A charging load prediction method for an unmanned vehicle

Abstract

This invention relates to the field of intelligent traffic load prediction technology, and particularly to a method for predicting the charging load of autonomous vehicles. The method includes: acquiring the traffic network topology of the operating area, historical travel order data, and vehicle energy consumption characteristic parameters; generating a set of simulated travel tasks for a prediction period using an improved driving trajectory simulation algorithm with dynamic weight correction based on traffic conditions; calculating the energy consumption of each task sequentially to construct a cumulative energy consumption curve for a single vehicle; and determining the charging start time and required charging amount based on the remaining battery power threshold. Finally, integrating all vehicle charging demand trigger points and considering the distribution and capacity constraints of charging facilities, a spatiotemporal distribution prediction of the regional charging load is generated. This method can improve the consistency between trajectory simulation and actual paths, refine the accuracy of energy consumption calculation, accurately determine the timing and amount of charging demand, and enhance the realism and reliability of the spatiotemporal distribution prediction of charging load.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation load forecasting. In particular, it relates to a method for forecasting the charging load of autonomous vehicles. Background Technology

[0002] Current methods for predicting charging load for autonomous vehicles largely rely on historical charging data statistics, fixed path planning, and overall average energy consumption estimation. Driving trajectory simulation uses a static weighted random path selection algorithm, with path selection probability weights remaining constant. Charging demand is triggered by a fixed energy ratio or fixed time period, and charging load prediction simply sums the charging demands of multiple vehicles without considering the distribution of charging facilities and capacity constraints for spatiotemporal adaptation calculations.

[0003] The static weighted random path selection algorithm does not incorporate real-time traffic state parameters, and path weights cannot be dynamically adjusted, resulting in discrepancies between simulated travel trajectories and actual vehicle paths. Energy consumption is roughly calculated using interval averages instead of being calculated for each individual trip, leading to insufficient accuracy in cumulative energy consumption figures. The charging demand triggering conditions are set too simplistically, failing to match the energy decay patterns of continuous vehicle travel, and resulting in low accuracy in determining the charging start time and required charging amount. The combined effects of trajectory simulation distortion, coarse energy consumption calculations, and charging demand positioning errors cause significant discrepancies between the predicted spatiotemporal distribution of regional charging load and the actual situation, making it difficult to adapt to dynamic traffic environments and charging facility capacity limitations.

[0004] Random route selection lacks a dynamic weight correction mechanism linked to traffic conditions, resulting in insufficient relevance to simulated travel tasks. It cannot generate a continuous cumulative energy consumption curve through task-by-task energy consumption calculations, making it difficult to accurately determine the charging demand trigger time and corresponding charging amount by combining battery remaining power thresholds, and thus failing to generate a charging load spatiotemporal distribution that accurately matches actual conditions. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a method for predicting the charging load of autonomous vehicles.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for predicting the charging load of unmanned vehicles, comprising: Acquire traffic network topology, historical trip order data, and vehicle energy consumption characteristics parameters of the operating area of ​​autonomous vehicles; Based on the traffic network topology and the historical travel order data, an improved vehicle trajectory simulation algorithm is used to generate a set of simulated travel tasks for vehicles within a predetermined prediction period. The improved vehicle trajectory simulation algorithm dynamically adjusts the static weight random path selection algorithm according to the current traffic conditions. The energy consumption of each simulated travel task is calculated, and the cumulative energy consumption curve of each autonomous vehicle in the predetermined prediction period is generated by combining the vehicle energy consumption characteristic parameters. Based on the cumulative energy consumption curve and the preset vehicle battery remaining power threshold, the simulated charging demand trigger point for each autonomous vehicle is determined. The simulated charging demand trigger point includes the charging start time and the required charging amount. By integrating the simulated charging demand trigger points of all autonomous vehicles, and based on the distribution and capacity constraints of charging facilities, a spatiotemporal distribution prediction of charging load in the autonomous vehicle operating area within the predetermined prediction period is generated.

[0007] As a further aspect of the present invention, the step of generating a set of simulated travel tasks for a vehicle within a predetermined prediction period using an improved vehicle trajectory simulation algorithm includes: Based on the historical travel order data, the probability distribution matrix of the origin and destination points and the probability distribution function of the travel distance are extracted for different date types and time periods; Based on the real-time traffic information of the traffic network topology, a dynamic travel time weight is assigned to each road segment in the road network. The dynamic travel time weight is calculated based on the historical average speed of the road segment and the traffic flow at the current moment. Using the improved vehicle trajectory simulation algorithm, a series of consecutive trip origin-destination pairs are randomly generated for each simulated vehicle within the predetermined prediction period, based on the trip origin-destination probability distribution matrix. For each of the trip origin-destination pairs, the expected trip distance is determined according to the trip distance probability distribution function, and a simulated travel path is planned in the traffic network topology based on the dynamic trip time weight and the shortest time path search strategy. The simulated travel routes, corresponding origin and destination points, and departure times of all simulated vehicles are combined to form the simulated travel task set.

[0008] As a further aspect of the present invention, the improved vehicle trajectory simulation algorithm dynamically adjusts the static weight random path selection algorithm based on the current traffic conditions, including: In the static weighted random path selection algorithm, the probability of a vehicle choosing the next path segment at a path intersection depends only on the static length of the path or the free-flow time. The improved vehicle trajectory simulation algorithm introduces the dynamic travel time weight as a core decision variable; When a simulated vehicle arrives at a path intersection, the algorithm obtains the real-time dynamic travel time weights from the path intersection to each downstream road segment. Calculate the reciprocal of the dynamic travel time required to reach each downstream road segment, and normalize the reciprocal to obtain the real-time probability of the vehicle choosing each downstream road segment; Based on the real-time probabilities, the roulette wheel selection method is used to determine the next road segment that the simulated vehicle will actually travel on. This dynamic correction process enables the simulated vehicle's route selection to respond in real time to changes in traffic congestion, thereby generating the simulated driving path that more closely resembles the actual traffic flow.

[0009] As a further aspect of the present invention, the energy consumption of the simulated travel task set is calculated on a task-by-task basis, and the cumulative energy consumption curve of each autonomous vehicle within the predetermined prediction period is generated by combining the vehicle energy consumption characteristic parameters, including: The energy consumption benchmark value per unit mileage, the air conditioning energy consumption coefficient, and the correlation function between vehicle speed and energy consumption are obtained from the vehicle energy consumption characteristic parameters. For each simulated travel task in the set of simulated travel tasks, extract the total length of the simulated travel path, the predicted average speed of the road segments traversed, and the predicted ambient temperature during task execution. The basic driving energy consumption of the simulated travel task is calculated based on the total length and the energy consumption benchmark value per unit mileage. Based on the predicted average vehicle speed, the correlation function between vehicle speed and energy consumption is queried to obtain the vehicle speed correction factor, which is then used to correct the basic driving energy consumption. Based on the deviation between the predicted ambient temperature and the comfort temperature threshold, and in conjunction with the air conditioning energy consumption coefficient, the additional energy consumption for temperature control in the simulated travel task is calculated. The estimated energy consumption of the simulated travel task is obtained by summing the corrected basic driving energy consumption of the simulated travel task with the additional energy consumption of temperature control. The estimated power consumption of an autonomous vehicle for all simulated travel missions is accumulated in chronological order to generate the cumulative power consumption curve of the autonomous vehicle over time within the predetermined prediction period.

