A vehicle-pile road network charging load prediction method, system, device and medium

Abstract

The application discloses a kind of vehicle pile road network charging load prediction method, system, equipment and medium, belong to load prediction technical field, comprising: the sparse state tensor of vehicle is obtained.According to sparse state tensor and the region contribution function of preestablished, construct charging intention distribution diagram, wherein, region contribution function simulates each vehicle to the charging intention density contribution of each position in space electric pile of each position.Electric pile is extracted from charging intention distribution diagram Space attraction degree, and space attraction degree is substituted into the access utility function of preestablished, to obtain access utility value.According to access utility value, normalization is handled, to obtain the probability distribution diagram of vehicle access electric pile.Probability distribution diagram indicates that each vehicle selects the behavior probability of any electric pile in future time period.According to probability distribution diagram and preestablished average charging power, the actual predicted load data of each electric pile is calculated, improve the accuracy of electric pile load prediction.

Description

Technical Field

[0001] This invention relates to the field of load forecasting technology, specifically to a method, system, equipment, and medium for forecasting the charging load of vehicle-charging infrastructure networks. Background Technology

[0002] The charging behavior of new energy vehicles exhibits a significant imbalance in terms of time and space distribution. In certain areas, charging demand is highly concentrated at certain times, resulting in queuing and overload of charging stations, while in other areas, charging stations remain idle for extended periods. This supply-demand mismatch not only affects the charging experience of car owners but also places higher demands on the load balancing of the power grid.

[0003] Existing charging load forecasting methods mainly rely on historical charging pile usage data for prediction, which has obvious limitations: it is difficult to capture dynamic factors in vehicle charging decisions, such as real-time power status, route planning, and differences in charging pile power, and it is prone to prediction distortion when dealing with hotspot areas, making it difficult to provide effective support for power supply scheduling. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention provides a method, system, equipment and medium for predicting the charging load of vehicle-charging network.

[0005] Therefore, the technical problem solved by this invention is: how to achieve accurate estimation from individual charging demand to system-level charging pile load by integrating vehicle status data, charging pile status data and path travel distance through modeling, while suppressing the problem of hot spot charging pile prediction distortion.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for predicting the charging load of a vehicle-charging network, comprising,

[0007] Obtain the sparse state tensor of the vehicle;

[0008] A charging intention distribution map is constructed based on the sparse state tensor and a preset regional contribution function; wherein, the regional contribution function simulates the charging intention density contribution of each vehicle to a charging pile at each location in space.

[0009] The spatial attractiveness of the charging piles is extracted from the charging intention distribution map, and the spatial attractiveness is substituted into a preset access utility function to obtain the access utility value;

[0010] The access utility value is normalized to obtain a probability distribution map of vehicle access to charging stations; wherein, the probability distribution map represents the probability of each vehicle selecting any charging station in a future time period.

[0011] The actual predicted load data for each charging pile is calculated based on the probability distribution diagram and the preset average charging power.

[0012] The sparse state tensor is a three-dimensional tensor composed of vehicle dimension, charging pile dimension and feature dimension, which respectively correspond to vehicle index, charging pile index and joint state feature vector dimension of each vehicle-charging pile pair. Each element of the sparse state tensor is the joint state feature vector between the corresponding vehicle and the corresponding charging pile. The joint state feature vector includes path travel distance, vehicle current battery level, charging pile current state and charging pile maximum charging power. The elements between vehicles and charging piles that are not accessible by path are empty.

[0013] The charging intention distribution map is a continuous distribution of charging intention density at each geographical location, with geographical location in the urban space as the independent variable.

[0014] The spatial attraction is the value of the charging intention distribution map at the geographical location of the charging pile;

[0015] The access utility function is used to calculate the access utility value of a vehicle accessing a charging station, using the spatial attractiveness, vehicle individual status, charging station attributes, traffic conditions, and behavioral preferences as variables.

[0016] As a preferred embodiment of the vehicle-charging network charging load prediction method of the present invention, after calculating the actual predicted load data of each charging pile based on the probability distribution map and the preset average charging power, the method further includes:

[0017] Obtain the charging pile power suppression factor, the regional power supply suppression factor, and the access behavior density regularization term;

[0018] The minimum value among the following is selected: the product of the charging pile power suppression factor and the preset maximum charging power of the charging pile, the product of the regional power supply suppression factor and the preset maximum regional power supply capacity, and the difference between the actual predicted load data and the access behavior density regularization term.

[0019] The largest value between zero and the minimum value is selected as the corrected predicted load data for the charging pile.

[0020] As a preferred embodiment of the vehicle-charging network charging load prediction method of the present invention, wherein: obtaining the charging pile power suppression factor, the regional power supply suppression factor, and the access behavior density regularization term includes:

[0021] Calculate the first ratio of the actual predicted load data to the sum of the maximum charging power of the charging pile and a small positive constant, calculate the first difference between 1 and the first ratio, and perform non-negative truncation on the first difference to obtain the charging pile power suppression factor.

[0022] The total actual predicted load data is obtained by summing all the actual predicted load data in the region.

[0023] Calculate the second ratio of the total actual predicted load data to the sum of the maximum power supply capacity of the region and the small positive constant, calculate the second difference between 1 and the second ratio, and perform non-negative truncation on the second difference to obtain the power supply suppression factor of the region.

[0024] The access behavior density regularization term is constructed based on the probability distribution diagram and the preset adjustment coefficient;

[0025] The non-negative truncation process is as follows: when the first difference or the second difference is less than 0, the result is 0; when the first difference or the second difference is not less than 0, the result is retained.

