Power demand prediction method and system based on energy dual-carbon gpt

Abstract

The invention discloses an energy dual-carbon gpt-based power demand prediction method and system, and the method comprises the following steps: obtaining historical power data which comprises historical charging data and historical power demand data of an electric vehicle; extracting charging demand characteristics based on historical charging data of the electric vehicle; constructing an electric vehicle charging demand prediction model according to the charging demand characteristics, and outputting a charging demand prediction value by the model; calculating a power demand-charging demand correlation coefficient between the historical power demand data and the historical charging demand data of the electric vehicle; training a preset power demand prediction model, inputting the charging demand prediction value into the trained power demand prediction model, and outputting a power demand prediction result of the same timestamp by the model; and calculating an error between the power demand prediction value and a historical actual value, comparing the error with a preset standard range, and if the error is within the standard range, applying the prediction method to the power demand prediction system.

Description

Technical Field

[0001] The present invention relates to the technical field of power demand forecasting, and in particular to a power demand forecasting method and system based on energy dual-carbon GPT. Background Art

[0002] As the global climate change problem becomes increasingly serious, the frequent occurrence of global warming and extreme climate events have made reducing greenhouse gas emissions and controlling global warming a consensus of the international community. At the United Nations General Assembly in September 2020, China proposed to achieve carbon peak by 2030 and carbon neutrality by 2060. Achieving the dual carbon goals will help reduce greenhouse gas emissions, protect ecosystem integrity and human well-being, reduce climate change risks, and promote the comprehensive transformation of the economy and society to green and low-carbon development.

[0003] Energy dual carbon GPT refers to the "dual carbon" goals in the energy field, namely carbon peak and carbon neutrality. Carbon peak means that at a certain point in time, carbon dioxide emissions no longer increase and reach a peak, and then gradually decline, marking a historical turning point from an increase to a decrease in carbon emissions, and carbon emissions are decoupled from economic development; Carbon neutrality refers to the total amount of greenhouse gas emissions directly or indirectly generated by enterprises, groups or individuals within a certain period of time, and then offset their own carbon dioxide emissions through afforestation, energy conservation and emission reduction, etc., to achieve "zero emissions" of carbon dioxide.

[0004] Under the framework of the dual carbon targets, electricity demand forecasts not only need to consider the supply and demand of electricity, but also pay attention to the carbon emissions of different energy sources, especially the impact of renewable energy, energy storage technology, clean energy and other factors. Under the background of the "dual carbon targets", new energy will become the main source of electricity supply, especially with the advancement of technology, the cost of renewable energy such as wind and solar energy will gradually decrease, and will gradually replace traditional energy.

[0005] As electric vehicles become more popular, charging demand will also become an important part of future electricity demand. Changes in the charging time, location, and amount of electric vehicles will affect electricity demand, especially at different times and in different regions. This requires combining new energy with the charging mode of electric vehicles to analyze electricity demand.

[0006] In the prior art, there is a problem that the traditional electricity demand forecasting method mainly relies on the use of traditional energy, but does not consider the impact of new energy on future electricity supply. Summary of the invention

[0007] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a method and system for predicting electricity demand based on energy dual-carbon GPT to solve the problem in the prior art that the traditional electricity demand forecasting method does not consider the impact of new energy on future electricity supply.

[0008] To achieve the above object, the present invention provides the following technical solutions: A method for predicting electricity demand based on energy dual-carbon GPT includes the following steps: acquiring historical electricity data, the historical electricity data including historical charging data of electric vehicles and historical electricity demand data; extracting charging demand characteristics based on the historical charging data of electric vehicles, the charging demand characteristics including charging demand characteristics during peak hours, seasonal variation characteristics and climate condition correlation characteristics; constructing an electric vehicle charging demand prediction model according to the charging demand characteristics, and the model outputs a charging demand prediction value; aligning the historical electricity demand data with the historical charging demand data of electric vehicles using timestamps as markers, and calculating the electricity demand-charging demand correlation coefficient between the historical electricity demand data and the historical charging demand data of electric vehicles; training a preset electricity demand prediction model in combination with the electricity demand-charging demand correlation coefficient, inputting the charging demand prediction value into the trained electricity demand prediction model, and the model outputs an electricity demand prediction result of the same timestamp; analyzing the electricity demand at historical time for the predicted electricity demand prediction result, calculating the error between the electricity demand prediction value and the historical actual value, and comparing the error with a preset standard range, and if the error is within the standard range, applying the prediction method to the electricity demand prediction system.