[0010] As a further aspect of the present invention, based on the cumulative energy consumption curve and a preset vehicle battery remaining power threshold, the simulated charging demand trigger point for each autonomous vehicle is determined, including: Set the initial battery state of charge when the vehicle begins to execute the set of simulated travel tasks; Moving along the time axis, the energy consumption at the corresponding moment on the cumulative energy consumption curve is continuously subtracted from the initial battery state of charge to calculate the remaining battery power in real time. When the remaining charge of the simulated battery first falls below the preset vehicle battery remaining charge threshold, the charging start time triggered by the current charging demand is recorded. Based on the total capacity of the vehicle battery, the current remaining charge of the simulated battery, and the preset target state of charge, the required amount of electrical energy replenishment from the triggering time is calculated, and the amount of electrical energy replenishment is used as the required charging amount. The charging start time and the required charging amount are recorded together to form a simulated charging demand trigger point; If the remaining charge of the simulated battery falls below the vehicle battery's remaining charge threshold again before the end of the predetermined prediction period, then the process returns to the step of recording the current time as the charging start time to determine the next simulated charging demand trigger point.

[0011] As a further aspect of the present invention, the step of generating a spatiotemporal distribution prediction of charging load in the operating area of ​​the autonomous vehicle within the predetermined prediction period based on the distribution and capacity constraints of charging facilities includes: Obtain the physical location information, charging power specification information, and current occupancy status information of all charging facilities within the operating area of ​​the autonomous vehicle; For each of the simulated charging demand trigger points, based on its charging start time and the vehicle's current location, a set of accessible candidate charging facilities is searched in the spatial and temporal dimensions. Based on the required charging amount and the charging power specifications of the candidate charging facilities, the estimated time required to complete charging at each of the candidate charging facilities is determined. A charging facility selection optimization model is constructed with the goal of minimizing total waiting time and travel cost. The charging facility selection optimization model must meet the charging facility capacity constraint, that is, the number of charging tasks allocated to the same charging facility in the same time period shall not exceed the number of its idle charging interfaces. Solve the charging facility selection optimization model to assign a specific charging facility and an expected charging time period to each simulated charging demand trigger point; The charging power demand generated by all charging tasks within their respective allocated charging time periods and on the corresponding charging facilities is summarized to form a spatiotemporal distribution prediction of the charging load with time as the horizontal axis and spatial location as the vertical axis.

[0012] As a further aspect of the present invention, the construction of the charging facility selection optimization model with the objective of minimizing total waiting time and travel cost includes: Define a decision variable, which is a binary variable, representing whether a certain simulated charging demand trigger point is allocated to a certain charging time window of a certain charging facility; Construct an objective function, which is a weighted sum of the total time cost of traveling to the charging facility, the expected queuing time cost at the charging facility, and the additional travel time cost caused by deviating from the original planned route due to charging. The objective is to minimize the value of the objective function. The constraints include: each simulated charging demand trigger point must be assigned to one and only one charging time window; for any charging time window of any charging facility, the total number of charging interfaces required for the charging task assigned to the charging window shall not exceed the number of available charging interfaces of the charging facility within the charging window; the start time of the charging task shall not be earlier than the charging demand trigger time. The optimization model is solved using a mixed-integer linear programming solver to obtain the optimal solution for the decision variables.

[0013] As a further aspect of the present invention, based on real-time traffic information of the traffic network topology, a dynamic travel time weight is assigned to each road segment in the road network, including: The system obtains historical average speed data of all road segments in the traffic network topology within a predetermined prediction period from the real-time traffic information platform, as well as traffic flow data of each road segment at the current moment collected in real time by sensors or vehicle communication devices deployed on the roadside. For each road segment in the road network, the historical average speed data is used as the benchmark free-flow speed for that road segment, and a benchmark traffic capacity is set for the road segment according to the road grade and design standards. At each dynamically updated simulation moment, based on the collected traffic flow data of each road segment at the current moment, the ratio of the current traffic flow of the processed road segment to its baseline traffic capacity is calculated, i.e., the current saturation. Based on the current saturation, query the preset saturation-velocity reduction relationship table to obtain the velocity reduction coefficient corresponding to the current saturation; Multiply the baseline free-flow velocity by the velocity reduction factor to obtain the estimated traffic speed of the processed road segment at the current simulation time; Based on the static length and estimated traffic speed of the processed road segment, the dynamic travel time required for vehicles to pass through the processed road segment is calculated, and this dynamic travel time is assigned as a dynamic travel time weight to the processed road segment. The calculated dynamic travel time weights for all road segments are updated in the corresponding road segment attributes of the traffic network topology, so that the improved vehicle trajectory simulation algorithm can call them during path planning.

[0014] As a further aspect of the present invention, based on the deviation between the predicted ambient temperature and the comfort temperature threshold, and in conjunction with the air conditioning energy consumption coefficient, the additional energy consumption for temperature control in the simulated travel task is calculated, including: Obtain a preset vehicle cabin comfort temperature threshold, which includes a lower temperature limit and a higher temperature limit; Read the predicted ambient temperature during the simulated travel mission execution period; Determine whether the predicted ambient temperature is lower than the lower limit of the comfort temperature threshold, or higher than the upper limit of the temperature threshold; If the predicted ambient temperature is lower than the lower limit of the temperature, then the first difference between the lower limit of the temperature and the predicted ambient temperature is calculated, and the first difference is multiplied by the preset heating condition air conditioning energy consumption coefficient to obtain the heating additional energy consumption, and the heating additional energy consumption is used as the temperature control additional energy consumption. If the predicted ambient temperature is higher than the upper limit of temperature, then calculate the second difference between the predicted ambient temperature and the upper limit of temperature, multiply the second difference by the preset cooling mode air conditioning energy consumption coefficient to obtain the additional cooling energy consumption, and use the additional cooling energy consumption as the additional temperature control energy consumption. If the predicted ambient temperature is within the comfort temperature threshold range, then the additional energy consumption for temperature control is set to zero.

[0015] As a further aspect of the present invention, the method further includes a predictive dynamic update step based on real-time vehicle status information: Receive real-time reports from the autonomous vehicle fleet on vehicle location, current battery charge status, and current task execution. The real-time reported vehicle location information and current battery charge status information are compared with the predicted status of the corresponding vehicle in the simulated travel task set to calculate the location deviation and battery charge deviation. When the position deviation or power deviation exceeds the preset update threshold, the prediction update process is triggered; The real-time reported vehicle geographic location information is used as a new starting point, the current battery state of charge information is used as a new initial battery state, and the subsequent travel task assumptions are updated based on the current task execution information. The process is repeated from generating a set of simulated travel tasks for the vehicle within the remaining prediction period using an improved vehicle trajectory simulation algorithm until an updated prediction of the spatiotemporal distribution of charging load is generated.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: The static weighted random path selection algorithm is dynamically adjusted based on the current traffic conditions, changing the fixed weight calculation mode of the static weighted algorithm by adjusting the path selection probability distribution in real time. The path selection logic is linked and matched with the real-time operation status of the traffic network, simulating travel trajectories that closely match actual vehicle travel paths, reducing path and timing deviations caused by fixed weights. The simulated travel task set can realistically reflect the dynamic changes in regional traffic, improving the accuracy of travel task simulation.

