A long path charging planning method based on hierarchical space aggregation search

Abstract

The application relates to a long-path charging planning method based on hierarchical space aggregation search, which comprises the following steps: establishing a multi-level virtual space, wherein each level of the virtual space is composed of a plurality of mutually connected regular hexagons; initializing the position of a charging station in the virtual space; obtaining the ratio k of actual and theoretical power loss per kilometer; calculating the farthest position point that can be reached by the current vehicle according to the total power of the current vehicle and the ratio k; judging whether the farthest position point has reached the terminal point, if yes, prompting the user that the terminal point can be reached without charging; if not, performing hierarchical space search from small to large according to the layer number at the farthest position point in the virtual space, adding the charging station information meeting the condition into a charging planning scheme, and feeding back to the user, if no charging station meeting the condition is found, prompting the user that the road cannot be charged; compared with the prior art, the application has the advantages of high search efficiency.

Description

Technical Field

[0001] This invention relates to the field of new energy vehicle charging, and in particular to a long-path charging planning method based on hierarchical spatial aggregation search. Background Technology

[0002] With strong domestic support for new energy vehicles, more and more people are choosing them. New energy vehicles are primarily powered by electricity, and charging is the most common way to replenish their power. While charging is convenient and easy for most everyday travel scenarios, it becomes more complex and challenging for long-distance trips. For example, questions arise such as: where to charge along a journey from Beijing to Shanghai; how many times to charge; how long each charge takes; the availability of fast charging stations in each location; traffic congestion; the energy consumption of air conditioning in winter and summer; and the possibility of the car running out of power before reaching a charging station. These are all practical concerns, and the question is how to ensure a safe arrival at the destination while minimizing total charging time.

[0003] Currently, there is a lack of patented methods for charging route planning in China. Existing technologies generally employ greedy algorithms, calculating the charging location for each step based on the vehicle's range until all charging locations for the entire route are found. This method typically has some drawbacks:

[0004] First, because this method typically uses a step-by-step search approach, the spherical distance between each point on the path and all charging stations needs to be calculated, as well as the spherical distance between every two adjacent points on the path. As the number of charging stations increases, and the path becomes longer with more points, the computational load grows exponentially, making it unsuitable for practical applications.

[0005] Secondly, the actual driving range of a vehicle is quite complex. Even for the same model, battery life and actual driving range can vary significantly due to different usage patterns. Users' locations, driving habits, and the usage of electrical devices within the vehicle also differ, leading to substantial deviations in the overall actual driving range. This method cannot account for these real-world driving conditions.

[0006] Third, due to the large number of charging stations, the power and quality of the charging guns equipped at each station vary. Therefore, choosing suitable and high-quality charging stations is crucial for route charging planning. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a long-path charging planning method based on hierarchical spatial aggregation search.

[0008] The objective of this invention can be achieved through the following technical solutions:

[0009] A long-path charging planning method based on hierarchical spatial aggregation search includes the following steps:

[0010] S1. Obtain the driving route based on the starting point and ending point of this trip, and establish a multi-level virtual space based on the geographical space where the driving route is located. Each level of virtual space consists of several interconnected regular hexagons. The side length of each regular hexagon decreases with the level number. Initialize the position of the charging station in the virtual space.

[0011] S2. Input the basic information of the vehicle's most recent n complete power loss, the farthest path from the current starting point to the destination, the current road information, and the current vehicle information into the pre-trained deep neural network model to obtain the ratio k of the actual and theoretical power loss per kilometer.

[0012] S3. Based on the current total battery level of the vehicle and the ratio k, calculate the farthest point that the vehicle can travel to.

[0013] S4. Determine whether the farthest point has reached the destination. If yes, proceed to step S7; otherwise, proceed to step S5.

[0014] S5. At the farthest point in the virtual space, perform a layer-by-layer spatial search based on the layer number from smallest to largest. If a charging station that can be charged is found, proceed to step S6. If no charging station that can be charged is found at the current farthest point, move the farthest point along the driving path towards the starting point by a first length threshold and repeat step S5.

