Vehicle power replenishment methods, devices, electronic equipment and readable storage media
By acquiring the vehicle's multimodal perception parameters and using a scene decision tree model to make decisions, a multidimensional energy replenishment strategy is generated. This solves the problem that existing technologies cannot make optimal energy replenishment decisions for different scenarios, achieving comprehensive and intelligent vehicle energy replenishment and reducing the anxiety of car owners.
Patent Information
- Application Number
- CN202611278313.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-21
- Publication Date
- 2026-11-17
AI Technical Summary
Existing vehicle charging methods cannot make optimal charging decisions for different scenarios, resulting in car owners' charging anxiety not being fundamentally improved.
By acquiring the vehicle's multimodal perception parameters, using a pre-trained scene decision tree model to make scene decisions, determining the vehicle's current scene category and scene feature vector, and performing strategy analysis, a multi-dimensional power replenishment strategy is generated, including power replenishment warning, route strategy, and control strategy.
It improves the comprehensiveness and intelligence of vehicle refueling decisions, enabling vehicles to make optimal refueling decisions for different scenarios and reducing refueling anxiety for car owners.
Smart Images

Figure CN123104433A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of vehicle power replenishment technology, specifically to a vehicle power replenishment method, apparatus, electronic device, and readable storage medium. Background Technology
[0002] With the increasing popularity of new energy vehicles, users' anxiety about charging has become increasingly prominent. Current technologies typically provide charging reminders based on the vehicle's current State of Charge (SOC) or remaining range, and can combine this with in-vehicle navigation to search for and navigate to nearby charging stations when the SOC or range is low. However, this approach only considers a single parameter such as the vehicle's SOC or remaining range, and only searches for nearby charging stations based on the current location. This can easily lead to problems such as recommended charging stations not meeting user expectations (e.g., recommended charging stations are not on the user's actual navigation route or even in the opposite direction), or being unavailable in real-world scenarios (e.g., long queues at charging stations during holidays). Therefore, existing vehicle charging methods cannot make optimal charging decisions for different scenarios, resulting in no fundamental improvement in car owners' charging anxiety. Summary of the Invention
[0003] This application provides a vehicle charging method, device, electronic device, and readable storage medium to solve the problem that existing vehicle charging methods cannot make optimal charging decisions for different scenarios, resulting in the failure to fundamentally alleviate vehicle owners' charging anxiety.
[0004] In a first aspect, this application provides a vehicle refueling method, the method comprising: Obtain the multimodal perception parameters corresponding to the vehicle; The pre-trained scene decision tree model is used to make scene decisions on the multimodal perception parameters to obtain the current scene category and scene feature vector of the vehicle. A multi-dimensional energy replenishment strategy is obtained by performing strategy analysis on the multimodal perception parameters, the scene category, and the scene feature vector. The multi-dimensional energy replenishment strategy is output so that the vehicle owner can replenish the vehicle's energy based on the multi-dimensional energy replenishment strategy.
[0005] Optionally, the scene decision tree model includes a first-level decision node, a second-level decision node, and a third-level decision node, and the multimodal perception parameters include vehicle owner profile parameters, vehicle status parameters, environmental parameters, and vehicle owner interaction parameters. The step of using a pre-trained scene decision tree model to make scene decisions on the multimodal perception parameters to obtain the current scene category and scene feature vector of the vehicle includes: The vehicle owner profile parameters and the vehicle owner interaction parameters are input into the first-level decision node for scenario decision-making to obtain a first decision result. The first decision result is used to characterize the current scenario category of the vehicle. The scenario category is any one of commuting scenario category, long-distance scenario category, and general scenario category. The owner profile parameters, vehicle status parameters, environmental parameters and / or owner interaction parameters are input into the second-level decision node for scenario decision-making to obtain a second decision result, wherein the second decision result is used to characterize whether the vehicle currently has a need for recharging; If the current scenario category of the vehicle is the commuting scenario category and the vehicle currently has a need for recharging, the owner profile parameters are input into the third-level decision node for scenario decision-making to obtain a third decision result, wherein the third decision result is used to characterize whether the vehicle currently supports home charging; Based on a preset feature extraction strategy, features are extracted from the multimodal perception parameters, the first decision result, the second decision result, and / or the third decision result to obtain the scene feature vector.
[0006] Optionally, the step of inputting the vehicle owner profile parameters and the vehicle owner interaction parameters into the first-level decision node for scenario decision-making to obtain a first decision result includes: The vehicle owner profile parameters and the vehicle owner interaction parameters are input into the first-level decision node to determine whether the vehicle's navigation route is a commuting route and whether the vehicle's navigation distance is greater than a preset distance. The vehicle's navigation route and navigation distance belong to the vehicle owner interaction parameters, and the commuting route belongs to the vehicle owner profile parameters. If the navigation route of the vehicle is determined to be the commuting route, then the scene category is determined to be the commuting scene category; If the navigation distance of the vehicle is determined to be greater than the preset distance, the scenario category is determined to be the long-distance scenario category; If it is determined that the vehicle's navigation route is not the commuting route, and the vehicle's navigation distance is less than or equal to the preset distance, then the scene category is determined to be the general scene category.
[0007] Optionally, the step of inputting the vehicle owner profile parameters, the vehicle status parameters, the environmental parameters, and / or the vehicle owner interaction parameters into the second-level decision node for scenario decision-making to obtain a second decision result includes: If the current scenario category of the vehicle is determined to be the commuting scenario category, the vehicle status parameters and the driver interaction parameters are input into the second-level decision node. The second-level decision node is used to determine whether the target ratio is greater than a preset threshold, and based on the determination result, it is determined whether the vehicle currently has a need for refueling. The target ratio refers to the ratio between the vehicle's driving range and the distance between the vehicle and the destination. The vehicle's driving range belongs to the vehicle status parameters, and the distance between the vehicle and the destination belongs to the driver interaction parameters. If the current scenario category of the vehicle is determined to be the long-distance scenario category, the vehicle status parameters, the environmental parameters, and the driver interaction parameters are input into the second-level decision node. The second-level decision node is used to determine whether the actual reachable mileage of the vehicle is less than the navigation distance of the vehicle. Based on the determination result, it is determined whether the vehicle currently has a need for refueling. The actual reachable mileage of the vehicle is determined based on the vehicle status parameters and the environmental parameters, and the navigation distance of the vehicle is included in the driver interaction parameters. If the current scenario category of the vehicle is determined to be the general scenario category, the owner profile parameters and the vehicle status parameters are input into the second-level decision node. The second-level decision node is used to determine whether the current charge status of the vehicle is greater than the charge status of the owner's preferred starting charge. Based on the determination result, it is determined whether the vehicle currently has a charging need. The current charge status of the vehicle belongs to the vehicle status parameters, and the charge status of the owner's preferred starting charge belongs to the owner profile parameters.
[0008] Optionally, the actual achievable mileage of the vehicle is calculated using the following formula: E_actual = E_rated × (1 - α × T_congestion - β × ΔT_temp - γ ×P_AC - δ × P_heating); Wherein, E_actual represents the actual achievable range of the vehicle, E_rated represents the nominal driving range of the vehicle, T_congestion represents the route congestion coefficient, which is determined based on the road congestion level in the environmental parameters, ΔT_temp represents the difference between the external ambient temperature and the preset optimal battery operating temperature in the environmental parameters, P_heating represents the battery heating power coefficient, which is determined based on the battery temperature in the vehicle status parameters, P_AC represents the air conditioning power coefficient, which is determined based on the air conditioning on / off status in the vehicle status parameters, and α, β, γ and δ represent weighting coefficients.
[0009] Optionally, the step of performing strategy analysis on the multimodal perception parameters, the scene category, and the scene feature vector to obtain a multidimensional energy replenishment strategy includes: The multimodal perception parameters, the scene category, and the scene feature vector are input into a pre-trained reinforcement learning model to perform policy analysis and output the multidimensional charging strategy. The multidimensional charging strategy includes a charging warning strategy, a charging route strategy, and a control strategy. The charging warning strategy is used to characterize whether a charging warning needs to be generated and when to generate it. The charging route strategy is used to characterize the route taken by the vehicle to the target charging station. The control strategy includes at least one of the following: a battery heating control strategy for the vehicle, an energy recovery strategy for the vehicle, and a home charging station reservation strategy for the vehicle.