[0017] The system performs independent energy consumption calculations for each simulated travel task, sequentially accumulating these calculations to form a continuous cumulative energy consumption curve for each autonomous vehicle. This cumulative energy consumption curve fully reflects the battery depletion process during continuous travel, ensuring that the energy consumption values ​​are consistent with actual travel consumption. Combined with a preset battery remaining capacity threshold, the system can accurately pinpoint the timing of charging demand triggers and simultaneously determine the corresponding required charging amount, thus improving the accuracy of charging demand parameter determination.

[0018] By integrating all vehicle charging demand trigger points and considering the spatial distribution and capacity constraints of charging facilities, spatiotemporal calculations are performed to generate regional charging load spatiotemporal distribution results. The charging load is consistent with actual charging demand in terms of time series, spatial layout, and load amplitude, mitigating distribution biases caused by coarse energy consumption estimation, delayed demand triggering, and direct load superposition in static weighted prediction methods. The charging load prediction results are adaptable to actual operating constraints of charging facilities, and the spatiotemporal distribution closely matches the real charging behavior characteristics of autonomous vehicles. Attached Figure Description

[0019] Figure 1 This is a flowchart of a charging load prediction method for an autonomous vehicle according to the present invention. Figure 2 A flowchart for generating a set of simulated travel tasks for an improved vehicle trajectory simulation algorithm; Figure 3 A flowchart for generating cumulative energy consumption curves for task-by-task energy consumption calculation. Detailed Implementation

[0020] The following reference Figures 1 to 3This invention describes a method for predicting the charging load of an autonomous vehicle according to an embodiment of the present invention. In this description, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first" and "second" may explicitly or implicitly include at least one of that feature, that is, include one or more of that feature. In the description of the present invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified. When a feature "includes or contains" one or more of the features it encompasses, unless otherwise specifically described, this indicates that other features are not excluded and may be further included.

[0021] In the description of this embodiment, the terms "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0022] See Figure 1 This invention provides a method for predicting the charging load of autonomous vehicles, the overall implementation of which includes: The system acquires traffic network topology data, historical trip order datasets, and vehicle energy consumption characteristic parameter sets for the autonomous vehicle operating area. Based on the traffic network topology and historical trip order data, an improved vehicle trajectory simulation algorithm is used to generate a set of simulated trip tasks within a predetermined prediction period. This improved algorithm corrects the statically weighted random path selection by introducing dynamic trip time weights, ensuring that path selection reflects real-time traffic conditions. Energy consumption is calculated for each task in the simulated trip task set, and combined with the vehicle energy consumption characteristic parameters, a cumulative energy consumption curve is generated for each vehicle within the predetermined prediction period. Based on the cumulative energy consumption curve and a preset vehicle battery remaining power threshold, the simulated charging demand trigger point for each vehicle is determined. This trigger point includes the charging start time and the required charging amount. By integrating the simulated charging demand trigger points of all vehicles and combining the distribution information and capacity limitations of charging facilities, a spatiotemporal distribution prediction of charging load in the operating area within the predetermined prediction period is generated.

[0023] In one embodiment of the present invention, see [reference] Figure 2Based on historical travel order data, the probability distribution matrix of travel origin and destination points and the probability distribution function of travel distance are extracted for different date types and time periods. According to the traffic network topology, the historical average speed of each road segment's time slice is obtained from the real-time traffic information platform, and real-time traffic flow is collected through roadside sensors or vehicle-mounted equipment. For each road segment, its historical average speed is used as the baseline free-flow speed, and a baseline traffic capacity is set in conjunction with the road grade. The setting of the baseline traffic capacity is specifically based on the design capacity requirements for different road grades in the national urban road engineering design specifications. Corresponding baseline hourly traffic capacity is assigned to expressways, arterial roads, and secondary arterial roads, and the final baseline traffic capacity of the target road segment is determined based on the number of lanes. At each simulation update time, the ratio of the current traffic flow to the baseline capacity of the road segment is calculated to obtain the current saturation. The speed reduction coefficient is obtained by querying the preset saturation-speed reduction relationship table. The baseline free-flow speed is multiplied by the reduction coefficient to obtain the current estimated traffic speed. Then, the dynamic travel time weight is calculated based on the road segment length and updated in the road network attributes. The saturation-speed reduction table stores the correspondence between road saturation and vehicle speed reduction coefficients. The data is obtained by analyzing the historical average speed and traffic flow data of each road segment in the traffic network topology.

[0024] An improved vehicle trajectory simulation algorithm is used to randomly generate a continuous sequence of origin-destination pairs for each simulated vehicle within a predetermined prediction period, based on the probability distribution matrix of origin and destination points. For each pair, the expected travel distance is determined by combining the travel distance probability distribution function. This function is obtained through statistical analysis of the travel distances of a large number of trips in historical travel order data, and it characterizes the distribution pattern of travel distances under specific date types and time periods. When determining the expected travel distance, the algorithm calls the corresponding probability distribution function based on the spatiotemporal category of the currently simulated origin-destination pair, calculates the statistical average of the travel distance under that distribution, and uses this as the expected distance for this simulated trip. Furthermore, a shortest-time path search strategy is employed in the traffic network topology, planning the simulated travel path based on dynamic travel time weights. The shortest-time path search strategy refers to the algorithmic process of finding the route sequence with the minimum total travel time from the origin to the destination within the traffic network topology, using dynamic travel time weights as path costs. The core of this strategy lies in using the real-time updated dynamic travel time weights of road segments as edge weights in the graph search. Through appropriate shortest path algorithms, such as graph search algorithms based on greedy or dynamic programming, the algorithm finds the feasible path with the minimum total weight among all non-negative paths, which becomes the simulated driving path for the vehicle. In static weighted random path selection algorithms, the vehicle's selection probability at intersections is based solely on static length or free-flow time. The improved algorithm uses dynamic travel time weights as the core decision variable. When the simulated vehicle arrives at a path intersection, the algorithm obtains the real-time dynamic travel time weights of each downstream road segment, calculates the reciprocal of the dynamic travel time in each direction, and normalizes it to obtain the real-time probability of the vehicle selecting each downstream road segment. A roulette wheel selection method is then used to determine the next road segment to actually travel on. The specific steps of the roulette wheel selection method are as follows: Based on the calculated selection probabilities of each downstream road segment, a continuous interval is assigned to each road segment within a numerical range of zero to one. The length of the interval is equal to the selection probability of that road segment. A random number is generated and uniformly distributed within this numerical range. The downstream road segment corresponding to the interval in which the random number falls is determined as the next road segment to be traveled. This dynamic correction process allows the route selection to respond to changes in traffic congestion, generating a more realistic simulated driving path. The routes, origin and destination points, and departure times of all simulated vehicles together constitute the set of simulated travel tasks.