[0026] As a preferred embodiment of the vehicle-charging-road network charging load prediction method of the present invention, wherein: obtaining the sparse state tensor of the vehicle includes:

[0027] Acquire vehicle status data, charging station status data, and route travel distance;

[0028] The sparse state tensor is constructed based on the vehicle status data, the charging pile status data, and the path travel distance.

[0029] As a preferred embodiment of the vehicle-pile road network charging load prediction method of the present invention, the sparse state tensor includes the path travel distance, the current vehicle power, the current state of the charging pile, and the maximum charging power of the charging pile.

[0030] The step of constructing a charging intention distribution map based on the sparse state tensor and a preset region contribution function includes:

[0031] The vehicle's intention weight for the charging station is constructed based on the vehicle's current battery level, the charging station's current status, the desired charging amount, the charging station's maximum charging power, the path travel distance, and a preset Boolean flag.

[0032] The influence of vehicle intent on the spatial neighborhood is simulated based on a preset spatial diffusion kernel function;

[0033] The region contribution function is constructed based on the intention weight and the spatial diffusion kernel function, and the sparse state tensor is substituted into the region contribution function to construct the charging intention distribution map.

[0034] The spatial diffusion kernel function takes the distance between the location in space and the charging pile as the independent variable, and the value of the spatial diffusion kernel function characterizes the diffusion intensity of the vehicle's intention to charge the area where the charging pile is located at the location.

[0035] As a preferred embodiment of the vehicle-charging-pile road network charging load prediction method of the present invention, the step of constructing the vehicle's intention weight for the charging pile based on the vehicle's current battery level, the charging pile's current status, the desired charging capacity, the charging pile's maximum charging power, the path travel distance, and a preset Boolean flag includes:

[0036] The vehicle's current battery level is used to calculate the vehicle's current battery consumption.

[0037] The product of the preset adjustment cost impact intensity parameter and the path travel distance is defined as the first target product;

[0038] The target ratio is obtained by taking the ratio of the sum of the product of the maximum charging power of the charging pile and the preset value and the first target value.

[0039] The product of the preset weight coefficient and the Boolean flag is defined as the second target product;

[0040] The intention weight is obtained by weighting and summing the product of the vehicle's consumed electricity, the current status of the charging pile, the expected charging amount, the target ratio, and the second target.

[0041] As a preferred embodiment of the vehicle-charging-road network charging load prediction method of the present invention, the construction of the access utility function includes:

[0042] To obtain the degree of compatibility between the charging station's power and the vehicle's power requirements, and the deviation length between the vehicle's navigation path and the path to the charging station;

[0043] The access utility function is constructed based on the spatial attraction, the current state of the charging pile, the suitability, the path travel distance, the deviation length, and a preset local load suppression factor; wherein, the preset local load suppression factor is obtained by performing a local average convolution in the area near each charging pile based on the charging intention distribution map.

[0044] This invention provides a vehicle-charging-road network charging load prediction system.

[0045] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a vehicle-charging-road network charging load prediction system, comprising:

[0046] The acquisition unit is used to acquire the sparse state tensor of the vehicle.

[0047] A construction unit is used to construct a charging intention distribution map based on the sparse state tensor and a preset regional contribution function; wherein, the regional contribution function simulates the charging intention density contribution of each vehicle to a charging pile at each location in space.

[0048] An extraction unit is used to extract the spatial attractiveness of charging piles from the charging intention distribution map, and substitute the spatial attractiveness into a preset access utility function to obtain an access utility value;

[0049] The processing unit is used to perform normalization processing based on the access utility value to obtain a probability distribution map of vehicle access to charging piles; wherein, the probability distribution map represents the probability of each vehicle selecting any charging pile in a future time period.

[0050] The prediction unit is used to calculate the actual predicted load data of each charging pile based on the probability distribution map and the preset average charging power.

[0051] The sparse state tensor is a three-dimensional tensor composed of vehicle dimension, charging pile dimension and feature dimension, which respectively correspond to vehicle index, charging pile index and joint state feature vector dimension of each vehicle-charging pile pair. Each element of the sparse state tensor is the joint state feature vector between the corresponding vehicle and the corresponding charging pile. The joint state feature vector includes path travel distance, vehicle current battery level, charging pile current state and charging pile maximum charging power. The elements between vehicles and charging piles that are not accessible by path are empty.

[0052] The charging intention distribution map is a continuous distribution of charging intention density at each geographical location, with geographical location in the urban space as the independent variable.

[0053] The spatial attraction is the value of the charging intention distribution map at the geographical location of the charging pile;

[0054] The access utility function is used to calculate the access utility value of a vehicle accessing a charging station, using the spatial attractiveness, vehicle individual status, charging station attributes, traffic conditions, and behavioral preferences as variables.

[0055] The present invention provides an electronic device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the vehicle-charging network charging load prediction method.

[0056] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the vehicle-charging network charging load prediction method.

[0057] The beneficial effects of this invention are as follows: This invention constructs a sparse state tensor to fuse multi-source data such as vehicle battery power, charging pile power and path distance, and combines it with a spatial diffusion kernel function to establish a continuous distribution model of charging intention, thereby realizing the spatial expression of charging demand in urban areas.

[0058] In the probabilistic modeling of vehicle access to charging stations, spatial attraction, path deviation penalty, and local load suppression adjustment mechanisms are introduced to effectively improve the realism of behavior prediction and avoid excessive concentration of hotspot charging stations in the prediction. In the load calculation stage, the synergistic constraints of charging station power suppression factor, regional power supply suppression factor, and access behavior density regularization term ensure that the prediction results conform to the physical carrying capacity of charging stations and the power supply capacity limitations of the power grid, thus solving the problem of prediction distortion in hotspot areas in traditional methods. Attached Figure Description

[0059] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 The above is a flowchart of a method for predicting the charging load of a vehicle-to-charging-road network, provided as an embodiment of the present invention.