[0010] Optionally, after outputting the multi-dimensional energy replenishment strategy for the vehicle owner to replenish the vehicle's energy based on the multi-dimensional energy replenishment strategy, the method further includes: Receive feedback data from vehicle owners during the vehicle recharging process; The model parameters of the car owner profile construction model are fine-tuned based on the feedback data, wherein the car owner profile construction model is used to output the car owner profile parameters.
[0011] Secondly, this application provides a vehicle refueling device, the device comprising: The acquisition module is used to acquire the multimodal perception parameters corresponding to the vehicle. The scene decision module is used to make scene decisions on the multimodal perception parameters using a pre-trained scene decision tree model, so as to obtain the current scene category and scene feature vector of the vehicle. The strategy analysis module is used to perform strategy analysis on the multimodal perception parameters, the scene category, and the scene feature vector to obtain a multidimensional energy replenishment strategy. The output module is used to output the multi-dimensional energy replenishment strategy so that the vehicle owner can replenish the vehicle based on the multi-dimensional energy replenishment strategy.
[0012] Thirdly, this application provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; The processor, when executing a program stored in memory, implements the vehicle refueling method described in the first aspect.
[0013] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions for performing the vehicle refueling method described in the first aspect.
[0014] The beneficial effects of this application are: (1) This application uses a scenario decision tree model to determine the current scenario category and scenario feature vector of the vehicle, and uses the scenario category, scenario feature vector and the corresponding multimodal perception parameters of the vehicle to participate in the vehicle's energy replenishment decision, thereby improving the comprehensiveness of the vehicle's energy replenishment decision, enabling the vehicle to make the optimal energy replenishment decision for different scenarios, and thus fundamentally improving the owner's energy replenishment anxiety.
[0015] (2) After performing strategy analysis on multimodal perception parameters, scene categories and scene feature vectors, this application can obtain a multi-dimensional energy replenishment strategy, rather than just a single strategy such as recommending navigation routes for charging stations, thereby improving the intelligence level of vehicle energy replenishment. Attached Figure Description
[0016] Figure 1 A schematic flowchart illustrating a vehicle refueling method provided in an embodiment of this application; Figure 2 This is a schematic diagram of a vehicle refueling system provided in an embodiment of this application; Figure 3 This is a schematic diagram illustrating a process for making scene decisions based on multimodal perception parameters using a scene decision tree model, as provided in an embodiment of this application. Figure 4 This is a schematic diagram of the structure of a vehicle refueling device provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0017] The embodiments of this application will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be understood that the preferred embodiments are only for illustrating this application and are not intended to limit the scope of protection of this application.
[0018] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.
[0019] For ease of description, spatial relative terms may be used in the text to describe the relative position or movement of one element or feature relative to another element or feature, as shown in the figure. These relative terms include, for example, "inside," "outside," "middle," "outer," "below," "below," "above," "front," "back," etc. Such spatial relative terms are intended to include different orientations of the device in use or operation, other than those depicted in the figure. For example, if the device in the figure undergoes a positional flip, orientation change, or change of motion, these directional indications will change accordingly. For instance, an element described as "below other elements or features" or "below other elements or features" will subsequently be oriented "above other elements or features" or "above other elements or features." Therefore, the example term "below" can include both upper and lower orientations. The device may be otherwise oriented (rotated 90 degrees or in other directions), and the spatial relative descriptors used in the text will be interpreted accordingly.
[0020] To address the problem that existing vehicle charging methods cannot make optimal charging decisions for different scenarios, thus failing to fundamentally alleviate vehicle owners' charging anxiety, this application provides a vehicle charging method, device, electronic device, and readable storage medium that improves the comprehensiveness of vehicle charging decisions, enabling vehicles to make optimal charging decisions for different scenarios.
[0021] See Figure 1 , Figure 1 This is a schematic flowchart illustrating a vehicle refueling method provided in an embodiment of this application. Figure 1 As shown, the vehicle refueling method may include the following steps: Step S102: Obtain the multimodal perception parameters corresponding to the vehicle.
[0022] Specifically, the aforementioned multimodal perception parameters may include, but are not limited to, owner profile parameters, vehicle status parameters, environmental parameters, and owner interaction parameters. These multimodal perception parameters can be obtained through the vehicle's Controller Area Network (CAN) bus, onboard sensors, vehicle-mounted interface, cloud server, and third-party data service interfaces. By acquiring the vehicle's corresponding multimodal perception parameters, this application overcomes the limitations of traditional solutions that rely solely on a single SOC parameter, providing a comprehensive and accurate information foundation for subsequent energy replenishment decisions that align with the owner's actual needs and adapt to complex environments.
[0023] As an optional implementation, the vehicle owner profile parameters may include, but are not limited to, the vehicle owner's identification number (ID), price sensitivity (high / medium / low), preferred charging type (fast charging / slow charging), commuting-home address, commuting-work address, availability of a home charger, home charger power (7.0kW / 11.0kW / 21.0kW, etc.), and the vehicle owner's initial SOC preference. The price sensitivity parameter influences the price priority ranking of recommended charging stations and the selection of off-peak electricity strategies for home charger reservations in subsequent decisions. The preferred charging type parameter influences the decision-making bias between recommending fast or slow charging stations in subsequent decisions; the availability and power of a home charger are used to determine the energy replenishment strategy in commuting scenarios in subsequent decisions; and the vehicle owner's initial SOC preference parameter is used to determine the timing of reminding the vehicle owner to charge in subsequent decisions.
[0024] The vehicle status parameters may include, but are not limited to, the Vehicle Identification Number (VIN), current battery SOC, total battery capacity, battery temperature, estimated driving range, operating conditions, voltage platform (400V / 800V, etc.), maximum charging power, power type, battery type, remaining fuel level, cabin temperature, tire pressure, whether the charging gun is plugged in, charging status (not charging / slow charging / fast charging / charging completed, etc.), remaining charging time, whether discharging, and energy recovery. These vehicle status parameters can be collected in real time via the vehicle's CAN bus.
[0025] These environmental parameters may include, but are not limited to, vehicle location, weather (rain / snow / fog, etc.), outside temperature, time, whether it is a holiday or peak hour, route congestion, a list of nearby charging stations, and information about each charging station (including charging station location, charging price, number of occupied / idle charging piles, voltage compatibility, etc.). These environmental parameters are obtained through the vehicle's Global Positioning System (GPS), online weather services, map navigation services, and the charging operation platform's Application Programming Interface (API).
[0026] These driver interaction parameters may include, but are not limited to, navigation destination and estimated duration of stay at the destination. These parameters can be obtained in real time through the vehicle's navigation system and the driver's user interface.
[0027] Step S104: Use the pre-trained scene decision tree model to make scene decisions on the multimodal perception parameters to obtain the current scene category and scene feature vector of the vehicle.
[0028] Specifically, the aforementioned scene decision tree model can be a multi-level binary decision tree structure or other multi-level structure, which may include multi-level decision nodes. This scene decision tree model is used to make scene decisions based on multi-modal perception parameters, thereby obtaining the vehicle's current scene category and scene feature vector. The scene category here may include, but is not limited to, commuting scene category, long-distance scene category, and general scene category. The scene feature vector here is used to characterize the vehicle's key features in the corresponding scene.
[0029] For example, when the scene decision tree model receives parameters such as "navigation destination is home", "home charging available", and "SOC is 25%", it performs logical judgment and outputs the scene category as commuting scene. The corresponding scene feature vector can be represented as V_scene= [25, 80, 12, 7.0, 3], where each dimension represents: current SOC (25%), target SOC (80%), expected stay duration (12 hours), home charging power (7.0kW), and price sensitivity level (high, encoded as 3).
[0030] In this way, complex multimodal perception parameters can be efficiently summarized into specific and understandable scene categories through the scene decision tree model, achieving a balance between computational efficiency and interpretability of decision logic. The output scene categories and scene feature vectors effectively reduce the dimensionality and refine the original data, pointing the way for subsequent refined decisions and reducing the complexity of decision-making and computational resource consumption.
[0031] Step S106: Perform strategy analysis on multimodal perception parameters, scene categories, and scene feature vectors to obtain multidimensional energy replenishment strategies.
