A public charging pile demand prediction method and system based on differentiable symbolic learning

Abstract

The application provides a public charging pile demand prediction method and system based on differentiable symbolic learning. According to the charging pile use data of a city, a symbolic regression model is used to generate a corresponding expression between the charging pile use data and the demand parameters, and a call weight is set for each operator symbol to realize differentiable iteration, thereby accelerating the iteration speed, and then quickly obtaining the expression. Furthermore, the city is divided into a hexagonal network on the basis of the original partition, the charging piles are configured by using the hexagonal network to radiate the entire city range, and the configuration parameters of the charging piles in the hexagonal network region are determined, so as to as far as possible reduce the delivery cost of the charging piles and improve the utilization rate.

Description

Technical Field

[0001] This application relates to the field of charging pile deployment technology, specifically to a method and system for predicting the demand for public charging piles based on differentiable symbolic learning. Background Technology

[0002] With the continuous growth of the number of new energy vehicles, the demand for charging is also increasing, and the supply of charging piles has become a major factor restricting the development of new energy vehicles. Current public charging pile planning technologies are based on experience and cannot respond to dynamic changes in urban areas or effectively match user needs. Therefore, there is an urgent need for a charging pile layout planning method that can be tailored to different urban areas and user needs. Summary of the Invention

[0003] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a method and system for predicting the demand for public charging stations based on differentiable symbolic learning.

[0004] According to one aspect of this application, a method for predicting the demand for public charging piles based on differentiable symbolic learning is provided, comprising: acquiring charging pile usage data of a target city; inputting the charging pile usage data into a symbolic regression model to obtain charging pile demand parameters of the target city; wherein the symbolic regression model includes multiple operators, each operator corresponding to a call weight, and determining the probability of the corresponding operator being called based on the call weight; acquiring the boundaries of multiple target partitions of the target city; constructing multiple hexagonal network regions based on the multiple target partitions; wherein each target partition is divided into at least one hexagonal network region; and calculating charging pile configuration parameters of the hexagonal network region based on the regional information of the hexagonal network region and the charging pile demand parameters of the target city.

[0005] In one embodiment, the charging pile usage data includes user usage characteristics and macro-regional characteristics, and the symbolic regression model includes a parallel fully connected network layer and a one-dimensional convolutional layer; wherein, the step of inputting the charging pile usage data into the symbolic regression model to obtain the charging pile demand parameters of the target city includes: inputting the user usage characteristics and the macro-regional characteristics into the fully connected network layer and the one-dimensional convolutional layer respectively to obtain user usage characteristics and macro-regional characteristics; fusing the user usage characteristics and the macro-regional characteristics to obtain target characteristics; and obtaining the charging pile demand parameters of the target city based on the target characteristics and the computational expression; wherein, the computational expression is determined by the invoked operator.

[0006] In one embodiment, constructing multiple hexagonal network regions based on multiple target partitions includes: dividing the multiple target partitions into multiple hexagonal network regions based on the center point of the multiple target partitions and the maximum radius of the hexagonal network regions.

[0007] In one embodiment, calculating the charging pile configuration parameters of the hexagonal network area based on the area information of the hexagonal network area and the charging pile demand parameters of the target city includes: determining the charging pile demand levels of multiple hexagonal network areas based on the area information of the hexagonal network area; and calculating the charging pile configuration parameters of each hexagonal network area based on the charging pile demand levels of multiple hexagonal network areas and the charging pile demand parameters of the target city.

[0008] In one embodiment, calculating the charging pile configuration parameters for each hexagonal network region based on the charging pile demand levels of multiple hexagonal network regions and the charging pile demand parameters of the target city includes: calculating the charging pile demand parameters for each hexagonal network region based on the charging pile demand levels of multiple hexagonal network regions and the charging pile demand parameters of the target city; and calculating the charging pile configuration parameters for the hexagonal network region based on the charging pile demand parameters of the hexagonal network region.

[0009] In one embodiment, calculating the charging pile configuration parameters of the hexagonal network area based on the charging pile demand parameters of the hexagonal network area includes: calculating the charging pile configuration influence factor of the hexagonal network area based on the charging pile demand parameters of the hexagonal network area; and calculating the charging pile configuration parameters of the hexagonal network area based on the charging pile configuration influence factor of the hexagonal network area.

[0010] In one embodiment, the charging pile configuration influencing factors include a spatial weighting factor, a temporal distribution factor, and a spatial clustering factor; wherein, calculating the charging pile configuration parameters of the hexagonal network region based on the charging pile configuration influencing factors of the hexagonal network region includes: calculating the location and number of charging piles in the hexagonal network region based on the spatial weighting factor, the temporal distribution factor, and the spatial clustering factor.