[0061] Figure 2 This is a flowchart illustrating the acquisition of sparse state tensors of vehicles in a vehicle-pile-road network charging load prediction method according to an embodiment of the present invention.

[0062] Figure 3 This is a flowchart illustrating the construction of a charging intention distribution map for a vehicle-charging network charging load prediction method according to an embodiment of the present invention.

[0063] Figure 4 The present invention provides a load correction flowchart for a vehicle-to-charging-network charging load prediction method according to an embodiment of the present invention.

[0064] Figure 5 This is a schematic diagram of a vehicle-charging network charging load prediction system provided in one embodiment of the present invention.

[0065] In the diagram: 501, acquisition unit; 502, construction unit; 503, extraction unit; 504, processing unit; 505, prediction unit. Detailed Implementation

[0066] To make the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a 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 effort should fall within the protection scope of the present invention.

[0067] Example 1, referring to Figures 1-4This is one embodiment of the present invention, which provides a method for predicting the charging load of a vehicle-charging network, including:

[0068] Step S101: Obtain the sparse state tensor of the vehicle.

[0069] Step S102: Construct a charging intention distribution map based on the sparse state tensor and the preset regional contribution function; wherein, the regional contribution function simulates the charging intention density contribution of each vehicle to the charging pile at each location in space.

[0070] Step S103: Extract the spatial attraction of charging piles from the charging intention distribution map, and substitute the spatial attraction into the preset access utility function to obtain the access utility value.

[0071] Step S104: Normalize the access utility value to obtain the probability distribution map of vehicle access to charging pile; wherein, the probability distribution map represents the probability of each vehicle selecting any charging pile in the future time period.

[0072] Step S105: Calculate the actual predicted load data for each charging pile based on the probability distribution map and the preset average charging power.

[0073] like Figure 2 As shown, in some embodiments, step S101 includes steps S201 to S202:

[0074] Step S201: Obtain vehicle status data, charging pile status data, and route travel distance.

[0075] Step S202: Construct a sparse state tensor based on vehicle status data, charging pile status data, and path travel distance.

[0076] It should be noted that in step S201, the vehicle status data includes the data for each vehicle at time [time value missing]. Geographical location Its power level (State of Charge) and vehicle model This data is typically uploaded to the dispatch platform in real time via the vehicle's BMS system.

[0077] For example, taking a ride-hailing platform as an example, every 5 seconds, a vehicle uploads JSON-formatted data to the server containing fields such as location, battery level, and motor status. The battery level is displayed as a percentage (e.g., 0.43 represents 43% battery). The vehicle model information comes from vehicle registration data or VIN decoding, which is used for subsequent charging parameter mapping.

[0078] It should be noted that the status data of the charging piles includes the geographical coordinates of each charging pile. Is it available? (e.g., 1 for idle, 0 for occupied or faulty), and the maximum supported charging power of the charging station. This data is obtained through the charging pile management platform API. For example, a city-level platform provides a REST interface that supports querying charging pile information and operating status by region. The returned fields include location, voltage level, availability, and type. The power value is a fixed field, and the unit is numeric (e.g., 7 represents a 7 kilowatt power level).

[0079] Furthermore, the path travel distance is for vehicles. To the charging station The estimated travel distance is typically obtained dynamically through a route planning interface. For example, it can be obtained by calling the Amap WebAPI and inputting the starting point coordinates. and endpoint coordinates Output the path travel distance. Considering system deployment latency and efficiency for large-scale calls, it is recommended to build a local path query cache service to store the path information of common vehicle-charging station pairs in a time-segmented structure. Vehicle status data, charging station status data, and path travel distance all need to be normalized.

[0080] Understandably, in step S202, a three-dimensional sparse state tensor is constructed. , indicating vehicle With electric pile The joint state features between them. The three dimensions represent the vehicle dimension, charging pile dimension, and feature dimension, respectively, corresponding to the vehicle index, charging pile index, and the joint state feature vector dimension of each vehicle-charging pile pair.

[0081] It should be noted that each element of the tensor is a four-dimensional vector, as shown below:

[0082] ;

[0083] in, Indicates from vehicle To the charging station Path travel distance It is a vehicle The current battery percentage is uploaded by the BMS system. It is an electric pole. The current idle status is obtained by querying the operation platform, and the value is 0 or 1. It is an electric pole. The maximum supported charging power is derived from the charging station registration parameters and has been mapped to a standard power level. In actual urban spaces, the path between some vehicles and certain charging stations is inaccessible or impractical (e.g., charging stations more than 30 kilometers away). One of the items will be marked as empty to avoid misleading modeling or training bias caused by sparse padding.

[0084] Furthermore, it is necessary to obtain reference values ​​for vehicle power behavior in advance for subsequent load calculations. The average charging power for each type of vehicle can be calculated from the transaction history of the operating platform. and average charging time The expected charging capacity is calculated as follows:

[0085] ;

[0086] in, It is a vehicle The average charging power is calculated from historical transaction records, categorized by vehicle model. It is a vehicle The average charging time is also derived from historical data. Indicates vehicle The expected charging capacity.

[0087] For example, a certain model E100 vehicle in the historical records , (minutes) This value is used after normalization.

[0088] Through steps S201 to S202, the sparse state tensor uses coordinate matching during the construction process to ensure the spatial correspondence between vehicle, charging pile, and path states. Its sparsity preserves the natural gaps in some "unreachable" charging piles in the city-level charging pile network, avoiding deviations caused by artificial completion.

[0089] like Figure 3 As shown, in some embodiments, step S102 includes steps S301 to S303:

[0090] Step S301: Construct the vehicle's intention weight for the charging station based on the vehicle's current battery level, the charging station's current status, the desired charging capacity, the charging station's maximum charging power, the path travel distance, and a preset Boolean flag.

