Dynamic evolution analysis method suitable for graphical user side demand response strategy

Abstract

The invention discloses a graphical user side demand response strategy dynamic evolution analysis method. The method comprises the steps of obtaining demand response resource original data and constructing an initial resource model; constructing a resource flow dynamic structure model, and generating a resource flow topological structure and a resource association intensity matrix; real-time monitoring and evaluation of cognitive loads are realized, and user cognitive load indexes are calculated; interface element self-adaptive adjustment is executed, and interface parameter configuration is generated; and carrying out resource flow dynamic layout optimization to generate a resource flow visual layout. Through cognitive load dynamic evaluation and interface adaptive adjustment, the problem of cognitive load during arrangement of a large number of resources is solved, the strategy arrangement efficiency is improved, and the operation error rate is reduced.

Description

Technical Field

[0001] The present invention belongs to the field of user-side demand response, and in particular, to a method for dynamically evolving analysis of user-side demand response strategies applicable to graphical representation. Background Art

[0002] Demand response in the power system is an important means to achieve energy supply-demand balance and improve system resilience. With the increasing penetration of distributed energy and renewable energy, user-side demand response resources have shown the characteristics of a sharp increase in quantity, diverse types, and wide distribution. How to efficiently manage and orchestrate these demand response resources to form effective response strategies has become a key link in the construction of the energy Internet and smart grid. Especially in extreme weather and peak electricity consumption periods, the ability to quickly formulate and flexibly adjust demand response strategies is directly related to the safe and stable operation of the power grid and the guarantee of electricity reliability.

[0003] Current research mainly focuses on the static management and simple visual display of demand response resources. Common methods include resource management interfaces based on list or grid layouts, using simple color coding and icons to distinguish different types of resources, and using basic drag-and-drop mechanisms to transfer resources from one area to another. Some systems introduce grouping and filtering functions based on tags or attributes, as well as an interactive method of displaying detailed resource attributes through tooltips. In terms of expressing the relationships between resources, mainly static connections or simple hierarchical tree structures are used, and the temporal correlation and operation relevance between resources are not fully considered.

[0004] However, these methods have obvious deficiencies when users need to manage a large number of resources (>20) simultaneously for strategy orchestration. First, static grouping and filtering cannot adapt to the dynamic changes of resource states. When resource parameters are updated in real time, it is difficult for users to quickly perceive the changes and adjust the strategies. Second, when the display area is fixed and the number of resources increases, the visual representation of resources inevitably overlaps and becomes crowded, making it difficult for users to accurately select and operate specific resources. Third, the multi-dimensional relationships between resources (such as superior-subordinate, dependence, mutual exclusion, etc.) are difficult to comprehensively and intuitively express on a traditional two-dimensional plane. Users need to frequently switch views or remember multiple relationships, increasing the cognitive burden. Most critically, the existing interfaces lack the ability to perceive the user's cognitive state and cannot dynamically adjust the interface complexity and interaction assistance level according to the user's cognitive load, resulting in visual fatigue and attention dispersion for users after long-term operation, affecting the accuracy and efficiency of strategy orchestration. Summary of the Invention

[0005] The object of the invention is to provide a method for dynamically evolving analysis of user-side demand response strategies applicable to graphical representation, in order to solve at least one technical problem existing in the prior art.

[0006] Technical solution, applicable to a graphical user-side demand response strategy dynamic evolution analysis method, including: Obtain the original data of demand response resources and construct an initial resource model, generate resource basic characterization data, and based on this, construct a resource flow dynamic structure model, generate a resource flow topology structure and a resource association strength matrix; Realize real-time monitoring and evaluation of cognitive load according to user real-time interaction data, calculate user cognitive load indicators and perform adaptive adjustment of interface elements based on them, and generate interface parameter configurations; Based on the resource flow topology structure, the resource association strength matrix and the interface parameter configuration, optimize the dynamic layout of the resource flow and generate a visual layout of the resource flow.

[0007] Beneficial effects: The present invention realizes the accurate perception of user cognitive load and the intelligent adjustment of the interface, effectively solves the cognitive and operation challenges during the simultaneous orchestration of a large number of resources, reduces the user cognitive load, improves the operation efficiency, reduces the operation error rate, enhances the user satisfaction, and provides effective technical support for the visual orchestration of user-side demand response strategies. Description of the drawings

[0008] Figure 1 It is a step flow chart of a graphical user-side demand response strategy dynamic evolution analysis method provided by an embodiment of the present application.

[0009] Figure 2 It is a step flow chart of generating a resource flow topology structure provided by an embodiment of the present application.

[0010] Figure 3 It is a step flow chart of generating a time-sequential resource structure provided by an embodiment of the present application.

[0011] Figure 4 It is a step flow chart of generating user cognitive load indicators provided by an embodiment of the present application. Detailed implementation manners

[0012] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0013] It should be noted that to clearly illustrate the steps of this application, serial numbers are assigned to each step in the specification. These serial numbers are for illustrative purposes only and do not limit the order in which the steps must be executed. In actual operation, depending on the technical requirements of the specific implementation scenario, the steps may be executed in a different order than shown in the specification, and in some cases, parallel processing between steps may be implemented.

[0014] like Figure 1 As shown, a graphical user-side demand response strategy dynamic evolution analysis method includes the following steps: S1. Obtain the original data of demand response resources and build an initial resource model to generate resource basic representation data; Specifically, the raw data of demand response resources includes resource identification, resource type, historical load curve, adjustment range, response time, and superior-subordinate relationship. The resource identification is the identity information of the resource, such as the number or name; the resource type is the classification of the resource, such as whether it is a residential power resource or an industrial power resource; the historical load curve is the record of the resource's past power consumption, showing the changing trend in the form of a curve; the adjustment range is the range of power usage that the resource can adjust, such as the minimum and maximum load; the response time is the length of time the resource can participate in demand response, such as how long the adjustment can continue; the superior-subordinate relationship is the hierarchical relationship between resources, such as whether one resource is subordinate to another resource.