[0032] Specifically, the aforementioned multi-dimensional energy replenishment strategies may include, but are not limited to, energy replenishment warning strategies, energy replenishment route strategies, and control strategies. Energy replenishment warning strategies may include low battery warning strategies and home charging station reminder strategies. Energy replenishment route strategies may include strategies for recommending convenient energy replenishment stations along the way, long-distance energy replenishment route planning strategies, and integrated energy replenishment and dining route recommendation strategies that combine long-distance travel with meal times. Control strategies may include vehicle battery heating control strategies, vehicle energy recovery strategies, and vehicle home charging station reservation strategies.
[0033] When performing policy analysis on multimodal perception parameters, scene categories, and scene feature vectors, pre-trained reinforcement learning models or other neural network models can be used; this application does not impose specific limitations on the embodiments. Through in-depth policy analysis, a multi-layered and comprehensive charging solution is generated. It not only tells car owners "whether they need to charge" and "where to charge," but also covers better strategies such as "when to charge," "how to charge more efficiently," and even "how to integrate with service area living scenarios," greatly improving the intelligence level and practical value of charging decision-making.
[0034] Step S108: Output the multi-dimensional energy replenishment strategy so that car owners can replenish the vehicle's energy based on the multi-dimensional energy replenishment strategy.
[0035] Specifically, after obtaining the multi-dimensional energy replenishment strategy, the strategy can be output, such as broadcasting low battery warnings, displaying recommended routes for energy replenishment stations, and controlling the vehicle battery to heat up. In this way, car owners can replenish their vehicles based on the multi-dimensional energy replenishment strategy.
[0036] In this way, by using a scene decision tree model to determine the vehicle's current scene category and scene feature vector, and incorporating this scene category, scene feature vector, and the vehicle's corresponding multimodal perception parameters into the vehicle's refueling decision, the comprehensiveness of the vehicle's refueling decision is improved. This allows the vehicle to make optimal refueling decisions for different scenarios, thereby fundamentally alleviating the owner's refueling anxiety. Furthermore, after strategy analysis of the multimodal perception parameters, scene category, and scene feature vector, a multi-dimensional refueling strategy can be obtained, rather than a single strategy such as recommending navigation routes to charging stations, thus improving the intelligence level of vehicle refueling.
[0037] In one optional embodiment, the scene decision tree model includes a first-level decision node, a second-level decision node, and a third-level decision node, and the multimodal perception parameters include vehicle owner profile parameters, vehicle status parameters, environmental parameters, and vehicle owner interaction parameters. Step S104 above involves using a pre-trained scene decision tree model to make scene decisions on the multimodal perception parameters, obtaining the vehicle's current scene category and scene feature vector, including: The vehicle owner profile parameters and vehicle owner interaction parameters are input into the first-level decision node to make a scenario decision and obtain the first decision result. The first decision result is used to represent the current scenario category of the vehicle. The scenario category can be any one of the following: commuting scenario category, long-distance scenario category, and general scenario category. The owner profile parameters, vehicle status parameters, environmental parameters and / or owner interaction parameters are input into the second-level decision node to make a scenario decision and obtain a second decision result. The second decision result is used to characterize whether the vehicle currently has a need for recharging. If the current scenario category of the vehicle is commuting and the vehicle currently has a need for recharging, the owner profile parameters are input into the third-level decision node for scenario decision-making to obtain the third decision result. The third decision result is used to characterize whether the vehicle currently supports home charging. Based on a preset feature extraction strategy, features are extracted from multimodal perception parameters, first decision results, second decision results, and / or third decision results to obtain scene feature vectors.
[0038] Specifically, when making scenario decisions based on multimodal perception parameters, the vehicle owner profile parameters and vehicle owner interaction parameters can be input into the first-level decision node to obtain the first decision result, which is the current scenario category of the vehicle (such as commuting scenario, long-distance scenario, or general scenario). Next, the vehicle owner profile parameters, vehicle status parameters, environmental parameters, and / or vehicle owner interaction parameters can be input into the second-level decision node to obtain the second decision result, which is whether the vehicle currently has a charging need. If the current scenario category of the vehicle is not a commuting scenario, and / or the vehicle currently has no charging need, the process ends and proceeds to the next-level decision node for scenario decision. If the current scenario category of the vehicle is a commuting scenario, and the vehicle currently has a charging need, the vehicle owner profile parameters can be input into the third-level decision node to obtain the third decision result, which is whether the vehicle currently supports home charging. Once all scenario decision-making processes are completed, features can be extracted from the multimodal perception parameters, the first decision result, the second decision result, and / or the third decision result based on a preset feature extraction strategy to obtain the scenario feature vector of the vehicle in the corresponding scenario, which can serve as an important basis for subsequent energy replenishment strategy analysis.
[0039] In this way, through the first-level decision nodes, second-level decision nodes, and third-level decision nodes in the scenario decision tree model, the step-by-step decision-making of "scenario classification → energy replenishment demand determination → home charging condition determination" can be realized, thereby achieving the classification of complex multimodal perception parameters and scenario understanding.
[0040] In an optional embodiment, the above steps, including inputting the vehicle owner profile parameters and vehicle owner interaction parameters into the first-level decision node for scenario decision-making to obtain a first decision result, include: The vehicle owner profile parameters and vehicle owner interaction parameters are input into the first-level decision node to determine whether the vehicle's navigation route is a commuting route and whether the vehicle's navigation distance is greater than the preset distance. The vehicle's navigation route and vehicle's navigation distance are vehicle owner interaction parameters, and the commuting route is a vehicle owner profile parameter. If the vehicle's navigation route is determined to be a commuter route, the scene category is determined to be a commuter scene category; If the navigation distance of the vehicle is determined to be greater than the preset distance, the scenario category is determined to be a long-distance scenario. If the vehicle's navigation route is determined to be a non-commuter route and the vehicle's navigation distance is less than or equal to a preset distance, the scenario category is determined to be a general scenario category.
[0041] Specifically, the aforementioned owner interaction parameters include at least the current vehicle's navigation route information and navigation distance information. The navigation route information refers to the navigation destination set by the owner through the vehicle's navigation system and the corresponding planned route data; the navigation distance information represents the total mileage from the vehicle's current location to the navigation destination along the planned route, in kilometers (km). The aforementioned owner profile parameters include at least commuting route information. This commuting route information specifically refers to the address data pre-entered and marked as "home" and "work" by the owner in the vehicle system, along with their corresponding navigation routes. This information serves as a digital representation of the owner's frequent daily travel patterns, assisting the system in determining whether the current navigation behavior belongs to the owner's regular commuting activities. The core judgment logic of the aforementioned first-level decision node is: using the navigation destination and navigation distance as classification criteria to determine the current scenario category of the vehicle.
[0042] When making scenario decisions using the first-level decision node, it can first determine whether the vehicle's navigation route is a commuter route and whether the vehicle's navigation distance is greater than a preset distance. If the vehicle's navigation route is a commuter route, the scenario category can be determined as a commuter scenario category; if the vehicle's navigation distance is greater than the preset distance, the scenario category can be determined as a long-distance scenario category; if the vehicle's navigation route is not a commuter route and the vehicle's navigation distance is less than or equal to the preset distance, the scenario category can be determined as a general scenario category.
[0043] For example, after get off work, a car owner enters the destination "XX City, XX Road, XX Community" into the vehicle's navigation system. The owner's profile parameters record "Home Address" as "XX City, XX Road, XX Community" and "Company Address" as "XX City, YY Road, YY Building". The first-level decision node matches the navigation destination with the home address, confirming they match. Therefore, it determines the navigation route is a commuting route, and the first decision result is a commuting scenario. If the car owner enters the navigation destination "XX City, YY Road, YY Building" on Monday morning, the first-level decision node again recognizes the match between the navigation destination and the company address, similarly classifying it as a commuting scenario.
[0044] For example, a car owner plans to drive from city A to city B, and the navigation distance shows 350 kilometers. In this case, the first-level decision node first determines whether the navigation destination is a commuter route (assuming the commuter route address is the home and office address in city A, which does not match city B), and determines it to be a non-commuter route. Then, the first-level decision node can compare the navigation distance of 350 kilometers with a preset distance threshold of 100 kilometers. Since 350 > 100, the scenario category is determined to be a long-distance scenario.