[0011] In one embodiment, calculating the location and number of charging piles in the hexagonal network region based on the spatial weighting factor, the temporal distribution factor, and the spatial aggregation factor includes: calculating a baseline number of charging piles in the hexagonal network region based on the spatial weighting factor, the temporal distribution factor, and the spatial aggregation factor; and calculating the location and number of charging piles in the hexagonal network region based on the baseline number of charging piles in the hexagonal network region.

[0012] In one embodiment, calculating the location and number of charging piles in the hexagonal network region based on the baseline number of charging piles in the hexagonal network region includes: determining the location and number of cluster centers in the hexagonal network region using a clustering algorithm based on the baseline number of charging piles in the hexagonal network region; and determining the location and number of charging piles in the hexagonal network region based on the location and number of cluster centers.

[0013] According to another aspect of this application, a public charging pile demand prediction system based on differentiable symbolic learning is provided, comprising: a city data acquisition module for acquiring charging pile usage data of a target city; a city demand determination module for inputting the charging pile usage data into a symbolic regression model to obtain charging pile demand parameters of the target city; wherein the symbolic regression model includes multiple operators, each operator corresponding to a call weight, and the probability of the corresponding operator being called is determined based on the call weight; a target partition determination module for acquiring the boundaries of multiple target partitions of the target city; a network region construction module for constructing multiple hexagonal network regions based on the multiple target partitions; wherein each target partition is divided into at least one hexagonal network region; and a configuration parameter calculation module for calculating charging pile configuration parameters of the hexagonal network regions based on the regional information of the hexagonal network regions and the charging pile demand parameters of the target city.

[0014] This application provides a method and system for predicting public charging pile demand based on differentiable symbolic learning. The method involves: acquiring charging pile usage data for a target city; inputting the charging pile usage data into a symbolic regression model to obtain charging pile demand parameters for the target city; wherein the symbolic regression model contains multiple operators, each corresponding to a call weight, and determining the probability of the corresponding operator being invoked based on the call weight; acquiring the boundaries of multiple target partitions in the target city; constructing multiple hexagonal network regions based on the multiple target partitions; wherein each target partition is divided into at least one hexagonal network region; and using the regional information of the hexagonal network regions... Based on the charging pile demand parameters of the target city, the charging pile configuration parameters of the hexagonal network area are calculated. According to the charging pile usage data of the city, a symbolic regression model is used to generate the corresponding expression between the charging pile usage data and the demand parameters. A call weight is set for each operator to achieve differentiable iteration, thereby accelerating the iteration speed and quickly obtaining the expression. The city is divided into hexagonal networks based on the original partitions. The charging piles are configured using the hexagonal network to cover the entire city area. At the same time, the configuration parameters of the charging piles within the hexagonal network area are determined, thereby minimizing the deployment cost of charging piles and improving the utilization rate. Attached Figure Description

[0015] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 This is a flowchart illustrating a public charging pile demand prediction method based on differentiable symbolic learning, provided in an exemplary embodiment of this application.

[0017] Figure 2 This is a schematic diagram of the structure of a public charging pile demand prediction system based on differentiable symbolic learning provided in an exemplary embodiment of this application.

[0018] Figure 3 This is a structural diagram of an electronic device provided in an exemplary embodiment of this application. Detailed Implementation

[0019] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0020] Figure 1 This is a flowchart illustrating an exemplary embodiment of the public charging pile demand prediction method based on differentiable symbolic learning provided in this application. Figure 1 As shown, the public charging pile demand prediction method based on differentiable symbolic learning includes the following steps:

[0021] Step 110: Obtain charging pile usage data for the target city.

[0022] Charging pile usage data can include: dynamic population distribution data (daytime and nighttime density changes, commuting patterns), electric vehicle behavior data (mileage distribution, charging hotspots), and land function constraint data (land use codes, spatial development intensity), etc. Preferably, this application can also use interpolation to fill data gaps and establish a dual index relationship between geographic grids and timestamps to ensure the continuity of multidimensional data in the spatiotemporal dimensions.

[0023] Step 120: Input the charging pile data into the symbolic regression model to obtain the charging pile demand parameters for the target city.

[0024] The symbolic regression model contains multiple operators, each corresponding to a call weight. The probability of an operator being called is determined based on these call weights. Internally, the model includes a predefined library of mathematical symbols. This library forms the foundational set of atomic operations for constructing the final mathematical expression. The library contains the following elements:

[0025] Binary operators include addition, multiplication, subtraction, and protective division (pdiv). Protective division is defined as pdiv(a,b) = a / (b+ε), where ε is a very small positive number used to prevent division by zero errors during training and ensure numerical stability.

[0026] Elementary functions in one variable: square, square root, natural logarithm, and exponential function. These functions are crucial for capturing nonlinear relationships between variables.

[0027] Learnable constants: a set of trainable parameter scalars C1, C2, ..., C k These constants, which serve as coefficients, exponents, or thresholds in the formula, will have their optimal values ​​learned automatically during training via gradient descent, and are a key manifestation of the model's adaptability.

