Vehicle charging planning method and system based on large language model

Abstract

The invention provides a vehicle charging planning method and system based on a large language model, and belongs to the technical field of vehicle energy management, and the vehicle charging planning method comprises the steps: obtaining a charging planning instruction for a vehicle; generating a charging planning target and a charging planning constraint condition according to the charging planning instruction; acquiring current vehicle data, current power grid charging station data and current environment data; and inputting the current vehicle data, the current power grid charging station data and the current environment data into the large language model for charging planning by taking the charging planning target and the charging planning constraint conditions as model output targets and constraint conditions, and generating at least one charging planning scheme. According to the invention, the charging plan of the vehicle can be dynamically optimized, the charging demand of the user is met, the mileage anxiety of the user is reduced, and the charging time and the charging cost of the user are reduced.

Description

Technical Field

[0001] This invention relates to the field of vehicle energy management technology, and in particular to a vehicle charging planning method and system based on a large language model. Background Technology

[0002] In related technologies, vehicle charging planning is typically based on static planning using rule engines, resulting in slow system response, low fault tolerance, and difficulty in dynamically integrating multi-dimensional data such as user habits, real-time electricity prices, and grid load. Fixed charging strategies cannot adapt to sudden road conditions (such as congestion or accidents) and changes in user demand (such as trip changes). Furthermore, the large peak-valley differences in grid load hinder coordinated optimization between vehicle charging behavior and grid scheduling. Summary of the Invention

[0003] This invention aims to solve at least one of the technical problems existing in the prior art, and proposes a vehicle charging planning method and system based on a large language model.

[0004] In a first aspect, embodiments of the present invention provide a vehicle charging planning method based on a large language model, comprising:

[0005] Obtain charging planning instructions for the vehicle;

[0006] Based on the charging planning instructions, charging planning objectives and charging planning constraints are generated;

[0007] Acquire current vehicle data, current power grid charging station data, and current environmental data;

[0008] Using the charging planning objectives and constraints as the model output objectives and constraints, the current vehicle data, current power grid charging station data, and current environmental data are input into the large language model to perform charging planning, generating at least one charging planning scheme.

[0009] Secondly, embodiments of the present invention provide a vehicle charging planning system based on a large language model, comprising:

[0010] The instruction acquisition unit is used to acquire charging planning instructions for the vehicle.

[0011] The planning target matching unit is used to generate charging planning targets and charging planning constraints according to the charging planning instructions.

[0012] The data acquisition unit is used to acquire current vehicle data, current power grid charging station data, and current environmental data;

[0013] The charging planning unit is used to take the charging planning objectives and charging planning constraints as the model output objectives and constraints, input the current vehicle data, the current power grid charging station data and the current environmental data into the large language model to perform charging planning, and generate at least one charging planning scheme.

[0014] The vehicle charging planning method based on a large language model provided by this invention generates charging planning objectives and constraints by issuing charging planning instructions for the vehicle. These objectives and constraints are then used as the output objectives and constraints of the large language model. The large language model performs real-time charging planning based on real-time collected vehicle data, grid charging station data, and environmental data. This facilitates dynamic optimization of vehicle charging planning, meets user charging needs, reduces user range anxiety, and decreases user charging time and costs. It also achieves multi-objective collaborative optimization of charging strategies, effectively improving user experience and energy network efficiency. Attached Figure Description

[0015] Figure 1 A flowchart illustrating a vehicle charging planning method based on a large language model, provided for an embodiment of the present invention;

[0016] Figure 2 This is a flowchart illustrating one implementation of step S12 in an embodiment of the present invention.

[0017] Figure 3 A structural block diagram of a vehicle charging planning system based on a large language model is provided for an embodiment of the present invention;

[0018] Figure 4 This is a schematic diagram of the system architecture of a vehicle charging planning system provided in an embodiment of the present invention;

[0019] Figure 5 This is a structural block diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solutions of the present invention, exemplary embodiments of the present invention are described below in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0021] Where there is no conflict, the various embodiments of the present invention and the features thereof may be combined with each other.

