Charging revenue improvement intelligent regulation and control method and system fusing user behavior and operation data

Abstract

The invention provides a charging revenue improvement intelligent regulation and control method and system fusing user behaviors and operation data, and the method comprises the steps: carrying out the load prediction modeling of the charging time period distribution of each user group through a long-short-term memory network, and obtaining the charging revenue improvement intelligent regulation and control information of each user group according to the user group label and the time sequence data; obtaining an expected load distribution curve of each charging pile in the next 24 hours, and judging the load difference degree of the peak period and the valley period; real-time charging pile state information and user charging request data are obtained, if the predicted load distribution shows that the supply and demand mismatching degree in a certain time period exceeds a preset threshold value, a dynamic resource configuration mechanism is started, and the number of users needing load transfer and a target time period are determined; and according to the flexible user characteristics in the user group tag, a personalized off-peak charging suggestion is generated by adopting a differential incentive strategy, and an optimal charging period recommendation scheme for each user is obtained by calculating electricity price discounts and charging point rewards in different periods.

Description

Technical Field

[0001] This invention belongs to the field of information processing technology, and in particular relates to an intelligent control method and system for improving charging revenue by integrating user behavior and operational data. Background Technology

[0002] The rapid development of the electric vehicle charging industry has placed higher demands on energy management and infrastructure utilization, making increased charging revenue a crucial driver for the industry's sustainable development. Optimizing charging pile utilization efficiency, balancing grid load, and improving user experience are core objectives in this field. Current charging management methods largely rely on fixed scheduling or single operational data analysis, making it difficult to dynamically adapt to the diversity and complexity of user behavior. For example, some methods only predict demand based on historical data, ignoring the dynamic changes in real-time user behavior, leading to uneven allocation of charging pile resources, congestion during peak hours and idle equipment during off-peak hours, making it difficult to achieve a dual improvement in revenue and efficiency. Furthermore, existing solutions lack effective user interaction mechanisms when coordinating user behavior and operational data, failing to fully utilize behavioral differences among users to optimize resource allocation.

[0003] The key challenge in this field lies in establishing a dynamic resource allocation mechanism by integrating user behavior and operational data. First, the heterogeneity of user charging behavior makes it difficult for the system to accurately match the charging preferences of different users. For example, some users prefer slow charging at night, while others need to fast charge during the day due to time constraints. If this behavioral difference cannot be effectively coordinated, it will lead to a shortage of charging stations during peak hours and wasted resources during off-peak hours. Second, behavioral heterogeneity further exacerbates the difficulty of building a collaborative mechanism among users. Without incentive mechanisms to guide users to adjust their charging habits, the system cannot achieve a dynamic balance of off-peak charging. For example, a charging station may operate at full capacity during weekday lunch hours due to concentrated charging by commuters, but experience low user usage at night, resulting in low resource utilization.

[0004] Therefore, designing a dynamic user mutual assistance mechanism based on the heterogeneity of user charging behavior to incentivize off-peak charging and optimize the allocation of charging pile resources has become a key issue in improving charging revenue and system efficiency. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide a method and system for intelligent regulation and control of charging revenue that integrates user behavior and operational data.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for intelligently controlling charging revenue by integrating user behavior and operational data includes:

[0008] Step S1: Identify user behavior patterns from historical charging data;

[0009] Step S2: Based on the charging time distribution of each user group according to the user behavior pattern, obtain the expected load distribution curve of each charging pile in the next 24 hours;

[0010] Step S3: Based on the expected load distribution curve, real-time charging pile status information, and user charging request data, determine the number of users and target time periods that need to be transferred to the load.

[0011] Step S4: By calculating the electricity price discount and charging points rewards for different time periods, the optimal charging time recommendation scheme for each user is obtained.

[0012] As a preferred option, in step S1, the K-means clustering algorithm is used to identify user behavior patterns in historical charging data. Based on the feature vectors of three dimensions—charging time preference, charging frequency, and charging duration—different types of user group labels are obtained. If a user's charging time is concentrated during weekdays, they are marked as an urgent user; if their charging time is distributed at night or on weekends, they are marked as a flexible user.