[0025] In its implementation, this embodiment relies on a pre-built traffic network topology and accumulated historical travel order data. This data covers both weekdays and weekends, and is divided into time periods with 15-minute intervals to statistically analyze travel patterns. The origin-destination probability distribution matrix extracted from the historical travel order data records the probability of a passenger departing from any traffic zone i and arriving at traffic zone j within each time period. The travel distance probability distribution function fits the frequency distribution of mileage between different origins and destinations. This function is constructed through statistical analysis of the massive amount of completed trip distance data recorded in the historical travel order dataset. Specifically, historical data is categorized according to date type and time period. For each category, statistical fitting techniques, such as parameter estimation or kernel density estimation, are used to characterize the distribution characteristics of the trip distance, thus obtaining the travel distance probability distribution function for that category. During the simulation, the algorithm calls the corresponding probability distribution function based on the date type and time period corresponding to the current simulation task to determine the expected travel distance. In its implementation, the improved vehicle trajectory simulation algorithm generates a set of simulated travel tasks based on the above probability distribution and introduces dynamic trip time weighting to correct path selection logic. The modification to the path selection logic specifically involves the algorithm no longer relying on fixed static road network attributes when the simulated vehicle reaches each intersection. Instead, it acquires the dynamic travel time weights of each downstream road segment in real time. The reciprocal of the dynamic travel time required to reach each downstream road segment is calculated and normalized to obtain a set of path selection probabilities closely related to real-time traffic conditions. Based on this probability distribution, a roulette wheel selection method is used to determine the next road segment the vehicle will actually travel on, enabling the simulated vehicle's path selection to dynamically respond to changes in traffic congestion.

[0026] Historical average speed data for all road segments in the traffic network topology at the same time over the past 30 days is obtained from a real-time traffic information platform and set as the baseline free-flow speed for each road segment. Simultaneously, baseline traffic capacities are set for expressways, arterial roads, and secondary arterial roads according to urban road design specifications. Traffic flow data for each road segment at the current time is collected using loop detectors deployed on the roadside and vehicle-mounted GPS devices. For each road segment in the network, at each dynamically updated simulation time, the ratio of the current traffic flow to the baseline traffic capacity is calculated to obtain the current saturation level. Based on a preset saturation-speed reduction table, the speed reduction coefficient is set to 0.9 when the current saturation level is 0.5, and to 0.6 when the current saturation level is 0.8. The baseline free-flow speed is multiplied by the speed reduction coefficient to obtain the estimated travel speed at the current simulation time. Then, the dynamic travel time is calculated based on the road segment length and the estimated travel speed, and this dynamic travel time is assigned as a dynamic travel time weight to the corresponding road segment. In specific implementation, the formula for calculating the dynamic travel time weight is expressed as: in: This represents the dynamic travel time weight of the road segment connecting node i and node j at time t. Represents the static length of the road segment. The baseline free-flow velocity representing the road segment. This represents the current saturation level of the road segment at time t. This represents the saturation-speed reduction mapping function. After calculation, the dynamic travel time weights of all road segments are updated in real time to the attribute table of the traffic network topology.

[0027] In practical implementation, continuous travel tasks within a predetermined prediction period are generated for each simulated vehicle. Based on the current date type and time period, the probability distribution matrix of the origin and destination points is queried, and a continuous sequence of origin-destination pairs is randomly generated. For a single origin-destination pair, the travel distance probability distribution function is queried to determine the expected travel distance. A path search algorithm is then invoked in the traffic network topology with the goal of minimizing travel time. During the search process, the link cost at each step uses the latest dynamic travel time weight. While the statically weighted random path selection algorithm allocates selection probabilities at intersections based on fixed geometric lengths, in the improved vehicle trajectory simulation algorithm, when the simulated vehicle arrives at a path intersection, the system obtains the real-time dynamic travel time weights from the intersection to each downstream segment. The reciprocal of the dynamic travel time required to reach each downstream segment is calculated, and the reciprocals for all directions are normalized so that the sum of the probabilities for each direction is 1. Assuming there are three downstream directions at the intersection, with dynamic travel time weights of 10 minutes, 15 minutes, and 20 minutes respectively, the calculated normalized selection probabilities are 0.5, 0.33, and 0.17. Based on these real-time probabilities, a roulette wheel selection method is used to determine the actual travel segment for the simulated vehicle. Specifically, a conceptual probability roulette wheel is constructed based on the calculated real-time selection probabilities of each downstream segment, where the size of the sector occupied by each segment is proportional to its probability value. A random number within the range [0,1) is generated and mapped onto the roulette wheel; the downstream segment corresponding to the area where the random number lands is selected as the next actual travel segment for the simulated vehicle. This allows the route selection behavior to fluctuate with traffic congestion conditions, generating a more realistic simulated travel path. After all the routes, origins, destinations, and departure times for all simulated vehicles have been generated, they are combined to form a complete set of simulated travel tasks.

[0028] In one embodiment of the present invention, see [reference] Figure 3The system extracts the baseline energy consumption per unit mile, air conditioning energy consumption coefficient, and the correlation function between vehicle speed and energy consumption from vehicle energy consumption characteristic parameters. For each task in the simulated travel task set, it extracts the total length of the simulated travel path, the predicted average speed of the traversed road segments, and the predicted ambient temperature during the task execution period. Based on the total length and the baseline energy consumption per unit mile, the system calculates the basic driving energy consumption for the task. Using the predicted average speed, it queries the correlation function between speed and energy consumption to obtain a speed correction factor, which is then used to correct the basic driving energy consumption. Using the predicted average speed of the road segments traversed by the current simulated travel task as the query key, it performs a matching query in the discrete mapping table to directly obtain the corresponding value, which is the speed correction factor. Multiplying the calculated basic driving energy consumption by this speed correction factor yields the corrected basic driving energy consumption.

[0029] Obtain a preset vehicle cabin comfort temperature threshold, which includes a lower and upper temperature limit. Read the predicted ambient temperature during the simulated travel task execution period and determine whether it is below the lower temperature limit or above the upper temperature limit. If it is below the lower limit, calculate the first difference between the lower limit and the predicted value, and multiply this difference by a preset heating mode air conditioning energy consumption coefficient to obtain the heating additional energy consumption, which is used as the temperature control additional energy consumption. If it is above the upper limit, calculate the second difference between the predicted value and the upper limit, and multiply this difference by a preset cooling mode air conditioning energy consumption coefficient to obtain the cooling additional energy consumption, which is used as the temperature control additional energy consumption. If the ambient temperature is within the comfort threshold range, set the temperature control additional energy consumption to zero. Add the corrected base driving energy consumption to the temperature control additional energy consumption to obtain the estimated power consumption of the simulated task. Accumulate the estimated consumption of all tasks for a single vehicle in chronological order to generate its cumulative power consumption curve within a predetermined prediction period.