[0091] Step S302: Simulate the impact of vehicle intent on the spatial neighborhood according to the preset spatial diffusion kernel function.

[0092] Step S303: Construct a region contribution function based on the intention weight and the spatial diffusion kernel function, and substitute the sparse state tensor into the region contribution function to construct a charging intention distribution map;

[0093] It should be noted that the spatial diffusion kernel function takes the distance between the location in space and the charging pile as the independent variable, and the value of the spatial diffusion kernel function characterizes the diffusion intensity of the vehicle's intention to charge the area where the charging pile is located at the location.

[0094] It should be noted that in step S301, the intention weight not only expresses the strength of the vehicle's charging intention, but also incorporates considerations of key decision-making variables in urban traffic operations, such as charging station preference, queue tolerance, and power capacity differences. The vehicle's current battery level is... At that time, the vehicle's electricity consumption was The lower the vehicle's existing battery power, the stronger its intention to charge. Adjusting the parameters related to the impact of toll costs. and path distance The product of these is the first objective product. When the preset value is 1, the charging pile's maximum charging power is... The target ratio is the ratio of the sum of the products of the preset value and the first target. Preset weighting coefficients Boolean symbols The product is defined as the second objective product. The intention weight is obtained by weighting and summing the vehicle's consumed electricity, the current status of the charging station, the desired charging amount, the target ratio, and the product of the second target, as shown below:

[0095] ;

[0096] It should be noted that the above formula is used to characterize the relative intensity of a vehicle's intention to charge a certain charging station in the current state. Its core idea is to map the various factors that affect the charging decision onto a unified proportional space for weighted combination.

[0097] Understandably, the vehicle has already used up its electricity through This indicates the vehicle's urgent charging need; the lower the vehicle's remaining battery power, the higher this value, and the more significant its impact on the overall intent weighting. The current status of the charging station... As a feasibility constraint factor, this item is 0 when the charging station is occupied or in a faulty state, which makes the corresponding intention weight directly invalid, thereby preventing vehicles from selecting unavailable charging stations.

[0098] Furthermore, vehicles Expected charging capacity Used to characterize the scale of a vehicle's single charging demand, giving higher weight to vehicles with high energy demand in intent calculation; maximum charging power of charging piles. Path distance item This constitutes the target ratio, reflecting the trade-off between "high-power attraction" and "travel costs" for vehicles. The longer the path distance, the smaller this ratio, thereby mitigating the impact of distant charging stations on intent; Additional Items This is used to incorporate historical preferences or operational strategy factors, so that, without changing the main decision-making logic, appropriate preferential treatment is given to charging piles with historical usage preferences or preferential strategies.

[0099] Furthermore, regarding parameter settings, the parameter for adjusting the intensity of the impact of toll costs can be adjusted. Take non-negative real numbers, whose range of values ​​can be set as follows: The default value is 1, which is used to establish a balance between power attraction and travel distance. When the system is deployed in high-density urban areas or in scenarios where users are highly sensitive to route detours, the value can be increased appropriately. To enhance the distance penalty effect; when deployed in highway service areas or areas with sparsely distributed charging stations, the penalty can be reduced. To mitigate the influence of distance. Weighting coefficients. This is used to control the proportion of historical preferences in the overall intent weight, and its value range can be set to... The default value is 0.1, used to introduce weak preference correction without dominating the decision outcome; when the system needs to reflect obvious user brand preferences or operational guidance strategies, it can be appropriately increased. The value of .

[0100] For example, for a ride-hailing vehicle with 0.3% battery remaining, the target charging station is idle, has a power rating of 7, and is 0.2 meters away. And had previously charged the device to the brand of the charging station ( ),but It can be approximated as follows:

[0101] ;

[0102] It should be noted that even if the distance is not the shortest, high power and preference can significantly enhance the weight of intent distribution, so that the final intent heatmap has the dual expression capability of "focus + preference".

[0103] In step S302, This is a spatial diffusion kernel function used to simulate the effect of vehicle intent on the spatial neighborhood. This step uses a Gaussian kernel with sparse control, as shown below:

[0104] ;

[0105] in, This indicates the Euclidean distance between the current location and the target charging station. It is the diffusion scale parameter, used to control the propagation width of the Gaussian term. This is the spatial attenuation coefficient, used to control the linear suppression rate of the diffusion process. Its value is a non-negative real number, preferably ranging from 0.5 to 2.0, with a default value of 1.0. When the system is deployed in urban environments with uneven distribution of charging piles or obvious hotspot areas, this value can be appropriately increased. To enhance spatial sparsity and suppress excessive diffusion of intent; when deployed in areas with uniformly distributed charging piles and a wide range of vehicle charging options, the size can be appropriately reduced. To maintain the continuity of the intended distribution. When Exceed hour, The value will be truncated to zero, achieving automatic convergence of the spatial propagation range and avoiding misleading diffusion results in the intention heatmap where there are weak values ​​even though the target pile is far away. This improves the focus and representation efficiency of the charging intention distribution map.

[0106] It should be noted that this spatial diffusion kernel function is used to simulate the diffusion intensity of a vehicle's charging intention towards its target charging station in its spatial neighborhood. Compared to a conventional Gaussian kernel, this function introduces a linear decay factor in addition to the standard Gaussian term, which is achieved by multiplying by... To achieve sparse control over the diffusion range.

[0107] In step S303, the vehicle's intention distribution needs to be mapped to geospatial space. Considering that in reality, car owners do not precisely target a single charging station, but rather make behavioral decisions within a "certain area," a charging intention distribution map is needed. It should be constructed as a continuous function of urban space. To this end, this step uses a differentiable regional contribution function to simulate each vehicle. For each location in space The contribution of charging intention density is specifically expressed as follows:

[0108] ;

[0109] in, Indicates vehicle Electricity pile Intent weight, It is a spatial diffusion kernel function. Let represent the set of vehicles at time t. Let represent the set of charging piles at time t. This represents the coordinates of any geographical location point in the urban space. Indicates the electric pile Its geographical location.

