A virtual power plant front-end layout dynamic optimization method and system based on user behavior driving

Abstract

This application relates to the field of front-end page layout technology, specifically to a method and system for dynamic optimization of the front-end layout of a virtual power plant based on user behavior. The method includes: generating dynamic behavior tags through cluster analysis based on interaction behavior data and combined with the user's predefined static business role information, and establishing a user interface preference model to form a unified multi-level user profile that simultaneously includes static roles and dynamic behaviors; generating and outputting a personalized front-end layout through a dynamic layout engine by: calculating the display priority of each menu item and functional component; automatically matching the interface layout template most suitable for the current user profile from a preset layout template library; and performing differentiated display and hiding control and arrangement optimization of functional components within the interface according to the display priority. This solves the problems of rigid interfaces, low operational efficiency, and lack of intelligent perception and evolution capabilities in existing systems.

Description

Technical Field

[0001] This application relates to the field of front-end page layout technology, specifically to a method and system for dynamic optimization of the front-end layout of a virtual power plant based on user behavior. Background Technology

[0002] Virtual power plant management systems are core platforms for coordinating distributed energy resources and grid operations, with users encompassing various professional roles such as dispatchers, traders, and energy owners. The job responsibilities and core task flows of these different roles differ fundamentally; however, existing systems generally employ static, uniform layouts in their front-end interfaces, failing to adapt to these diverse professional needs, resulting in low operational efficiency and high learning costs.

[0003] Specifically, the interface is rigid, with frequently used functions often buried in multi-level menus, requiring users to click multiple times to access them, resulting in lengthy navigation paths. It cannot self-optimize based on actual user behavior, lacks personalized dynamic adjustment capabilities, and the user experience stagnates. Existing technologies include solutions that adjust webpage templates by analyzing user browsing behavior. However, when applied to complex virtual power plant systems, this approach reveals fundamental flaws: firstly, its optimization relies solely on generalized browsing behaviors such as clicking and scrolling, completely ignoring the context of "business roles," leading to a lack of precision in the initial layout; secondly, its optimization granularity remains at the page-level, failing to perceive and understand the cross-functional, coherent task flow of "monitoring-analysis-scheduling" within the virtual power plant, and thus unable to achieve deep, component-level intelligent aggregation.

[0004] Therefore, there is an urgent need in this field for a dynamic layout solution that can deeply integrate business roles and real-time behavioral data, and has the ability to continuously learn and optimize, in order to improve the professionalism and intelligent interactive experience of virtual power plant management systems. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a method and system for dynamic optimization of the front-end layout of a virtual power plant based on user behavior.

[0006] In a first aspect, the technical solution of the present invention provides a method for dynamic optimization of the front-end layout of a virtual power plant based on user behavior, comprising the following steps:

[0007] S1. Collect user interaction behavior data in the virtual power plant management system; the interaction behavior data includes at least menu item click events, function module access and exit events, and interface jump events, wherein each data record has a corresponding user identifier and timestamp;

[0008] S2. Based on the collected interaction behavior data and combined with the user's predefined static business role information, dynamic behavior tags are generated through cluster analysis, and a user interface preference model is established to form a unified multi-level user profile that includes both static roles and dynamic behaviors.

[0009] S3. Based on multi-layered user profiles, perform the following operations using a dynamic layout engine to generate and output personalized front-end layouts:

[0010] (a) Calculate the display priority of each menu item and functional component;

[0011] (b) Automatically match the interface layout template that best suits the current user profile from the preset layout template library;

[0012] (c) Based on the display priority, perform differentiated display control and arrangement optimization of functional components within the interface;

[0013] S4. Collect subsequent interaction data of the user with the personalized front-end layout and feed it back to S1 to update the user profile and optimize the layout decision.

[0014] As a further limitation of the technical solution of the present invention, step S1 includes:

[0015] S11. By embedding data acquisition probes into the front-end interface components of the virtual power plant management system, user clicks, browsing duration, and page jump events are captured in real time.

[0016] S12. Convert the captured events into structured behavior logs, which include at least user ID, timestamp, operation object, and operation type;

[0017] S13. Based on the behavior log, sessions are divided according to user ID and timestamp, and a business rule-based cleaning algorithm is used to filter out meaningless erroneous operation events and heartbeat keep-alive events to construct a user operation sequence that represents a complete operation task.

[0018] This ensures that the collected interaction data is high-quality, structured, and rich in semantic information. This lays a reliable data foundation for the subsequent construction of accurate user profiles, avoids interference from noisy data on the analysis results, and improves the accuracy and reliability of the entire system optimization.

[0019] As a further limitation of the technical solution of this invention, in S2, the predefined static business role information includes at least dispatchers, traders, and energy owners; the system pre-sets differentiated initial interface layout templates and functional permission sets for each role. This clarifies that core business roles such as dispatchers, traders, and energy owners are the starting point for system design, and differentiated initial templates and permissions are pre-set for different roles. This ensures that the system can provide a professional and accurate basic working interface from the beginning, achieving coarse-grained personalization based on roles and solving the problem that the interface cannot meet the needs of professional division of labor.

[0020] As a further limitation of the technical solution of the present invention, step S2 includes:

[0021] S21. Extract user behavior feature vectors within a set statistical period from the interactive behavior data. The vectors include at least menu item click frequency, function usage duration, and operation sequence pattern.

[0022] S22. Calculate the behavioral feature vector using an unsupervised clustering algorithm to divide users into different behavioral clusters; and assign a dynamic behavioral label to each behavioral cluster, the label including at least high-frequency scheduling type, data browsing type, and transaction-driven type;

[0023] S23. For each type of dynamic behavior label, assign a set of interface component preference weight vectors, wherein the weight vectors quantify the degree of preference of the user for different functional modules, data panels and operation controls.

