A dynamic recommendation method for charging stations based on user preferences

Abstract

The present invention discloses a method for dynamically recommending charging stations based on user preferences, comprising: calculating a reachable distance according to the remaining power of a user's vehicle, screening charging stations marked on a map according to the current user's vehicle position and the reachable distance, and obtaining a recommended candidate pool; calculating the charging cost of each charging station in the recommended candidate pool, screening several charging stations with the smallest charging costs in the recommended candidate pool according to the charging cost to form a charging station candidate list; obtaining the user's preference tags and preference tag weights, scoring and sorting the charging stations in the charging station candidate list according to the preference tags and their types and preference tag weights, and obtaining a user preference recommendation list.

Description

Technical Field

[0001] The present invention relates to the technical field of dynamic recommendation of charging stations, and in particular to a method for dynamic recommendation of charging stations based on user preference. Background Art

[0002] With the popularization and promotion of new energy vehicles, charging stations, as an important supporting infrastructure, have also entered a period of rapid construction and development in recent years. When new energy vehicle owners have charging needs, they often need to use mobile phone software to find nearby charging stations.

[0003] There are two main ways. One is to search through map software. Car owners can quickly check the distance of charging stations. Generally, they filter by the relative distance between the charging station and the vehicle. However, if you want to know the specific information of the charging station, you have to click on the options of each charging station separately. The operation is cumbersome, and you may not be able to get the information you are interested in due to incomplete charging station data. The other is to find charging stations through charging assistant software developed by charging station operators. This type of software contains relatively complete surrounding data, but the software can only view the charging station information of a single or cooperative operating company, which is not convenient for users to make horizontal comparisons.

[0004] Furthermore, some existing softwares have functions that can filter recommendations based on user preferences, but users can only choose from a few fixed preference tags and cannot dynamically choose based on their actual situation. In addition, since user preference tags have different data types, if the unified preference recommendation calculation method in the prior art is used, there may be recommendation calculation deviations, and fixed calculation weights will also lead to recommendation calculation deviations.

[0005] In view of the above problems in the prior art, the purpose of the present invention is to design a dynamic recommendation method for charging stations based on user preferences. Summary of the invention

[0006] In view of this, an object of the present invention is to propose a method for dynamically recommending charging stations based on user preferences, which can solve the above-mentioned problems.

[0007] The present invention provides a method for dynamically recommending charging stations based on user preferences, comprising:

[0008] Calculate the reachable distance based on the remaining power of the user's vehicle, filter the charging stations marked on the map based on the current user's vehicle location and reachable distance, and obtain the recommended alternative pool;

[0009] Calculate the charging cost of each charging station in the recommended candidate pool, and select several charging stations with the lowest charging cost in the recommended candidate pool according to the charging cost to form a charging station candidate list;

[0010] Obtain the user's preference tags and preference tag weights, score and sort the charging stations in the charging station candidate list according to the preference tags, their types, and preference tag weights, and obtain the user's preference recommendation list. Specifically:

[0011] Determine the type of the preference tag, where the types of the preference tag include: Boolean, numerical, text, and conditional, and score the charging stations according to the type of the preference tag to obtain a score for each type of preference tag for each charging station;

[0012] The total user preference score of each charging station is calculated according to the preference tag weight and the preference tag score, and the charging stations are sorted according to the total user preference score of each charging station to obtain a user preference recommendation list.

[0013] Further, the method of filtering the charging stations marked on the map according to the current user vehicle location and the reachable distance to obtain a recommended candidate pool includes:

[0014] Get the current user's vehicle location, take the current user's vehicle location as the center and the reachable distance as the radius, retrieve the charging stations marked on the map, and form a charging station candidate pool with all the charging stations within the circular area;

[0015] The actual distance from each charging station in the charging station candidate pool to the current user vehicle position is calculated, and charging stations with actual distances greater than the accessible distance are removed from the charging station candidate pool to obtain a recommended candidate pool.

[0016] Furthermore, the calculation of the charging cost of each charging station in the recommended candidate pool includes:

[0017] The distance cost Cd is calculated by the actual distance between the current user vehicle location and each charging station in the recommended candidate pool;

[0018] The charging cost Ce is calculated by the real-time electricity price and service fee of the charging station, and the waiting time cost Ct is calculated by the idle situation of the charging station;

[0019] The charging cost C is calculated based on the distance cost Cd, charging fee cost Ce, and waiting time cost Ct.

