A remote-controlled vehicle intelligent one-key charging method and system

Abstract

The application discloses a remote control vehicle intelligent one-key charging method and system, and belongs to the technical field of electric vehicle charging control. The method comprises the following steps: after receiving a user remote one-key charging instruction, obtaining user historical behavior, preference mode and charging habit, and performing intention inference to obtain a behavior intention vector; collecting real-time running state of a controlled energy unit and deeply representing the running state to obtain a running state representation; performing action influence prediction based on the behavior intention vector and the running state representation, dynamically optimizing a preset strategy graph to form an adaptive strategy graph structure, generating a candidate action sequence through strategy search and combination, and performing executability verification to obtain a target action sequence. The target action sequence is screened and mapped for a target vehicle, and a control instruction is issued to the controlled energy unit, so that remote intelligent charging is realized. The scheme dynamically optimizes the charging strategy in combination with user intention and real-time running state, generates a control instruction executable for the target vehicle, and improves charging matching degree and execution efficiency.

Description

Technical Field

[0001] This application belongs to the field of electric vehicle charging control technology, specifically relating to a remotely controlled intelligent one-button charging method and system for vehicles. Background Technology

[0002] With the widespread adoption of the energy internet and smart devices, users' demands for flexibility and personalization in energy systems have increased significantly. In scenarios such as electric vehicles, users want efficient charging while also considering the status of energy units, such as grid load and battery health, to avoid resource waste and equipment damage. Dynamically sensing user intent and coordinating with energy unit status to achieve precise adaptive charging control has become crucial for improving energy efficiency.

[0003] Existing electric vehicle charging control mostly adopts a preset rule base driven mode. When a user initiates a charging command for a specific vehicle, the system directly reads the current remaining power and charging power of the energy unit associated with the vehicle, determines the charging parameters according to fixed rules, such as starting fast charging mode when the remaining power is less than 20%, and generates control commands by combining preset strategies such as staged constant current charging.

[0004] However, preset rules cannot dynamically adapt to the differentiated characteristics of users' charging habits, such as common time periods and power preferences, resulting in a mismatch between control commands and actual needs. At the same time, static strategies are difficult to respond to changes in the real-time status of energy units, such as grid load fluctuations and battery temperature changes, which can easily lead to overcharging risks or efficiency degradation, ultimately causing the dual problems of low energy utilization and increased equipment wear. Summary of the Invention

[0005] To overcome the above-mentioned shortcomings, this invention is proposed to provide a solution, or at least a partial solution, to the technical problem that the static rules of the prior art are difficult to adapt to users' charging habits and real-time changes in the state of energy units, which easily leads to a mismatch between charging instructions and actual needs, reduces energy utilization efficiency, and accelerates equipment wear and tear.

[0006] In a first aspect, the present invention provides a remotely controlled intelligent one-button charging method for vehicles, the method comprising:

[0007] If a one-click charging command for the vehicle is received remotely from a user, the user's historical behavior, preference patterns, and charging habits are obtained. Based on the user's historical behavior, preference patterns, and charging habits, the user's intent is inferred to obtain the user's behavioral intent vector.

[0008] Acquire real-time operating status data of the controlled energy unit, and perform in-depth characterization processing based on the real-time operating status data to obtain an operating status representation;

[0009] Action impact prediction is performed based on behavioral intention vectors and operational state representations to obtain action prediction results. Based on behavioral intention vectors, operational state representations, and action prediction results, the preset strategy graph structure is dynamically optimized to obtain an adaptive strategy graph structure.

[0010] Based on the adaptive strategy graph structure, strategy search and combination are performed to obtain at least one candidate action sequence. Based on the candidate action sequence, executability verification is performed to obtain the target action sequence acting on the controlled energy unit.

[0011] The target vehicle is determined based on the one-click charging command. Based on the target vehicle, the target action sequence is filtered and mapped to obtain the charging control action acting on the target vehicle. The control command is then sent to the controlled energy unit according to the charging control action to provide charging energy to the target vehicle.

[0012] In a second aspect, the present invention provides a remotely controlled intelligent one-button charging system for vehicles, the system comprising:

[0013] The instruction receiving module is used to obtain the user's historical behavior, preference patterns and charging habits if it receives a one-click charging instruction for the vehicle sent remotely by the user, and to infer the user's intention based on the user's historical behavior, preference patterns and charging habits to obtain the user's behavioral intention vector.

[0014] The deep characterization module is used to acquire real-time operating status data of the controlled energy unit, and perform deep characterization processing on the real-time operating status data to obtain an operating status representation.

[0015] The dynamic optimization module is used to predict the impact of actions based on behavioral intention vectors and running state representations, obtain action prediction results, and dynamically optimize the preset strategy graph structure based on behavioral intention vectors, running state representations, and action prediction results to obtain an adaptive strategy graph structure.

[0016] The verification module is used to perform policy search and combination based on the adaptive policy graph structure to obtain at least one candidate action sequence, and to perform executability verification based on the candidate action sequence to obtain the target action sequence acting on the controlled energy unit.

[0017] The charging control module is used to determine the target vehicle based on the vehicle's one-click charging command, perform action filtering and mapping on the target action sequence based on the target vehicle to obtain the charging control action acting on the target vehicle, and issue control commands to the controlled energy unit according to the charging control action to provide charging energy to the target vehicle.

[0018] In a third aspect, an electronic device is provided, comprising a processor, a memory, and a program or instructions stored in the memory and executable on the processor, the program or instructions being loaded and run by the processor to perform the steps of the aforementioned remotely controlled vehicle intelligent one-button charging method.

[0019] In a fourth aspect, a computer-readable storage medium is provided, wherein a plurality of program codes are stored therein, the program codes being adapted to be loaded and run by a processor to perform the steps of the aforementioned remotely controlled vehicle intelligent one-button charging method.

[0020] The above-described technical solutions of the present invention have at least one or more of the following beneficial effects:

[0021] In implementing the technical solution of this invention, the charging strategy is dynamically optimized by combining user intent and real-time operating status, and executable control commands are generated for the target vehicle, thereby improving the matching degree and execution efficiency of charging. Attached Figure Description

[0022] The disclosure of this invention will become more readily understood with reference to the accompanying drawings. It will be readily understood by those skilled in the art that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. Furthermore, similar numbers in the drawings are used to denote similar components, wherein:

[0023] Figure 1 This is a schematic diagram of the vehicle charging control steps of a remotely controlled intelligent one-button charging method for vehicles according to an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of the charging anomaly handling steps of a remotely controlled intelligent one-button charging method for vehicles according to an embodiment of the present invention.

[0025] Figure 3 This is a schematic diagram of the main structure of a remotely controlled intelligent one-button charging system for vehicles according to an embodiment of the present invention.

[0026] Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0027] Some embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0028] In the description of this invention, "module" and "processor" can include hardware, software, or a combination of both. A module can include hardware circuitry, various suitable sensors, communication ports, memory, and may also include software components, such as program code, or a combination of software and hardware. A processor can be a central processing unit, microprocessor, image processor, digital signal processor, or any other suitable processor. The processor has data and / or signal processing capabilities. The processor can be implemented in software, in hardware, or a combination of both. Non-transitory computer-readable storage media include any suitable medium capable of storing program code, such as magnetic disks, hard disks, optical disks, flash memory, read-only memory, random access memory, etc. The term "A and / or B" means all possible combinations of A and B, such as only A, only B, or A and B. The terms "at least one A or B" or "at least one of A and B" have a similar meaning to "A and / or B" and can include only A, only B, or A and B. The singular terms "a" or "this" can also include plural forms.

[0029] See appendix Figure 1 , Figure 1 This is a schematic flowchart illustrating the vehicle charging control steps of a remotely controlled intelligent one-button charging method for vehicles according to an embodiment of the present invention. Figure 1 As shown, a remotely controlled vehicle intelligent one-button charging method in an embodiment of the present invention mainly includes the following steps S101-S105.

[0030] Step S101: If a one-click charging command for the vehicle is received remotely from the user, the user's historical behavior, preference patterns, and charging habits are obtained. Based on the user's historical behavior, preference patterns, and charging habits, the user's intent is inferred to obtain the user's behavioral intent vector.