[0045] For example, if a car owner plans to go shopping at a mall 25 kilometers from their home, the first-level decision node first determines whether the navigation destination is a commuting route (assuming the commuting route addresses are home and office, and the mall address does not match either), classifying it as a non-commuting route. The first-level decision node then compares the navigation distance of 25 kilometers with a preset distance threshold of 100 kilometers. Since 25 < 100, the scenario category is determined to be a general scenario category.
[0046] In this way, the first-level decision node can accurately determine the current scenario category of the vehicle, which will facilitate the subsequent second-level decision node to determine the energy replenishment demand based on the scenario category.
[0047] In an optional embodiment, the above steps, including inputting the vehicle owner profile parameters, vehicle status parameters, environmental parameters, and / or vehicle owner interaction parameters into the second-level decision node for scenario decision-making to obtain a second decision result, include: If the current scenario of the vehicle is determined to be a commuting scenario, the vehicle status parameters and the owner interaction parameters are input into the second-level decision node. The second-level decision node is used to determine whether the target ratio is greater than a preset threshold. Based on the judgment result, it is determined whether the vehicle currently has a need for refueling. The target ratio is the ratio between the vehicle's range and the distance between the vehicle and the destination. The vehicle's range is a vehicle status parameter, and the distance between the vehicle and the destination is a owner interaction parameter. If the current scenario category of the vehicle is determined to be a long-distance scenario, the vehicle status parameters, environmental parameters, and owner interaction parameters are input into the second-level decision node. The second-level decision node is used to determine whether the actual reachable mileage of the vehicle is less than the navigation distance of the vehicle. Based on the judgment result, it is determined whether the vehicle currently has a need for recharging. The actual reachable mileage of the vehicle is determined based on the vehicle status parameters and environmental parameters, while the navigation distance of the vehicle is a parameter of the owner interaction. If the current scenario category of the vehicle is determined to be a general scenario category, the owner profile parameters and vehicle status parameters are input into the second-level decision node. The second-level decision node is used to determine whether the current charge status of the vehicle is greater than the charge status of the owner's preferred starting charge. Based on the judgment result, it is determined whether the vehicle currently has a need for charging. The current charge status of the vehicle belongs to the vehicle status parameters, and the charge status of the owner's preferred starting charge belongs to the owner profile parameters.
[0048] Specifically, when the first-level decision node determines that the vehicle's current scenario category is a commuting scenario, the second-level decision node obtains vehicle status parameters and driver interaction parameters as input to perform a refueling demand determination. Specifically, the second-level decision node reads the vehicle's remaining range from the vehicle status parameters and the distance from the vehicle to the destination from the driver interaction parameters. The second-level decision node calculates the ratio R between the two, i.e., target ratio R = vehicle's remaining range / vehicle's distance to the destination. This target ratio R represents the "safe margin" that the vehicle, in its current energy state and measured by its current range, can cover the remaining journey. Subsequently, the second-level decision node compares the target ratio R with a preset threshold. This preset threshold is a pre-set empirical value used to determine whether the vehicle faces a refueling risk. As an optional embodiment of the present invention, the preset threshold ranges from 1.5 to 2.0, meaning that the vehicle's remaining range is considered "safe" if it is at least 1.5 to 2.0 times the remaining journey distance. This value takes into account energy consumption fluctuations during actual driving (such as actual energy consumption exceeding nominal energy consumption due to factors like traffic congestion, air conditioning use, and changes in driving style), ensuring sufficient margin for the vehicle to reach its destination. Based on the comparison results, the second-level decision node determines whether the vehicle currently needs refueling. If R > the preset safety threshold, it indicates that the vehicle's remaining range is far greater than the distance required to reach the destination, with sufficient energy reserves and no risk of running out of power midway. In this case, the second-level decision node determines that the vehicle currently does not need refueling. If R ≤ the preset safety threshold, it indicates that the vehicle's remaining range is close to or insufficient to safely cover the remaining journey, posing a risk of running out of power midway. In this case, the second-level decision node determines that the vehicle currently needs refueling.
[0049] Once the first-level decision node determines that the vehicle's current scenario category is a long-distance scenario, the second-level decision node takes vehicle status parameters, environmental parameters, and owner interaction parameters as input to perform a refueling demand determination. Unlike commuting scenarios, energy consumption in long-distance scenarios is significantly affected by external environmental factors (such as temperature and road conditions) and vehicle equipment usage (such as air conditioning and battery heating), thus requiring a more refined energy consumption prediction model. Specifically, the second-level decision node first calculates the vehicle's actual achievable range E_actual based on vehicle status parameters and environmental parameters. This actual achievable range refers to the maximum distance the vehicle can actually travel under the current external environment and vehicle status, taking into account various energy consumption influencing factors; it differs from the nominal estimated range displayed on the vehicle's dashboard. After calculating the actual reachable mileage E_actual, the second-level decision node compares it with the vehicle's navigation distance (obtained from the owner's interaction parameters): If E_actual ≥ navigation distance, it indicates that even after considering factors affecting energy consumption such as congestion, low temperature, and air conditioning, the vehicle's actual reachable mileage can still cover the entire navigation journey, and there is no risk of running out of power midway. In this case, the second-level decision node determines that the vehicle currently has no need for refueling. If E_actual < navigation distance, it indicates that under actual driving conditions, the vehicle's range is insufficient to cover the entire journey, and refueling will inevitably be required along the way. In this case, the second-level decision node determines that the vehicle currently has a need for refueling.
[0050] If the first-level decision node determines that the current scenario category of the vehicle is a general scenario category, the second-level decision node obtains the owner profile parameters and vehicle status parameters as input to perform a power replenishment demand determination.
[0051] Specifically, the second-level decision node reads the vehicle's current charge state (SOC_current) from the vehicle status parameters and the owner's preferred initial charging charge state (SOC_pref) from the owner profile parameters. Here, the owner's preferred initial charging charge state refers to the SOC value that the owner habitually or expects to start charging when the vehicle's battery level drops to a specific level, reflecting the owner's personal charging habits and risk preferences. The second-level decision node compares the vehicle's current charge state (SOC_current) with the owner's preferred initial charging charge state (SOC_pref). If SOC_current > SOC_pref, it indicates that the vehicle's current battery level is higher than the owner's charging trigger preference, and the owner subjectively believes that charging is not yet necessary. In this case, the second-level decision node determines that the vehicle currently has no charging requirement. If SOC_current ≤ SOC_pref, it indicates that the vehicle's current battery level has reached or fallen below the owner's charging trigger preference, and the second-level decision node determines that the vehicle currently has a charging requirement.
[0052] By implementing adaptive charging demand determination logic in different scenario categories through the second-level decision node, the following beneficial effects were achieved: First, in commuting scenarios, mechanical low-battery disturbances were avoided. When the driver is home from get off work and has enough remaining range to safely get home (even if the absolute SOC value is low), low-battery reminders will not be triggered, effectively reducing interference from invalid information and improving the continuity and comfort of the driving experience. Second, in long-distance scenarios, the urgent need for charging was accurately identified. By comprehensively considering multiple energy consumption factors such as congestion, low temperature, air conditioning, and battery heating, the accuracy of charging demand determination was significantly improved, effectively preventing the risk of running out of power during the journey due to inflated nominal range, and ensuring the safety of long-distance travel. Third, in general scenarios, personalized triggering was achieved based on driver preferences, realizing personalized charging reminder timing for each individual and meeting the differentiated needs of different drivers for sufficient battery reserve.
[0053] In one optional embodiment, the actual achievable mileage of the vehicle is calculated using the following formula: E_actual = E_rated × (1 - α × T_congestion - β × ΔT_temp - γ ×P_AC - δ × P_heating); Wherein, E_actual represents the vehicle's actual achievable range, E_rated represents the vehicle's nominal range, T_congestion represents the route congestion coefficient, which is determined based on the road congestion level in the environmental parameters, ΔT_temp represents the difference between the external ambient temperature and the preset optimal battery operating temperature in the environmental parameters, P_heating represents the battery heating power coefficient, which is determined based on the battery temperature in the vehicle status parameters, P_AC represents the air conditioning power coefficient, which is determined based on the air conditioning on / off status in the vehicle status parameters, and α, β, γ, and δ represent weighting coefficients.