[0013] Preferably, in step S2, a load prediction model is performed on the charging time distribution of each user group using a long short-term memory network. Based on the user group labels and time series data, the expected load distribution curve of each charging pile in the next 24 hours is obtained, and the degree of load difference between peak and off-peak periods is determined.

[0014] Preferably, in step S3, real-time charging pile status information and user charging request data are obtained. If the predicted load distribution shows that the supply and demand mismatch exceeds a preset threshold during a certain period, a dynamic resource allocation mechanism is activated to determine the number of users and the target time period that need to be transferred.

[0015] Preferably, in step S4, based on the flexible user characteristics in the user group tags, a differentiated incentive strategy is adopted to generate personalized off-peak charging suggestions. By calculating the electricity price discount and charging points reward for different time periods, the optimal charging time recommendation scheme for each user is obtained.

[0016] This invention also provides an intelligent control system for improving charging revenue by integrating user behavior and operational data, comprising:

[0017] The user behavior pattern recognition module is used to identify user behavior patterns in historical charging data using the K-means clustering algorithm. Based on the feature vectors of three dimensions—charging time preference, charging frequency, and charging duration—it obtains labels for different types of user groups. If a user's charging time is concentrated during weekdays, they are marked as an urgent user; if their charging time is distributed at night or on weekends, they are marked as a flexible user.

[0018] The load forecasting modeling module is used to perform load forecasting modeling on the charging time distribution of each user group through a long short-term memory network. Based on the user group labels and time series data, it obtains the expected load distribution curve of each charging pile in the next 24 hours and judges the degree of load difference between peak and off-peak periods.

[0019] The dynamic resource configuration module is used to obtain real-time charging pile status information and user charging request data. If the predicted load distribution shows that the supply and demand mismatch exceeds the preset threshold during a certain period, the dynamic resource configuration mechanism is activated to determine the number of users and the target time period that need to be transferred.

[0020] The differentiated incentive strategy module is used to generate personalized off-peak charging suggestions based on the flexible user characteristics in the user group tags, and obtain the optimal charging time recommendation scheme for each user by calculating the electricity price discount and charging points reward for different time periods.

[0021] This invention addresses the supply-demand imbalance and low resource utilization efficiency in electric vehicle charging scenarios. It analyzes historical charging data using a K-means clustering algorithm, classifying users into urgent and flexible groups based on charging time preferences, frequency, and duration characteristics, thus accurately identifying user behavior patterns. By combining this with a Long Short-Term Memory (LSTM) network, it forecasts the charging time distribution for each group, generating a projected 24-hour charging pile load curve and determining the load difference between peak and off-peak periods. When the forecast indicates a supply-demand mismatch exceeding a threshold during a certain period, this invention activates a dynamic resource allocation mechanism to determine the number of users requiring load transfer and the target time period. For flexible users, it designs differentiated incentive strategies, calculating the optimal charging time recommendation scheme through electricity price discounts and points rewards. This invention achieves efficient allocation of charging resources, significantly reduces peak-valley load differences, improves grid stability and charging pile utilization, and balances user charging experience with economic benefits through the seamless integration of behavioral pattern recognition, accurate load forecasting, and personalized incentives. Attached Figure Description

[0022] Figure 1 The flowchart illustrates the intelligent control method for improving charging revenue by integrating user behavior and operational data, as described in this invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0024] Example 1:

[0025] like Figure 1As shown, this embodiment of the invention provides a method for intelligent control of charging revenue improvement that integrates user behavior and operational data, including:

[0026] S101. The K-means clustering algorithm is used to identify user behavior patterns in historical charging data. Based on the feature vectors of three dimensions—charging time preference, charging frequency, and charging duration—different types of user group labels are obtained. If the user's charging time is concentrated during weekdays, they are marked as an urgent user. If the charging time is distributed at night or on weekends, they are marked as a flexible user.