[0028] Feature tokens: refer to the various dimensions H1, H2, ..., H1 output by the upper-layer feature encoder. d In this layer, each dimension feature is itself a non-linear combination of the original input features, and this layer will learn which high-level feature combinations contribute the most to the prediction.

[0029] This application simulates the forward propagation process of a mathematical symbol library as a differentiable expression tree evaluation. Its core is a trainable weight tensor W, which represents the probability of each operator being invoked. Specifically, for the input feature H, its association strength with elements in the symbol library is calculated through linear transformation and a softmax function: A = softmax(H·W), where each element A[i,j] of the weight matrix A represents the importance of the j-th element in the symbol library (corresponding to the j-th operator) for the i-th sample in constructing the expression. Based on the weights A, this layer softly executes all possible operations in the symbol library in a weighted summation manner. This process can be predefined with a fixed computation graph structure (a two-layer expression tree). The computation starts from the leaf nodes (constants and feature tokens) and recursively evaluates upwards to the root node (a binary operator). The entire computation process consists of a series of differentiable operations, ensuring that the gradient can propagate smoothly backwards. Finally, the root node outputs the layer's output, which is a differentiable approximation of the target symbolic expression under the current parameters.

[0030] To ensure the model ultimately yields a concise and interpretable expression, this application imposes explicit sparsity constraints during training. Specifically, a joint loss function is incorporated into the training of this layer:

[0031] L total =L pred (y pred y true )+λ·L reg ;

[0032] Among them, L pred (y pred y true ) is the prediction loss (such as the predicted value y) pred and the true value y true The mean squared error between L and L is used to ensure the accuracy of the model. reg The regularization term is crucial for achieving interpretability. Here, L1 regularization is applied to the weight tensor of the symbolic layer. reg =||W||, where λ is a hyperparameter used to balance prediction accuracy and model simplicity.

[0033] After model training is complete, parsing is required to transform the learned parameters into human-readable mathematical formulas. A thresholding process is applied to the trained weight tensor W, retaining only the k connections with the largest absolute values ​​or weights greater than a certain threshold. This determines the invoked symbols and their connections, constructing a clear expression tree structure. The generated expression is then algebraically simplified to its simplest form, which is the final interpretable expression.

[0034] By inputting different training data, the network output can include multi-dimensional prediction results: overall charging demand density, peak demand periods, average charging time, user type demand decomposition (BEV demand and PHEV demand), time distribution characteristics (morning peak demand ratio, off-peak demand ratio, evening peak demand ratio), spatial clustering coefficient (reflecting the degree of demand concentration), etc.

[0035] Step 130: Obtain the boundaries of multiple target partitions in the target city.

[0036] This application utilizes GIS spatial analysis technology to identify and delineate the boundaries and levels of the target city's central system by integrating multiple indicators. Specifically, by identifying characteristics such as the density of commercial and business facilities, the density of employment positions, the density and passenger flow of rail transit stations, the distribution of high-level public service facilities (municipal hospitals, large cultural venues), and peak land price areas, the boundaries and influence range of the target city's main urban center (CBD), sub-center, regional center, and community center are clarified, thereby obtaining multiple target zones (such as commercial districts, residential areas, and administrative districts).

[0037] Step 140: Construct multiple hexagonal network regions based on multiple target partitions.

[0038] Each target partition is divided into at least one hexagonal network region. This application uses the center point of each target partition as a generator to generate at least one hexagon to divide each target partition. The distance from any point within the hexagon to the generator (center point) within the polygon is less than the distance to any other generator.

[0039] Step 150: Based on the regional information of the hexagonal network area and the charging pile demand parameters of the target city, calculate the charging pile configuration parameters of the hexagonal network area.

[0040] After constructing multiple hexagonal network regions to divide target cities, this application comprehensively determines the charging pile configuration parameters for each hexagonal network region based on the regional information of the hexagonal network regions and the charging pile demand parameters of the target cities, that is, the charging pile configuration plan for each hexagonal network region.

[0041] Among them, the charging pile demand parameter of the target city can be the total demand baseline of the target city: N total =y demand ×S, where N total y is the baseline for total demand. demand S represents the overall charging demand density, and S represents the urban area of ​​the target city.

[0042] This application provides a method for predicting the demand for public charging piles based on differentiable symbolic learning. The method involves acquiring charging pile usage data for a target city; inputting this data into a symbolic regression model to obtain charging pile demand parameters for the target city; wherein the symbolic regression model contains multiple operators, each corresponding to a call weight, and the probability of each operator being invoked is determined based on the call weight; obtaining the boundaries of multiple target partitions in the target city; constructing multiple hexagonal network regions based on these partitions; wherein each target partition is divided into at least one hexagonal network region; calculating charging pile configuration parameters for the hexagonal network regions based on the regional information of the hexagonal network regions and the charging pile demand parameters of the target city; generating a corresponding expression between the charging pile usage data and the demand parameters using the symbolic regression model based on the city's charging pile usage data, and assigning a call weight to each operator to achieve differentiable iteration, thereby accelerating the iteration speed and quickly obtaining the expression; furthermore, dividing the city into hexagonal networks based on the original partitions, and using these hexagonal networks to configure charging piles to cover the entire city area, while simultaneously determining the configuration parameters of charging piles within the hexagonal network regions, thereby minimizing the deployment cost of charging piles and improving utilization.

