New energy vehicle power supply load prediction method and system

Abstract

The invention belongs to the technical field of data processing, and particularly relates to a power supply load prediction method and system for a new energy vehicle, and the method comprises the steps: obtaining real-time operation parameters and a historical power supply load data set of the new energy vehicle; for any moment, determining the load fluctuation sensitivity according to the offset of the power supply load relative to the historical power supply load in the same period and the difference of the power supply load relative to the previous moment; determining the load demand potential by combining the difference between the residual electric quantity of the on-line controlled vehicle and the environment temperature; and determining a prediction weight based on the two indexes, extracting a long or short period data set in combination with a magnitude relationship between the prediction weight and a preset weight threshold, and realizing prediction by using an exponential smoothing method. According to the method, the problems of response lag and insufficient precision of a traditional prediction algorithm when a charging scene is highly dynamic are effectively solved, and high-precision real-time tracking of unsteady state load characteristics is realized.

Description

Technical Field

[0001] This invention relates to the field of data processing technology. More specifically, this invention relates to a method and system for predicting the power supply load of new energy vehicles. Background Technology

[0002] With the deepening of global energy structure transformation and green travel policies, the new energy vehicle industry has experienced explosive growth, and its large-scale grid connection has become an irreversible trend. In the context of the integration of modern intelligent transportation and smart energy, accurately predicting the power supply load of new energy vehicles is of decisive significance for grid dispatch optimization, rational allocation of charging infrastructure, and ensuring the stability of power supply.

[0003] Traditional load forecasting typically employs time-series-based statistical models or simple regression analysis methods. These methods often analyze the trend and periodic terms of historical load data and use moving averages or autoregressive algorithms to calculate load forecasts for future times.

[0004] However, in the practical application of power supply forecasting for new energy vehicles, the charging behavior of new energy vehicles is affected by a variety of dynamic factors such as ambient temperature, remaining vehicle power, and the randomness of the user's destination, resulting in a highly nonlinear and volatile load curve. When dealing with such scenarios, traditional load forecasting algorithms, due to their fixed time-series weight allocation, cannot effectively capture the non-steady-state characteristics of the load caused by random events. Especially in specific areas, when a large number of vehicles generate concentrated charging demand due to the same environmental factors, traditional models often fail to identify the instantaneous variation of the current data characteristics, resulting in a significant lag in the prediction results. This prediction bias not only reduces the response efficiency of the power grid but may also lead to overload of charging stations or local distribution network voltage instability due to insufficient load forecasting. Summary of the Invention

[0005] To address the technical problem that traditional load forecasting methods cannot effectively capture data fluctuation characteristics and instantaneous variability in the highly dynamic charging scenarios of new energy vehicles, resulting in low forecasting accuracy and delayed response, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a method for predicting the power supply load of new energy vehicles, comprising: acquiring real-time operating parameters of new energy vehicles in a charging station and extracting historical power supply load datasets; for any given moment: extracting the power supply load corresponding to all historical moments of the same period at that moment; determining the load fluctuation sensitivity based on the deviation of the power supply load at that moment from its historical average power supply load for the same period and the difference from the power supply load at the previous moment; determining the load demand potential based on the difference between the remaining battery power of all online controlled vehicles at that moment, the ambient temperature, and the preset optimal reference temperature of the battery; weighting and fusing the load fluctuation sensitivity and the load demand potential to determine the prediction weights; adaptively switching the prediction mode and extracting the corresponding dataset based on the relationship between the normalized value of the prediction weights and the preset weight threshold; and performing exponential smoothing prediction on the dataset based on the normalized value of the prediction weights to obtain the predicted power supply load value of the charging station at the next moment.

[0007] This invention introduces load fluctuation sensitivity and physical demand potential indicators, enabling the forecasting system to anticipate potential risks of sudden load increases in advance. Through the coordinated intervention of nonlinear enhanced weights and multi-mode switching, it maximizes the real-time tracking capability of vehicle charging demand while ensuring the stability of long-term forecasts. It also eliminates the phase lag phenomenon of traditional smoothing forecasting algorithms in high-dynamic scenarios, thereby ensuring the stability of power supply.

[0008] Preferably, the real-time operating parameters of the new energy vehicles in the charging station include: power supply load, rated power supply load, all online controlled vehicles in the charging station, and the remaining power of the controlled vehicles.