[0020] Furthermore, the charging station scoring is performed according to the type of preference tag to obtain the preference tag score of each type of each charging station, including:

[0021] If the preference tag is of Boolean type, the preference tag is assigned a value, a result that is favorable to vehicle charging is assigned a value of 1, a result that is unfavorable to vehicle charging is assigned a value of 0, and a score is calculated according to the result assignment;

[0022] If the preference tag is a numerical value, the ranking and scoring are performed according to the specific numerical value type of the preference tag, and the numerical value type includes a dynamic numerical value and a static numerical value;

[0023] If the preference label is text type, the text of the preference label is graded to obtain a level value, and a weighted sum is performed according to the level value and the level value weight to obtain the numerical value of the text type preference label, and the text type preference label is sorted and scored according to the numerical value;

[0024] If the preference tag is of conditional type, the content of the preference tag is converted into a numerical value for sorting and scoring according to the type of conditional content.

[0025] Furthermore, the text of the preference label is graded to obtain a level value, and a value of the text type preference label is obtained by weighted summation according to the level value and the level value weight, and before sorting and scoring by the value of the text type preference label:

[0026] Determine whether the text of the preference tag is real-time text, if so, obtain real-time text data, and classify according to the real-time text;

[0027] Otherwise, extract keywords from the text in the data set and perform semantic classification to obtain text keywords and their classifications, perform graded word frequency statistics on the text keywords and their classifications, obtain text descriptions based on the average evaluation values, and perform grading based on the text descriptions.

[0028] Furthermore, the preference label weight and level value weight are obtained through the following steps:

[0029] Obtain user feedback data on the preference recommendation list, extract a weighted feedback vector according to the type of feedback data, input the weighted feedback vector and the charging station preference label into a pre-trained preference label weight optimization model, and obtain the preference label weight;

[0030] The feedback data of the preference recommendation list containing text labels is obtained, and the label level feedback vector is extracted according to the type of feedback data. The label level feedback and the charging station preference label are input into the pre-trained level value weight optimization model to obtain the level value weight.

[0031] Furthermore, the weight feedback vector and the label level feedback vector are obtained by the following steps:

[0032] Determine whether the selected charging station belongs to the first or second charging station in the user's preference recommendation list. If yes, mark the weight feedback vector and the label level feedback vector as positive feedback; otherwise, mark the weight feedback vector and the label level feedback vector as negative feedback;

[0033] If the selected charging station is not in the user's preferred recommendation list, the weight feedback vector and the label level feedback vector are marked as severe negative feedback;

[0034] The user's evaluation of the preference recommendation list is obtained. If the evaluation is satisfactory, the weighted feedback vector and the label-level feedback vector are marked as positive feedback. If the evaluation is unsatisfactory, the weighted feedback vector and the label-level feedback vector are marked as negative feedback.

[0035] Further, after obtaining the user preference recommendation list, the following steps are performed:

[0036] The preference labels of charging stations are modified based on the correction information provided by users.

[0037] Further, the modifying the preference label of the charging station according to the modification information fed back by the user includes:

[0038] Obtain the correction information of the question preference label fed back by the user, and determine whether the question preference label is a dynamic numerical type or a real-time text type. If so, no correction is made; otherwise, push the correction information of the question preference label fed back by the user to several users who have recently gone to the charging station to charge, and collect user replies;

[0039] Determine whether the number of user replies meets the reply threshold. If yes, calculate the number of user replies that are adopted, otherwise continue to push the revised information of the question preference label;

[0040] It is determined whether the number of user replies that adopt the modification exceeds half of the total number of user replies. If so, the question preference label is modified; otherwise, the question preference label is not modified.

[0041] Beneficial effects of the present invention:

[0042] First, a pool of charging station candidates can be quickly obtained through dual distance screening, and the actual reachable range of the vehicle with the remaining power can be accurately calculated. From the numerous charging stations, reasonable options that users can actually reach can be selected to avoid users being unable to reach the charging station due to insufficient power when looking for it.

[0043] The second is to quantify and compare the comprehensive use costs of different charging stations, including distance, charges and waiting time. By accurately calculating factors such as distance cost, charging fee, and waiting time cost, the most economical charging option can be determined.

