Energy complementing path generation method and system based on big data

Abstract

The invention belongs to the technical field of path generation, and particularly relates to an energy complementing path generation method and system based on big data, and the method comprises the steps: screening out all candidate charging stations when a vehicle needs energy complementing, obtaining a predicted waiting time mean value and a standard deviation of each candidate charging station through a trained Gaussian process regression model, and obtaining an energy complementing path; calculating the waiting time uncertainty factor of each candidate charging station, calculating the risk influence coefficient of each candidate charging station in combination with the task time margin of the vehicle, further calculating the risk penalty factor of each candidate charging station, and calculating the basic recommendation score of each candidate charging station through an analytic hierarchy process based on the predicted waiting time mean value; and correcting the basic recommendation score of each candidate charging station through the risk penalty factor, taking the candidate charging station with the maximum final comprehensive recommendation score as an optimal energy complementing path point, and planning the optimal energy complementing path point into the driving route of the vehicle. According to the invention, the reliability and safety of energy complementation decision making are improved.

Description

Technical Field

[0001] The present invention relates to the field of path generation technology. More specifically, the present invention relates to a method and system for generating energy replenishment paths based on big data. Background Art

[0002] In the modern logistics system, the application of new energy electric vehicles is becoming more and more widespread, and their on-the-way recharging planning has become a key link affecting operational efficiency: when a vehicle needs to be recharged, it is necessary to select the best one from multiple candidate charging stations.

[0003] The Analytic Hierarchy Process (AHP) is a commonly used multi-criteria decision analysis method. It can quantify multiple complex criteria such as the distance to charging stations, charging prices, and charging speeds. By establishing a hierarchical model and calculating the weight scores of each option, it provides decision support for charging station selection.

[0004] However, when applying the analytic hierarchy process, existing technologies usually directly use a single deterministic value output by various prediction models as a decision input, such as "estimated waiting time"; this expected waiting time is a predicted value obtained by analyzing historical data and real-time data, which itself has certain volatility and uncertainty.

[0005] Inputting an uncertain prediction value into the AHP model as a certain value will cause the model to be unable to assess the hidden risks behind different plans: Plan A has an estimated waiting time of 10 minutes, but the prediction results fluctuate greatly, while Plan B has an estimated waiting time of 12 minutes, but the prediction results are very stable. Existing technology will prioritize Plan A, which has a higher risk, because 10 minutes is shorter than 12 minutes, thereby bringing delay risks to actual logistics distribution tasks and affecting the reliability and safety of decision-making. Summary of the Invention

[0006] In order to solve the above-mentioned technical problem of decision-making risks caused by not considering the uncertainty of the predicted input value, the present invention provides solutions in the following aspects.

[0007] In a first aspect, the present invention provides a method for generating a recharging path based on big data, comprising: when a vehicle needs recharging, based on the vehicle's current GPS coordinates and the estimated remaining range, screening out all candidate charging stations within the vehicle's reach from a charging station database; obtaining the expected waiting time mean and expected waiting time standard deviation of each candidate charging station through a trained Gaussian process regression model; calculating the basic recommendation score of each candidate charging station through a hierarchical analysis method based on the expected waiting time mean; taking the ratio of the expected waiting time standard deviation to the expected waiting time mean of each candidate charging station as the waiting time uncertainty factor of each candidate charging station, and calculating the risk impact coefficient of each candidate charging station in combination with the vehicle's mission time margin; calculating the risk penalty factor of each candidate charging station through the risk impact coefficient; correcting the basic recommendation score of each candidate charging station through the risk penalty factor to obtain a final comprehensive recommendation score for each candidate charging station, taking the candidate charging station with the largest final comprehensive recommendation score as the optimal recharging path point, and planning it into the vehicle's driving route.

[0008] The present invention simultaneously obtains the mean and standard deviation of the waiting time through the Gaussian process regression model, which represents the uncertainty of the prediction result. The uncertainty is combined with the vehicle's mission time margin, and the basic recommendation score obtained based on the traditional hierarchical analysis method is corrected through a dynamic risk penalty mechanism. The recommendation of the charging path is no longer a static selection of the charging station with the shortest theoretical waiting time, but can be adaptively adjusted according to the urgency of the task, thereby effectively avoiding the delay risk that may be caused by ignoring the prediction volatility in the existing technology, and significantly improving the reliability and safety of charging decision-making in complex real-world application scenarios.

