Battery replacement recommendation method and system based on user behavior and load, and storage medium

Abstract

The invention discloses a user behavior and load-based battery replacement recommendation method and system and a storage medium. The method comprises the steps of obtaining user battery replacement electricity charge data, user position data, battery replacement station battery replacement electricity charge data, battery replacement station electric quantity data, battery replacement station position data and user battery replacement time data; processing and analyzing the obtained data to obtain the characteristics of the online learning recommendation model of the input value based on the UCB algorithm, inputting the characteristics to the trained online learning recommendation model based on the UCB algorithm, and recommending the position of the optimal battery swap station of the user at the moment; information is fed back according to the battery replacing station recommended by the user, parameters of the online learning recommendation model based on the UCB algorithm are corrected, and a better battery replacing station is recommended when the user replaces the battery next time. The method can provide the user with the most electricity-changing station selection which can save electricity charge during electricity changing, and can maintain the economic benefit requirement of the electricity-changing station while reducing the electricity-changing cost of the user.

Description

Technical Field

[0001] The present invention relates to the field of battery swapping technology, and particularly to a battery swapping recommendation method based on user behavior and load. Background Art

[0002] With the increasing popularity of battery swapping for two-wheelers, users are paying more and more attention to the selection of battery swapping stations. Existing battery swapping services usually lack an intelligent recommendation solution that comprehensively considers user behavior patterns, site load conditions, and changes in electricity prices. In addition, traditional static recommendation methods cannot adapt to rapidly changing market conditions and user needs. At the same time, the distance between different battery swapping stations also affects the total cost of users. Therefore, an intelligent recommendation system that can comprehensively consider these factors is needed, and it is necessary to take into account the electricity cost of the battery swapping station and the utilization rate of the battery swapping station. While meeting the needs of users, it is necessary to meet the economic requirements of the battery swapping station. Summary of the Invention

[0003] Object of the Invention: In order to overcome the deficiencies in the prior art, the present invention provides a battery swapping recommendation method, system, and storage medium based on user behavior and load. By collecting data and using advanced online learning algorithms to dynamically adjust the recommendation strategy, a better battery swapping station location can be recommended to users.

[0004] Technical Solution: To achieve the above object, the battery swapping recommendation method based on user behavior and load of the present invention includes the following steps:

[0005] Step 1: Obtain the electricity cost data of user battery swapping, user location data, electricity cost data of battery swapping at the battery swapping station, battery amount data of the battery swapping station, location data of the battery swapping station, the number of battery swapping times at the battery swapping station, and the time data when the user needs to swap the battery.

[0006] Step 2: Calculate the average value of the electricity cost data of user battery swapping within a period of time, the average value of the electricity cost data of battery swapping at each battery swapping station, and the average battery swapping frequency of each battery swapping station. Calculate the utilization rate of the battery swapping station based on the battery amount data and the battery swapping frequency of the battery swapping station.

[0007] Step 3: Calculate the distance between the user's current location and the locations of each battery swapping station, and associate the distance between the user's current location and the locations of each battery swapping station, the battery amount data and the battery swapping frequency in each battery swapping station to obtain a number of battery swapping behavior association data.

[0008] Step 4: Initialize an online learning recommendation model based on the UCB algorithm, and use the average electricity cost of user battery swapping, the average electricity cost of battery swapping at each battery swapping station, the battery swapping frequency of each battery swapping station, the utilization rate of the battery swapping station, the battery swapping behavior association data, and the time data when the user needs to swap the battery as features and input them into the trained online learning recommendation model based on the UCB algorithm to recommend the location of the optimal battery swapping station for the user at this time.

[0009] Step 5: Provide feedback information based on the swapping station points recommended by the user, and correct the parameters of the online learning recommendation model based on the UCB algorithm to recommend a better swapping station point for the user during the next battery swapping.

[0010] Furthermore, in Step 2, based on the electricity bills of several swapping stations and users in different time periods, the changing pattern of electricity prices is obtained. The calculation processes for the average value of the electricity bill data of users within each time period T and the average value of the electricity bill data of each swapping station point are as follows:

[0011]

[0012] In the formula, E s，t represents the electricity bill expenditure of swapping station point s at time t, and E u，t represents the electricity bill expenditure of user u at time t;

[0013] The swapping frequency F s，T of swapping station point s within each time period T is calculated as follows:

[0014]

[0015] In the formula, F s，T represents the swapping frequency of swapping station point s within time period T, and R s，t represents the number of swapping times of swapping station point s at time t.