[0054] Specifically, the vehicle's nominal driving range E_rated is the estimated driving range displayed on the vehicle's instrument panel, in km. The route congestion coefficient T_congestion can range from 0 to 1, and can be determined based on real-time traffic conditions along the navigation route. For example, 0.8 for severe congestion, 0.5 for moderate congestion, 0.2 for light congestion, and 0 for smooth traffic. When the ambient temperature is lower than the preset optimal battery operating temperature (e.g., 25℃), the difference ΔT_temp between the ambient temperature and the preset optimal battery operating temperature (e.g., 25℃) is positive, in °C; when the ambient temperature is higher than the preset optimal battery operating temperature (e.g., 25℃), the difference ΔT_temp is negative, in °C. When the cabin air conditioning is on, the air conditioning power coefficient P_AC is 1; otherwise, it is 0. When the battery temperature is below a preset threshold (e.g., 10°C), the battery heating power coefficient P_heating is set to 1; otherwise, the battery heating power coefficient P_heating is set to 0. The weighting coefficients α, β, γ, and δ can be determined through regression analysis of historical driving data. As an optional implementation, α = 0.3, β = 0.02, γ = 0.15, and δ = 0.10.
[0055] The above formula reduces the nominal driving range E_rated by multiple penalty terms (congestion, low temperature, air conditioning, battery heating) to more accurately approximate the vehicle's actual achievable range under real-world driving conditions. Of course, as another alternative implementation, the above formula can be simply modified to calculate the vehicle's actual achievable range, such as ignoring the effects of ΔT_temp or P_heating.
[0056] In this way, the above formula can be used to comprehensively consider multiple energy consumption factors such as congestion, low temperature, air conditioning, and battery heating, significantly improving the accuracy of energy replenishment demand determination, effectively preventing the risk of running out of power during the journey due to an inflated nominal range, and ensuring the safety of long-distance travel.
[0057] In an optional embodiment, step S106 above, which involves performing strategy analysis on multimodal perception parameters, scene categories, and scene feature vectors to obtain a multidimensional energy replenishment strategy, includes: Multimodal perception parameters, scene categories, and scene feature vectors are input into a pre-trained reinforcement learning model to perform policy analysis and output a multidimensional power replenishment strategy. The multidimensional power replenishment strategy includes a power replenishment warning strategy, a power replenishment route strategy, and a control strategy. The power replenishment warning strategy is used to characterize whether a power replenishment warning needs to be generated and when to generate it. The power replenishment route strategy is used to characterize the route taken by the vehicle to the target charging station. The control strategy includes at least one of the following: a battery heating control strategy for the vehicle, an energy recovery strategy for the vehicle, and a home charging station reservation strategy for the vehicle.
[0058] Specifically, the aforementioned reinforcement learning model is an intelligent decision-making model built using a deep reinforcement learning (DRL) architecture. Deep reinforcement learning is a machine learning paradigm that combines the perceptual capabilities of deep learning with the decision-making capabilities of reinforcement learning. It learns the optimal decision-making strategy through continuous interaction between the agent and the environment, maximizing rewards through trial and error. The training process of this model is as follows: 1. Simulation Environment Construction: Based on large-scale historical driving data, charging records, and vehicle owner behavior data, a high-fidelity simulation environment is constructed. This simulation environment can simulate the vehicle's driving process, energy consumption dynamics, charging station queuing status, and the vehicle owner's acceptance or rejection of various charging strategies under different scenarios (commuting, long-distance travel, emergency, destination stop, etc.). 2. Reward Function Design: A multi-dimensional reward function is designed to evaluate the quality of the charging strategies output by the model. The reward function includes at least the following dimensions: whether the car owner accepts the recommended strategy (positive reward), actual charging cost (negative reward), charging waiting time (negative reward), and whether the risk of running out of power is avoided (significant positive reward). The model continuously tries different decision-making strategies in the simulation environment and adjusts its neural network parameters based on the feedback from the reward function to maximize the cumulative reward. 3. Strategy Convergence: Through millions of simulation iterations, the model's decision-making strategy gradually converges to an optimal or near-optimal state, meaning it can generate strategy combinations that maximize car owner satisfaction and charging efficiency in various scenarios. After training, the model can be deployed on the vehicle's infotainment system or in the cloud, and during actual driving, it performs rapid inference based on real-time input parameters to output the optimal multi-dimensional charging strategy.
[0059] The aforementioned multi-dimensional energy replenishment strategy includes an energy replenishment warning strategy, an energy replenishment route strategy, and a control strategy. The energy replenishment warning strategy characterizes whether an energy replenishment warning needs to be generated and when it should be generated. The warning types output by the energy replenishment warning strategy include at least one or more of the following: low battery charging reminder and home charging station reservation reminder. The low battery charging reminder is applicable to scenarios where the vehicle is at risk of running out of battery power during the current journey; the home charging station reservation reminder is applicable to scenarios where the vehicle is commuting home and has access to home charging.
[0060] The charging route strategy is used to characterize the route planning that guides the vehicle to the target charging station. The output of this strategy is a navigation route plan or set of plans, specifically including: In commuting or general scenarios, if the vehicle needs to recharge en route, the model recommends charging stations along the current navigation route and plans a route for recharging along the way; in long-distance scenarios, the model plans a complete multi-recharging plan along the route, including the location of each recharge point, the recharge sequence, and the expected energy replenishment; in long-distance scenarios, the model combines the owner's meal times (e.g., lunch / dinner) to recommend charging stations and corresponding navigation routes that integrate with catering services, achieving seamless integration of charging and daily life scenarios. The control strategy is the active control command for the vehicle's underlying actuators. Its output is parameter setting or on / off commands that can be directly sent to the vehicle control system, specifically including: 1. When a decision is made to initiate fast charging in a long-distance scenario, the model determines the start time and target heating temperature of the battery heating based on the vehicle's current battery temperature, distance to the target charging station, and estimated arrival time, and sends the command to the Battery Management System (BMS) for execution. This strategy aims to ensure the battery arrives at the charging station in the optimal fast-charging temperature range, thereby shortening charging time. 2. The model determines whether to adjust the energy recovery intensity (e.g., from standard mode to strong mode) based on the vehicle's estimated range, road conditions, and driving mode, and sends a command to the Vehicle Control Unit (VCU) for execution. This strategy aims to increase range by increasing energy recovery during coasting and braking. 3. In the commuting-home charging scenario, the model calculates the required charging time and optimal charging period (e.g., off-peak hours at night) based on the owner's price sensitivity, home charging power, current SOC, and target SOC, and generates a setting command for reserving a home charging station. This command is then sent to the smart charging station via the vehicle's infotainment system or a mobile application (APP) for execution.
[0061] In this way, reinforcement learning models can be used to output multi-dimensional energy replenishment strategies, forming a complete service chain from information prompts and navigation guidance to vehicle execution, realizing a paradigm upgrade from passive response to proactive service.
[0062] In an optional embodiment, after step S108, where the multi-dimensional energy replenishment strategy is output for the vehicle owner to replenish the vehicle's energy based on the multi-dimensional energy replenishment strategy, the method further includes: Receive feedback data from vehicle owners during the vehicle refueling process; The model parameters of the car owner profile construction model are fine-tuned based on feedback data. The car owner profile construction model is used to output the car owner profile parameters.
[0063] Specifically, during the process of a vehicle owner replenishing the vehicle's power using a multi-dimensional replenishment strategy, the owner can choose to accept, modify, or reject the recommended replenishment strategy through the vehicle's infotainment system. These interactions are recorded as feedback data. If the owner accepts the recommended strategy, it is recorded as positive feedback, and the model receives a positive reward; if the owner rejects or modifies the recommended strategy, it is recorded as negative feedback, and the model receives a negative reward. The reward signal is used to adjust the model's reward function weights or directly for local incremental fine-tuning of the model.
[0064] Based on the long-term positive / negative feedback patterns of car owners, the model parameters of the car owner profile construction model can be fine-tuned so that the energy replenishment strategy output by the model gradually approaches the personalized optimal strategy of a specific car owner.
[0065] In this way, through continuous interaction with car owners' feedback, we can learn the deep preferences of specific car owners (such as their true trade-offs regarding price, time, and convenience), thereby generating personalized charging strategies that conform to individual habits, achieving precise services tailored to each individual, and significantly improving car owner satisfaction.