[0031] One-click charging commands are charging request commands sent by users via remote terminals such as mobile apps, in-vehicle systems, or cloud platforms to initiate the automatic charging process for a designated vehicle. For example, a user can send a one-click charging command via a mobile app on their way home from get off work, and the system will schedule it in advance. Once the vehicle arrives home and is plugged in, charging will begin automatically. Alternatively, a user can remotely click one-click charging from the office, and vehicles that are already plugged in but not powered will immediately begin charging. At night, users can remotely set up one-click charging, and the system will take advantage of off-peak electricity pricing, automatically starting charging at the set time once the vehicle is connected. It's also convenient in cold weather; users can remotely click one-click charging in advance, and once the vehicle is connected, the system will preheat the battery while charging begins, resulting in higher charging efficiency and greater safety.

[0032] User history behavior is the behavioral data left by users in previous charging-related operations, such as the specific charging time, charging frequency, how long each charging took, and the charging power selected.

[0033] Preference patterns are the regular preferences that users exhibit when charging, such as liking to charge at night, using specific charging stations, or prioritizing low-cost options when charging.

[0034] Charging habits are the charging habits that users develop over a long period of time, including charging at fixed times, frequently used charging locations, and the desired battery life after charging.

[0035] A behavioral intent vector is a numerical representation derived from a user's historical behavior, preference patterns, and charging habits, inferred through intent. It describes the user's current charging needs and priorities.

[0036] Upon receiving a one-click charging command from a user, the system first parses the command to identify the target vehicle, desired charging time, and charging method, then converts this information into a standardized data format. Next, it retrieves and cleans and structures the user's historical behavior data. Based on this processed data, it performs statistical analysis of historical user behavior, extracting key features such as charging time distribution, power distribution, and charging frequency to form preference patterns. Simultaneously, it infers the user's charging intentions in specific situations by combining long-established charging habits. In the intention inference stage, it integrates historical behavior features, preference patterns, and charging habits for comprehensive analysis, generating a behavioral intention vector through rule-based calculation or probabilistic inference. Each dimension of this vector corresponds to different charging preferences or constraints, such as priority charging time, target range, and cost sensitivity. For example, a user might typically charge at home at 80% power between 10 PM and 1 AM, aiming for a 300km range. After receiving the one-click charging command for the vehicle, by combining historical data such as nighttime charging peaks, 80% power preference, and frequently used home charging stations, along with long-term habits, it can infer behavioral intent vectors containing dimensions such as "prioritize nighttime charging", "80% power", and "300km range".

[0037] Step S102: Obtain real-time operating status data of the controlled energy unit, and perform deep characterization processing based on the real-time operating status data to obtain an operating status representation.

[0038] Controlled energy units are intelligently controlled energy supply or storage devices, such as the energy storage modules in charging piles. Their core function is to provide adjustable amounts of electricity to vehicles that need to be charged.

[0039] Real-time operating status data records the current working status of the controlled energy unit, such as how much power is currently output, how much electricity is remaining, and whether each control channel is occupied. By looking at this data, you can clearly see how many usable resources the energy unit has and how heavy the load is.

[0040] Operational status representation is the result obtained by processing, extracting features, or encoding real-time operational status data. For example, normalization, vectorization, or deep representation methods are used to transform the data into numerical form so that the system can recognize it.

[0041] Real-time operating status data of the controlled energy unit can be collected from sensors or monitoring interfaces within the energy unit. The collection frequency must be high enough to ensure data accuracy and timeliness. The collected data is first checked and filtered, removing abnormal, missing, or obviously inconsistent data. Then, data units are standardized and timestamps are aligned to ensure that different parameters are comparable and consistent at the same point in time. The checked data is then preprocessed, such as normalized, smoothed, and standardized, so that data can be compared during subsequent processing and noise interference with decision-making is reduced. The processed parameters are then integrated to form a high-dimensional feature vector, which contains both current instantaneous data and sliding window data over a period of time.

[0042] After obtaining the feature vectors, deep representation techniques are used to process them again, extracting more abstract and easily distinguishable state features. For example, autoencoders, time-series convolutional networks, or recurrent neural networks are used to encode the original vectors, resulting in a lower-dimensional representation that retains all key information. These representations clearly reflect the current load status of the controlled energy unit, how much energy it has stored, and its potential response capabilities. Finally, the results of deep representation are used as the operational state representation.

[0043] Step S103: Based on the behavioral intention vector and the running state representation, predict the action impact to obtain the action prediction result. Based on the behavioral intention vector, the running state representation and the action prediction result, dynamically optimize the preset strategy graph structure to obtain an adaptive strategy graph structure.

[0044] Action prediction results are data obtained by combining the current user behavior intent vector and the operating state representation of the controlled energy unit to estimate the charging actions that may be performed in the future, and the impact of these actions on the state of the controlled energy unit. This includes information such as the power changes, energy consumption, and whether the control channel is occupied that each action may trigger.

[0045] The pre-defined strategy graph structure is a pre-defined action planning framework used to clarify various charging actions, their execution order, and the dependencies and constraints between them. Nodes in the graph represent specific charging actions, and edges represent the order or dependencies between actions.

[0046] The adaptive strategy graph structure is an optimized graph formed by dynamically adjusting the attributes or weights of nodes and edges based on the original preset strategy graph structure, combined with the current user behavior intent vector, the operating status of the controlled energy unit, and the action prediction results. It can rearrange the execution order of actions, adjust the priority of action selection, or modify the dependencies between actions, thereby forming an action planning framework that better fits the actual operating conditions and user needs.

[0047] After receiving the user's behavioral intent vector and the operating status representation of the controlled energy unit, the system first extracts the user's historical behavioral characteristics, preference patterns, and charging habits from the behavioral intent vector, and then extracts indicators such as real-time power, state of charge, control channel occupancy, and response time from the operating status representation. Based on this information, a set of candidate actions that can be executed at present is generated. For example, the system compares the user's expected charging power level with the upper limit of the power allowed by the controlled energy unit for each action, and directly eliminates actions whose power requests exceed the upper limit. In addition, it combines the user's set target driving range, the vehicle's current state of charge, and the charging increment per unit time under the corresponding charging power to determine whether each action can achieve the target driving range within the preset charging period, and only retains those actions that can meet the driving range requirements within the executable time window. After such constraint filtering and accessibility judgment, a preliminary set of candidate actions that both meet the current operating conditions and satisfy the user's charging needs is finally formed.

[0048] Next, action impact prediction is performed for each candidate action. This is done through numerical simulation or a physical constraint model based on dynamics and energy conservation, simulating power changes, state of charge changes, control channel occupancy, and response time at various future time steps, generating state change curves for the action at different time steps. Then, each state curve is matched with the preset power, energy, and control channel constraints of the controlled energy unit, eliminating actions that do not meet the requirements. Subsequently, a comprehensive analysis is performed combining historical behavior, preference patterns, and charging habits from the user's behavioral intent vector to determine the execution order and priority of actions. Specifically, the power, state of charge, and channel occupancy of candidate actions at each time step are compared with the user's historical behavior to see if they conform to past charging time, power selection, and range requirements. Then, actions are matched with user preference patterns; for example, actions that match these preferences are prioritized if the user tends to charge at night, frequently chooses specific charging stations, or is cost-sensitive. Finally, long-term charging habits are considered, such as fixed charging time periods or target range requirements, prioritizing actions that better fit these habits. Through these analyses, execution priorities are assigned to each candidate action, and the action order is adjusted according to priority, ultimately forming a complete action prediction result.

[0049] After obtaining the action prediction results, the preset strategy graph structure is dynamically optimized based on the user's behavioral intent vector, the operating state representation of the controlled energy unit, and the action prediction results, thereby generating an adaptive strategy graph structure. Specifically, each candidate action in the action prediction results is first mapped to a node in the preset strategy graph structure. The power, state of charge, channel occupancy, and response characteristics of each action at different time steps are analyzed. Then, combined with historical behavioral characteristics, preference patterns, and charging habits in the behavioral intent vector, the execution order and priority of actions are evaluated. Next, potential constraint conflicts or low-priority action nodes in the strategy graph are identified, such as power over-limit, insufficient state of charge, or channel occupancy conflicts, and these nodes are marked or reordered. Following this, the dependencies between nodes are adjusted based on the action prediction results and operating state representation to ensure that the action sequence is coherent and executable, while also satisfying user preferences and habits as much as possible. If there are multiple executable action paths for the same node, the path that best matches the user's intent, optimizes energy utilization, and minimizes resource conflicts is selected. Finally, the order of nodes, dependencies, and action selection are updated as a whole to generate the adaptive strategy graph structure.

[0050] Based on the above technical solution, optionally, action impact prediction can be performed based on behavioral intention vectors and operational state representations to obtain action prediction results, including:

[0051] Based on the behavioral intent vector, the operating state representation, and the preset multi-objective constraints, the candidate actions and their parameter ranges corresponding to the behavioral intent of the controlled energy unit under the current operating conditions are determined, and the set of feasible action states of the controlled energy unit is determined based on each candidate action and its parameter range.