[0016] Furthermore, in Step 2, according to the number of swapping times of each swapping station point and the electricity quantity data of the swapping station point, the electricity quantity supply-demand balance index BAL s，t , that is, the utilization rate of the swapping station point, is defined. The calculation formula is as follows:

[0017] BAL s，t = B s，t - R s，t

[0018] In the formula, B s，t represents the number of batteries of swapping station point s at time t.

[0019] Furthermore, in Step 3, according to the user location data and the swapping station point location data, the distance D u，s from the current location of user u to each swapping station point s is calculated. The distance D u，s , the electricity quantity data and the swapping frequency F s，t in the swapping station point are associated to obtain the swapping behavior association data. The calculation process is as follows:

[0020]

[0021] Where D u，s represents the distance from user u to the battery swapping station s at time t; α is a weight parameter; B s，t represents the number of batteries at the battery swapping station s at time t, that is, the power of the battery swapping station s at time t.

[0022] Furthermore, perform correlation analysis and feature importance evaluation on all features input to the online learning recommendation model based on the UCB algorithm, and filter out the features that affect the recommendation results of the online learning recommendation model based on the UCB algorithm; calculate the correlation between all features, and the correlation between two features is measured by the Pearson correlation coefficient. When the absolute value of the calculated correlation value is greater than 0.8, it is determined that the correlation between the two features is too high, and one of the two features is removed;

[0023] Sort according to the standardized coefficients after standardizing the feature coefficients fitted from all features to obtain the ranking y of the final features; according to the feature x n coefficient β n After standardizing the absolute value of the coefficient β n reflects the importance of feature x n When the coefficient β

[0024] y = β0 + β1x1 + β2x2 + … + β n x n + ∈

[0025] Where xn is the nth feature, βn is the coefficient of the nth feature, and ∈ is the error term.

[0026] Furthermore, in step four, when initializing the online learning recommendation model based on the UCB algorithm, initialize the number of attempts Ns(0) = 0 and the cumulative reward Hs(0) = 0 for each battery swapping station s;

[0027] At each time step t, for each battery swapping station s, calculate its average reward

[0028]

[0029] Calculate the upper bound UCBs(t) of the confidence interval:

[0030]

[0031] Where c is a hyperparameter that controls the degree of exploration, and c > 0; is the standard UCB exploration reward term, Ds is the distance from the user's current location to the battery swapping station s; λ is the distance penalty coefficient, and λ > 0.

[0032] Further, in step 4, the online learning recommendation model based on the UCB algorithm recommends a swapping station based on the input features, including the following steps:

[0033] S1. Input the obtained features into the online learning recommendation model based on the UCB algorithm for analysis, and select a swapping station according to the set upper confidence bound UCBs(t);

[0034] S2. Obtain the selected swapping station s* = argmaxUCBs(t), and recommend it to the user;

[0035] S3. The user goes to the swapping station s* for battery swapping, and feeds back the actual cost after the battery swapping;

[0036] S4. Generate cost data based on the fed-back actual cost and update the number of attempts of the swapping station and the cumulative reward

[0037] S5. Increment the time step t = t + 1, and repeat the above steps until the stop condition is met;

[0038]

[0039] In the formula, Ethreshold is the electricity cost threshold, and Es(t) is the electricity expense of the swapping station.

[0040] Further, the information fed back by the user includes the swapping stations that receive the recommendation and the swapping stations that reject the recommendation; when the user receives the recommended swapping station, record and store it according to the location of the recommended swapping station, the average electricity cost of the swapping station, the number of battery swaps at the swapping station, and the user's ID information; and when the user conducts the next battery swap, recommend a swapping station that is the same as or similar to the recommended information this time;

[0041] When the user rejects the recommended swapping station, provide the reason for rejection at this time and give feedback; store the information fed back by the user, the recommended information, and the user's ID information; analyze the information fed back by the user, and generate relevant feedback data; adjust the relevant parameters in the recommendation model based on the feedback data, and recommend a swapping station for this user based on the adjusted recommendation model.