[0044] Third, by classifying the preference tags selected by users and identifying different types of preference tags, different scoring strategies can be applied to standardize and quantify complex and diversified preference information. By analyzing user choices and feedback, the system can identify which tags and features are more in line with user preferences, and then through neural network training, the weights of these tags can be dynamically adjusted to better reflect changes in user preferences. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 It is a flow chart of the method in embodiment 1.

[0047] Figure 2 This is a schematic diagram of the first distance screening in the first embodiment.

[0048] Figure 3 This is a schematic diagram of the second distance screening in the first embodiment.

[0049] Figure 4 This is a schematic diagram of a user selecting a label in the first embodiment.

[0050] Figure 5 This is a scoring flow chart of the text-based preference tags in the first embodiment.

[0051] Figure 6 This is a flowchart of charging station information correction in the first embodiment. DETAILED DESCRIPTION

[0052] To facilitate understanding by those skilled in the art, the structure of the present invention is now further described in detail with reference to the embodiments and the accompanying drawings. It should be understood that the steps mentioned in the present embodiment, unless otherwise specified in their order, can be adjusted in order according to actual needs, and can even be executed simultaneously or partially simultaneously.

[0053] Embodiment 1

[0054] like Figure 1 As shown, an embodiment of the present invention provides a method for dynamically recommending charging stations based on user preferences, including:

[0055] S1 calculates the reachable distance based on the remaining power of the user's vehicle, and selects the charging stations marked on the map based on the current user's vehicle location and the reachable distance to obtain a recommended candidate pool;

[0056] Specifically, the calculation formula for calculating the reachable distance based on the remaining power of the user's vehicle is as follows:

[0057] L = E × λ,

[0058] Among them, L is the reachable distance, E is the remaining power of the user's vehicle, and λ is the empirical coefficient.

[0059] In this step, the empirical coefficient λ can be 60%. New energy vehicles generally have a situation where the actual cruising range caused by actual working conditions is inconsistent with the theoretical cruising range. Therefore, this solution adopts a conservative method to estimate the reachable distance.

[0060] S101 obtains the current user vehicle position, takes the current user vehicle position as the center of the circle and the reachable distance as the radius, searches for charging stations marked on the map, and forms a charging station candidate pool with all charging stations within the circular area;

[0061] In this step, the user's current vehicle location is taken as the center of the circle and the reachable distance is taken as the radius to retrieve the charging stations on the map and perform the first distance screening. The charging stations within the circular area form a charging station candidate pool, such as Figure 2 As shown, the user's surrounding charging stations can be quickly obtained through the first distance screening.

[0062] S102 calculates the actual distance from each charging station in the charging station candidate pool to the current user vehicle location, removes charging stations with actual distances greater than accessible distances from the charging station candidate pool, and obtains a recommended candidate pool.

[0063] In this step, the actual distance from each charging station in the charging station candidate pool to the user's current location is calculated. The actual length of the road that the vehicle can drive on will be considered when calculating the distance. Figure 3 As shown, it is ensured that the charging stations in the alternative pool are accessible to the vehicle before the battery is exhausted and the vehicle will not break down along the way.

[0064] S2 calculates the charging cost of each charging station in the recommended candidate pool, and selects several charging stations with the lowest charging cost in the recommended candidate pool according to the charging cost to form a charging station candidate list;

[0065] S201 calculates the distance cost Cd by the actual distance between the current user vehicle location and each charging station in the recommended candidate pool;

[0066] In this step, the longer the distance between the vehicle and the charging station, the greater the distance cost. Assume that the actual distance between the vehicle and the charging station is d (km), the distance cost coefficient is α, and the distance cost Cd = α·d. The distance cost coefficient is an empirical value that adjusts the distance to a value that is easy to compare with other costs. The default value is 10.

[0067] S202 calculates the charging cost Ce according to the real-time electricity price and service fee of the charging station, and calculates the waiting time cost Ct according to the idle situation of the charging station;

[0068] In this step, the charging cost Ce is the fee paid by the user to the charging station, including electricity price and service fee, which is estimated by reading the charging station charging information and preferential activities. If the vehicle needs to queue up to charge at the charging station, this part of the cost will be incurred, otherwise the cost is 0. Let the waiting time coefficient be β, the vehicle waiting time be t (minutes), and the waiting time cost Ct = β·t. The waiting time coefficient is an empirical value, which adjusts the time to a value that is easy to compare with other costs. The default value is 1.