[0015] S2. Based on the resource base representation data, a dynamic structural model of resource flow is constructed to generate the resource flow topology structure and resource association intensity matrix; Specifically, the resource flow dynamic structure model reflects resource mobility and dynamic relationships, and can be used to represent interactions, flow paths, and dynamic changes between resources. The resource flow topology is a network-like representation that graphically displays resource connections and flow directions. For example, nodes and lines can be used to illustrate how resources are interconnected and how flows are distributed. The resource association strength matrix quantitatively describes the strength of associations between resources, demonstrating the extent to which some resources are closely connected while others are less so.

[0016] S3. Realize real-time monitoring and evaluation of cognitive load based on real-time user interaction data and calculate user cognitive load indicators; Specifically, real-time user interaction data includes mouse movement trajectory, dwell time, click frequency, and drag operation characteristics. Based on this real-time interaction data, we analyze whether the user experiences excessive cognitive load during the current task, such as whether the task is too complex or contains too much information, leading to decreased efficiency or increased errors. We then assess the user's cognitive load level, for example, by comparing it to normal interaction behavior to infer the user's current stress level.

[0017] S4. Based on the user cognitive load index, perform adaptive adjustment of interface elements to generate interface parameter configurations. Specifically, the interface is intelligently adjusted according to the user's load level, that is, it will automatically adapt to the user's current state. For example, if the user's load is too high, the system may hide some unnecessary information and simplify the interface; if the user's load is low, the system may present more information or increase the interaction density.

[0018] S5. Based on the resource flow topology structure, resource association strength matrix, and interface parameter configurations, perform optimization of the dynamic layout of resource flows to generate a visual layout of resource flows.

[0019] Specifically, the layout of resources is optimized, and the purpose of the optimization is to make the resources presented on the interface more reasonable and clearer. For example, avoid nodes being too crowded or make the flow path more intuitive.

[0020] This embodiment can track the topology structure and association strength of resource flows in real time to ensure the accuracy and efficiency of data flow; can optimize the user interface, adaptively adjust interface parameters according to the user's cognitive load, improve the interaction experience and work efficiency; dynamically adjust the visual layout of resource flows to make information presentation more intuitive and the hierarchy clearer, improve the user's ability to understand and process complex data; solve the cognitive load problem during a large number of resource orchestrations, improve the strategy orchestration efficiency, and reduce the operation error rate.

[0021] According to one aspect of the present application, the steps of constructing an initial resource model and generating resource basic characterization data include: S11. Read the original data of demand response resources, including resource identifiers, resource types, historical load curves, adjustment ranges, response durations, and superior-subordinate relationships. After standardization and cleaning, obtain the cleaned load data. S12. Based on the cleaned load data, calculate the response potential indicators of each resource in different time periods to generate a resource response potential feature matrix. S13. Integrate the cleaned load data and the resource response potential feature matrix, construct a resource logical association graph, calculate the importance weights of each resource, and generate resource basic characterization data.

[0022] Specifically, read the original data of demand response resources, extract and standardize through a data connector to generate a standardized resource data set; apply an outlier detection algorithm to the historical load curve data in the standardized resource data set to identify and correct abnormal fluctuation points to generate the cleaned load data.

[0023] According to one aspect of the present application, the steps of generating a resource response potential feature matrix include: The cleaned load data is divided into typical daily load patterns, classified based on date type and seasonal characteristics, and a classified load dataset is generated; based on the classified load dataset, load characteristics are statistically analyzed and basic response potential index is extracted through a basic response potential evaluation model; Time sensitivity analysis is performed on the basic response potential index, the response potential changes of resources at different time periods are calculated, and a time-varying response potential curve is generated; The actual response performance of resources in historical demand response events is analyzed, the deviation rate between the expected response and the actual response is calculated, and a response reliability index is generated; Combining the time-varying response potential curve and the response reliability index, a multi-dimensional response potential evaluation matrix is constructed to generate a resource response potential characteristic matrix.

[0024] Specifically, read the cleaned load data, divide the historical load curve of each resource into typical daily load patterns, classify them based on date type (weekday, weekend, holiday) and seasonal characteristics (such as applying the time series segmentation algorithm), and generate a classified load dataset. Apply the statistical feature extraction algorithm to each type of load curve in the classified load dataset, calculate features such as peak load, valley load, average load, and load volatility, and generate a load statistical feature set. Combine the regulation characteristic data of the resources (such as adjustable ratio, response speed, minimum adjustment step size, etc.) to construct a basic response potential evaluation model, which considers the regulation ability differences of resources at different load levels and generates a basic response potential index. Apply time sensitivity analysis to the basic response potential index, calculate the response potential changes of resources at different time periods in a day through the moving window method, identify high-response potential periods and low-response potential periods, and generate a time-varying response potential curve. Combine the historical demand response event data, analyze the actual response performance of resources in past events, calculate the deviation rate between the expected response and the actual response as the resource reliability index, and generate a response reliability index. Combine the time-varying response potential curve and the response reliability index to construct a multi-dimensional response potential evaluation matrix, which includes the response ability and reliability of resources under different time periods and different load conditions, and generate a resource response potential characteristic matrix.

[0025] According to one aspect of the present application, the steps of generating the resource flow topology structure and the resource association strength matrix include: S21. Read the basic resource characterization data, construct a directed graph of resource relationships based on resource time series correlation and operation relevance, and generate a resource backbone structure diagram; S22. Perform time series mapping on the resource backbone structure diagram, map the resources to the time axis according to the resource time attributes (response start time, duration, end time), and generate a time-series resource structure; S23. Calculate the association strength between resource pairs based on the time-ordered resource structure, taking into account factors such as the overlap degree of regulation timings, the correlation of response quantities, and the mutual exclusivity of regulations, and generate a resource association strength matrix; S24. Combine the resource association strength matrix with the time-ordered resource structure, and apply an improved Sankey diagram algorithm to construct a resource flow model and generate a resource flow topology structure.