[0009] In a preferred embodiment, the peak period charging demand characteristics are obtained by: cleaning the historical charging data of the electric vehicle, and aggregating the historical charging data of the electric vehicle at different time scales to generate a time series of charging demand; establishing a line graph based on the time series of charging demand to identify the peak time of charging demand.

[0010] In a preferred embodiment, the seasonal variation characteristics are obtained by combining the timestamp of the time series of charging demand with the season information to generate a seasonal label; and performing aggregation analysis on different seasons to obtain the difference in charging amount for each season.

[0011] In a preferred embodiment, the climate condition correlation feature is obtained by: obtaining a climate type time series signal, cutting the electric vehicle historical charging data according to the weather type time series signal, and obtaining a number of electric vehicle historical charging data belonging to the same climate type; classifying the several electric vehicle historical charging data belonging to the same climate type into a first signal set; performing variational modal decomposition processing on the first signal set to obtain a discontinuous modal component signal, smoothing the discontinuous modal component signal to obtain the volatility of the charging amount with climate change, and using it as the climate condition correlation feature.

[0012] In a preferred embodiment, the electric vehicle charging demand prediction model is constructed according to the charging demand characteristics, and the model outputs the charging demand prediction value, specifically: the charging demand characteristics are predicted by time, and the first charging demand prediction values ​​of several time granularities are obtained after the time-sharing prediction; the first charging demand prediction value of each time granularity corresponds to each node of the initial electric vehicle charging demand prediction model, and the many-to-many relationship between each node is calculated; based on the many-to-many relationship between the nodes, the prediction value of each time point is combined with the prediction results of other nodes to obtain the final prediction value of the charging demand.

[0013] In a preferred embodiment, the first charging demand prediction value of each time granularity corresponds to each node of the initial electric vehicle charging demand prediction model, and the many-to-many relationship between each node is calculated, specifically: the first charging demand prediction value of each time granularity is matched one-to-one with each node of the initial electric vehicle charging demand prediction model; the similarity between the first charging demand prediction value of each time granularity and the first charging demand prediction value of other time granularities is calculated; the similarities calculated by matching all nodes one by one are averaged, and the average value is used as the first threshold; two nodes whose similarity is greater than the first threshold are marked, and the time dependency between the marked nodes is further calculated.

[0014] In a preferred embodiment, the calculation of the power demand-charging demand correlation coefficient between the historical power demand data and the historical charging demand data of the electric vehicle is specifically as follows: checking the timestamps of the historical power demand data and the historical charging demand data of the electric vehicle, if the timestamps are inconsistent, performing data resampling; if the timestamps are consistent, calculating the ratio of the covariance of the historical power demand data and the historical charging demand data of the electric vehicle to the standard deviation in combination with the Pearson correlation coefficient, to obtain the power demand-charging demand correlation coefficient.