[0015] S6. Add the charging station information from step S5 to the charging plan and provide feedback to the user. If no charging station is found, prompt the user that the road is not available for charging.

[0016] S7 prompts the user that they can reach their destination without needing to recharge.

[0017] Furthermore, after obtaining a charging station that meets the preset conditions, a new path is formed between the charging station and the destination, and the above steps S1 to S7 are repeated.

[0018] Furthermore, the aggregation scale of each layer of the virtual space is inversely proportional to the side length of the regular hexagon.

[0019] Furthermore, the search method in step S5 is as follows:

[0020] Locate the hexagon containing the current farthest point in the virtual space with the smallest layer, and search for charging stations that meet preset conditions within this hexagon. The preset conditions include that the vehicle can reach the station within a preset time and that the charging station has a charging pile that matches the vehicle. If no charging station that meets the preset conditions is found in the current layer, proceed to the next layer of virtual space and search within the hexagon containing the current farthest point.

[0021] Furthermore, the training process of the deep neural network model is as follows:

[0022] A1. Obtain complete historical power failure information, vehicle information, and road information of vehicles from the database, and set the model convergence threshold and initial neural network weights;

[0023] A2. Input the complete historical power outage information into the first input layer, and input the vehicle information and road information into the second input layer. After convolution and pooling the information from the first and second input layers respectively, input them together into the connection layer, and classify them through Softmax to obtain the classification result of the ratio of actual and theoretical power consumption per kilometer of the vehicle.

[0024] A3. Continue executing step S2 until the model loss converges to within the model convergence threshold range, and output the deep neural network model.

[0025] Furthermore, after initializing the location of the charging station in the virtual space, based on the charging station's usage records for the past month, the basic information of the charging guns in the charging station, and the abnormal information reported in the past two weeks, the charging station with poor overall quality is removed from the virtual space according to the charging station quality assessment method.

[0026] Furthermore, the charging station quality assessment method includes: setting a charging station utilization rate threshold and a charging station abnormality number threshold; if the charging station utilization rate is lower than the set utilization rate threshold or the number of abnormalities is higher than the abnormality number threshold, the charging station is marked as a poor quality charging station.

[0027] Furthermore, the road information includes weather information and road condition information.

[0028] Furthermore, the vehicle information includes average vehicle speed information, vehicle SOC information, and vehicle owner information.

[0029] Furthermore, the inequality for calculating the farthest location point is as follows:

[0030]

[0031] Based on the above inequality, when the values ​​at both ends are equal, the value at either end is the distance from the farthest point to the starting point of the path, where E is the vehicle's current total battery level, k represents the ratio of actual to theoretical battery loss per kilometer, i represents the layer number of the virtual space, M represents the side length of the virtual space, β represents the calculation loss error, and D... n This represents the spherical distance between adjacent points in the path, and n represents the number of points in the path.

[0032] Compared with the prior art, the present invention has the following advantages:

[0033] 1. This invention employs a hierarchical space composed of multiple regular hexagons with varying side lengths for spatial search, combined with different aggregation scales, which significantly reduces the computational complexity of the search and improves its efficiency. Simultaneously, a deep neural network model is used to obtain vehicle power loss information, ensuring the accuracy of the vehicle's required charging time and further enhancing the precision of charging station location. Furthermore, the hexagonal splicing method allows the spliced ​​hierarchical space to form a spherical arc surface, more closely resembling the Earth's surface rather than a single plane, which helps improve search accuracy.

[0034] 2. This invention incorporates road and vehicle information during the training of the deep neural network model, determining accurate power loss information based on the environment, vehicles, and the vehicle owner, making it more realistic and further improving the accuracy of charging time.

[0035] 3. This invention has screened charging stations multiple times to ensure that the charging station equipment is in good condition and that vehicles can easily reach the geographical location of the charging station, thus providing a guarantee for vehicle charging. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the process of the present invention.