[0024] S24. Associate and map the preference weight vector with the corresponding dynamic behavior labels to construct a preliminary user interface preference model;

[0025] S25. Using the predefined static business role information as the base layer, integrate the dynamic behavior tags and the corresponding preliminary user interface preference model, and encapsulate them together into a unified user profile data entity that can be directly called by the dynamic layout engine, thus forming a unified multi-level user profile.

[0026] S26. Analyze the timestamp information in the interaction behavior data to identify the user's function usage patterns on weekdays and holidays, and during peak and off-peak electricity consumption periods; use the identified time patterns as a correction factor to optimize the classification results of dynamic behavior tags in S22, and / or to optimize the weight allocation of the interface component preference weight vector in S23.

[0027] It can sense and adapt to patterns in user behavior that change with business rhythms (such as peak and off-peak electricity consumption, weekdays / holidays). This makes user profiles and layout recommendations no longer static, but dynamically adaptable to real business scenarios, improving the accuracy of layout recommendations in complex business environments.

[0028] As a further limitation of the technical solution of the present invention, in step S3(a), the display priority of each menu item and functional component is calculated, which is specifically achieved by the following formula:

[0029]

[0030] in, To display priority scores, Use frequency weights for functions Weighting based on feature usage time Weighting of business roles To normalize the click-through rate, To normalize usage time, The role-function relevance is dynamically adjusted based on the dynamic behavior tags in the user profile, including the weights for function usage frequency, function usage duration, and business role. The role-function relevance is obtained by querying a predefined role-function weight mapping table, which is established based on the business rules of the virtual power plant and the responsibilities and permissions of different roles. The weight values ​​in the mapping table are periodically and dynamically adjusted based on the historical average behavior data of the user group under the same role.

[0031] By incorporating the weight of business roles and the correlation between role functions as core variables in the calculation, the layout decision-making process takes into account both the user's immediate behavioral preferences and the long-term rigid requirements determined by their job responsibilities, achieving an intelligent balance between dynamic needs and static responsibilities.

[0032] As a further limitation of the technical solution of the present invention, in step S3(b), automatically matching the most suitable interface layout template from the preset layout template library specifically includes:

[0033] S31. Based on the predefined static business role information in the user profile, select a set of candidate templates that meet the role criteria from the template library;

[0034] S32. Based on the interface component preference weight vector in the user profile, calculate the cosine similarity with the preset component layout of each template in the candidate template set.

[0035] S33. Select the template with the highest cosine similarity to the interface component preference weight vector as the most suitable interface layout template.

[0036] By quickly narrowing down the template pool through role-based matching, the professionalism of the basic framework is ensured. Then, by calculating cosine similarity, the template that best matches the user's personal preferences is selected from the candidate templates. This method balances matching accuracy and efficiency, avoiding the performance overhead of global matching in a large template library.

[0037] As a further limitation of the technical solution of the present invention, in step S3 (c), according to the display priority, the functional components within the interface are subject to differentiated display / hiding control and arrangement optimization, specifically including:

[0038] S34. In the main workspace of the most suitable interface layout template, prioritize rendering and displaying the top N functional components with the highest priority.

[0039] S35: Aggregate the functional components with a priority ranking of Nth or lower into the secondary menu;

[0040] S36. Within the same area, functional components are arranged in descending order according to their display priority scores.

[0041] The rules specify detailed interface optimization principles, allowing high-frequency functions to be rendered directly in the core area while low-frequency functions are aggregated into secondary menus. This reduces user navigation steps and time spent searching for functions, fundamentally improving operational efficiency.

[0042] As a further limitation of the technical solution of the present invention, after the execution of step S3, the system provides a multi-level configuration collaboration mechanism. This collaboration mechanism allows users to manually fine-tune the personalized layout automatically generated by the system, and the user's manual configuration has the highest priority and is recorded as feedback data in the user profile for optimizing subsequent automatic layout generation.

[0043] Above the automated layout, the highest priority for manual user intervention is introduced. This not only respects the user's personal choices and enhances the user's sense of control over the system, but more importantly, it transforms the user's manual adjustments into valuable feedback data for the system's learning, achieving collaborative and closed-loop optimization between human and machine intelligence.

[0044] As a further limitation of the technical solution of the present invention, step S4 includes:

[0045] S41. Collect the user's subsequent interaction behavior with the personalized front-end layout. The subsequent interaction behavior includes at least the click rate of rearranged functional components, the access depth of functions in the sub-menu, and the user's manual adjustment of the layout.

[0046] S42. Compare subsequent interactive behaviors with the user profile and layout decisions used to generate the personalized front-end layout, and evaluate the effectiveness of the layout decisions.

[0047] S43. Based on the evaluation results, dynamically adjust the preference weight vector of relevant functional components in the user interface preference model; and / or, modify the function usage frequency weight, function usage duration weight, and business role weight in the calculation formula of step S3(a).

[0048] The feedback loop works by evaluating the effectiveness of layout decisions and then adjusting the user preference model and layout generation algorithm accordingly. This ensures the system can track changes in user preferences, continuously iterate and optimize, and provide a consistent and accurate personalized experience.

[0049] Secondly, the present invention also provides a user behavior-driven dynamic optimization system for the front-end layout of a virtual power plant, used to implement the method described in the first aspect, comprising:

[0050] The behavior data acquisition module is used to collect user interaction behavior data in the virtual power plant management system. The interaction behavior data includes at least menu item click events, function module access and exit events, and interface jump events, and each data record has a corresponding user identifier and timestamp.

[0051] The user profile building module is used to generate dynamic behavior tags through cluster analysis based on the collected interaction behavior data and combined with the user's predefined static business role information, and to establish a user interface preference model to form a unified multi-level user profile that includes both static roles and dynamic behaviors.