[0009] Preferably, the training process of the Gaussian process regression model includes: obtaining daily feature data and daily average waiting time of each candidate charging station in each training cycle to form a training set, wherein the daily average waiting time is equal to the mean of the waiting time of all vehicles, and the waiting time of each vehicle is equal to the difference between the time when charging starts and the time when each vehicle arrives at the charging station; using the daily feature data of each candidate charging station in the training set as input and the daily average waiting time of each candidate charging station in the training set as output, and training a Gaussian process regression model, wherein the Gaussian process regression model constructs a probability distribution model about the input-output relationship by learning the distribution therein, including the mean and standard deviation of the expected waiting time.

[0010] The present invention uses a Gaussian process regression model to predict waiting time, which can not only predict the mean of the waiting time, but also output the standard deviation of the predicted value to indicate the confidence or uncertainty of the prediction result, providing data input for subsequent risk assessment.

[0011] Preferably, the characteristic data of the candidate charging station for each day in each training cycle include: peak charging time and low charging time in a day, whether it is a holiday, the total number of charging guns, the station type, the maximum charging power of the charging station, and the number of idle charging guns per hour of the day.

[0012] Preferably, the basic recommendation score of each candidate charging station is calculated by the hierarchical analysis method, including: establishing a hierarchical model, the hierarchical model including a target layer, a criterion layer and a solution layer, wherein the target layer refers to selecting the optimal charging station; the criterion layer includes time cost, economic cost and convenience, and the solution layer includes all the screened candidate charging stations; constructing a criterion layer judgment matrix; solving the quantitative weight of each criterion from the criterion layer judgment matrix and performing a consistency check; taking the average expected waiting time, maximum charging power, electricity price strategy and detour distance of each candidate charging station as input, and calculating the score of each candidate charging station under each criterion; and performing a weighted summation of the scores of each candidate charging station under each criterion according to the quantified weight of each criterion to obtain the basic recommendation score of each candidate charging station.

[0013] Preferably, the criterion layer includes time cost, economic cost and convenience, specifically: time cost is equal to the sum of the average expected waiting time, detour time and expected charging time; economic cost is equal to the charging price, which is obtained according to the expected charging time and the electricity price strategy of the charging station; convenience is used to measure the convenience of reaching the charging station, and convenience is equal to the inverse of the detour distance.

[0014] Preferably, the task time margin of the vehicle includes: locating the last task point that needs to be delivered from the task sequence of the vehicle, and taking the latest time point of the delivery time window of the last task point that needs to be delivered in the task sequence of the vehicle as the planned latest delivery time of all subsequent delivery points of the vehicle. Through the multi-waypoint route calculation function of the route planning service API, an optimal driving route connecting all the task points in the task sequence of the vehicle is planned, and the total driving time required to complete the route is given as the predicted remaining driving time for all subsequent delivery points of the vehicle. The product of the number of all task points in the vehicle's task sequence and the average service time is used as the predicted remaining service time for all subsequent delivery points of the vehicle. ; Vehicle mission time margin ,in, Indicates the current time.

[0015] The present invention obtains the vehicle's current mission time margin available to handle additional delays through the latest delivery time of the last mission point and the estimated driving and service time, thereby grasping the urgency of the mission in real time and providing dynamically changing input parameters for the calculation of the risk impact coefficient, so that the refueling decision is closely linked to the transportation mission goal.

[0016] Preferably, the step of calculating the risk impact coefficient of each candidate charging station in combination with the vehicle's mission time margin includes: Where, For the Risk impact coefficient of each candidate charging station; For the The uncertainty factor of the waiting time of candidate charging stations; The mission time margin for the vehicle; is the adjustment constant; is the natural exponential function.

[0017] The present invention dynamically combines the waiting time uncertainty factor of the charging station with the vehicle's mission time margin to calculate the risk impact coefficient. The system's sensitivity to risk will automatically increase with the increase in mission urgency, making risk assessment closely related to specific mission scenarios, greatly improving the scenario adaptability of decision-making.

[0018] Preferably, calculating the risk penalty factor of each candidate charging station by using the risk impact coefficient includes: Where, For the Risk penalty factor for candidate charging stations; For the The risk impact coefficient of each candidate charging station.