[0026] As Figure 2 shown, according to one aspect of the present application, the steps of applying an improved Sankey diagram algorithm to construct a resource flow model and generate a resource flow topology structure include: Read the time-ordered resource structure and the resource association strength matrix, perform association strength threshold filtering and weighted network modularity analysis, and generate resource functional clusters; Combine the resource functional clusters and the resource time attributes, construct an initial structure of a directed flow graph, arrange the resources in chronological order while maintaining the internal aggregation of the clusters, and generate an initial flow graph skeleton; Merge the connections with parallel directions and similar targets in the initial flow graph skeleton into weighted flow bundles to generate an optimized flow connection structure; combine the optimized flow connection structure with the resource association strength matrix, calculate the width of the flow bundles, generate a visually weighted flow graph, and perform node position fine-tuning and edge path optimization to obtain a resource flow topology structure.

[0027] Specifically, read the time-ordered resource structure and the resource association strength matrix, perform association strength threshold filtering, retain the connections of resource pairs with association strength exceeding the minimum threshold, and form a simplified association network. Perform weighted network modularity analysis based on association strength (such as applying a community detection algorithm) on the simplified association network to identify highly relevant resource clusters and generate resource functional clusters. Combine the resource functional clusters and the resource time attributes, construct an initial structure of a directed flow graph, apply a hierarchical layout strategy to arrange the resources in chronological order, and at the same time maintain the relative aggregation of resources within the functional clusters to generate an initial flow graph skeleton. Merge multiple connections with parallel directions and similar targets in the initial flow graph skeleton into weighted flow bundles (such as applying an edge bundling optimization algorithm) to reduce visual complexity while retaining the semantic information of the connections and generate an optimized flow connection structure. Calculate the width of each flow bundle in the optimized flow connection structure according to the resource association strength and resource importance (such as applying a flow mapping algorithm) to ensure that the visual representation is proportional to the actual association degree and generate a visually weighted flow graph. Perform node position fine-tuning and edge path optimization (such as applying a topology crossing minimization algorithm) on the visually weighted flow graph to reduce unnecessary streamline crossings and improve the readability of the graph, and generate the final resource flow topology structure.

[0028] As Figure 3 shown, according to one aspect of the present application, the steps of performing time mapping and generating a time-ordered resource structure include: Read the resource backbone structure diagram and resource time attribute data, build the resource time axis coordinate system, and generate the time axis mapping framework; Normalize the time attributes of all resources in the timeline mapping framework to generate normalized time attributes; Based on the normalized time attribute, the horizontal coordinate value is calculated to generate the initial horizontal position of the resource; Combine the hierarchical relationship and functional classification information of resources to calculate the vertical coordinate value and generate the initial vertical position of the resource; Combining the initial horizontal and vertical positions of resources, an initial two-dimensional layout is constructed, and resource overlap is resolved through force-directed local adjustments to generate adjusted resource positions. The time correlation of the adjusted resource positions is maintained to obtain a time-series resource structure.

[0029] Specifically, the resource backbone structure diagram and resource time attribute data are read to extract the response start time, duration, and end time of each resource. A resource timeline coordinate system is constructed, with the horizontal axis representing time and the vertical axis representing resource hierarchy. This generates a timeline mapping framework. The time attributes of all resources are normalized, converting absolute times to offsets relative to the response event start time to ensure consistency in time representation. Normalized time attributes are then generated. Based on these normalized time attributes, the initial position of each resource on the timeline is calculated. Time values are converted to horizontal coordinates (e.g., using a linear mapping) to generate the resource's initial horizontal position. Combining the resource's hierarchical relationships and functional classification information, the vertical position of each resource is calculated, ensuring appropriate vertical spacing between superior and subordinate resources (e.g., using a hierarchical algorithm) and that related resources are close in vertical position. This generates the resource's initial vertical position. Combining these initial horizontal and vertical positions, an initial two-dimensional layout is constructed. Resource overlap is detected and resolved, ensuring sufficient display space for each resource (e.g., using a force-directed local adjustment algorithm), and the adjusted resource positions are generated. Maintain the temporal correlation of the adjusted resource positions to ensure that the strong correlation between the resource time attributes and their horizontal positions is maintained after resolving the overlap problem. Make trade-off adjustments when necessary to generate a time-series resource structure.

[0030] According to one aspect of the present application, the step of calculating the user cognitive load index includes: S31. Capture real-time user interaction data on the interface, including mouse movement trajectory, dwell time, click frequency, and drag operation characteristics, and generate an original interaction data stream; S32. Apply sliding window analysis to the original interactive data stream to extract the operation speed change rate, operation accuracy, and repeated operation frequency, and generate an operation feature vector; S33. Based on the operation feature vector, perform real-time cognitive load assessment, calculate the cognitive load index, and generate the user's cognitive load index.

[0031] As Figure 4 shown, according to one aspect of the present application, the steps of performing real-time cognitive load assessment and generating the user's cognitive load index include: Decompose the operation feature vector into three sub-vectors: operation trajectory (such as mouse trajectory, swipe trajectory), operation time, and operation frequency feature (such as through a feature separation algorithm) to generate a classification feature set; Analyze the operation trajectory features in the classification feature set, calculate the trajectory smoothness, direction change frequency, and acceleration volatility (such as applying a curvature analysis algorithm), quantify the trajectory uncertainty (such as using an interval statistical method), and generate a trajectory uncertainty index; Analyze the operation time features in the classification feature set, identify operation pause points and acceleration points (such as applying an attention fluctuation analysis algorithm based on peak detection), calculate the operation rhythm stability, and generate an attention dispersion index; Analyze the operation frequency features in the classification feature set, identify repeated operation sequences and error-correction patterns (such as applying an abnormal pattern detection algorithm), calculate the operation redundancy, and generate an operation efficiency index; Integrate the trajectory uncertainty, attention dispersion, and operation efficiency indices, perform non-linear weighted mapping and time-series smoothing processing according to mental load, time pressure, and frustration level, generate a smoothed cognitive load curve and map it to discrete cognitive load levels, and at the same time calculate the current load trend to form the user's cognitive load index.