[0052] The dynamic layout engine module is used to calculate the display priority of each menu item and functional component; automatically match the interface layout template that best suits the current user profile from the preset layout template library; and perform differentiated display and hiding control and arrangement optimization of functional components in the interface according to the display priority to generate and output a personalized front-end layout.

[0053] The closed-loop feedback module is used to collect subsequent interaction data of the user with the personalized front-end layout and feed it back to the behavior data collection module for updating the user profile and optimizing layout decisions.

[0054] As a further limitation of the technical solution of the present invention, the behavioral data acquisition module includes:

[0055] The front-end probe unit is used to capture user clicks, browsing duration, and page jump events in real time by embedding data acquisition probes into the front-end interface components of the virtual power plant management system.

[0056] The log structuring unit is used to transform captured events into structured behavior logs, which at least include user ID, timestamp, operation object, and operation type.

[0057] The session construction and cleaning unit is used to divide sessions according to user ID and timestamp based on the behavior log, and to filter out meaningless erroneous operation events and heartbeat keep-alive events using a business rule-based cleaning algorithm, so as to construct a user operation sequence that represents a complete operation task.

[0058] As a further limitation of the technical solution of the present invention, the user profile construction module includes:

[0059] The role management unit is used to store and manage the predefined static business role information and its corresponding initial interface layout template and function permission set;

[0060] The behavior analysis unit is used to extract user behavior feature vectors from the interaction behavior data within a set statistical period. These vectors include at least menu item click frequency, function usage duration, and operation sequence patterns. An unsupervised clustering algorithm is used to calculate the behavior feature vectors, dividing users into different behavior clusters. A dynamic behavior label is assigned to each behavior cluster, including at least high-frequency scheduling, data browsing, and transaction-driven labels. The unit analyzes the timestamp information in the interaction behavior data to identify user function usage patterns during weekdays and holidays, and peak and off-peak electricity consumption periods. The identified time patterns are used as a correction factor to optimize the classification results of the dynamic behavior labels and / or to optimize the weight allocation of the interface component preference weight vector.

[0061] The preference model construction unit is used to assign a set of interface component preference weight vectors to each type of dynamic behavior label. The weight vectors quantify the degree of preference of users of this type for different functional modules, data panels and operation controls. The preference weight vectors are associated and mapped with the corresponding dynamic behavior labels to construct a preliminary user interface preference model.

[0062] The profile encapsulation unit is used to integrate the dynamic behavior tags and the corresponding preliminary user interface preference model with the predefined static business role information as the base layer, and encapsulate them together into a unified user profile data entity that can be directly called by the dynamic layout engine, thus forming a unified multi-level user profile.

[0063] As a further limitation of the technical solution of the present invention, the dynamic layout engine module includes:

[0064] The priority calculation unit is used to calculate the display priority of each menu item and functional component based on the multi-level user profile and a preset algorithm.

[0065] The template matching unit is used to automatically match the most suitable interface layout template from the preset layout template library based on the multi-level user profile.

[0066] The interface optimization unit is used to perform differentiated display and hiding control and arrangement optimization of functional components in the interface according to the display priority.

[0067] A multi-level configuration collaboration unit is used to execute a multi-level configuration collaboration mechanism and process user manual fine-tuning commands. The multi-level configuration collaboration mechanism allows users to manually fine-tune the personalized layout automatically generated by the system, and the user's manual configuration has the highest priority and is recorded as feedback data in the user profile for optimizing subsequent automatic layout generation.

[0068] As a further limitation of the technical solution of the present invention, the closed-loop feedback module includes:

[0069] The feedback behavior collection unit is used to collect users' subsequent interaction behaviors with the personalized front-end layout;

[0070] The decision evaluation unit is used to compare subsequent interactive behaviors with the user profile and layout decisions used to generate the personalized front-end layout, and to evaluate the effectiveness of the layout decisions.

[0071] The model and algorithm update unit is used to dynamically adjust the preference weight vector of relevant functional components in the user interface preference model based on the evaluation results; and / or, correct the function usage frequency weight, function usage duration weight, and business role weight in the calculation formula.

[0072] As can be seen from the above technical solutions, this application has the following advantages: it constructs a complete technical closed loop from data collection to profile building, layout generation, and feedback optimization. It achieves deep integration of static business roles and dynamic behavioral data in the virtual power plant front-end layout and introduces a continuous self-optimization mechanism. It solves the problems of rigid interfaces, low operational efficiency, and lack of intelligent perception and evolutionary capabilities in existing systems, realizing a truly personalized and adaptive interface, and improving the efficiency and satisfaction of users with different professional roles. Attached Figure Description

[0073] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0074] Figure 1 This is a flowchart illustrating the method provided in an embodiment of the present invention.

[0075] Figure 2 A block diagram of a system provided in an embodiment of the present invention. Detailed Implementation

[0076] To make the purpose, features, and advantages of this application more apparent and understandable, specific embodiments and accompanying drawings will be used to clearly and completely describe the technical solution protected by this application. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0077] Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this application and in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0078] like Figure 1 As shown, this embodiment of the invention provides a method for dynamic optimization of the front-end layout of a virtual power plant based on user behavior, including the following steps:

[0079] S1. Collect user interaction data in the virtual power plant management system; the interaction data includes at least menu item click events, function module access and exit events, and interface jump events, wherein each data item records a corresponding user identifier and timestamp; this step specifically includes:

[0080] S11. By embedding data acquisition probes into the front-end interface components of the virtual power plant management system, user clicks, browsing duration, and page jump events are captured in real time.

[0081] In one specific embodiment of the present invention, the data acquisition probe is integrated into the front-end visualization framework of the virtual power plant management system in the form of a JavaScript code package. This probe listens for user interaction events with all interface components, for example:

[0082] Click events: Listen for the onClick event of interactive elements such as menu items, buttons, and charts.