[0030] In practical implementation, the energy consumption calculation module reads vehicle energy consumption characteristic parameters from a pre-configured parameter library, specifically including the baseline energy consumption per unit mile, the air conditioning energy consumption coefficient, and the correlation function between vehicle speed and energy consumption. The baseline energy consumption per unit mile is set to 0.18 kWh per kilometer. The air conditioning energy consumption coefficient is divided into heating and cooling modes. The correlation function between vehicle speed and energy consumption is stored in a discrete mapping table, recording an energy consumption correction factor of 1.0 for a speed of 60 km / h, 1.15 for 80 km / h, and 1.35 for 100 km / h. For each simulated travel task in the simulated travel task set, the data processing unit extracts the total length of the simulated driving path, the predicted vehicle speed for each road segment, and the predicted ambient temperature for the task execution period provided by weather forecasts. For example, a simulated travel task might have a total length of 12 kilometers, an average predicted vehicle speed of 80 km / h, and a predicted ambient temperature of -5 degrees Celsius.

[0031] In practice, the basic driving energy consumption is calculated by multiplying the total length of the simulated trip by the baseline energy consumption per unit mileage. For a trip with a total length of 12 kilometers, the basic driving energy consumption equals 12 multiplied by 0.18, resulting in 2.16 kWh. The correlation function between vehicle speed and energy consumption is then looked up based on the predicted average vehicle speed to obtain the corresponding energy consumption correction factor. When the predicted average vehicle speed is 80 kilometers per hour, the energy consumption correction factor is found to be 1.15. This correction factor is used to adjust the basic driving energy consumption. The corrected basic driving energy consumption equals the basic driving energy consumption multiplied by the energy consumption correction factor, i.e., 2.16 multiplied by 1.15, resulting in 2.484 kWh.

[0032] It is understandable that the vehicle cabin comfort temperature threshold is set to a lower limit of 18 degrees Celsius and an upper limit of 26 degrees Celsius. The predicted ambient temperature during the simulated travel task execution period is read, and its relationship with the comfort temperature threshold is determined. If the predicted ambient temperature is lower than the lower limit, the first difference between the lower limit and the predicted value is calculated, and this first difference is multiplied by the air conditioning energy consumption coefficient for heating mode to obtain the additional heating energy consumption. If the predicted ambient temperature is higher than the upper limit, the second difference between the predicted value and the upper limit is calculated, and this second difference is multiplied by the air conditioning energy consumption coefficient for cooling mode to obtain the additional cooling energy consumption. If the predicted ambient temperature is within the threshold range, the additional temperature control energy consumption is set to zero. In this example, the predicted ambient temperature is -5 degrees Celsius, lower than the lower limit of 18 degrees Celsius, and the first difference is 23 degrees Celsius. Assuming the air conditioning energy consumption coefficient for heating mode is 0.05 kWh per degree Celsius, and the task duration is 0.25 hours, the formula for calculating the additional temperature control energy consumption is: in: This represents the additional energy consumption for temperature control. The energy consumption coefficient of an air conditioner represents its heating operation. This represents the lower limit of the comfortable temperature. Represents the predicted ambient temperature. This represents the duration of the task. Substituting the numerical values, the additional energy consumption for temperature control equals 0.05 multiplied by 23 and then multiplied by 0.25, resulting in 0.2875 kWh. In some embodiments, the energy consumption coefficient of the air conditioner in cooling mode differs from that in heating mode to adapt to high-temperature scenarios in summer.

[0033] Optionally, the corrected base driving energy consumption and the additional energy consumption for temperature control are summed to obtain the estimated energy consumption for the simulated task. Under the above data conditions, the estimated energy consumption equals 2.484 plus 0.2875, totaling 2.7715 kWh. The estimated energy consumption of all simulated travel tasks for the same autonomous vehicle within the predetermined prediction period is accumulated chronologically to generate the vehicle's cumulative energy consumption curve. The horizontal axis of the cumulative energy consumption curve represents time, and the vertical axis represents cumulative energy consumption. The curve exhibits a stepped upward trend. Specifically, when the vehicle is in a driving state and consuming energy, the cumulative energy consumption gradually increases over time, represented by rising steps on the curve; when the vehicle is in a parking or charging state, the cumulative energy consumption remains constant, represented by horizontal steps on the curve. This accurately reflects the physical process of intermittent energy consumption during actual vehicle operation.

[0034] In one embodiment of the present invention, an initial battery state of charge (SBC) is set when the vehicle begins executing a set of simulated travel tasks. Moving along the time axis, the energy consumption at corresponding moments on the cumulative energy consumption curve is continuously subtracted from the initial SBC to calculate the remaining simulated battery capacity in real time. When the remaining simulated battery capacity first falls below a preset vehicle battery capacity threshold, the current moment is recorded as the charging start time triggering the charging demand. Based on the vehicle battery's total capacity, the current remaining simulated battery capacity, and the preset target SBC, the required energy replenishment from the trigger time is calculated, and this replenishment is taken as the required charging amount. Recording the charging start time and the required charging amount constitutes a simulated charging demand trigger point. If the remaining simulated battery capacity falls below the threshold again before the end of the predetermined prediction period, the above recording process is repeated to determine the next simulated charging demand trigger point.

[0035] In practice, the prerequisite is that the energy consumption calculation for each simulated travel task set has been completed, and the cumulative energy consumption curve for each autonomous vehicle within the predetermined prediction period has been generated. The initial battery state of charge when the vehicle begins executing the simulated travel task set is set to 80%, and the preset threshold for remaining battery power is 20%. The cumulative energy consumption curve records the vehicle's cumulative energy consumption from the start time, with a time resolution of minutes. In practice, the system progresses along the time axis, continuously subtracting the energy consumption at the corresponding moment on the cumulative energy consumption curve from the initial battery state of charge, and calculates the simulated battery power in real time. The change in simulated battery power over time is shown in Table 1. Table 1: Changes in Remaining Battery Capacity over Time In practice, when the simulated battery's remaining charge first falls below a preset vehicle battery remaining charge threshold, the system records the current moment as the charging start time triggered by the charging demand. According to Table 1, at 280 minutes, the simulated battery's remaining charge drops to 21.5%, falling below the 20% threshold for the first time; therefore, the charging start time is recorded as the 280th minute. Based on the vehicle battery's total capacity, the current simulated battery's remaining charge, and the preset target state of charge, the required energy replenishment from the trigger moment is calculated, and this energy replenishment is taken as the required charging amount. The formula for calculating the required charging amount is: in: This represents the amount of charge required. Represents the total capacity of the vehicle's battery. This represents the preset target state of charge. This represents the simulated remaining battery charge at the trigger point. Assuming the vehicle battery has a total capacity of 100kWh, a target state of charge of 80%, and a simulated remaining battery charge of 21.5% at the trigger point, the required charge is 100 multiplied by (80% minus 21.5%), resulting in 58.5kWh. Recording the charging start time and the required charge together constitutes a simulated charging demand trigger point.

[0036] Understandably, if the simulated battery charge falls below the vehicle battery charge threshold again before the end of the predetermined prediction period, the process returns to the step of recording the current time as the charging start time to determine the next simulated charging demand trigger point. Assuming the vehicle's charge continues to decrease during the subsequent journey and the remaining charge drops to 18% again after 500 minutes, the charging start time for the second charging demand trigger is recorded as the 500th minute. Based on the current remaining charge and the target state, the new required charging amount is calculated, generating the second simulated charging demand trigger point.