[0042] Further, a battery swapping recommendation system based on user behavior and load, which is used to implement the electricity cost optimization battery swapping recommendation method based on the UCB algorithm, includes:

[0043] Data collection module: used to obtain the electricity fee data for users' battery swapping, user location data, electricity fee data for battery swapping at battery swapping stations, electricity quantity data of battery swapping stations, location data of battery swapping stations, number of battery swapping times at battery swapping stations, and time data when users need battery swapping; Data processing and analysis module: used to analyze and process the obtained data; Online learning recommendation algorithm calculation module based on multi-armed bandit: used to recommend the optimal battery swapping station according to the processed data; User feedback module: used to feedback user feedback information to the online learning recommendation algorithm calculation module based on multi-armed bandit.

[0044] Furthermore, a storage medium stores an executable program, and the executable program, when executed by a processor, can implement the battery swapping recommendation method based on user behavior and load.

[0045] Beneficial effects: The battery swapping recommendation method, system, and storage medium based on user behavior and load of the present invention collect various battery swapping electricity fee data, number of battery swapping times, etc., and use an advanced online learning algorithm to dynamically adjust the recommendation strategy to optimize the user's battery swapping cost; can continuously learn and improve according to the user's real-time feedback, and provide the user with the battery swapping station selection that saves the most electricity fees; introduces a distance penalty term to comprehensively consider the distance factor of the battery swapping station and achieve global optimal recommendation; while reducing the user's battery swapping cost, it maintains the economic revenue requirements of the battery swapping station. Description of the Drawings

[0046] Figure 1 It is a block diagram of the battery swapping recommendation method based on user behavior and load. Detailed Embodiments

[0047] The present invention will be further described below with reference to the drawings.

[0048] As Figure 1 shown, the battery swapping recommendation method based on user behavior and load includes the following steps:

[0049] Step 1: Obtain data such as the electricity fee data for users' battery swapping, user location data, electricity fee data for battery swapping at battery swapping stations, electricity quantity data of battery swapping stations, location data of battery swapping stations, number of battery swapping times at battery swapping stations, and time data when users need battery swapping, etc.; The electricity fee data for users' battery swapping records the specific electricity fee for each battery swapping, electricity consumption * cabinet electricity fee unit price; User location data is the real-time location data of the user when the user currently needs battery swapping; The electricity fee data for battery swapping at battery swapping stations is the electricity fee data for battery swapping at each battery swapping station; The electricity quantity data of battery swapping stations is the available electricity quantity data of each battery swapping station. This data uses the number of available batteries for battery swapping in the battery swapping station, that is, to obtain the load of the battery swapping station; The location data of battery swapping stations is the location data of each battery swapping station.

[0050] Step 2: Calculate the average value of the electricity charges for users' battery swapping within a certain period of time, the average value of the electricity charges for battery swapping at each swapping station, and the average battery swapping frequency of each swapping station. Calculate the utilization rate of the swapping station based on the electricity quantity data and the battery swapping frequency of the swapping station; use the data processing and analysis module to analyze the electricity charges, the number of battery swaps, and the site load.

[0051] Step 3: Calculate the distances between the user's current location and the locations of each swapping station, and associate the distances between the user's current location and the locations of each swapping station, the electricity quantity data and the battery swapping frequency in each swapping station to obtain a number of battery swapping behavior association data.

[0052] Step 4: Initialize an online learning recommendation model based on the UCB algorithm, train the model on the historical data training set, evaluate the generalization ability of the model through the cross-validation method to avoid overfitting; evaluate the performance of the model through A / B testing, use the key indicators of recommendation accuracy, user click-through rate, and electricity charge savings, compare the effects of different models, and select the best model; input the average electricity charge for the user's battery swapping, the average electricity charge for battery swapping at each swapping station, the battery swapping frequency of each swapping station, the utilization rate of the swapping station, the battery swapping behavior association data, and the time data when the user needs to swap the battery as features into the trained online learning recommendation model based on the UCB algorithm to recommend the location of the optimal swapping station for the user at this time; recommend the optimal swapping station for the user when the user needs to swap the battery, and at the same time, the optimal swapping station can be recommended according to the time when the user selects to swap the battery; the online learning recommendation model based on the UCB algorithm, where the UCB algorithm refers to the algorithm that the UCB algorithm selects the best arm by calculating the upper confidence bound of each arm, which is the upper confidence bound algorithm, and the upper confidence bound is calculated based on the historical rewards of the arm and the number of times it is selected; it is a heuristic method for solving the multi-armed bandit problem, that is, the online learning recommendation algorithm calculation module based on the multi-armed bandit.

