Intelligent interaction dynamic resource recommendation method for shared electric bicycle and battery charging and replacing integrated base station

Abstract

The invention is suitable for the technical field of resource recommendation, and provides a shared electric bicycle and battery charging and replacing integrated base station intelligent interaction dynamic resource recommendation method, which comprises the following steps: acquiring battery state data, user behavior data and charging demand identification of a shared electric bicycle in real time; collecting resource state data of the charging and battery replacing integrated base station; evaluating availability characteristics of available batteries of the base stations, and generating recommendation priorities of the base stations; a resource locking request containing dynamic reserved duration is sent to a target base station with the highest priority, the dynamic reserved duration is dynamically adjusted according to the real-time position of a user and navigation estimated time, and when it is detected that a path is blocked or user requirements are changed, the recommendation priority is updated; and after the battery replacement operation is completed, a recommendation strategy is optimized based on the actual performance data of the battery and user operation feedback. According to the invention, highly intelligent and personalized base station recommendation can be provided for the user, the resource configuration of the whole system is optimized, and the user experience and the operation efficiency are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of resource recommendation, and particularly to an intelligent interactive dynamic resource recommendation method for shared electric bicycles and integrated charging and battery swapping stations. Background Art

[0002] Currently, the shared electric bicycle system has formed a technical architecture with electric bicycle terminals, user mobile applications, back-end management platforms, and charging and battery swapping infrastructures as the core. Electric bicycles can report their positions and battery status in real time by integrating GPS positioning, battery management systems, and wireless communication modules; charging and battery swapping stations are equipped with charging piles, battery swapping cabinets, and environmental monitoring sensors to support battery replacement and charging services. Users can find available vehicles or stations nearby through the APP, and the back-end system allocates resources based on set rules. However, due to the rules being set based on static factors, there are problems such as delayed update or large errors in battery status information, difficulty in timely dispatching low-battery vehicles by operation and maintenance, and the situation where users may rent inferior batteries.

[0003] In addition, some systems have basic data analysis functions, but still rely on manual intervention in dynamic scheduling and intelligent recommendation. This leads to uneven load distribution of charging and battery swapping stations, and users may face problems such as full charging positions or insufficient batteries after arrival, resulting in low resource utilization. Moreover, there is a lack of in-depth interaction among current electric bicycles, stations, user APPs, and back-end systems, and it is impossible to optimize the charging strategy based on real-time power grid load, environmental parameters, and user behavior.

[0004] In view of this, an intelligent interactive dynamic resource recommendation method for shared electric bicycles and integrated charging and battery swapping stations is proposed. Summary of the Invention

[0005] The present invention provides an intelligent interactive dynamic resource recommendation method for shared electric bicycles and integrated charging and battery swapping stations to solve the problem of low resource utilization.

[0006] The present invention provides an intelligent interactive dynamic resource recommendation method for shared electric bicycles and integrated charging and battery swapping stations, including: Real-time collecting battery status data, user behavior data, and charging demand identifiers of shared electric vehicles; meanwhile, collecting resource status data of integrated charging and battery swapping stations; Evaluating the availability characteristics of available batteries at each station based on the resource status data, where the availability characteristics include battery health, available number of charging positions, and environmental adaptability; Generating the recommended priorities of each station according to the battery status data, user behavior data, charging demand identifiers of shared electric vehicles, and the availability characteristics; Send a resource lock request containing the dynamic reservation duration to the target base station with the highest priority. The dynamic reservation duration is dynamically adjusted according to the user's real-time location and the navigation estimated time. If the user does not arrive within the dynamic reservation time, the resource lock will be automatically released; When it is detected that the path is blocked or the user's demand changes, re-match the base station that meets the lowest availability characteristics and has the highest arrival efficiency, and update the recommended priority; After the battery replacement operation is completed, optimize the recommendation strategy based on the actual performance data of the newly replaced battery and the user operation feedback. The actual performance data of the new battery includes the battery health, charge and discharge efficiency, and temperature stability corresponding to the availability characteristics.

[0007] Furthermore, evaluating the availability characteristics of the available batteries at each base station based on the resource status data. The availability characteristics include battery health, the available number of charging positions, and environmental adaptability, including: Where: is the battery health index, is the current actual capacity of the battery, is the nominal capacity, is the cumulative number of charge and discharge cycles, is the designed maximum cycle life, is the total number of charge and discharge cycles, is the th deviation between the average temperature and the optimal temperature during the th charge and discharge cycle, is the temperature reference benchmark period, , and are the weight coefficients respectively, based on the priority allocation of capacity attenuation dominance, cycle times secondary, and long-term temperature impact in the shared electric vehicle scenario.