[0019] The present invention converts the risk impact coefficient into a risk penalty factor through a smooth nonlinear function. When the risk impact is small, the recommendation score is only slightly affected. When the risk impact increases, the penalty intensity increases rapidly, which can reduce the weight of those options that have high basic scores but also high potential risks, thereby filtering out false optimal solutions in the final ranking and ensuring that the solution finally recommended to the user is a robust choice that takes into account both benefits and risks.

[0020] Preferably, the method filters out all candidate charging stations within the reach of the vehicle from the charging station database based on the vehicle's current GPS coordinates and the estimated remaining range, including: in the charging station database, with the vehicle's current GPS coordinates as the center and the estimated remaining range as the radius, performing a fast circular area query to obtain a rough screening list of candidate charging stations; traversing each candidate charging station in the rough screening list of candidate charging stations, with the vehicle's current GPS coordinates as the starting point and the GPS coordinates of each candidate charging station as the end point, calling the path planning service API to estimate the driving navigation distance of the vehicle to each candidate charging station; multiplying the vehicle's current GPS coordinates and the estimated remaining range by a safety factor to obtain the vehicle's estimated remaining safe range; and retaining all candidate charging stations whose driving navigation distance is less than or equal to the estimated remaining safe range.

[0021] In a second aspect, the present invention provides a big data-based energy replenishment path generation system, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned big data-based energy replenishment path generation method is implemented.

[0022] By adopting the above technical solution, the above-mentioned method for generating a replenishment path based on big data is generated into a computer program and stored in a memory so as to be loaded and executed by a processor, thereby making a terminal device based on the memory and the processor for easy use.

[0023] The beneficial effects of the present invention are: The present invention simultaneously obtains the mean and standard deviation of the waiting time through the Gaussian process regression model, which represents the uncertainty of the prediction result. The uncertainty is combined with the vehicle's mission time margin, and the basic recommendation score obtained based on the traditional hierarchical analysis method is corrected through a dynamic risk penalty mechanism. The recommendation of the charging path is no longer a static selection of the charging station with the shortest theoretical waiting time, but can be adaptively adjusted according to the urgency of the task, thereby effectively avoiding the delay risk that may be caused by ignoring the prediction volatility in the existing technology, and significantly improving the reliability and safety of charging decision-making in complex real-world application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a flow chart schematically illustrating a method for generating an energy replenishment path based on big data in the present invention; Figure 2 is a flowchart schematically illustrating step S4; Figure 3 is a diagram schematically showing a comparison of the recommendation scores of each candidate charging station before and after correction when the vehicle's mission time margin is 20 minutes; Figure 4FIG. 1 is a diagram schematically showing a comparison of the recommendation scores of candidate charging stations before and after correction when the vehicle's mission time margin is 5 minutes. DETAILED DESCRIPTION

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0026] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0027] The embodiment of the present invention discloses a method for generating energy replenishment paths based on big data, referring to Figure 1 , including steps S1 to S5: S1: When the vehicle needs to be recharged, all candidate charging stations within the vehicle's reach are screened from the charging station database based on the vehicle's current GPS coordinates and estimated remaining range.

[0028] It should be noted that in order to make a comprehensive decision, all basic data related to the decision criteria must be obtained first, which includes not only traditional decision factors but also key data for assessing decision risks, namely the volatility of waiting time prediction results.

[0029] Specifically, when a vehicle needs to recharge, it filters all candidate charging stations within the vehicle's reach from the charging station database based on the vehicle's current GPS coordinates and estimated remaining range. This involves two stages, coarse and fine screening, to balance computational efficiency and accuracy. The specific method is as follows: (1) Rapid coarse screening based on straight-line distance, including: in the charging station database, with the vehicle's current GPS coordinates as the center and the estimated remaining range as the radius, a fast circular area query is performed. All charging stations falling into this circular area are preliminarily selected to form a coarse screening list of candidate charging stations; it can quickly eliminate a large number of charging stations that are obviously too far away, reducing the number of candidate charging stations from tens of thousands to dozens or hundreds, reducing the burden of accurate calculations in the next stage.

