Charging station intelligent operation method and system based on charging user behavior characteristics

Abstract

The invention relates to the field of electric vehicle charging facility operation and safety management, in particular to a charging station intelligent operation method and system based on charging user behavior characteristics. A charging station intelligent operation method based on charging user behavior characteristics comprises the following steps: S1, when a vehicle initiates a charging request, obtaining a vehicle identification code of a requesting vehicle; based on the vehicle identification code, thermal management capability parameters of the vehicle are obtained through analysis, and the thermal management capability parameters comprise theoretical maximum continuous charging power and a thermal management type. Through fusion of vehicle hardware thermal management capability identification, user thermal behavior preference mining and multi-agent collaborative reinforcement learning, collaborative optimization of safety, efficiency and personalized experience in a charging network high-heat risk scene is realized, thermal safety management is converted into beforehand prevention and global collaboration, and the safety and efficiency of the charging network are improved. And a complete closed loop from accurate perception and intelligent decision to personalized service is formed.

Description

Technical Field

[0001] This invention relates to the field of electric vehicle charging facility operation and safety management, and in particular to a smart operation method and system for charging stations based on the behavioral characteristics of charging users. Background Technology

[0002] With the increasing popularity of electric vehicles, the operational efficiency and safety of charging networks are becoming increasingly prominent issues. Existing charging operation methods mainly rely on historical charging demand distribution, real-time electricity prices, or simple user classification for resource allocation and recommendations. These methods have the following limitations: First, they fail to fully consider the inherent differences in battery thermal management hardware capabilities of different vehicles, such as heat dissipation type and maximum continuous charging power, resulting in a lack of differentiated perception and prevention capabilities for thermal safety risks during the charging process. Second, the analysis of user charging behavior patterns focuses mainly on macro-characteristics such as time, location, and electricity consumption, without deeply exploring users' personalized decision-making preferences when facing thermal risks such as high temperatures, such as their preference for charging speed and the shielding attributes of charging stations, resulting in insufficient matching between operational strategies and users' actual needs. Third, existing multi-charging station collaborative optimization models mostly aim to improve overall utilization or revenue, lacking explicit modeling and collaborative control mechanisms for the key constraint of thermal safety. Therefore, in harsh environments such as high temperatures, existing technologies struggle to achieve a dynamic balance between safety, efficiency, and user experience in charging networks, posing risks of thermal safety hazards and resource misallocation. Summary of the Invention

[0003] To overcome the shortcomings of existing charging operation technologies, which neglect differences in vehicle hardware thermal management, user thermal behavior preferences, and multi-station thermal safety coordination, resulting in insufficient operational safety and inaccurate resource allocation in high-thermal-risk scenarios, this invention provides a smart charging station operation method and system based on charging user behavior characteristics.

[0004] The technical solution of this invention is: a smart operation method for charging stations based on charging user behavior characteristics, comprising the following steps: S1: When a vehicle initiates a charging request, obtain the vehicle identification code of the requesting vehicle; based on the vehicle identification code, parse and obtain the thermal management capability parameters of the vehicle, the thermal management capability parameters including the theoretical maximum continuous charging power and the thermal management type; S2: Obtain the historical charging behavior data of the requested vehicle; construct and train a hidden Markov model based on the historical charging behavior data to infer the thermal behavior preferences of the requested vehicle, and decode to obtain the type of thermal behavior preferences implicit in the request of the requested vehicle. S3: Construct a multi-agent reinforcement learning environment with charging stations in the region as agents; each agent's state space includes its own operating state, current ambient temperature, and a thermal risk situation vector aggregated from the thermal management capability parameters and thermal behavior preference types of vehicles waiting to be charged in the region; each agent outputs operating actions based on its own operating state, current ambient temperature, and thermal risk situation vector; each agent generates station-level guidance strategies through collaborative training. S4: For a single request vehicle, integrate the thermal management capability parameters of the single request vehicle, the decoded thermal behavior preference type and the current ambient temperature, calculate the thermal compatibility score for going to the charging station in the area, and generate and output the thermal compatibility charging guidance sequence in combination with the station-level guidance strategy.

[0005] Preferably, the step of parsing and obtaining the vehicle's thermal management capability parameters based on the vehicle identification code includes: interacting with the manufacturer's data platform to obtain the vehicle's battery type, battery rated capacity, and thermal management configuration information based on the vehicle identification code; determining the thermal management type based on the thermal management configuration information, wherein the type includes active liquid cooling, forced air cooling, and natural cooling; and determining the vehicle's theoretical maximum continuous charging power under standard operating conditions based on the battery type and rated capacity, in conjunction with the battery technical specifications.

[0006] Preferably, the step of constructing and training a hidden Markov model based on the historical charging behavior data to infer the thermal behavior preferences of the requesting vehicle includes: the historical charging behavior data includes the ambient temperature at the time of the historical charging event, the ratio of the actual charging power to the theoretical maximum continuous charging power at the time of the historical charging event, and the physical shading attributes of the charging station at the time of the historical charging event; defining the user's thermal behavior preference hidden state set as... ,in The total number of states. Indicates by A set of hidden states, each representing a user's attitude and behavioral pattern towards charging thermal risks, including at least two of the following: thermal risk indifference, thermal risk avoidance, and thermal sensitivity with speed priority; define the user's observation sequence. ,in The length of the observation sequence represents the total number of historical charging events for the user. Indicates by A sequence of observations, each observation... Including the The features of the historical charging record include the ambient temperature during charging, the ratio of the actual average charging power to the vehicle's theoretical maximum continuous charging power, and the shading attribute identifier of the charging station; the parameters of the hidden Markov model are... ,in The state transition probability matrix is... For the observation probability matrix, The initial state distribution is used; the forward-backward algorithm and the Baum-Welch algorithm are employed, utilizing the observed sequence. For model parameters The model is trained to obtain a personalized hot behavior preference model for users.