[0037] In one embodiment of the present invention, the physical location, charging power specifications, and current occupancy status information of all charging facilities within the operating area of ​​the autonomous vehicle are obtained. For each simulated charging demand trigger point, based on its charging start time and the vehicle's current location, a set of accessible candidate charging facilities is selected in a spatiotemporal dimension. The core of the spatiotemporal selection is to set a maximum allowable detour time threshold, and only facilities whose estimated travel time from the trigger location to the charging facility is less than this threshold are included in the candidate set. Based on the required charging amount and the charging power specifications of the candidate facilities, the occupancy time required to complete charging at each facility is estimated, which is directly calculated by dividing the required charging amount by the rated charging power of the candidate charging facility. A charging facility selection optimization model is constructed, with binary decision variables representing whether a charging demand trigger point is allocated to a specific charging time window of a charging facility. This model is a system optimization and resource scheduling model based on the mathematical framework of Mixed Integer Linear Programming (MILP). It treats the charging demands generated by all vehicles in a region as a series of tasks that need to be scheduled on limited charging facility resources. The core is to construct a mixed integer linear programming problem, formalizing the "task-resource-time" allocation relationship by defining binary decision variables. The optimization objective is to minimize the total system time cost, strictly adhering to three basic constraints: each demand must be satisfied, charging facility capacity must not be exceeded, and charging start time must be reasonable. This allows for the solution of the optimal charging facility and charging time slot allocation scheme. In other words, by establishing a mixed integer linear programming problem, under the constraints of strictly satisfying each charging demand, each charging pile serving at most one vehicle at a time, and charging start not earlier than the demand generation time, a charging facility and time window allocation scheme that minimizes the total additional travel and waiting costs for all vehicles is found. The objective function is a weighted sum of the time cost of traveling to the facility, the expected queuing time cost, and the additional travel time cost due to deviation from the original path, aiming to minimize the total cost. Constraints include: each trigger point must be assigned to one and only one charging time window; the total number of interfaces required for charging tasks assigned to any facility within any time window does not exceed its available interface count; and the charging start time is no earlier than the demand trigger time. A mixed-integer linear programming solver is used to solve the model to obtain the optimal allocation scheme. The solution process involves inputting the defined decision variables, objective function, and all constraints into the mixed-integer linear programming solver. The solver automatically executes its internal mathematical programming algorithm, searching for and outputting the decision variable assignments that minimize the objective function value while satisfying all constraints; this is the optimal solution. Based on this solution, the specific charging facility and charging time window assigned to each simulated charging demand trigger point can be determined. The power demand generated by all charging tasks in the allocated time periods and facilities is summarized to form a spatiotemporal distribution prediction of charging load with time as the horizontal axis and spatial location as the vertical axis.

[0038] In practical implementation, this embodiment assumes that all simulated charging demand trigger points have been obtained, including the charging start time and required charging amount for each autonomous vehicle. The physical location information, charging power specifications, and real-time updated current occupancy status information of all charging facilities within the autonomous vehicle's operating area are obtained. Assuming there are three charging stations within the operating area, their specific attributes are shown in Table 2. Table 2: Charging Facility Attribute Table For each simulated charging demand trigger point, a set of accessible candidate charging facilities is selected in the spatiotemporal dimension based on the charging start time and the vehicle's current location. For example, a vehicle located at coordinates (x_v, y_v) triggers a charging demand at 14:00, requiring 58.5 kWh of charging. Setting a maximum allowable detour time of 15 minutes, the travel time to CS001, CS002, and CS003 is calculated. If all three charging stations are within 15 minutes, they are included in the candidate set. Based on the required charging amount and the charging power specifications of the candidate charging facilities, the estimated time required to complete charging is calculated. For CS001, the time required is equal to the required charging amount divided by the charging power specification, i.e., 58.5 divided by 120, approximately 0.4875 hours, or about 29.25 minutes.

[0039] In practical implementation, an optimization model for selecting charging facilities is constructed to solve the allocation problem of multiple vehicles and multiple charging stations. Decision variables are defined. Let be a binary variable, representing whether the simulated charging demand trigger point p is allocated to a specific charging time window w of the charging facility f. Construct the objective function. ,in This represents the total time cost of driving to a charging facility. This represents the expected waiting time cost at charging facilities. The Z-value represents the additional travel time cost incurred due to deviation from the original planned path, and the objective is to minimize this value. Constraints include: each simulated charging demand trigger point must be assigned to one and only one charging time window; for any charging facility and any charging time window, the total number of charging interfaces required for a charging task assigned to that window must not exceed the number of available charging interfaces at that facility within the window; and the start time of a charging task must not be earlier than the charging demand trigger time. A mixed-integer linear programming solver is used to solve the model, outputting the optimal allocation scheme.

[0040] It is understandable that the charging power demand generated by all charging tasks within their allocated charging time periods and on their corresponding charging facilities is aggregated to form a spatiotemporal distribution prediction of charging load. Assuming that during the time period from 15:00 to 15:30, CS001 is assigned 2 charging tasks, CS002 is assigned 3 charging tasks, and CS003 is assigned 1 charging task, then the predicted charging load for CS001 during this time period is 240kW, for CS002 it is 540kW, and for CS003 it is 150kW. Mapping these data onto a spatial grid and a time axis generates a two-dimensional matrix-style spatiotemporal distribution prediction of charging load.

[0041] In one embodiment of the present invention, the system receives real-time reports from a fleet of autonomous vehicles regarding their geographical location, current battery state of charge, and ongoing task information. The reported location and battery level information are compared with the predicted states of the corresponding vehicles in the simulated travel task set, and location and battery level deviations are calculated. When either deviation exceeds a preset update threshold, a dynamic prediction update process is initiated. The real-time reported vehicle geographical location is used as a new starting point, the current battery state of charge is used as a new initial state, and subsequent travel task assumptions are adjusted based on the currently executing task information. The entire process, from generating the simulated travel task set for the remaining prediction period to energy consumption calculation, determining charging demand trigger points, and generating updated spatiotemporal distribution predictions of charging load, is re-executed.

[0042] In practice, the prediction dynamic update step establishes a communication connection between a prediction server deployed in the cloud and the autonomous vehicle fleet to receive status messages uploaded by the vehicles in real time. These status messages include the vehicle ID, real-time geographic coordinates, current battery state of charge percentage, and the order number of the currently executing task. The prediction server maintains prediction status records for the corresponding vehicles in the simulated travel task set, including predicted location coordinates and predicted battery state of charge. In this specific implementation, a real-time message is received from vehicle ID V101, with real-time geographic coordinates (3150, 4210), a current battery state of charge of 45%, and the currently executing task being order O789. The prediction status of vehicle V101 is retrieved from the simulated travel task set, showing predicted location coordinates (3120, 4180) and a predicted battery state of charge of 52%.

[0043] In practice, the Euclidean distance formula and absolute value subtraction are used to calculate the position deviation and electrical deviation. The formula for calculating the position deviation is: in: Represents positional deviation. and Represents real-time geographic location coordinates. and This represents the predicted position coordinates. Substituting the numerical values, the position deviation equals... The result is approximately 42.43 meters. The charge deviation equals the real-time battery state of charge minus the absolute value of the predicted battery state of charge, i.e. The preset update threshold is set to a position deviation of 100 meters or a battery level deviation of 10%. Since the position deviation of 42.43 meters is less than 100 meters and the battery level deviation of 7% is less than 10%, the trigger condition was not met. It can be understood that if the real-time battery state of charge reported by vehicle V101 is 38%, then the battery level deviation becomes... The update threshold exceeds 10%.