[0052] Based on the set of feasible action states, the control object and its action parameters corresponding to each candidate action within the controlled energy unit are determined, and each candidate action and its corresponding control object and action parameters are represented in a structured manner to form the action relationship description information of each candidate action.

[0053] Based on the description of the interaction relationship and the representation of the operating state, a single-step causal inference is performed on the state changes caused by each candidate action under the current operating conditions to obtain the state change results corresponding to each candidate action.

[0054] Based on the state change results and the preset prediction time range, the state changes of the controlled energy unit under each candidate action are simulated in multiple steps to obtain the action prediction results.

[0055] In this scheme, the pre-defined multi-objective constraints are a set of rules established before formal operation, based on the safety standards, stable operation requirements, and maximum operating capacity of the controlled energy unit. These rules must simultaneously govern key dimensions such as power, energy, control channel occupancy, and response sequence, clearly defining the boundaries of actions and related parameter values.

[0056] Current operating conditions refer to the actual working state of the controlled energy unit at the moment of the predicted action, providing a direct reflection of the energy unit's operational status. It corresponds to the operating status representation and generally includes real-time power output, remaining power, whether the control channel is occupied, and the unit's current ability to respond to control requests.

[0057] Candidate actions are a set of actions that, under the current operating conditions, conform to the user's behavioral intent, satisfy preset multi-objective constraints, and are logically and practically feasible for the controlled energy unit. Each candidate action is a specific control method, such as charging at a specific power level during a certain time period or adjusting the energy output.

[0058] The parameter range is the defined interval of control parameters for each candidate action. It clarifies the extent to which the action can be adjusted and what boundary limitations exist. For example, the power level, duration, energy change, and adjustment speed are all specified within the parameter range.

[0059] The feasible state set is a set of states that a controlled energy unit can safely reach or stabilize in, calculated based on current operating conditions, candidate actions, and their parameter ranges. It primarily describes the states a controlled energy unit might be in after executing compliant actions.

[0060] The controlled object is a specific operating component or control node within the controlled energy unit that will directly respond to an action command. Examples include power regulation circuits, energy conversion components, control channels, and operating mode switching units.

[0061] Action parameters are the specific control values ​​given to the corresponding controlled object when a candidate action is executed. They explain how the action affects the controlled object and the magnitude of that effect. Examples include the target power, which control channel to use, the duration of the action, and the adjustment speed.

[0062] Action relationship description information clearly explains the correspondence between candidate actions, controlled objects, and action parameters in a structured way. The core is to clearly explain which controlled objects each action will act on and what parameters will be used to control them.

[0063] State change results refer to the state changes that occur in a controlled energy unit within a unit of time after executing a candidate action under current operating conditions. These changes typically manifest as variations in power output, energy consumption or increase, changes in remaining charge, and changes in control channel occupancy. These changes directly reflect the immediate impact of the action.

[0064] The preset prediction time range is the length of the time window for extrapolating the state changes of the controlled energy unit when predicting the impact of the action. This time range can be set to a continuous duration.

[0065] First, the behavioral intent vector is broken down to extract key intent elements, including users' historical charging behavior characteristics, preference patterns, and charging habits. After decomposition, these intent elements are transformed into unified numerical or categorical fields, creating a structured intent element table to facilitate subsequent action matching and analysis. For example, "nighttime charging preference" is mapped to specific time period labels, and "target 80% battery life" is transformed into a numerical range of state of charge.

[0066] Next, key indicators such as real-time power output, state of charge, control channel occupancy, and adjustable margin of the controlled energy unit are extracted from the operating status representation and organized into a status field table. Then, the intent element table, status field table, and preset multi-objective constraints are compared and logically matched one by one to filter out a preliminary set of candidate actions. When filtering actions, both the user's behavioral intent and these constraints must be met simultaneously. For example, if the user wants to charge at 50 kW, but the maximum allowable power of the controlled energy unit is only 40 kW, then actions exceeding this limit are directly eliminated. If the target range corresponds to an 80% state of charge, the current state of charge is 60%, and the charging rate is 20 kW, then the achievable state of charge range is calculated to be 60%-80%, and actions exceeding this range are also eliminated. If the preset constraints clearly state that the load cannot exceed the total power limit during a specific period, then actions related to that period will be adjusted or deleted. For time continuity requirements, candidate actions are sorted according to the user's preferred time period, while actions that conflict with time constraints are eliminated. This yields candidate actions and their parameter ranges that simultaneously meet the user's intent and constraints under the current operating conditions.

[0067] After identifying candidate actions, the nature and requirements of each action are analyzed to find its corresponding control object within the controlled energy unit. Specifically, the required power level, charging time period, number of channels, and other attributes of the action are clarified. These attributes are then matched with the currently available hardware resources of the controlled energy unit, such as the charging channel number, adjustable power modules, and adjustable time windows. All available channels and power units are checked one by one, and those that can meet the power and time requirements of the action are selected as the control object. For example, if a candidate action requires charging at 30 kW for 20 minutes at night, and both channels 1 and 2 are idle, but channel 1 has a maximum power of 35 kW and channel 2 only has 25 kW, then channel 1 will be selected as the control object because channel 2 cannot meet the power requirement.

[0068] Once the controlled object is identified, the action parameters are calculated based on the action requirements and the actual capabilities of the controlled energy unit. These parameters include power output, duration, and channel occupancy. The calculation first considers the specified power and time for the action, then adjusts these parameters in conjunction with the channel's available power limit and charging rate to ensure the action can be executed within a feasible range. Simultaneously, channel occupancy and time periods are set based on channel availability to avoid conflicts with other actions. For example, for the action "charging 30 kW for 20 minutes at night," the available power of the selected channel is first confirmed to be no less than 30 kW. Then, the action parameters are set to power of 30 kW, duration of 20 minutes, and channel occupancy is recorded as single-channel occupancy.

[0069] Next, the actions, controlled objects, and action parameters are structurally correlated to form action relationship description information. Then, combining the action relationship description information with the operating state representation of the controlled energy unit, a single-step causal inference is performed on the state changes that each candidate action may cause under the current operating conditions. During the inference, the action parameters of the candidate action are first read, such as power output, duration, and channel occupancy, and key indicators such as real-time power, state of charge, channel occupancy, and adjustable margin are retrieved from the operating state representation. Then, using physical constraint models, energy conservation models, or numerical simulation methods based on historical data, the specific changes in power output, state of charge, and channel occupancy after one step of the action are calculated, generating a single-step state change curve, and recording the various state indicators after the action is completed as the single-step state change result.

[0070] After obtaining the single-step state change results for each candidate action, a multi-step time-series simulation is performed to predict the future impact of the action. During the simulation, the single-step state change results are used as initial conditions. Following a preset prediction time range and time step, the power output, state of charge change, and channel occupancy of the candidate action are calculated step-by-step for each future time step. Each step uses the state of the previous step as input, combining the action parameters and operational constraints in the interaction description information to determine whether the action can still be executed and whether parameter adjustments are needed. Simultaneously, it checks for potential power overruns, state of charge overruns, or channel conflicts, correcting actions or marking them as infeasible if necessary. This iterative process continues until the preset prediction time range is covered, such as the entire charging cycle or a specified prediction window. After the multi-step simulation, the state change trajectory of each candidate action within the preset prediction time range is compiled into an action prediction result, including power output over time curves, state of charge change curves, channel occupancy over time changes, and indicators of action execution feasibility.

[0071] This solution can accurately match user intent with energy unit status and constraints, filter actions and clarify parameters, predict their short-term and long-term impacts, form reliable results, and ensure efficient and safe charging.

[0072] Based on the above technical solution, optionally, an adaptive strategy graph structure is obtained by dynamically optimizing the preset strategy graph structure based on the behavioral intent vector, the running state representation, and the action prediction results, including:

[0073] Under the pre-defined multi-objective constraints, the non-executable actions in the pre-defined strategy graph structure are filtered based on the action prediction results, behavioral intention vectors, and running state representations to obtain a set of available action nodes;

[0074] Calculate the numerical impact of each action in the available action node set on the preset multi-objective indicators, and obtain the multi-objective indicator data of each action under the current operating conditions;

[0075] Based on multi-objective index data, the node actions in the preset strategy graph structure are prioritized and optimized to obtain an adaptive strategy graph structure.