[0008] Furthermore, evaluating the availability characteristics of the available batteries at each base station based on the resource status data. The availability characteristics include battery health, the available number of charging positions, and environmental adaptability, further including: Where: is the available number of charging positions, is the current number of idle charging positions, is the total number of charging positions, is the total number of charging positions, is the historical usage frequency of the th charging position, is the current time and the th nearest usage time of the charging position The interval is the attenuation factor Assign a higher weight to recent usage Is the confidence level for demand prediction in a future time period 、 、 Is the balance coefficient to ensure a reasonable proportion of real-time idle rate, historical attenuation, and prediction uncertainty

[0009] Furthermore, evaluating the availability characteristics of the available batteries of each base station based on the resource status data, the availability characteristics include battery health, the number of available charging positions, and environmental adaptability, and further include: Wherein: Is the environmental adaptability index Is the real-time temperature Is the safety temperature threshold Is the real-time grid load Is the reference load Is the meteorological warning level 、 、 Is the dynamic weight, representing the priority allocation of high temperature, grid load, and meteorological warning 、 Is the standard deviation, statistically based on historical data

[0010] Furthermore, generating the recommended priority of each base station according to the battery status data, user behavior data, charging demand identifier, and the availability characteristics of shared electric vehicles includes: Obtain the current battery power and battery health index, and generate a battery status score based on the battery health quantification model, which integrates voltage, cycle times, and temperature historical data Generate a user behavior adaptation score according to the historical riding path preference, time sensitivity, and battery selection tendency in the user behavior data Parse the charging demand identifier and generate a charging demand emergency score according to the urgency of the user order Combined with the calculated And the calculated Perform weighted summation on the battery status score, user behavior adaptation score, and charging demand emergency score to generate the recommended total score of each base station

[0011] Furthermore, generating a user behavior adaptation score according to the historical riding path preference, time sensitivity, and battery selection tendency in the user behavior data includes: Define the user behavior state space, including the set of historical riding paths, the time sensitivity interval, and the battery health selection preference Define the action space as the base station selection behavior, and the reward function is dynamically calculated according to the matching degree between the base station finally selected by the user and the base station recommended by the system, the order completion efficiency, and the battery life satisfaction; Train a preset reinforcement learning model through user behavior data to output the preference probability distribution of the user for the base station; Divide the user behavior patterns into high-frequency optimization type, regular type, and random type according to the preference probability distribution, and map them to the corresponding behavior adaptation scores; among them, the high-frequency optimization type has the highest score, and the random type has the lowest score.

[0012] Furthermore, the generation of the recommended priority of each base station includes: Sort all base stations according to the total recommendation score and divide the priority levels; Optimize the priority result based on the dynamic weight adjustment strategy, and the strategy includes increasing the weight of charging position availability during peak hours, increasing the weight of environmental adaptability under extreme weather conditions, and increasing the weight of user behavior adaptation for VIP users; Conduct multi-dimensional verification on the priority result, and the verification conditions include whether the battery health meets the minimum threshold, the probability of charging position reservation conflict, and the path accessibility; Output the final priority list, and push the real-time status and navigation path of the recommended base station to the user terminal.

[0013] Furthermore, the resource locking request including the dynamic reservation duration is sent to the target base station with the highest priority, and the dynamic reservation duration is dynamically adjusted according to the user's real-time position and the navigation estimated time. If the user does not arrive within the reserved time, the resource lock will be automatically released, including: Collect the historical usage data of the target base station, including the charging position occupancy rate, user arrival interval, and battery replacement time at different times; Construct a time series model, extract periodic features and trend features through the sliding window algorithm, and predict the congestion probability of the base station within a preset future period; Dynamically calculate the reservation duration according to the congestion probability and the navigation estimated time from the user's real-time position to the base station; Among them, if the predicted congestion probability of the target base station during the user's arrival period is higher than the preset threshold, the reservation duration will be shortened to avoid resource idleness, otherwise the reservation duration will be extended.

[0014] Furthermore, the dynamic reservation duration is dynamically adjusted according to the user's real-time position and the navigation estimated time. If the user does not arrive within the reserved time, the resource lock will be automatically released, including: Real-time monitor the user's travel speed, traffic condition changes, and base station resource status updates; If the user's traveling speed is lower than a preset ratio of the navigation estimated value, the reserved duration is extended according to the delay ratio, and a new estimated arrival time is pushed through the user terminal. If the update of the base station resource status shows that the reserved charging position is unexpectedly occupied, available base stations are rematched and the reserved duration is updated. If an external event triggers a sudden change in the base station load, the reserved duration is compressed or extended according to the real-time load rate.

[0015] Furthermore, when it is detected that the path is blocked or the user's demand changes, rematching the base station that meets the minimum availability characteristics and has the highest arrival efficiency and updating the recommended priority includes: Based on a preset minimum availability threshold, candidate base stations that meet the battery health ≥ preset health threshold, charging position idle rate ≥ preset idle rate threshold, and environmental adaptability ≥ preset adaptability threshold are screened. According to the positions of the candidate base stations, the real-time traffic conditions, and the user's traveling speed, the shortest estimated arrival time of each base station is calculated. The candidate base stations are sorted in ascending order according to the shortest estimated arrival time, an updated recommended priority list is generated, and the adjusted recommended result and navigation path are pushed to the user terminal. If the original target base station no longer meets the conditions, its reserved resources are released, and a dynamic resource locking request is sent to the base station with the highest new priority.