[0015] In a preferred embodiment, the system of the electricity demand forecasting method based on energy dual-carbon GPT is characterized by including a historical electricity data acquisition module, a charging demand feature extraction module, a charging demand forecasting model construction module, a correlation calculation module, a power demand forecasting module, and a forecasting method application module: a historical electricity data acquisition module, used to acquire historical electricity data, the historical electricity data includes historical charging data and historical electricity demand data of electric vehicles; a charging demand feature extraction module, used to extract charging demand features based on historical charging data of electric vehicles, the charging demand features include peak period charging demand features, seasonal variation features, and climate condition correlation features; a charging demand forecasting model construction module, used to construct an electric vehicle charging demand forecasting model based on the charging demand features, and the model outputs a charging demand forecast value A correlation calculation module is used to align the historical power demand data with the historical charging demand data of the electric vehicle using timestamps as markers, and calculate the power demand-charging demand correlation coefficient between the historical power demand data and the historical charging demand data of the electric vehicle; a power demand prediction module is used to train a preset power demand prediction model in combination with the power demand-charging demand correlation coefficient, input the charging demand prediction value into the trained power demand prediction model, and the model outputs the power demand prediction result of the same timestamp; a prediction method application module is used to analyze the power demand of the predicted power demand prediction result in historical time, calculate the error between the power demand prediction value and the historical actual value, compare the error with a preset standard range, and if the error is within the standard range, apply the prediction method to the power demand prediction system.

[0016] Technical effects and advantages of the power demand forecasting method and system based on energy dual-carbon GPT of the present invention: 1. By comprehensively considering the charging data of electric vehicles and the characteristics of charging demand, a more accurate electricity demand forecasting model is provided, which enables the electricity demand forecast to take into account the impact of charging demand on the power system, and effectively solves the problem in the existing technology that the traditional electricity demand forecasting method mainly relies on the use of traditional energy without considering the impact of new energy on future power supply.

[0017] 2. By introducing the variational mode decomposition method, the volatility characteristics of climate condition correlation can be effectively extracted from complex electric vehicle charging demand data, thereby capturing the changing trend of charging demand under different climate conditions and improving the accuracy of climate feature extraction.

[0018] 3. By combining the collaborative filtering algorithm to capture the similarities between different time points in the charging demand data, the many-to-many relationship between time nodes is identified and weighted averaged, thereby generating a more accurate charging demand forecast value. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A flow chart of a method for predicting electricity demand based on energy dual-carbon GPT provided in an embodiment of the present application; Figure 2 A system module diagram of a system for a method for predicting electricity demand based on energy dual-carbon GPT provided in an embodiment of the present application. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0022] Embodiment 1, Figure 1 The present invention provides a method for predicting electricity demand based on energy dual-carbon GPT, comprising the following steps: S1, acquiring historical power data, wherein the historical power data includes historical charging data of electric vehicles and historical power demand data, wherein the historical charging data of electric vehicles includes charging amount in different time periods, charging rate and charging power; S2, extracting charging demand characteristics based on historical charging data of electric vehicles, wherein the charging demand characteristics include charging demand characteristics during peak hours, seasonal variation characteristics, and climate condition correlation characteristics; S3, constructing an electric vehicle charging demand prediction model based on the charging demand characteristics, and the model outputs a charging demand prediction value; S4, aligning the historical power demand data with the historical charging demand data of the electric vehicle using the timestamp as a marker, and calculating the power demand-charging demand correlation coefficient between the historical power demand data and the historical charging demand data of the electric vehicle; S5, training a preset power demand prediction model in combination with the power demand-charging demand correlation coefficient, inputting the charging demand prediction value into the trained power demand prediction model, and the model outputs the power demand prediction result at the same timestamp; S6, analyzing the power demand in historical time based on the predicted power demand forecast results, calculating the error between the power demand forecast value and the historical actual value, comparing the error with a preset standard range, and if the error is within the standard range, applying the forecasting method to the power demand forecasting system.

[0023] The historical power demand data refers to the total power demand at different time points in the power grid during a period of historical time.

[0024] The charging demand characteristics are extracted based on the historical charging data of electric vehicles, as follows: The peak period charging demand characteristics are obtained by: Clean the historical charging data of electric vehicles and aggregate them at different time scales to generate a time series of charging demand; Create a line chart based on the time series of charging demand to identify the peak time of charging demand.

[0025] The seasonal variation characteristics are obtained by: Combine the timestamp of the time series of charging demand with the seasonal information to generate seasonal labels; Aggregate analysis is performed on different seasons to obtain the difference in charging capacity in each season.