[0022] As used herein, the term “and / or” includes any and all combinations of one or more related enumerated entries.

[0023] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used herein, the singular forms “a” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that when the terms “comprising” and / or “made of” are used in this specification, the presence of the stated feature, integral, step, operation, element, and / or component is specified, but the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof is not excluded. Terms such as “connected” or “linked” are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect.

[0024] Unless otherwise specified, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having the meaning consistent with their meaning in the context of the relevant art and the invention, and will not be interpreted as having an idealized or overly formal meaning unless expressly so defined herein.

[0025] In related technologies, vehicle charging planning is typically based on static planning using rule engines, resulting in slow system response, low fault tolerance, and difficulty in dynamically integrating multi-dimensional data such as user habits, real-time electricity prices, and grid load. Fixed charging strategies cannot adapt to sudden road conditions (such as congestion or accidents) and changes in user demand (such as trip changes). Furthermore, the large peak-valley differences in grid load hinder coordinated optimization between vehicle charging behavior and grid scheduling.

[0026] This invention provides a vehicle charging planning method based on a large language model. Figure 1 A flowchart illustrating a vehicle charging planning method based on a large language model, as provided in an embodiment of the present invention, is shown below. Figure 1 As shown, the vehicle charging planning method includes:

[0027] Step S11: Obtain charging planning instructions for the vehicle.

[0028] Step S12: Generate charging planning objectives and charging planning constraints according to the charging planning instructions.

[0029] Step S13: Obtain current vehicle data, current power grid charging station data, and current environmental data.

[0030] Step S14: Using the charging planning objectives and constraints as the model output objectives and constraints, input the current vehicle data, current power grid charging station data, and current environmental data into the large language model to perform charging planning and generate at least one charging planning scheme.

[0031] According to the vehicle charging planning method of the present invention, charging planning objectives and constraints are generated by charging planning instructions for vehicles. The charging planning objectives and constraints are used as the output objectives and constraints of a large language model. The large language model performs charging planning in real time based on real-time collected vehicle data, power grid charging station data, and environmental data. This method is conducive to dynamically optimizing vehicle charging planning, meeting user charging needs, reducing user range anxiety, reducing user charging time and charging costs, achieving multi-objective collaborative optimization of charging strategies, and effectively improving user experience and energy network efficiency.

[0032] In some embodiments, obtaining the charging planning instruction for the vehicle in step S11 above may further include receiving the charging planning instruction input by the user terminal.

[0033] The charging planning command input by the user terminal can be a voice command or a text command. Users can initiate the charging planning command through voice input or text input.

[0034] In some embodiments, obtaining the charging planning instruction for the vehicle in step S11 above may further include: generating the charging planning instruction in response to detecting that the state of charge (SOC) of the vehicle battery is lower than a state threshold.

[0035] Among them, the battery state of charge (SOC) represents the percentage of the electric vehicle's current remaining power relative to its total capacity. The state threshold can be set according to actual needs, and this embodiment of the invention does not impose any special restrictions on it. For example, the state threshold can be set to 20%.

[0036] Figure 2 This is a flowchart illustrating one implementation of step S12 in an embodiment of the present invention. In some embodiments, generating charging planning targets and charging planning constraints according to charging planning instructions in step S12 may further include:

[0037] Step S21: Parse the charging planning instruction to obtain the corresponding charging demand information.

[0038] In some embodiments, charging demand information includes, but is not limited to, charging location demand, charging time demand, charging cost demand, charging brand demand, and fast / slow charging demand.

[0039] In some embodiments, the charging planning instruction is an instruction actively input by the user through a user terminal, which carries the user's charging demand information. For example, if the user inputs the instruction "Arrive at the destination before 6 PM, budget 50 yuan for charging," the user's charging demand information, including charging address requirements and charging cost requirements, can be parsed from the instruction.

[0040] In some embodiments, the charging planning instruction is generated by the system when it detects that the vehicle battery's state of charge is below a threshold. When the system detects that the vehicle battery's state of charge is below the threshold, it can generate the charging planning instruction based on the system's default charging demand information, or based on the user's preferred charging demand information. For example, the system's default charging demand information includes prioritizing the charging address closest to the user, prioritizing the charging station with the lowest electricity price, etc.