[0066] In one optional embodiment, the vehicle refueling method provided in this application can be applied to a vehicle refueling system. The structure of the vehicle refueling system is as follows: Figure 2 As shown, the vehicle refueling system comprises, from top to bottom, a perception layer, a decision-making and planning layer, and an execution output layer. The perception layer acquires four types of parameters: owner profile parameters, vehicle status parameters, environmental parameters, and owner interaction parameters. The decision-making and planning layer employs a two-layer heterogeneous decision architecture, including a scene decision tree model and a reinforcement learning model. The scene decision tree model performs scene recognition and classification based on the four types of parameters acquired by the perception layer, outputting scene categories and scene feature vectors. The hierarchical reinforcement learning model generates the optimal refueling strategy under the constraint of the scene category output by the scene decision tree model, based on the input parameters acquired by the perception layer and the scene feature vectors. The workflow of the vehicle refueling system is as follows: First, the perception layer acquires four types of input parameters; then, the decision-making and planning layer performs scene recognition and refueling strategy decision based on the acquired input parameters; finally, the execution output layer outputs a multi-dimensional complementary refueling strategy.
[0067] Specifically, the process of using a scene decision tree model to make scene decisions based on multimodal perception parameters is as follows: Figure 3 As shown, the specific steps include the following: Step S302: Obtain multimodal sensing parameters.
[0068] Step S304: Make a first-level decision based on the navigation route and navigation distance.
[0069] Step S306: Enter the commuting scenario branch.
[0070] Step S308: Determine whether the target ratio is greater than the preset threshold, i.e., make a second-level decision.
[0071] If the target ratio is greater than the preset threshold, proceed to step S310; if the target ratio is less than or equal to the preset threshold, proceed to step S312. Here, the target ratio refers to the ratio between the vehicle's driving range and the distance between the vehicle and its destination.
[0072] Step S310: Enter the Commute - Fully Charged Branch.
[0073] Step S312: Enter the Commute - Energy Replenishment Branch.
[0074] Step S314: Determine if there is a home charger available and whether the vehicle's SOC has reached the owner's preferred initial charging SOC, i.e., make a third-level decision.
[0075] If a home charger is available and the vehicle's SOC reaches the owner's preferred initial charging SOC, then proceed to step S316; if there is no home charger and / or the vehicle's SOC does not reach the owner's preferred initial charging SOC, then proceed to step S318.
[0076] Step S316: Generate a home charging reminder and make recommendations based on off-peak electricity prices.
[0077] Step S318: Do not issue a charging reminder.
[0078] Step S320: Generate a recommendation reminder for energy replenishment along the way.
[0079] Step S322: Enter the long-distance scene branch.
[0080] Step S324: Determine whether the vehicle's actual reachable mileage is less than the vehicle's navigation distance, i.e., make a second-level decision.
[0081] If the vehicle's actual reachable mileage is less than the vehicle's navigation distance, then proceed to step S326; if the vehicle's actual reachable mileage is greater than or equal to the vehicle's navigation distance, then proceed to step S328.
[0082] Step S326: Enter the long-distance route - energy replenishment required.
[0083] Step S328: Enter Long Distance - No Energy Replenishment Required Branch.
[0084] Step S330: Calculate the energy replenishment needs and number of times, and recommend charging stations based on the expected energy replenishment time and meal time.
[0085] Step S332: Calculate whether it is necessary to adjust the kinetic energy recovery and preheat the battery based on the distance to the charging station.
[0086] Step S334: Do not issue a charging reminder.
[0087] Step S336: Enter the general scenario branch.
[0088] Step S338: Determine whether the vehicle's current SOC is greater than the owner's preferred initial charging SOC, i.e., make a second-level decision.
[0089] If the vehicle's current SOC is greater than the owner's preferred initial charging SOC, then proceed to step S340; if the vehicle's current SOC is greater than or equal to the owner's preferred initial charging SOC, then proceed to step S342.
[0090] Step S340: Enter the General - No Energy Replenishment Branch.
[0091] Step S342: Enter the General - Must Replenish Energy branch.
[0092] Step S344: Generate a recommendation reminder for energy replenishment along the way.
[0093] Step S346: Generate scene feature vectors.
[0094] To better understand the technical solution of this application, examples of commuting and long-distance scenarios are provided below.
[0095] 1. The intelligent energy replenishment decision-making in commuting scenarios is as follows: Assuming it's 6:00 PM on a weekday (non-holiday), during peak hours, the vehicle's status is as follows: Owner ID: U10086, Vehicle VIN: LS6D3E136SA550149, Navigation destination: "Home" (XX residential area), Estimated parking time: 12 hours. Current SOC: 25%, Total Battery Capacity: 89kWh, Battery Temperature: 25℃, Rated Voltage: 400V, Maximum Charging Power: 80kW. Environmental conditions: External temperature: 16℃, Sunny weather, Light traffic congestion. Owner profile shows price sensitivity as "High" and preferred charging type as "Slow Charging". Vehicle status parameters show "Home Charger Available" is true, with a home charger power of 7.0kW.
[0096] The decision-making and planning layer first obtains three core parameters: "Current SOC 25%", "Remaining range 90km", "Navigation destination - home", and "Whether there is a home charging station". By querying map data, it calculates that the navigation distance from the current location to home is 15 kilometers, with an estimated energy consumption of 3kWh. Therefore, it can be determined that the current battery level (25% * 89kWh ≈ 22kWh) is much higher than the energy required to reach home.
[0097] Based on the above decision-making logic, when "commuting route = true", "home charging available = true", and "remaining range / distance to home > safety threshold (e.g., 1.5)", home charging is an option. Due to the car owner's "high price sensitivity", the following charging decision will occur: Warning: "Low battery charging reminder" will not be triggered.
[0098] Personalized Recommendation: Outputs "Home Charger Reservation Reminder". Considering the estimated stop time at the destination (12 hours) and the home charger power (7.0kW), the system calculates that charging from 25% to 80% will only take approximately (89kWh * 0.55) / 7.0kW ≈ 7 hours. Charging can be scheduled during off-peak electricity price periods.
[0099] Navigation route parameters: Do not change the existing navigation route.
[0100] Active control parameters: No intervention is performed. Meanwhile, since the destination is a known home, the system will maintain the energy recovery strategy in the owner's preferred "standard" mode to provide a more comfortable driving experience.
[0101] In this way, the vehicle's central control screen will not display a charging reminder. When the owner arrives at their home parking space, parks the car, and plugs in the charging gun, the vehicle's system or mobile app will prompt: "Home has been identified. Based on your electricity price preference, charging has been scheduled to start during off-peak hours at night (00:00-08:00 the next day), and is expected to reach 80% before 8:00 in the morning."
[0102] By employing the above methods, we can successfully avoid pushing invalid fast charging suggestions along the way when car owners already have a better charging solution (home charging), thus eliminating disturbances to car owners; and through intelligent scheduling, we can further meet the high price sensitivity needs of car owners and achieve optimal cost.
[0103] 2. The intelligent power replenishment decision-making in long-distance scenarios is as follows: The time is 2:30 PM on January 15, 2026 (the eve of the Spring Festival, a holiday). The owner is driving a pure electric vehicle with a ternary lithium battery and an 800V voltage platform. The navigation destination is "XX County, XX Community", a distance of 350 kilometers. The current SOC is 80%, and the estimated range is 400km (WLTC cycle). The outside temperature is -5℃ (winter), and the weather is light snow. The vehicle is currently on a highway, and the traffic congestion is "severe". The owner's profile shows a "moderate" price sensitivity but a preference for "fast charging". The vehicle's current "air conditioning on" setting is true, and the cabin temperature is set to 26℃.
[0104] The decision-making and planning layer received strong signals such as "holiday = true", "severe congestion", "low external temperature", and "long-distance navigation". Combined with "congested road conditions" and "air conditioning on", the actual range was recalculated and estimated. The model predicted that due to congested traffic and air conditioning energy consumption, the actual range would drop to about 280 kilometers, making it impossible to safely reach the destination 350 kilometers away.
[0105] At this point, the long-distance energy replenishment planning is initiated, and the decision-making process is as follows: Identify the urgent need for refueling: Calculations show that at least one refueling is required during the journey.