[0037] Figure 2 This is a diagram of the deep neural network model training architecture of the present invention. Detailed Implementation

[0038] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0039] This embodiment provides a long-path charging planning method based on hierarchical spatial aggregation search, such as... Figure 1 As shown, the specific steps include:

[0040] Step S1: Obtain the driving path L based on the starting point and ending point of this trip, and establish a multi-level virtual space based on the geographical space where the driving path is located.

[0041] Each layer of the virtual space consists of several interconnected regular hexagons. The side length of the hexagons decreases with each layer number. In this embodiment, there are four layers of the virtual space: the first layer has a side length of 22km, the second layer has 8km, the third layer has 3km, and the fourth layer has 1km. The purpose of setting different side lengths for the hexagons is to provide different search aggregation scales. A larger aggregation scale (i.e., smaller hexagon side lengths) results in a more refined search, but a slower search speed; conversely, a smaller aggregation scale results in a faster search speed, but a weaker search depth. Simultaneously, the path points on the path and the positions of the charging stations in the virtual space are initialized, and their indexes in the virtual space are created.

[0042] After initializing the location of charging stations in the virtual space, based on the charging station's usage records for the past month, basic information about the charging guns within the station, and reported anomalies from the past two weeks, charging stations with poor overall quality are removed from the virtual space according to the charging station quality assessment method. This method includes setting a charging station utilization rate threshold and a charging station anomaly frequency threshold. If a charging station's utilization rate is lower than the set threshold or its anomaly frequency exceeds the threshold, the charging station is marked as a poor-quality charging station.

[0043] Step S2: Obtain the basic information of the vehicle's most recent n complete power loss through vehicle historical data, and input the basic information of power loss, together with the longest path from the current starting point to the destination, the current road information, and the current vehicle information, into the pre-trained deep neural network model to obtain the ratio k of actual and theoretical power loss per kilometer.

[0044] The road information includes weather and road condition information, while the vehicle information includes average vehicle speed, vehicle SOC information, and vehicle owner information. By using real-time road conditions and vehicle driving habits as input to the neural network, the output can be more consistent with the actual situation and the judgment can be more accurate.

[0045] Specifically, the main network structures of deep neural networks are as follows: Figure 2 As shown, N represents the number of such layers.

[0046] The training process of a deep neural network model is as follows:

[0047] Step A1: Obtain complete historical power failure information, vehicle information, and road information of the vehicle from the database, and set the model convergence threshold and initial neural network weights.

[0048] Step A2: Input the complete historical power loss information into the first input layer, and input the vehicle information and road information into the second input layer. Perform multiple convolutions and pooling operations with kernels of 5, 3, and 1 on the first input layer, and perform multiple convolutions and pooling operations with kernels of 3 and 1 on the second input layer. Input the outputs of both into the connection layer, specifically by using the Concat function to connect the outputs of both, and then inputting them into the fully connected layer Dense. Finally, perform classification using Softmax to obtain the classification result of the ratio of actual to theoretical power consumption per kilometer of the vehicle.

[0049] Step A3: Continue executing step S2 until the model loss converges to within the model convergence threshold range, and output the deep neural network model.

[0050] Step S3: Based on the current total battery level and ratio k, calculate the farthest possible destination Pm that the vehicle can reach. The calculation method for the maximum distance is as follows:

[0051]

[0052] Based on the above inequalities, when the values ​​at both ends are equal, the value at either end is the distance from the farthest point Pm to the starting point of the path, where E is the vehicle's current total battery level, i represents the layer number of the virtual space, M represents the side length of the virtual space, β represents the calculation loss error, and D... n This represents the spherical distance between adjacent points in the path, where n represents the number of points in the path. A detailed illustration follows:

[0053] Assume the current total available power is the theoretical value E, which is 750 kW;

[0054] The current ratio of actual to theoretical power loss per kilometer, k, is 0.805 km / kW.