[0027] Historical charging data was acquired, and outliers and missing values ​​were removed using data cleaning techniques to obtain a standardized charging dataset. Charging time preference, charging frequency, and charging duration were extracted from the standardized charging dataset to construct feature vectors. K-means clustering was used to cluster the feature vectors to determine user group labels. Based on the clustering results, users whose charging time is concentrated during weekday daytime were labeled as urgent users; those whose charging time is distributed at night or on weekends were labeled as flexible users, resulting in user classification results. The feature vector distributions of urgent and flexible users were obtained from the user classification results to construct user behavior patterns. The relationship between charging time preference and charging station load was analyzed through user behavior patterns to obtain a charging station load prediction model. Using the charging station load prediction model, charging station resource allocation was optimized, and a resource scheduling scheme was determined.

[0028] In one possible implementation, the process of acquiring historical charging data can begin by retrieving user charging records from the charging station management system.

[0029] For example, data from the past year might be collected from 100 charging stations in a city, including fields such as user ID, charging start time, end time, and charging amount. Data cleaning is a crucial step, requiring the removal of outliers and missing values.

[0030] For example, records with charging times exceeding 24 hours may be due to equipment malfunction and should be removed; records lacking charging end times can be filled in by estimating the average charging time. After standardization, the dataset is stored in a unified format, such as converting timestamps to hourly units and unifying the unit of charging quantity to kilowatt-hours. This cleaning process ensures data consistency and provides a reliable foundation for subsequent analysis.

[0031] Specifically, charging time preferences, charging frequency, and charging duration can be extracted through statistical analysis.

[0032] For example, analysis shows that User A charges 10 times between 8:00 AM and 12:00 PM on weekdays, with an average charging time of 2 hours and a charging volume of 10 kWh, indicating a preference for charging during weekday daytime. User B charges 3 times a week, mainly on weekend nights, with an average charging time of 4 hours and a charging volume of 15 kWh. Based on these metrics, a feature vector is constructed, such as [user_id, percentage of charging during weekday daytime, percentage of charging during weekend nights, average charging time, weekly charging frequency], for example, [UserA, 0.8, 0.1, 2, 10].

[0033] In one embodiment, the K-means clustering algorithm can be used to cluster feature vectors. Assuming K=3, after clustering, one type of user is identified as a user with a high charging rate during weekdays, and is marked as an urgent user; the other type of user is identified as a user with a high charging rate at night or on weekends, and is marked as a flexible user.

[0034] For example, the feature vector for users with urgent needs might be [0.9, 0.05, 1.5, 12], reflecting their high-frequency, short-term charging requirements; while for flexible users it might be [0.2, 0.7, 4, 3], showing a preference for long-term, low-frequency charging. The clustering results clearly distinguish user groups, facilitating subsequent behavioral analysis.

[0035] For example, analyzing the feature vector distributions of users with urgent needs and those with flexible needs can reveal behavioral patterns. Users with urgent needs are mostly commuters, with charging time concentrated between 9:00 and 17:00 on weekdays, resulting in significant peak loads at charging stations; while users with flexible needs have more dispersed charging times, leading to smoother loads. Based on this, a charging station load prediction model can be constructed.

[0036] For example, a charging station may reach 80% load at 10:00 AM on weekdays, but only 20% at night. By using regression analysis, peak load can be predicted, and resource allocation can be optimized.

[0037] Specifically, resource scheduling schemes can be optimized through predictive models.

[0038] For example, for charging stations with a high concentration of users in urgent need of charging, the number of daytime charging piles can be increased, such as from 10 to 15; for users with flexible charging needs, the nighttime charging discount period can be extended to encourage off-peak charging. Such scheduling schemes reduce peak load pressure, improve charging station utilization, reduce user waiting time, and enhance service efficiency.

[0039] In one possible implementation, the benefit of the above method is to improve the operational efficiency of charging stations.