[0016] From the above technical solutions, it can be seen that the present invention has the following advantages: The present invention constructs a multi-source dynamic database by collecting battery status, user behavior, and base station resource data in real time, providing a high-timeliness and high-precision data basis for subsequent analysis; converting heterogeneous data into standardized indicators based on a quantization model; generating personalized recommended priorities by dynamically weighted fusion of user needs, battery health, and base station availability characteristics, effectively improving the accuracy of resource matching; dynamically adjusting the reserved duration of resources through time series prediction and real-time monitoring mechanisms to balance resource utilization efficiency and user experience; when the path is blocked or the demand changes, quickly rematching the base station through threshold screening and shortest arrival time calculation to ensure system flexibility and service continuity; finally, based on the battery performance data and user feedback after battery replacement, using reinforcement learning to iteratively optimize the recommendation strategy. The present invention can provide highly intelligent and personalized base station recommendations for users, while optimizing the overall system resource configuration, significantly improving user experience and operation efficiency. Description of the Drawings

[0017] Figure 1 It is a schematic flowchart of an embodiment of a method for intelligent interaction dynamic resource recommendation between a shared electric bicycle and a charging and swapping integrated base station in the present invention. Detailed Embodiments

[0018] The terms "first", "second", "third", "fourth", etc. (if any) in the description of this application and the above-mentioned drawings are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments of the application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "corresponding to" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that comprises a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0019] Embodiment 1 In this embodiment, the implementation method can be realized in a system, in a server, or in a terminal, and no specific limitation is made. Below, from the perspective of system implementation, the intelligent interaction dynamic resource recommendation method between shared electric bicycles and integrated charging and battery swapping stations in this application will be introduced. Please refer to Figure 1 , the method provided by the embodiment of this application includes the following steps: S11. Real-time collect the battery status data, user behavior data, and charging demand identifiers of shared electric vehicles; at the same time, collect the resource status data of the integrated charging and battery swapping station; In this embodiment, the battery status data, user behavior data, and corresponding parameters of the charging demand identifier are collected in real time through the battery management system built in the electric bicycle and transmitted to the background management platform via the cellular network or Bluetooth. Under normal conditions, it is reported once every minute, and real-time reporting is triggered in case of abnormal conditions. The battery status data includes the current remaining battery power percentage, reflecting the available endurance; the degree of battery capacity attenuation and life status; the real-time voltage and current during charging and discharging, used to monitor the battery working status; the internal and external environmental temperatures of the battery to avoid over-temperature risks; the cumulative number of charging and discharging times of the battery to evaluate the aging degree; abnormal conditions such as overvoltage, undervoltage, and short circuit detected by the battery management system.

[0020] The user behavior data records the riding trajectory, battery swapping operation log, and order data through the user APP, analyzes the user behavior pattern in combination with the background database, and generates structured behavior tags, such as "high-frequency optimization type users" and "time-sensitive users". The user behavior data includes the user's commonly used riding routes, the distribution of starting and ending points; the average battery swapping interval duration and preferred battery swapping time period of the user; the on-time rate of the user's order and the range of riding speed fluctuations; the proportion of the user's selection of high-health batteries; the frequency of the user's battery swapping requests initiated through the APP and the operation response time.

[0021] The charging demand identifier is a classification mark for the user's current charging demand, which is used to dynamically adjust the resource allocation strategy. The system automatically generates the identifier based on the real-time battery level and the user's APP operations (such as checking "urgent battery replacement"). The identifier is transmitted to the background management platform through the API interface. Among them, the urgent demand is automatically triggered by the system or manually marked by the user, with the highest priority; the high demand is when the battery level is between 15% - 30% and there is no urgent mark, and it is automatically generated by the system; the normal demand is when the battery level is higher than 30% and there is no special mark, with the lowest priority.

[0022] The resource status data of the integrated charging and battery replacement base station is used to monitor the physical status in real time through a sensor network such as optoelectronic sensors, temperature and humidity sensors, and current transformers at the base station terminal, and obtain real-time load and electricity price data through the power grid interface. The data is uploaded to the background management platform through the communication module of the base station controller. Among them, the resource status data includes the current number of idle charging positions and the occupancy duration, the health, battery level, and cycle count of the replaceable batteries in the base station; the internal temperature, humidity, and smoke detection status of the base station; the real-time grid power, time-of-use electricity price information, and power supply stability; the charging pile fault code and the abnormal alarm of the battery compartment lock status.

[0023] S12. Evaluate the availability characteristics of the available batteries in each base station based on the resource status data. The availability characteristics include battery health, the available number of charging positions, and environmental adaptability; This step quantifies and evaluates the availability characteristics of the integrated charging and battery replacement base station based on the battery status data, user behavior data, and base station resource status data collected in real time in the above step S11. Specifically, it includes three core indicators: battery health, the available number of charging positions, and environmental adaptability. Through mathematical modeling and dynamic weight allocation, complex multi-source data is converted into comparable quantitative indicators. The calculation formulas are as follows: Battery health: Where: is the battery health index, is the current actual capacity of the battery, is the nominal capacity, is the cumulative cycle count, is the designed maximum cycle life, is the total number of charge and discharge cycles, is the th deviation between the average temperature and the optimal temperature during the charge and discharge cycle, is the temperature reference base period, , and is the weight coefficient, which is assigned based on the priority of capacity attenuation dominance, cycle times secondary, and long-term temperature impact in the shared electric vehicle scenario. For example, the weight setting of 60% for capacity attenuation reflects the current available capacity of the battery and directly determines the endurance ability; the weight setting of 30% for cycle life characterizes the degree of battery aging, and the closer the cycle times are to the designed life, the lower the score; the weight setting of 10% for temperature deviation accumulates the duration of temperature deviation from the optimal value (such as 25°C). High temperature accelerates battery aging and requires long-term correction.