[0110] Substituting the sparse state tensor into the above formula, we obtain the charging intention distribution map. , The structure is of size The floating-point density map can be integrated with a map system for heatmap display or used as model input.

[0111] It should be noted that in actual engineering deployment, It can be discretized into a two-dimensional floating-point matrix with a fixed resolution (e.g., 100×100), where each position represents the potential future charging density level. Spatial diffusion calculations can be efficiently performed using vectorization acceleration frameworks (e.g., TensorRT).

[0112] It is understandable that future moments are constructed through steps S301 to S303. Urban space charging intention distribution map This serves as a crucial intermediary link in the "reasoning from behavioral state to demand." This prediction is not aimed at a single vehicle, but rather at the potential cluster of collective spatial intentions that all vehicles may form in their current state.

[0113] Furthermore, it is necessary not only to make full use of the sparse state tensor that integrates path travel distance, vehicle current battery level, and charging pile current status in the aforementioned steps, but also to consider specific "soft constraints" and "physical guidance" factors in the urban charging pile service system, such as user preferences such as car owners avoiding low-power charging piles and having limited tolerance for queuing.

[0114] Furthermore, a regional contribution function is introduced, which combines spatial kernel feedback weighting, vehicle power expectation regulation and distance attenuation to achieve an explicit expression of the intent generation mechanism. A sparse spatial filter is used to ensure the behavioral focus of the heatmap, thereby improving the practicality and deployability of the results.

[0115] It should be noted that in step S103, the electrical pile... Spatial attraction You can directly see the charging intention distribution map Extract from, as shown below:

[0116] ;

[0117] Furthermore, the access utility function is constructed, and based on this, a probability distribution map of vehicle access to charging stations is established. The access utility function integrates multiple influencing factors such as city-level intent goals, individual vehicle status, charging station attributes, traffic conditions, and behavioral preferences. Its design objective is to meet the following engineering requirements:

[0118] It should be noted that excessive concentration of load on urban hotspot charging stations should be avoided to adapt to the load balance of the power grid; user preferences of "high power priority" and "low waiting time priority" should be reasonably expressed; the deviation of the actual navigation path of the vehicle should be limited to improve the realism of the behavior simulation; and the analytical nature of the model structure and the controllability of deployment should be maintained.

[0119] Specifically, the degree of adaptation between the charging station's power and the vehicle's power demand, and the deviation length between the vehicle's navigation path and the path to the charging station are obtained. The access utility function is constructed based on spatial attractiveness, the current state of the charging station, the degree of adaptation, the path travel distance, the deviation length, and a preset local load suppression factor. That is, the access utility function is as follows:

[0120] ;

[0121] in, Indicates the utility access value. Indicates spatial attractiveness. This indicates the degree to which the charging pile's power meets the vehicle's power requirements. It is an electric pole. Maximum supported charging power It is a vehicle The average charging power. Indicates whether the charging station is currently available; the value is 0 or 1. Indicates vehicle To the charging station The path travel distance. This indicates the deviation length between the vehicle's navigation path and the path to the charging station, derived from the navigation engine's trajectory comparison. The local load suppression factor for charging piles is derived from the local density convolution calculation of the area surrounding the charging pile in the charging intention distribution map. If there is a large concentration of vehicles intending to charge in the vicinity of the charging pile, its attractiveness is suppressed. When there is a large concentration of vehicles intending to charge in the vicinity of the charging pile... The value of increases accordingly, thus inhibiting the overall attractiveness of the charging pile in the access utility function; when the surrounding area of ​​the charging pile has a low density or is scattered, Smaller values ​​tend to have a weaker impact on access utility. This represents the empirical coefficient for adjusting distance attenuation. The empirical coefficient representing the path penalty. An empirical coefficient representing the intensity of load suppression. The value ranges from 0.5 to 3.0, with a default value of 1.0, and is used to reflect the vehicle's sensitivity to changes in travel distance; The value ranges from 0.5 to 2.0, with a default value of 1.0, and is used to characterize the vehicle's tolerance level to detour behavior; The value ranges from 0.1 to 1.0, with a default value of 0.3. It is used to control the intensity of the suppression of the attraction of the electric pile by the local intentional aggregation.

[0122] It should be noted that the low surrounding intention density refers to the low density of electric piles during the forecast period. Within the preset spatial neighborhood centered on the target, the local density statistics of the charging intention distribution map are at a low level. Specifically, the weighted average of the intention density within this neighborhood is lower than a preset threshold or lower than the average level of the global intention density, and its normalized value is usually less than 0.1. "Dispersed intention distribution" means that although a certain number of charging intentions exist within the neighborhood, these intentions do not form a clear spatial concentration, but rather exhibit a multi-point, uniform, or discrete distribution. Their local density peaks are not significant, resulting in relatively small local density values ​​calculated through convolution or weighted statistics. In both of these cases, the local load suppression factor... The values ​​of are all in the low range, with the preferred range being 0-0.3, so that the inhibitory effect on the attraction of the electric pile in the access utility function is weak or basically has no inhibitory effect.

[0123] It should be noted that, The calculation method essentially involves performing a local average convolution on the charging intention distribution map in the vicinity of each charging station, as shown below:

[0124] ;

[0125] in, Indicates the use of electric piles Centered on, with radius The spatial neighborhood. The weighted kernel function (such as a normalized Gaussian kernel) is used to simulate the "competitive suppression effect" of the density of nearby intentions on the attractiveness of the electric pile. This is a diagram showing the distribution of charging intentions.