[0040] For example, after optimization, the peak load at a charging station dropped from 90% to 70%, and the average waiting time for users decreased from 20 minutes to 10 minutes. This technological improvement significantly enhances the user experience while reducing equipment wear and tear and extending the lifespan of the charging station.

[0041] S102. The load prediction model is performed on the charging time distribution of each user group through a long short-term memory network. Based on the user group labels and time series data, the expected load distribution curve of each charging pile in the next 24 hours is obtained, and the degree of load difference between peak and off-peak periods is judged.

[0042] By using user group tags and historical time-series data, a Long Short-Term Memory (LSTM) network model is constructed to generate load forecast results for each charging station. The load distribution curve for the next 24 hours is extracted from the load forecast results to determine the load change trend of each charging station. Based on the load change trend, peak and off-peak periods are identified in the load distribution curve. If the load value during a peak period exceeds a preset threshold, the difference between it and the load value during an off-peak period is calculated to obtain the degree of load difference. Based on the degree of load difference, the time-period load distribution characteristics of each charging station are generated to determine the operating status of the charging stations. The operating status of each charging station is obtained to generate a load allocation plan for the next 24 hours, optimizing charging station resource scheduling. Based on the optimized load allocation plan, the time-series data is adjusted, the input of the LSM network model is updated, and new load forecast results are generated.

[0043] It should be noted that by constructing a long short-term memory network model using user group tags and historical time series data, the time dependence of charging behavior can be effectively captured.

[0044] For example, based on the labels of users with urgent needs and those with flexible charging habits, and combined with charging data from the past 30 days, including the start time, end time, and charging amount of each charge, input features can be generated. Users with urgent needs may concentrate their charging between 8:00 AM and 12:00 PM on weekdays, with an average charging amount of 20 kWh; while flexible users may charge between 10:00 PM and 2:00 AM or on weekends, with an average charging amount of 15 kWh. The Long Short-Term Memory (LSTM) network model analyzes these time-series features to learn the periodic changes in user charging behavior and generate load forecasts for each charging station for the next 24 hours.

[0045] Specifically, the load forecast results can be extracted as hourly load distribution curves.

[0046] For example, a charging station has 10 charging piles. Prediction shows that charging pile A experiences a load of 80kW during weekdays (12:00-14:00) but only 10kW during nighttime (2:00-4:00). Based on this, the peak period can be identified as 12:00-14:00, and the off-peak period as 2:00-4:00. If the preset threshold is 60kW, and the peak period load exceeds the threshold, the difference between the peak and off-peak periods is calculated, yielding a load difference of 70kW. This difference reflects the resource scarcity of charging piles during peak hours.

[0047] In one embodiment, the time-period load distribution characteristics of each charging station are generated by the degree of load difference.

[0048] For example, if charging pile A experiences 70% peak load and only 10% off-peak load, its operating status is "high peak load." Conversely, if charging pile B's peak load is only 50kW, below the threshold, its operating status is "stable load." These characteristics help determine whether charging piles need priority resource allocation. Based on the operating status, a load allocation plan for the next 24 hours can be generated.

[0049] For example, peak-hour charging demand can be redirected to low-load charging stations, such as transferring the charging needs of some users in urgent need from charging station A to charging station B, thereby reducing pressure during peak hours.

[0050] For example, an optimized load allocation scheme might reduce the load on charging station A from 80kW to 50kW during the 12:00-14:00 period, and allocate it to charging stations B and C. The adjusted time-series data, such as the new charging times and amounts, are then re-input into the Long Short-Term Memory (LSTM) network model to update the prediction results.

[0051] For example, the new forecast shows that the peak load of charging pile A has dropped to 55kW, and the overall load distribution is more balanced. This approach continuously optimizes the model's prediction accuracy by dynamically adjusting the input data.

[0052] In one possible implementation, the optimization of the load distribution scheme can also be combined with user behavior patterns.