[0007] Preferably, the decoding to obtain the hot behavior preference type implicit in the request of the vehicle includes: based on user... Second historical charging observation sequence As input, the Viterbi algorithm is used to calculate the most probable hidden state sequence; the final state of the hidden state sequence is taken as the user's current hot behavior preference type. .

[0008] Preferably, the thermal risk situation vector, which is an aggregation of thermal management capability parameters and thermal behavior preference types of vehicles waiting to be charged within the area, includes: the thermal risk situation vector. It is an intelligent agent The state space is composed of components, each dimension corresponding to a statistical measure of a thermal risk-related attribute; the thermal risk situation vector is constructed as follows: ;in, Indicating in intelligent agents Within the service area, the proportion of vehicles with natural cooling and forced air cooling as the thermal management type to the total number of vehicles; Indicating in intelligent agents Within the service scope, the proportion of users with the "hot risk indifference" behavior preference type to the total number of users; Represents intelligent agents Average ambient temperature within the service area.

[0009] Preferably, each agent generates a station-level guidance strategy through collaborative training, including: the operational actions include guidance strategies directly related to thermal risk control; the operational actions of the guidance strategies directly related to thermal risk control include adjusting the recommended priority of charging piles; generating and sending instruction packages to users containing delayed charging suggestions, transfer guidance information, and incentive compensation; and controlling the active cooling of charging piles and adjusting their operating power; the station-level guidance strategy content includes the adjustment of the recommended priority of charging piles generated based on thermal risk control, the instruction packages sent to users, and the instructions for controlling charging piles.

[0010] Preferably, the construction of a multi-agent reinforcement learning environment with charging stations within the region as agents includes: the region being a configurable operation and management unit; and each agent in the multi-agent reinforcement learning environment... reward function Designed as follows: in, For intelligent agents Benefits during the decision-making cycle; For intelligent agents The average utilization rate of charging piles; , respectively intelligent agents , Load rate; For intelligent agents The set of neighboring agents; For intelligent agents The number of high-risk charging sessions served within the decision-making cycle; a high-risk charging session is defined as a charging request where the vehicle's thermal management type is natural cooling, the user's thermal behavior preference is thermal risk indifference, and the actual requested charging power is higher than the theoretical maximum continuous charging power by a certain proportion. These are preset positive weighting coefficients.

[0011] Preferably, the step of calculating a thermal compatibility score for a single requesting vehicle by integrating the vehicle's thermal management capability parameters, decoded thermal behavior preference type, and current ambient temperature, includes: the thermal compatibility score. The calculation formula is: ;in, As an efficiency factor, and with users Heading to the charging station Estimated driving time and charging stations The current queuing time is negatively correlated; Economic factors, and charging stations The service fees are negatively correlated; The thermal matching factor is calculated using the thermal matching function. The calculation formula is as follows: ;in, Indicates user Vehicle thermal management capability rating Indicates user Hot behavioral preference score Indicates ambient temperature. Indicates charging station The score for the heat dissipation capacity of the pile end. Indicates charging station Current load; function The design aims to give a high matching score to charging stations with strong heat dissipation capabilities and low load when the vehicle's heat dissipation capacity is weak, user preferences are aggressive, and the ambient temperature is high.

[0012] Preferably, the step of generating and outputting the thermal adaptation charging guidance sequence in conjunction with the station-level guidance strategy includes: sorting all candidate charging stations in descending order according to the thermal adaptation score; selecting the top K charging stations to form the guidance sequence; for each charging station option in the sequence, when a station-level guidance strategy is attached, converting the station-level guidance strategy into user-perceptible guidance information and outputting it together with the basic information of the charging station.

[0013] Preferably, the intelligent operation system for charging stations based on charging user behavior characteristics includes: Vehicle thermal profile construction module: When a vehicle initiates a charging request, the vehicle identification code of the requesting vehicle is obtained; based on the vehicle identification code, the thermal management capability parameters of the vehicle are parsed and obtained, including the theoretical maximum continuous charging power and the thermal management type. User Hot Preference Analysis Module: Obtains historical charging behavior data of the requested vehicle; constructs and trains a Hidden Markov Model based on the historical charging behavior data to infer the hot behavior preferences of the requested vehicle, and decodes to obtain the type of hot behavior preferences implicit in the request of the requested vehicle. Thermal Risk Collaborative Decision-Making Module: Constructs a multi-agent reinforcement learning environment with charging stations within the region as agents; each agent's state space includes its own operating state, current ambient temperature, and a thermal risk situation vector aggregated from the thermal management capability parameters and thermal behavior preference types of vehicles waiting to be charged within the region; each agent outputs operational actions based on its own operating state, current ambient temperature, and thermal risk situation vector; each agent generates station-level guidance strategies through collaborative training. Personalized thermal adaptation guidance module: For a single request vehicle, it integrates the thermal management capability parameters of the single request vehicle, the decoded thermal behavior preference type and the current ambient temperature, calculates the thermal adaptation score of the charging station in the area, and generates and outputs the thermal adaptation charging guidance sequence in combination with the station-level guidance strategy. Data communication and interface module: used to respond to charging requests initiated by vehicles, obtain vehicle identification code and vehicle status information from user terminals, obtain operational status data from charging pile equipment, obtain ambient temperature data from meteorological service providers, and distribute vehicle status information, operational status data and ambient temperature data to the corresponding processing processes; at the same time, it receives the thermal adaptation charging guidance sequence and outputs it to the user terminal.