[0126] It should be noted that this item reflects the modeling of the hotspot aggregation suppression mechanism to prevent the charging piles in the city center from being selected by all vehicles at the same time due to excessive density, which would cause model estimation bias and extreme load concentration.

[0127] In step S104, based on the utility access value The probability distribution of vehicle access to charging stations is obtained by using softmax normalization, as shown below:

[0128] ;

[0129] in, This represents a probability distribution diagram, indicating the probability distribution of each vehicle in the future time period. The probability of selecting any charging pile is represented by a two-dimensional array with dimension 1. . Indicates vehicle Electricity pile Access utility value. Represents a set of charging piles Any of the charging piles in the network, This represents the set of all available charging stations at time t. Indicates vehicle Electricity pile Access utility value.

[0130] For example, a taxi with 30% battery charge and a target power of 7 is 0.15 meters away from a charging station in the city center. What score does this charging station receive on the charging intention distribution map? However, due to the presence of three high-density target areas in the surrounding area, If the charging pile deviates from its predetermined path by 0.3 units, the overall utility will be reduced. The rate of decrease is significant, thus giving the vehicle a higher probability of turning towards a nearby suboptimal charging station.

[0131] In step S105, based on the probability distribution diagram Based on the vehicle's average charging power, predict future moments. Actual load data for each charging pile The actual predicted load data for each charging station is shown below:

[0132] ;

[0133] in, This represents the actual predicted load data. Represents a probability distribution graph. Indicates vehicle Average charging power, Let represent the set of vehicles at time t.

[0134] It should be noted that steps S101 to S105 of the embodiments of this application involve obtaining the sparse state tensor of the vehicle. A charging intention distribution map is constructed based on the sparse state tensor and a preset regional contribution function, where the regional contribution function simulates the charging intention density contribution of each vehicle to charging piles at each location in space. The spatial attraction of charging piles is extracted from the charging intention distribution map, and this spatial attraction is substituted into a preset access utility function to obtain an access utility value. The access utility value is then normalized to obtain a probability distribution map of vehicle access to charging piles. This probability distribution map represents the probability of each vehicle selecting any charging pile within a future time period. The actual predicted load data for each charging pile is calculated based on the probability distribution map and a preset average charging power, improving the accuracy of charging pile load prediction.

[0135] like Figure 4 As shown, in some embodiments, after step S105, a vehicle-charging network charging load prediction method further includes steps S401 to S402:

[0136] Step S401: Obtain the charging pile power suppression factor, the area power supply suppression factor, and the access behavior density regularization term, wherein the charging pile power suppression factor indicates that the closer the charging pile's power load is to the limit, the stronger the correction.

[0137] Step S402: Select the minimum value among the product of the charging pile power suppression factor and the preset maximum charging power of the charging pile, the product of the regional power supply suppression factor and the regional maximum power supply capacity, and the difference between the actual predicted load data and the access behavior density regularization term. Select the largest value from zero and the minimum value as the corrected predicted load data of the charging pile.

[0138] In step S401, due to numerous limitations in actual operation, such as the maximum power carrying capacity of charging piles, the capacity limitations of power grid feeders, and excessively high vehicle access density, only [the following is used]. This could lead to unrealistic predictions. Therefore, three limiting mechanisms are designed to limit the actual load within deployable limits by introducing adjustment terms:

[0139] It should be noted that the charging pile power suppression factor This is used to adjust whether the actual predicted load data exceeds the physical power limit of the charging pile itself. Specifically, based on the first ratio of the actual predicted load data to the sum of the charging pile's maximum charging power and a small positive constant, the first difference between 1 and the first ratio is calculated, and the first difference is truncated non-negatively to obtain the charging pile power suppression factor.

[0140] Furthermore, the regional power supply suppression factor This study simulates a load reduction mechanism when multiple power piles in the same area experience simultaneous high loads. Specifically, it calculates the total actual predicted load data by summing all actual predicted load data within the area. Then, it calculates the second difference between 1 and the second ratio of the total actual predicted load data to the sum of the area's maximum power supply capacity and a small positive constant. Finally, it performs non-negative truncation on this second difference to obtain the area's power supply suppression factor.

[0141] Furthermore, access behavior density regularization terms ,measure How many vehicles access it frequently and Extremely high density (indicating intense competition) is used to adjust the load to prevent excessive congestion when the density is too high. This involves constructing a density regularization term for access behavior based on a probability distribution map and preset adjustment coefficients.

[0142] In step S402, the minimum value among the following is selected: the product of the charging pile power suppression factor and the preset maximum charging power of the charging pile, the product of the regional power supply suppression factor and the regional maximum power supply capacity, and the difference between the actual predicted load data and the access behavior density regularization term; the largest value among zero and the minimum value is selected as the corrected predicted load data of the charging pile, as shown below:

[0143] ;

[0144] in, This indicates a correction to the predicted load data. This represents the actual predicted load data. Indicates the power suppression factor of the charging pile. This indicates that the closer the electrical load of the charging pile is to its limit, the stronger the correction. This represents the actual predicted load data. This indicates the maximum charging power of the charging station. This represents a small positive constant used to prevent the denominator from being zero or the calculation result from being too sensitive. Indicates the regional power supply suppression factor. This means that the closer the total load of electrical piles in the same feeder or power supply area is to the upper limit of the regional power supply capacity, the smaller the value of the regional power supply suppression factor, thus imposing a more stringent downward adjustment limit on the load result of individual electrical piles in subsequent predictions. For the purpose of connecting with the charging pile A group of electrical piles belonging to the same feeder line. Represents any electric pile in the set of electric piles. Indicates the electric pile The maximum power supply capacity of the area. This represents the access behavior density regularization term. , The squared term, which is a preset adjustment coefficient, emphasizes the high-density concentration effect and reflects the "predictive expansion brought about by high-frequency access".