[0044] When the location deviation or power deviation exceeds a preset update threshold, the system triggers the prediction update process. The real-time reported vehicle geographic location information is used as the new starting point, replacing the original prediction starting point coordinates. The current battery state of charge information is used as the new initial battery state, replacing the initial values ​​in the original simulated travel task set. Subsequent travel task assumptions are updated based on the currently executing task information. If the vehicle reports a task change, unexecuted simulated tasks in the original set are deleted, and new tasks are planned based on the new destination. In some embodiments, if the vehicle completes its task early and enters an empty state, a cruise task generation logic is added to subsequent simulations. The process re-executes, starting from generating the simulated travel task set for the vehicle within the remaining prediction period using an improved vehicle trajectory simulation algorithm, sequentially performing task-by-task energy consumption calculations, generating cumulative energy consumption curves, determining simulated charging demand trigger points, until an updated spatiotemporal distribution prediction of the charging load is generated. Optionally, the update operation of the prediction server adopts a transaction mechanism, locking the vehicle's state records during the update to prevent conflicts between old and new data. After the update is completed, the lock is released, and the scheduling system is notified to load the new spatiotemporal distribution prediction data of the charging load. In some embodiments, to prevent system oscillations caused by frequent updates, a minimum update interval of 5 minutes is set, meaning that even if the deviation exceeds the limit again within 5 minutes after an update is triggered, no action is taken.

[0045] The flowchart provided in this embodiment is not intended to indicate that the operations of the method will be performed in any particular order, or that all operations of the method are included in every case. Furthermore, the method may include additional operations. Within the scope of the technical concept provided by the method in this embodiment, additional variations can be made to the above method.

[0046] It should be understood that in some embodiments, the components may be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods may be implemented using software or firmware stored in memory and executed by a suitable instruction execution system.

[0047] Therefore, those skilled in the art should recognize that although numerous exemplary embodiments of the present invention have been shown and described in detail herein, many other variations or modifications conforming to the principles of the present invention can be directly determined or derived from the disclosure of the present invention without departing from the spirit and scope of the invention. Thus, the scope of the present invention should be understood and construed as covering all such other variations or modifications.

Claims

1. A method for predicting the charging load of autonomous vehicles, characterized in that, include: Acquire traffic network topology, historical trip order data, and vehicle energy consumption characteristics parameters of the operating area of ​​autonomous vehicles; Based on the traffic network topology and the historical travel order data, an improved vehicle trajectory simulation algorithm is used to generate a set of simulated travel tasks for vehicles within a predetermined prediction period. The improved vehicle trajectory simulation algorithm dynamically adjusts the static weight random path selection algorithm according to the current traffic conditions. The energy consumption of each simulated travel task is calculated, and the cumulative energy consumption curve of each autonomous vehicle in the predetermined prediction period is generated by combining the vehicle energy consumption characteristic parameters. Based on the cumulative energy consumption curve and the preset vehicle battery remaining power threshold, the simulated charging demand trigger point for each autonomous vehicle is determined. The simulated charging demand trigger point includes the charging start time and the required charging amount. By integrating the simulated charging demand trigger points of all autonomous vehicles, and based on the distribution and capacity constraints of charging facilities, a spatiotemporal distribution prediction of charging load in the autonomous vehicle operating area within the predetermined prediction period is generated.

2. The charging load prediction method for an unmanned vehicle according to claim 1, characterized in that, The method of generating a set of simulated travel tasks for a vehicle within a predetermined prediction period using an improved vehicle trajectory simulation algorithm includes: Based on the historical travel order data, the probability distribution matrix of the origin and destination points and the probability distribution function of the travel distance are extracted for different date types and time periods; Based on the real-time traffic information of the traffic network topology, a dynamic travel time weight is assigned to each road segment in the road network. The dynamic travel time weight is calculated based on the historical average speed of the road segment and the traffic flow at the current moment. Using the improved vehicle trajectory simulation algorithm, a series of consecutive trip origin-destination pairs are randomly generated for each simulated vehicle within the predetermined prediction period, based on the trip origin-destination probability distribution matrix. For each of the trip origin-destination pairs, the expected trip distance is determined according to the trip distance probability distribution function, and a simulated travel path is planned in the traffic network topology based on the dynamic trip time weight and the shortest time path search strategy. The simulated travel routes, corresponding origin and destination points, and departure times of all simulated vehicles are combined to form the simulated travel task set.

3. The charging load prediction method for an unmanned vehicle according to claim 2, characterized in that, The improved vehicle trajectory simulation algorithm dynamically adjusts the static weighted random path selection algorithm based on the current traffic conditions, including: In the static weighted random path selection algorithm, the probability of a vehicle choosing the next path segment at a path intersection depends only on the static length of the path or the free-flow time. The improved vehicle trajectory simulation algorithm introduces the dynamic travel time weight as a core decision variable; When a simulated vehicle arrives at a path intersection, the algorithm obtains the real-time dynamic travel time weights from the path intersection to each downstream road segment. Calculate the reciprocal of the dynamic travel time required to reach each downstream road segment, and normalize the reciprocal to obtain the real-time probability of the vehicle choosing each downstream road segment; Based on the real-time probabilities, the roulette wheel selection method is used to determine the next road segment that the simulated vehicle will actually travel on. This dynamic correction process enables the simulated vehicle's route selection to respond in real time to changes in traffic congestion, thereby generating the simulated driving path that more closely resembles the actual traffic flow.

4. The charging load prediction method for an unmanned vehicle according to claim 3, characterized in that, For the simulated travel task set, a task-by-task energy consumption calculation is performed. Combined with the vehicle energy consumption characteristic parameters, a cumulative energy consumption curve for each autonomous vehicle within the predetermined prediction period is generated, including: The energy consumption benchmark value per unit mileage, the air conditioning energy consumption coefficient, and the correlation function between vehicle speed and energy consumption are obtained from the vehicle energy consumption characteristic parameters. For each simulated travel task in the set of simulated travel tasks, extract the total length of the simulated travel path, the predicted average speed of the road segments traversed, and the predicted ambient temperature during task execution. The basic driving energy consumption of the simulated travel task is calculated based on the total length and the energy consumption benchmark value per unit mileage. Based on the predicted average vehicle speed, the correlation function between vehicle speed and energy consumption is queried to obtain the vehicle speed correction factor, which is then used to correct the basic driving energy consumption. Based on the deviation between the predicted ambient temperature and the comfort temperature threshold, and in conjunction with the air conditioning energy consumption coefficient, the additional energy consumption for temperature control in the simulated travel task is calculated. The estimated energy consumption of the simulated travel task is obtained by summing the corrected basic driving energy consumption of the simulated travel task with the additional energy consumption of temperature control. The estimated power consumption of an autonomous vehicle for all simulated travel missions is accumulated in chronological order to generate the cumulative power consumption curve of the autonomous vehicle over time within the predetermined prediction period.