[0053] For example, flexible users can be incentivized to charge during off-peak hours, such as offering a discounted electricity price of 0.5 yuan / kWh at night, reducing peak-hour load. Users with urgent needs are prioritized for high-power charging stations to ensure their fast charging needs are met. This resource scheduling method based on user tags and load forecasting can effectively balance the operating efficiency of charging stations.

[0054] S103. Obtain real-time charging pile status information and user charging request data. If the predicted load distribution shows that the supply and demand mismatch exceeds the preset threshold during a certain period, then start the dynamic resource allocation mechanism to determine the number of users and the target time period that need to be transferred.

[0055] The system acquires real-time charging pile status data and user charging request information. Status and request parameters are extracted from charging pile sensors and user-submitted requests via a data acquisition system to obtain a real-time data set. Based on this data set, time series analysis is used to predict load distribution for each time period, resulting in a load distribution curve. If the load distribution curve shows that the supply-demand imbalance exceeds a preset threshold for a certain time period, a dynamic resource allocation mechanism is triggered to determine the time period requiring adjustment. Based on the required adjustment time period, load transfer demand is calculated, and a linear regression model is used to determine the number of users requiring transfer. Using the number of transferred users and charging pile status data, available charging pile resources are matched to determine the allocation scheme for the target time period. An optimized scheduling algorithm is used to adjust the execution time of user charging requests according to the allocation scheme for the target time period, resulting in an optimized charging plan. The optimized charging plan is then used to update the charging pile status data, generating a dynamic resource allocation result.

[0056] For example, when acquiring real-time charging pile status data and user charging request information, sensors deployed on the charging piles can collect status parameters such as current, voltage, and charging duration. Simultaneously, request parameters such as user location, desired charging time, and vehicle battery demand can be extracted from charging requests submitted via user mobile applications. Assuming a city has 100 charging piles, with sensors uploading data every minute, generating a real-time data set containing charging pile ID, current load, and operating status, and user request data including user ID, request time, and estimated charging duration, the data acquisition system integrates this information into a unified data set, providing a foundation for subsequent analysis.

[0057] Specifically, when using time series analysis to predict load distribution in different time periods, the load change trend over the past 24 hours can be extracted using a sliding window technique based on historical and real-time data.

[0058] For example, a charging station typically experiences high load during weekdays from 8:00 to 10:00, averaging 80% capacity, while the load drops below 20% during midnight to 4:00. By analyzing real-time data sets, the load distribution curve for the next 24 hours is predicted, identifying 8:00-10:00 as a potential peak period. If the curve shows that the supply-demand imbalance exceeds a preset threshold during a certain period, such as the load exceeding 90%, a dynamic resource allocation mechanism is triggered.

[0059] In one embodiment, when determining the time period requiring adjustment, the overload period of 8:00-9:00 can be identified by analyzing the load distribution curve. Combined with real-time data, the load transfer demand is calculated.

[0060] For example, if 10 charging stations are overloaded during a certain period, 20% of the load needs to be diverted, which is roughly equivalent to the charging needs of 10 users. Using a linear regression model based on historical user behavior data, the number of users that need to be diverted is predicted. The model analysis assumes that diverting 8 users can reduce the peak-hour load to a safe level.

[0061] For example, when matching available charging pile resources, idle charging piles during low-load periods can be selected based on charging pile status data. Assuming that 30 charging piles have a load below 20% between 0:00 and 4:00 AM, they can be prioritized for allocation to users who are relocating. The optimized scheduling algorithm adjusts the charging times for 8 users to late-night hours based on user location and charging pile distribution, generating an optimized charging plan.

[0062] For example, a user who originally planned to charge at 8:30 was rescheduled to 1:00 and matched with the nearest available charging station.

[0063] Specifically, when updating the charging pile status data, the system adjusts the load distribution of the charging piles in real time according to the optimized charging plan.

