vehicles.

[0069] The standard form of the Gompertz model is as follows:

[0070] N (x) =N0·exp(-h·exp(-r·x)) (7)

[0071] Where: N (x) represents the number of electric vehicles under the independent variable x; N0 is the initial number of electric vehicles; h and r are the parameters of the model, which control the shape and change rate of the growth curve; h controls the height of the growth curve, r controls the growth rate, and x is the per capita GDP variable.

[0072] 3. Comprehensive prediction model: To calculate the weights of the three factors of policy, technology, and economy on the number of electric vehicles, a variety of methods can be used. The entropy weight method is a common and effective tool. The core idea of ​​the entropy weight method is to measure the amount of information of each indicator according to the size of the information entropy, and use this to determine the importance of the indicator. The smaller the entropy value, the greater the amount of information, and the higher the weight should be. This method can effectively avoid the interference of subjective factors and make the determination of weights more scientific and objective. According to the analysis of influencing factors, the weights of each indicator should be different in different periods. Therefore, the weights of various influencing factors in the later stage of fuel vehicle development are used as the weights for predicting the later stage of electric vehicle development, and correction factors are introduced to smoothly adjust the changes in the weights of various factors.

[0073] Dynamic weight adjustment:

[0074] Initial weights: Technology, policy, and economic weights are calculated using the entropy weight method (Table 1)

[0075] Table 1 Weights of various influencing factors in the initial stage (2023)

[0076] Method Logistic GM Gompertz Ratio 0.41 0.25 0.34

[0077] Final period weight: set with reference to fuel vehicle development data (Table 2).

[0078] Table 2 Weights of various influencing factors at the end of the period (2025)

[0079] Method Logistic GM Gompertz Ratio 0.30 0.20 0.50

[0080] Correction factor: The impact of technology maturity and policy regulation on electric vehicle ownership decreases as the forecast year increases, while the weight of economic environment factors increases, which is consistent with the analysis of the impact of each factor at different development stages. The correction factor is introduced using the weight coefficient of the Gopmertz model as a variable, and its calculation formula is as follows:

[0081]

[0082] Among them, t0 is the initial time (2023), θ and d are formula parameters, which are solved by substituting the weight values ​​of the Gopmertz model predicted in 2023 and 2025; and the changes in the weight coefficients of the Logistic model and the GM model are proportionally reduced by the negative growth of the weight coefficient of the Gopmertz model.

[0083] Comprehensive prediction: weighted sum of the results of each model, outputting a dynamically adjusted predicted value of inventory. The formula is as follows:

[0084] P(i)=C gm (i)P gm (i)+C log(i)P log (i)+C gom (i)P gom (i) (9)

[0085] Among them, P(i) represents the comprehensive forecast value of year i, P gm (i) P log (i) P gom (i) is the predicted value of each model in year i, C gm (i) C log (i) C gom (i) represents the weight of each model in year i, which is calculated as follows:

[0086] C gom (i) = C 0gom ·Y(i) (10)

[0087]

[0088] Among them, C 0gom represents the weight of the Gompertz model for the first year of prediction, Y(i) is the correction factor for the i-th year, ΔC gm , ΔC log is the change in weight value of GM model and Logistic model at the beginning and end of prediction.

[0089] In summary, the actual number of electric vehicles in 2023 and 2024 is compared with the prediction results of each single model and the comprehensive prediction model results of the present invention, as shown in the following table:

[0090] Table 3 Relative errors of various prediction models in 2023

[0091]

[0092] Table 4 Relative errors of various prediction models in 2024

[0093]

[0094] The relative errors in 2023 and 2024 (1.8% and 3.3%) were significantly lower than those of a single model. This model, through a weight adjustment mechanism, adapts to the changing weights of technology, policy, and economic factors throughout the industry lifecycle. This model clarifies the dominant factors at each stage, enhances interpretability, and provides a scientific basis for policymaking and grid planning.

Claims

1. A long-term prediction method for electric vehicle ownership considering dynamic weights of multiple factors, characterized by: The following steps are involved: Obtain data on electric vehicle ownership, per capita GDP, policy subsidy amounts, and technology R&D investment; Establish a logistic forecasting model based on the number of electric vehicles in use and technology R&D investment; According to the policy subsidy amount, a grey prediction model GM(1,1) is established; A Gompertz model is established based on the number of electric vehicles and GDP per capita; According to the established Logistic prediction model, grey prediction model GM(1,1) and Gompertz model, correction factors are introduced to establish a comprehensive prediction model for long-term prediction of electric vehicle ownership.

2. The long-term prediction method for electric vehicle ownership considering dynamic weights of multiple factors according to claim 1 is characterized in that: The Logistic prediction model is: Among them, P (t) represents the number of electric vehicles at time t, C is the maximum number of electric vehicles, z is the slope of the growth rate, and t0 is the time point when the growth rate reaches half.

3. The long-term prediction method for electric vehicle ownership considering dynamic weights of multiple factors according to claim 1 or 2 is characterized in that: The specific process of establishing the grey prediction model GM(1,1) is as follows: Get the original time series dataset X=(x1,x2,…x n ) Among them, x1, x2, … x n The number of electric vehicles in the first to n years; Accumulate the original data once to generate a new sequence: Y(1)=(y1,y2,…y n ) in: y1=x1 Establish the grey difference equation through the new sequence: y′ k +is k =b(k=2,3,…n) Where a and b are constants.

4. The long-term prediction method for electric vehicle ownership considering dynamic weights of multiple factors according to claim 1 is characterized in that: The Gompertz model is: N (x) =N0·exp(-h·exp(-r·x)) Where: N (x) represents the number of electric vehicles under the independent variable x; N0 is the initial number of electric vehicles; h and r are the parameters of the model, h controls the height of the growth curve, r controls the growth rate, and x is the per capita GDP variable.

5. The long-term prediction method for electric vehicle ownership considering dynamic weights of multiple factors according to claim 1 is characterized in that: The correction factor is: Where t is time, t0 is the initial time, and θ and d are formula parameters.

6. The long-term prediction method for electric vehicle ownership considering dynamic weights of multiple factors according to claim 1 or 5 is characterized in that: The comprehensive prediction model is: P(i)=C gm (i)·P gm (i)+C log (i)·P log (i)+C gom (i)·P gom (i) C gom (i)=C 0gom ·Y(i) Among them, P(i) represents the comprehensive forecast value of year i, P gm (i) P log (i) P gom (i) are the predicted values ​​of the grey prediction model GM(1,1), the Logistic prediction model, and the Gompertz model in year i, respectively. gm (i) C log (i) C gom (i) are the weights of the grey prediction model GM(1,1), the logistic prediction model, and the Gompertz model in year i, respectively. 0gom represents the weight of the Gompertz model for the first year of prediction, Y(i) is the correction factor for the i-th year, ΔC gm , ΔC log is the change in weight value of the grey prediction model GM(1,1) and the Logistic prediction model at the beginning and end of the prediction.

Images