[0106] Multi-objective optimization: Combining "time 14:30", "estimated driving time 4-5 hours (including traffic jams)" and "owner's preference for fast charging", it is inferred that the owner will need to recharge around 18:30-19:30 (dinner time).
[0107] Active policy generation: Personalized Recommendation: Generates "Recommended Highway Charging Stations for Long-Distance Highway Trips Based on Meal Times". The system filters out a city charging station near a service area approximately 220 kilometers from the current location (estimated arrival time: 18:00). This station has many available charging piles (not a busy time during holidays), high charging power (compatible with 800V), and restaurants nearby.
[0108] Navigation route parameters: Create a new navigation route "Long-distance charging station subsidy navigation route" to guide the vehicle to exit at the designated highway exit and complete the "charging + dining" integration.
[0109] Active control parameters: When the distance to the planned charging station is about 30 kilometers, the system automatically issues a "battery heating" command, using the residual heat of the motor or the energy consumption of the battery itself to heat the battery temperature from the current -5℃ to the optimal fast charging temperature (such as 25℃).
[0110] The vehicle's central control screen then displays: "Current range may not be sufficient to reach your destination. The optimal charging plan has been planned for you: charging at the ZZ charging station near YY Restaurant in XX City, expected arrival time is 18:30, and charging will reach 80% in 20 minutes. Battery preheating has been activated for you, so you can enjoy the fastest charging speed upon arrival. Do you accept this plan?" After the owner confirms, the vehicle will execute the plan automatically.
[0111] By employing the methods described above, charging risks can be predicted in advance, and charging, dining, and battery preheating in winter can be creatively integrated into a seamless trip. Compared to traditional methods (which only recommend highway service areas, where drivers may face issues such as queuing, slow charging, and having to stop alone to eat), this invention significantly improves the overall efficiency and experience of long-distance travel.
[0112] See Figure 4 , Figure 4This is a schematic diagram of a vehicle refueling device provided in an embodiment of this application. Figure 4 As shown, the vehicle refueling device 400 may include: The acquisition module 402 is used to acquire the multimodal perception parameters corresponding to the vehicle; Scene decision module 404 is used to make scene decisions on multimodal perception parameters using a pre-trained scene decision tree model to obtain the current scene category and scene feature vector of the vehicle. The strategy analysis module 406 is used to perform strategy analysis on multimodal perception parameters, scene categories and scene feature vectors to obtain multidimensional energy replenishment strategies. Output module 408 is used to output the multi-dimensional energy replenishment strategy so that the vehicle owner can replenish the vehicle's energy based on the multi-dimensional energy replenishment strategy.
[0113] Furthermore, the scene decision tree model includes first-level decision nodes, second-level decision nodes, and third-level decision nodes; the multimodal perception parameters include owner profile parameters, vehicle status parameters, environmental parameters, and owner interaction parameters; the scene decision module 404 includes: The first input submodule is used to input the vehicle owner profile parameters and vehicle owner interaction parameters into the first-level decision node for scenario decision-making and obtain the first decision result. The first decision result is used to represent the current scenario category of the vehicle. The scenario category can be any one of the commuting scenario category, long-distance scenario category, and general scenario category. The second input submodule is used to input the vehicle owner profile parameters, vehicle status parameters, environmental parameters and / or vehicle owner interaction parameters into the second-level decision node for scenario decision-making and to obtain the second decision result. The second decision result is used to characterize whether the vehicle currently has a need for recharging. The third input submodule is used to input the owner profile parameters into the third-level decision node to make a scenario decision when the current scenario category of the vehicle is commuting and the vehicle currently has a need for recharging, and to obtain the third decision result. The third decision result is used to characterize whether the vehicle currently supports home charging. The feature extraction submodule is used to extract features from multimodal perception parameters, first decision results, second decision results and / or third decision results based on a preset feature extraction strategy, so as to obtain scene feature vectors.
[0114] Furthermore, the first input submodule includes: The judgment unit is used to input the vehicle owner profile parameters and vehicle owner interaction parameters into the first-level decision node, so as to use the first-level decision node to determine whether the vehicle's navigation route is a commuting route and whether the vehicle's navigation distance is greater than a preset distance. Here, the vehicle's navigation route and vehicle's navigation distance are vehicle owner interaction parameters, and the commuting route is a vehicle owner profile parameter. The first determining unit is used to determine the scene category as a commuting scene category when the vehicle's navigation route is determined to be a commuting route. The second determining unit is used to determine the scenario category as a long-distance scenario category when the navigation distance of the vehicle is determined to be greater than the preset distance. The third determining unit is used to determine the scene category as a general scene category when it is determined that the vehicle's navigation route is not a commuter route and the vehicle's navigation distance is less than or equal to a preset distance.
[0115] Furthermore, the second input submodule includes: The first input unit is used to input vehicle status parameters and driver interaction parameters into the second-level decision node when the current scenario category of the vehicle is determined to be a commuting scenario category. The second-level decision node is used to determine whether the target ratio is greater than a preset threshold and to determine whether the vehicle currently has a need for refueling based on the judgment result. The target ratio is the ratio between the vehicle's driving range and the distance between the vehicle and the destination. The vehicle's driving range is a vehicle status parameter and the distance between the vehicle and the destination is a driver interaction parameter. The second input unit is used to input vehicle status parameters, environmental parameters, and driver interaction parameters into the second-level decision node when the current scenario category of the vehicle is determined to be a long-distance scenario category. The second-level decision node is used to determine whether the actual reachable mileage of the vehicle is less than the navigation distance of the vehicle, and to determine whether the vehicle currently has a need for recharging based on the judgment result. The actual reachable mileage of the vehicle is determined based on the vehicle status parameters and environmental parameters, and the navigation distance of the vehicle is a driver interaction parameter. The third input unit is used to input the owner profile parameters and vehicle status parameters into the second-level decision node when the current scenario category of the vehicle is determined to be a general scenario category. The second-level decision node is used to determine whether the current charge state of the vehicle is greater than the charge state of the initial charging preferred by the owner, and to determine whether the vehicle currently has a need for recharging based on the judgment result. The current charge state of the vehicle is a vehicle status parameter, and the charge state of the initial charging preferred by the owner is a owner profile parameter.
[0116] Furthermore, the actual achievable mileage of the vehicle is calculated using the following formula: E_actual = E_rated × (1 - α × T_congestion - β × ΔT_temp - γ ×P_AC - δ × P_heating); Wherein, E_actual represents the vehicle's actual achievable range, E_rated represents the vehicle's nominal range, T_congestion represents the route congestion coefficient, which is determined based on the road congestion level in the environmental parameters, ΔT_temp represents the difference between the external ambient temperature and the preset optimal battery operating temperature in the environmental parameters, P_heating represents the battery heating power coefficient, which is determined based on the battery temperature in the vehicle status parameters, P_AC represents the air conditioning power coefficient, which is determined based on the air conditioning on / off status in the vehicle status parameters, and α, β, γ, and δ represent weighting coefficients.
[0117] Furthermore, the strategy analysis module 406 includes: The strategy analysis submodule is used to input multimodal perception parameters, scene categories, and scene feature vectors into a pre-trained reinforcement learning model to perform strategy analysis using the reinforcement learning model and output a multidimensional power replenishment strategy. The multidimensional power replenishment strategy includes a power replenishment warning strategy, a power replenishment route strategy, and a control strategy. The power replenishment warning strategy is used to characterize whether a power replenishment warning needs to be generated and when to generate the power replenishment warning. The power replenishment route strategy is used to characterize the route for the vehicle to navigate to the target charging station. The control strategy includes at least one of the following: a battery heating control strategy for the vehicle, an energy recovery strategy for the vehicle, and a home charging station reservation strategy for the vehicle.
[0118] Furthermore, the vehicle refueling device 400 also includes: The receiving module is used to receive feedback data from vehicle owners during the vehicle recharging process; The fine-tuning module is used to fine-tune the model parameters of the car owner profile building model based on feedback data. The car owner profile building model is used to output the car owner profile parameters.
[0119] It should be noted that the vehicle refueling device 400 can implement the vehicle refueling method provided in any of the aforementioned method embodiments and achieve the same technical effect, which will not be elaborated here.