[0041] Step S22: Based on the charging demand information, generate charging planning objectives and charging planning constraints.

[0042] In some embodiments, the semantic understanding capabilities of a Large Language Model (LLM) are used to identify the user's charging planning instructions and extract entities, thereby extracting charging demand information and transforming unstructured charging demand information into structured charging planning objectives and charging planning constraints.

[0043] Intent recognition refers to recognizing the user's intent in instructions through the semantic understanding capabilities of LLM.

[0044] Entity extraction refers to identifying key entities (such as location, time, preferred brand) and operation types (such as "avoid highways" or "prioritize low prices") in instructions.

[0045] For example, if a user inputs a charging planning instruction such as "I need to charge midway and avoid highways," the LLM can parse the user's natural language instruction, extract the user's charging requirement information "avoid highways," and generate structured constraints as follows:

[0046] {"route_type":"city_road"}.

[0047] For example, if a user inputs a charging planning instruction of "arrive before 6 PM, charging budget not exceeding 50 yuan", LLM can parse the user's natural language instruction to extract the user's charging demand information "arrive before 6 PM" and "charging cost less than or equal to 50 yuan". This information is then structured to obtain the charging planning objective and charging constraints:

[0048] {"deadline":"18:00","max_cost":50,"route_preference":"shortest_time"}

[0049] In some embodiments, the structured charging planning objectives and charging planning conditions are in JSON format.

[0050] In some embodiments, the Embedding module of LLM (such as GPT-4 or a self-developed model) is used to encode instructions, and combined with domain adaptation, the recognition accuracy of charging scenario terms (such as SOC, V2G) is improved.

[0051] In some embodiments, during step S13, current vehicle data, current grid charging station data, and current environmental data can be continuously acquired in real time while the vehicle is in motion. Current vehicle data includes the vehicle's battery state of charge, remaining driving range, vehicle location, and user trip planning. Current grid charging station data includes: regional grid load, charging station location, charging pile status (e.g., idle, occupied, power status), dynamic electricity price, and whether the charging station is equipped with V2G (Vehicle to Grid) functionality. Current environmental data includes: road traffic data (e.g., traffic flow, road speed limits) and weather data (e.g., weather data for the next two hours).

[0052] Among them, the vehicle's battery state of charge, remaining driving range, and vehicle location can be obtained through onboard sensors; user trip planning can be obtained through onboard maps; information such as regional power grid load, dynamic electricity prices, and whether charging stations are equipped with V2G (Vehicle to Grid) functionality can be obtained through regional EMS (Energy Management System), which is used to monitor and optimize the operation of the regional power grid; and charging station locations, charging pile status (such as idle status, occupied status, and power status), and road traffic data can be obtained through high-precision maps.

[0053] In some embodiments, the charging planning scheme output by the model includes, but is not limited to: charging station recommendation information, charging path, estimated arrival time, estimated charging duration and charging cost, and grid load impact index. The charging station recommendation information includes the identifier of the recommended charging station, the charging path refers to the path from the vehicle's location to the location of the recommended charging station, and the estimated arrival time is the estimated time to reach the recommended charging station.

[0054] In some embodiments, before inputting the current vehicle data, current power grid charging station data, and current environmental data into the Large Language Model for charging planning, using charging planning objectives and charging planning constraints as model output objectives and constraints, the method further includes: encoding the real-time collected current vehicle data, current power grid charging station data, and current environmental data through the Embedding module of the LLM model, and then concatenating and merging the encoded data to construct LLM input context information, which is then input into the context window of the LLM model.

[0055] In some embodiments, before encoding the data, a dedicated Tokenizer and Positional Encoding can be used to preprocess the data. The Tokenizer supports mixed data type input (text + numeric + timestamp), and Positional Encoding can preserve the spatiotemporal order of the data.

[0056] By constructing LLM context information in the above manner, multimodal data fusion and context construction can be achieved.