[0064] For example, after the adjustment, the load during 8:00-9:00 AM decreased from 90% to 70%, while the load during late-night hours increased from 20% to 40%, achieving a balance between supply and demand. The dynamic resource allocation results are communicated to users via the user application, notifying them of the new charging times and simultaneously updating the charging pile management system to ensure efficient resource utilization. This approach, through real-time data analysis and dynamic adjustments, effectively balances load distribution, improves charging pile operating efficiency, and optimizes the user experience.

[0065] S104. Based on the flexible user characteristics in the user group tags, a differentiated incentive strategy is adopted to generate personalized off-peak charging suggestions. By calculating the electricity price discount and charging points reward for different time periods, the optimal charging time recommendation scheme for each user is obtained.

[0066] Obtain the labeled feature data of flexible users, extract the user's charging behavior, electricity consumption preferences, and historical charging records from the pre-established user database, and determine the user labeled feature set. According to the user labeled feature set, use the K-means clustering algorithm to group users, generate the initial grouping of the differentiated incentive strategy, and obtain the user grouping result. For the user grouping result, obtain the electricity price discount data and charging integral rules for each period, extract the real-time electricity price and integral policy from the power grid data interface, and determine the incentive parameters for each period. Through the weighted average algorithm, combine the electricity price discount and charging integral, calculate the comprehensive incentive value for each period, and obtain the incentive score for each period. If the incentive score for a certain period is higher than the preset threshold, mark that period as a candidate charging period and generate a list of candidate charging periods. According to the list of candidate charging periods, use the greedy algorithm, combine the electricity consumption preferences and historical charging records in the user labeled features, match the optimal charging period, and determine the personalized peak-shaving charging recommendation. By associating the personalized peak-shaving charging recommendation with the user labeled features, generate a recommendation plan and store it in the user database for subsequent push.

[0067] It should be noted that obtaining the labeled feature data of <result flexible> users, such as data, is based on the users established for realizing personalized charging pre-recommendations. The labeled feature data of flexible users is usually extracted from the user database. The labels of users generally include charging feature data, such as behavior, electricity consumption preferences, and historical charging records.

[0068] For example, the charging behavior, electricity consumption preference, and historical charging record of a certain user are as follows: The user has the habit of charging at night from 22:00 to 6:00 the next day. The charging behavior may include charging frequency, the electricity demand per charge, and the commonly used low electricity price period. The historical charging record shows that the user's monthly average charging is about 3 times, and each charge is about 30 kWh. These data are extracted from the user database, forming the preference for low electricity price periods, the preference for fast charging, charging frequency, and electricity demand. The historical charging record can provide the charging time, location, and electricity consumption of the user in the past three months. This method ensures the user portrait.

[0069] For example, comprehensive user data provides accurate information for subsequent user grouping. For instance, records show that using the K-means clustering algorithm, users are grouped twice a week for charging. When grouping, an average of 30 kWh is charged each time. Based on the charging time and demand, users are divided into nighttime (00:00-02:00), daytime (00:00-02:00), and random pricing charging groups. Assuming a time period, user A's data is often extracted for charging during off-peak hours, providing accurate basic information for monthly user groups. In a scenario with a possible 50 kWh, categorized as nighttime, the K-means clustering algorithm is used for nighttime charging; user B is grouped for daytime charging, with a charging frequency of 20 kWh based on demand, categorized into daily and off-peak charging groups. After grouping by time, strategies can be developed for nighttime charging characteristics such as high off-peak pricing, while daytime and off-peak charging groups can be categorized by charging points and flexible reward charging strategies. This classification. The group approach assumes that the user database contains incentive strategies for 1000 users, making them more user-friendly. Through targeting, analysis is improved, and user participation is increased. Specifically, in terms of flexibility, charging group electricity price discount data is obtained because its charging time and charging point rules are flexible and require preference through the real-time time period of the grid's low electricity price data interface. Point extraction. For example, after grouping, the system designs a discount of 0.3 points per kWh for the electricity price group of high-frequency charging from 22:00 to 6:00 on a certain day, with a reward strategy of 0.3 points per kWh. For charging, the flexible point rules provide 1 point for each kWh of dynamic electricity accumulation for the charging group. Price assumption discount at a certain time. This segmented electricity price group discount ensures the targeting of the incentive strategy at 0.5 points per kWh. Specifically, in terms of point rules, the electricity price discount data for each time period is obtained at 2 points per kWh. When calculating the comprehensive value through a weighted average algorithm and charging point rules, it can be obtained from the incentive grid data value. If the interface extracts real-time information with a price weight of 0.