[0030] (2) Precise screening based on road network navigation: traverse each candidate charging station in the rough screening list of candidate charging stations, take the current GPS coordinates of the vehicle as the starting point and the GPS coordinates of each candidate charging station as the end point, call the path planning service API, and the path planning service API will estimate the driving navigation distance of the vehicle to each candidate charging station based on the real-time road conditions; multiply the current GPS coordinates of the vehicle and the estimated remaining cruising range by a safety factor of 0.8 to cope with the risks of additional power consumption such as uphill and traffic jams, and obtain the estimated remaining safe cruising range of the vehicle; retain all candidate charging stations that meet the driving navigation distance less than or equal to the estimated remaining safe cruising range of the vehicle, and the list of charging stations screened out in this way is the final candidate charging station that the vehicle can safely reach with the current power level.

[0031] S2: Obtain the expected waiting time mean and expected waiting time standard deviation of each candidate charging station through the trained Gaussian process regression model.

[0032] Specifically, the characteristic data of each candidate charging station in the past 24 hours is obtained and input into the trained Gaussian process regression model, which outputs two key information of each candidate charging station: the mean expected waiting time and the standard deviation of the expected waiting time. Among them, the mean expected waiting time is the core output of the model and is the predicted result of the expected waiting time of the candidate charging station. The standard deviation of the expected waiting time is the confidence level of the predicted result.

[0033] Among them, the daily characteristic data of the charging station includes: the peak and low charging times of the day, whether it is a holiday, the total number of charging guns, the station type, the maximum charging power of the charging station, and the number of idle charging guns per hour of the day; whether it is a holiday, divided into weekdays, weekends, and holidays, and unique-hot encoding is performed; the station types include logistics parks, highway service areas, and public parking lots, and are unique-hot encoded.

[0034] The training process of the Gaussian process regression model is as follows: obtain the daily feature data and average daily waiting time of each candidate charging station in each training cycle to form a training set, where the average daily waiting time is equal to the mean waiting time of all vehicles, and the waiting time of each vehicle is equal to the difference between the time when charging starts and the time when each vehicle arrives at the charging station. The training cycle is one month; use the daily feature data of each candidate charging station in the training set as input and the average daily waiting time of each candidate charging station in the training set as output to train a Gaussian process regression model. The Gaussian process regression model constructs a probability distribution model about the input-output relationship by learning the distribution therein, including the mean and standard deviation of the expected waiting time.

[0035] S3: Based on the mean expected waiting time, the basic recommendation score of each candidate charging station is calculated using the hierarchical analysis method.

[0036] It should be noted that the traditional hierarchical analysis method is used to calculate a basic score that does not take risk into consideration. This score reflects the pros and cons of the plan under ideal forecasting conditions and can be used as a benchmark to facilitate targeted risk correction in subsequent steps.

[0037] The basic recommendation score for each candidate charging station is calculated using the Analytic Hierarchy Process. The entire process can be broken down into the following five specific steps: 1. Establish a hierarchical model.

[0038] The decision-making problem of selecting the optimal charging station is decomposed into a target layer, a criterion layer, and a solution layer, so as to structure and hierarchize the complex decision-making problem. The target layer refers to the final decision-making goal, that is, selecting the optimal charging station. The criterion layer sets multiple key criteria for evaluating the pros and cons of the goal, including time cost, economic cost, and convenience. The solution layer includes all objects to be decided, that is, all candidate charging stations screened in step S1.

[0039] Among them, at the criterion level: time cost is used to comprehensively measure the total time consumed in the energy replenishment process, and the time cost is equal to the sum of the expected waiting time average, detour time and expected charging time. The expected charging time is obtained based on the vehicle's remaining power, target power and the maximum charging power of the charging station; economic cost is used to comprehensively measure the direct costs incurred in the energy replenishment process. The economic cost is equal to the charging price. The charging price is obtained based on the expected charging time and the electricity price strategy of the charging station; convenience is used to measure the convenience of reaching the charging station. Convenience is equal to the inverse of the detour distance.

[0040] 2. Construct the criterion layer judgment matrix.

[0041] By quantifying the relative importance of different criteria, a basis for calculating weights is provided. The construction of the criterion layer judgment matrix is ​​based on experience by the fleet manager, who compares each criterion in the criterion layer pairwise to determine the relative importance of each two criteria, and the value is [1,9], where 1 means that the two criteria are equally important, 3 means that the former is slightly more important than the latter, 5 means that the former is obviously more important than the latter, 7 means that the former is strongly more important than the latter, 9 means that the former is extremely more important than the latter, and 2, 4, 6, and 8 represent the intermediate values ​​between the above adjacent judgments. If the relative importance scale of criterion 1 and criterion 2 is 3, then the relative importance scale of criterion 2 and criterion 1 is .