[0076] In this solution, an unexecutable action is one that does not conform to system rules or user needs, based on the current user's charging intention, the energy unit's operating status, and the predicted action effect. Examples include exceeding the energy unit's power limit or failing to reach the target charging range.

[0077] The set of available action nodes is formed by filtering actions and retaining those that can still be executed under the current conditions. Each node represents a specific action and includes its parameter range. These nodes form the basis for subsequent optimization strategies and action sequence generation.

[0078] Preset multi-objective metrics are evaluation criteria considered when optimizing the strategy graph structure or the execution sequence of actions. Examples include charging power consumption, charging time, whether it achieves the user's desired battery life, cost sensitivity, and control channel occupancy. These are standards used to quantify the effectiveness of actions or strategies, such as power output, energy utilization efficiency, completion time, and cost; their core function is to determine whether an action is good or bad. Preset multi-objective constraints, on the other hand, are restrictions set by the system that must be followed, such as power not exceeding the upper limit, state of charge within a specified range, and control channel occupancy not conflicting. Their key function is to determine whether an action can be executed. In short, metrics assess "goodness" and constraints assess "feasibility."

[0079] Multi-objective index data is the specific value calculated for each available action node according to the preset multi-objective index. It shows the performance of each action on various indicators.

[0080] After obtaining the action prediction results, behavioral intent vectors, and operational status representations, each action node in the preset strategy graph structure is verified one by one. Specifically, referring to the multi-step temporal state trajectory in the action prediction results, the power changes, state of charge changes, and control channel occupancy status for each action in the future are extracted. These change trajectories are then matched with preset multi-objective constraints using rule matching and threshold judgment to identify actions that cannot be executed at the physical or logical level. At the same time, combined with the current power adjustment margin, the number of available control channels, and the equipment operating boundaries in the operational status representation, it is further confirmed whether the action has a basis for execution at the present moment.

[0081] In addition to this, the rationality of the action execution needs to be further judged by combining the user's historical behavior characteristics, preference patterns, and charging habits in the behavioral intent vector. For example, check whether the action exceeds the user's usual charging time window, deviates from the user's commonly used charging power range, and whether it can meet the user's unspoken but default target range expectation. If an action exceeds the power limit, exceeds the safe state of charge range, conflicts with the mutual exclusion constraint of the control channel, or is obviously inconsistent with the user's behavioral intent at any prediction time step, it will be marked as an unexecutable action and deleted from the strategy graph; the remaining actions that meet both the operational constraints and the behavioral intent will be removed. Figure 1 Consistent action nodes are retained to form a set of available action nodes. For example, if an action predicts a continuous output of 45 kilowatts of power over the next 10 minutes, but the current operating status shows that the maximum available power of the controlled energy unit is only 40 kilowatts, then this action will be determined as an unexecutable action during constraint verification; another action that consistently meets the power and channel constraints during prediction will remain in the set of available action nodes.

[0082] With a set of available action nodes, quantitative calculations are performed on each action across preset multi-objective indicator dimensions. Specifically, based on the state change trajectory in the action prediction results, methods such as energy accumulation calculation, time integration, and cost estimation are used to calculate the numerical impact of the action execution on the operating state of the controlled energy unit, such as cumulative energy change, duration change, power utilization change ratio, and cost change. These numerical impacts are then mapped to preset multi-objective indicator dimensions to obtain specific indicator values, ultimately forming multi-objective indicator data that reflects the current comprehensive performance of each action.

[0083] Then, by combining the user's focus reflected in the behavioral intent vector, the numerical impact of different indicators is normalized, and corresponding weights are assigned to different indicators, allowing for comparative analysis of numerical impacts with different dimensions and emphases within a unified framework. For example, for a nighttime charging action, the total charging energy is calculated by integrating the predicted power curve, the charging time is accumulated over time steps, and then the unit energy cost is calculated by combining the nighttime electricity price. If the user's behavioral intent shows sensitivity to cost, the weight of the cost indicator is increased in the multi-objective indicator data, thus more accurately reflecting the comprehensive value of the action.

[0084] After generating multi-objective metrics data, the action nodes in the strategy graph are re-prioritized. Specifically, multi-objective sorting or weighted scoring methods are used to rank the actions in the available action node set based on their comprehensive scores in metrics such as efficiency, cost, and response speed. Then, the execution order and path weights of the nodes in the strategy graph are adjusted according to the sorting results, prioritizing actions with higher comprehensive scores while reducing the selection probability or execution priority of suboptimal actions. During this process, the sequential dependencies between nodes are also checked simultaneously to ensure that the adjusted strategy graph still satisfies action continuity and constraints, ultimately forming an adaptive strategy graph structure. For example, in the current running state, one action charges quickly but at a high cost, while another charges slightly slower but at a much lower cost. If the user prefers cost-priority, the latter will be made a high-priority node in the strategy graph, making the strategy search more inclined to select this action path.

[0085] In this solution, under the premise of ensuring safety and constraints, the strategy graph is optimized by combining user intent and action prediction, and the best path is selected first to improve the charging decision-making effect.

[0086] Step S104: Based on the adaptive policy graph structure, perform policy search and combination to obtain at least one candidate action sequence, and perform executability verification based on the candidate action sequence to obtain the target action sequence acting on the controlled energy unit.

[0087] Candidate action sequences are combinations of possible actions generated under the guidance of an adaptive policy graph structure, combining user behavior intent vectors, the operating state of the controlled energy unit, and action prediction results. Each sequence contains multiple consecutive actions, each specifying a particular charging operation, such as adjusting charging power, selecting a charging period, or using a control channel.

[0088] The target action sequence is the final action sequence that meets the requirements after the candidate action sequences have undergone executability verification. The verification mainly checks whether it meets the constraints such as the power limit of the controlled energy unit, the state of charge requirement, and the control channel capacity.

[0089] After obtaining the adaptive strategy graph structure, the individual action nodes and their dependencies are expanded to create a set of candidate paths for executable actions. For each path, action information is extracted, such as charging power settings, charging time periods, and required control channels. This information is then combined with state change curves and priority information from the action prediction results for initial sorting. Specifically, based on priority scores from the action prediction results, high-priority actions are placed at the beginning of the sequence, while also considering dependencies between actions, prioritizing actions that must be executed first to ensure logical order. If actions have resource conflicts or overlapping times, their order is adjusted during sorting to prevent interference in time and resource usage. When actions have the same priority, their order is determined by historical execution habits and user preferences, such as fixed charging time periods and frequently selected charging stations. This results in a pre-sorted sequence of candidate actions.

[0090] Next, an executability check will be performed on each candidate action sequence. Each action will be checked individually to see if it can be executed under the current real-time operating state of the controlled energy unit. For example, whether the power output exceeds the limit, whether the change in state of charge can meet the user's objectives, whether there will be conflicts in control channel occupancy, and whether the action response time is within the allowable range. Actions that do not meet the constraints will be deleted or replaced, and finally the target action sequence will be obtained.

[0091] Based on the above technical solution, optionally, policy search and combination can be performed based on the adaptive policy graph structure to obtain at least one candidate action sequence, including:

[0092] Obtain the action information of each node in the adaptive strategy graph structure and the dependencies between nodes. Perform a combined search in the adaptive strategy graph structure based on the action information and the dependencies between nodes to obtain at least two preliminary action sequences.

[0093] The execution order of each action in the preliminary action sequence and the state connection conditions between adjacent actions are checked for consistency to obtain at least one candidate action sequence.

[0094] In this solution, action information is a data set attached to each action node in the adaptive policy graph, used to describe how the action should be executed and what its execution characteristics are. This includes the action type, the object the action affects, the corresponding control parameters, and the resource conditions required to execute the action. This information helps determine whether an action can be selected and how it should be combined with other actions during policy search.

[0095] Inter-node dependencies are the constraints between different action nodes in an adaptive strategy graph. They involve temporal order, state prerequisites, or resource usage, and are used to specify how actions can be connected. They clarify the prerequisite actions or state conditions that must be met before an action can be executed.

[0096] The initial action sequence is the result of an action arrangement obtained by combining action information and dependencies between nodes in the adaptive policy graph, and its continuity has not yet been verified. Multiple action nodes in this sequence are connected according to dependencies, structurally satisfying the path constraints of the policy graph, but state continuity and execution feasibility still need further verification.

[0097] The execution sequence is the chronological arrangement of the action nodes in the initial action sequence, clarifying the actual order in which the actions are executed. It determines the triggering time of actions within the controlled energy unit, as well as the way actions influence each other.