[0043] In one embodiment, the charging pile usage data includes user usage characteristics and macro-regional characteristics, and the symbolic regression model includes a parallel fully connected network layer and a one-dimensional convolutional layer; wherein, the specific implementation of the above step 120 can be as follows: inputting the user usage characteristics and macro-regional characteristics into the fully connected network layer and the one-dimensional convolutional layer respectively to obtain the user usage characteristics and macro-regional characteristics; fusing the user usage characteristics and macro-regional characteristics to obtain the target characteristics; and obtaining the charging pile demand parameters of the target city based on the target characteristics and the operational expression; wherein, the operational expression is determined by the called operator.

[0044] After obtaining charging pile usage data in the target city, this application can generate user usage characteristics and macro-regional characteristics based on the charging pile usage data. The user usage characteristics include: vehicle energy type code = {0:BEV1:PHEV}, driving range characteristics = [driving range (km)] / standard driving range, battery health index = battery health (0-1), charging frequency (times / month) / base frequency, charging duration (hour) / average duration, distance from home to charging pile (km) / average city distance, distance from workplace to charging pile (km) / average city distance, self-driving commuting distance (km) / standard commuting distance, self-driving weekend travel distance (km) / standard weekend travel distance, self-driving holiday travel distance (km) / standard holiday travel distance, time preference characteristics ∈ {0: morning 1: noon 2: evening}, battery level at start of charging (%) / 100, battery level at end of charging (%) / 100, (end charge - start charge) / 100, etc.; Macro-regional characteristics include: regional BEV ownership / total number of city BEVs, regional PHEV ownership / total number of city PHEVs, [BEV sales in the next 1 year, BEV sales in the next 2 years, BEV sales in the next 3 years] / baseline sales, [PHEV sales in the next 1 year, PHEV sales in the next 2 years, PHEV sales in the next 3 years] / baseline sales, [BEV model 1 range, BEV model 2 range, ...] / standard range, [PHEV model 1 electric range, PHEV model 2 electric range, ...] / standard electric range, etc.

[0045] Different pathways are employed to address user usage features and macro-region features. Pathway 1 is an MLP sub-network consisting of 2-3 fully connected network layers, using the ReLU activation function to process user usage features and output a dense user usage feature. Pathway 2 is a one-dimensional convolutional network that processes the temporal and distributional features in the macro-region features, outputting a macro-region feature that integrates macro information. Finally, the macro-region features are concatenated and fused to form the final target feature vector, which serves as the final abstract representation of the entire deep network and is fed into the symbolic layer. The dimension of the target feature defines the complexity of the symbolic layer's search space.

[0046] In one embodiment, step 140 can be implemented by dividing the multiple target partitions into multiple hexagonal network regions based on the center point of the multiple target partitions and the maximum radius of the hexagonal network region.

[0047] This application divides the target area into multiple hexagonal network regions by setting a maximum radius of a hexagonal network region, which is the radius of radiation starting from the center point of the hexagonal network region.

[0048] In one embodiment, step 150 can be implemented as follows: based on the regional information of the hexagonal network area, determine the charging pile demand level of multiple hexagonal network areas; based on the charging pile demand level of multiple hexagonal network areas and the charging pile demand parameters of the target city, calculate the charging pile configuration parameters of each hexagonal network area.

[0049] After dividing the hexagonal network areas into multiple zones, this application determines the charging pile demand level for each zone based on its regional information. This regional information includes functional tags, which categorize the hexagonal network areas into high-intensity mixed-use zones (core business districts, central business districts), medium-to-high-intensity residential zones (mature residential communities), emerging industrial zones (high-tech zones, science parks), and general built-up areas (traditional industrial zones, transportation hubs, etc.). After determining the functional tags for each hexagonal network area, the application calculates the charging pile configuration parameters for each zone by combining these tags with the charging pile demand parameters of the target city.

[0050] In one embodiment, step 150 can be implemented as follows: based on the charging pile demand levels of multiple hexagonal network areas and the charging pile demand parameters of the target city, calculate the charging pile demand parameters of each hexagonal network area; based on the charging pile demand parameters of the hexagonal network areas, calculate the charging pile configuration parameters of the hexagonal network areas.