[0014] Beneficial Effects: This invention deeply integrates vehicle engineering data with user behavior analysis to construct a complete technical closed loop from risk identification to collaborative control and personalized guidance. It breaks through the limitations of traditional single hardware parameters by accurately parsing the theoretical maximum continuous charging power and thermal management type using vehicle identification codes. Combined with a Hidden Markov Model, it decodes user thermal behavior preferences from historical charging behavior, thus achieving accurate prediction of potential high-risk charging demands. By constructing a multi-agent reinforcement learning environment with charging stations as the agents, it integrates the aggregated thermal risk situation vector into the state space and encodes the number of high-risk charging sessions into the reward function. This drives each charging station agent to generate a station-level guidance strategy that balances its own benefits with regional thermal load balance during collaborative training, achieving an upgrade from passive response to proactive collaborative prevention. For individual vehicles, it integrates their thermal management capabilities, behavioral preferences, and ambient temperature to calculate thermal adaptability scores. Combined with station-level guidance strategies, it outputs personalized charging guidance sequences, improving user acceptance and operational efficiency while ensuring thermal safety, enabling charging stations to possess adaptive and continuously optimized intelligent operation capabilities. Attached Figure Description

[0015] Figure 1 This is a flowchart of the intelligent operation method for charging stations based on charging user behavior characteristics according to the present invention. Figure 2 This is a schematic diagram of the intelligent operation system for charging stations based on the behavioral characteristics of charging users, as described in this invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1: A smart operation method for charging stations based on charging user behavior characteristics, such as Figure 1 As shown, it includes the following steps: S1: When a vehicle initiates a charging request, obtain the vehicle identification code of the requesting vehicle; based on the vehicle identification code, parse and obtain the thermal management capability parameters of the vehicle, the thermal management capability parameters including the theoretical maximum continuous charging power and the thermal management type; S2: Obtain the historical charging behavior data of the requested vehicle; construct and train a hidden Markov model based on the historical charging behavior data to infer the thermal behavior preferences of the requested vehicle, and decode to obtain the type of thermal behavior preferences implicit in the request of the requested vehicle. S3: Construct a multi-agent reinforcement learning environment with charging stations in the region as agents; each agent's state space includes its own operating state, current ambient temperature, and a thermal risk situation vector aggregated from the thermal management capability parameters and thermal behavior preference types of vehicles waiting to be charged in the region; each agent outputs operating actions based on its own operating state, current ambient temperature, and thermal risk situation vector; each agent generates station-level guidance strategies through collaborative training. S4: For a single request vehicle, integrate the thermal management capability parameters of the single request vehicle, the decoded thermal behavior preference type and the current ambient temperature, calculate the thermal compatibility score for going to the charging station in the area, and generate and output the thermal compatibility charging guidance sequence in combination with the station-level guidance strategy.

[0018] By interacting with the manufacturer's data platform, the battery type, rated capacity, and thermal management configuration information of the vehicle are obtained based on the vehicle identification number; based on the thermal management configuration information, the thermal management type is determined, including active liquid cooling, forced air cooling, and natural cooling; based on the battery type and rated capacity, and in conjunction with the battery technical specifications, the theoretical maximum continuous charging power of the vehicle under standard operating conditions is determined.

[0019] It needs to be explained that after obtaining the vehicle identification number (VIN), the system interacts with the automotive manufacturer's data platform by calling an application programming interface (API). The VIN is used as a unique query identifier to initiate a request. The automotive manufacturer's data platform returns structured vehicle configuration information, including battery type, battery rated capacity, and thermal management configuration description text. This interaction follows the general application programming interface protocol of the vehicle data open platform. The thermal management configuration description text is a string of information describing the vehicle's battery cooling method, and it is parsed according to the automotive manufacturer's pre-defined classification rules. Taking the VIN LSGJC5CM0JG123456 as an example, an HTTPS GET request is initiated to the service address, with the request body containing {"vin": "LSGJC5CM0JG123456"}. A successful response returns JSON format data, which includes fields... The value represents "dual-pump active liquid cooling circulation system"; when the text contains keywords indicating the presence of a liquid cooling circulation pump and direct refrigerant cooling, it is determined to be an active liquid cooling type; when the text contains keywords indicating the presence of a high-power cooling fan but no liquid circulation loop, it is determined to be a forced air cooling type; when the text does not mention any active cooling device and is explicitly described as natural cooling, it is determined to be a natural cooling type; the theoretical maximum continuous charging power is determined by combining the obtained battery type and battery rated capacity and consulting recognized battery technical specifications; the battery technical specifications refer to the technical standard documents on the sustainable charging performance of various power batteries publicly released by the automotive industry association; based on the battery type and the standard continuous charging rate corresponding to the battery type in the specifications, the rate is multiplied by the battery rated capacity to calculate the vehicle's theoretical maximum continuous charging power.

[0020] The historical charging behavior data includes the ambient temperature at the time of the historical charging event, the ratio of the actual charging power to the theoretical maximum continuous charging power at the time of the historical charging event, and the physical shielding attributes of the charging station at the time of the historical charging event; the user's thermal behavior preference latent state set is defined as follows: ,in The total number of states. Indicates by A set of hidden states, each representing a user's attitude and behavioral pattern towards charging thermal risks, including at least two of the following: thermal risk indifference, thermal risk avoidance, and thermal sensitivity with speed priority; define the user's observation sequence. ,in The length of the observation sequence represents the total number of historical charging events for the user. Indicates by A sequence of observations, each observation... Including the The features of the historical charging record include the ambient temperature during charging, the ratio of the actual average charging power to the vehicle's theoretical maximum continuous charging power, and the shading attribute identifier of the charging station; the parameters of the hidden Markov model are... ,in The state transition probability matrix is... For the observation probability matrix, The initial state distribution is used; the forward-backward algorithm and the Baum-Welch algorithm are employed, utilizing the observed sequence. For model parameters The model is trained to obtain a personalized hot behavior preference model for users.