[0026] The method for obtaining the climate condition correlation characteristics is specifically as follows: Acquire a weather type time series signal, and cut the electric vehicle historical charging data according to the weather type time series signal to obtain a plurality of electric vehicle historical charging data belonging to the same weather type; Classifying the plurality of electric vehicle historical charging data belonging to the same climate type into a first signal set; The first signal set is subjected to variational modal decomposition processing to obtain discontinuous modal component signals, and the discontinuous modal component signals are smoothed to obtain the volatility of the charging amount with climate change, which is used as a climate condition correlation feature.

[0027] Variational Mode Decomposition (VMD) is a technique used for signal analysis and processing, which aims to decompose complex signals into multiple modes (patterns) with different frequency components. This method achieves signal decomposition through variational derivation, has good signal local characteristics and adaptability, and is widely used in signal denoising, feature extraction, fault diagnosis and other fields.

[0028] Variational mode decomposition has the following advantages and characteristics: Strong adaptability: VMD can adaptively select the center frequency and bandwidth of the mode according to the characteristics of the signal, independent of the preset frequency range; Good decomposition effect: VMD can effectively avoid the overlap of frequency components in the signal, so it has more advantages than traditional decomposition methods (such as EEMD and EMD) when processing complex signals; Noise suppression: VMD can effectively suppress noise components and retain effective information of the signal during signal decomposition; High frequency resolution: VMD can accurately decompose the frequency components of the signal and has good frequency resolution, which is suitable for the analysis of various dynamic and non-stationary signals.

[0029] The electric vehicle charging demand prediction model is constructed according to the charging demand characteristics, and the model outputs the charging demand prediction value, which is specifically: Perform time-sharing prediction on charging demand characteristics, and obtain first charging demand prediction values ​​of several time granularities after time-sharing prediction; The first charging demand prediction value of each time granularity corresponds to each node of the initial electric vehicle charging demand prediction model, and the many-to-many relationship between each node is calculated; Based on the many-to-many relationship between nodes, the predicted value at each time point is combined with the prediction results of other nodes to obtain the final predicted value of the charging demand.

[0030] The first charging demand forecast value at each time granularity (such as the hourly forecast value) may be affected by multiple factors, and there are complex relationships between these factors. Therefore, it is necessary to calculate the many-to-many relationship between nodes.

[0031] The first charging demand prediction value of each time granularity corresponds to each node of the initial electric vehicle charging demand prediction model, and the many-to-many relationship between each node is calculated, specifically: The first charging demand prediction value of each time granularity is matched one by one with each node of the initial electric vehicle charging demand prediction model; Calculate the similarity between the first charging demand prediction value of each time granularity and the first charging demand prediction values ​​of other time granularities; Calculate the average value of similarities calculated by matching all nodes one by one, and use the average value as the first threshold; Two nodes whose similarity is greater than the first threshold are marked, and the time dependency between the marked nodes is further calculated.

[0032] It should be noted that for the predicted value at each time point, the similarity between the predicted values ​​at other time points is calculated, and the similarity between different time granularities can be calculated by cosine similarity; considering the change of charging demand over time, the collaborative filtering algorithm can be used to capture the dependency between different time nodes in the time series data.

[0033] The similarity between the first charging demand prediction value of each time granularity and the first charging demand prediction value of other time granularities is calculated by cosine similarity. The specific calculation formula is as follows:

[0034] Where: For one of the nodes to calculate similarity, Another node for calculating similarity, is the similarity calculation formula between two nodes, Represents the modulo calculation of a node.

[0035] Based on the many-to-many relationship between nodes, the predicted value at each time point is combined with the prediction results of other nodes to obtain the final predicted value of the charging demand, specifically:

[0036] Where: is the final predicted value of the charging demand of node A, n is the total number of time granularities, that is, the number of summary points of the model; is the similarity between two nodes, is the first charging demand prediction value of node B.