[0009] Preferably, the extraction of historical power load dataset includes: extracting historical power load data for a preset number of days from the power distribution monitoring system and the charging cloud platform database with a fixed sampling step size, and constructing a historical power load dataset.

[0010] Preferably, the step of extracting all historical contemporaneous moments of the current moment includes: extracting moments from the historical power load dataset that are exactly the same as the current moment in the daily cycle position, as all historical contemporaneous moments of the current moment.

[0011] Preferably, the load fluctuation sensitivity satisfies the expression: In the formula, For the first Sensitivity to load fluctuations at any given time; , For the first Time and the The power supply load value at any given time; , For the first The mean and standard deviation of all historical power supply loads during the same period at any given time; To avoid zero constant; The rated power load of the charging station; To take the absolute value.

[0012] This invention comprehensively assesses abnormal load conditions by calculating the deviation between real-time load and historical average for the same period, and superimposing the fluctuation amplitude within a short period of time. This ensures that power consumption fluctuations caused by sudden events are captured and converted into mode switching signals, avoiding the dilution of real sudden load characteristics by massive historical statistical data and enhancing the system's sensitivity to abnormal signals.

[0013] Preferably, the load demand potential satisfies the expression: In the formula, For the first The potential for load demand at any given moment; For all online controlled vehicles in the charging area in the first The normalized average of the remaining battery power at any given time; For the first The ambient temperature at that moment; This is the optimal reference temperature for the battery. This is the influence coefficient; It is the natural logarithm function; It is a natural exponential function.

[0014] This invention utilizes the linear transformation characteristics of the vehicle's remaining battery power mapped by a logarithmic function, and combines this with the excitation effect of ambient temperature on energy consumption by a Gaussian mapping, to achieve the prediction of upcoming charging behavior; even if the current load has not yet fully reached its peak, the system can predict the trend of load growth in the next moment, providing physical-level predictability for forecasting.

[0015] Preferably, the prediction weights satisfy the expression: In the formula, , The first The forecast weights at different times, load fluctuation sensitivity, and load demand potential; This is the balance coefficient; It is a natural exponential function.

[0016] This invention uses a nonlinear enhancement mechanism to sum and exponentially process the load fluctuation sensitivity and load demand potential. When the compensation voltage and load fluctuation increase simultaneously, the prediction weight will rise rapidly and tend to saturate, guiding the subsequent prediction model to track the current real-time data, thereby ensuring that the model still has extremely high tracking accuracy in non-steady-state charging scenarios.

[0017] Preferably, the step of adaptively switching the prediction mode and extracting the corresponding dataset based on the relationship between the normalized value of the prediction weight and the preset weight threshold includes: using max-min normalization to normalize the first... The prediction weights at time 0 are mapped to the range of 0-1, and the normalized value of the prediction weights at that time is denoted as [value]. The preset weight threshold is ;like When a peak demand period is identified, the power supply load data for the short period preceding that time is extracted as the dataset; if When the load is determined to be in a stable load zone, the power supply load data of the same period in the long-term historical data before that moment is extracted as the dataset.

[0018] This invention achieves adaptive switching of dataset extraction logic by setting weight thresholds. When an abnormal peak period is detected, the system automatically narrows the data focus to the core short-cycle interval covering the peak charging period of vehicles, thereby eliminating the phase lag caused by traditional long-cycle prediction. During the stable period, the system maintains the reference of long-cycle data to ensure that the prediction results will not frequently change due to minor noise.

[0019] Preferably, the short period ranges from 1 hour to 4 hours, and the long period ranges from 7 days to 30 days.

[0020] Secondly, the present invention provides a power supply load prediction system for new energy vehicles, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned power supply load prediction method for new energy vehicles is implemented.

[0021] By adopting the above technical solution, a computer program for predicting the power supply load of a new energy vehicle is generated and stored in a memory so that it can be loaded and executed by a processor. This allows for the creation of a terminal device based on the memory and processor, making it convenient to use.