[0057] In some embodiments, before encoding the current vehicle data, current grid charging station data, and current environmental data, multi-frequency data (such as second-level vehicle GPS positioning and minute-level electricity price) in the current vehicle data, current grid charging station data, and current environmental data are aligned to a unified timestamp through time series interpolation and sliding window aggregation.

[0058] In some embodiments, before inputting current vehicle data, current power grid charging station data, and current environmental data into the large language model for charging planning, using charging planning objectives and charging planning constraints as model output objectives and constraints, the method further includes: employing Few-shot Prompting technology to perform LLM prompt word engineering based on the charging planning objectives and charging planning constraints, and constructing and injecting prompt word templates (Prompts) of the charging planning objectives and charging planning constraints into the LLM model.

[0059] Among them, the Few-shot Prompting technique guides the LLM to generate decision outputs that conform to the format through a small number of examples, reducing the dependence on training data; for the LLM output, the Output Schema is used to ensure the legality of the JSON format, and the Temperature coefficient is set to 0.3 to balance the diversity and stability of the generated results.

[0060] In some embodiments, in the step of using charging planning objectives and constraints as model output objectives and constraints, and inputting current vehicle data, current power grid charging station data, and current environmental data into the Large Language Model (LLM) for charging planning, the prompt word template generated based on the charging planning objectives and constraints, and the context information generated based on the current vehicle data, current power grid charging station data, and current environmental data, are input into the Large Language Model (LLM) for charging planning reasoning. The Large Language Model (LLM) outputs at least one structured charging planning scheme.

[0061] In some embodiments, the structured charging plan output by LLM is in JSON format.

[0062] For example, a prompt template for entering LLM:

[0063] [Charging planning objective] Minimize total travel time;

[0064] [Charging planning constraints] Battery SOC must not be lower than 20%, and priority should be given to cooperating brand charging stations.

[0065] Enter the context information for the LLM:

[0066] [Vehicle Data] SOC=25%, Location=Location coordinates (116.4, 39.9), Remaining range=80km;

[0067] [Grid Charging Station Data] Regional grid load = 85%, V2G availability of charging station = Yes, Charging station A: Idle, Electricity price = 0.8 yuan / kWh, Charging station B: Occupancy rate 70%.

[0068] LLM output:

[0069] Structured charging planning scheme (JSON format): includes recommended charging station information, charging route, estimated arrival time, estimated charging duration and cost, and grid load impact index.

[0070] In some embodiments, when the LLM performs charging planning inference, it can interact with the power grid to perform multi-objective collaborative optimization of the charging planning scheme. Specifically, when the LLM performs charging planning inference, it interacts with the regional energy management system (EMS) to obtain real-time power grid load data. When it is determined from the real-time power grid load data that the power grid load is at its peak, the LLM automatically adjusts its output recommendation strategy, giving priority to recommending charging stations equipped with V2G (vehicle-to-grid) functionality, and guiding vehicles to charging stations equipped with V2G functionality for discharge.

[0071] In some embodiments, in step S14 above, the charging planning objective and charging planning constraints are used as the model output objective and constraints. Current vehicle data, current power grid charging station data, and current environmental data are input into the large language model for charging planning, including:

[0072] In response to the vehicle meeting the charging planning trigger conditions, the charging planning objectives and constraints are used as the model output objectives and constraints. The current vehicle data, current power grid charging station data, and current environmental data are input into the large language model for charging planning.

[0073] The charging planning trigger conditions include at least one of the following:

[0074] The current time is more than a preset time threshold since the last charging plan;

[0075] The charging station scheduled in the last charging session has been detected as occupied.

[0076] The current battery state of charge has been detected to be decreasing more than expected.

[0077] Changes to the current user's travel plans have been detected.

[0078] In some embodiments, in response to the vehicle not meeting the charging plan triggering conditions, the charging plan obtained from the previous charging plan remains unchanged, and the vehicle performs charging planning according to the original charging plan.