[0055] The current search space has a regular hexagon with side length M. i It is 22km;

[0056] The current road section's calculated loss error β is -10km;

[0057] The journey from Hangzhou to Yalong Bay National Tourist Resort in Haikou involves a total of 64 road segments {Unnamed Road, Gonglian Road, IoT Street, Jiangling Road, Jiangnan Avenue, Zhongxing Interchange, Shidai Elevated Road, Rainbow Interchange, Rainbow Expressway, Changshen Expressway, Hangchang Expressway, Lining Expressway, Lutangjiao Interchange, Hukun Expressway, Yingtan Interchange, Jigang Expressway, Guangchang Expressway, Unnamed Road…, Yalong Bay Road, Bohou Road, Bohou North Road, Unnamed Road, Bohou Road}, with a total distance of 22,162 spherical distances D. n, such as: {26.9m, 11.5m, 3m, 40m, 17m, 7m, 5m, 16m, 117m, 11m, 26m, 38m, 25m, 84m, 53m, 11m, 17m, 15m, 24m, 8m, 51m, 141m, 27m, 18m, 29m, 46m, 75m, 19m, 20m, 6m, 84m, 24m , 60m, 89m, 32m, 33m, 22m, 24m, 55m, 63m, 64m, 22m, 100m, 18m, 103m, 326m, 8m, 31m , 105m, 294m, 8m, 23m, 30m, 90m, 98m, 16m, 76m, 52m, 108m,..., 15m, 45m, 53m, 55m};

[0058] The maximum distance to the charging station during the first driving search is 750*0.805-22 / 2+(-10)=582km;

[0059] By solving the inequality using the formula, the value of the maximum position point Pm can be obtained as m = n = 2562.

[0060] Once a suitable charging station is found, the values ​​of E, k, and β are recalculated for subsequent road sections, and the maximum location point is then solved using the same formula.

[0061] Step S4: Determine whether the farthest point has reached the destination. If yes, the vehicle can drive directly to the destination, and proceed to step S7; otherwise, proceed to step S5.

[0062] Step S5: At the farthest point in the virtual space, perform a layer-by-layer spatial search based on the layer number from smallest to largest. If a charging station that can be charged is found, proceed to step S6. If no charging station that can be charged is found at the current farthest point, move the farthest point along the driving path towards the starting point by a first length threshold and repeat step S5.

[0063] Specifically, a charging station that meets preset conditions is defined as one that meets certain criteria. These conditions include the vehicle's ability to reach the station within a preset timeframe and the availability of a compatible charging station. For example, when a car is traveling on a highway, a charging station found using a hexagon with a larger side length might be located in a city near the highway. In this case, reaching the station would require finding an exit off the highway, which is inconvenient. Therefore, this situation is classified as a charging station that does not meet the preset conditions. Thus, by progressively decreasing the side length of the hexagon during the search, the effectiveness of the search can be improved.

[0064] In addition, the first length threshold is half the side length of the regular hexagon of the current virtual space layer.

[0065] Step S6: Add the charging station information from step S5 to the charging plan and provide feedback to the user. If no charging station is found, prompt the user that the road is not available for charging and do not recommend driving.

[0066] Step S7: Inform the user that they can reach the destination without needing to recharge.

[0067] Finally, after obtaining a charging station that meets the preset conditions and generating a charging plan, a new path is formed between the charging station and the destination. Steps S1 to S7 are repeated to ensure that the charging plan for the entire path can be completed.