[0098] State connection conditions are requirements for two adjacent actions: the system state at the end of the previous action must be consistent with the state required to start the next action.

[0099] After obtaining the adaptive strategy graph structure, we first traverse all nodes in the graph, extracting the action information corresponding to each node. This action information includes action type, preset power or control parameters, duration, expected energy change, and execution conditions. We also record the inter-node dependencies between each node and other nodes, such as order requirements, resource sharing constraints, and conditional triggering rules. This ensures that the action sequence is feasible and logically coherent during combinatorial search. To organize this information into structured data, we save each action and its dependent nodes in a node-edge graph data format, and also generate an action attribute table and a dependency matrix.

[0100] After obtaining complete action information and inter-node dependencies, a combinatorial search algorithm is used to generate preliminary action sequences. Specifically, according to the topological order of inter-node dependencies in the policy graph, action nodes are combined recursively or iteratively to generate multiple possible action execution paths. Each path records the execution order and timing of nodes, and verifies whether each action meets resource constraints, power limits, and channel occupancy conditions. For example, starting from the initial node in the graph, dependent nodes are selected sequentially to form a sequence. Once a node is selected, the available resources and power budget are updated to ensure that subsequent nodes can execute within the resource limits. In this way, at least two preliminary action sequences are generated, each of which satisfies the policy graph constraints and node dependency conditions, ensuring logical feasibility.

[0101] After obtaining the initial action sequence, a consistency check is performed on each sequence. First, the actual execution order is determined based on the arrangement of action nodes in the sequence and mapped onto the timeline. Then, the state connection conditions of adjacent actions in the sequence are checked one by one. For example, whether the termination power, state of charge, and channel occupancy of the previous action can meet the start requirements of the next action, and whether there are resource conflicts or discontinuous time intervals between actions. If any pair of adjacent actions in a sequence does not meet these state connection conditions, the sequence will be excluded. After the consistency check, at least one action sequence that meets the continuous execution requirements is retained to form the final candidate action sequence for subsequent execution or optimization. For example, the strategy graph has nighttime charging action nodes and morning charging action nodes. When the nighttime action ends, the power drops to 20 kW and the state of charge reaches 50%, while the morning action requires the power to not exceed 25 kW and the state of charge to not be lower than 50%. In this case, the state connection conditions are met, and the two actions can form a candidate action sequence. If the state of charge does not reach 50% when the nighttime action ends, the sequence will be excluded.

[0102] In this solution, action information is extracted from the strategy graph and node dependencies are considered. Multiple action paths are generated through combinatorial search. Candidate action sequences are then selected through state connection and continuity checks to ensure that the sequences meet resource constraints and action logic, thereby improving the reliability of action selection and the efficiency of strategy execution.

[0103] Based on the above technical solution, optionally, an executability verification is performed on the candidate action sequence to obtain the target action sequence acting on the controlled energy unit, including:

[0104] According to the preset control interface constraints, each candidate action sequence is executed sequentially in the simulation environment, and the power output data, state of charge data and control channel occupancy data of the controlled energy unit after each action in each candidate action sequence are recorded.

[0105] If there is a candidate action sequence whose power output data exceeds the preset power output threshold, this candidate action sequence is removed, and the remaining candidate action sequences are taken as candidate action sequences that meet the power constraints.

[0106] If, among the candidate action sequences that meet the power constraints, there is a candidate action sequence whose state of charge data exceeds the preset state of charge threshold, this candidate action sequence will be removed, and the remaining candidate action sequences will be taken as candidate action sequences that meet both the power constraints and the state of charge constraints.

[0107] If, among the candidate action sequences that meet both power constraints and state of charge constraints, there is a candidate action sequence whose control channel occupancy exceeds the preset control channel capacity, this candidate action sequence will be eliminated, and the remaining candidate action sequences will be used as the target action sequences.

[0108] In this solution, the preset control interface constraints are the operational limitations and specifications set for the controlled energy unit's action execution interface. These include, for example, the allowed command types, action triggering conditions, maximum execution frequency, and communication protocol requirements. This ensures that when executing candidate action sequences, the sent commands and the actual execution process do not exceed the operational range acceptable to the device.

[0109] A simulation environment is a virtual or simulated scenario used to test and validate candidate action sequences. It includes the power output model, state-of-charge model, and control channel model of the controlled energy unit.

[0110] Power output data is the instantaneous or average power output value of the controlled energy unit recorded during the simulation of each candidate action.

[0111] State of charge (SCC) data is the record of how the state of charge of a controlled energy unit (battery or energy storage unit) changes over time during the simulation of each candidate action.

[0112] Control channel occupancy data is the occupancy status of the control channels within the controlled energy unit recorded when each candidate action is simulated.

[0113] The preset power output threshold is the upper limit set for the power output of the controlled energy unit. Exceeding this value may damage the equipment or violate operating safety regulations.

[0114] The preset state of charge threshold is a safe state of charge range defined for a battery or energy storage unit, including the maximum charging state of charge and the minimum discharging state of charge.

[0115] The preset control channel capacity is the maximum number of control channels that can be occupied simultaneously within the controlled energy unit, such as the upper limit of the number of charging channels or the upper limit of execution resources.

[0116] Upon receiving the candidate action sequence, the system first executes each sequence sequentially within the simulation environment according to preset control interface constraints. Specifically, the candidate action sequence is first mapped to the controlled energy unit model in the simulation environment, including the power output model, battery state of charge model, and control channel model. Then, control commands are issued step by step according to the action execution order. During execution, the power output data, state of charge data, and control channel occupancy data after each action are completed are recorded.

[0117] Next, sequences are filtered according to power constraints. The recorded power output data is compared step by step with the preset power output threshold. If the power output of any action in a candidate action sequence exceeds the allowable threshold, the sequence is determined to be non-compliant with the power constraints and is directly eliminated. The remaining sequences are the candidate action sequences that meet the power constraints. Then, a state of charge constraint check is performed on these sequences. The state of charge data after the execution of each action in the sequence is compared with the preset state of charge threshold. If any value exceeds the maximum state of charge or falls below the minimum state of charge, the sequence is eliminated. The remaining sequences are the candidate action sequences that meet both power and state of charge constraints.

[0118] Finally, control channel occupancy constraint verification is performed on sequences that meet the first two types of constraints. Specifically, the channel usage of each action in each sequence on the time axis is analyzed, and the control channel occupancy data is compared with the preset control channel capacity. If a sequence is found to exceed the channel capacity limit at any time step, that sequence is removed. After these three levels of constraint filtering, the remaining candidate action sequences are the target action sequences.

[0119] In this scheme, through hierarchical constraint screening and simulation verification, the final target action sequence meets the system's safe operation requirements in terms of power, state of charge, and control channel usage, thereby improving the reliability and controllability of action execution and reducing risks and resource conflicts in actual operation.

[0120] Based on the above technical solution, optionally, if there are candidate action sequences whose power output data exceeds a preset power output threshold, after eliminating these candidate action sequences, the method further includes:

[0121] If there are no remaining candidate action sequences, the preset minimum action sequence will be used as the target action sequence.

[0122] Accordingly, if, among the candidate action sequences that meet the power constraints, there exists a candidate action sequence whose state of charge data exceeds a preset state of charge threshold, after removing this candidate action sequence, the method further includes:

[0123] If there are no remaining candidate action sequences, the preset minimum action sequence will be used as the target action sequence.

[0124] Accordingly, if, among the candidate action sequences that meet both power constraints and state of charge constraints, there exists a candidate action sequence whose control channel occupancy exceeds the preset control channel capacity, after removing this candidate action sequence, the method further includes:

[0125] If there are no remaining candidate action sequences, the preset minimum action sequence will be used as the target action sequence.

[0126] In this scheme, the preset backup action sequence is a standard or safety action sequence pre-set by the system. If all candidate action sequences are eliminated due to constraints, this sequence can be used to ensure the safe and stable operation of the controlled energy unit and to complete the basic tasks.

[0127] If, after any step in the selection process of candidate action sequences, there are no remaining candidate action sequences, then the backup action sequence is taken as the final target action sequence and replaces the original candidate action sequence.

[0128] In this scheme, even if all candidate action sequences are eliminated due to constraints, the controlled energy unit can still operate safely and stably and complete the minimum charging or control tasks, thereby avoiding system stagnation or safety risks.

[0129] Step S105: Determine the target vehicle based on the one-click charging command, perform action filtering and mapping on the target action sequence based on the target vehicle to obtain the charging control action acting on the target vehicle, and issue control commands to the controlled energy unit according to the charging control action to provide charging energy to the target vehicle.