[0079] In some embodiments, at least one charging planning scheme includes multiple charging planning schemes. After generating at least one charging planning scheme, the vehicle charging planning method further includes: determining a target charging planning scheme from the multiple charging planning schemes; and performing charging planning for the vehicle according to the target charging planning scheme.

[0080] In some embodiments, a target charging plan is determined from multiple charging plan schemes according to preset weights.

[0081] The preset weight is the degree of attention paid to each charging plan information in the charging plan scheme, that is, the importance of each charging plan information. For example, if the charging cost accounts for 60% and the grid load impact index accounts for 40%, then the charging plan scheme with the lowest charging cost will be selected first. If the charging costs are the same, then the charging plan scheme with the lower grid load impact index will be selected first.

[0082] In some embodiments, multiple charging plans generated by the plan are pushed to the user terminal; in response to the user's selection of a charging plan through the user terminal, the user-selected charging plan is taken as the target charging plan.

[0083] In some embodiments, charging planning for the vehicle according to a target charging plan includes: guiding the vehicle to a recommended charging station via a navigation system according to the charging path; and sending a charging plan (such as start time and expected power demand) to the power grid EMS.

[0084] In some embodiments, after planning the charging of the vehicle according to the target charging plan, the method further includes: in response to a charging station failure, invoking a backup charging plan for charging.

[0085] Among them, the backup charging plan can be a recommended charging station that is close to the target charging plan among the above multiple charging plans, or a new charging plan can be made.

[0086] In some embodiments, after planning the charging of the vehicle according to the target charging plan, the method further includes: in response to receiving an emergency grid dispatch request, the request including information on a forced switch to V2G discharge mode, confirming with the user terminal whether to force a switch to V2G discharge mode; in response to receiving an operation from the user allowing a forced switch to V2G discharge mode, switching the charging mode to V2G discharge mode; and in response to receiving an operation from the user refusing a forced switch to V2G discharge mode, recording user feedback information.

[0087] In some embodiments, after planning the charging of the vehicle according to the target charging plan, the method further includes: obtaining user feedback information, regional power grid load data, and actual vehicle charging data; and optimizing the model parameters of the large language model using a reinforcement learning model based on the user feedback information, regional power grid load data, and actual vehicle charging data.

[0088] By using reinforcement learning models, the model parameters of the LLM model, such as the model's decision weights, are adjusted based on real-time feedback data such as user feedback information (e.g., user adoption rate, user satisfaction), regional power grid load data (e.g., power grid load smoothness), and actual vehicle charging data (e.g., actual charging time, cost deviation, charging efficiency).

[0089] Reward function design for reinforcement learning model: Reward = 0.6 × User satisfaction + 0.3 × Grid load smoothness + 0.1 × Charging efficiency.

[0090] Among them, user satisfaction: whether users adopt the charging plan and the number of times users manually adjust the plan; grid load smoothness: the degree of matching between the charging time distribution and the grid peak and valley.

[0091] In some embodiments, the reinforcement learning model employs the Q-learning algorithm and utilizes the proximal policy optimization (PPO) algorithm to update the parameters of the last layer of the LLM, fine-tuning only 1% to 5% of the model parameters to reduce computational overhead.

[0092] The vehicle charging planning method according to embodiments of the present invention employs a multimodal data fusion architecture. It structures user-input charging planning instructions into LLM-recognizable charging planning objectives and constraints, and fuses collected vehicle data, grid charging station data, and environmental data to construct an LLM model input context. The LLM dynamic decision model is then used to make charging planning decisions, generating a multi-objective optimized charging planning scheme. At the user level, it minimizes travel time and charging costs while prioritizing user preferences (such as branded charging stations). At the grid level, it guides vehicles to charge during low-load periods, mitigating grid fluctuations. For the LLM model, a lightweight LLM fine-tuning framework is used, dynamically optimizing model weights through reinforcement learning (RL), enabling it to adapt to different urban grid policies and user behavior patterns.

[0093] During the charging planning process, vehicle data, grid charging station data, and environmental data can be continuously received and collected to periodically evaluate the charging plan. When deviations from the original charging plan are detected (such as increased power consumption due to congestion or user changes to their trip), dynamic planning can be automatically triggered, and updated planning suggestions can be pushed through the vehicle's HMI or mobile app.