[0022] The beneficial effects of this invention are as follows: This invention constructs an adaptive switching prediction system by leveraging the psychological mapping of users' urgency to replenish power and the incentive effect of ambient temperature on energy consumption. The system selectively extracts long and short period datasets for exponential smoothing prediction. This prediction mode ensures the stability of long period prediction in stable load areas, while achieving high-precision tracking of transient anomalies during abnormal peak periods. This significantly improves the prediction accuracy of power supply load and ensures the flexibility of power grid dispatch. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a method for predicting the power supply load of a new energy vehicle according to the present invention; Figure 2 This is a schematic diagram showing the comparison of predicted power supply loads for new energy vehicles. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0026] This invention discloses a method for predicting the power supply load of new energy vehicles, referring to... Figure 1 This includes steps S1-S5: S1. Obtain the real-time operating parameters of new energy vehicles in the charging station and extract the historical power load dataset.

[0027] It should be noted that the load of new energy vehicles is the result of the coupling of macro-level electricity consumption patterns and micro-level physical states. Historical load datasets reflect the background noise and long-term inertia of regional charging demand, while real-time vehicle operating parameters reflect the instantaneous driving force of load generation. By constructing a multi-source heterogeneous data system, we can provide underlying data support for subsequent analysis of users' sense of urgency for charging and behavioral variations.

[0028] Specifically, through smart gateways and data acquisition terminals deployed at charging stations, real-time operating parameters of new energy vehicles in the charging station are obtained in real time via standard communication protocols, including: power supply load, rated power supply load, all online controlled vehicles in the charging station, and the remaining power of the controlled vehicles; the sampling frequency of the smart gateways and data acquisition terminals is preset to a fixed sampling step size, which is 15 minutes in this embodiment of the invention, and the implementers can adjust it according to the actual situation.

[0029] Meanwhile, the power supply load data of the charging station for a preset number of days is extracted from the power distribution monitoring system and the charging cloud platform database with a fixed sampling step size to construct a historical power supply load dataset. In this embodiment of the invention, the historical power supply load dataset contains a power supply load sequence within a historical 30 days with a fixed sampling step size of 15 minutes, and interpolation algorithms are used to smooth and repair missing data to ensure the integrity and temporal consistency of the dataset.

[0030] At this point, the real-time operating parameters of the new energy vehicles in the charging station are obtained.

[0031] S2. For any given moment: Extract the power supply load corresponding to all historical moments of the same period at that moment; determine the load fluctuation sensitivity based on the deviation of the power supply load at that moment from its historical average power supply load and the difference from the power supply load at the previous moment.

[0032] It should be noted that the charging load of new energy vehicles is highly susceptible to sudden triggers, such as extreme weather or large-scale gatherings, which can cause transient peaks in the power supply load curve. In the early stages, this anomaly manifests as a deviation of the value from historical statistical patterns. If traditional smoothing algorithms are used directly, the variation characteristics will be diluted by a massive amount of historical samples. Therefore, it is necessary to analyze the degree of deviation of the load point from the stable trend in real time and determine the fluctuation sensitivity as a trigger signal for the model to switch from steady-state mode to dynamic capture mode, so that the model can sense whether the current environment is in an abnormal state.

[0033] Specifically, for any given moment, extract the moments that are exactly the same in the daily cycle from the historical power load dataset, and use them as all historical moments of the same period; then obtain the power load corresponding to all historical moments of the same period. For example, if the current moment is 11:15 on December 10th, the moments that are exactly the same in the daily cycle are 11:15 on each day in the historical power load dataset.

[0034] The load fluctuation sensitivity is determined based on the deviation of the power supply load at any given time from its historical average power supply load for the same period, and the difference from the power supply load at the previous time. This load fluctuation sensitivity satisfies the expression:

[0035] In the formula, For the first Sensitivity to load fluctuations at any given time; , For the first Time and the The power supply load value at any given time; , For the first The mean and standard deviation of all historical power supply loads during the same period at any given time; To avoid zero constant; The rated power load of the charging station; To take the absolute value.

[0036] in, For the first The normalized deviation of the power supply load value at time 1 from the mean of the power supply load values ​​for all historical periods reflects the value of the load at time 2. The degree of outlier of the power supply load at a given time relative to the same historical period; the larger this value, the more outlier the load at that time. The more significantly the electricity consumption deviates from the normal level at any given time, the greater the sensitivity to load fluctuations. For the dynamic correction operator, when the first The greater the difference in power supply load between a given moment and the previous moment, i.e. the larger the dynamic correction operator, the more likely there is a short-term change in power supply load at that moment. This indicates a possible sudden change in the scale of electricity consumption, which significantly increases the sensitivity to load fluctuations and helps the model detect possible abnormal states.