[0042] For example, based on experience, fleet managers believe that time cost is slightly more important than economic cost. Therefore, the scale of the relative importance of time cost to economic cost is 2. Time cost is significantly more important than convenience. Therefore, the scale of the relative importance of time cost to convenience is 4. Economic cost is slightly more important than convenience. Therefore, the scale of the relative importance of economic cost to convenience is 3. From this, the criterion layer judgment matrix can be constructed: The horizontal axis of the matrix is ​​time cost, economic cost and convenience, and the vertical axis is time cost, economic cost and convenience.

[0043] 3. From the criterion layer judgment matrix, solve the quantitative weight of each criterion and perform consistency test.

[0044] 4. Take the average estimated waiting time, maximum charging power, electricity price strategy, and detour distance of each candidate charging station obtained in S1 as input and calculate the score of each candidate charging station under each criterion.

[0045] 5. Using the quantified weight of each criterion, the scores of each candidate charging station under each criterion are weighted and summed to obtain the basic recommendation score for each candidate charging station.

[0046] S4: Based on the expected waiting time mean and standard deviation of each candidate charging station, the waiting time uncertainty factor of each candidate charging station is calculated. Combined with the vehicle's mission time margin, the risk impact coefficient of each candidate charging station is calculated, which is used to calculate the risk penalty factor of each candidate charging station.

[0047] It should be noted that the uncertainty of the predicted data is used as a risk penalty, and the risk penalty is used to correct the idealized basic score obtained in S2, so that the decision-making results can effectively avoid risks.

[0048] Step S4 flow chart reference Figure 2 , including steps S401 to S404, specifically: S401: Calculate the waiting time uncertainty factor of each candidate charging station based on the expected waiting time mean and standard deviation of each candidate charging station.

[0049] It should be noted that, in order to perform a horizontal comparison between waiting times of different lengths, the present invention constructs a waiting time uncertainty factor based on the mean and standard deviation of the waiting time predictions to measure the degree of uncertainty.

[0050] Specifically, based on the estimated waiting time mean and standard deviation of each candidate charging station obtained in S1, the waiting time uncertainty factor of each candidate charging station is calculated. The specific calculation formula is: ; Where, For the The uncertainty factor of the waiting time of candidate charging stations; For the The standard deviation of the expected waiting time for each candidate charging station; For the The average expected waiting time for candidate charging stations.

[0051] The formula converts absolute fluctuations into relative fluctuations by calculating the ratio of standard deviation to mean: When it increases, even if the standard deviation Unchanged, waiting time uncertainty factor This reflects that for longer waiting times, the relative risk brought about by fluctuations of the same absolute length is lower.

[0052] It should be noted that this factor realizes the normalized measurement of the predicted risk of waiting time at different charging stations, so that the two situations of "waiting for 5 minutes, fluctuation of 2 minutes" and "waiting for 30 minutes, fluctuation of 2 minutes" can be compared horizontally.

[0053] S402: Calculate the mission time margin of the vehicle.

[0054] It should be noted that the actual impact of the risks brought by uncertainty on the task is closely related to the time requirements of the task itself. A task with tight time constraints has extremely low tolerance for any uncertainty.

[0055] Specifically, the vehicle's mission time margin is calculated as follows: from the vehicle's mission sequence, locate the last mission point that needs to be delivered, and use the latest time point of the delivery time window of this mission point as the planned latest delivery time of all subsequent delivery points of the vehicle. Through the multi-waypoint route calculation function of the route planning service API, an optimal driving route connecting all the task points in the task sequence of the vehicle is planned, and the total driving time required to complete the route is given as the predicted remaining driving time for all subsequent delivery points of the vehicle. The product of the number of all task points in the vehicle's task sequence and the average service time is used as the predicted remaining service time for all subsequent delivery points of the vehicle. The average service time refers to the average time spent on unloading, handover, and signing, which is set to 15 minutes.

[0056] Furthermore, according to the latest planned delivery time of all subsequent delivery points of the vehicle , predict remaining driving time , predict remaining service time , we get the mission time margin of the vehicle, then the mission time margin of the vehicle is ,in, Indicates the current time; the obtained mission time margin indicates that the vehicle can withstand a maximum time of Additional delays, such as for charging, dealing with sudden traffic jams, etc., if the total time spent on detours, waiting and charging for a charging solution exceeds , it will inevitably lead to task delays.