[0094] In practical applications, knowledge distillation can compress large LLM models into lightweight versions that can run on in-vehicle terminals. Complex calculations (such as power grid load forecasting) are handled in the cloud, while real-time decisions are executed locally. A full-parameter LLM is deployed in the cloud, responsible for model training and power grid co-optimization, while a lightweight LLM (<1B parameters) is deployed on the in-vehicle terminal, supporting offline semantic parsing and fast inference. Low-Rank Adaptation (LoRA) technology can be used for efficient parameter fine-tuning, and quantization can reduce model weights from FP32 precision to INT8 precision, reducing memory usage by 75%.

[0095] In some embodiments, after the LLM generates a solution, the status of the charging pile can be queried a second time to confirm its availability. When an abnormal situation occurs, a fault tolerance mechanism can be triggered.

[0096] For example, when an anomaly is detected (such as an incorrect charging station status), the LLM's rapid replanning capability is triggered.

[0097] When an anomaly is detected (such as a timeout returned by the charging pile API), the current state can be predicted using historical charging pile data.

[0098] When an execution anomaly is detected (such as the user not following the navigation), the location can be re-collected and local planning can be triggered.

[0099] In some embodiments, if the confidence level of the LLM output is lower than a threshold, the execution engine of the charging plan is switched to a rule engine (such as the Dijkstra algorithm).

[0100] The vehicle charging planning method of this invention has the following technical advantages:

[0101] 1. Dynamic adaptability: Through real-time data and reinforcement learning (RL) optimization, it can effectively cope with rapid changes in road conditions, power grid, and user behavior.

[0102] 2. Multi-objective collaboration: Breaking through the limitations of single optimization in traditional rule engines, achieving a win-win situation for users and the power grid.

[0103] 3. Low resource overhead: Lightweight technology enables LLM to be deployed in vehicle terminals, reducing response latency by at least 2 seconds.

[0104] 4. User-friendly interaction: Supports natural language interaction and visual adjustment of solutions, reducing the learning cost for users.

[0105] This invention also provides a vehicle charging planning system based on a large language model. Figure 3 A structural block diagram of a vehicle charging planning system based on a large language model is provided for an embodiment of the present invention, as shown below. Figure 3 As shown, the vehicle charging planning system 300 includes:

[0106] The instruction acquisition unit 301 is used to acquire charging planning instructions for the vehicle.

[0107] The planning target matching unit 302 is used to generate charging planning targets and charging planning constraints according to the charging planning instructions.

[0108] The data acquisition unit 303 is used to acquire current vehicle data, current power grid charging station data, and current environmental data.

[0109] The charging planning unit 304 is used to take the charging planning objectives and charging planning constraints as the model output objectives and constraints, input the current vehicle data, the current power grid charging station data and the current environmental data into the large language model to perform charging planning, and generate at least one charging planning scheme.

[0110] The vehicle charging planning system based on a large language model in this embodiment of the invention is used to implement the vehicle charging planning method described above. For a detailed description of each functional unit in the system, please refer to the relevant description in the vehicle charging planning method of the above embodiment, which will not be repeated here.

[0111] Figure 4 This is a schematic diagram of the system architecture of a vehicle charging planning system provided in an embodiment of the present invention, such as... Figure 4As shown, the system includes: a data acquisition layer, a data processing layer, a decision layer, and an output layer.

[0112] like Figure 4 As shown, the data acquisition layer includes vehicle-mounted sensors, user terminals, regional EMS, and third-party data terminals.

[0113] Among them, the on-board sensors are responsible for collecting vehicle data such as SOC, remaining driving range, and vehicle location (GPS) in real time.

[0114] User terminal: Responsible for receiving charging planning instructions (such as "prioritize free parking and charging stations") via voice / text input.

[0115] Regional EMS: Responsible for providing grid data such as grid load, dynamic electricity price, and V2G availability status.