[0070] For example, in a certain city, the integral weight of the power grid is 0.4 for nighttime 22, the comprehensive incentive is 00-06, and the electricity value = price discount is 0. 50.4 × yuan / kilowatt-hour 0.6 + 2 watt-hours, and daytime × 0.4 = 1 day. 1. If the preset threshold value of yuan / kilowatt-hour is 1, then the integral rule is that the time period is marked as a segment, and each charging time is a candidate 10 kilowatt charging time segment. This 1 integral method ensures that the system incentive score is calculated based on these data to determine the actual attractiveness. In each time period, the comprehensive incentive implementation value is based on the candidate.

[0071] For example, in the list of nighttime charging times, the segment uses a greedy low-algorithm to match the electricity price and the highest integral, thus optimizing the charging time and comprehensively incentivizing the segment.

[0072] For example, the bias value of user A reaches 0.85 at night for charging. According to historical high records, it is usually preset to charge at 0.7:00. Suppose the candidate time period is marked as the candidate time period from 22:00 to 0:00. This way of marking 22:00 - 0:00 and 2:00 - 4:00 helps the greedy algorithm to preferentially select the time when the user chooses to charge with a more economical and incentive value and conforms to the time. In one embodiment of user preference, when using the greedy algorithm to match the best charging time period as 22:00 - 0:00, the system will give priority to this time period. This matching method takes into account both user preference and the incentive effect score of the candidate time period.

[0073] For example, a small increase in the recommended acceptance rate of Zhang's preference for night charging.

[0074] For example, after the system selects a recommendation from the list of candidate personalized off-peak charging time periods generated, it associates it with the user's label features of 22:00 - 23:00, and stores it in the database because of its highest incentive and score matching. Suppose the recommendation for user A's habit is the same as Zhang's. The system generates a recommendation for charging at 23:00, records the time period of his preference for charging, and expects historical data, which can save 20% of the charging text message cost for subsequent push. This kind of personalized notification for charging or App. This storage and recommendation method improves the user experience in tracking and user response to the power grid, and optimizes the accuracy of future recommendations. For example, when storing the recommendation plan to the user database, it can be through hierarchical design from the user portrait to the off-peak charging recommendation and incentive label score features, and then record the time period matching, forming a strict logic of preference for night charging and low electricity price system tendency. Each recommendation link includes specific support for each other, ensuring that the charging time and expected savings cost both meet the user's needs and can effectively guide the off-peak charging to be pushed to the user through the App, improving the utilization efficiency of grid resources. This way ensures the sustainability and traceability of the recommendation result, providing data support for subsequent optimization.

[0075] Embodiment 2:

[0076] The embodiment of the present invention also provides a smart charging revenue improvement control system that integrates user behavior and operation data, including:

[0077] A user behavior pattern recognition module, which is used to recognize user behavior patterns from historical charging data by using the K-means clustering algorithm, and obtain different types of user group labels according to the feature vectors of three dimensions: charging time preference, charging frequency, and charging duration. If the user's charging time is concentrated during the day on weekdays, it is marked as an urgent user. If the charging time is distributed at night or on weekends, it is marked as a flexible user;

[0078] The load forecasting modeling module is used to perform load forecasting modeling on the charging time distribution of each user group through a long short-term memory network. Based on the user group labels and time series data, it obtains the expected load distribution curve of each charging pile in the next 24 hours and judges the degree of load difference between peak and off-peak periods.