5. The charging load prediction method for an unmanned vehicle according to claim 4, characterized in that, Based on the cumulative energy consumption curve and the preset vehicle battery remaining power threshold, the simulated charging demand trigger point for each autonomous vehicle is determined, including: Set the initial battery state of charge when the vehicle begins to execute the set of simulated travel tasks; Moving along the time axis, the energy consumption at the corresponding moment on the cumulative energy consumption curve is continuously subtracted from the initial battery state of charge to calculate the remaining battery power in real time. When the remaining charge of the simulated battery first falls below the preset vehicle battery remaining charge threshold, the charging start time triggered by the current charging demand is recorded. Based on the total capacity of the vehicle battery, the current remaining charge of the simulated battery, and the preset target state of charge, the required amount of electrical energy replenishment from the triggering time is calculated, and the amount of electrical energy replenishment is used as the required charging amount. The charging start time and the required charging amount are recorded together to form a simulated charging demand trigger point; If the remaining charge of the simulated battery falls below the vehicle battery's remaining charge threshold again before the end of the predetermined prediction period, then the process returns to the step of recording the current time as the charging start time to determine the next simulated charging demand trigger point.

6. The charging load prediction method for an unmanned vehicle according to claim 5, characterized in that, The step of generating a spatiotemporal distribution prediction of charging load in the autonomous vehicle's operating area within the predetermined prediction period, based on the distribution and capacity constraints of charging facilities, includes: Obtain the physical location information, charging power specification information, and current occupancy status information of all charging facilities within the operating area of ​​the autonomous vehicle; For each of the simulated charging demand trigger points, based on its charging start time and the vehicle's current location, a set of accessible candidate charging facilities is searched in the spatial and temporal dimensions. Based on the required charging amount and the charging power specifications of the candidate charging facilities, the estimated time required to complete charging at each of the candidate charging facilities is determined. A charging facility selection optimization model is constructed with the goal of minimizing total waiting time and travel cost. The charging facility selection optimization model must meet the charging facility capacity constraint, that is, the number of charging tasks allocated to the same charging facility in the same time period shall not exceed the number of its idle charging interfaces. Solve the charging facility selection optimization model to assign a specific charging facility and an expected charging time period to each simulated charging demand trigger point; The charging power demand generated by all charging tasks within their respective allocated charging time periods and on the corresponding charging facilities is summarized to form a spatiotemporal distribution prediction of the charging load with time as the horizontal axis and spatial location as the vertical axis.

7. The charging load prediction method for an unmanned vehicle according to claim 6, characterized in that, The construction of the charging facility selection optimization model, which aims to minimize total waiting time and travel cost, includes: Define a decision variable, which is a binary variable, representing whether a certain simulated charging demand trigger point is allocated to a certain charging time window of a certain charging facility; Construct an objective function, which is a weighted sum of the total time cost of traveling to the charging facility, the expected queuing time cost at the charging facility, and the additional travel time cost caused by deviating from the original planned route due to charging. The objective is to minimize the value of the objective function. The constraints include: each simulated charging demand trigger point must be assigned to one and only one charging time window; for any charging time window of any charging facility, the total number of charging interfaces required for the charging task assigned to the charging window shall not exceed the number of available charging interfaces of the charging facility within the charging window; the start time of the charging task shall not be earlier than the charging demand trigger time. The optimization model is solved using a mixed-integer linear programming solver to obtain the optimal solution for the decision variables.

8. The charging load prediction method for an unmanned vehicle according to claim 7, characterized in that, Based on real-time traffic information of the aforementioned traffic network topology, a dynamic travel time weight is assigned to each road segment in the network, including: The system obtains historical average speed data of all road segments in the traffic network topology within a predetermined prediction period from the real-time traffic information platform, as well as traffic flow data of each road segment at the current moment collected in real time by sensors or vehicle communication devices deployed on the roadside. For each road segment in the road network, the historical average speed data is used as the benchmark free-flow speed for that road segment, and a benchmark traffic capacity is set for the road segment according to the road grade and design standards. At each dynamically updated simulation moment, based on the collected traffic flow data of each road segment at the current moment, the ratio of the current traffic flow of the processed road segment to its baseline traffic capacity is calculated, i.e., the current saturation. Based on the current saturation, query the preset saturation-velocity reduction relationship table to obtain the velocity reduction coefficient corresponding to the current saturation; Multiply the baseline free-flow velocity by the velocity reduction factor to obtain the estimated traffic speed of the processed road segment at the current simulation time; Based on the static length and estimated traffic speed of the processed road segment, the dynamic travel time required for vehicles to pass through the processed road segment is calculated, and this dynamic travel time is assigned as a dynamic travel time weight to the processed road segment. The calculated dynamic travel time weights for all road segments are updated in the corresponding road segment attributes of the traffic network topology, so that the improved vehicle trajectory simulation algorithm can call them during path planning.

9. The charging load prediction method for an unmanned vehicle according to claim 8, characterized in that, Based on the deviation between the predicted ambient temperature and the comfort temperature threshold, and in conjunction with the air conditioning energy consumption coefficient, the additional energy consumption for temperature control in the simulated travel task is calculated, including: Obtain a preset vehicle cabin comfort temperature threshold, which includes a lower temperature limit and a higher temperature limit; Read the predicted ambient temperature during the simulated travel mission execution period; Determine whether the predicted ambient temperature is lower than the lower limit of the comfort temperature threshold, or higher than the upper limit of the temperature threshold; If the predicted ambient temperature is lower than the lower limit of the temperature, then the first difference between the lower limit of the temperature and the predicted ambient temperature is calculated, and the first difference is multiplied by the preset heating condition air conditioning energy consumption coefficient to obtain the heating additional energy consumption, and the heating additional energy consumption is used as the temperature control additional energy consumption. If the predicted ambient temperature is higher than the upper limit of temperature, then calculate the second difference between the predicted ambient temperature and the upper limit of temperature, multiply the second difference by the preset cooling mode air conditioning energy consumption coefficient to obtain the additional cooling energy consumption, and use the additional cooling energy consumption as the additional temperature control energy consumption. If the predicted ambient temperature is within the comfort temperature threshold range, then the additional energy consumption for temperature control is set to zero.

10. The charging load prediction method for an unmanned vehicle according to claim 9, characterized in that, The method also includes a predictive dynamic update step based on real-time vehicle status information: Receive real-time reports from the autonomous vehicle fleet on vehicle location, current battery charge status, and current task execution. The real-time reported vehicle location information and current battery charge status information are compared with the predicted status of the corresponding vehicle in the simulated travel task set to calculate the location deviation and battery charge deviation. When the position deviation or power deviation exceeds the preset update threshold, the prediction update process is triggered; The real-time reported vehicle geographic location information is used as a new starting point, the current battery state of charge information is used as a new initial battery state, and the subsequent travel task assumptions are updated based on the current task execution information. The process is repeated from generating a set of simulated travel tasks for the vehicle within the remaining prediction period using an improved vehicle trajectory simulation algorithm until an updated prediction of the spatiotemporal distribution of charging load is generated.

Images