[0037] It should be added that the parameters This is used to prevent computational overflow when historical loads are extremely stable, i.e., the standard deviation is close to zero. The value range is... In this implementation, we use 0.001 to ensure numerical robustness under large-scale concurrent computing.

[0038] At this point, the load fluctuation sensitivity at each moment is obtained.

[0039] S3. Determine the load demand potential based on the difference between the remaining battery power of all online controlled vehicles at that moment, the ambient temperature, and the preset optimal reference temperature of the battery.

[0040] It should be noted that when the state of charge of vehicles in the region is generally low, it means that there is a high amount of unreleased pressure to compensate for the power shortage within the system. If the ambient temperature deviates from the optimal temperature control range at this time, such as severe cold conditions causing an increase in battery internal resistance and a surge in air conditioning heating power, the intensity of this demand will be further amplified. Although this physical demand backlog is not fully reflected in the real-time power supply load at the current moment, it foreshadows the slope and peak value of the power supply load surge in the next stage. Therefore, by analyzing the critical changes of these leading indicators, the a priori capability of the system's prediction can be ensured, thereby achieving accurate prediction of potential load surges.

[0041] Specifically, for any given moment, the remaining battery power of all online controlled vehicles at the charging station is extracted and normalized to eliminate the influence of dimensions. The normalization refers to the ratio of the remaining battery power of each online controlled vehicle at that moment to the sum of the remaining battery power of all online controlled vehicles at that moment. The normalized remaining battery power of each online controlled vehicle at that moment is obtained. Simultaneously, the ambient temperature and the battery's optimal efficiency reference temperature at that moment are obtained through the sensor interface.

[0042] Based on the remaining battery power of all online controlled vehicles at any given time, and the difference between the ambient temperature and the preset optimal battery reference temperature, the load demand potential is determined. The load demand potential satisfies the following expression:

[0043] In the formula, For the first The potential for load demand at any given moment; For all online controlled vehicles in the charging area in the first The normalized average of the remaining battery power at any given time; For the first The ambient temperature at that moment; This is the optimal reference temperature for the battery. This is the influence coefficient; It is the natural logarithm function; It is a natural exponential function.

[0044] in, By utilizing the psychological characteristic that the probability of a user charging their vehicle increases non-linearly and rapidly as the battery level decreases, based on a logarithmic function mapping, when all online controlled vehicles in the charging area are in the [missing information] phase... The lower the average normalized remaining charge at any given time, the better. The higher the value, the better the representation of the first... The more urgent the need for recharging, the better; considering that actual users complete charging when the vehicle has a remaining battery level, it is impossible for all controlled vehicles to run out of power. Numbers greater than 0; This is the environment correction term based on Gaussian mapping, when the... When the temperature at any given time is near the preset optimal reference temperature for the battery... As the value approaches 1, the environmental correction term approaches 0. At this point, the incentive effect of the environment on energy consumption is minimal, and load growth is entirely driven by the vehicle's remaining battery power. When the temperature deviates significantly from the preset optimal battery reference temperature, the energy consumption per unit mileage increases significantly due to the increased internal resistance of the battery and the surge in power consumption of the vehicle's air conditioning. The rapid decay to 0 causes the environmental correction term to climb towards its maximum value of 1, significantly increasing the [economic impact]. The indicator represents a high degree of pent-up, environment-induced urgency for power replenishment within the system, suggesting that the first... The higher the potential load demand at any given moment.

[0045] It should be added that, Extracted from existing new energy vehicle design parameter standards The range of values ​​is The temperature used in this implementation is 22 degrees Celsius, but the implementers may adjust it according to the actual situation. Using 10, in degrees Celsius, a reference temperature for the battery's optimal operating temperature is defined. Centered on, spanning approximately The 10-degree Celsius steady-state environmental range is considered to have only a slight impact on the charging load, and the implementers can adjust it according to the actual situation.

[0046] At this point, the potential load demand at each moment is obtained.

[0047] S4. Weight and integrate load fluctuation sensitivity with load demand potential to determine forecast weights.