[0083] Browsing duration: The effective dwell time of a user within a single functional page or module is calculated by listening to the page's onLoad and onBeforeUnload events, or by listening to the visibilitychange API.

[0084] Page navigation: Listen for changes in the front-end route (such as the afterEach hook in Vue Router) to capture all interface switching behaviors.

[0085] The captured raw event data is immediately encapsulated into a data packet containing the event type, trigger element ID, current timestamp, and user session ID.

[0086] S12. Convert the captured events into structured behavior logs, which include at least user ID, timestamp, operation object, and operation type;

[0087] The front-end probe captures data packets and sends them to the log collection server in real time via HTTP requests or WebSocket connections. A server-side log processing service (e.g., based on Logstash or a custom Node.js service) receives this data and standardizes it, transforming it into structured behavioral logs in a uniform JSON format.

[0088] S13. Based on the behavior log, sessions are divided according to user ID and timestamp, and a business rule-based cleaning algorithm is used to filter out meaningless erroneous operation events and heartbeat keep-alive events to construct a user operation sequence that represents a complete operation task.

[0089] This step is performed by a dedicated data preprocessing service. This service first groups the data by userID, and then sorts the data for each user by timestamp.

[0090] Session segmentation: A classic method based on timeout thresholds is used. If the time difference between two consecutive log entries exceeds 30 minutes, the user is considered to have started a new session. Each session represents a complete login usage cycle for the user.

[0091] Data cleaning: Logs within a session are filtered using a cleaning algorithm based on business rules. Rules include:

[0092] Filtering out consecutive click events with extremely short durations (such as less than 300 milliseconds) is generally considered a mistake.

[0093] Identify and filter out heartbeat requests that are automatically initiated by the system to maintain the connection.

[0094] Filter out user behavior that involves entering and exiting the same page within a very short period of time (e.g., within 3 seconds), as this is usually considered navigation exploration rather than a valid task.

[0095] After cleaning, the meaningful sequence of events retained within a session is constructed as a user action sequence. This sequence clearly represents the series of actions performed by the user to complete a specific task, providing high-quality data input for subsequent behavior sequence analysis.

[0096] S2. Based on the collected interaction behavior data and combined with the user's predefined static business role information, dynamic behavior tags are generated through cluster analysis, and a user interface preference model is established to form a unified multi-level user profile that includes both static roles and dynamic behaviors; the predefined static business role information includes at least dispatchers, traders and energy owners; the system has a differentiated initial interface layout template and function permission set for each role.

[0097] It should be further noted that the steps in S2 include:

[0098] S21. Extract user behavior feature vectors within a set statistical period from the interactive behavior data. The vectors include at least menu item click frequency, function usage duration, and operation sequence pattern.

[0099] In this embodiment, the system sets a statistical period. The data processing service aggregates user interaction behavior data within this period and calculates the following features to construct a behavior feature vector:

[0100] Menu item click frequency: Count the total number of times users click on each core menu item, and normalize the data to obtain a frequency vector that reflects the breadth and concentration of function usage.

[0101] Function usage time: Calculate the cumulative time users spend in each major functional module, and normalize it to form a time preference vector.

[0102] Operation sequence pattern: Using N-gram model or sequence embedding technology, the operation path frequently performed by users (such as "alarm list -> alarm details -> historical data comparison") is transformed into a fixed-length vector representation to capture their workflow habits.

[0103] Finally, the above vectors are concatenated to form a fixed-length behavioral feature vector for each user, which is then used as input for the clustering algorithm.

[0104] S22. Calculate the behavioral feature vector using an unsupervised clustering algorithm to divide users into different behavioral clusters; and assign a dynamic behavioral label to each behavioral cluster, the label including at least high-frequency scheduling type, data browsing type, and transaction-driven type;

[0105] This step is performed by an analytics engine running on the server. This engine uses an unsupervised K-means clustering algorithm to cluster the behavioral feature vectors of all users. The optimal number of clusters is determined using the elbow rule or silhouette coefficient (K=3 in this embodiment). After clustering, combined with business expert knowledge, each behavioral cluster is assigned a dynamic behavioral label with business meaning.

[0106] High-frequency scheduling type: This cluster of users is characterized by frequent clicks on functions such as real-time monitoring, alarm handling, and dispatch command issuance, with working hours far exceeding the average.

[0107] Data browsing type: Users in this cluster frequently access modules such as historical data reports, power generation forecast analysis, and market information, but rarely perform active operations.

[0108] Trading-driven: This group of users is characterized by their activity during electricity trading hours and their frequent use of functions such as quantity and price quotation, transaction result inquiry, and position management.

[0109] S23. For each type of dynamic behavior label, assign a set of interface component preference weight vectors, wherein the weight vectors quantify the degree of preference of the user for different functional modules, data panels and operation controls.

[0110] The system maintains a global component-tag preference mapping library. Based on the dynamic behavior tags obtained in S22, a set of preset interface component preference weight vectors is assigned to each tag. This weight vector is an array, where each element corresponds to an interface function component (such as "real-time power curve graph", "alarm list card", "trading order panel"), and the element value (weight) quantifies the degree of preference of that type of user for that component.

[0111] For example, in the weight vector assigned to high-frequency scheduling labels, the weight values ​​of the real-time alarm list and the whole network power flow graph are preset to 0.95 (very high), while the weight of the monthly electricity bill report is preset to 0.1 (very low).

[0112] S24. Associate and map the preference weight vector with the corresponding dynamic behavior labels to construct a preliminary user interface preference model;

[0113] The correspondence between the multiple dynamic behavior labels and preference weight vectors generated in S23 is organized into a key-value pair mapping table in memory or a database, or encapsulated into an independent model object. The core function of this model object is: input a dynamic behavior label, and it will output the corresponding complete interface component preference weight vector. This model is the preliminary user interface preference model.