[0069] S203 calculates the charging cost C according to the distance cost Cd, the charging fee cost Ce, and the waiting time cost Ct.

[0070] In this step, the charging cost C is calculated as: C = Cd + Ce + Ct = α·d + Ce + β·t. These three cost dimensions can comprehensively measure the actual consumption and risks of users going to a charging station from the three core aspects of economy, time and distance.

[0071] S3 obtains the user's preference tags and preference tag weights, scores and sorts the charging stations in the charging station candidate list according to the preference tags and their types and preference tag weights, and obtains the user's preference recommendation list.

[0072] In this step, the user selects 1-5 tags that he / she considers most important and concerned about from the tag pool, such as Figure 4 As shown. Figure 4 In the figure, dark colors represent the tags selected by the user. When calculating the preference score, only the information corresponding to the selected tags is considered, and the rest of the information is not considered.

[0073] Assume that a user is an electric taxi owner and hopes to find a charging station with a rest area, unobstructed surrounding roads, close to a transportation hub, large flow of people in the surrounding area, and no additional parking fees, so he checks 5 tags. There are 5 charging stations in the charging station candidate list. The specific information of the 5 selected tags of the 5 charging stations is shown in Table 1:

[0074] Table 1 Table of user preference labels corresponding to each charging station

[0075]

[0076] S301 determines the type of the preference tag, where the types of the preference tag include: Boolean, numerical, text, and conditional, and scores the charging stations according to the type of the preference tag to obtain a score for each type of preference tag for each charging station;

[0077] In this step, by identifying different types of preference tags, different scoring strategies can be applied to standardize and quantify complex and diversified preference information. The maximum score of the user preference score can be set to 100, and the maximum score of each tag is obtained by averaging the number of tags selected by the user and multiplying it by the tag weight: .

[0078] S3011 If the preference tag is a Boolean type, assign a value to the preference tag, assigning a value of 1 to a result that is favorable for vehicle charging, and assigning a value of 0 to a result that is unfavorable for vehicle charging, and scoring according to the result assignments;

[0079] In this step, the information corresponding to the Boolean preference tag has only two values: "yes" and "no". First, assign a value to the tag. The result that is beneficial to the vehicle charging or the owner is assigned a value of 1, and the unfavorable result is assigned a value of 0. When scoring, the score of the Boolean preference tag also has only two values. If the tag value is 1, the tag is full score, otherwise it is 0.

[0080] The following uses the "rest area" tag as an example to illustrate the scoring method. If the charging station has a rest area, the corresponding information of the tag is "yes", the value is 1, and the tag score is full. Otherwise, the score is 0. Figure 4 Among them, “non-open-air parking lot”, “with fast charging pile”, “with rest area”, “with toilet”, and “with convenience store” are all Boolean labels. The scores of the five charging stations on the label of “with rest area” are shown in Table 2.

[0081] Table 2 Scores of charging stations on Boolean preference labels

[0082]

[0083] S3012: if the preference tag is a numerical value, ranking and scoring are performed according to the specific numerical value type of the preference tag, where the numerical value type includes a dynamic numerical value and a static numerical value;

[0084] In this step, the information corresponding to the numerical preference tag is a specific numerical value, which can be further divided into static numerical tags and dynamic numerical tags according to whether the numerical value is static data or real-time updated data.

[0085] Static numeric preference tags:

[0086] The value is static data and will not change over a period of time. For example, the information corresponding to the "Near Transportation Hub" label is the distance between the charging station and the nearest transportation hub, such as an airport or high-speed rail station. This value usually does not change and is a static value.

[0087] The charging stations are ranked according to the values ​​on the labels, with unified dimensions. The score = ranking percentage × full score of the label.

[0088] When ranking, the options are arranged in order of whether they are beneficial to vehicle charging or car owners from small to large or from large to small. Figure 4 The “near transportation hub”, “near highway intersection” and “near commercial district” in the figure are all static numerical labels. The scores of the five charging stations on the “near transportation hub” label are shown in Table 3.

[0089] Table 3 Scores of charging stations on static numerical preference labels

[0090]

[0091] Dynamic numerical preference tags:

[0092] The values ​​are dynamic data, and the real-time data will be read from the data interface for statistical analysis each time the score is calculated. For example, the information corresponding to the label "Nearby roads are unobstructed" is the ratio of the length of unobstructed roads to the total distance from the vehicle to the charging station. Whether the road is unobstructed is dynamic, and the system reads the real-time road traffic conditions from the map software for statistics, ranking, and scoring. Figure 4 The “smooth surrounding roads” and “many repeat customers” in the figure are both dynamic numerical labels. The scores of the five charging stations on the “smooth surrounding roads” label are shown in Table 4.