[0053] Step 5: Provide feedback information to the user feedback module according to the recommended swapping station for the user, and correct the parameters of the online learning recommendation model based on the UCB algorithm to recommend a better swapping station for the user during the next battery swapping.

[0054] Obtain real-time data such as the electricity cost data of users' battery swapping, users' location data, the electricity cost data of battery swapping at battery swapping stations, the electricity quantity data of battery swapping stations, the number of battery swapping times at battery swapping stations, and the location data of battery swapping stations, and perform data cleaning and data normalization; clean the collected original data, remove duplicate records, fill in missing values, and handle outliers; for example, for missing electricity cost data, mean filling or interpolation using data at adjacent time points can be adopted. At the same time, use statistical methods to identify and eliminate significantly abnormal electricity costs or the number of battery swapping times to ensure the accuracy and reliability of the data. Standardize data from different sources so that they can be compared under the same dimension. For numerical features such as electricity costs, use normalization or standardization methods to convert the data into values between 0 and 1 or a distribution with a mean of 0 and a variance of 1 to improve the convergence speed and accuracy of the model.

[0055] The battery swapping recommendation system based on user behavior and load is composed of a data collection module, a data processing and analysis module, an online learning recommendation algorithm calculation module based on multi-armed bandit, and a user feedback module; the data collection module collects data such as the electricity cost data of users' battery swapping, users' location data, the electricity cost data of battery swapping at battery swapping stations, the electricity quantity data of battery swapping stations, the location data of battery swapping stations, and the time data when users need to swap batteries, and transmits various data to the data processing and analysis module for analysis and processing. The results output by the data processing and analysis module are used as features and transmitted to the online learning recommendation algorithm calculation module and the user feedback module based on multi-armed bandit for calculation and processing to recommend the optimal battery swapping station to the user. The user feeds back to the online learning recommendation algorithm calculation module based on multi-armed bandit through the user feedback module according to the recommended battery swapping station to optimize and update the model, and then the online learning recommendation algorithm calculation module based on multi-armed bandit recommends the battery swapping station again.

[0056] Every time a user swaps a battery again, the system will perform a data update, continuously updating the relevant data of users and battery swapping stations, including real-time electricity costs and user behavior records, to ensure that the recommendation model makes decisions based on the latest data; the system continuously and real-time updates the relevant data of users and stations, which covers multiple key aspects; the real-time electricity cost data changes continuously with factors such as time and market supply and demand, accurately recording the electricity cost prices of each station at different times. User behavior records include information such as users' charging history, the time and frequency of visiting stations, providing a basis for comprehensively understanding user habits. By continuously updating these data, ensure that the recommendation model always makes decisions based on the latest information, thereby improving the accuracy and practicality of the recommendation.

[0057] Meanwhile, the system is equipped with system performance monitoring: regularly monitor the performance indicators of each module, including the latency of data collection and the efficiency of algorithm operation. Anomaly detection: introduce an anomaly detection mechanism to identify abnormal situations during system operation in real time and automatically alarm. Feedback mechanism: combine user feedback with system performance, and regularly adjust the recommendation strategy to improve the user experience. Regular evaluation and optimization: regularly generate system evaluation reports, analyze the accuracy of recommendations, user feedback, and electricity cost savings, and dynamically adjust algorithm parameters based on the evaluation results to adapt to market changes and user needs.

[0058] In step 2, based on the electricity bills of several battery swapping stations and users at different time periods, the changing pattern of electricity prices is obtained. The calculation process for the average value of user electricity bill data and the average value of electricity bill data for each battery swapping station within each time period T is as follows:

[0059]

[0060] In the formula, E s，t represents the electricity bill expenditure of battery swapping station s at time t, and E u，t represents the electricity bill expenditure of user u at time t;

[0061] The battery swapping frequency F of battery swapping station s within each time period T s，T , and the calculation process is as follows:

[0062]

[0063] In the formula, F s，T represents the battery swapping frequency of battery swapping station s within time period T, and R s，t represents the number of battery swaps of battery swapping station s at time t.

[0064] In step 2, according to the number of battery swaps of each battery swapping station and the power data of the battery swapping station, the power supply-demand balance index BAL s，t is defined, which is the utilization rate of the battery swapping station. The smaller the value of the power supply-demand balance index, the higher the utilization rate of the battery swapping station. The calculation formula is as follows:

[0065] BAL s，t = B s，t - R s，t

[0066] In the formula, B s，t represents the number of batteries of battery swapping station s at time t.