[0130] The target vehicle is the specific vehicle that needs to be charged, as specified by the user through a one-click charging command.

[0131] Charging control actions are specific charging operations performed on a target vehicle, such as setting the charging power, scheduling charging time periods, and selecting the charging mode. These actions are filtered and mapped from the target action sequence, ensuring that the operation meets both user needs and constraints.

[0132] Control commands are specific execution instructions issued to the controlled energy unit to implement charging control actions and enable the target vehicle to actually begin charging. They include information such as action parameters, execution time, and sequence.

[0133] Upon receiving a one-click charging command from a user, the system first parses the command to extract the target vehicle information, including the vehicle's unique identifier, location, and the user-defined desired charging time and method. Then, it compares the target vehicle information with the existing vehicle management database to confirm whether the vehicle can be charged. Simultaneously, it acquires real-time vehicle status data, such as current state of charge, connection status, charging interface type, and maximum acceptable power range.

[0134] Once the target vehicle and its status are identified, the target action sequence is filtered and mapped. First, each action in the target action sequence is matched with the real-time status and characteristics of the target vehicle. For example, it checks whether the charging power of the action is within the vehicle's allowable range, whether the timing of the action matches the user's desired time, and whether the control channel required for the action is idle. Next, combining historical behavioral characteristics, preference patterns, and charging habits from the user's behavioral intent vector, executable actions are prioritized. For example, actions that prioritize nighttime charging, frequently used charging stations, or actions that align with the range target are given priority. Through this comprehensive matching and prioritization, the target action sequence is mapped into charging control actions specifically suited to the target vehicle.

[0135] After generating charging control actions, these actions are converted into executable control commands. These commands contain information such as specific charging power, charging time period, charging mode, and action execution sequence. The control commands are then sent to the controlled energy unit. Upon receiving the commands, this unit schedules energy resources to perform the actual charging operation on the target vehicle. For example, a target action sequence exists, containing actions to charge the vehicle at a specified power during different time periods at night and morning. Upon receiving the vehicle's one-click charging command, the system first confirms that the target vehicle is vehicle A, then obtains vehicle A's current state of charge and maximum allowable power. Next, the target action sequence is matched with vehicle A's actual state, removing actions that exceed the power limit, adjusting the charging time period based on vehicle A's situation, and retaining actions that match the user's usual charging preferences. This forms a charging control sequence specifically for vehicle A. These charging control actions are then converted into specific control commands and sent to the controlled energy unit. Upon receiving the commands, the controlled energy unit begins charging vehicle A.

[0136] Before executing a one-click charging command, the electric vehicle and the controlled energy unit have already established an identifiable and schedulable association within the system. This association describes the logical correspondence between the electric vehicle and at least one controlled energy unit that can supply it, without requiring them to be physically connected or in a charging state. For example, this system-level association can be established through user account binding, the correspondence between the vehicle's unique identifier and the charging device's identifier, historical charging records, reservation information, or a list of controllable energy units at the charging station where the vehicle is located. In this way, when the system receives a one-click charging command, it can clearly identify the target vehicle and which controlled energy units can supply it.

[0137] In actual use, when a user remotely sends a one-click charging command for their vehicle, the electric vehicle does not need to be plugged into the charging gun. This command simply expresses the user's desire to charge. Upon receiving the command, the system first performs an executability check, verifying whether the target vehicle and the controlled energy unit are connected. If they are not connected, and the user agrees to schedule or pre-start charging, the system stores this charging request in a pending state, generating a "waiting for connection—automatic start after connection" action flow, while continuously identifying the conditions for plugging in the charging gun or charging readiness. If a connection is detected, the system directly generates and issues a control command to start charging. In this way, even if the vehicle is not yet connected to the charging equipment, the system can respond appropriately to the remote command and automatically complete the charging control once the conditions are met.

[0138] Based on the above steps S101-S105, the charging strategy is dynamically optimized by combining user intent and real-time operating status, and executable control commands are generated for the target vehicle, thereby improving the matching degree and execution efficiency of charging.

[0139] See appendix Figure 2 , Figure 2 This is a schematic flowchart illustrating the charging anomaly handling steps of a remotely controlled intelligent one-button charging method for vehicles according to an embodiment of the present invention. Figure 2 As shown, a remotely controlled intelligent one-button charging method for vehicles in an embodiment of the present invention mainly includes the following steps S201-S209.

[0140] Step S201: If a one-click charging command for the vehicle is received remotely from the user, the user's historical behavior, preference patterns, and charging habits are obtained. Based on the user's historical behavior, preference patterns, and charging habits, the user's intent is inferred to obtain the user's behavioral intent vector.

[0141] Step S202: Obtain real-time operating status data of the controlled energy unit, and perform deep characterization processing based on the real-time operating status data to obtain an operating status representation.

[0142] Step S203: Based on the behavioral intention vector and the running state representation, predict the action impact to obtain the action prediction result. Based on the behavioral intention vector, the running state representation and the action prediction result, dynamically optimize the preset strategy graph structure to obtain an adaptive strategy graph structure.

[0143] Step S204: Based on the adaptive policy graph structure, perform policy search and combination to obtain at least one candidate action sequence, and perform executability verification based on the candidate action sequence to obtain the target action sequence acting on the controlled energy unit.

[0144] Step S205: Determine the target vehicle based on the one-click charging command, perform action filtering and mapping on the target action sequence based on the target vehicle to obtain the charging control action acting on the target vehicle, and issue control commands to the controlled energy unit according to the charging control action to provide charging energy to the target vehicle.

[0145] Step S206: Continuously collect execution feedback data from the controlled energy unit, and extract charging execution features that reflect the execution status of charging control actions based on the execution feedback data.

[0146] Execution feedback data refers to the real-time status and performance data generated by the controlled energy unit when performing charging control actions. This includes information such as power output, state of charge changes, whether the control channel is occupied, charging time, and response delays.

[0147] Charging execution characteristics are key indicators selected from execution feedback data to describe the execution of charging control actions. Examples include average charging power, actual charging time, change in state of charge, and energy utilization efficiency.

[0148] Once the charging control process begins, the operating status of the controlled energy unit is collected in real time. Specifically, data can be read at pre-defined time intervals or when triggered by specific events. The data read includes the energy unit's power output, state of charge, control channel occupancy, voltage, current, and response delay. Each data point is then timestamped and stored sequentially, forming a continuous execution feedback data stream. To ensure data reliability and usability, outliers, noise, and sensor drift must be addressed during collection. This is achieved through methods such as moving averages and Kalman filtering for filtering and correction. If any data is missing, it is supplemented using interpolation or padding.

[0149] Once continuous and properly processed execution feedback data is obtained, charging execution characteristics can be extracted. First, the power output data is integrated or averaged to obtain the average charging power, peak power, and power fluctuation range. Then, the increment of the state of charge change is analyzed to calculate how much electricity was actually charged, how fast the state of charge changed, and the charging efficiency. The occupancy of the control channels is statistically analyzed to summarize the occupancy time, the number of channels used, and whether channel conflicts occurred. Finally, combined with voltage, current, and response delay data, the response of the action, operational stability, and load adaptability are calculated.

[0150] After adjusting these features to a unified dimension, different weights can be assigned to different features based on user preferences or desired optimization goals, ultimately integrating them into structured charging execution features. For example, during nighttime charging, power output and state of charge changes are collected every second. After processing with a moving average filter, the average power is calculated to be 30 kW, with a total charging of 12 kWh. Simultaneously, the control channel occupancy rate is calculated to be 80%, and the response latency is 0.2 seconds. These data, when combined, constitute the charging execution features describing the effectiveness of this nighttime charging action.

[0151] Step S207: Perform a consistency check between the charging execution features and the preset operating constraints. If there is a situation where the preset operating constraints are not met, it is determined that there is a charging execution abnormality.

[0152] Pre-defined operating constraints are various restrictions and rules set in advance to ensure safety, prevent malfunctions, and maintain stable performance during charging control. These include things like the maximum power limit, the maximum and minimum state of charge, the maximum load the control channel can handle, the charging speed not being too fast or too slow, the safe voltage and current ranges, the unacceptable slowness of the response, and the specified charging time periods.

[0153] When verifying charging execution characteristics, each indicator in the feature vector must be compared with its corresponding preset operating constraints. Specifically, the average power and peak power are compared with the power upper limit, the rate of change of state of charge is compared with the safety range requirements, and the channel occupancy rate is compared with the control channel capacity limit. Then, for each feature indicator, the difference between the actual value and the constraint is calculated. This difference can be calculated as a specific number, a percentage, or expressed as a standardized fraction. For example, if the maximum actual output power is 42 kW, but the specified upper limit is 40 kW, then the difference is 2 kW, or 5% over the limit.