[0114] S25. Using the predefined static business role information as the base layer, integrate the dynamic behavior tags and the corresponding preliminary user interface preference model, and encapsulate them together into a unified user profile data entity that can be directly called by the dynamic layout engine, thus forming a unified multi-level user profile.

[0115] The system creates a standardized user profile data entity (e.g., a JSON object or a database record). This entity has a clearly defined three-level structure:

[0116] Baseline layer: Directly records the user's predefined static business role information (such as "role: dispatcher") and associates it with its initial interface layout template ID.

[0117] Classification layer: Records dynamic behavior labels obtained through S22 analysis (e.g., "Label: High-frequency scheduling type").

[0118] Preference layer: By querying the preliminary preference model constructed by S24, load the complete interface component preference weight vector corresponding to the dynamic behavior label.

[0119] This data entity, which contains a three-layer structure, is a unified multi-layered user profile, which is a complete digital description of a user's identity.

[0120] S26. Analyze the timestamp information in the interaction behavior data to identify the user's function usage patterns on weekdays and holidays, and during peak and off-peak electricity consumption periods; use the identified time patterns as a correction factor to optimize the classification results of dynamic behavior tags in S22, and / or to optimize the weight allocation of the interface component preference weight vector in S23.

[0121] The system adds a time pattern analyzer, whose execution cycle is synchronized with S21. This analyzer:

[0122] Analyze the timestamps in users' historical behavior data to identify patterns. For example, it was found that a user frequently accessed the load forecasting function during 9:00-11:00 on weekdays (morning peak electricity consumption), and frequently used the transaction module during 14:00-16:00 (transaction hours).

[0123] If a user's behavior pattern during a specific time period is found to be significantly inconsistent with their current dynamic behavior label (e.g., a "data browsing" user exhibits strong trading behavior during a trading period), a "trading-driven" label component can be temporarily superimposed on them during that time period, or a re-clustering can be triggered.

[0124] Based on the baseline weight vector allocated in S23, the weights of specific functional components are dynamically increased according to the identified time patterns. For example, during the identified trading period, the weight of the order placement panel component for all users (regardless of their primary label) is temporarily increased by an offset (e.g., +0.2) to ensure that the interface layout can respond sensitively to changes in business rhythm.

[0125] S3. Based on multi-layered user profiles, perform the following operations using a dynamic layout engine to generate and output personalized front-end layouts:

[0126] (a) Calculate the display priority of each menu item and functional component;

[0127]

[0128] in, To display priority scores, Use frequency weights for functions Weighting based on feature usage time Weighting of business roles To normalize the click-through rate, To normalize usage time, The role-function relevance is dynamically adjusted based on the dynamic behavior tags in the user profile, including the weights for function usage frequency, function usage duration, and business role. The role-function relevance is obtained by querying a predefined role-function weight mapping table, which is established based on the business rules of the virtual power plant and the responsibilities and permissions of different roles. The weight values ​​in the mapping table are periodically and dynamically adjusted based on the historical average behavior data of the user group under the same role.

[0129] Normalized click-through rate (CTR) is used to eliminate the magnitude difference in the total number of clicks among different users, making the click frequencies of different users comparable.

[0130] Within a defined statistical period (e.g., one week), the normalized click-through rate (CTR) for a specific user is calculated using the following steps:

[0131] The total number of clicks made by the user on all menu items and functional components in the system is recorded as follows: Statistics on the user's usage of specific functional components The number of clicks is recorded as .

[0132] Calculate the user's preference for a specific functional component Original click-through rate: .

[0133] To perform cross-user comparisons, for all users Normalization is performed, mapping it to the interval [0, 1]. The min-max normalization method is used:

[0134]

[0135] For the statistical period, all users of this specific functional component raw click-through rate The maximum value.

[0136] For the statistical period, all users of this specific functional component raw click-through rate The minimum value.

[0137] This method allows users to access specific functional components. Normalized click-through rate This intuitively reflects the user's understanding of the function. Click preferences, and their relative position among all users.

[0138] Normalized usage time is designed to eliminate differences in total online time among different users and to fairly measure users' time investment preferences across different functions.

[0139] Similar to click-through rate, for a specific user and a specific functional component:

[0140] The total effective session duration of this user within the statistical period is recorded as follows: .

[0141] Statistics on this user's usage of specific functional components The cumulative duration of stay is recorded as follows: .

[0142] Calculate the original proportion of usage time: .

[0143] Using all users' energy components within the statistical period The maximum and minimum values ​​of the original proportion of usage time are used to obtain the normalized usage time through cross-user comparison using the min-max normalization method. .

[0144] Role-function association measures the importance of a functional module to a business role.

[0145] Initial predefined: The system maintains a role-function association matrix. This matrix is ​​set by domain experts during system initialization, assigning a basic association degree to each pair (role, function). The value range is [0, 1].

[0146] For example: (Role = Dispatcher, Function = Real-time Load Forecasting) -> = 0.9;

[0147] (Role = Dispatcher, Function = Electricity Trading Quote) -> = 0.2;

[0148] Dynamic correction (dynamic part): The system periodically (e.g., monthly) analyzes the historical behavior data of all users under the same role and fine-tunes the basic correlation.

[0149] Calculate the average heatmap value of a certain function for all users under this role. The overall heat index of individual users is calculated by weighting normalized click-through rate and duration.

[0150] Correction formula:

[0151] in It is a mixing factor (0 < <1), used to control the weighting of business rules and group behavior data. For example, =0.7 indicates greater reliance on expert-defined basic rules, while making 30% adjustments based on group behavior.

[0152] (b) Automatically match the interface layout template that best suits the current user profile from the preset layout template library; specifically including:

[0153] S31. Based on the predefined static business role information in the user profile, select a set of candidate templates that meet the role criteria from the template library;

[0154] S32. Based on the interface component preference weight vector in the user profile, calculate the cosine similarity with the preset component layout of each template in the candidate template set.