[0037] The two nodes whose similarity is greater than the first threshold are marked, and the time dependency between the marked nodes is further calculated. The time dependency is manifested as node i occurring after node j, and the predicted value of node i depends on the predicted value of node j. The specific calculation formula is as follows:

[0038] Where: is the time dependency between nodes i and j, is the weight of time dependency, which is learned from historical data and represents the delay relationship between the time point of node i and the time point of node j. The similarity calculation formula between two nodes.

[0039] The basic principle of collaborative filtering algorithm is as follows: In recommendation systems, collaborative filtering is often used to analyze user ratings of items. For charging demand prediction, we can regard different time points as "items" and different charging stations or electric vehicles as "users". Each item in the matrix represents the charging demand at that time point; if some charging stations show similar demand patterns in similar time periods, the demand of these charging stations can be used to predict the demand of current charging stations; by analyzing the similarities between different time periods, we can find time periods with similar demand in the past and apply their demand patterns to the current prediction.

[0040] The purpose of calculating the many-to-many relationship between each node is to determine the relationship between the predicted values ​​at different time points. By calculating the similarity between each time point, it can be determined which time points have a greater impact on each other. For example, the charging demand forecast value of a certain period may be highly correlated with the demand forecast value of the previous period.

[0041] By calculating the similarity between every two time nodes, the strength of the relationship between each time node can be obtained, and then a weighted average method or combination strategy is used to weight the prediction results of multiple related nodes to obtain the final prediction value. For example, the prediction results of each time node can be weighted so that nodes with stronger relationships have a greater impact on the final result.

[0042] The electricity demand-charging demand correlation coefficient refers to the statistical correlation between electricity demand and charging demand, and reflects the coefficient of the impact of charging demand on overall electricity demand.

[0043] The calculation of the power demand-charging demand correlation coefficient between the historical power demand data and the historical charging demand data of the electric vehicle is specifically: Check the timestamps of historical electricity demand data and historical electric vehicle charging demand data. If the timestamps are inconsistent, resample the data. If the timestamps are consistent, combined with the Pearson correlation coefficient, the ratio of the covariance and standard deviation of the historical electricity demand data and the historical electric vehicle charging demand data is calculated to obtain the electricity demand-charging demand correlation coefficient.

[0044] The Pearson correlation coefficient is combined to calculate the ratio of the covariance of the two variables to the standard deviation to obtain the power demand-charging demand correlation coefficient. The specific calculation formula is as follows: The covariance calculation formula of historical electricity demand data and electric vehicle historical charging demand data is as follows:

[0045] Where: is the covariance calculation formula, is the electric vehicle charging demand value at the i-th timestamp, is the power demand value at the i-th timestamp, is the average value of the electric vehicle charging demand, is the average value of the power demand value at each timestamp, and n is the total number of timestamps.

[0046] The formula for calculating the standard deviation of historical electricity demand data and historical electric vehicle charging demand data is as follows:

[0047] Where: is the standard deviation of the historical charging demand data of electric vehicles, is the standard deviation of historical electricity demand data, is the electric vehicle charging demand value at the i-th timestamp, is the power demand value at the i-th timestamp, is the average value of the electric vehicle charging demand, is the average value of the power demand value at each timestamp, and n is the total number of timestamps.

[0048] The power demand-charging demand correlation coefficient is calculated as follows:

[0049] Where: is the covariance calculation formula, is the standard deviation of the historical charging demand data of electric vehicles, is the standard deviation of historical electricity demand data.

[0050] The alignment process is to align the power demand and electric vehicle charging demand data based on timestamps to ensure that there are corresponding power demand and charging demand data at the same time point.

[0051] It should be noted that the electricity demand-charging demand correlation coefficient indicates the strength of the relationship between electricity demand and electric vehicle charging demand. If the electricity demand-charging demand correlation coefficient is close to 1 or -1, it means that the relationship is strong, and if it is close to 0, it means that the linear relationship between the two is weak.