[0145] It should be noted that in the calculation of the revised predicted load data, the "min" operation ensures that the physical and network levels do not exceed the upper limit. First, the minimum value is taken for the power constraint value of the charging piles, the regional power supply constraint value, and the actual predicted load data after correction by the access behavior density regularization term. Then, the minimum value is compared with 0, and the larger of the two is taken as the revised predicted load data. Therefore, when the difference between the actual predicted load data and the access behavior density regularization term is less than 0, the revised predicted load data is 0, thus ensuring that the final output revised predicted load data is non-negative, meeting the basic fault tolerance requirements of urban power distribution systems. It can be connected to city-level power distribution systems or operation platform dispatch systems for real-time early warning, power allocation, load balancing and other operations.

[0146] For example, if a certain charging station of ,and , And its feeder The total projected load is 45, calculated as follows: , ,at the same time The final revised forecast load data is as follows:

[0147] ;

[0148] This indicates that the charging pile will be marked as an overloaded and unschedulable charging pile by the system to prevent unrealistic predictions from entering the dispatching system.

[0149] In this embodiment, steps S401 to S402 involve correcting the actual predicted load data to obtain corrected predicted load data, thereby further improving the accuracy of load prediction for power piles.

[0150] Example 2, refer to Figure 5 As one embodiment of the present invention, this embodiment provides a vehicle-charging-road network charging load prediction system, comprising:

[0151] Acquisition unit 501 is used to acquire the sparse state tensor of the vehicle.

[0152] The construction unit 502 is used to construct a charging intention distribution map based on the sparse state tensor and the preset regional contribution function; wherein, the regional contribution function simulates the charging intention density contribution of each vehicle to the charging pile at each location in space.

[0153] Extraction unit 503 is used to extract the spatial attractiveness of charging piles from the charging intention distribution map, and substitute the spatial attractiveness into a preset access utility function to obtain the access utility value.

[0154] The processing unit 504 is used to perform normalization processing based on the access utility value to obtain a probability distribution map of the vehicle accessing the charging station; wherein, the probability distribution map represents the probability of each vehicle selecting any charging station in the future time period.

[0155] Prediction unit 505 is used to calculate the actual predicted load data of each charging pile based on the probability distribution map and the preset average charging power;

[0156] The sparse state tensor is a three-dimensional tensor composed of vehicle dimension, charging pile dimension and feature dimension. Each element of the sparse state tensor is a joint state feature vector between the corresponding vehicle and the corresponding charging pile. The joint state feature vector includes the path travel distance, the current battery level of the vehicle, the current state of the charging pile and the maximum charging power of the charging pile. The elements between vehicles and charging piles that are not accessible by path are empty.

[0157] The charging intention distribution map is a continuous distribution of charging intention density at each geographical location, with geographical location in the urban space as the independent variable.

[0158] The spatial attraction is the value of the charging intention distribution map at the geographical location of the charging pile;

[0159] The access utility function is used to calculate the access utility value of a vehicle accessing a charging station, using the spatial attractiveness, vehicle individual status, charging station attributes, traffic conditions, and behavioral preferences as variables.

[0160] This embodiment also provides an electronic device applicable to a vehicle-charging-road network charging load prediction method, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the vehicle-charging-road network charging load prediction method proposed in the above embodiment.

[0161] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements a vehicle-charging network charging load prediction method as proposed in the above embodiments.

[0162] The storage medium proposed in this embodiment belongs to the same inventive concept as the method for predicting the charging load of a vehicle-pile road network proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0163] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause an electronic device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0164] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for predicting the charging load of a vehicle-charging network, characterized in that: include, Obtain the sparse state tensor of the vehicle; A charging intention distribution map is constructed based on the sparse state tensor and a preset regional contribution function; wherein, the regional contribution function simulates the charging intention density contribution of each vehicle to a charging pile at each location in space. The spatial attractiveness of the charging piles is extracted from the charging intention distribution map, and the spatial attractiveness is substituted into a preset access utility function to obtain the access utility value; The access utility value is normalized to obtain a probability distribution map of vehicle access to charging stations; wherein, the probability distribution map represents the probability of each vehicle selecting any charging station in a future time period. The actual predicted load data for each charging pile is calculated based on the probability distribution diagram and the preset average charging power. The sparse state tensor is a three-dimensional tensor composed of vehicle dimension, charging pile dimension and feature dimension, which respectively correspond to vehicle index, charging pile index and joint state feature vector dimension of each vehicle-charging pile pair. Each element of the sparse state tensor is the joint state feature vector between the corresponding vehicle and the corresponding charging pile. The joint state feature vector includes path travel distance, vehicle current battery level, charging pile current state and charging pile maximum charging power. The elements between vehicles and charging piles that are not accessible by path are empty. The charging intention distribution map is a continuous distribution of charging intention density at each geographical location, with geographical location in the urban space as the independent variable. The spatial attraction is the value of the charging intention distribution map at the geographical location of the charging pile; The access utility function is used to calculate the access utility value of a vehicle accessing a charging station, using the spatial attractiveness, vehicle individual status, charging station attributes, traffic conditions, and behavioral preferences as variables.

2. The method for predicting vehicle-charging network charging load as described in claim 1, characterized in that: After calculating the actual predicted load data for each charging pile based on the probability distribution map and the preset average charging power, the method further includes: Obtain the charging pile power suppression factor, the regional power supply suppression factor, and the access behavior density regularization term; The minimum value among the following is selected: the product of the charging pile power suppression factor and the preset maximum charging power of the charging pile, the product of the regional power supply suppression factor and the preset maximum regional power supply capacity, and the difference between the actual predicted load data and the access behavior density regularization term. The largest value between zero and the minimum value is selected as the corrected predicted load data for the charging pile.