[0024] Number of available charging positions: Where: is the number of available charging positions, is the number of currently idle charging positions, is the total number of charging positions, is the total number of charging positions, is the historical usage frequency of the th charging position, is the current time and the time interval since the most recent use of the th charging position, and is the decay factor Higher weight is given to recent usage, is the confidence level of demand prediction for a future time period, and are balance coefficients to ensure a reasonable proportion of real-time idle rate, historical decay, and prediction uncertainty. For example, the real-time idle rate (main item) directly reflects the current availability, and the score is the highest when the idle rate ≥ 50%; historical usage decay (secondary item) gives higher weight to charging positions with high recent usage frequency to avoid interference from old data; prediction confidence correction (secondary item) reduces the score if the uncertainty of demand prediction for a future time period is high to prevent resource overload.

[0025] Environmental adaptability: Where: is the environmental adaptability index, is the real-time temperature, is the safety temperature threshold, is the real-time grid load, is the reference load, is the meteorological warning level, and and are dynamic weights representing the priority allocation of high temperature, grid load, and meteorological warning, and is the standard deviation, statistically based on historical data. For example, when 50% of the real-time temperature exceeds the safety threshold (such as 40°C) in the temperature safety setting, the score is significantly reduced, triggering the cooling system or derating charging; when 30% of the grid load is higher than the reference value (such as the historical average) in the grid load setting, the score is reduced, and the charging cost is optimized in combination with time-of-use electricity prices; when 20% of the corresponding extreme weather such as heavy rain and strong wind occurs in the meteorological warning setting, the score is reduced, and the protected base stations are preferentially recommended.

[0026] The quantization index of this step can comprehensively reflect the comprehensive performance of the base station, avoiding single-dimensional deviation. Through quantization scoring, the system can preferentially recommend high-health batteries, low-load base stations, and resources with stable environments, improving the user experience and operation efficiency.

[0027] S13. Generate the recommended priorities of each base station according to the battery status data, user behavior data, charging demand identification, and availability characteristics of shared electric vehicles; This step is based on the battery status, user behavior, charging demand identification collected in step S11, and the base station availability characteristics quantified in step S12, that is, battery health, available number of charging positions, and environmental adaptability. The recommended priorities of each base station are generated through a multi-dimensional scoring model.

[0028] S131. Obtain the current battery power and battery health index, and generate a battery status score based on the battery health quantization model. The quantization model integrates voltage, cycle times, and temperature historical data; The battery status score is based on the battery power, health, voltage, cycle times, and temperature historical data collected in real time in step S11. The quantization model uses the above battery health quantization model to calculate the battery health index , and preferentially recommend batteries with high battery life and low aging degree, reducing the riding interruption caused by battery problems for users.

[0029] S132. Generate a user behavior adaptation score according to the historical riding path preference, time sensitivity, and battery selection tendency in the user behavior data; 1. Define the user behavior state space, including the set of historical riding paths, time sensitivity interval, and battery health selection preference; 2. Define the action space as the base station selection behavior, and the reward function is dynamically calculated according to the matching degree between the base station finally selected by the user and the base station recommended by the system, order completion efficiency, and battery life satisfaction; 3. Train a preset reinforcement learning model through the user's historical behavior data, and output the preference probability distribution of the user for the base station; 4. Divide the user behavior pattern into high-frequency optimization type, regular type, and random type according to the preference probability distribution, and map it to the corresponding behavior adaptation score; among them, the high-frequency optimization type has the highest score, and the random type has the lowest score.

[0030] Specifically, the user behavior data includes the historical riding path, time sensitivity, and battery selection preference collected in step S11. The historical riding path set includes common routes and frequently visited base stations; the time sensitivity interval includes a high sensitivity when the on-time rate ≥ 95%; the battery health selection preference is the proportion of users who choose high SoH batteries. The action space is the user's selection behavior of base stations. The reward function dynamically calculates the reward value based on the matching degree between the base station finally selected by the user and the system recommendation, the order completion efficiency (such as the on-time delivery rate), and the actual battery life satisfaction (the remaining battery percentage after the user rides).

[0031] The total reward value is composed of the matching reward , the order efficiency reward and the battery life satisfaction reward through weighted calculation. Among them, if the base station finally selected by the user is consistent with the system recommendation, , otherwise it is ; ; .

[0032] The deep Q reinforcement learning model is trained with user behavior data to output the preference probability distribution of the user for base stations. Among them, the state space is the set of the user's historical riding paths, the time sensitivity interval, and the battery health selection preference; the action space is the user's selection behavior of base stations, that is, all optional base station identifiers in the recommendation list; the input layer of the Q network structure is the user state vector (dimension = the number of historical paths + the number of time sensitivity levels + the number of battery preference parameters); the hidden layer is a 3-layer connected neural network, and the activation function is ReLU; the output layer is the Q value of each base station. The training process of the model includes: 1. Experience recycling: storing the user state, action, reward, and next state tuple in the recycling pool; 2. Action selection: using the ε-greedy strategy; 3. Network update: sampling batch data from the replay pool every 1000 steps, calculating the TD error and updating the network parameters through the Adam optimizer; 4. Target network synchronization: copying the main network parameters to the target network every 10000 steps.