[0048] It should be noted that during periods of smooth load fluctuations, the system should maintain the dominance of historical statistical patterns and use long-term historical datasets for forecasting to ensure the stability of the forecast results. However, during periods of high sensitivity or high physical potential variation, the correlation weight of real-time feedback information should be rapidly increased to ensure the accuracy of the forecast. Therefore, establishing adaptive forecast weights enables the forecasting system to switch forecasting modes according to the degree of change in current load characteristics, thereby effectively solving the problem of forecast failure caused by instantaneous shifts in data distribution.

[0049] Specifically, for any given moment, the load fluctuation sensitivity and load demand potential at that moment are weighted and fused to determine the prediction weights, which satisfy the following expression:

[0050] In the formula, , The first The forecast weights at different times, load fluctuation sensitivity, and load demand potential; This is the balance coefficient; It is a natural exponential function.

[0051] in, The weighted summation method was used to achieve the... The integration of moment-to-moment load fluctuation sensitivity and load demand potential; A non-linear weight enhancement mechanism is implemented through exponential form: when the first... When the sensitivity to load fluctuations at any given moment and the potential for load demand have a positive superposition effect, The model will rapidly break through the linear growth range and tend to saturate, leading to a significant reduction in the reference proportion of historical long-term averages, and instead emphasizing real-time tracking of the current non-stationary sequence.

[0052] It should be added that the balance coefficient The range of values ​​is In this implementation, 0.6 is used to balance the proportion of instantaneous disturbances and physical potential in the weighting decision. The implementers can adjust it according to the scale of the regional charging stations.

[0053] At this point, the prediction weights for each time step are obtained.

[0054] S5. Based on the relationship between the normalized value of the prediction weight and the preset weight threshold, adaptively switch the prediction mode and extract the corresponding dataset; perform exponential smoothing prediction on the dataset based on the normalized value of the prediction weight to obtain the predicted power supply load of the charging station at the next moment.

[0055] It should be noted that by applying prediction weights to the power supply load prediction scenario, the final predicted value of the power supply load of new energy vehicles in the target area is obtained by weighted synthesis of historical inertia and real-time trends. Through the intervention of prediction weights, the prediction model realizes the switch from static fitting to dynamic adaptive mode, thereby eliminating the response lag of the traditional model when the load changes suddenly, and significantly enhancing the accuracy of load prediction.

[0056] Specifically, the prediction weights are mapped to the range of 0-1 using max-min normalization to obtain the normalized value of the prediction weights at any given time.

[0057] Let the normalized value of the prediction weight at this moment be... The preset weight threshold is ;like When a peak demand period is identified, the power load data from the short period preceding that time is extracted as the dataset; if When the load is determined to be in a stable load zone, the power supply load data of the same period in the long-term historical data before that moment is extracted as the dataset.

[0058] It should be added that if the preset weight threshold is set too low, slight environmental fluctuations or sampling errors may cause frequent changes in the prediction. In this embodiment, the preset weight threshold is set to 0.75. The implementer can adjust it based on the actual prediction results. 0.75 is a relatively high filtering threshold, which can ensure that in the load stable area, the model is still dominated by the historical mean, and maintain the long-term smoothness of the prediction results.

[0059] The extraction range of the dataset is scalable for charging stations of different sizes: the range of short-term values ​​is set according to the average charging time in the charging station, usually ranging from 1 hour to 4 hours; in this embodiment, 2 hours is selected to cover the core response interval from vehicle access to peak load; the range of long-term values ​​should cover at least one complete electricity consumption cycle, usually ranging from 7 days to 30 days; in this embodiment, 30 days is selected to fully extract the periodic patterns of the same time period within the day and smooth out random fluctuations outside of holidays; implementers can adjust the period values ​​according to the actual prediction results.

[0060] Furthermore, based on the normalized value of the prediction weight at that moment, the dataset is subjected to exponential smoothing prediction to obtain the predicted power supply load value of the charging station at the next moment.

[0061] It should be added that exponential smoothing forecasting is a well-known technique, and will not be elaborated on here.

[0062] For example, Figure 2 The graph shows the comparison of power supply load prediction results for new energy vehicles, with the horizontal axis representing time and the vertical axis representing load values. Traditional algorithms predict power supply load using exponential smoothing, which results in severe phase lag during periods of load abrupt changes. When the actual load has reached its peak, the predicted value is still on an upward trend with insufficient peak height, failing to effectively respond to non-steady-state charging characteristics. In contrast, the adaptive prediction curve of this invention identifies load variation points through load fluctuation sensitivity and performs physical state perception based on load demand potential. With dynamic intervention and weight correction, it achieves near-identity with the measured curve, maximizing the fulfillment of vehicle charging needs and fully verifying its high-precision tracking capability and robustness in complex dynamic scenarios.