[0068] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A long-path charging planning method based on hierarchical spatial aggregation search, characterized in that, Includes the following steps: S1. Obtain the driving route based on the starting point and ending point of this trip, and establish a multi-level virtual space based on the geographical space where the driving route is located. Each level of virtual space consists of several interconnected regular hexagons. The side length of each regular hexagon decreases with the level number. Initialize the position of the charging station in the virtual space. S2. Input the basic information of the vehicle's most recent n complete power loss, the farthest path from the current starting point to the destination, the current road information, and the current vehicle information into the pre-trained deep neural network model to obtain the ratio k of the actual and theoretical power loss per kilometer. S3. Based on the current total battery level of the vehicle and the ratio k, calculate the farthest point that the vehicle can travel to. S4. Determine whether the farthest point has reached the destination. If yes, proceed to step S7; otherwise, proceed to step S5. S5. At the farthest point in the virtual space, perform a layer-by-layer spatial search based on the layer number from smallest to largest. If a charging station that can be charged is found, proceed to step S6. If no charging station that can be charged is found at the current farthest point, move the farthest point along the driving path towards the starting point by a first length threshold and repeat step S5. S6. Add the charging station information from step S5 to the charging plan and provide feedback to the user. If no charging station is found, prompt the user that the road is not available for charging. S7 prompts the user that they can reach their destination without needing to recharge.

2. The long-path charging planning method based on hierarchical spatial aggregation search according to claim 1, characterized in that, After obtaining a charging station that meets the preset conditions, a new path is formed between the charging station and the destination, and the above steps S1 to S7 are repeated.

3. The long-path charging planning method based on hierarchical spatial aggregation search according to claim 1, characterized in that, The aggregation scale of each layer of the virtual space is inversely proportional to the side length of the regular hexagon.

4. The long-path charging planning method based on hierarchical spatial aggregation search according to claim 1, characterized in that, The specific search method in step S5 is as follows: Locate the hexagon in the virtual space with the fewest layers where the current farthest point is located, and search for charging stations that meet preset conditions in the hexagon. The preset conditions include that the vehicle can reach the station within a preset time and that the charging station has a charging pile that matches the vehicle. If no charging station that meets the preset conditions is found in the current layer, the system will enter the next layer of virtual space and search within the regular hexagon where the current farthest point is located.

5. The long-path charging planning method based on hierarchical spatial aggregation search according to claim 1, characterized in that, The training process of the deep neural network model is as follows: A1. Obtain complete historical power failure information, vehicle information, and road information of vehicles from the database, and set the model convergence threshold and initial neural network weights; A2. Input the complete historical power outage information into the first input layer, and input the vehicle information and road information into the second input layer. After convolution and pooling the information from the first and second input layers respectively, input them together into the connection layer, and classify them through Softmax to obtain the classification result of the ratio of actual and theoretical power consumption per kilometer of the vehicle. A3. Continue executing step S2 until the model loss converges to within the model convergence threshold range, and output the deep neural network model.

6. The long-path charging planning method based on hierarchical spatial aggregation search according to claim 1, characterized in that, After initializing the location of the charging station in the virtual space, based on the charging station's usage records for the past month, the basic information of the charging guns in the charging station, and the abnormal information reported in the past two weeks, the charging station with poor overall quality is removed from the virtual space according to the charging station quality assessment method.

7. The long-path charging planning method based on hierarchical spatial aggregation search according to claim 6, characterized in that, The charging station quality assessment method includes: setting a charging station utilization rate threshold and a charging station abnormality number threshold. If the charging station utilization rate is lower than the set utilization rate threshold or the number of abnormalities is higher than the abnormality number threshold, the charging station is marked as a poor quality charging station.

8. The long-path charging planning method based on hierarchical spatial aggregation search according to claim 1, characterized in that, The road information includes weather information and road condition information.

9. The long-path charging planning method based on hierarchical spatial aggregation search according to claim 1, characterized in that, The vehicle information includes average vehicle speed, vehicle SOC (State of Charge) information, and vehicle owner information.

10. The long-path charging planning method based on hierarchical spatial aggregation search according to claim 1, characterized in that, The inequality for calculating the farthest point is as follows: Calculations are performed based on the above inequalities. When the values ​​on both sides are equal, the value on either side represents the distance from the farthest point to the starting point of the path. Here, E is the vehicle's current total battery level, k represents the ratio of actual to theoretical battery loss per kilometer, i represents the layer number of the virtual space, and M represents the side length of the virtual space. This indicates the calculation error of the loss. This represents the spherical distance between adjacent points in the path. Indicates the number of path points.

Images