[0033] Finally, according to the preference probability distribution, the users are divided into: high-frequency optimization type corresponding to a preference probability ≥ 70%, score +30%; regular type corresponding to 30% ≤ probability < 70%, score +15%; random type corresponding to a probability < 30%, score +0%.

[0034] S133. Analyze the charging demand identifier and generate a charging demand emergency score according to the urgency of the user's order; The charging demand identifier is generated by step S11 according to the user's real-time power and the urgency of the order. Among them, the emergency demand corresponds to a score increase of +25%, when SoC < 20% or the remaining order time < 10 minutes; the high demand corresponds to a score increase of +10%, when 20% ≤ SoC < 50% and there is no emergency mark; the normal demand corresponds to a score increase of +0%, when SoC ≥ 50%. By responding to the user's urgency, it ensures that low-power or urgent orders are preferentially allocated resources.

[0035] S134. Combine the calculated and the calculated , and perform a weighted sum of the battery status score, the user behavior adaptation score, and the charging demand emergency score to generate the recommended total score for each base station.

[0036] Perform a weighted sum of the battery status score (weight 30%), the user behavior adaptation score (weight 20%), the charging demand emergency score (weight 25%), the charging position availability score (weight 15%), and the environmental adaptation score (weight 10%). By dynamically adapting the weights to the real-time scenario requirements, the resource allocation efficiency is optimized. Specifically, the recommended priorities for each charging pile are optimized as follows: 1. Sort all base stations according to the recommended total score and divide them into priority levels; 2. Optimize the priority results based on the dynamic weight adjustment strategy. The strategies include increasing the weight of the charging position availability during peak hours, increasing the weight of the environmental adaptation under extreme weather conditions, and increasing the weight of the user behavior adaptation for VIP users; 3. Perform multi-dimensional verification on the priority results. The verification conditions include whether the battery health meets the minimum threshold, the probability of charging position reservation conflict, and the path reachability; 4. Output the final priority list and push the real-time status and navigation path of the recommended base stations to the user terminal.

[0037] Specifically, the dynamic screening includes the user identity priority: VIP users are automatically promoted to the top 10% of the list; real-time traffic conditions: if the path is congested or under construction, the priority of the corresponding base station is reduced; meteorological warning level: when there is a rainstorm warning, priority is given to recommending base stations with rain shelters or indoor ones. The battery health threshold excludes base stations with SoH < 75%; the charging position reservation conflict means that if the reserved position of the target base station is preempted, switch to the second-best base station; the path reachability is verified through the navigation API whether the path is closed or restricted. Finally, push the final priority list to the user terminal, which includes the base station score, the estimated arrival time, and the navigation link, and lock the resources of the highest-priority base station.

[0038] The above steps combine user behavior preferences, battery health, and real-time requirements to provide personalized base station recommendations; adapt to complex scenarios such as traffic congestion and extreme weather through weight adjustment and multi-dimensional verification; and can avoid the allocation of low-health batteries and base station overload, improving operation efficiency and user satisfaction.

[0039] S14. Send a resource lock request containing a dynamic reservation duration to the target base station with the highest priority. The dynamic reservation duration is adjusted dynamically according to the user's real-time location and the estimated navigation time. If the user does not arrive within the reserved time, the resource lock is automatically released. Based on the recommended priority list generated in step S13, this step sends a resource lock request to the target base station with the highest priority and dynamically adjusts the reservation duration according to the user's real-time location and the estimated navigation time to ensure efficient resource utilization. If the user does not arrive within the reserved time, the lock is automatically released to avoid resource idling. Specifically, it includes the following: S141. Collect the historical usage data of the target base station, including the occupancy rate of charging positions, the arrival interval of users, and the battery replacement time at different time periods. Collect the historical usage data of the target base station, including the occupancy ratio of charging positions at different time periods (such as morning rush hour and evening rush hour), the average time interval for users to arrive at the base station, and the average time taken for a single battery replacement operation.

[0040] S142. Construct a time series model, extract periodic features and trend features through a sliding window algorithm, and predict the congestion probability of the base station within a preset future time period. Construct a time series model using a sliding window algorithm. The specific operations are as follows: 1. Extract periodic features (such as the peak occupancy rate during the daily peak hours); 2. Extract trend features (such as the charging demand on weekends increasing by 20% compared to weekdays); 3. Predict the congestion probability of the base station within a preset future time period (such as within the next hour). For example, if the predicted congestion probability for the next hour is 80%, it is determined that the base station may be overloaded.

[0041] S143. Dynamically calculate the reservation duration according to the congestion probability and the estimated navigation time from the user's real-time location to the base station. Among them, if the predicted congestion probability of the target base station during the user's arrival period is higher than the preset threshold, the reservation duration is shortened to avoid resource idling, and vice versa, the reservation duration is extended.

[0042] The reservation duration is calculated based on the time required for the user to reach the base station calculated in real-time through the map API, the basic reservation buffer time, and the congestion probability critical value: Reservation duration = Estimated navigation time + Basic reservation buffer time (1 - congestion probability / preset threshold), where the preset threshold is the congestion probability critical value (e.g., 60%) and is used to determine whether to shorten or extend the reservation duration. The threshold is optimized through historical data analysis to balance resource utilization and user experience. If the preset threshold is exceeded, the reservation duration is shortened. When the high congestion probability ≥ 60%, the reservation duration is shortened. For example, if the user is expected to arrive in 10 minutes, the reservation duration is set to 8 minutes; when the low congestion probability < 60%, the reservation duration is extended. For example, if the user is expected to arrive in 10 minutes, the reservation duration is set to 15 minutes.