[0116] Third-party data sources include high-precision map APIs and weather APIs;

[0117] High-precision map API: Responsible for uploading road traffic data such as traffic flow, road speed limits, charging station locations, and charging pile status.

[0118] Weather API: Responsible for acquiring weather data.

[0119] The data processing layer includes a semantic parsing module, a time-series alignment module, and a data fusion module.

[0120] The semantic parsing module is used to parse user natural language instructions through LLM and generate structured goals and constraints.

[0121] Time alignment module: used to align multi-frequency data (such as second-level GPS positioning, minute-level electricity price) in the collected data to a unified timestamp.

[0122] Data fusion module: used to combine and collect data such as vehicle SOC, charging pile status, and grid load to construct LLM input context information.

[0123] At the decision-making level, this includes the LLM inference engine and the reinforcement learning tuning module.

[0124] The LLM inference engine is used to generate a structured, multi-objective optimized charging plan based on the prompt word templates provided by Few-shot Prompting and the context information input from the context window.

[0125] Reinforcement Learning Tuning Module: Used to dynamically adjust the LLM model weights based on historical decision performance (such as user adoption rate, grid load smoothness, etc.) through reinforcement learning.

[0126] The output layer includes a dynamic visualization module and a power grid collaborative execution module.

[0127] The dynamic visualization module is responsible for displaying the charging path, charging station details, and real-time update prompts of the target charging plan in the in-vehicle HMI or mobile APP. Users can manually adjust the charging plan through the dynamic visualization module (such as extending the charging time to obtain a lower electricity price).

[0128] The power grid coordination execution module is responsible for interacting with the power grid control system through an encrypted API and sending vehicle group charging scheduling suggestions to the EMS (such as delaying the charging time of 10% of vehicles in a certain area).

[0129] Based on the same inventive concept, embodiments of the present invention also provide an electronic device. Figure 5 This is a structural block diagram of an electronic device provided in an embodiment of the present invention. Figure 5 As shown, an embodiment of the present invention provides an electronic device including: one or more processors 501, a memory 502, and one or more I / O interfaces 503. The memory 502 stores one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement any of the vehicle charging planning methods described in the above embodiments; the one or more I / O interfaces 503 are connected between the processors and the memory, configured to enable information interaction between the processors and the memory.

[0130] Among them, processor 501 is a device with data processing capabilities, including but not limited to central processing unit (CPU); memory 502 is a device with data storage capabilities, including but not limited to random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory (FLASH); I / O interface (read-write interface) 503 is connected between processor 501 and memory 502, and can realize information interaction between processor 501 and memory 502, including but not limited to data bus (Bus).

[0131] In some embodiments, the processor 501, memory 502, and I / O interface 503 are interconnected via bus 504, and thus connected to other components of the computing device.

[0132] In some embodiments, the one or more processors 501 include a field-programmable gate array.

[0133] This invention also provides a computer-readable medium. The computer-readable medium stores a computer program, which, when executed by a processor, implements the steps of any of the vehicle charging planning methods described in the above embodiments. The computer-readable storage medium may be volatile or non-volatile.

[0134] This invention also provides a computer program product, including computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code. When the computer-readable code is run in the processor of an electronic device, the processor in the electronic device executes the above-described vehicle charging planning method.

[0135] Those skilled in the art will understand that all or some of the steps, systems, and apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software can be distributed on a computer-readable storage medium, which may include computer storage media (or non-transitory media) and communication media (or transient media).

[0136] As is known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable program instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), static random access memory (SRAM), flash memory or other memory technologies, portable compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, it is known to those skilled in the art that communication media typically contain computer-readable program instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0137] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0138] The computer program instructions used to perform the operations of this invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing state information from the computer-readable program instructions. This electronic circuitry can execute the computer-readable program instructions to implement various aspects of the invention.