[0079] The dynamic resource configuration module is used to obtain real-time charging pile status information and user charging request data. If the predicted load distribution shows that the supply and demand mismatch exceeds the preset threshold during a certain period, the dynamic resource configuration mechanism is activated to determine the number of users and the target time period that need to be transferred.

[0080] The differentiated incentive strategy module is used to generate personalized off-peak charging suggestions based on the flexible user characteristics in the user group tags, and obtain the optimal charging time recommendation scheme for each user by calculating the electricity price discount and charging points reward for different time periods.

[0081] The above description is merely a specific implementation of this specification. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the scope of protection of this specification is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this specification, and these modifications or substitutions should all be covered within the scope of protection of this specification.

Claims

1. A method for intelligent control of charging revenue improvement that integrates user behavior and operational data, characterized in that, include: Step S1: Identify user behavior patterns from historical charging data; Step S2: Based on the charging time distribution of each user group according to the user behavior pattern, obtain the expected load distribution curve of each charging pile in the next 24 hours; Step S3: Based on the expected load distribution curve, real-time charging pile status information, and user charging request data, determine the number of users and target time periods that need to be transferred to the load. Step S4: By calculating the electricity price discount and charging points rewards for different time periods, the optimal charging time recommendation scheme for each user is obtained.

2. The intelligent control method for improving charging revenue by integrating user behavior and operational data as described in claim 1, characterized in that, In step S1, the K-means clustering algorithm is used to identify user behavior patterns in historical charging data. Based on the feature vectors of three dimensions—charging time preference, charging frequency, and charging duration—different types of user group labels are obtained. If a user's charging time is concentrated during weekdays, they are marked as an urgent user; if their charging time is distributed at night or on weekends, they are marked as a flexible user.

3. The intelligent control method for improving charging revenue by integrating user behavior and operational data as described in claim 1, characterized in that, In step S2, a load prediction model is performed on the charging time distribution of each user group using a long short-term memory network. Based on the user group labels and time series data, the expected load distribution curve of each charging pile in the next 24 hours is obtained, and the degree of load difference between peak and off-peak periods is determined.

4. The intelligent control method for improving charging revenue by integrating user behavior and operational data as described in claim 1, characterized in that, In step S3, real-time charging pile status information and user charging request data are obtained. If the predicted load distribution shows that the supply and demand mismatch exceeds a preset threshold during a certain period, the dynamic resource allocation mechanism is activated to determine the number of users and the target time period that need to be transferred.

5. The intelligent control method for improving charging revenue by integrating user behavior and operational data as described in claim 1, characterized in that, In step S4, based on the flexible user characteristics in the user group tags, a differentiated incentive strategy is adopted to generate personalized off-peak charging suggestions. By calculating the electricity price discount and charging points reward for different time periods, the optimal charging time recommendation scheme for each user is obtained.

6. A smart control system for improving charging revenue by integrating user behavior and operational data, characterized in that, include: The user behavior pattern recognition module is used to identify user behavior patterns in historical charging data using the K-means clustering algorithm. Based on the feature vectors of three dimensions—charging time preference, charging frequency, and charging duration—it obtains labels for different types of user groups. If a user's charging time is concentrated during weekdays, they are marked as an urgent user; if their charging time is distributed at night or on weekends, they are marked as a flexible user. The load forecasting modeling module is used to perform load forecasting modeling on the charging time distribution of each user group through a long short-term memory network. Based on the user group labels and time series data, it obtains the expected load distribution curve of each charging pile in the next 24 hours and judges the degree of load difference between peak and off-peak periods. The dynamic resource configuration module is used to obtain real-time charging pile status information and user charging request data. If the predicted load distribution shows that the supply and demand mismatch exceeds the preset threshold during a certain period, the dynamic resource configuration mechanism is activated to determine the number of users and the target time period that need to be transferred. The differentiated incentive strategy module is used to generate personalized off-peak charging suggestions based on the flexible user characteristics in the user group tags, and obtain the optimal charging time recommendation scheme for each user by calculating the electricity price discount and charging points reward for different time periods.

Images