[0043] Specifically, the dynamic reservation duration is adjusted according to the user's real-time location and the navigation estimated time. If the user does not arrive within the reserved time, the resource lock is automatically released, including the following: 1. Real-time monitoring of the user's traveling speed, traffic condition changes, and base station resource status updates; 2. If the user's traveling speed is lower than a preset percentage of the navigation estimated value, the reservation duration is extended according to the delay percentage, and a new estimated arrival time is pushed to the user terminal; 3. If the base station resource status update shows that the reserved charging position is unexpectedly occupied, available base stations are re-matched and the reservation duration is updated; 4. If an external event triggers a sudden change in the base station load, the reservation duration is compressed or extended according to the real-time load rate.

[0044] The real-time monitoring parameters include the user's traveling speed, that is, by comparing the GPS positioning data with the historical speed, calculating the deviation between the actual speed and the estimated speed; accessing the traffic API to obtain real-time congestion, construction, accident information, and other traffic condition changes; also monitoring the occupancy of charging positions, battery health, and environmental parameters.

[0045] If the actual speed is lower than a preset percentage (80%) of the estimated value, the reservation duration is extended according to the delay percentage. The formula is set as: new reservation duration = original reservation duration ⋅ (1 + (estimated time - actual estimated time)); for example, if the user originally needed 10 minutes to arrive and actually needs 15 minutes due to congestion, the reservation duration is extended to 15 minutes, and an update reminder is pushed. If the reserved position is occupied by other users or system failures, the resources are immediately released and available base stations are re-matched; if a sudden heavy rain causes a sharp increase in the base station load rate, the reservation duration is compressed to the lowest threshold to accelerate resource turnover. If a red alert for rainstorm or high temperature is triggered, protected base stations are preferentially recommended and their reservation durations are extended; if the power grid load exceeds the safety threshold, the charging operation is suspended and the reservation strategy is adjusted.

[0046] S15. When it is detected that the path is blocked or the user's demand changes, re-match the base station that meets the minimum availability characteristics and has the highest arrival efficiency, and update the recommendation priority; This step is used to re-screen eligible base stations and update the recommendation priority when path obstructions such as traffic congestion and road construction are detected, or user demand changes such as order emergency status adjustment and sudden power drop occur, ensuring that users always obtain the optimal battery replacement resources. Specifically as follows: 1. Based on a preset minimum availability threshold, screen candidate base stations that meet the battery health ≥ preset health threshold, charging position idle rate ≥ preset idle rate threshold, and environmental adaptability ≥ preset adaptability threshold; Path obstruction detection obtains traffic event data in real time through an integrated map service API, and combines user GPS positioning data to determine whether the current navigation path is feasible. If the estimated arrival time exceeds a preset deviation ratio of the original value, it is determined that the path is obstructed. User demand change detection updates the charging demand flag to a higher emergency level by monitoring user manual operations or system automatically triggered events.

[0047] Candidate base stations are screened based on the minimum availability threshold to ensure that the recommended base stations meet the basic availability requirements. Among them, the battery health threshold is analyzed based on historical data. Batteries with SoH≥75% can guarantee at least 80% of the nominal cruising range, avoiding interruption of user rides, and are used to exclude base stations with SoH<75%; the charging position idle rate threshold ensures that at least 1 idle charging position is available to prevent users from having no position to replace the battery when they arrive, and is used to screen base stations with an idle rate≥10%; the environmental adaptability threshold is set as follows: temperature≤40℃ and humidity≤80% are the battery safe operating conditions, exceeding the threshold may cause performance degradation or failure, and are used to exclude base stations with temperature>40℃ or humidity>80%.

[0048] 2. Calculate the shortest estimated arrival time for each base station based on the location of the candidate base stations, real-time traffic conditions, and user travel speed; Based on the geographical location of the above-mentioned candidate base stations, real-time traffic conditions, and user real-time travel speed, call the map API (such as Amap path planning) to obtain the real-time navigation path and estimated time. If the user travel speed is lower than a preset ratio (such as 80%) of the historical average, the estimated time is corrected proportionally. Calculation formula: shortest estimated time = navigation API time ⋅ (1 + ), for example, if the navigation API estimates an 8-minute arrival and the user's real-time speed is only 70% of the historical average, the corrected time is 10.4 minutes.

[0049] 3. Sort the candidate base stations in ascending order of the shortest estimated arrival time, generate an updated recommended priority list, and push the adjusted recommendation results and navigation path to the user terminal; Arrange candidate base stations in ascending order according to the shortest estimated arrival time to generate an updated priority list; push the adjusted recommendation results to the user terminal, including: the battery health and charging status of the new base station; the revised navigation path and estimated arrival time; and the reason why the original base station is unavailable.

[0050] 4. If the original target base station no longer meets the conditions, its reserved resources are released and a dynamic resource locking request is sent to the base station with the highest new priority.

[0051] If the original target base station no longer meets the conditions due to path obstruction or resource occupation, its reservation lock is immediately released; a dynamic resource locking request is sent to the base station with the highest new priority, and the reservation duration is recalculated according to the logic of step S14.