[0032] Specifically, integrate the trajectory uncertainty index, attention dispersion index, and operation efficiency index to construct a cognitive load assessment function. This function is based on an improved multiple resource theory model, performs non-linear weighted mapping of each index according to three dimensions: mental load, time pressure, and frustration level, and generates a dimensionalized cognitive load index. Apply time-series smoothing processing to the dimensionalized cognitive load index, use the exponential weighted moving average algorithm to eliminate short-term fluctuations while retaining medium- and long-term trends, and avoid the additional cognitive burden caused by frequent interface adjustments to generate a smoothed cognitive load curve. Based on the smoothed cognitive load curve, apply a multi-threshold piecewise function to map continuous cognitive load values to discrete cognitive load levels (low, medium, high, ultra-high), and at the same time calculate the current load trend (rising, stable, falling), and combine them to form the user's cognitive load index.

[0033] According to one aspect of the present application, the steps of performing adaptive adjustment of interface elements and generating interface parameter configurations include: S41. Receive the user's cognitive load index, determine the priority of interface parameters to be adjusted, and form a list of adjustment parameter priorities; S42. Combine the adjusted parameter priority list with the preset interface parameter adjustment rules to calculate the target values of interface parameters, including information density, visual discrimination, interaction assistance intensity, and animation smoothness; S43. Based on the target values of interface parameters, adopt a progressive parameter adjustment method to avoid cognitive interference caused by interface mutations and generate interface parameter configurations.

[0034] According to one aspect of the present application, the steps of generating an interface parameter configuration by adopting a progressive parameter adjustment method include: Read the target values of interface parameters and the current interface parameter status, calculate the parameter difference, and generate a parameter adjustment requirement vector; Assign weights to the parameter adjustment requirement vector according to the influence degree of different parameters on the cognitive load to generate a weighted adjustment requirement vector; According to the weighted adjustment requirement vector and the user cognitive load index, construct a parameter adjustment rate function, reduce the adjustment rate in a high cognitive load state, and generate an adaptive adjustment rate table; Combine the adaptive adjustment rate table with the weighted adjustment requirement vector, calculate the parameter increment value, ensure that the change is smooth and does not exceed the human perception threshold, and generate a parameter increment vector; Add the current parameter value to the parameter increment vector, and perform parameter constraint and parameter coordination check to generate an interface parameter configuration.

[0035] Specifically, read the target values of the interface parameters and the current status of the interface parameters, calculate the parameter difference vector, including the gap between the current value and the target value of each parameter, and generate a parameter adjustment requirement vector. Apply cognitive sensitivity analysis to the parameter adjustment requirement vector, assign weights according to the influence degree of different parameters on the cognitive load, and important parameters (such as information density, contrast) obtain higher weights to generate a weighted adjustment requirement vector. According to the weighted adjustment requirement vector and the current user cognitive load index, construct a parameter adjustment rate function, which reduces the adjustment rate in the high cognitive load state to avoid sudden changes, and increases the adjustment rate in the low cognitive load state to accelerate convergence, and generate an adaptive adjustment rate table. Combine the weighted adjustment requirement vector and the adaptive adjustment rate table, calculate the incremental value of each parameter in the next frame, ensure that the change is smooth and does not exceed the human perception threshold, and generate a parameter increment vector. Add the current parameter value to the parameter increment vector to generate a new interface parameter value, and at the same time apply parameter constraint checking to ensure that all parameter values are within the valid range, and generate updated parameter values. Apply parameter coordination checking to the updated parameter values to ensure that the proportional relationship between related parameters conforms to the interface design specifications, such as the text size and the container size, the contrast between the foreground color and the background color, etc., and perform coordination adjustment if necessary to generate coordinated parameter values. Calculate the difference between the coordinated parameter values and the target parameter values, determine whether the convergence condition is reached, if not converged, return to the step of applying cognitive sensitivity analysis to continue iteration, if converged, output the interface parameter configuration, and complete the progressive adjustment process.

[0036] According to one aspect of the present application, the steps of performing resource flow dynamic layout optimization and generating a resource flow visualization layout include: S51. Integrate the resource flow topology structure, the resource association strength matrix, and the interface parameter configuration, construct a multi-objective layout optimization problem, define an optimization function including cognitive load, visual aesthetics, and operation efficiency, and generate a layout optimization model; S52. Apply an improved simulated annealing algorithm to the layout optimization model to optimize the positions of resource nodes and generate optimized coordinates of resource nodes; S53. Integrate the optimized coordinates of resource nodes and visual performance attributes into a complete visual description object, generate a resource flow visualization layout, and implement real-time interaction response.

[0037] According to one aspect of the present application, the steps of constructing a multi-objective layout optimization problem include: Read the resource flow topology structure, the resource association strength matrix, and the interface parameter configuration, construct a layout evaluation function including a cognitive load evaluation sub-function, a visual aesthetics evaluation sub-function, and an operation efficiency evaluation sub-function, and generate a multi-objective evaluation function; Calculate the visual entropy, the uniformity of element density distribution, and the clarity of visual guidance paths in the layout, and generate a cognitive load score; Evaluate the balance, symmetry, regularity, and rhythm of the layout to generate a visual aesthetics score; Based on the operation path analysis, evaluate the length, complexity, and target reachability of the key operation path to generate an operation efficiency score; Based on the current user cognitive load index, dynamically adjust the weights of the three sub-functions in the multi-objective evaluation function to generate an adaptive weighted evaluation function; Combine the topological constraints in the resource flow topology structure and the display parameters in the interface parameter configuration to construct a set of layout constraint conditions, including the minimum distance between nodes, the preference for non-crossing of flow lines, and the requirement for the cohesion of resource clusters, etc., to generate a layout constraint model; Integrate the adaptive weighted evaluation function and the layout constraint model to construct an optimization problem including decision variables (node coordinates), objective function (maximizing the comprehensive score), and constraint conditions to generate a layout optimization model.

[0038] Specifically, in the multi-objective evaluation function, the cognitive load evaluation sub-function calculates the visual entropy, the uniformity of element density distribution, and the clarity of the visual guidance path in the layout based on information theory, quantifies the impact of the layout on cognitive load, and generates a cognitive load score. In the multi-objective evaluation function, the visual aesthetics evaluation sub-function evaluates the balance, symmetry, regularity, and rhythm of the layout based on the principles of computational aesthetics, quantifies the aesthetic quality of the layout, and generates a visual aesthetics score. In the multi-objective evaluation function, the operation efficiency evaluation sub-function evaluates the length, complexity, and target reachability of the key operation path based on Fitts' law and operation path analysis, quantifies the impact of the layout on operation efficiency, and generates an operation efficiency score. Based on the current user cognitive load index, dynamically adjust the weights of the three sub-functions in the multi-objective evaluation function, increasing the weight of the cognitive load score when the cognitive load is high and increasing the weight of the visual aesthetics score when the cognitive load is low, to generate an adaptive weighted evaluation function.