[0155] S33. Select the template with the highest cosine similarity to the interface component preference weight vector as the most suitable interface layout template.

[0156] (c) Based on the display priority, perform differentiated display control and arrangement optimization of functional components in the interface; S34. In the main work area of ​​the most suitable interface layout template, prioritize rendering the top N functional components with the highest display priority.

[0157] S35: Aggregate the functional components with a priority ranking of Nth or lower into the secondary menu;

[0158] S36. Within the same area, functional components are arranged in descending order according to their display priority scores.

[0159] After the S3 steps are executed, the system provides a multi-level configuration collaboration mechanism. This mechanism allows users to manually fine-tune the personalized layout automatically generated by the system. The user's manual configuration has the highest priority and is recorded as feedback data in the user profile to optimize subsequent automatic layout generation.

[0160] S4. Collect subsequent user interaction data on the personalized front-end layout and feed it back to S1 for updating the user profile and optimizing layout decisions. It should be noted that this step specifically includes:

[0161] S41. Collect the user's subsequent interaction behavior with the personalized front-end layout. The subsequent interaction behavior includes at least the click rate of rearranged functional components, the access depth of functions in the sub-menu, and the user's manual adjustment of the layout.

[0162] After rendering the personalized layout, the system immediately initiates a new round of front-end data collection, focusing on interactive behaviors that can verify the layout effect. These behaviors are encapsulated as feedback logs, whose core metrics include:

[0163] (1) Click-through rate (CTR) of rearranged functional components:

[0164] Click-through rate = Number of clicks on a specific component / Total number of impressions of that component

[0165] The system records the number of times each component whose position is dynamically adjusted (such as being moved to the main workspace) is displayed and clicked in its new position.

[0166] (2) Access depth to functions within sub-menus:

[0167] This counts the number of subsequent steps a user takes after clicking a submenu and accessing a hidden function. For example, if a user clicks "More" -> selects "Historical Data Report" -> clicks "Export," the access depth is counted as 3 steps.

[0168] The shallower the depth, the more difficult it is for users to find the function, and the greater the negative effects of this hidden decision may be.

[0169] (3) User-manual layout adjustment operation:

[0170] The system captures user actions that directly intervene in the layout, such as dragging and dropping components to move them in / out of the main workspace, and fixing or unfixing components. It also records the target component, operation type, and timestamp of the operation.

[0171] S42. Compare subsequent interactive behaviors with the user profile and layout decisions used to generate the personalized front-end layout, and evaluate the effectiveness of the layout decisions, i.e., evaluate the effectiveness of the behavior; specifically including:

[0172] If the click-through rate of a high-priority feature component that is highlighted is lower than the expected threshold during the display period, it is determined that there is a deviation in the priority calculation of that feature component in this layout decision.

[0173] High-priority components have a low click-through rate (misjudged as high priority):

[0174] For a component that is determined to be high priority by the S3 engine and placed in the main workspace, the system sets an expected click-through rate threshold (e.g., 5%). If the actual click-through rate of the component is lower than the expected click-through rate threshold within an evaluation period (set to 24 hours in this embodiment), a misjudgment of high priority is triggered.

[0175] If a user frequently moves a hidden functional component to the main workspace manually, it is determined that the layout decision has a deviation in controlling the visibility of that functional component.

[0176] Users frequently manually restore hidden features (misjudging low priority):

[0177] For a component hidden in a secondary menu by the S3 engine, the system records the number of times the user manually moves it back to the main workspace. A manual restoration frequency threshold is set (e.g., if the same user performs this action twice within 3 days). If the frequency exceeds the threshold, a low-priority misjudgment is triggered.

[0178] S43. Based on the results of the behavioral effectiveness evaluation, dynamically adjust the preference weight vectors of relevant functional components in the user interface preference model; and / or modify the weight parameters in the formula used to calculate the display priority score. This step involves incremental updates to the model and parameters, specifically including:

[0179] (1) Dynamically adjust the weight vector in the user interface preference model

[0180] Directly correct the interface component preference weight vector assigned to dynamic behavior labels in S23.

[0181] Adjust the formula:

[0182] For components that are mistakenly identified as having high priority: reduce their weight.

[0183]

[0184] in, The attenuation coefficient is set to 0.2 in this embodiment of the invention, which means that the weight is reduced by 20%.

[0185] For components that are mistakenly classified as low priority (frequent manual recovery): increase their weight.

[0186]

[0187] in, To enhance the coefficient, it is set to 0.3 in this embodiment of the invention, which means that the weight is increased by 30%.

[0188] (2) Correct the weight parameter in the formula for calculating the priority score.

[0189] The priority calculation formula for S3(a) is revised. The weight parameters include: function usage frequency weight. Weighting of Function Usage Time Business role weight .

[0190] Fine-tuning is performed using the gradient descent approach. First, the loss of the current decision is calculated.

[0191] Loss function definition: Loss = (Number of incorrectly predicted components) / (Total number of evaluated components)

[0192] Based on loss, by learning rate The weight parameters are updated to a value of 0.05. The update direction is to reduce the overall loss. The following rules are adopted:

[0193] If errors caused by misjudgment of click-through rate are the dominant factor, then appropriately reduce... and correspondingly improve and .

[0194]

[0195]

[0196]

[0197] If errors caused by manual recovery misjudgments dominate, it indicates that the system relies too heavily on historical behavior. , ), while ignoring the character's rigid requirements ( At this point, the increase should be made. .

[0198]

[0199] Once all adjustments are complete, the new weight vectors and formula parameters will take effect immediately and be used in the decision-making process for generating layouts for the user in the next iteration, thereby enabling the system to evolve itself.