[0052] The following is an explanation of specific scenarios: Scenario 1: The user originally planned to go to base station A (SoH=80%, idle rate 20%), but the route was blocked due to road construction. The system screened out base stations B (SoH=78%, idle rate 15%) and base station C (SoH=85%, idle rate 25%), calculated that base station C has the shortest arrival time (8 minutes), updated the recommendation, and locked the resources. Scenario 2: The user's power level dropped sharply from 25% to 8%, and the demand changed to urgent. The system immediately screened base stations with SoH≥75% and idle rate≥10%, and gave priority to recommending base station D, which is the closest and has the lowest load.

[0053] S16. After the battery replacement operation is completed, optimize the recommendation strategy based on the actual battery performance data and user operation feedback.

[0054] In this embodiment, the actual battery performance data such as the battery health SoH, charge and discharge efficiency, and temperature stability after battery replacement are collected in real time, and user operation feedback such as battery replacement time, interface operation satisfaction score, and whether to manually adjust the recommendation results are collected; if the battery health of a base station is continuously lower than the preset threshold, it is marked as a "low-quality resource" and its recommendation weight is reduced; if the user frequently refuses to recommend a base station, such as refusing to accept the recommendation due to complex paths or insufficient charging positions, it is recorded as a "low-trust base station" to reduce its exposure priority; a reinforcement learning model is used to dynamically adjust the score weight according to the match between the user's actual choice and the recommendation result. For example, when the user chooses a non-recommended base station, the distance weight is reduced and the battery health weight is increased; the user retention rate, order completion efficiency, and resource utilization under the new and old strategies are compared to select the optimal model version; finally, the model is retrained regularly to integrate the latest battery performance data and user feedback to generate an updated recommendation strategy; the effect of the new strategy is gradually verified through grayscale release to avoid global risks.

[0055] Through data-driven dynamic resource allocation, the above embodiments improve the utilization rate of the charging and swapping resources of shared electric bicycles and the user experience. At the same time, through intelligent algorithms and real-time response capabilities, the operation and maintenance costs are significantly reduced, and the adaptability of the system to complex urban scenarios is enhanced.

[0056] It can be understood that those skilled in the art can, under the guidance of the above embodiments, combine various implementation manners in the above respective embodiments to obtain technical solutions of various implementation manners.

[0057] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for intelligent interaction and dynamic resource recommendation between shared electric bicycles and integrated charging and battery replacement base stations, characterized in that, Including: Real-time collecting battery status data, user behavior data, and charging demand identifiers of shared electric vehicles; Simultaneously collecting resource status data of charging and swapping integrated base stations; Evaluating the availability characteristics of available batteries at each base station based on the resource status data, where the availability characteristics include battery health, available number of charging positions, and environmental adaptability; Generating recommended priorities for each base station according to the battery status data, user behavior data, charging demand identifiers, and the availability characteristics of shared electric vehicles; Sending a resource locking request including a dynamic reservation duration to the target base station with the highest priority, where the dynamic reservation duration is dynamically adjusted according to the user's real-time location and the estimated time of navigation. If the user does not arrive within the dynamic reservation time, the resource lock will be automatically released; When detecting that the path is blocked or the user's demand changes, re-matching the base station that meets the minimum availability characteristics and has the highest arrival efficiency, and updating the recommended priority; After the battery swapping operation is completed, optimizing the recommendation strategy based on the actual performance data of the replaced new battery and the user operation feedback, where the actual performance data of the new battery includes battery health, charge and discharge efficiency, and temperature stability corresponding to the availability characteristics.

2. The intelligent interactive dynamic resource recommendation method for shared electric bicycles and integrated charging and battery swapping stations according to claim 1, wherein The evaluating the availability characteristics of available batteries at each base station based on the resource status data, where the availability characteristics include battery health, available number of charging positions, and environmental adaptability, includes: Wherein: is the battery health index, is the current actual capacity of the battery, is the nominal capacity, is the cumulative number of charge and discharge cycles, is the designed maximum cycle life, is the total number of charge and discharge cycles, is the th deviation between the average temperature and the optimal temperature during the charge and discharge cycle, is the temperature reference base period, , and are the weight coefficients, which are assigned based on the priority of capacity attenuation dominance, cycle number secondary, and long-term temperature impact in the shared electric vehicle scenario.

3. The intelligent interaction dynamic resource recommendation method for shared electric bicycles and integrated charging and battery swapping base stations according to claim 2, wherein The evaluating the availability characteristics of available batteries at each base station based on the resource status data, where the availability characteristics include battery health, available number of charging positions, and environmental adaptability, further includes: Wherein: is the available number of charging positions, is the current number of idle charging positions, is the total number of charging positions, is the total number of charging positions, is the historical usage frequency of the th charging position, is the current time and the th charging position's most recent usage time interval, and is the decay factor assigns higher weight to recent usage, is the confidence level of the demand forecast for a certain future period, , , are balance coefficients to ensure a reasonable proportion of real-time idle rate, historical decay, and prediction uncertainty.

4. The intelligent interaction dynamic resource recommendation method for shared electric bicycles and integrated charging and battery swapping stations according to claim 3, wherein, The evaluating the availability characteristics of available batteries at each base station based on the resource status data, where the availability characteristics include battery health, available number of charging positions, and environmental adaptability, further includes: Wherein: is the environmental adaptability index, is the real-time temperature, is the safety temperature threshold, is the real-time grid load, is the reference load, is the meteorological warning level, and and are dynamic weights, representing the priority distribution of high temperature, grid load, and meteorological warning, and are standard deviations, statistically based on historical data.