[0039] In a specific embodiment of the present application, a method for dynamically evolving the analysis of the user-side demand response strategy applicable to graphing includes the following steps: Step 1: Obtain the demand response resource data and construct an initial resource model.

[0040] 1.1 Read the original demand response resource data.

[0041] The system reads the original data of 10 demand response resources from the database of the power demand response management platform, including: Resource identifiers: R001 to R010; Resource types: industrial users (R001 - R003), commercial buildings (R004 - R007), residential user clusters (R008 - R010); Historical load curves: 24-hour load data of 96 points (one point every 15 minutes) for each resource, with the unit of kW; Regulation range: the maximum adjustable capacity, such as 200 kW for R001, 150 kW for R002, etc.; Response duration: the longest sustainable response time, such as 2 hours for R001, 1.5 hours for R002, etc.; Superior-subordinate relationship: for example, R001 is the superior, and R004 and R005 are its subordinate resources. Through the data connector, the system extracts and normalizes the above various types of data to generate a standardized resource dataset SD.

[0042] 1.2. Apply the outlier detection algorithm.

[0043] Apply the outlier detection algorithm based on moving average to the historical load curve data in the standardized resource dataset SD: For the load curve L(i, t) of resource Ri, calculate the moving average value MA(i, t) = (1 / w)·∑(j=t - w / 2 to t + w / 2)L(i, j); calculate the standard deviation SD(i, t) = sqrt((1 / w)·∑(j=t - w / 2 to t + w / 2)(L(i, j) - MA(i, t)) 2 ); If |L(i, t) - MA(i, t)| > 3·SD(i, t), then the point L(i, t) is considered an outlier and is replaced by MA(i, t); where: L(i, t) is the load value of resource i at time point t; MA(i, t) is the moving average value; SD(i, t) is the standard deviation; w is the moving window width, with a value of 6; t is the time point index.

[0044] For example, the original load value of resource R001 at t = 37 is 210 kW, while its moving average value is 150 kW and the standard deviation is 15 kW. Since |210 - 150| > 3×15, the system identifies this point as an outlier and replaces it with 150 kW to generate the cleaned load data CL.

[0045] 1.3. Calculate the resource response potential index.

[0046] Based on the cleaned load data CL, calculate the response potential index of each resource at different time periods: The response potential index RP(i, t) of resource i at time period t = α·NL(i, t)·AR(i)·(1 - LV(i, t) / ML(i))·RF(i); where NL(i, t) is the load normalization value of resource i at time period t, calculated as CL(i, t) / ML(i); AR(i) is the adjustable ratio of resource i; LV(i, t) is the load volatility of resource i at time period t; ML(i) is the maximum historical load value of resource i; RF(i) is the response reliability factor of resource i; α is the adjustment coefficient, with a value of 0.8.

[0047] For example, the calculation process of resource R001 during the peak load period (t = 48 - 60): NL(R001, 54) = 180kW / 200kW = 0.9; AR(R001) = 0.25; LV(R001, 54) = 0.05; ML(R001) = 200kW; RF(R001) = 0.95; RP(R001, 54) = 0.8×0.9×0.25×(1 - 0.05)×0.95 = 0.162. Through similar calculations, the response potential indexes of all resources at 96 time points in 24 hours are obtained, forming the resource response potential characteristic matrix RPM.

[0048] 1.4. Construct a resource logical association graph.

[0049] Based on the comprehensively cleaned load data CL and the resource response potential characteristic matrix RPM, construct a resource logical association graph and calculate the importance weights of each resource: The importance weight IW(i) of resource i = β1·CAP(i) / ∑(j = 1 to n)CAP(j)+ β2·ARP(i) / ∑(j = 1 to n)ARP(j) + β3·HC(i) / MAX(HC); where: CAP(i) is the maximum adjustable capacity of resource i; ARP(i) is the average response potential of resource i, calculated as ∑(t = 1 to T)RP(i, t) / T; HC(i) is the hierarchical centrality of resource i, reflecting its position in the resource upper and lower level structure; β1, β2, β3 are weight coefficients, which are 0.4, 0.4, 0.2 respectively; n is the total number of resources; T is the total number of time points.

[0050] Calculation result example: IW(R001) = 0.4×(200 / 1500) + 0.4×(0.15 / 1.2) + 0.2×(0.8 / 1.0) = 0.16; According to the above calculations, generate the resource basic characterization data BD containing resource basic information, response potential characteristics, and importance weights.

[0051] Step 2: Construct a dynamic structure model of resource flow.

[0052] 2.1. Construct a directed graph of resource relationships.

[0053] Read the resource basic characterization data BD, and construct a directed graph of resource relationships based on resource time series correlation and operation relevance: The association strength CS(i, j) between resources i and j = γ1·TS(i, j) + γ2·OS(i, j) + γ3·HS(i, j); where TS(i, j) is the time series correlation, calculated as the Pearson correlation coefficient of the load curves of resources i and j; OS(i, j) is the operation relevance, and the value is the normalized value of the sequential operation frequency of resources i and j by the user; HS(i, j) is the hierarchical similarity, which is 1 when i and j are at the same level, and otherwise decreases according to the hierarchical difference; γ1, γ2, and γ3 are weight coefficients, which are 0.3, 0.5, and 0.2 respectively.

[0054] For example, the calculation of the association strength between R001 and R004: TS(R001, R004) = 0.75 (load curve correlation coefficient); OS(R001, R004) = 0.85 (operation relevance); HS(R001, R004) = 0.6 (with a difference of one level); CS(R001, R004) = 0.3×0.75 + 0.5×0.85 + 0.2×0.6 = 0.775. By calculating the association strength of all resource pairs, a main structure diagram MSG of resources is generated.