[0093] Table 4 Scores of charging stations on dynamic numerical preference labels

[0094]

[0095] S3013: If the preference tag is text type, the text of the preference tag is graded to obtain a level value, and a value of the text type preference tag is obtained by weighted summation according to the level value and the level value weight, and the text type preference tag is sorted and scored according to the value;

[0096] Furthermore, the text of the preference label is graded to obtain a level value, and a value of the text type preference label is obtained by weighted summation according to the level value and the level value weight, and before sorting and scoring by the value of the text type preference label:

[0097] Determine whether the text of the preference tag is real-time text, if so, obtain real-time text data, and classify according to the real-time text;

[0098] Otherwise, extract keywords from the text in the data set and perform semantic classification to obtain text keywords and their classifications, perform graded word frequency statistics on the text keywords and their classifications, obtain text descriptions based on the average evaluation values, and perform grading based on the text descriptions.

[0099] In this step, text-based preference labels are usually unstructured data that is difficult to use directly for calculation and comparison. By grading these texts, they can be converted into a computable numerical form, so that they can be considered together with other quantitative indicators. The information corresponding to the text-based label is text, which can be divided into real-time text preference labels and long-term text preference labels according to the content of the text. The text preference label scoring process is as follows: Figure 5 shown.

[0100] Real-time text label:

[0101] Real-time text needs to be updated frequently, and its content can quickly reflect the dynamic changes in the current situation. Real-time classification ensures that such changes can be immediately reflected in the value of the preference tag. Read the real-time data at each scoring, and grade and score according to the real-time data. The following uses the "surrounding area traffic" label as an example to explain the scoring method. Read the real-time traffic information around the charging station from the data interface of the map software, and the returned information is a description of the traffic. Divide the traffic into 5 levels, and define the level according to the principle of being beneficial to vehicle charging or to the owner: very small, small, medium, large, and very large. The level value and weight of each level are shown in Table 5.

[0102] Table 5 Level value and weight of each level of real-time text preference label

[0103]

[0104] The initial level weight is 1. As users use the system, the system will collect user feedback and continuously adjust the weight. Score = level weight × level value × label full score. Figure 4 The “traffic flow in the surrounding area” in the figure belongs to the real-time text preference label. The values ​​and scores of the five charging stations on the label “traffic flow in the surrounding area” are shown in Table 6.

[0105] Table 6 Assignment and score of real-time text preference labels for charging stations

[0106]

[0107] Long-term text preference tags:

[0108] Long-term text has a low frequency of change, and its processing can focus more on trend analysis and comprehensive evaluation of historical data. Therefore, keywords and semantic classification are used for more in-depth classification and weighting. The user evaluation information in the long-term accumulated map software and charging software operated by various charging companies is used as a data set. By extracting keywords, semantic classification, word frequency statistics and other methods, a relatively stable text description of a certain label of a charging station over a period of time is obtained.

[0109] For example, for the label “facility maintenance status”, keywords are extracted from user evaluation information. Through semantic classification, similar evaluations with similar meanings such as “the facilities are very good”, “charging is fast”, or “the charging guns are all broken”, “the machine often fails”, etc. are aggregated. According to the statistical distribution of the evaluations, the interval of the mean is calculated to obtain the corresponding description, as shown in Table 7.

[0110]

[0111] Table 7 Correspondence table of evaluation mean and interval

[0112]

[0113] Since users are usually more sensitive to negative reviews, this application adjusts the negative review multiplier through a step function to amplify the impact of negative reviews.

[0114] An example of a step function design:

[0115]

[0116] Assume that there are 132 reviews for a charging station, of which 16 reviews are positive reviews of "facility maintenance" through semantic classification, and 2 reviews are negative reviews of "facility maintenance". The above method can be used to calculate the average review value of 0.56, and the text description is "medium". The same method as the real-time text label is used for scoring. The level value and weight of each level of the long-term text preference label are shown in Table 8. When the initial value of the level weight is 1, the full score of the label is 20 points, and the score of the charging station is 10 points. Figure 4 The "service attitude", "fuel truck occupancy", "facility maintenance" and "site environmental sanitation" are all long-term text labels.