5. The intelligent interaction dynamic resource recommendation method for shared electric bicycles and integrated charging and battery swapping base stations according to any one of claims 1-4, characterized in that, The generating recommended priorities for each base station according to the battery status data, user behavior data, charging demand identifiers, and the availability characteristics, includes: Obtaining the current battery power and battery health index, and generating a battery status score based on a battery health quantification model that integrates voltage, cycle times, and temperature history data; Generating a user behavior adaptation score according to the historical riding path preference, time sensitivity, and battery selection tendency in the user behavior data; Parsing the charging demand identifier and generating a charging demand urgency score according to the urgency of the user's order; Combined calculation And calculated , perform weighted summation on the battery status score, user behavior adaptation score, and charging demand urgency score to generate the recommended total score for each base station.

6. The intelligent interaction dynamic resource recommendation method for shared electric bicycles and integrated charging and battery swapping base stations according to claim 5, characterized in that The generating a user behavior adaptation score according to the historical riding path preference, time sensitivity, and battery selection tendency in the user behavior data, includes: Defining a user behavior state space, including a set of historical riding paths, a time sensitivity interval, and a battery health selection preference; Defining the action space as the base station selection behavior, and the reward function is dynamically calculated according to the matching degree between the base station finally selected by the user and the base station recommended by the system, the order completion efficiency, and the battery endurance satisfaction; Training a preset reinforcement learning model through user behavior data, and outputting the preference probability distribution of the user for the base station; Dividing the user behavior pattern into high-frequency optimization type, regular type, and random type according to the preference probability distribution, and mapping it to the corresponding behavior adaptation score; among them, the high-frequency optimization type has the highest score, and the random type has the lowest score.

7. The intelligent interaction dynamic resource recommendation method for shared electric bicycles and integrated charging and battery swapping base stations according to claim 5, wherein The generating recommended priorities for each base station, includes: Sort all base stations according to the total recommended score and divide them into priority levels; Optimize the priority results based on a dynamic weight adjustment strategy, which includes increasing the weight of charging position availability during peak hours, increasing the weight of environmental adaptability under extreme weather conditions, and increasing the weight of user behavior adaptability for VIP users; Conduct multi-dimensional verification on the priority results, and the verification conditions include whether the battery health meets the minimum threshold, the probability of charging position reservation conflict, and path accessibility; Output the final priority list and push the real-time status and navigation path of the recommended base stations to the user terminal.

8. The intelligent interaction dynamic resource recommendation method for shared electric bicycles and integrated charging and battery swapping base stations according to claim 1, wherein Send a resource locking request containing the dynamic reservation duration to the target base station with the highest priority. The dynamic reservation duration is dynamically adjusted according to the user's real-time position and the navigation estimated time. If the user does not arrive within the reserved time, the resource lock will be automatically released, including: Collect the historical usage data of the target base station, including the charging position occupancy rate, user arrival interval, and battery replacement time at different times; Construct a time series model, extract periodic features and trend features through a sliding window algorithm, and predict the congestion probability of the base station within a preset future period; Dynamically calculate the reservation duration according to the congestion probability and the navigation estimated time from the user's real-time position to the base station; Among them, if it is predicted that the congestion probability of the target base station during the user's arrival period is higher than the preset threshold, the reservation duration will be shortened to avoid resource idleness, otherwise the reservation duration will be extended.

9. The intelligent interactive dynamic resource recommendation method for shared electric bicycles and integrated charging and battery swapping base stations according to claim 8, characterized in that, The dynamic reservation duration is dynamically adjusted according to the user's real-time position and the navigation estimated time. If the user does not arrive within the reserved time, the resource lock will be automatically released, including: Real-time monitor the user's traveling speed, traffic condition changes, and base station resource status updates; If the user's traveling speed is lower than a preset ratio of the navigation estimated value, extend the reservation duration according to the delay ratio and push a new estimated arrival time to the user terminal; If the base station resource status update shows that the reserved charging position is unexpectedly occupied, re-match the available base station and update the reservation duration; If an external event triggers a sudden change in the base station load, compress or extend the reservation duration according to the real-time load rate.

10. The intelligent interaction dynamic resource recommendation method for shared electric bicycles and integrated charging and battery swapping base stations according to claim 1, wherein When it is detected that the path is blocked or the user's demand changes, re-match the base station that meets the minimum availability characteristics and has the highest arrival efficiency and update the recommended priority, including: Based on a preset minimum availability threshold, screen candidate base stations that meet the battery health ≥ preset health threshold, charging position vacancy rate ≥ preset vacancy rate threshold, and environmental adaptability ≥ preset adaptability threshold; Calculate the shortest estimated arrival time of each base station according to the position of the candidate base station, the real-time traffic condition, and the user's traveling speed; Sort the candidate base stations in ascending order of the shortest estimated arrival time, generate an updated recommended priority list, and push the adjusted recommended results and navigation path to the user terminal; If the original target base station no longer meets the conditions, release its reserved resources and send a dynamic resource locking request to the base station with the highest new priority.

Images