[0055] 2.2. Apply the time series mapping algorithm.

[0056] Apply the time series mapping algorithm to the main structure diagram MSG of resources, and map the resources to the time axis: The coordinate X(i) of resource i on the horizontal axis = Xmin + (RT(i)-RTmin) / (RTmax-RTmin)×(Xmax-Xmin); The coordinate Y(i) of resource i on the vertical axis = Ymin + HL(i) / HLmax×(Ymax-Ymin) + Δ(i); where RT(i) is the response start time of resource i, normalized to the interval [0, 1]; RTmin and RTmax are the earliest and latest response start times among all resources; Xmin and Xmax are the boundary values of the horizontal display area; HL(i) is the hierarchical value of resource i; HLmax is the maximum hierarchical value; Ymin and Ymax are the boundary values of the vertical display area; Δ(i) is a fine-tuning value used to avoid overlapping of resources at the same level.

[0057] For example, for resource R001, the calculation process is as follows: RT(R001) = 0.2 (normalized response start time); RTmin = 0, RTmax = 1; Xmin = 50, Xmax = 950; X(R001) = 50 + (0.2 - 0) / (1 - 0)×(950 - 50) = 230; HL(R001) = 1 (top-level resource); HLmax = 3; Ymin = 50, Ymax = 550; Y(R001) = 50 + 1 / 3×(550 - 50) + 5 = 221.67. After the initial calculation, the force-directed local adjustment algorithm is applied to solve the overlap problem: for overlapping resources i and j, calculate the repulsive force F(i, j) = k / (d(i, j)) 2 ; where d(i, j) is the Euclidean distance between resources i and j in the two-dimensional plane; k is the proportionality coefficient with a value of 1000; when d(i, j) < dmin, the repulsive force is applied to adjust the resource positions. Through multiple rounds of iterative adjustment, the time-sequenced resource structure TSS is finally generated.

[0058] 2.3. Calculate the resource association strength matrix.

[0059] Based on the time-sequenced resource structure TSS, calculate the association strength matrix between resource pairs: the association strength RC(i, j) between resources i and j = λ1·TOL(i, j) + λ2·RCC(i, j) + λ3·EMC(i, j); where TOL(i, j) is the time-sequence overlap degree, calculated as the overlap degree of the response periods of the two resources; RCC(i, j) is the response quantity correlation, calculated as the correlation coefficient of the response capacity changes of the two resources; EMC(i, j) is the regulation mutual exclusivity, which is 0 when the two resources cannot be regulated simultaneously, otherwise 1; λ1, λ2, and λ3 are weight coefficients, which are 0.4, 0.4, and 0.2 respectively.

[0060] For example, the calculation of the association strength between R001 and R002: TOL(R001, R002) = 0.6 (period overlap degree); RCC(R001, R002) = 0.2 (negative correlation); EMC(R001, R002) = 1 (can be regulated simultaneously); RC(R001, R002) = 0.4×0.6 + 0.4×0.2 + 0.2×1 = 0.48. By calculating all resource pairs, a 10×10 resource association strength matrix RCM is generated.

[0061] 2.4. Construct a resource flow model.

[0062] Combined with the time-series resource structure TSS and the resource association strength matrix RCM, an improved Sankey diagram algorithm is applied to construct a resource flow model: First, the association strength threshold filtering algorithm is applied to retain the connections of resource pairs with an association strength exceeding 0.3, forming a simplified association network SRN. Then, community detection based on the Louvain algorithm is applied to the simplified association network SRN: The community membership of resource i is optimized according to modularity Q, where Q = (1 / 2m)·∑(i,j)[RC(i,j) - k(i)·k(j) / (2m)]·Δ(c(i),c(j)); here, m is the sum of the weights of all edges in the network; k(i) is the sum of the weights of all edges connected to resource i; c(i) is the community to which resource i belongs; Δ(c(i),c(j)) is 1 when c(i)=c(j), otherwise 0. After calculation, 10 resources are divided into 3 functional clusters: Cluster 1: R001, R004, R005; Cluster 2: R002, R006, R007, R008; Cluster 3: R003, R009, R010.

[0063] Next, construct the initial structure of the directed flow graph, and apply the hierarchical layout strategy to arrange the resources in chronological order to generate the initial flow graph skeleton IFS. Then, apply the edge bundling optimization algorithm to the initial flow graph skeleton IFS: For a set of edges E with similar directions, generate a main path P, and the mapping strength MS(e,P) of edge e to the main path P = exp(-d(e,P) / σ); where d(e,P) is the average distance from edge e to the main path P; σ is a control parameter with a value of 0.5; after completion of the mapping, calculate the flow bundle width W(P) = ∑(e∈E)RC(e)·MS(e,P); RC(e) is the association strength of the resource pair represented by edge e. Through edge bundling optimization, the originally 18 independent connections are merged into 7 main flow bundles, generating an optimized flow connection structure OFS. Finally, apply the topological crossing minimization algorithm, and through node position fine-tuning and edge path optimization, reduce unnecessary streamline crossings: Calculate the crossing cost CC(e1,e2) of each pair of crossing edges (e1,e2) = RC(e1)·RC(e2); optimize according to the total crossing cost TCC = ∑(e1,e2∈E,e1 crosses e2)CC(e1,e2); through the simulated annealing algorithm, adjust the node positions while maintaining the temporal relationship to minimize TCC. After multiple rounds of optimization, the total crossing cost is reduced from the initial 2.46 to 0.72, generating the final resource flow topological structure RFTS.

[0064] Step 3: Implement real-time monitoring and evaluation of cognitive load.

[0065] 3.1. Capture real-time user interaction data.

[0066] The system captures the user's real-time interaction data on the interface, including mouse movement trajectories, dwell times, click frequencies, and drag operation characteristics. Sampling is performed every 100 ms to generate the original interaction data stream RID. Data example (within a 5-second sampling period): Operation trajectory coordinates: [(232, 145), (235, 148),..., (350, 220)]; Dwell points: [(235, 148, 0.5s), (350, 220, 1.2s)]; Click events: [(235, 148), (350, 220)]; Drag operations: [(235, 148) → (350, 220)].