[0154] Then, based on this difference and the pre-defined allowable deviation range, the compliance of each indicator is determined. If any indicator exceeds the allowable range, it means that the preset operational constraints are not met, and an anomaly record is created, specifying which indicator is problematic, how much it exceeds the limit, and when it occurred. If any execution feature fails to meet the constraints, it is determined that this action or action sequence has encountered an anomaly during charging, and it is directly marked as an anomaly.

[0155] Step S208: Obtain the real-time status information of the controlled energy unit, and select the starting action that matches the real-time status information from the preset backup action sequence based on the real-time status information of the controlled energy unit.

[0156] Real-time status information refers to the current operating status data of the controlled energy unit, which directly reflects what resources are available and what operating conditions it has at this moment. For example, it includes the current output power, state of charge, number of available control channels, voltage and current levels, and the limits of the equipment's operation.

[0157] The initial action is the first action in the preset backup action sequence that best matches the current real-time state of the controlled energy unit, serving as the beginning of the entire sequence execution.

[0158] It is necessary to continuously collect real-time status information of the controlled energy unit to determine its current available resources and operating conditions. Specifically, this involves sampling and real-time monitoring of key indicators such as power output, state of charge, number of available control channels, voltage and current values, and equipment operating limits. Power output data is acquired through real-time power sensors or energy monitoring interfaces, and then filtered and denoised to remove the impact of instantaneous fluctuations. State of charge data is calculated by the battery management system through charge counting and voltage curves, and then improved in accuracy through state fusion correction. The occupancy status of control channels is obtained through signal monitoring or scheduling records, reflecting which channels are currently available for action. Voltage, current, and other boundary parameters, after being collected by sensors or control interfaces, are normalized to form data in a unified format.

[0159] Once complete real-time status information is obtained, it is matched against the execution requirements of each action in the preset backup action sequence. Specifically, parameters such as the required power level, state of charge range, channel occupancy requirement, and duration for each action in the backup action sequence are extracted. These parameters are then compared one by one with the real-time status data to calculate the differences or degree of matching. During matching, methods such as Euclidean distance, weighted scoring, or logical condition judgment can be used to quantify whether each action can be executed in the current state.

[0160] Then, based on the matching score or feasibility assessment, the action that best matches the real-time status information is selected from the backup action sequence as the starting action. The selection must meet the following criteria: the power required by the action cannot exceed the currently available power, the state of charge must be within the allowable range, and the channel occupancy cannot exceed the number of available channels. Actions that match historical execution patterns and user preferences are prioritized to ensure a smooth start-up of the action sequence. Once the starting action is selected, it becomes the first action, and subsequent actions are executed sequentially according to the backup action sequence. This ensures the safe and stable operation of the controlled energy unit even without other candidate actions.

[0161] Step S209: Starting from the initial action, control commands are sent to the controlled energy unit in the order of the preset guaranteed action sequence to continue to provide charging energy to the target vehicle.

[0162] Once the initial action is determined, control commands are issued to the controlled energy unit step by step according to the preset sequence of backup actions, starting with this action to ensure the target vehicle can continue charging. Specifically, the execution parameters of the initial action are first read, such as the target power, charging time period, required control channels, and related constraints such as upper and lower limits of state of charge and voltage and current limits. These parameters are then converted into control command formats that the controlled energy unit can recognize, such as PWM signal amplitude, communication protocol command frames, or power setting commands. During the conversion process, parameter verification and unit standardization are performed to ensure that the commands accurately convey the action requirements.

[0163] The control commands must be issued in the order of the backup action sequence. For each action, the current state of the controlled energy unit is first checked to see if the power output, state of charge, and channel occupancy meet the activation conditions of the action. If the conditions are met, the control command is sent to the controlled energy unit, and the time and parameter values ​​of the command are recorded. Then, a timer or a dedicated process is started to record the execution progress of the action. During the execution of the action, feedback data from the controlled energy unit is continuously collected, such as actual output power, changes in state of charge, whether the channel is occupied, voltage and current values, and response speed. This data is then compared with the action requirements in real time to determine whether the action is being executed as expected. If any deviation or abnormality is found, the parameters of the subsequent actions are adjusted, or the issuance of the command for the next action is delayed to ensure that the charging process can proceed stably. After the current action is completed, the next action is selected in the order of the backup action sequence, and the previous parameter conversion, command issuance, status monitoring, and feedback verification steps are repeated until all actions in the backup action sequence are executed.

[0164] In this embodiment, the charging process can be monitored and controlled in a closed loop in real time: abnormalities can be identified in a timely manner through feedback data, and continuous charging safety can be ensured by relying on the backup sequence, thereby improving the overall reliability and controllability.

[0165] It should be noted that although the steps in the above embodiments are described in a specific order, those skilled in the art will understand that in order to achieve the effects of the present invention, different steps do not necessarily have to be executed in such an order. They can be executed simultaneously (in parallel) or in other orders, and these variations are all within the scope of protection of the present invention.

[0166] Furthermore, the present invention also provides a remotely controlled intelligent one-button charging system for vehicles.

[0167] See appendix Figure 3 , Figure 3This is a main structural block diagram of a remotely controlled intelligent one-button charging system for vehicles according to an embodiment of the present invention. Figure 3 As shown, it specifically includes:

[0168] The instruction receiving module 301 is used to obtain the user's historical behavior, preference patterns and charging habits if it receives a one-click charging instruction for the vehicle sent remotely by the user, and to infer the user's intention based on the user's historical behavior, preference patterns and charging habits to obtain the user's behavioral intention vector.

[0169] The deep characterization module 302 is used to acquire real-time operating status data of the controlled energy unit, and perform deep characterization processing based on the real-time operating status data to obtain an operating status representation.

[0170] The dynamic optimization module 303 is used to predict the impact of actions based on the behavioral intention vector and the running state representation, obtain the action prediction result, and dynamically optimize the preset strategy graph structure based on the behavioral intention vector, the running state representation and the action prediction result to obtain an adaptive strategy graph structure.

[0171] The verification module 304 is used to perform policy search and combination based on the adaptive policy graph structure to obtain at least one candidate action sequence, and to perform executability verification based on the candidate action sequence to obtain the target action sequence acting on the controlled energy unit.

[0172] The charging control module 305 is used to determine the target vehicle according to the vehicle's one-button charging command, perform action filtering and mapping on the target action sequence based on the target vehicle to obtain the charging control action acting on the target vehicle, and issue control commands to the controlled energy unit according to the charging control action to provide charging energy to the target vehicle.

[0173] The remote-controlled intelligent one-button charging system for vehicles provided in this application embodiment can achieve... Figure 1 The various processes implemented in the method implementation examples will not be described again here to avoid repetition.

[0174] Those skilled in the art will understand that all or part of the processes in the method of the above embodiment of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium can include any entity or device capable of carrying the computer program code, a medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc. It should be noted that the content included in the computer-readable storage medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable storage medium does not include electrical carrier signals and telecommunication signals.

[0175] Furthermore, the present invention also provides an electronic device 400, including a processor 401, a memory 402, and a program or instructions stored in the memory 402 and executable on the processor 401. When the program or instructions are executed by the processor 401, they implement the various processes of the above-described embodiment of a remotely controlled vehicle intelligent one-button charging method and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0176] It should be noted that the electronic devices in the embodiments of this application include the mobile electronic devices and non-mobile electronic devices described above.

[0177] Furthermore, the present invention also provides a computer-readable storage medium. In one embodiment of the computer-readable storage medium according to the present invention, the computer-readable storage medium can be configured to store a program for executing a remotely controlled intelligent one-button charging method for a vehicle according to the above-described method embodiments. This program can be loaded and run by a processor to implement the aforementioned remotely controlled intelligent one-button charging method for a vehicle. For ease of explanation, only the parts related to the embodiments of the present invention are shown; for specific technical details not disclosed, please refer to the method section of the embodiments of the present invention. The computer-readable storage medium can be a storage device comprising various electronic devices. Optionally, in the embodiments of the present invention, the computer-readable storage medium is a non-transitory computer-readable storage medium.

[0178] Furthermore, it should be understood that since the various modules are only provided to illustrate the functional units of the device of the present invention, the physical devices corresponding to these modules may be the processor itself, or a part of the processor's software, hardware, or a combination of software and hardware. Therefore, the number of modules shown in the figures is merely illustrative.