[0051] This application calculates the charging pile demand parameters for each hexagonal network area based on the charging pile demand levels of multiple hexagonal network areas and the charging pile demand parameters of the target city. Based on the charging pile demand parameters of the hexagonal network areas, it calculates the charging pile configuration parameters of the hexagonal network areas, that is, it allocates the charging pile demand parameters of the target city to each hexagonal network area to obtain the charging pile configuration parameters of each hexagonal network area.

[0052] In one embodiment, the specific implementation of step 150 above may be: calculating the charging pile configuration influence factor of the hexagonal network area based on the charging pile demand parameters of the hexagonal network area; and calculating the charging pile configuration parameters of the hexagonal network area based on the charging pile configuration influence factor of the hexagonal network area.

[0053] This application calculates the influence factors affecting the configuration of charging piles in a hexagonal network area, and calculates the configuration parameters of charging piles in the hexagonal network area based on the influence factors.

[0054] In one embodiment, the charging pile configuration influencing factors include spatial weighting factors, temporal distribution factors, and spatial clustering factors; wherein, the specific implementation of the above step 150 may be: calculating the location and number of charging piles in the hexagonal network area based on the spatial weighting factors, temporal distribution factors, and spatial clustering factors.

[0055] Specifically, the charging pile demand for the j-th hexagonal network region is:

[0056] ,

[0057] Where, N j For the charging pile demand in the j-th hexagonal network region, W j Let Adj be the spatial weight factor of the j-th hexagonal network region. j is the multidimensional adjustment coefficient for the j-th hexagonal network region.

[0058] The spatial weighting factor is calculated as follows:

[0059] W j =(w1·P j +w2·R j +w3·M j +w4·L j +w5·T j ) / Z,

[0060] Among them, P j Let R be the daytime population density of the j-th hexagonal network region. j Let M be the road network density of the j-th hexagonal network region. j L represents the functional mixing degree of the j-th hexagonal network region (calculated based on the land use code in the "Land Function Constraint Data"). j Let T be the land use intensity of the j-th hexagonal network region. j Let Z represent the traffic accessibility of the j-th hexagonal network region, and Z be the normalization factor to ensure that the sum of all spatial weight factors equals 1. w1, w2, w3, w4, and w5 are the weight coefficients of each indicator, respectively.

[0061] The calculation method for the multidimensional adjustment coefficient is as follows:

[0062] Adj j =Time j ×S j ×U j ,

[0063] Among them, Time j S is the time distribution factor of the j-th hexagonal network region. j U is the spatial clustering factor of the j-th hexagonal network region. j This is the user type adjustment factor for the j-th hexagonal network region.

[0064] The time distribution factor is calculated as follows:

[0065] Time j =1+α1×(y tj -0.33)×f j ,

[0066] Among them, y tj f represents the peak demand percentage of the j-th hexagonal network region. j Let α1 be the land function weight of the j-th hexagonal network region (commercial area = 1.2, residential area = 0.8, industrial area = 1.0), and let α1 be the first coefficient.

[0067] The spatial aggregation factor is calculated as follows:

[0068] S j =1+α2×(y sj -0.5),

[0069] Among them, y sj α1 is the spatial clustering coefficient of the j-th hexagonal network region, and α2 is the second coefficient.

[0070] The user type adjustment factor is calculated as follows:

[0071] U j =1+α3×(r j ×h j ),

[0072] r j =y bj / (y bj +y pj ),

[0073] Where, r j h represents the proportion of BEV demand in the j-th hexagonal network region. j y is the adjustment factor for the income level or land price of the j-th hexagonal network region. bj For the BEV demand of the j-th hexagonal network region, y pj α3 represents the PHEV requirement for the j-th hexagonal network region, and α3 is the third coefficient.

[0074] In one embodiment, step 150 can be implemented as follows: calculate the reference number of charging piles in the hexagonal network region based on the spatial weighting factor, the temporal distribution factor, and the spatial aggregation factor; and calculate the location and number of charging piles in the hexagonal network region based on the reference number of charging piles in the hexagonal network region.

[0075] This application calculates the baseline number of charging piles in a hexagonal network area based on spatial weighting factors, temporal distribution factors, and spatial clustering factors, and calculates the location and number of charging piles in the hexagonal network area based on the baseline number of charging piles in the hexagonal network area.

[0076] The formula for calculating the service capacity requirement of the j-th hexagonal network region is as follows:

[0077] Service j =(N j ×y aj ) / (T s ×η),

[0078] Among them, y aj T is the average charging time of the j-th hexagonal network region. s Let η be the daily effective service duration of a single pile in the j-th hexagonal network region, and η be the target utilization rate (0.7-0.8).

[0079] The formula for calculating the fast charging ratio of the j-th hexagonal network region is:

[0080] R fj =0.4+0.3×tanh[β×(y timej -12)+γC j ],

[0081] Among them, y timej For the peak demand period of the j-th hexagonal network region, C j Let γ be the commercial function density of the j-th hexagonal network region, and let γ be the set commercial coefficient.