[0067] 3.2. Extract operation characteristics.

[0068] Apply sliding window analysis to the original interaction data stream RID to extract operation characteristics: within the sliding window W, calculate the operation speed variation rate SVR = std(v) / mean(v); operation precision AP = 1 - ∑(d(ci, ti)) / MAX_D; repeat operation frequency ROF = Nr / Nt; where v is the speed vector of each sampling point within the window; std(v) is the standard deviation of the speed; mean(v) is the average speed; d(ci, ti) is the distance between the actual click position ci and the target position ti; MAX_D is the normalization parameter; Nr is the number of repeated operations; Nt is the total number of operations.

[0069] For example, the calculation within a certain 5-second window: Mouse speed sequence: [5.2, 6.1, 4.8, 7.2, 3.9, 8.5, 4.2] pixels / 100ms; Average speed: 5.7 pixels / 100ms; Speed standard deviation: 1.67 pixels / 100ms; SVR = 1.67 / 5.7 = 0.293; Click accuracy: The average deviation between the target position and the actual click position is 8 pixels, MAX_D is 50; AP = 1 - 8 / 50 = 0.84; There is 1 repeated operation out of 5 operations; ROF = 1 / 5 = 0.2. By extracting the characteristics of each window, an operation feature vector OFV is generated.

[0070] 3.3. Apply the cognitive load dynamic assessment model.

[0071] Based on the operation feature vector OFV, apply the cognitive load dynamic assessment model to calculate the user's cognitive load index: Decompose the operation feature vector OFV into three sub-vectors: operation trajectory feature MTF, operation time feature OTF, and operation frequency feature OFF, to generate a classification feature set CFS. Apply the curvature analysis algorithm to the operation trajectory feature MTF: Trajectory curvature K(i) = |v(i) × a(i)| / |v(i)| 3; The trajectory smoothness TS = 1 - avg(K) / (K_max); the direction change frequency DCF = Nd / L; the acceleration volatility AV = std(|a|) / avg(|a|), where v(i) and a(i) are the velocity and acceleration vectors at point i respectively; avg(K) is the average curvature; K_max is the maximum allowable curvature value; Nd is the number of direction changes; L is the total length of the trajectory; std(|a|) is the standard deviation of the acceleration magnitude; avg(|a|) is the average acceleration magnitude. Calculate the trajectory uncertainty index TUI = (1 - TS)×0.4 + DCF×0.4 + AV×0.2. For example, the analysis results of a certain section of the trajectory: TS = 0.65 (trajectory smoothness); DCF = 0.28 (direction change frequency); AV = 0.42 (acceleration volatility); TUI = (1 - 0.65)×0.4 + 0.28×0.4 + 0.42×0.2 = 0.28.

[0072] Analyze the operation time feature OTF: Identify the operation pause points and acceleration points through the peak detection algorithm; the operation rhythm stability ORS = 1 - std(ΔT) / avg(ΔT); the attention dispersion index ADI = (1 - ORS)×0.6 + SP×0.4, where ΔT is the time interval between adjacent operations; std(ΔT) is the standard deviation of the time interval; avg(ΔT) is the average time interval; SP is the pause point frequency, calculated as the ratio of the number of pause points to the total number of operation points. For example, the analysis results of a certain section of the operation: ORS = 0.72 (operation rhythm stability); SP = 0.25 (pause point frequency); ADI = (1 - 0.72)×0.6 + 0.25×0.4 = 0.268.

[0073] Analyze the operation frequency feature OFF: Identify the repeated operation sequences and error - correction patterns, calculate the operation redundancy OR = (Ntotal - Neffective) / Ntotal; the operation efficiency index OEI = OR×0.7 + EC×0.3, where Ntotal is the total number of operations; Neffective is the number of effective operations; EC is the frequency of the error - correction pattern. For example, the analysis results of a certain section of the operation: OR = 0.32 (operation redundancy); EC = 0.15 (error - correction frequency); OEI = 0.32×0.7 + 0.15×0.3 = 0.269.

[0074] Combining three indicators, a cognitive load assessment function is constructed: Mental Load ML = TUI×0.4 + ADI×0.6; Time Pressure TP = OEI×0.3 + FCR×0.7; Frustration Degree FD = OEI×0.5 + UIA×0.5; Comprehensive Cognitive Load CL = ML×w1 + TP×w2 + FD×w3; where FCR is the operation frequency change rate; UIA is the interface adjustment attempt frequency; w1, w2, w3 are weight coefficients, which are 0.4, 0.3, and 0.3 respectively. For example, the calculation of the cognitive load at a certain moment: ML = 0.28×0.4 + 0.268×0.6 = 0.273; TP = 0.269×0.3 + 0.35×0.7 = 0.326; FD = 0.269×0.5 + 0.2×0.5 = 0.235; CL = 0.273×0.4 + 0.326×0.3 + 0.235×0.3 = 0.277.

[0075] The calculated cognitive load value is smoothed using the exponentially weighted moving average algorithm: Smoothed Cognitive Load SCL(t) = α·CL(t) + (1-α)·SCL(t-1); where CL(t) is the cognitive load value at time t; SCL(t-1) is the smoothed cognitive load value at time t-1; α is the smoothing coefficient, with a value of 0.3. The continuous cognitive load values are mapped to discrete cognitive load levels using a multi-threshold piecewise function: When SCL < 0.3, the cognitive load level is low; when 0.3 ≤ SCL < 0.5, the cognitive load level is medium; when 0.5 ≤ SCL < 0.7, the cognitive load level is high; when SCL ≥ 0.7, the cognitive load level is extremely high; at the same time, the trend T(t) = sign(SCL(t)-SCL(t-n)) is calculated; where n is the trend window size, with a value of 5. After calculation, the system generates a user cognitive load index UCI = {level: "medium", trend: "rising"} that includes the cognitive load level and trend.

[0076] Step Four: Perform adaptive adjustment of interface elements.

[0077] 4.1. Determine the priority of interface parameters to be adjusted.