[0021] It should be explained that the determination of each thermal behavior preference type is based on the statistical analysis and quantification of user historical charging behavior data. In the historical charging records of users with thermal risk indifference, the ratio of actual average charging power to theoretical maximum continuous charging power is higher than a threshold. The proportion of charging events exceeds the first preset proportion, and the proportion of charging events where users choose unshaded charging stations when the ambient temperature is higher than the temperature threshold exceeds the second preset proportion; for users who are heat risk averse, the ratio of actual average charging power to theoretical maximum continuous charging power is lower than the threshold. The proportion of charging events exceeds the third preset proportion, and the proportion of charging events where users actively choose charging stations with physical shielding attributes exceeds the fourth preset proportion; the first, second, third, and fourth preset proportions are used to perform unsupervised clustering analysis on user behavior feature vectors based on historical operation databases. The feature vectors include power ratio, shielding selection, and ambient temperature, and the main behavior pattern clusters are identified; the statistical distribution of the corresponding behaviors in each typical pattern cluster is statistically analyzed, and the high probability interval boundary of the distribution is used as the value of each preset proportion; the behavior characteristics of heat-sensitive speed-priority users show pattern switching under different ambient temperatures; The first quantile threshold of the ratio of the actual average charging power to the vehicle's theoretical maximum continuous charging power in events that trigger an overheat safety warning is taken. The second quantile threshold is the ratio of the actual average charging power to the vehicle's theoretical maximum continuous charging power in normal events where no warning is triggered. The specific quantile values ​​in the first and second quantile thresholds are determined by a grid search within a predefined quantile value candidate set on historical data samples, aiming to maximize the accuracy of warning event identification. The optimal quantile values ​​obtained from the search are set as the first and second quantile values, respectively. The user's thermal behavior preference latent state set is a finite set containing M abstract states, each representing a stable internal attitude and decision-making pattern of the user when facing thermal risks during charging. Among them, thermal risk indifference refers to a user's tendency to ignore and be willing to bear the risk of battery overheating in exchange for the fastest charging speed; thermal risk avoidance refers to a user's tendency to take proactive actions to reduce thermal risks, such as choosing charging stations with better heat dissipation conditions or actively reducing charging power; and thermal sensitivity speed priority refers to a user's concern about thermal risks but prioritizing high charging speeds within a controllable risk range. The behavioral patterns of power efficiency; the construction of the observation sequence is based on the processing of users' historical charging records. The observation values ​​are multi-dimensional vectors, including ambient temperature data during charging, which is obtained by associating the timestamp of the charging event with geographical location information. By calling the standardized application programming interface of commercial meteorological service providers, the corresponding historical hourly temperature data is queried based on the timestamp of the charging event and the latitude and longitude of the charging station; the ratio of actual average charging power to the vehicle's theoretical maximum continuous charging power is obtained by calculating the quotient of the total amount of electricity and the total duration of the charging session, and dividing the average power by the obtained theoretical maximum continuous charging power of the vehicle; the shading attribute identifier of the charging station is a categorical variable used to describe the physical conditions of the charging station having sunshade and indoor parking. The shading attribute identifier of the charging station is obtained by querying the charging station database, which is a database maintained by the charging station operator that records detailed facility attributes of the charging station; Hidden Markov Model parameters. From the state transition probability matrix Observation probability matrix and initial state distribution The training process employs the Baum-Welch algorithm, which updates iteratively. Interpreting the observation sequence The first step in the training process is to calculate the expected value using the current parameters. and the entire observation sequence The forward-backward algorithm is used to calculate the first... The first step is to record the probability of a user being in each hidden state in the previous charging history, and the expected probability of a transition between hidden states between adjacent time points; the second step is parameter reestimation, which uses all the expected probability values ​​calculated in the first step to re-evaluate the parameters. Update the state transition probability matrix. Each element is updated with the normalized value of the corresponding state transition expectation count, and the observation probability matrix. Each element is updated to the normalized value of the expected count of the joint occurrence of the corresponding state and observations, with the initial state distribution... The probability expectation of each hidden state at the start of the sequence is updated; the training process is repeated until the parameters are updated. The convergence yields a personalized hot behavior preference model for users.

[0022] With users Second historical charging observation sequence As input, the Viterbi algorithm is used to calculate the most probable hidden state sequence; the final state of the hidden state sequence is taken as the user's current hot behavior preference type. .

[0023] It needs to be explained that the parameters based on the Hidden Markov Model When it is necessary to determine the user's thermal behavior preferences when initiating a charging request, the system obtains the user's most recent N historical charging records. Following the method described in the previous embodiment, each record is converted into an observation value that includes ambient temperature, the ratio of actual average charging power to the vehicle's theoretical maximum continuous charging power, and the shading attribute of the charging station, forming a fixed-length recent observation sequence. N is a preset positive integer used to provide a statistically significant recent behavior time window; based on recent observation sequences As input, based on the parameters of the hidden Markov model The Viterbi algorithm is used to calculate the optimal path among all the hidden state sequences of hot behavior preferences. Starting from the first observation in the sequence, the algorithm recursively calculates the cumulative maximum probability of reaching each time step and being in each hidden state, along with the corresponding preceding state, until all N observations in the sequence have been processed. After the calculation is complete, the entire recent observation sequence is found by backtracking. The corresponding most probable hidden state sequence, where the last hidden state represents the most likely hot behavior preference type after considering the user's recent N charging behavior characteristics, outputs the hot behavior preference type as the user's current hot behavior preference type. .

[0024] The thermal risk situation vector It is an intelligent agent The state space is composed of components, each dimension corresponding to a statistical measure of a thermal risk-related attribute; the thermal risk situation vector is constructed as follows: ;in, Indicating in intelligent agents Within the service area, the proportion of vehicles with natural cooling and forced air cooling as the thermal management type to the total number of vehicles; Indicating in intelligent agents Within the service scope, the proportion of users with the "hot risk indifference" behavior preference type to the total number of users; Represents intelligent agents Average ambient temperature within the service area.

[0025] It needs to be explained that the thermal risk situation vector The construction is based on intelligent agents The real-time service range; the service range is a geographical area centered on the charging station corresponding to the intelligent agent, such as a circular area with a radius of five kilometers; real-time statistics are performed on all vehicles and users whose current location is within the service range and who have initiated a charging request but have not yet started charging; each dimension of the vector is calculated as follows: Dimension 1: Proportion of vehicles with natural cooling type. This refers to the proportion of vehicles waiting to be charged within the service area that have natural cooling and forced air cooling as their thermal management type, as determined by the vehicle identification number (VIN). The data source is the thermal management type obtained from the VIN analysis. Dimension Two: Proportion of users exhibiting "heat risk indifference" behavior. This refers to the proportion of users whose current hot behavior preference type is "heat risk indifference" among the vehicles waiting to be charged within the service area, as determined in real-time using a Hidden Markov Model. The data source is hot behavior preference types. ; Dimension 3: Average ambient temperature. The average ambient temperature within the service area is calculated. The data comes from real-time temperature data obtained from the meteorological service platform through the application programming interface. Based on the geographical location of each vehicle and charging station within the service area, the corresponding temperature readings are obtained and averaged.