[0117] Table 8 Level value and weight of each level of long-term text preference label

[0118]

[0119] S3014: If the preference tag is of conditional type, the content of the preference tag is converted into a numerical value for sorting and scoring according to the type of conditional content.

[0120] In this step, conditional tags are tags with conditions in their content, and different conditions correspond to different information. First, determine which condition the vehicle is charging at the charging station under, read the information corresponding to the condition, and then process it according to the above method based on whether the information is Boolean, text, or numeric type.

[0121] The following uses the "Extra Parking Fee" label as an example to explain the scoring method. The system first estimates the vehicle charging time and selects the value corresponding to the condition. Assuming that the vehicle charging time is 45 minutes, the corresponding charge of charging station 1 is 5 yuan, and the remaining charging stations are free. At this time, the result has only two values, which are processed as Boolean type, which is conducive to the result of the charging station being assigned a value of 1, and the full score of the label is obtained, as shown in Table 9. Figure 4 The "additional parking fees" and "overtime parking fees" are both conditional labels.

[0122] Table 9 Scores of charging stations on conditional preference labels

[0123]

[0124] S302 calculates the total user preference score of each charging station according to the preference tag weight and the preference tag score, and sorts the charging stations according to the total user preference score of each charging station to obtain a user preference recommendation list.

[0125] In this step, the preference tag weight is used to adjust the proportion of each tag in the total score. The initial preference tag weight is 1. As users use it, the system will collect user feedback and continuously revise the weight value so that the scoring system can be as close to user preferences as possible.

[0126] Assuming that this is the user’s first use, 5 tags are selected, the initial weight is 1, and the full score of each tag is 20. The scores and totals of the 5 charging stations on the 5 tags are shown in Table 10.

[0127] Table 10 Preference label scores and sums for each charging station

[0128]

[0129] Sort by total score, and sort the stations with the same score by charging cost from small to large to get the final charging station recommendation list. Assuming that the order of charging station cost from small to large is charging station 1 < charging station 2 < charging station 3 < charging station 4 < charging station 5, the user preference recommendation list in this example is shown in Table 11.

[0130] Table 11 User preference recommendation list

[0131]

[0132] Furthermore, the preference label weight and level value weight are obtained through the following steps:

[0133] Obtain user feedback data on the preference recommendation list, extract a weighted feedback vector according to the type of feedback data, input the weighted feedback vector and the charging station preference label into a pre-trained preference label weight optimization model, and obtain the preference label weight;

[0134] The feedback data of the preference recommendation list containing text labels is obtained, and the label level feedback vector is extracted according to the type of feedback data. The label level feedback and the charging station preference label are input into the pre-trained level value weight optimization model to obtain the level value weight.

[0135] Specifically, the weight feedback vector and the label level feedback vector are obtained through the following steps:

[0136] Determine whether the selected charging station belongs to the first or second charging station in the user's preference recommendation list. If yes, mark the weight feedback vector and the label level feedback vector as positive feedback; otherwise, mark the weight feedback vector and the label level feedback vector as negative feedback;

[0137] If the selected charging station is not in the user's preferred recommendation list, the weight feedback vector and the label level feedback vector are marked as severe negative feedback;

[0138] The user's evaluation of the preference recommendation list is obtained. If the evaluation is satisfactory, the weighted feedback vector and the label-level feedback vector are marked as positive feedback. If the evaluation is unsatisfactory, the weighted feedback vector and the label-level feedback vector are marked as negative feedback.

[0139] In this step, after receiving the recommendation list, the user chooses the charging station to go to. If the user chooses the charging station ranked 1st or 2nd in the recommendation list, the system records it as positive feedback. If the user chooses the other charging stations, the system records it as negative feedback. If the user does not choose a charging station in the recommendation list, the system records it as severe negative feedback. If the user chooses a charging station in the recommendation list to charge, he can evaluate the recommended service after charging. If he is satisfied with the recommendation result, the system records it as positive feedback, otherwise it records it as negative feedback. The system regularly trains through neural network methods to optimize adjustable parameters such as label weights and level weights that affect user preference scoring.