[0067] In step 3, according to the user location data and the battery swapping station location data, calculate the distance D from the current location of user u to each battery swapping station s u，s , and the distance D u，s, the power data and the battery replacement frequency F at the battery replacement station s，t Perform related operations to obtain the battery replacement behavior correlation data. The calculation process is as follows:

[0068]

[0069] In the formula, D u，s represents the distance from user u to the battery replacement station s at time t; a is a weight parameter used to adjust the influence of the replacement point frequency on the comprehensive index; B s，t represents the number of batteries at the battery replacement station s at time t, that is, the power of the battery replacement station s at time t.

[0070] Perform correlation analysis and feature importance evaluation on all features input to the online learning recommendation model based on the UCB algorithm, screen out the features that affect the recommendation results of the online learning recommendation model based on the UCB algorithm, reduce redundant features, and improve the performance and interpretability of the model; calculate the correlation between all features, and the correlation between two features is measured by the Pearson correlation coefficient. When the absolute value of the calculated correlation value is greater than 0.8, it is determined that the correlation between the two features is too high, and one of the two features is removed;

[0071]

[0072] In the formula, x i is the i-th x feature, y i is the i-th y feature, and are the average values of the x feature and the y feature respectively.

[0073] Sort according to the standardized coefficients after standardizing the feature coefficients fitted from all features to obtain the final feature ranking y; according to the coefficient β n of the feature x n After standardizing the absolute value of, it reflects the importance of the feature x n When the coefficient β n is close to 0, the feature importance is too low, and this feature is removed;

[0074] y = β0 + β1x1 + β2x2 + … + β nn x n + ∈

[0075] In the formula, xn is the n-th feature, βn is the coefficient of the n-th feature, and ∈ is the error term, representing the model error and other random factors and the unexplained part.

[0076] In step 4, the online learning recommendation model based on the UCB algorithm is initialized. For each swap station s, the initial number of attempts Ns(0) = 0 and the cumulative reward Hs(0) = 0 are initialized. At the same time, the initial parameter of the model alpha = 0.1;

[0077] At each time step t, for each swap station s, calculate its average reward

[0078]

[0079] Calculate the upper bound of the confidence interval UCBs(t):

[0080]

[0081] In the formula, c is a hyperparameter that controls the degree of exploration, and c > 0; is the standard UCB exploration reward term, Ds is the distance from the user's current location to the swap station s; λ is the distance penalty coefficient, and λ > 0, which is used to balance the relationship between electricity cost savings and driving costs.

[0082] In step 4, the online learning recommendation model based on the UCB algorithm recommends a swap station based on the input features, including the following steps:

[0083] S1. Input the obtained features into the online learning recommendation model based on the UCB algorithm for analysis, and select a swap station according to the set upper bound of the confidence interval UCBs(t);

[0084] S2. Obtain the selected swap station s* = argmaxUCBs(t), and recommend it to the user;

[0085] S3. The user goes to the swap station s* to swap the battery, and feedback the actual cost after swapping the battery;

[0086] S4. Generate cost data based on the feedback of the actual cost And update the number of attempts of the swap station and the cumulative reward

[0087] S5. Increment the time step t = t + 1, and repeat the above steps until the stop condition is met;

[0088]

[0089] In the formula, Ethreshold is the electricity cost threshold, which is a pre-determined reference value used to measure whether the electricity cost of the battery swapping station is within an acceptable range; Es(t) is the electricity expense of the battery swapping station; when the stop condition is met, the location of the battery swapping station output is based on the reduction of the user's battery swapping cost and ensuring that the battery swapping cost of the battery swapping station will not result in losses for the operation of the battery swapping station.

[0090] The information fed back by the user includes the battery swapping stations that receive the recommendation and the battery swapping stations that reject the recommendation; when the user accepts the recommended battery swapping station, the location of the recommended battery swapping station, the average electricity cost of the battery swapping station, the real-time electricity cost of the battery swapping station, the number of battery swaps at the battery swapping station, and the user's ID information are recorded and stored. At the same time, it can also; and based on the information of this recommendation, when the user conducts a battery swap next time, recommend a battery swapping station that is the same as or similar to the information of this recommendation; the location of the recommended battery swapping station, the average electricity cost of the battery swapping station, the real-time electricity cost of the battery swapping station, the number of battery swaps at the battery swapping station, and the user's ID information are used as features and input into an online learning recommendation model based on the UCB algorithm. Based on these data, the model parameters are updated by minimizing the loss function, so as to re-recommend a new optimal battery swapping station recommendation plan when the user conducts a battery swap next time;