[0026] The operational actions include guidance strategies directly related to thermal risk control; the operational actions of the guidance strategies directly related to thermal risk control include adjusting the recommended priority of charging piles; generating and sending instruction packages to users containing suggestions for delayed charging, transfer guidance information, and incentive compensation; and controlling the active cooling of charging piles and adjusting their operating power. The station-level guidance strategy includes adjusting the recommended priority of charging piles based on thermal risk control, sending instruction packages to users, and controlling the charging piles.

[0027] It should be explained that the aforementioned operational actions specifically include the following three types of executable instructions: The first category involves adjusting the recommendation priority of charging piles by setting dynamic weights for different charging piles in the background recommendation algorithm. For example, when the agent determines that the vehicle needs to be guided to use resources with better heat dissipation, the priority weight of charging piles equipped with liquid cooling in the station is increased so that they are ranked higher in the recommendation list generated for the user. The decision is based on the agent's judgment of the current thermal risk situation. The second category involves generating and sending instruction packets to specific user groups. These instruction packets are structured data messages, with content dynamically generated by the intelligent agent based on the control objectives. The instruction packets include delayed charging suggestions, advising users to avoid peak hours; relocation guidance information, recommending users to go to nearby charging stations with lower loads and better heat dissipation; and incentive compensation, such as electronic coupons, to encourage users to follow the guidance. The generation logic of the instruction packets and the selection of the target user group are based on real-time matching of the user's current thermal behavior preferences and the vehicle's thermal management capabilities. The third category is commands to control the activation of active cooling in charging piles. For charging pile smart devices that support active cooling, control commands are directly issued. The commands include activating active cooling and adjusting the operating power level. The issuance of commands is based on the charging pile device's ability to receive and execute remote control protocols.

[0028] The region is a configurable operation and management unit; in the multi-agent reinforcement learning environment, each agent... reward function Designed as follows: ;in, For intelligent agents Benefits during the decision-making cycle; For intelligent agents The average utilization rate of charging piles; , respectively intelligent agents , Load rate; For intelligent agents The set of neighboring agents; For intelligent agents The number of high-risk charging sessions served within the decision-making cycle; a high-risk charging session is defined as a charging request where the vehicle's thermal management type is natural cooling, the user's thermal behavior preference is thermal risk indifference, and the actual requested charging power is higher than the theoretical maximum continuous charging power by a certain proportion. These are preset positive weighting coefficients.

[0029] It needs to be explained that the intelligent agent reward function The composition includes revenue items ,in For intelligent agents The total service fee revenue obtained within a decision-making cycle is based on order settlement data from charging stations; equipment health incentive items. middle, For intelligent agents The thermal damage risk coefficient of charging equipment is calculated by statistical intelligent agents. The ratio of the service time lost by the charging piles under its jurisdiction due to thermal protection mechanisms to the theoretical total available time. Thermal protection mechanisms include active power reduction and forced interruption. The awards recognize and encourage low risk of thermal damage and high equipment health; load balancing items. The penalty is the load difference between the agent and neighboring charging stations, which encourages a smoother load distribution within the area; among which... and respectively intelligent agents and neighboring agents The load factor is two ratios representing the current power consumption to the total installed capacity. This represents the difference in load rates between two adjacent smart agents. The load difference between adjacent charging stations is quantified by calculating the difference. The load difference represents the degree of imbalance in instantaneous operational pressure between adjacent charging stations and measures the uneven distribution of charging service demand in local space. The sum of the squares of the differences is applied with a negative penalty. The larger the difference, the greater the reduction in reward value. For intelligent agents The purpose of setting up an adjacent set is to ensure that the loads of adjacent sites significantly influence each other, making multi-agent reinforcement learning more likely to converge. Predefined by the regional network topology; thermal security penalty item Incorporate thermal safety objectives into the learning framework. For intelligent agents The number of high-risk charging sessions served during the decision-making cycle; the determination is based on three real-time data points: the vehicle's thermal management type obtained from vehicle identification code parsing, and the user's thermal behavior preference type obtained from Hidden Markov Model decoding. The actual power demand requested; a charging session is considered high-risk if and only if the vehicle's thermal management type is natural cooling, the user's thermal behavior preference type is thermal risk indifference, and the requested charging power exceeds the fifth preset ratio threshold of the vehicle's theoretical maximum continuous charging power; the fifth preset ratio threshold is set based on the safety margin guidance value for overcharge protection in the national electric vehicle power battery safety technical specifications, statistically analyzes battery overheating warning events recorded in historical operating data, obtains the percentage distribution of actual requested power exceeding the vehicle's theoretical maximum continuous charging power at the time of occurrence, and, based on the statistical characteristics of the percentage distribution, performs parameter optimization within a preset optimization interval centered on the basic ratio threshold. The optimization objective is to achieve optimal differentiation between warning events and normal charging events in historical data, and to determine the fifth preset ratio threshold; in the formula These are preset positive weighting coefficients used to balance the relative importance of the four objectives—profitability, operational efficiency, regional load balancing, and thermal safety—in the agent's decision-making process. The numerical values ​​are determined by constructing a simulation environment based on historical operational data, and using Bayesian optimization within a predefined value space, such as... The parameters are optimized to maximize long-term cumulative rewards; finally, the optimal combination of parameters obtained through optimization is used as the preset weight coefficient.