[0120] like Figure 5 As shown, this application embodiment also provides an electronic device, including a processor 511, a communication interface 512, a memory 513 and a communication bus 514, wherein the processor 511, the communication interface 512 and the memory 513 communicate with each other through the communication bus 514. Memory 513 is used to store computer programs; In one embodiment of this application, the processor 511, when executing the program stored in the memory 513, implements the vehicle refueling method provided in any of the foregoing method embodiments.
[0121] In addition, embodiments of this application also provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the vehicle refueling method provided in any of the foregoing method embodiments.
[0122] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.
[0123] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, 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 ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0124] It should be understood that the terminology used herein is for the purpose of describing particular exemplary embodiments only and is not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “described” as used herein may also include the plural forms. The terms “comprising,” “including,” “containing,” and “having” are inclusive and therefore indicate the presence of the stated features, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, and / or combinations thereof. The method steps, processes, and operations described herein are not construed as requiring them to be performed in a particular order described or illustrated unless the order of performance is explicitly indicated. It should also be understood that additional or alternative steps may be used.
[0125] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.
Claims
1. A method for replenishing vehicle energy, characterized in that, The method includes: Obtain the multimodal perception parameters corresponding to the vehicle; The pre-trained scene decision tree model is used to make scene decisions on the multimodal perception parameters to obtain the current scene category and scene feature vector of the vehicle. A multi-dimensional energy replenishment strategy is obtained by performing strategy analysis on the multimodal perception parameters, the scene category, and the scene feature vector. The multi-dimensional energy replenishment strategy is output so that the vehicle owner can replenish the vehicle's energy based on the multi-dimensional energy replenishment strategy.
2. The method according to claim 1, characterized in that, The scenario decision tree model includes a first-level decision node, a second-level decision node, and a third-level decision node. The multimodal perception parameters include vehicle owner profile parameters, vehicle status parameters, environmental parameters, and vehicle owner interaction parameters. The step of using a pre-trained scene decision tree model to make scene decisions on the multimodal perception parameters to obtain the current scene category and scene feature vector of the vehicle includes: The vehicle owner profile parameters and the vehicle owner interaction parameters are input into the first-level decision node for scenario decision-making to obtain a first decision result. The first decision result is used to characterize the current scenario category of the vehicle. The scenario category is any one of commuting scenario category, long-distance scenario category, and general scenario category. The owner profile parameters, vehicle status parameters, environmental parameters and / or owner interaction parameters are input into the second-level decision node for scenario decision-making to obtain a second decision result, wherein the second decision result is used to characterize whether the vehicle currently has a need for recharging; If the current scenario category of the vehicle is the commuting scenario category and the vehicle currently has a need for recharging, the owner profile parameters are input into the third-level decision node for scenario decision-making to obtain a third decision result, wherein the third decision result is used to characterize whether the vehicle currently supports home charging; Based on a preset feature extraction strategy, features are extracted from the multimodal perception parameters, the first decision result, the second decision result, and / or the third decision result to obtain the scene feature vector.
3. The method according to claim 2, characterized in that, The step of inputting the vehicle owner profile parameters and the vehicle owner interaction parameters into the first-level decision node for scenario decision-making to obtain a first decision result includes: The vehicle owner profile parameters and the vehicle owner interaction parameters are input into the first-level decision node to determine whether the vehicle's navigation route is a commuting route and whether the vehicle's navigation distance is greater than a preset distance. The vehicle's navigation route and navigation distance belong to the vehicle owner interaction parameters, and the commuting route belongs to the vehicle owner profile parameters. If the navigation route of the vehicle is determined to be the commuting route, then the scene category is determined to be the commuting scene category; If the navigation distance of the vehicle is determined to be greater than the preset distance, the scenario category is determined to be the long-distance scenario category; If it is determined that the vehicle's navigation route is not the commuting route, and the vehicle's navigation distance is less than or equal to the preset distance, then the scene category is determined to be the general scene category.
4. The method according to claim 2, characterized in that, The step of inputting the vehicle owner profile parameters, vehicle status parameters, environmental parameters, and / or vehicle owner interaction parameters into the second-level decision node for scenario decision-making to obtain a second decision result includes: If the current scenario category of the vehicle is determined to be the commuting scenario category, the vehicle status parameters and the driver interaction parameters are input into the second-level decision node. The second-level decision node is used to determine whether the target ratio is greater than a preset threshold, and based on the determination result, it is determined whether the vehicle currently has a need for refueling. The target ratio refers to the ratio between the vehicle's driving range and the distance between the vehicle and the destination. The vehicle's driving range belongs to the vehicle status parameters, and the distance between the vehicle and the destination belongs to the driver interaction parameters. If the current scenario category of the vehicle is determined to be the long-distance scenario category, the vehicle status parameters, the environmental parameters, and the driver interaction parameters are input into the second-level decision node. The second-level decision node is used to determine whether the actual reachable mileage of the vehicle is less than the navigation distance of the vehicle. Based on the determination result, it is determined whether the vehicle currently has a need for refueling. The actual reachable mileage of the vehicle is determined based on the vehicle status parameters and the environmental parameters, and the navigation distance of the vehicle is included in the driver interaction parameters. If the current scenario category of the vehicle is determined to be the general scenario category, the owner profile parameters and the vehicle status parameters are input into the second-level decision node. The second-level decision node is used to determine whether the current charge status of the vehicle is greater than the charge status of the owner's preferred starting charge. Based on the determination result, it is determined whether the vehicle currently has a charging need. The current charge status of the vehicle belongs to the vehicle status parameters, and the charge status of the owner's preferred starting charge belongs to the owner profile parameters.
5. The method according to claim 4, characterized in that, The actual achievable mileage of the vehicle is calculated using the following formula: E_actual = E_rated × (1 - α × T_congestion - β × ΔT_temp - γ × P_AC- δ × P_heating); Wherein, E_actual represents the actual achievable range of the vehicle, E_rated represents the nominal driving range of the vehicle, T_congestion represents the route congestion coefficient, which is determined based on the road congestion level in the environmental parameters, ΔT_temp represents the difference between the external ambient temperature and the preset optimal battery operating temperature in the environmental parameters, P_heating represents the battery heating power coefficient, which is determined based on the battery temperature in the vehicle status parameters, P_AC represents the air conditioning power coefficient, which is determined based on the air conditioning on / off status in the vehicle status parameters, and α, β, γ and δ represent weighting coefficients.
6. The method according to claim 1, characterized in that, The step of performing strategy analysis on the multimodal perception parameters, the scene category, and the scene feature vector to obtain a multidimensional energy replenishment strategy includes: The multimodal perception parameters, the scene category, and the scene feature vector are input into a pre-trained reinforcement learning model to perform policy analysis and output the multidimensional charging strategy. The multidimensional charging strategy includes a charging warning strategy, a charging route strategy, and a control strategy. The charging warning strategy is used to characterize whether a charging warning needs to be generated and when to generate it. The charging route strategy is used to characterize the route taken by the vehicle to the target charging station. The control strategy includes at least one of the following: a battery heating control strategy for the vehicle, an energy recovery strategy for the vehicle, and a home charging station reservation strategy for the vehicle.
7. The method according to claim 1, characterized in that, After outputting the multi-dimensional energy replenishment strategy for vehicle owners to replenish the vehicle's energy based on the strategy, the method further includes: Receive feedback data from vehicle owners during the vehicle recharging process; The model parameters of the car owner profile construction model are fine-tuned based on the feedback data, wherein the car owner profile construction model is used to output the car owner profile parameters.
8. A vehicle refueling device, characterized in that, The device includes: The acquisition module is used to acquire the multimodal perception parameters corresponding to the vehicle. The scene decision module is used to make scene decisions on the multimodal perception parameters using a pre-trained scene decision tree model, so as to obtain the current scene category and scene feature vector of the vehicle. The strategy analysis module is used to perform strategy analysis on the multimodal perception parameters, the scene category, and the scene feature vector to obtain a multidimensional energy replenishment strategy. The output module is used to output the multi-dimensional energy replenishment strategy so that the vehicle owner can replenish the vehicle based on the multi-dimensional energy replenishment strategy.
9. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; The processor, when executing a program stored in memory, implements the vehicle refueling method according to any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The device stores computer-executable instructions for performing the vehicle refueling method according to any one of claims 1-7.