[0179] Those skilled in the art will understand that the various modules in the device can be adaptively split or combined. Such splitting or combining of specific modules will not cause the technical solution to deviate from the principles of the present invention; therefore, the technical solutions after splitting or combining will fall within the protection scope of the present invention.

[0180] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A remote-controlled vehicle intelligent one-key charging method, characterized in that, The method comprises: If a vehicle one-key charging instruction remotely sent by a user is received, user historical behaviors, preference patterns and charging habits are acquired, intention inference is performed based on the user historical behaviors, preference patterns and charging habits, and a behavior intention vector of the user is obtained; Real-time running state data of the controlled energy unit is acquired, and deep characterization processing is performed based on the real-time running state data, and a running state representation is obtained; Action influence prediction is performed based on the behavior intention vector and the running state representation, an action prediction result is obtained, the preset strategy graph structure is dynamically optimized based on the behavior intention vector, the running state representation and the action prediction result, and an adaptive strategy graph structure is obtained; wherein, under the preset multi-objective constraint condition, unexecutable actions in the preset strategy graph structure are filtered according to the action prediction result, the behavior intention vector and the running state representation, and a set of available action nodes is obtained; The numerical influence of each action in the set of available action nodes on the preset multi-objective index is calculated, and multi-objective index data of each action under the current running condition is obtained; The node actions in the preset strategy graph structure are prioritized according to the multi-objective index data, and an adaptive strategy graph structure is obtained; Strategy search and combination are performed based on the adaptive strategy graph structure, at least one candidate action sequence is obtained, executability verification is performed based on the candidate action sequence, and a target action sequence acting on the controlled energy unit is obtained; wherein, action information of each node in the adaptive strategy graph structure and the inter-node dependency relationship are acquired, combination search is performed in the adaptive strategy graph structure according to the action information and the inter-node dependency relationship, and at least two preliminary action sequences are obtained; Consistency verification is performed on the execution order of each action in the preliminary action sequence and the state connection condition between adjacent actions, and at least one candidate action sequence is obtained; A target vehicle is determined according to the vehicle one-key charging instruction, action filtering and mapping are performed on the target action sequence based on the target vehicle, a charging control action acting on the target vehicle is obtained, and a control instruction is issued to the controlled energy unit according to the charging control action to provide charging energy to the target vehicle.

2. The method of claim 1, wherein, Wherein, Action influence prediction is performed based on the behavior intention vector and the running state representation, and an action prediction result is obtained, which comprises: According to the behavior intention vector, the running state representation and the preset multi-objective constraint condition, each candidate action and its parameter range corresponding to the behavior intention of the controlled energy unit under the current running condition are determined, and the action feasible state set of the controlled energy unit is determined according to each candidate action and its parameter range; According to the action feasible state set, the control object and its action parameter corresponding to each candidate action inside the controlled energy unit are determined, and each candidate action and the corresponding control object and its action parameter are structured to form the action relationship description information of each candidate action; According to the action relationship description information and the running state representation, the state change caused by each candidate action under the current running condition is single-step causally deduced, and a state change result corresponding to each candidate action is obtained; According to the state change result and a preset prediction time range, state changes of the controlled energy unit under each candidate action are performed multi-step time sequence deduction, and an action prediction result is obtained.

3. The method of claim 1, wherein, In the method, Based on the candidate action sequence, performability verification is performed to obtain a target action sequence acting on the controlled energy unit, including: According to the preset control interface constraint, each candidate action sequence is executed in the simulation environment in turn, and the power output data, the state of charge data and the control channel occupation data of the controlled energy unit after each action in each candidate action sequence is executed are recorded; If there is a candidate action sequence whose power output data exceeds the preset power output threshold, the candidate action sequence is removed, and the remaining candidate action sequences are taken as candidate action sequences that meet the power constraint; If there is a candidate action sequence whose state of charge data exceeds the preset state of charge threshold in the candidate action sequence that meets the power constraint, the candidate action sequence is removed, and the remaining candidate action sequences are taken as candidate action sequences that meet the power constraint and the state of charge constraint; If there is a candidate action sequence whose control channel occupation data exceeds the preset control channel capacity in the candidate action sequence that meets the power constraint and the state of charge constraint, the candidate action sequence is removed, and the remaining candidate action sequence is taken as the target action sequence.

4. The intelligent one-key charging method for a remote-controlled vehicle according to claim 3, characterized in that, In the method, After the candidate action sequence whose power output data exceeds the preset power output threshold is removed, the method further includes: If there is no remaining candidate action sequence, a preset bottom action sequence is taken as the target action sequence; Correspondingly, after the candidate action sequence whose state of charge data exceeds the preset state of charge threshold is removed in the candidate action sequence that meets the power constraint, the method further includes: If there is no remaining candidate action sequence, a preset bottom action sequence is taken as the target action sequence; Correspondingly, after the candidate action sequence whose control channel occupation data exceeds the preset control channel capacity is removed in the candidate action sequence that meets the power constraint and the state of charge constraint, the method further includes: If there is no remaining candidate action sequence, a preset bottom action sequence is taken as the target action sequence.

5. The method of claim 1, wherein, In the method, After the control instruction of the charging control action is issued to the controlled energy unit to provide charging energy to the target vehicle, the method further includes: Continuously collect execution feedback data of the controlled energy unit, extract charging execution features reflecting the execution of the charging control action according to the execution feedback data; Perform consistency verification on the charging execution features and the preset operation constraint condition, and if there is a case that does not meet the preset operation constraint condition, it is determined that there is a charging execution exception; Obtain real-time state information of the controlled energy unit, and select a starting action matching the real-time state information in the preset bottom action sequence according to the real-time state information of the controlled energy unit; From the starting action, control instructions are issued to the controlled energy unit in the order of the preset bottom action sequence to continue providing charging energy to the target vehicle.

6. A remote operated vehicle intelligent one key charging system, characterized in that, The system includes: The instruction receiving module is configured to, if receiving a vehicle one-key charging instruction remotely sent by a user, acquire historical behaviors, preference modes, and charging habits of the user, perform intention inference based on the historical behaviors, the preference modes, and the charging habits of the user, and obtain a behavior intention vector of the user. The deep representation module is configured to acquire real-time running state data of the controlled energy unit, perform deep representation processing based on the real-time running state data, and obtain a running state representation. The dynamic optimization module is configured to perform action influence prediction based on the behavior intention vector and the running state representation, obtain an action prediction result, dynamically optimize a preset strategy graph structure based on the behavior intention vector, the running state representation, and the action prediction result, and obtain an adaptive strategy graph structure. The dynamic optimization includes filtering unexecutable actions in the preset strategy graph structure according to the action prediction result, the behavior intention vector, and the running state representation under a preset multi-objective constraint condition, and obtaining a set of available action nodes. The dynamic optimization module is configured to perform action influence prediction based on the behavior intention vector and the running state representation, obtain an action prediction result, dynamically optimize a preset strategy graph structure based on the behavior intention vector, the running state representation, and the action prediction result, and obtain an adaptive strategy graph structure. The dynamic optimization includes filtering unexecutable actions in the preset strategy graph structure according to the action prediction result, the behavior intention vector, and the running state representation under a preset multi-objective constraint condition, and obtaining a set of available action nodes. The dynamic optimization module is configured to perform action influence prediction based on the behavior intention vector and the running state representation, obtain an action prediction result, dynamically optimize a preset strategy graph structure based on the behavior intention vector, the running state representation, and the action prediction result, and obtain an adaptive strategy graph structure. The dynamic optimization includes filtering unexecutable actions in the preset strategy graph structure according to the action prediction result, the behavior intention vector, and the running state representation under a preset multi-objective constraint condition, and obtaining a set of available action nodes. The charging control module is configured to determine a target vehicle according to the vehicle one-key charging instruction, perform action filtering and mapping on the target action sequence based on the target vehicle, obtain a charging control action acting on the target vehicle, and issue a control instruction to the controlled energy unit according to the charging control action to provide charging energy to the target vehicle. The program or instruction is adapted to be loaded and run by the processor to perform the remote-controlled vehicle intelligent one-key charging method in any one of claims 1 to 5. The program code is adapted to be loaded and run by the processor to perform the remote-controlled vehicle intelligent one-key charging method in any one of claims 1 to 5.

7. An electronic device comprising a processor, a memory, and a program or instructions stored on the memory and executable on the processor, characterized in that, ​ 8. A computer readable storage medium having stored therein a plurality of program codes, characterized in that, ​

Images