[0063] This invention also discloses a power supply load prediction system for new energy vehicles, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a power supply load prediction method for new energy vehicles according to this invention is implemented.

[0064] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A method for predicting the power supply load of new energy vehicles, characterized in that, include: Obtain real-time operating parameters of new energy vehicles in charging stations and extract historical power load datasets; For any given moment: extract the power supply load corresponding to all historical moments of the same period at that moment; The load fluctuation sensitivity is determined based on the deviation of the power supply load at this moment from its historical average power supply load for the same period, and the difference from the power supply load at the previous moment. Based on the difference between the remaining battery power of all online controlled vehicles at that moment, the ambient temperature, and the preset optimal battery reference temperature, the load demand potential is determined; The load fluctuation sensitivity and load demand potential are weighted and integrated to determine the forecast weight; Based on the relationship between the normalized value of the prediction weight and the preset weight threshold, the prediction mode is adaptively switched and the corresponding dataset is extracted. The dataset is subjected to exponential smoothing based on the normalized values ​​of the prediction weights to obtain the predicted power load of the charging station at the next moment.

2. The method for predicting the power supply load of a new energy vehicle according to claim 1, characterized in that, The real-time operating parameters of the new energy vehicles in the charging station include: Power supply load, rated power supply load, all online controlled vehicles in the charging station, and the remaining power of the controlled vehicles.

3. The method for predicting the power supply load of a new energy vehicle according to claim 1, characterized in that, The extraction of historical power load dataset includes: Historical power load data for a preset number of days is extracted from the power distribution monitoring system and the charging cloud platform database using a fixed sampling step size, and a historical power load dataset is constructed.

4. The power supply load prediction method for new energy vehicles according to claim 1, characterized in that, The extraction of all historical contemporaneous moments at that moment includes: Extract the times that are exactly the same as the current time in the daily cycle from the historical power load dataset, and use them as all historical times of the same period.

5. The method for predicting the power supply load of a new energy vehicle according to claim 1, characterized in that, The load fluctuation sensitivity satisfies the expression: ； In the formula, For the first Sensitivity to load fluctuations at any given time; , For the first Time and the The power supply load value at any given time; , For the first The mean and standard deviation of all historical power supply loads during the same period at any given time; To avoid zero constant; The rated power load of the charging station; To take the absolute value.

6. The method for predicting the power supply load of a new energy vehicle according to claim 1, characterized in that, The load demand potential satisfies the expression: ； In the formula, For the first The potential for load demand at any given moment; For all online controlled vehicles in the charging area in the first The normalized average of the remaining battery power at any given time; For the first The ambient temperature at that moment; This is the optimal reference temperature for the battery. This is the influence coefficient; It is the natural logarithm function; It is a natural exponential function.

7. The method for predicting the power supply load of a new energy vehicle according to claim 1, characterized in that, The prediction weights satisfy the expression: ； In the formula, , The first The forecast weights at different times, load fluctuation sensitivity, and load demand potential; This is the balance coefficient; It is a natural exponential function.

8. The power supply load prediction method for new energy vehicles according to claim 1, characterized in that, The step of adaptively switching prediction modes and extracting the corresponding dataset based on the relationship between the normalized value of the prediction weight and the preset weight threshold includes: Using max-min normalization to the th The prediction weights at time 0 are mapped to the range of 0-1, and the normalized value of the prediction weights at that time is denoted as [value]. The preset weight threshold is ;like When a peak demand period is identified, the power supply load data for the short period preceding that time is extracted as the dataset; if When the load is determined to be in a stable load zone, the power supply load data of the same period in the long-term historical data before that moment is extracted as the dataset.

9. The power supply load prediction method for a new energy vehicle according to claim 8, characterized in that, The short period ranges from 1 hour to 4 hours, and the long period ranges from 7 days to 30 days.

10. A power supply load prediction system for new energy vehicles, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a power supply load prediction method for a new energy vehicle according to any one of claims 1-9.

Images