3. The method for predicting vehicle-charging network charging load as described in claim 2, characterized in that: The acquisition of the charging pile power suppression factor, the regional power supply suppression factor, and the access behavior density regularization term includes: Calculate the first ratio of the actual predicted load data to the sum of the maximum charging power of the charging pile and a small positive constant, calculate the first difference between 1 and the first ratio, and perform non-negative truncation on the first difference to obtain the charging pile power suppression factor. The total actual predicted load data is obtained by summing all the actual predicted load data in the region. Calculate the second ratio of the total actual predicted load data to the sum of the maximum power supply capacity of the region and the small positive constant, calculate the second difference between 1 and the second ratio, and perform non-negative truncation on the second difference to obtain the power supply suppression factor of the region. The access behavior density regularization term is constructed based on the probability distribution diagram and the preset adjustment coefficient; The non-negative truncation process is as follows: when the first difference or the second difference is less than 0, the result is 0; when the first difference or the second difference is not less than 0, the result is retained.

4. The method for predicting vehicle-charging network charging load as described in claim 3, characterized in that: The acquisition of the sparse state tensor of the vehicle includes: Acquire vehicle status data, charging station status data, and route travel distance; The sparse state tensor is constructed based on the vehicle status data, the charging pile status data, and the path travel distance.

5. The method for predicting vehicle-charging network charging load as described in claim 4, characterized in that: The sparse state tensor includes the path travel distance, the vehicle's current battery level, the current status of the charging station, and the maximum charging power of the charging station. The step of constructing a charging intention distribution map based on the sparse state tensor and a preset region contribution function includes: The vehicle's intention weight for the charging station is constructed based on the vehicle's current battery level, the charging station's current status, the desired charging amount, the charging station's maximum charging power, the path travel distance, and a preset Boolean flag. The influence of vehicle intent on the spatial neighborhood is simulated based on a preset spatial diffusion kernel function; The region contribution function is constructed based on the intention weight and the spatial diffusion kernel function, and the sparse state tensor is substituted into the region contribution function to construct the charging intention distribution map. The spatial diffusion kernel function takes the distance between the location in space and the charging pile as the independent variable, and the value of the spatial diffusion kernel function characterizes the diffusion intensity of the vehicle's intention to charge the area where the charging pile is located at the location.

6. The method for predicting vehicle-charging network charging load as described in claim 5, characterized in that: The process of constructing the vehicle's intent weight for the charging station based on the vehicle's current battery level, the charging station's current status, the desired charging capacity, the charging station's maximum charging power, the path travel distance, and a preset Boolean flag includes: The vehicle's current battery level is used to calculate the vehicle's current battery consumption. The product of the preset adjustment cost impact intensity parameter and the path travel distance is defined as the first target product; The target ratio is obtained by taking the ratio of the sum of the product of the maximum charging power of the charging pile and the preset value and the first target value. The product of the preset weight coefficient and the Boolean flag is defined as the second target product; The intention weight is obtained by weighting and summing the product of the vehicle's consumed electricity, the current status of the charging pile, the expected charging amount, the target ratio, and the second target.

7. The method for predicting vehicle-charging network charging load as described in claim 6, characterized in that: The construction of the access utility function includes: To obtain the degree of compatibility between the charging station's power and the vehicle's power requirements, and the deviation length between the vehicle's navigation path and the path to the charging station; The access utility function is constructed based on the spatial attraction, the current state of the charging pile, the suitability, the path travel distance, the deviation length, and a preset local load suppression factor; wherein, the preset local load suppression factor is obtained by performing a local average convolution in the area near each charging pile based on the charging intention distribution map.

8. A vehicle-charging-road network charging load prediction system, employing the vehicle-charging-road network charging load prediction method as described in any one of claims 1 to 7, characterized in that, include: The acquisition unit is used to acquire the sparse state tensor of the vehicle. A construction unit is used to construct a charging intention distribution map based on the sparse state tensor and a preset regional contribution function; wherein, the regional contribution function simulates the charging intention density contribution of each vehicle to a charging pile at each location in space. An extraction unit is used to extract the spatial attractiveness of charging piles from the charging intention distribution map, and substitute the spatial attractiveness into a preset access utility function to obtain an access utility value; The processing unit is used to perform normalization processing based on the access utility value to obtain a probability distribution map of vehicle access to charging piles; wherein, the probability distribution map represents the probability of each vehicle selecting any charging pile in a future time period. The prediction unit is used to calculate the actual predicted load data of each charging pile based on the probability distribution map and the preset average charging power. The sparse state tensor is a three-dimensional tensor composed of vehicle dimension, charging pile dimension and feature dimension, which respectively correspond to vehicle index, charging pile index and joint state feature vector dimension of each vehicle-charging pile pair. Each element of the sparse state tensor is the joint state feature vector between the corresponding vehicle and the corresponding charging pile. The joint state feature vector includes path travel distance, vehicle current battery level, charging pile current state and charging pile maximum charging power. The elements between vehicles and charging piles that are not accessible by path are empty. The charging intention distribution map is a continuous distribution of charging intention density at each geographical location, with geographical location in the urban space as the independent variable. The spatial attraction is the value of the charging intention distribution map at the geographical location of the charging pile; The access utility function is used to calculate the access utility value of a vehicle accessing a charging station, using the spatial attractiveness, vehicle individual status, charging station attributes, traffic conditions, and behavioral preferences as variables.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the vehicle-pile-road network charging load prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the vehicle-pile-road network charging load prediction method according to any one of claims 1 to 7.

Images