[0091] When the user rejects the recommended battery swapping station, the reason for the rejection is provided and feedback is given at this time; the reason may be that the distance is too far, the cost of the battery swap is too high, or the battery swapping station is too crowded, etc.; the information fed back by the user, the recommended information, and the user's ID information are stored; the information fed back by the user is analyzed and relevant feedback data is generated; based on the feedback data, the relevant parameters in the recommendation model are adjusted, and a battery swapping station is recommended to the user based on the adjusted recommendation model. The system will not only record the reason for the rejection, but also use all the data collected during this period as features and input into an online learning recommendation model based on the UCB algorithm. During the model update process, the reason for the rejection is incorporated into the optimization algorithm as a constraint condition. For example, assuming that the reason for dissatisfaction is that a certain station is too far away, a penalty term related to the distance can be added to the loss function; or the parameters of the model can be optimized by adjusting the coefficients of the penalty term and the reward term. The model adjusts the parameters by minimizing the new loss function to avoid the reasons that cause user dissatisfaction again, and then re-recommends a new optimal battery swapping station recommendation plan.

[0092] A battery swapping recommendation system based on user behavior and load, which is used to implement the battery swapping recommendation method based on user behavior and load, and is characterized by including:

[0093] Data collection module: used to obtain the electricity fee data for users' battery swapping, users' location data, electricity fee data for battery swapping at the battery swapping station, electricity quantity data of the battery swapping station, location data of the battery swapping station, the number of battery swapping times at the battery swapping station, and the time data when users need battery swapping; Data processing and analysis module: used to analyze and process the obtained data; Online learning recommendation algorithm calculation module based on multi-armed bandit: used to recommend the optimal battery swapping station according to the processed data; User feedback module: used to feedback user feedback information to the online learning recommendation algorithm calculation module based on multi-armed bandit, and optimize the model in the online learning recommendation algorithm calculation module based on multi-armed bandit.

[0094] A storage medium stores an executable program, and when the executable program is executed by a processor, the battery swapping recommendation method based on user behavior and load can be implemented.

[0095] Take the user's feedback as a new data point and input it into the online learning algorithm to update the model parameters to reflect the latest user preferences and market dynamics. This feedback loop ensures that the model is continuously optimized during long-term operation. At the same time, based on the user's feedback information, construct a user profile to achieve personalized recommendations; for example, for users who often swap batteries during off-peak hours, stations with lower electricity fees can be preferentially recommended; according to the recorded user ID information, adjust the parameters in the recommendation model differently for different users, and recommend battery swapping stations for different users; provide dynamic and personalized battery swapping recommendations to optimize the user's battery swapping experience.

[0096] When recommending the optimal battery swapping station for users, consider the discounts on users' electricity fee expenditures and the shortening of the distance for users to travel to the battery swapping station. At the same time, also consider the utilization rate of the battery swapping station and the electricity fee expenditure of the battery swapping station; different battery swapping stations will have different electricity fees due to different locations, and different total electricity quantities and battery swapping times of the battery swapping stations will result in different utilization rates of the battery swapping stations; therefore, when recommending the battery swapping station with the lowest battery swapping cost for users, also consider the utilization rate of the battery swapping station and the electricity fee expenditure of the battery swapping station; recommend a battery swapping station that is close to the user, has a relatively low battery swapping cost, and has a low utilization rate of the battery swapping station through an online learning recommendation model based on the UCB algorithm. At this time, both the utilization rate of the battery swapping station is improved and the user's battery swapping cost is reduced, and at the same time, the congestion of user battery swapping is avoided. This is the optimal recommendation scheme; of course, at the same time, it is necessary to avoid recommending battery swapping stations at the time when users choose to swap batteries to avoid congestion during the peak period of battery swapping at this battery swapping station.