[0057] S403: Calculate the risk impact coefficient of each candidate charging station by combining the waiting time uncertainty factor of each candidate charging station and the mission time margin of the vehicle.

[0058] Specifically, the risk impact coefficient of each candidate charging station is calculated by combining the waiting time uncertainty factor of each candidate charging station and the vehicle's mission time margin. The specific calculation formula is: ; Where, For the Risk impact coefficient of each candidate charging station; For the The uncertainty factor of the waiting time of candidate charging stations; is the mission time margin of the vehicle, in minutes; It is an adjustment constant used to control the decay rate of the impact of task time margin on risk; is the natural exponential function.

[0059] in, It is an adjustment constant used to control the decay rate of the impact of task time margin on risk, representing the system's sensitivity to time margin: A larger value means that even with a small amount of time margin, the risk impact will be rapidly attenuated, and the system behaves as risk tolerant. The smaller the value, the less sensitive the system is to the time margin. Even with a longer time margin, the risk impact is still greater, and the system behaves as a risk-averse system. The value range of is set to [0.05, 0.5]. In the present invention, Set to 0.1.

[0060] The risk impact coefficient of the candidate charging station is calculated by associating the risk with the mission scenario: when the mission time margin is When it is large, the exponential term Approaching 0, the overall risk impact coefficient It also approaches 0, indicating that the impact of uncertainty can be ignored. On the contrary, when the task time margin is When it is very small or even negative, the exponential term approaches 1, and the risk impact coefficient is mainly determined by the waiting time uncertainty factor decision, indicating that uncertainty will have a significant impact.

[0061] It should be noted that the risk impact coefficient enables risk assessment to have scenario-adaptive capabilities. For tasks of different urgency levels, the system can dynamically evaluate the different levels of risk brought about by the uncertainty of the same charging station.

[0062] S404: Calculate the risk penalty factor of each candidate charging station based on the risk impact coefficient.

[0063] Specifically, the risk penalty factor of each candidate charging station is calculated based on the risk impact coefficient of each candidate charging station. The specific calculation formula is: ; Where, For the Risk penalty factor for candidate charging stations; For the The risk impact coefficient of each candidate charging station.

[0064] The calculation formula of the risk penalty factor is a smooth function with a range of values ​​between (0,1): when the risk impact coefficient When it is 0, the risk penalty factor When it is 1, it means no penalty is imposed. As the risk impact coefficient increases, the risk penalty factor decreases smoothly and approaches 0, indicating that the penalty imposed is getting stronger and stronger.

[0065] It should be noted that the risk penalty factor provides a basis for the subsequent revision of the basic recommendation score. The size of the risk penalty factor intuitively reflects the degree of discount in the recommendation degree caused by uncertainty.

[0066] S5: Using the risk penalty factor, the basic recommendation score of each candidate charging station is modified to obtain the final comprehensive recommendation score of each candidate charging station. The candidate charging station with the largest final comprehensive recommendation score is selected as the optimal charging path point and is planned into the vehicle's driving route.

[0067] It should be noted that the basic recommendation score of the candidate charging station is combined with the risk assessment to arrive at the final decision result.

[0068] Specifically, the basic recommendation score of each candidate charging station is adjusted by the risk penalty factor of each candidate charging station, that is, the product of the risk penalty factor of each candidate charging station and its basic recommendation score is used as the final comprehensive recommendation score of each candidate charging station.

[0069] Furthermore, the final comprehensive recommendation scores of all candidate charging stations are ranked, and the candidate charging station with the highest score is selected as the optimal charging path point and planned into the vehicle's driving route.

[0070] It should be noted that by multiplying the basic recommendation score with the risk penalty factor, the scores of those solutions with high uncertainty in waiting time prediction and tight vehicle mission time are significantly reduced, while those solutions with stable predictions and low risks are retained or their rankings are improved. The final recommended charging station is the optimal choice after combining expected benefits and risk control.

[0071] It should be noted that the method for recommending energy replenishment paths in the present invention can be adaptively adjusted according to the urgency of the task: when the task time is sufficient, a certain degree of uncertainty can be tolerated in exchange for better cost or time; when the task is urgent, priority is given to charging stations with more stable prediction results and lower risks, thereby effectively avoiding the risk of delays that may be caused by ignoring prediction volatility in existing technologies, and significantly improving the reliability and safety of energy replenishment decisions in complex real-world application scenarios.