[0052] The error between the predicted power demand value and the historical actual value is calculated as follows:

[0053] Where: is the error between the predicted value of electricity demand and the historical actual value, is the predicted value of electricity demand, is the historical actual value of electricity demand.

[0054] The charging demand characteristics include peak period charging demand characteristics, seasonal variation characteristics, and climate condition correlation characteristics; A charging demand prediction model building module is used to build an electric vehicle charging demand prediction model based on charging demand characteristics, and the model outputs a charging demand prediction value; A correlation calculation module is used to align the historical power demand data with the historical charging demand data of the electric vehicle using the timestamp as a marker, and calculate the power demand-charging demand correlation coefficient between the historical power demand data and the historical charging demand data of the electric vehicle; A power demand prediction module, used to train a preset power demand prediction model in combination with the power demand-charging demand correlation coefficient, input the charging demand prediction value into the trained power demand prediction model, and the model outputs the power demand prediction result at the same timestamp; The prediction method application module is used to analyze the power demand in historical time based on the predicted power demand forecast results, calculate the error between the power demand forecast value and the historical actual value, compare the error with a preset standard range, and if the error is within the standard range, apply the prediction method to the power demand forecast system.

[0055] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0056] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0057] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0058] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0059] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0060] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. 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.

Claims

1. The power demand forecasting method based on energy dual-carbon gpt is characterized by: The following steps are involved: Acquiring historical power data, wherein the historical power data includes historical charging data of electric vehicles and historical power demand data; Extracting charging demand characteristics based on historical charging data of electric vehicles, wherein the charging demand characteristics include peak period charging demand characteristics, seasonal variation characteristics, and climate condition correlation characteristics; Build an electric vehicle charging demand prediction model based on charging demand characteristics, and the model outputs the charging demand prediction value; Align the historical power demand data with the historical charging demand data of electric vehicles using timestamps as markers, and calculate the power demand-charging demand correlation coefficient between the historical power demand data and the historical charging demand data of electric vehicles; The preset power demand prediction model is trained in combination with the power demand-charging demand correlation coefficient, the charging demand prediction value is input into the trained power demand prediction model, and the model outputs the power demand prediction result at the same timestamp; The predicted power demand forecast results are used to analyze the power demand in historical time, calculate the error between the power demand forecast value and the historical actual value, compare the error with a preset standard range, and if the error is within the standard range, apply the prediction method to the power demand forecast system.

2. The method for predicting electricity demand based on energy dual-carbon GPT as claimed in claim 1 is characterized in that: The peak period charging demand characteristics are obtained by: Clean the historical charging data of electric vehicles and aggregate them at different time scales to generate a time series of charging demand; Create a line chart based on the time series of charging demand to identify the peak time of charging demand.

3. The method for predicting electricity demand based on energy dual-carbon GPT as claimed in claim 2 is characterized in that: The seasonal variation characteristics are obtained by: Combine the timestamp of the time series of charging demand with the seasonal information to generate seasonal labels; Aggregate analysis is performed on different seasons to obtain the difference in charging capacity in each season.

4. The method for predicting electricity demand based on energy dual-carbon GPT as claimed in claim 3 is characterized in that: The method for obtaining the climate condition correlation characteristics is specifically as follows: Acquire a weather type time series signal, and cut the electric vehicle historical charging data according to the weather type time series signal to obtain a plurality of electric vehicle historical charging data belonging to the same weather type; Classifying the plurality of electric vehicle historical charging data belonging to the same climate type into a first signal set; The first signal set is subjected to variational modal decomposition processing to obtain discontinuous modal component signals, and the discontinuous modal component signals are smoothed to obtain the volatility of the charging amount with climate change, which is used as a climate condition correlation feature.