[0097] The collected data is analyzed, processed and calculated, and then used as features to be input into the online learning recommendation model based on the UCB algorithm, which significantly reduces the user's battery swapping cost. Through the intelligent recommendation algorithm, users can find a battery swapping station closer to their current location during the period with lower electricity costs, thus effectively reducing the overall battery swapping cost. During the pre-test pilot period, the average battery swapping cost of users decreased by about 15% compared with before. At the same time, the battery swapping waiting time is shortened. Since the system can monitor the site load situation in real time and recommend sites with lower loads accordingly, the waiting time for users to queue at the battery swapping station is greatly reduced, improving the battery swapping efficiency. The user satisfaction is improved. According to the user feedback information collected by the system, more than 80% of the users are satisfied with the recommendation results of the new system, especially those users who often swap batteries at night. They generally believe that the recommended sites are more in line with their travel habits. The operator's revenue increases: Since the utilization rate of the battery swapping station is optimized, the operator can increase revenue through more reasonable resource allocation; at the same time, by reducing the electricity cost expenditure, the overall operating cost of the operator also decreases. Contribution to energy conservation and emission reduction. This system indirectly promotes the balanced use of power resources by guiding users to swap batteries during off-peak hours, which helps to reduce energy waste during the peak period of the power grid load and plays a positive role in environmental protection. The resource utilization rate is improved. Through the dynamic analysis of the load status of the battery swapping station, the system helps the operator better allocate resources and avoids the situation where some sites are overcrowded while other sites have low utilization rates.

[0098] The above is only a description of the preferred embodiments of the present invention. Those of ordinary skill in the art can make several modifications and optimizations based on the above disclosure without departing from the basic principle content. These improvements and optimizations should be regarded as the protection scope understood by the present invention.

Claims

1. A method for battery swapping recommendation based on user behavior and load, characterized in that: It includes the following steps: Step 1: Obtain the electricity fee data of the user's battery swapping, the user location data, the electricity fee data of the battery swapping at the battery swapping station, the power quantity data of the battery swapping station, the location data of the battery swapping station, the number of battery swapping times at the battery swapping station, and the time data when the user needs to swap the battery; Step 2: Calculate the average value of the electricity fee data of the user's battery swapping within a period of time, the average value of the electricity fee data of the battery swapping at each battery swapping station, and the average battery swapping frequency of each battery swapping station. Calculate the utilization rate of the battery swapping station according to the power quantity data and the battery swapping frequency of the battery swapping station; Step 3: Calculate the distances between the user's current location and the locations of each battery swapping station, and associate the distances between the user's current location and the locations of each battery swapping station, the power quantity data and the battery swapping frequency in each battery swapping station to obtain a number of battery swapping behavior association data; Step 4: Initialize the online learning recommendation model based on the UCB algorithm. Use the average electricity fee of the user's battery swapping, the average electricity fee of the battery swapping at each battery swapping station, the battery swapping frequency of each battery swapping station, the utilization rate of the battery swapping station, the battery swapping behavior association data, and the time data when the user needs to swap the battery as features and input them into the trained online learning recommendation model based on the UCB algorithm to recommend the location of the optimal battery swapping station for the user at this time; Step 5: Provide feedback information according to the recommended battery swapping station for the user, correct the parameters of the online learning recommendation model based on the UCB algorithm, and recommend a better battery swapping station when the user swaps the battery next time.

2. The battery swapping recommendation method based on user behavior and load according to claim 1, wherein: In the said Step 2, the change rule of the electricity fee price is obtained through the electricity fee expenditures of several battery swapping stations and users in different time periods. The calculation process of the average value of the user's electricity fee data and the average value of the electricity fee data of each battery swapping station within each time period T is as follows: where, E s，t represents the electricity cost of the battery swapping station s at time t, and E u，t represents the electricity cost of the user u at time t; The battery swapping frequency F of the battery swapping station s within each time period T s，T , and the calculation process is as follows: where F s，T represents the battery swapping frequency of the swapping station s within the time period T, and R s，t represents the number of battery swapping times of the swapping station s at time t.

3. The battery swapping recommendation method based on user behavior and load according to claim 1, characterized in that: In the second step, according to the battery swapping times and power data of each battery swapping station, define the power supply and demand balance index BAL s，t , which is the utilization rate of the battery swapping station, and the calculation formula is as follows: BAL s，t = B s，t - R s，t Where B s，t represents the number of batteries at the battery swapping station s at time t.

4. The battery swapping recommendation method based on user behavior and load according to claim 1, wherein: In the third step, the distance D from the current location of user u to each battery swapping station s is calculated according to the user location data and the battery swapping station location data u，s , and the distance D u，s , the power data and the battery swapping frequency F in the battery swapping stations s，t are associated to obtain the battery swapping behavior association data. The calculation process is as follows: where D u，s represents the distance from user u to the swapping station s at time t; α is the weight parameter; B s，t represents the number of batteries at the swapping station s at time t, that is, the power of the swapping station s at time t.