[0030] The thermal compatibility score The calculation formula is: ;in, As an efficiency factor, and with users Heading to the charging station Estimated driving time and charging stations The current queuing time is negatively correlated; Economic factors, and charging stations The service fees are negatively correlated; The thermal matching factor is calculated using the thermal matching function. The calculation formula is as follows: ;in, Indicates user Vehicle thermal management capability rating Indicates user Hot behavioral preference score Indicates ambient temperature. Indicates charging station The score for the heat dissipation capacity of the pile end. Indicates charging station Current load; function The design aims to give a high matching score to charging stations with strong heat dissipation capabilities and low load when the vehicle's heat dissipation capacity is weak, user preferences are aggressive, and the ambient temperature is high.

[0031] It needs to be explained that the thermal compatibility score The calculation is a weighted sum of three independent factors, with weight coefficients... The preset values ​​are used to adjust the relative importance of different dimensions. The preset weight coefficients are determined by collecting historical operational data on the charging stations actually selected by users after charging, constructing a feature set for each selection. This feature set includes the efficiency factor, economic factor, and thermal matching factor values ​​of the user's candidate charging stations. Using the user's actual selection as the dependent variable and the efficiency factor, economic factor, and thermal matching factor values ​​of each candidate station as independent variables, a linear regression model is used for analysis. By fitting the historical dataset, regression coefficients for the efficiency factor, economic factor, and thermal matching factor values ​​are obtained. These regression coefficients are then normalized, and the resulting value is the weight coefficient. The three factors are efficiency factor, economic factor, and thermal matching factor; efficiency factor Taking into account time costs, among which users Heading to the charging station The estimated travel time is calculated by calling the route planning application programming interface of commercial navigation maps, based on the user's real-time location, the location of charging stations, and current road conditions; charging stations The current queue time is obtained by querying the real-time operation and management status of the charging station, and is estimated based on the current number of vehicles in the queue and the average charging time; efficiency factor The calculation logic is for the user Heading to the charging station Estimated driving time and charging station The shorter the sum of current queue times, the higher the efficiency factor score; the economic factor... Considering charging costs, charging stations The service fee is dynamically set by the charging station operator and obtained in real time from the charging station pricing information database; the lower the service fee, the higher the economic factor score; heat matching factor Used to assess the relationship between charging stations and users in the current environment. The degree of matching in the thermal safety dimension is determined by the thermal matching function. The heat matching function is calculated. Among the parameters, the vehicle thermal management capability score Mapping is performed based on the thermal management type obtained from vehicle analysis; performance data of the thermal management system under standard temperature rise test conditions is collected, wherein the performance data is the continuous heat dissipation power that the battery can support within the target temperature range; normalization is performed using the highest performance data as a benchmark value, and the performance data of each type of thermal management system is divided by the benchmark value, with the ratio being the corresponding mapping value; user thermal behavior preference scoring is also performed. Based on the user's real-time decoding of thermal behavior preference types Mapping is performed; the mapping value is determined by filtering charging event records that trigger high-temperature warnings based on historical operational data. For users with specific heat behavior preferences, the proportion of events in which users actually choose to reduce charging power or switch to other charging stations after the warning is issued is calculated out of the total number of warning events. This yields the warning response rate for users with this heat behavior preference type. The difference is used as the calibration baseline value. The calibration baseline values ​​of users with different thermal behavior preference types are normalized and determined as the corresponding mapping values; ambient temperature coefficient. Current ambient temperature Normalization was performed, and a reference high-temperature threshold was set with reference to the inflection point temperature of typical power battery high-temperature performance. Calculate at 35°C ; The coefficient value is between 0 and 1, and the higher the ambient temperature... The closer the score is to 1, the greater the thermal environment pressure; pile tip heat dissipation capacity rating According to the charging station The attributes of the charging pile group equipment were determined, and the temperature rise data of key components of different types of charging piles were obtained when they were continuously operated at rated power to thermal equilibrium. The lowest temperature rise data in the type to be evaluated was used as a benchmark for normalization. The temperature rise data of each type of charging pile was divided by the benchmark value, and the reciprocal of the result was the corresponding mapping value. The charging station load factor was also determined. The calculation formula is: ,in This represents the current load rate of the charging station, with a value between 0 and 1. A coefficient value between 0 and 1 indicates a lower charging station load. The closer to 1, the greater the service margin; hot matching factor The calculation formula is: Formula numerator The comprehensive characterization of a charging station's thermal risk absorption capacity, namely its heat dissipation capacity and load margin, is shown in the denominator. The comprehensive characterization of the potential thermal risks that users may generate in the current environment is as follows: the weaker the vehicle's heat dissipation capacity, the more aggressive the user's risk appetite, and the higher the ambient temperature, the greater the potential. The higher the value, the stronger the charging station's absorption capacity relative to the user's risk potential, and the higher the heat matching degree.

[0032] All candidate charging stations are sorted in descending order based on their thermal adaptability scores; the top K charging stations are selected to form a guidance sequence; for each charging station option in the sequence, if a station-level guidance strategy is attached, the station-level guidance strategy is converted into user-perceptible guidance information and output along with the basic information of the charging station.

[0033] It should be explained that candidate charging stations are scored according to their thermal adaptability. The charging stations are sorted in descending order, with the highest-rated charging stations at the top. The maximum number K of the guidance sequence is determined by counting the total number of candidate charging stations with a thermal adaptability score greater than zero under the current request, and taking the smaller value between the total number of candidate charging stations and a preset maximum number of display items. The preset maximum number of display items is determined by statistically analyzing the probability distribution of users' final selections falling at different ranking positions in the recommendation list of different lengths. A minimum list length that can cover the selections of the vast majority of users is selected, and this length is set as the maximum number of display items, where the vast majority is, for example, 95% or more. The top K charging stations are selected from the sorted list to form an ordered recommendation list, which is the final guidance sequence. The station-level guidance strategy is integrated to generate user-side display information for the charging station options in the guidance sequence, and the candidate charging stations are checked. The station-level guidance strategy, when a charging station is detected When a station-level guidance strategy exists, a preset strategy-information conversion rule library is invoked to convert the abstract station-level guidance strategy instructions into natural language descriptions of rights identifiers. The strategy-information conversion rule library is a preset mapping table that maps the standardized strategy instructions generated by the intelligent agent to the corresponding natural language description templates. This is pre-configured by the operator based on the type of guidance strategy and the expected user interaction information. For example, when the strategy instruction is to provide incentive compensation, it is converted into "Recommendation reason: This station provides you with thermal safety management protection and offers an additional charging discount"; when the strategy instruction includes a delayed charging suggestion, it is converted into "Tip: It is currently peak time, and it is recommended to go 30 minutes later for a better experience". The converted personalized guidance information is encapsulated with basic information about the charging station, including location, distance, real-time electricity price, and estimated queuing time, and output to the user's in-vehicle terminal interface.