[0082] In one embodiment, the specific implementation of step 150 above may be as follows: based on the baseline number of charging piles in the hexagonal network region, a clustering algorithm is used to determine the location and number of cluster centers in the hexagonal network region; based on the location and number of cluster centers, the location and number of charging piles in the hexagonal network region are determined.

[0083] This application employs a clustering algorithm, using the center point of a hexagonal network region as the cluster center point and the demand for charging piles as the weight for spatial clustering. The corresponding objective function is:

[0084] ,

[0085] in, u ik For the first i Configured in each hexagonal network region k The penalty coefficient for each charging station, x i For the first i The location of the center point of each hexagonal network region c k For the first k The location of a charging station.

[0086] By solving the above objective function, the optimal location and number of charging stations in each hexagonal network region are obtained. Furthermore, the number of fast charging devices and slow charging devices is calculated based on the ratio of fast charging piles to slow charging piles. And based on the number and area occupied by fast charging piles and slow charging piles, the total land area is calculated.

[0087] Figure 2 This is a schematic diagram of the structure of a public charging pile demand prediction system based on differentiable symbolic learning, provided in an exemplary embodiment of this application. Figure 2 As shown, the public charging pile demand prediction system 20 based on differentiable symbolic learning includes: a city data acquisition module 21, used to acquire charging pile usage data of the target city; a city demand determination module 22, used to input the charging pile usage data into a symbolic regression model to obtain the charging pile demand parameters of the target city; wherein, the symbolic regression model contains multiple operators, each operator corresponding to a call weight, and the probability of the corresponding operator being called is determined based on the call weight; a target partition determination module 23, used to acquire the boundaries of multiple target partitions of the target city; a network region construction module 24, used to construct multiple hexagonal network regions based on multiple target partitions; wherein, each target partition is divided into at least one hexagonal network region; and a configuration parameter calculation module 25, used to calculate the charging pile configuration parameters of the hexagonal network region based on the regional information of the hexagonal network region and the charging pile demand parameters of the target city.

[0088] This application provides a public charging pile demand prediction system based on differentiable symbolic learning. The system acquires charging pile usage data for a target city through a city data acquisition module 21; a city demand determination module 22 inputs the charging pile usage data into a symbolic regression model to obtain charging pile demand parameters for the target city; wherein the symbolic regression model contains multiple operators, each corresponding to a call weight, and the probability of the corresponding operator being called is determined based on the call weight; a target partition determination module 23 acquires the boundaries of multiple target partitions in the target city; and a network region construction module 24 constructs multiple hexagonal network regions based on the multiple target partitions; wherein each target partition is divided into at least one hexagonal network region. The configuration parameter calculation module 25 calculates the charging pile configuration parameters of the hexagonal network area based on the regional information of the hexagonal network area and the charging pile demand parameters of the target city. According to the city's charging pile usage data, a symbolic regression model is used to generate the corresponding expression between the charging pile usage data and the demand parameters. A call weight is set for each operator to achieve differentiable iteration, thereby accelerating the iteration speed and quickly obtaining the expression. Furthermore, the city is divided into hexagonal networks based on the original partitions, and charging piles are configured using the hexagonal network to cover the entire city area. At the same time, the configuration parameters of charging piles within the hexagonal network area are determined, thereby minimizing the deployment cost of charging piles and improving utilization.

[0089] In one embodiment, the charging pile usage data includes user usage characteristics and macro-regional characteristics, and the symbolic regression model includes a parallel fully connected network layer and a one-dimensional convolutional layer; wherein, the aforementioned city demand determination module 22 can be further configured to: input user usage characteristics and macro-regional characteristics into the fully connected network layer and the one-dimensional convolutional layer respectively to obtain user usage characteristics and macro-regional characteristics; fuse user usage characteristics and macro-regional characteristics to obtain target characteristics; and obtain charging pile demand parameters for the target city based on the target characteristics and the computational expression; wherein, the computational expression is determined by the invoked operator.

[0090] In one embodiment, the network region construction module 24 can be further configured to divide the multiple target partitions into multiple hexagonal network regions based on the center point of the multiple target partitions and the maximum radius of the hexagonal network region.

[0091] In one embodiment, the configuration parameter calculation module 25 can be further configured to: determine the charging pile demand level of multiple hexagonal network areas based on the regional information of the hexagonal network areas; and calculate the charging pile configuration parameters of each hexagonal network area based on the charging pile demand level of multiple hexagonal network areas and the charging pile demand parameters of the target city.

[0092] In one embodiment, the configuration parameter calculation module 25 can be further configured to: calculate the charging pile demand parameters for each hexagonal network area based on the charging pile demand levels of multiple hexagonal network areas and the charging pile demand parameters of the target city; and calculate the charging pile configuration parameters for the hexagonal network area based on the charging pile demand parameters of the hexagonal network area.