[0140] Specifically, the generated recommendation list is used as a data set, each list is a piece of data, and the labels in the list are features. The label information of each charging station is converted into a feature value, and the label score is converted into a feature vector. At regular intervals, the system collects all user feedback using the same label combination collected during this period to form a weighted feedback vector. The weighted feedback vector and the data set are used as input together, and the label weight is adjusted through parameter fitting. At regular intervals, the system collects all user feedback collected during this period that selects a certain text-type label to form a label-level feedback vector. The label-level feedback vector and the data set are used as input together, and the level weight of a certain text-type label is adjusted through parameter fitting.

[0141] Different users may attach different importance to different tags. Through neural network training, the weights of these tags can be dynamically adjusted to better reflect changes in user preferences. Text tags (such as user reviews, service quality descriptions, etc.) are usually difficult to quantify. By analyzing user feedback, the level weights of these text tags can be adjusted, making the recommendation system's understanding and use of text information more accurate.

[0142] Further, after obtaining the user preference recommendation list, the following steps are performed:

[0143] The preference labels of charging stations are modified based on the correction information provided by users.

[0144] Specifically, the step of modifying the preference label of the charging station using the modification information fed back by the user includes:

[0145] Obtain the correction information of the question preference label fed back by the user, and determine whether the question preference label is a dynamic numerical type or a real-time text type. If so, no correction is made; otherwise, push the correction information of the question preference label fed back by the user to several users who have recently gone to the charging station to charge, and collect user replies;

[0146] Determine whether the number of user replies meets the reply threshold. If yes, calculate the number of user replies that are adopted, otherwise continue to push the revised information of the question preference label;

[0147] It is determined whether the number of user replies that adopt the modification exceeds half of the total number of user replies. If so, the question preference label is modified; otherwise, the question preference label is not modified.

[0148] In this step, if Figure 6 As shown, if the user believes that the data of a label of a charging station in the recommended list is inaccurate, he can apply to the system for correction by clicking on the label. If the label is a dynamic numerical label or a real-time text label, since the value corresponding to the label is read in real time, the system will re-read the data every time the score is calculated. Therefore, these two types of labels do not support user modification of information and will return to the user: modification of this type of information is not supported. If the label belongs to other types, the system pops up an input box and the user enters the corrected information. The system pushes the user's corrected information to several users who recently went to the charging station to charge. The invited user can choose to reply "Adopt the modification" or "Do not accept the modification". The system starts a timer, and the reply before the end of the timer is considered a valid reply, otherwise it is an invalid reply. The timing length should not be too short, the default is 24 hours.

[0149] The number of valid replies should exceed a certain proportion of the number of users invited to reply and be rounded up. For example, if the number of invited users is 30 and the proportion is 60%, the number of valid replies should exceed 18. If the number of valid replies is insufficient, the replies will continue to be pushed to other users who have visited the charging station to charge in order of the date of their visit to the charging station from recent to far, based on the number of valid replies that are still missing. The replies will be pushed for a maximum of several times, with the default being 3 times. After several times, the push will stop regardless of whether the number of valid replies is met. Statistics are kept to see if the number of replies that choose "adopt and modify" among the valid replies exceeds half. If so, the system will modify the label data of the charging station, otherwise the information will not be modified. A closed loop from discovery, feedback, verification to modification has been established to form a stable user feedback mechanism and continuously improve system data.

[0150] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0151] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0152] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0153] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0154] It should be noted that in the claims, any reference signs placed between brackets shall not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claim. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention may be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third etc. does not indicate any order. These words may be interpreted as names.

[0155] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0156] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

[0157] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0158] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms should not be understood as necessarily referring to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