[0034] Example 2: Based on Example 1, a smart operation system for charging stations based on user behavior characteristics, such as... Figure 2 As shown, it includes: Vehicle thermal profile construction module: When a vehicle initiates a charging request, the vehicle identification code of the requesting vehicle is obtained; based on the vehicle identification code, the thermal management capability parameters of the vehicle are parsed and obtained, including the theoretical maximum continuous charging power and the thermal management type. User Hot Preference Analysis Module: Obtains historical charging behavior data of the requested vehicle; constructs and trains a Hidden Markov Model based on the historical charging behavior data to infer the hot behavior preferences of the requested vehicle, and decodes to obtain the type of hot behavior preferences implicit in the request of the requested vehicle. Thermal Risk Collaborative Decision-Making Module: Constructs a multi-agent reinforcement learning environment with charging stations within the region as agents; each agent's state space includes its own operating state, current ambient temperature, and a thermal risk situation vector aggregated from the thermal management capability parameters and thermal behavior preference types of vehicles waiting to be charged within the region; each agent outputs operational actions based on its own operating state, current ambient temperature, and thermal risk situation vector; each agent generates station-level guidance strategies through collaborative training. Personalized thermal adaptation guidance module: For a single request vehicle, it integrates the thermal management capability parameters of the single request vehicle, the decoded thermal behavior preference type and the current ambient temperature, calculates the thermal adaptation score of the charging station in the area, and generates and outputs the thermal adaptation charging guidance sequence in combination with the station-level guidance strategy. Data communication and interface module: used to respond to charging requests initiated by vehicles, obtain vehicle identification code and vehicle status information from user terminals, obtain operational status data from charging pile equipment, obtain ambient temperature data from meteorological service providers, and distribute vehicle status information, operational status data and ambient temperature data to the corresponding processing processes; at the same time, it receives the thermal adaptation charging guidance sequence and outputs it to the user terminal.

[0035] The present application has been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present application. Therefore, the content of this specification should not be construed as a limitation of the present application.

Claims

1. A smart operation method for charging stations based on charging user behavior characteristics, characterized in that, Includes the following steps: S1: When a vehicle initiates a charging request, obtain the vehicle identification code of the requesting vehicle; based on the vehicle identification code, parse and obtain the thermal management capability parameters of the vehicle, the thermal management capability parameters including the theoretical maximum continuous charging power and the thermal management type; S2: Obtain the historical charging behavior data of the requested vehicle; Based on the historical charging behavior data, a hidden Markov model is constructed and trained to infer the thermal behavior preferences of the requesting vehicle, and the type of thermal behavior preference implied by the requesting vehicle in the request is decoded. S3: Construct a multi-agent reinforcement learning environment with charging stations in the region as agents; each agent's state space includes its own operating state, current ambient temperature, and a thermal risk situation vector aggregated from the thermal management capability parameters and thermal behavior preference types of vehicles waiting to be charged in the region; each agent outputs operating actions based on its own operating state, current ambient temperature, and thermal risk situation vector; each agent generates station-level guidance strategies through collaborative training. S4: For a single request vehicle, integrate the thermal management capability parameters of the single request vehicle, the decoded thermal behavior preference type and the current ambient temperature, calculate the thermal compatibility score for going to the charging station in the area, and generate and output the thermal compatibility charging guidance sequence in combination with the station-level guidance strategy.

2. The intelligent operation method for charging stations based on charging user behavior characteristics according to claim 1, characterized in that, The step of parsing the vehicle's thermal management capability parameters based on the vehicle identification code includes: interacting with the manufacturer's data platform to obtain the vehicle's battery type, rated battery capacity, and thermal management configuration information based on the vehicle identification code; determining the thermal management type based on the thermal management configuration information, including active liquid cooling, forced air cooling, and natural cooling; and determining the vehicle's theoretical maximum continuous charging power under standard operating conditions based on the battery type and rated capacity, combined with battery technical specifications.

3. The intelligent operation method for charging stations based on charging user behavior characteristics according to claim 1, characterized in that, The construction and training of a Hidden Markov Model (HMM) based on the historical charging behavior data to infer the thermal behavior preferences of the requesting vehicle includes: the historical charging behavior data includes the ambient temperature at the time of the historical charging event, the ratio of the actual charging power to the theoretical maximum continuous charging power at the time of the historical charging event, and the physical shading attributes of the charging station at the time of the historical charging event; the hidden state set of the user's thermal behavior preferences is defined as follows: ,in The total number of states. Indicates by A set of hidden states, each representing a user's attitude and behavioral pattern towards charging thermal risks, including at least two of the following: thermal risk indifference, thermal risk avoidance, and thermal sensitivity with speed priority; define the user's observation sequence. ,in The length of the observation sequence represents the total number of historical charging events for the user. Indicates by A sequence of observations, each observation... Including the The features of the historical charging record include the ambient temperature during charging, the ratio of the actual average charging power to the vehicle's theoretical maximum continuous charging power, and the shading attribute identifier of the charging station; the parameters of the hidden Markov model are... ,in Here is the state transition probability matrix. For the observation probability matrix, The initial state distribution is used; the forward-backward algorithm and the Baum-Welch algorithm are employed, utilizing the observed sequence. For model parameters The model is trained to obtain a personalized hot behavior preference model for users.