[0093] In one embodiment, the configuration parameter calculation module 25 can be further configured to: calculate the charging pile configuration influence factor of the hexagonal network area based on the charging pile demand parameters of the hexagonal network area; and calculate the charging pile configuration parameters of the hexagonal network area based on the charging pile configuration influence factor of the hexagonal network area.

[0094] In one embodiment, the charging pile configuration influencing factors include spatial weighting factors, temporal distribution factors, and spatial clustering factors; wherein, the above-mentioned configuration parameter calculation module 25 can be further configured to: calculate the location and number of charging piles in the hexagonal network area based on the spatial weighting factors, temporal distribution factors, and spatial clustering factors.

[0095] In one embodiment, the configuration parameter calculation module 25 can be further configured to: calculate the reference number of charging piles in the hexagonal network area based on the spatial weight factor, the temporal distribution factor, and the spatial aggregation factor; and calculate the location and number of charging piles in the hexagonal network area based on the reference number of charging piles in the hexagonal network area.

[0096] In one embodiment, the configuration parameter calculation module 25 can be further configured to: determine the location and number of cluster centers of the hexagonal network region based on the baseline number of charging piles in the hexagonal network region using a clustering algorithm; and determine the location and number of charging piles in the hexagonal network region based on the location and number of cluster centers.

[0097] Below, for reference Figure 3 This application describes an electronic device according to embodiments thereof. The electronic device may be either or both of a first device and a second device, or a standalone device independent of them, which may communicate with the first device and the second device to receive acquired input signals from them.

[0098] Figure 3 A block diagram of an electronic device according to an embodiment of this application is illustrated.

[0099] like Figure 3 As shown, the electronic device 10 includes one or more processors 11 and memory 12.

[0100] The processor 11 may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.

[0101] The memory 12 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 11 may execute the program instructions to implement the methods of the various embodiments of this application described above and / or other desired functions. Various contents such as input signals, signal components, and noise components may also be stored in the computer-readable storage medium.

[0102] In one example, the electronic device 10 may also include an input device 13 and an output device 14, which are interconnected via a bus system and / or other forms of connection mechanism (not shown).

[0103] When the electronic device is a standalone device, the input device 13 can be a communication network connector for receiving the collected input signals from the first device and the second device.

[0104] In addition, the input device 13 may also include, for example, a keyboard, a mouse, etc.

[0105] The output device 14 can output various information to the outside, including determined distance information, direction information, etc. The output device 14 may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0106] Of course, for the sake of simplicity, Figure 3 Only some of the components of the electronic device 10 relevant to this application are shown in this illustration; components such as buses, input / output interfaces, etc., are omitted. In addition, the electronic device 10 may include any other suitable components depending on the specific application.

[0107] In addition to the methods and apparatus described above, embodiments of this application may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps in the methods according to various embodiments of this application described in the "Exemplary Methods" section above.

[0108] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0109] Furthermore, embodiments of this application may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps in the methods according to various embodiments of this application described in the "Exemplary Methods" section above.

[0110] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0111] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0112] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0113] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0114] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0115] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A method for predicting the demand for public charging stations based on differentiable symbolic learning, characterized in that, include: Obtain charging station usage data for the target city; The charging piles are input into a symbolic regression model to obtain the charging pile demand parameters for the target city; wherein, the symbolic regression model contains multiple operators, each operator corresponding to a call weight, and the probability of the corresponding operator being called is determined based on the call weight; Obtain the boundaries of multiple target partitions of the target city; Based on the multiple target partitions, multiple hexagonal network regions are constructed; wherein each target partition is divided into at least one hexagonal network region; Based on the regional information of the hexagonal network area and the charging pile demand parameters of the target city, the charging pile configuration parameters of the hexagonal network area are calculated; The charging pile usage data includes user usage characteristics and macro-regional characteristics. The symbolic regression model includes parallel fully connected network layers and one-dimensional convolutional layers. The step of inputting the charging pile usage data into the symbolic regression model to obtain the charging pile demand parameters for the target city includes: The user usage features and the macro-region features are respectively input into the fully connected network layer and the one-dimensional convolutional layer to obtain the user usage features and the macro-region features. By fusing the user usage characteristics and the macro-regional characteristics, the target characteristics are obtained; Based on the target features and the computational expression, the charging pile demand parameters of the target city are obtained; wherein, the computational expression is determined by the invoked operator. The charging pile configuration parameters of the hexagonal network area include the charging pile demand; the formula for calculating the charging pile demand is: ， Where, N j For the charging pile demand in the j-th hexagonal network region, W j Let Adj be the spatial weight factor of the j-th hexagonal network region. j is the multidimensional adjustment coefficient for the j-th hexagonal network region.