[0072] For example, the candidate charging stations for a vehicle include candidate charging station A, candidate charging station B, candidate charging station C, and candidate charging station D. Among them, candidate charging station A has a shorter estimated waiting time but extremely high uncertainty, making it a high-risk option; candidate charging station B has a slightly longer estimated waiting time but extremely low uncertainty, making it a very safe low-risk option; candidate charging station C has the lowest charging price but the longest waiting and detour times, making it a medium uncertainty option; candidate charging station D has the shortest detour distance and waiting time, but also high uncertainty, making it another high-risk option. The specific analysis process is as follows: 1. Calculate the basic recommendation score for each candidate charging station using the Analytic Hierarchy Process. The basic recommendation scores for candidate charging stations A to D are 0.741, 0.269, 0.4, and 0.76, respectively. Candidate charging station D has the highest basic recommendation score due to its lowest total time cost, followed closely by candidate charging station A. 2. When the vehicle's mission time margin is 20 minutes: Candidate charging stations A and D have high standard deviations of estimated waiting times, and their calculated waiting time uncertainty factors are also correspondingly high, resulting in low risk penalty factors and large score penalties. In contrast, candidate charging stations B and C have relatively low standard deviations of estimated waiting times and are relatively stable, so their risk penalty factors are close to 1, and their final comprehensive recommendation scores remain basically unchanged. Although candidate charging stations A and D have high base scores, after multiplying by the risk penalty factor, the final comprehensive recommendation score is significantly lowered. Although candidate charging station D still has the highest final comprehensive recommendation score, the gap with candidate charging station A has narrowed. The comparison of the recommendation scores of each candidate charging station before and after the correction is shown in the figure below. Figure 3As shown in the figure, the final comprehensive recommendation scores of candidate charging stations A to D are 0.665, 0.264, 0.389, and 0.675, respectively, and the final optimal charging path point is candidate charging station D.

[0073] 3. The optimal charging route point recommendation model of the present invention has strong scenario adaptability. When the vehicle's mission time margin is 5 minutes, the risk penalty factors of candidate charging stations A and D are smaller, resulting in heavier penalties. The comparison chart of the recommendation scores of each candidate charging station before and after correction is as follows: Figure 4 As shown in the figure, the final comprehensive recommendation scores of candidate charging stations A to D are 0.456, 0.249, 0.354, and 0.447, respectively, and the final optimal charging path point is candidate charging station A.

[0074] An embodiment of the present invention also discloses a big data-based energy replenishment path generation system, including a processor and a memory, wherein the memory stores computer program instructions. When the computer program instructions are executed by the processor, a big data-based energy replenishment path generation method according to the present invention is implemented.

[0075] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface. The configuration and functions of these components are known in the art and will not be described in detail here.

Claims

1. A method for generating energy replenishment paths based on big data, characterized in that: include: When the vehicle needs to recharge, it will filter all candidate charging stations within the vehicle's reach from the charging station database based on the vehicle's current GPS coordinates and estimated remaining range. The mean and standard deviation of the expected waiting time for each candidate charging station are obtained through the trained Gaussian process regression model. Based on the mean expected waiting time, the basic recommendation score of each candidate charging station is calculated using the analytic hierarchy process; The ratio of the expected waiting time standard deviation to the expected waiting time mean of each candidate charging station is used as the waiting time uncertainty factor of each candidate charging station. Combined with the vehicle's mission time margin, the risk impact coefficient of each candidate charging station is calculated. Calculate the risk penalty factor of each candidate charging station through the risk impact coefficient; The basic recommendation score of each candidate charging station is corrected through the risk penalty factor to obtain the final comprehensive recommendation score of each candidate charging station. The candidate charging station with the largest final comprehensive recommendation score is selected as the optimal charging path point and planned into the vehicle's driving route.

2. The method for generating energy replenishment paths based on big data according to claim 1, characterized in that: The training process of the Gaussian process regression model includes: The daily feature data and average daily waiting time of each candidate charging station in each training cycle are obtained to form a training set, where the average daily waiting time is equal to the mean of the waiting times of all vehicles, and the waiting time of each vehicle is equal to the difference between the time when charging starts and the time when each vehicle arrives at the charging station. The daily feature data of each candidate charging station in the training set is used as input, and the average daily waiting time of each candidate charging station in the training set is used as output to train a Gaussian process regression model. The Gaussian process regression model constructs a probability distribution model about the input-output relationship by learning the distribution thereof, including the mean and standard deviation of the expected waiting time.