[0200] like Figure 2 As shown, this embodiment of the invention also provides a user behavior-driven dynamic optimization system for the front-end layout of a virtual power plant, used to implement the methods described in the above embodiments, including:

[0201] The behavior data acquisition module is used to collect user interaction behavior data in the virtual power plant management system. The interaction behavior data includes at least menu item click events, function module access and exit events, and interface jump events, and each data record has a corresponding user identifier and timestamp.

[0202] The user profile building module is used to generate dynamic behavior tags through cluster analysis based on the collected interaction behavior data and combined with the user's predefined static business role information, and to establish a user interface preference model to form a unified multi-level user profile that includes both static roles and dynamic behaviors.

[0203] The dynamic layout engine module is used to calculate the display priority of each menu item and functional component; automatically match the interface layout template that best suits the current user profile from the preset layout template library; and perform differentiated display and hiding control and arrangement optimization of functional components in the interface according to the display priority to generate and output a personalized front-end layout.

[0204] The closed-loop feedback module is used to collect subsequent interaction data of the user with the personalized front-end layout and feed it back to the behavior data collection module for updating the user profile and optimizing layout decisions.

[0205] In some embodiments, the behavior data acquisition module includes:

[0206] The front-end probe unit is used to capture user clicks, browsing duration, and page jump events in real time by embedding data acquisition probes into the front-end interface components of the virtual power plant management system.

[0207] The log structuring unit is used to transform captured events into structured behavior logs, which at least include user ID, timestamp, operation object, and operation type.

[0208] The session construction and cleaning unit is used to divide sessions according to user ID and timestamp based on the behavior log, and to filter out meaningless erroneous operation events and heartbeat keep-alive events using a business rule-based cleaning algorithm, so as to construct a user operation sequence that represents a complete operation task.

[0209] In some embodiments, the user profile building module includes:

[0210] The role management unit is used to store and manage the predefined static business role information and its corresponding initial interface layout template and function permission set;

[0211] The behavior analysis unit is used to extract user behavior feature vectors from the interaction behavior data within a set statistical period. These vectors include at least menu item click frequency, function usage duration, and operation sequence patterns. An unsupervised clustering algorithm is used to calculate the behavior feature vectors, dividing users into different behavior clusters. A dynamic behavior label is assigned to each behavior cluster, including at least high-frequency scheduling, data browsing, and transaction-driven labels. The unit analyzes the timestamp information in the interaction behavior data to identify user function usage patterns during weekdays and holidays, and peak and off-peak electricity consumption periods. The identified time patterns are used as a correction factor to optimize the classification results of the dynamic behavior labels and / or to optimize the weight allocation of the interface component preference weight vector.

[0212] The preference model construction unit is used to assign a set of interface component preference weight vectors to each type of dynamic behavior label. The weight vectors quantify the degree of preference of users of this type for different functional modules, data panels and operation controls. The preference weight vectors are associated and mapped with the corresponding dynamic behavior labels to construct a preliminary user interface preference model.

[0213] The profile encapsulation unit is used to integrate the dynamic behavior tags and the corresponding preliminary user interface preference model with the predefined static business role information as the base layer, and encapsulate them together into a unified user profile data entity that can be directly called by the dynamic layout engine, thus forming a unified multi-level user profile.

[0214] In some embodiments, the dynamic layout engine module includes:

[0215] The priority calculation unit is used to calculate the display priority of each menu item and functional component based on the multi-level user profile and a preset algorithm.

[0216] The template matching unit is used to automatically match the most suitable interface layout template from the preset layout template library based on the multi-level user profile.

[0217] The interface optimization unit is used to perform differentiated display and hiding control and arrangement optimization of functional components in the interface according to the display priority.

[0218] A multi-level configuration collaboration unit is used to execute a multi-level configuration collaboration mechanism and process user manual fine-tuning commands. The multi-level configuration collaboration mechanism allows users to manually fine-tune the personalized layout automatically generated by the system, and the user's manual configuration has the highest priority and is recorded as feedback data in the user profile for optimizing subsequent automatic layout generation.

[0219] In some embodiments, the closed-loop feedback module includes:

[0220] The feedback behavior collection unit is used to collect users' subsequent interaction behaviors with the personalized front-end layout;

[0221] The decision evaluation unit is used to compare subsequent interactive behaviors with the user profile and layout decisions used to generate the personalized front-end layout, and to evaluate the effectiveness of the layout decisions.

[0222] The model and algorithm update unit is used to dynamically adjust the preference weight vector of relevant functional components in the user interface preference model based on the evaluation results; and / or, correct the function usage frequency weight, function usage duration weight, and business role weight in the calculation formula.