4. The intelligent operation method for charging stations based on charging user behavior characteristics according to claim 1, characterized in that, The decoding process obtains the hot behavior preference types implicit in the request of the vehicle, including: based on user... Second historical charging observation sequence As input, the Viterbi algorithm is used to calculate the most probable hidden state sequence; the final state of the hidden state sequence is taken as the user's current hot behavior preference type. .

5. The intelligent operation method for charging stations based on charging user behavior characteristics according to claim 1, characterized in that, The thermal risk situation vector, which is an aggregation of thermal management capability parameters and thermal behavior preference types of vehicles waiting to be charged within the region, includes: the thermal risk situation vector. It is an intelligent agent The state space is composed of components, each dimension corresponding to a statistical measure of a thermal risk-related attribute; the thermal risk situation vector is constructed as follows: ;in, Indicating in intelligent agents Within the service area, the proportion of vehicles with natural cooling and forced air cooling as the thermal management type to the total number of vehicles; Indicating in intelligent agents Within the service scope, the proportion of users with the "hot risk indifference" behavior preference type to the total number of users; Represents intelligent agents Average ambient temperature within the service area.

6. The intelligent operation method for charging stations based on charging user behavior characteristics according to claim 1, characterized in that, Each agent generates a station-level guidance strategy through collaborative training, including: the operational actions include guidance strategies directly related to thermal risk control; the operational actions of the guidance strategies directly related to thermal risk control include adjusting the recommended priority of charging piles; generating and sending instruction packages to users containing delayed charging suggestions, transfer guidance information, and incentive compensation; and controlling the charging piles to activate active cooling and adjust operating power; the station-level guidance strategy content includes the adjustment of the recommended priority of charging piles generated based on thermal risk control, the instruction packages sent to users, and the instructions to control the charging piles.

7. The intelligent operation method for charging stations based on charging user behavior characteristics according to claim 1, characterized in that, The construction of a multi-agent reinforcement learning environment with charging stations within a region as agents includes: the region being a configurable operation and management unit; and each agent in the multi-agent reinforcement learning environment... reward function Designed as follows: ;in, For intelligent agents The returns during the decision-making cycle; For intelligent agents The average utilization rate of charging piles; , respectively intelligent agents , Load rate; For intelligent agents The set of neighboring agents; For intelligent agents The number of high-risk charging sessions served within the decision-making cycle; a high-risk charging session is defined as a charging request where the vehicle's thermal management type is natural cooling, the user's thermal behavior preference is thermal risk indifference, and the actual requested charging power is higher than the theoretical maximum continuous charging power by a certain proportion. These are preset positive weighting coefficients.

8. The intelligent operation method for charging stations based on charging user behavior characteristics according to claim 1, characterized in that, For a single requesting vehicle, the thermal management capability parameters of the single requesting vehicle, the decoded thermal behavior preference type, and the current ambient temperature are integrated to calculate a thermal adaptability score for traveling to charging stations within the area. This includes: the thermal adaptability score. The calculation formula is: ;in, As an efficiency factor, and with users Heading to the charging station Estimated driving time and charging stations The current queuing time is negatively correlated; Economic factors, and charging stations The service fees are negatively correlated; The thermal matching factor is calculated using the thermal matching function. The calculation formula is as follows: ;in, Indicates user Vehicle thermal management capability rating Indicates user Hot behavioral preference score Indicates ambient temperature. Indicates charging station The score for the heat dissipation capacity of the pile end. Indicates charging station Current load; function The design aims to give a high matching score to charging stations with strong heat dissipation capabilities and low load when the vehicle's heat dissipation capacity is weak, user preferences are aggressive, and the ambient temperature is high.

9. The intelligent operation method for charging stations based on charging user behavior characteristics according to claim 1, characterized in that, The process of generating and outputting a thermally adapted charging guidance sequence by combining a station-level guidance strategy includes: sorting all candidate charging stations in descending order based on thermal adaptability scores; selecting the top K charging stations to form a guidance sequence; and for each charging station option in the sequence, when a station-level guidance strategy is attached, converting the station-level guidance strategy into user-perceptible guidance information and outputting it along with the basic charging station information.

10. A smart charging station operation system based on charging user behavior characteristics, used to implement the smart charging station operation method based on charging user behavior characteristics as described in any one of claims 1-9, characterized in that, include: Vehicle thermal profile construction module: When a vehicle initiates a charging request, the vehicle identification code of the requesting vehicle is obtained; based on the vehicle identification code, the thermal management capability parameters of the vehicle are parsed and obtained, including the theoretical maximum continuous charging power and the thermal management type. User Hot Preference Analysis Module: Obtains historical charging behavior data of the requested vehicle; Based on the historical charging behavior data, a hidden Markov model is constructed and trained to infer the thermal behavior preferences of the requesting vehicle, and the type of thermal behavior preference implied by the requesting vehicle in the request is decoded. Thermal Risk Collaborative Decision-Making Module: Constructs a multi-agent reinforcement learning environment with charging stations within the region as agents; each agent's state space includes its own operating state, current ambient temperature, and a thermal risk situation vector aggregated from the thermal management capability parameters and thermal behavior preference types of vehicles waiting to be charged within the region; each agent outputs operational actions based on its own operating state, current ambient temperature, and thermal risk situation vector; each agent generates station-level guidance strategies through collaborative training. Personalized thermal adaptation guidance module: For a single request vehicle, it integrates the thermal management capability parameters of the single request vehicle, the decoded thermal behavior preference type and the current ambient temperature, calculates the thermal adaptation score of the charging station in the area, and generates and outputs the thermal adaptation charging guidance sequence in combination with the station-level guidance strategy. Data communication and interface module: used to respond to charging requests initiated by vehicles, obtain vehicle identification code and vehicle status information from user terminals, obtain operational status data from charging pile equipment, obtain ambient temperature data from meteorological service providers, and distribute vehicle status information, operational status data and ambient temperature data to the corresponding processing processes; at the same time, it receives the thermal adaptation charging guidance sequence and outputs it to the user terminal.

Images