3. The method for generating energy replenishment paths based on big data according to claim 2, characterized in that: The characteristic data of the candidate charging station every day in each training cycle includes: Peak and low charging times of the day, whether it is a holiday, total number of charging guns, station type, maximum charging power of the charging station, and number of idle charging guns per hour of the day.

4. The method for generating energy replenishment paths based on big data according to claim 1, characterized in that: The calculation of the basic recommendation score of each candidate charging station by the analytic hierarchy process includes: A hierarchical model was established, consisting of a target layer, a criterion layer, and a solution layer. The target layer was to select the optimal charging station; the criterion layer included time cost, economic cost, and convenience; and the solution layer included all selected candidate charging stations. A judgment matrix for the criterion layer was constructed; the quantitative weight of each criterion was solved from the criterion layer judgment matrix, and a consistency check was performed. The average estimated waiting time, maximum charging power, electricity price strategy, and detour distance of each candidate charging station are used as input to calculate the score of each candidate charging station under each criterion. By quantifying the weight of each criterion, the scores of each candidate charging station under each criterion are weighted and summed to obtain the basic recommendation score of each candidate charging station.

5. The method for generating energy replenishment paths based on big data according to claim 4, characterized in that: The criteria include time cost, economic cost and convenience, specifically: The time cost is equal to the sum of the average expected waiting time, detour time, and expected charging time; The economic cost is equal to the charging price, which is obtained based on the expected charging time and the electricity price strategy of the charging station; Convenience is used to measure the ease of reaching a charging station, and is equal to the inverse of the detour distance.

6. The method for generating energy replenishment paths based on big data according to claim 1, characterized in that: The mission time margin of the vehicle includes: From the vehicle's task sequence, locate the last task point that needs to be delivered, and use the latest time point of the delivery time window of the last task point that needs to be delivered in the vehicle's task sequence as the planned latest delivery time for all subsequent delivery points of the vehicle Through the multi-waypoint route calculation function of the route planning service API, an optimal driving route connecting all the task points in the task sequence of the vehicle is planned, and the total driving time required to complete the route is given as the predicted remaining driving time for all subsequent delivery points of the vehicle. The product of the number of all task points in the vehicle's task sequence and the average service time is used as the predicted remaining service time for all subsequent delivery points of the vehicle. ; Vehicle mission time margin ,in, Indicates the current time.

7. The method for generating energy replenishment paths based on big data according to claim 1, characterized in that: The risk impact coefficient of each candidate charging station is calculated based on the vehicle's mission time margin, including: ； Where, For the Risk impact coefficient of each candidate charging station; For the The uncertainty factor of the waiting time of candidate charging stations; The mission time margin for the vehicle; is the adjustment constant; is the natural exponential function.

8. The method for generating energy replenishment paths based on big data according to claim 1, characterized in that: The risk penalty factor of each candidate charging station is calculated using the risk impact coefficient, including: ； Where, For the Risk penalty factor for candidate charging stations; For the The risk impact coefficient of each candidate charging station.

9. The method for generating energy replenishment paths based on big data according to claim 1, characterized in that: Based on the vehicle's current GPS coordinates and estimated remaining range, all candidate charging stations within the vehicle's reach are screened from the charging station database, including: In the charging station database, a quick circular area query is performed with the vehicle's current GPS coordinates as the center and the estimated remaining range as the radius to obtain a rough list of candidate charging stations. Traverse each candidate charging station in the rough screening list of candidate charging stations, starting with the vehicle's current GPS coordinates and the GPS coordinates of each candidate charging station as the end point, call the path planning service API, and estimate the driving navigation distance from the vehicle to each candidate charging station; multiply the vehicle's current GPS coordinates and the estimated remaining range by the safety factor to obtain the vehicle's estimated remaining safe range; retain all candidate charging stations whose driving navigation distance is less than or equal to the estimated remaining safe range.

10. A big data-based energy replenishment path generation system, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a method for generating an energy replenishment path based on big data according to any one of claims 1 to 9 is implemented.

Images