[0223] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for dynamic optimization of the front-end layout of a virtual power plant based on user behavior, characterized in that, Includes the following steps: S1. Collect user interaction behavior data in the virtual power plant management system; the interaction behavior data includes at least menu item click events, function module access and exit events, and interface jump events, wherein each data record has a corresponding user identifier and timestamp; S2. Based on the collected interaction behavior data and combined with the user's predefined static business role information, dynamic behavior tags are generated through cluster analysis, and a user interface preference model is established to form a unified multi-level user profile that includes both static roles and dynamic behaviors; specifically including: S21. Extracting the user's behavior feature vector within a set statistical period from the interaction behavior data, wherein the vector includes at least the menu item click frequency, function usage duration, and operation sequence pattern. S22. Calculate the behavioral feature vector using an unsupervised clustering algorithm to divide users into different behavioral clusters; and assign a dynamic behavioral label to each behavioral cluster, the label including at least high-frequency scheduling type, data browsing type, and transaction-driven type; S23. For each type of dynamic behavior label, assign a set of interface component preference weight vectors, wherein the weight vectors quantify the degree of preference of the user for different functional modules, data panels and operation controls. S24. Associate and map the preference weight vector with the corresponding dynamic behavior labels to construct a preliminary user interface preference model; S25. Using the predefined static business role information as the base layer, integrate the dynamic behavior tags and the corresponding preliminary user interface preference model, and encapsulate them together into a unified user profile data entity that can be directly called by the dynamic layout engine, thus forming a unified multi-level user profile. S26. Analyze the timestamp information in the interaction behavior data to identify the user's function usage patterns on weekdays and holidays, and during peak and off-peak electricity consumption periods; use the identified time patterns as a correction factor to optimize the classification results of dynamic behavior tags in S22, and / or to optimize the weight allocation of the interface component preference weight vector in S23. S3. Based on multi-layered user profiles, perform the following operations using a dynamic layout engine to generate and output personalized front-end layouts: (a) Calculate the display priority of each menu item and functional component; (b) Automatically match the interface layout template that best suits the current user profile from the preset layout template library; (c) Based on the display priority, perform differentiated display control and arrangement optimization of functional components within the interface; S4. Collect subsequent user interaction data on the personalized front-end layout and feed it back to S1 for updating user profiles and optimizing layout decisions; In step S3(a), calculate the display priority of each menu item and functional component, specifically through the following formula: in, To display priority scores, Use frequency weights for functions Weighting based on feature usage time Assigning weights to business roles To normalize the click-through rate, To normalize usage time, The role-function relevance is dynamically adjusted based on the dynamic behavior tags in the user profile, including the weights for function usage frequency, function usage duration, and business role. The role-function relevance is obtained by querying a predefined role-function weight mapping table, which is established based on the business rules of the virtual power plant and the responsibilities and permissions of different roles. The weight values ​​in the mapping table are periodically and dynamically adjusted based on the historical average behavior data of the user group under the same role.

2. The method for dynamic optimization of virtual power plant front-end layout based on user behavior as described in claim 1, characterized in that, Step S1 includes: S11. By embedding data acquisition probes into the front-end interface components of the virtual power plant management system, user clicks, browsing duration, and page jump events are captured in real time. S12. Convert the captured events into structured behavior logs, which include at least user ID, timestamp, operation object, and operation type; S13. Based on the behavior log, sessions are divided according to user ID and timestamp, and a business rule-based cleaning algorithm is used to filter out meaningless erroneous operation events and heartbeat keep-alive events to construct a user operation sequence that represents a complete operation task.

3. The method for dynamic optimization of virtual power plant front-end layout based on user behavior as described in claim 2, characterized in that, In S2, the predefined static business role information includes at least dispatchers, traders and energy owners; the system has a pre-defined differentiated initial interface layout template and function permission set for each role.

4. The method for dynamic optimization of virtual power plant front-end layout based on user behavior as described in claim 3, characterized in that, In step S3(b), the most suitable interface layout template is automatically matched from the preset layout template library, specifically including: S31. Based on the predefined static business role information in the user profile, select a set of candidate templates that meet the role criteria from the template library; S32. Based on the interface component preference weight vector in the user profile, calculate the cosine similarity with the preset component layout of each template in the candidate template set. S33. Select the template with the highest cosine similarity to the interface component preference weight vector as the most suitable interface layout template.

5. The method for dynamic optimization of virtual power plant front-end layout based on user behavior as described in claim 4, characterized in that, In step S3(c), based on the display priority, differentiated display and hiding control and arrangement optimization are performed on the functional components within the interface, specifically including: S34. In the main workspace of the most suitable interface layout template, prioritize rendering and displaying the top N functional components with the highest priority. S35: Aggregate the functional components with a priority ranking of Nth or lower into the secondary menu; S36. Within the same area, functional components are arranged in descending order according to their display priority scores.

6. The method for dynamic optimization of virtual power plant front-end layout based on user behavior as described in claim 5, characterized in that, After the S3 steps are executed, the system provides a multi-level configuration collaboration mechanism. This mechanism allows users to manually fine-tune the personalized layout automatically generated by the system. The user's manual configuration has the highest priority and is recorded as feedback data in the user profile to optimize subsequent automatic layout generation.

7. The method for dynamic optimization of virtual power plant front-end layout based on user behavior as described in claim 6, characterized in that, The steps in S4 include: S41. Collect the user's subsequent interaction behavior with the personalized front-end layout. The subsequent interaction behavior includes at least the click rate of rearranged functional components, the access depth of functions in the sub-menu, and the user's manual adjustment of the layout. S42. Compare subsequent interactive behaviors with the user profile and layout decisions used to generate the personalized front-end layout, and evaluate the effectiveness of the layout decisions. S43. Based on the evaluation results, dynamically adjust the preference weight vector of relevant functional components in the user interface preference model; and / or, modify the function usage frequency weight, function usage duration weight, and business role weight in the calculation formula of step S3(a).

8. A user-behavior-driven dynamic optimization system for the front-end layout of a virtual power plant, used to implement the method of any one of claims 1 to 7, characterized in that, include: The behavior data acquisition module is used to collect user interaction behavior data in the virtual power plant management system. The interaction behavior data includes at least menu item click events, function module access and exit events, and interface jump events, and each data record has a corresponding user identifier and timestamp. The user profile building module is used to generate dynamic behavior tags through cluster analysis based on the collected interaction behavior data and combined with the user's predefined static business role information, and to establish a user interface preference model to form a unified multi-level user profile that includes both static roles and dynamic behaviors. The dynamic layout engine module is used to calculate the display priority of each menu item and functional component; automatically match the interface layout template that best suits the current user profile from the preset layout template library; and perform differentiated display and hiding control and arrangement optimization of functional components in the interface according to the display priority to generate and output a personalized front-end layout. The closed-loop feedback module is used to collect subsequent interaction data of the user with the personalized front-end layout and feed it back to the behavior data collection module for updating the user profile and optimizing layout decisions.

Images