5. The method for predicting electricity demand based on energy dual carbon GPT as claimed in claim 4 is characterized in that: The electric vehicle charging demand prediction model is constructed according to the charging demand characteristics, and the model outputs the charging demand prediction value, which is specifically: Perform time-sharing prediction on charging demand characteristics, and obtain first charging demand prediction values ​​of several time granularities after time-sharing prediction; The first charging demand prediction value of each time granularity corresponds to each node of the initial electric vehicle charging demand prediction model, and the many-to-many relationship between each node is calculated; Based on the many-to-many relationship between nodes, the predicted value at each time point is combined with the prediction results of other nodes to obtain the final predicted value of the charging demand.

6. The method for predicting electricity demand based on energy dual carbon GPT as claimed in claim 5 is characterized in that: The first charging demand prediction value of each time granularity corresponds to each node of the initial electric vehicle charging demand prediction model, and the many-to-many relationship between each node is calculated, specifically: The first charging demand prediction value of each time granularity is matched one by one with each node of the initial electric vehicle charging demand prediction model; Calculate the similarity between the first charging demand prediction value of each time granularity and the first charging demand prediction values ​​of other time granularities; Calculate the average value of similarities calculated by matching all nodes one by one, and use the average value as the first threshold; Two nodes whose similarity is greater than the first threshold are marked, and the time dependency between the marked nodes is further calculated.

7. The method for predicting electricity demand based on energy dual carbon GPT as claimed in claim 6 is characterized in that: The calculation of the power demand-charging demand correlation coefficient between the historical power demand data and the historical charging demand data of the electric vehicle is specifically: Check the timestamps of historical electricity demand data and historical electric vehicle charging demand data. If the timestamps are inconsistent, resample the data. If the timestamps are consistent, combined with the Pearson correlation coefficient, the ratio of the covariance and standard deviation of the historical electricity demand data and the historical electric vehicle charging demand data is calculated to obtain the electricity demand-charging demand correlation coefficient.

8. The method for predicting electricity demand based on energy dual-carbon GPT as claimed in claim 7 is characterized in that: Based on the many-to-many relationship between nodes, the predicted value at each time point is combined with the prediction results of other nodes to obtain the final predicted value of the charging demand, specifically: Where: is the final predicted value of the charging demand of node A, n is the total number of time granularities, that is, the number of summary points of the model; is the similarity between two nodes, is the first charging demand prediction value of node B.

9. The method for predicting electricity demand based on energy dual carbon GPT as claimed in claim 8, characterized in that: The time dependency between the marked nodes is further calculated, and the specific calculation formula is as follows: Where: is the time dependency between nodes i and j, is the weight of time dependency, which is learned from historical data and represents the delay relationship between the time point of node i and the time point of node j. The similarity calculation formula between two nodes.

10. A system using the power demand forecasting method based on energy dual-carbon GPT as claimed in any one of claims 1 to 9, characterized in that: It includes historical power data acquisition module, charging demand feature extraction module, charging demand prediction model construction module, correlation calculation module, power demand prediction module, and prediction method application module: A historical power data acquisition module, used to acquire historical power data, wherein the historical power data includes historical charging data and historical power demand data of electric vehicles; A charging demand feature extraction module is used to extract charging demand features based on historical charging data of electric vehicles, wherein the charging demand features include peak period charging demand features, seasonal variation features, and climate condition correlation features; A charging demand prediction model building module is used to build an electric vehicle charging demand prediction model based on charging demand characteristics, and the model outputs a charging demand prediction value; A correlation calculation module is used to align the historical power demand data with the historical charging demand data of the electric vehicle using the timestamp as a marker, and calculate the power demand-charging demand correlation coefficient between the historical power demand data and the historical charging demand data of the electric vehicle; A power demand prediction module, used to train a preset power demand prediction model in combination with the power demand-charging demand correlation coefficient, input the charging demand prediction value into the trained power demand prediction model, and the model outputs the power demand prediction result at the same timestamp; The prediction method application module is used to analyze the power demand in historical time based on the predicted power demand forecast results, calculate the error between the power demand forecast value and the historical actual value, compare the error with a preset standard range, and if the error is within the standard range, apply the prediction method to the power demand forecast system.

Images