[0139] The computer program product described herein can be implemented specifically through hardware, software, or a combination thereof. In one alternative embodiment, the computer program product is specifically embodied in a computer storage medium; in another alternative embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0140] Various aspects of the present invention are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0141] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0142] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0143] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction, which contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0144] Example embodiments have been disclosed herein, and while specific terminology has been used, it is for illustrative purposes only and should be construed as such, and is not intended to be limiting. In some instances, it will be apparent to those skilled in the art that features, characteristics, and / or elements described in conjunction with particular embodiments may be used alone, or in combination with features, characteristics, and / or elements described in conjunction with other embodiments, unless otherwise expressly indicated. Therefore, those skilled in the art will understand that various changes in form and detail may be made without departing from the scope of the invention as set forth in the appended claims.

Claims

1. A vehicle charging planning method based on a large language model, characterized in that, include: Obtain charging planning instructions for the vehicle; Based on the charging planning instructions, charging planning objectives and charging planning constraints are generated; Acquire current vehicle data, current power grid charging station data, and current environmental data; Using the charging planning objectives and constraints as the model output objectives and constraints, the current vehicle data, current power grid charging station data, and current environmental data are input into the large language model to perform charging planning, generating at least one charging planning scheme.

2. The vehicle charging planning method according to claim 1, characterized in that, The process of obtaining charging planning instructions for the vehicle includes: Receive the charging planning instruction input by the user terminal; or, The charging planning instruction is generated in response to the detection that the vehicle battery state of charge is below a state threshold.

3. The vehicle charging planning method according to claim 1, characterized in that, The step of generating charging planning objectives and charging planning constraints according to the charging planning instructions includes: The charging planning instruction is parsed to obtain the corresponding charging demand information; Based on the charging demand information, the charging planning objective and the charging planning constraints are generated.

4. The vehicle charging planning method according to claim 1, characterized in that, The current vehicle data includes the vehicle's battery state of charge, remaining driving range, vehicle location, and user trip plan. The current power grid charging station data includes: regional power grid load, charging station location, charging pile status, dynamic electricity price, and whether the charging station is equipped with V2G functionality; The current environmental data includes: road traffic data and weather data.

5. The vehicle charging planning method according to claim 1, characterized in that, The charging planning scheme includes: recommended charging station information, charging route, estimated arrival time, estimated charging duration and charging cost, and grid load impact index.

6. The vehicle charging planning method according to claim 1, characterized in that, The process of using the charging planning objectives and constraints as the model output objectives and constraints, and inputting current vehicle data, current power grid charging station data, and current environmental data into the large language model for charging planning, includes: In response to the vehicle meeting the charging planning triggering conditions, the charging planning target and charging planning constraints are used as the model output target and constraints. The current vehicle data, the current power grid charging station data, and the current environmental data are input into the large language model for charging planning.

7. The vehicle charging planning method according to claim 6, characterized in that, The charging planning triggering conditions include at least one of the following: The current time is more than a preset time threshold since the last charging plan; The charging station scheduled in the last charging session has been detected as occupied. The current battery state of charge has been detected to be decreasing more than expected. Changes to the current user's travel plans have been detected.

8. The vehicle charging planning method according to claim 1, characterized in that, The at least one charging plan scheme includes multiple charging plan schemes. After generating at least one charging plan scheme, the method further includes: The target charging plan is determined from the plurality of charging plan schemes; The vehicle's charging plan is executed according to the target charging plan scheme.

9. The vehicle charging planning method according to claim 8, characterized in that, After planning the charging of the vehicle according to the target charging plan, the method further includes: Obtain user feedback information, regional power grid load data, and actual vehicle charging data; Based on user feedback, regional power grid load data, and actual vehicle charging data, the parameters of the large language model are optimized using a reinforcement learning model.

10. A vehicle charging planning system based on a large language model, characterized in that, include: The instruction acquisition unit is used to acquire charging planning instructions for the vehicle. The planning target matching unit is used to generate charging planning targets and charging planning constraints according to the charging planning instructions. The data acquisition unit is used to acquire current vehicle data, current power grid charging station data, and current environmental data; The charging planning unit is used to take the charging planning objectives and charging planning constraints as the model output objectives and constraints, input the current vehicle data, the current power grid charging station data and the current environmental data into the large language model to perform charging planning, and generate at least one charging planning scheme.

Images