2. The public charging pile demand prediction method based on differentiable symbolic learning according to claim 1, characterized in that, The construction of multiple hexagonal network regions based on multiple target partitions includes: Based on the center point of the multiple target partitions and the maximum radius of the hexagonal network region, the multiple target partitions are divided into multiple hexagonal network regions.

3. The public charging pile demand prediction method based on differentiable symbolic learning according to claim 1, characterized in that, The calculation of charging pile configuration parameters for the hexagonal network area based on the regional information of the hexagonal network area and the charging pile demand parameters of the target city includes: Based on the regional information of the hexagonal network region, the charging pile demand levels of multiple hexagonal network regions are determined; Based on the charging pile demand levels of multiple hexagonal network regions and the charging pile demand parameters of the target city, the charging pile configuration parameters for each hexagonal network region are calculated.

4. The public charging pile demand prediction method based on differentiable symbolic learning according to claim 3, characterized in that, The calculation of charging pile configuration parameters for each hexagonal network area, based on the charging pile demand levels of multiple hexagonal network areas and the charging pile demand parameters of the target city, includes: Based on the charging pile demand levels of multiple hexagonal network regions and the charging pile demand parameters of the target city, the charging pile demand parameters of each hexagonal network region are calculated. Based on the charging pile demand parameters of the hexagonal network area, the charging pile configuration parameters of the hexagonal network area are calculated.

5. The public charging pile demand prediction method based on differentiable symbolic learning according to claim 4, characterized in that, The calculation of charging pile configuration parameters for the hexagonal network area based on the charging pile demand parameters includes: Based on the charging pile demand parameters of the hexagonal network area, calculate the charging pile configuration influence factor of the hexagonal network area; Based on the charging pile configuration influence factor of the hexagonal network area, the charging pile configuration parameters of the hexagonal network area are calculated.

6. The public charging pile demand prediction method based on differentiable symbolic learning according to claim 5, characterized in that, The charging pile configuration influencing factors include spatial weighting factors, temporal distribution factors, and spatial clustering factors; wherein, the calculation of the charging pile configuration parameters of the hexagonal network region based on the charging pile configuration influencing factors of the hexagonal network region includes: Based on the spatial weighting factor, the temporal distribution factor, and the spatial clustering factor, the location and number of charging piles in the hexagonal network region are calculated.

7. The public charging pile demand prediction method based on differentiable symbolic learning according to claim 6, characterized in that, The calculation of the location and number of charging piles in the hexagonal network region based on the spatial weighting factor, the temporal distribution factor, and the spatial clustering factor includes: Based on the spatial weighting factor, the temporal distribution factor, and the spatial clustering factor, the baseline number of charging piles in the hexagonal network region is calculated. Based on the baseline number of charging piles in the hexagonal network area, the location and number of charging piles in the hexagonal network area are calculated.

8. The public charging pile demand prediction method based on differentiable symbolic learning according to claim 7, characterized in that, The calculation of the location and quantity of charging piles in the hexagonal network area based on the baseline number of charging piles in the hexagonal network area includes: Based on the baseline number of charging piles in the hexagonal network region, a clustering algorithm is used to determine the location and number of cluster centers in the hexagonal network region. Based on the location and number of the cluster centers, the location and number of charging piles in the hexagonal network region are determined.

9. A public charging pile demand prediction system based on differentiable symbolic learning, characterized in that, include: The city data acquisition module is used to acquire charging pile usage data for the target city. The city demand determination module is used to input the charging pile usage data into a symbolic regression model to obtain the charging pile demand parameters of the target city; wherein, the symbolic regression model contains multiple operators, and each operator corresponds to a call weight, and the probability of the corresponding operator being called is determined based on the call weight; The target partition determination module is used to obtain the boundaries of multiple target partitions of the target city; A network region construction module is used to construct multiple hexagonal network regions based on multiple target partitions; wherein each target partition is divided into at least one hexagonal network region; The configuration parameter calculation module is used to calculate the charging pile configuration parameters of the hexagonal network area based on the regional information of the hexagonal network area and the charging pile demand parameters of the target city. The charging pile usage data includes user usage characteristics and macro-regional characteristics; the urban demand determination module's symbolic regression model includes parallel fully connected network layers and one-dimensional convolutional layers; wherein, the further configuration is as follows: The user usage features and the macro-region features are respectively input into the fully connected network layer and the one-dimensional convolutional layer to obtain the user usage features and the macro-region features. By fusing the user usage characteristics and the macro-regional characteristics, the target characteristics are obtained; Based on the target features and the computational expression, the charging pile demand parameters of the target city are obtained; wherein, the computational expression is determined by the invoked operator. The charging pile configuration parameters of the hexagonal network area include the charging pile demand; the formula for calculating the charging pile demand is: ， Where, N j For the charging pile demand in the j-th hexagonal network region, W j Let Adj be the spatial weight factor of the j-th hexagonal network region. j is the multidimensional adjustment coefficient for the j-th hexagonal network region.

Images