Claims

1. A method for dynamically recommending charging stations based on user preferences, characterized in that: include: The reachable distance is calculated based on the remaining power of the user's vehicle, and the charging stations marked on the map are filtered according to the current user's vehicle location and the reachable distance to obtain the recommended candidate pool. Specifically: Get the current user's vehicle location, take the current user's vehicle location as the center and the reachable distance as the radius, retrieve the charging stations marked on the map, and form a charging station candidate pool with all the charging stations within the circular area; Calculate the actual distance from each charging station in the charging station candidate pool to the current user vehicle location, remove the charging stations with actual distance greater than the accessible distance from the charging station candidate pool, and obtain a recommended candidate pool; Calculate the charging cost of each charging station in the recommended candidate pool, and select several charging stations with the lowest charging cost in the recommended candidate pool according to the charging cost to form a charging station candidate list; Obtain the user's preference tags and preference tag weights, score and sort the charging stations in the charging station candidate list according to the preference tags, their types, and preference tag weights, and obtain the user's preference recommendation list. Specifically: Determine the type of preference tag, which includes: Boolean, numerical, text, and conditional. Score the charging station according to the type of preference tag to obtain the score of each type of preference tag for each charging station. Specifically: If the preference tag is of Boolean type, the preference tag is assigned a value, a result that is favorable to vehicle charging is assigned a value of 1, a result that is unfavorable to vehicle charging is assigned a value of 0, and a score is calculated according to the result assignment; If the preference tag is a numerical value, the ranking and scoring are performed according to the specific numerical value type of the preference tag, and the numerical value type includes a dynamic numerical value and a static numerical value; If the preference label is text type, the text of the preference label is graded to obtain a level value, and a weighted sum is performed according to the level value and the level value weight to obtain the numerical value of the text type preference label, and the text type preference label is sorted and scored according to the numerical value; Furthermore, the text of the preference label is graded to obtain a level value, and a value of the text type preference label is obtained by weighted summation according to the level value and the level value weight, and before sorting and scoring by the value of the text type preference label: Determine whether the text of the preference tag is real-time text, if so, obtain real-time text data, and classify according to the real-time text; Otherwise, extract keywords from the text in the data set and perform semantic classification to obtain text keywords and their classifications, perform graded word frequency statistics on the text keywords and their classifications, obtain text descriptions based on the evaluation averages, and perform grading based on the text descriptions; The data set is user evaluation information in the long-term accumulated map software and charging software operated by various charging companies; The calculation formula of the evaluation mean is: , Wherein, the negative evaluation multiplier is adjusted by a step function to amplify the impact of negative evaluation; If the preference tag is conditional, the content of the preference tag is converted into a numerical value for sorting and scoring according to the type of the conditional content; further, the preference tag weight and level value weight are obtained through the following steps: Obtain user feedback data on the preference recommendation list, extract a weighted feedback vector according to the type of feedback data, input the weighted feedback vector and the charging station preference label into a pre-trained preference label weight optimization model, and obtain the preference label weight; Obtain feedback data of the preference recommendation list containing text-type labels, extract label-level feedback vectors according to the type of feedback data, input the label-level feedback and charging station preference labels into a pre-trained level value weight optimization model, and obtain the level value weights; Furthermore, the weight feedback vector and the label level feedback vector are obtained by the following steps: Determine whether the selected charging station belongs to the first or second charging station in the user's preference recommendation list. If yes, mark the weight feedback vector and the label level feedback vector as positive feedback; otherwise, mark the weight feedback vector and the label level feedback vector as negative feedback; If the selected charging station is not in the user's preferred recommendation list, the weight feedback vector and the label level feedback vector are marked as severe negative feedback; Obtain the user's evaluation of the preference recommendation list. If the evaluation is satisfactory, mark the weighted feedback vector and the label-level feedback vector as positive feedback; if the evaluation is unsatisfactory, mark the weighted feedback vector and the label-level feedback vector as negative feedback; Calculate the total user preference score for each charging station based on the preference tag weight and the preference tag score, and sort the charging stations based on the total user preference score of each charging station to obtain a user preference recommendation list; Further, after obtaining the user preference recommendation list, the following steps are performed: Modify the preference labels of charging stations based on the correction information provided by users, specifically: Obtain the correction information of the question preference label fed back by the user, and determine whether the question preference label is a dynamic numerical type or a real-time text type. If so, no correction is made; otherwise, push the correction information of the question preference label fed back by the user to several users who have recently gone to the charging station to charge, and collect user replies; Determine whether the number of user replies meets the reply threshold. If yes, calculate the number of user replies that are adopted, otherwise continue to push the revised information of the question preference label; It is determined whether the number of user replies that adopt the modification exceeds half of the total number of user replies. If so, the question preference label is modified; otherwise, the question preference label is not modified.

2. A method for dynamically recommending charging stations based on user preferences according to claim 1, characterized in that: The calculation of the charging cost of each charging station in the recommended candidate pool includes: The distance cost Cd is calculated by the actual distance between the current user vehicle location and each charging station in the recommended candidate pool; The charging cost Ce is calculated by the real-time electricity price and service fee of the charging station, and the waiting time cost Ct is calculated by the idle situation of the charging station; The charging cost C is calculated based on the distance cost Cd, charging fee cost Ce, and waiting time cost Ct.

Images