5. The method for recommending battery swapping based on user behavior and load according to claim 4, wherein: Conduct a correlation analysis and an evaluation of the feature importance for all features input into the online learning recommendation model based on the UCB algorithm, and screen out the features that affect the recommendation result of the online learning recommendation model based on the UCB algorithm; calculate the correlation between all features. The correlation between two features is measured by the Pearson correlation coefficient. When the absolute value of the calculated correlation value is greater than 0.8, it is determined that the correlation between the two features is too high, and one of the two features is removed; Sort the standardized coefficients after standardizing the feature coefficients fitted according to all features to obtain the sorting y of the final features; according to the feature x n The coefficient β n The magnitude of the absolute value of reflects the importance of the feature x n When the coefficient β n is close to 0, the importance of the feature is too low, and this feature is removed; y = β0 + β1x1 + β2x2 + … + β n x n + ∈ In the formula, xn is the nth feature, βn is the coefficient of the nth feature, and ∈ is the error term.

6. The battery swapping recommendation method based on user behavior and load according to claim 1, wherein: In the said Step 4, when initializing the online learning recommendation model based on the UCB algorithm, initialize the number of trial times Ns(0)=0 and the cumulative reward Hs(0)=0 for each battery swapping station s; At each time step t, for each battery swapping station s, calculate its average reward Calculate the upper bound UCBs(t) of the confidence interval: where c is a hyperparameter that controls the exploration degree and c > 0; is the standard UCB exploration reward term, Ds is the distance from the user's current location to the swapping station s; λ is the distance penalty coefficient and λ >

0.

7. The method for battery swapping recommendation based on user behavior and load according to claim 6, wherein: In the said Step 4, the online learning recommendation model based on the UCB algorithm recommends a battery swapping station based on the input features, including the following steps: S1: Input the obtained features into the online learning recommendation model based on the UCB algorithm for analysis, and select a battery swapping station according to the set upper bound UCBs(t) of the confidence interval; S2: Obtain the selected battery swapping station s* = argmaxUCBs(t), and recommend it to the user; S3: The user goes to the battery swapping station s* to swap the battery, and feedback the actual cost after the battery swapping. S4. Generate cost data based on the actual feedback cost And update the number of attempts at the battery swapping station And the cumulative rewards S5. Increment the duration step \(t = t + 1\), and repeat the above steps until the stop condition is met; Where \(E_{threshold}\) is the electricity cost threshold, and \(E_s(t)\) is the electricity cost expenditure of the swapping station.

8. The battery swapping recommendation method based on user behavior and load according to claim 1, wherein: The information fed back by the user includes the swapping stations that receive the recommendation and the swapping stations that reject the recommendation; when the user accepts the recommended swapping station, record and store it according to the location of the recommended swapping station, the average electricity cost of the swapping station, the number of swapping times at the swapping station, and the user's ID information; And based on the information of this recommendation, when the user swaps electricity next time, recommend swapping stations that are the same as or similar to the information of this recommendation to the user; When the user rejects the recommended swapping station, provide the reason for rejection at this time and give feedback; Store the information fed back by the user, the recommended information, and the user's ID information; analyze the information fed back by the user and generate relevant feedback data; Adjust the relevant parameters in the recommendation model based on the feedback data, and recommend swapping stations for this user based on the adjusted recommendation model.

9. A battery swapping recommendation system based on user behavior and load, which is used to implement the battery swapping recommendation method based on user behavior and load according to any one of claims 1 to 8, characterized in that: Including: Data collection module: used to obtain the electricity cost data of the user's electricity swapping, the user location data, the electricity cost data of the swapping station's electricity swapping, the electricity quantity data of the swapping station, the location data of the swapping station, the number of swapping times at the swapping station, and the time data when the user needs to swap electricity; Data processing and analysis module: used to analyze and process the obtained data; Online learning recommendation algorithm calculation module based on multi-armed bandit: used to recommend the optimal swapping station according to the processed data; User feedback module: used to feedback the user feedback information to the online learning recommendation algorithm calculation module based on multi-armed bandit.

10. A storage medium, characterized in that: It stores an executable program, and when the executable program is executed by a processor, it can implement the electricity swapping recommendation method based on user behavior and load according to any one of claims 1 to 8.

Images