[0078] Receive the user cognitive load index UCI and determine the priority of interface parameters to be adjusted: According to the cognitive load level and trend, calculate the adjustment priority of parameter pi as PP(pi) = BW(pi)×(CL_weight(level) + 0.2×trend_value(trend)); where BW(pi) is the basic weight of parameter pi; CL_weight(level) is the weight coefficient corresponding to different cognitive load levels; trend_value(trend) is the trend value, 1 for rising, 0 for stable, and -1 for falling. For the cognitive load of "medium" level and the trend of "rising", calculate the priority of each parameter: Information density (ID): PP(ID) = 0.4×(0.6 + 0.2×1) = 0.32; Visual distinctiveness (VD): PP(VD) = 0.3×(0.6 + 0.2×1) = 0.24; Interaction assistance strength (IAS): PP(IAS) = 0.2×(0.6 + 0.2×1) = 0.16; Animation smoothness (AS): PP(AS) = 0.1×(0.6 + 0.2×1) = 0.08. According to the calculation results, form the adjustment parameter priority list PPL = [{param: "ID", priority: 0.32}, {param: "VD", priority: 0.24},...].

[0079] 4.2. Calculate the target values of interface parameters.

[0080] According to the adjustment parameter priority list PPL and the preset interface parameter adjustment rules, calculate the target values of interface parameters: The target value of parameter pi is TV(pi) = CV(pi) + ΔV(pi)×CF(level); where CV(pi) is the current value of parameter pi; ΔV(pi) is the basic adjustment amount; CF(level) is the adjustment coefficient corresponding to the cognitive load level, 0.5 for low level, 1 for medium level, 1.5 for high level, and 2 for ultra-high level. For the cognitive load of "medium" level, calculate the target values of each parameter: Information density: TV(ID) = 0.75 - 0.15×1 = 0.6; Visual distinctiveness: TV(VD) = 0.6 + 0.15×1 = 0.75; Interaction assistance strength: TV(IAS) = 0.4 + 0.2×1 = 0.6; Animation smoothness: TV(AS) = 0.7 - 0.1×1 = 0.6; Generate the interface parameter target value set TV = {ID: 0.6, VD: 0.75, IAS: 0.6, AS: 0.6}.

[0081] 4.3. Adopt the progressive parameter adjustment algorithm.

[0082] Adopt a progressive parameter adjustment algorithm to avoid cognitive interference caused by sudden interface changes: Read the target value TV of the interface parameters and the current interface parameter status CV, calculate the parameter difference vector DV = TV - CV, and generate the parameter adjustment requirement vector ARV = {ID: -0.15, VD: 0.15, IAS: 0.2, AS: -0.1}. Apply cognitive sensitivity analysis to the parameter adjustment requirement vector ARV, and assign weights according to the influence degree of different parameters on the cognitive load: The weighted adjustment requirement WARV(pi) of parameter pi = ARV(pi) × CS(pi); where CS(pi) is the cognitive sensitivity of parameter pi, the information density is 1.2, the visual discrimination is 1.0, the interaction assistance intensity is 0.8, and the animation smoothness is 0.6. Calculation results: WARV(ID) = -0.15 × 1.2 = -0.18; WARV(VD) = 0.15 × 1.0 = 0.15; WARV(IAS) = 0.2 × 0.8 = 0.16; WARV(AS) = -0.1 × 0.6 = -0.06. Generate the weighted adjustment requirement vector WARV = {ID: -0.18, VD: 0.15, IAS: 0.16, AS: -0.06}.

[0083] According to the weighted adjusted requirement vector WARV and the current user cognitive load index UCI, construct a parameter adjustment rate function: the adjustment rate AR(pi) of parameter pi = BASE_RATE×(1 - ω·CL); where BASE_RATE is the base adjustment rate, with a value of 0.1; CL is the current cognitive load value; ω is the weight coefficient, taking the value of 0.8 when the cognitive load is on the rise, 0.5 when stable, and 0.3 when falling. For the case where the cognitive load value is 0.277 and the trend is rising: AR(ID) = 0.1×(1 - 0.8×0.277) = 0.078; AR(VD) = AR(IAS) = AR(AS) = 0.078; generate an adaptive adjustment rate table ART = {ID: 0.078, VD: 0.078, IAS: 0.078, AS: 0.078}. Combining the weighted adjusted requirement vector WARV and the adaptive adjustment rate table ART, calculate the increment value of each parameter in the next frame: the increment value IV(pi) of parameter pi = sign(WARV(pi))·min(|WARV(pi)|, AR(pi)); where sign(x) is the sign function, which is 1 when x > 0, -1 when x < 0, and 0 when x = 0; min(a, b) returns the smaller value of a and b. Calculation results: IV(ID) = -1·min(0.18, 0.078) = -0.078; IV(VD) = 1·min(0.15, 0.078) = 0.078; IV(IAS) = 1·min(0.16, 0.078) = 0.078; IV(AS) = -1·min(0.06, 0.078) = -0.06; generate a parameter increment vector IV = {ID: -0.078, VD: 0.078, IAS: 0.078, AS: -0.06}.

[0084] Add the current parameter values to the parameter increment vector IV to generate new interface parameter values: New information density = 0.75 + (-0.078) = 0.672; New visual distinctiveness = 0.6 + 0.078 = 0.678; New interaction assistance intensity = 0.4 + 0.078 = 0.478; New animation smoothness = 0.7 + (-0.06) = 0.64; Generate the updated parameter values UPV = {ID: 0.672, VD: 0.678, IAS: 0.478, AS: 0.64}. Apply a parameter coordination check to the updated parameter values UPV to ensure that the proportional relationships between relevant parameters comply with the interface design specifications: Check whether the proportional relationship PR(pi, pj) = pi / pj is within the preset range [PRmin(pi, pj), PRmax(pi, pj)]; where PRmin(pi, pj) and PRmax(pi, pj) are the minimum and maximum allowable proportional values between parameters pi and pj. For example, the proportional relationship check between information density and visual distinctiveness: PR(ID, VD) = 0.672 / 0.678 = 0.991; PRmin(ID, VD) = 0.8, PRmax(ID, VD) = 1.2; 0.8 < 0.991 < 1.2, which meets the requirements. After comprehensive inspection, all parameter relationships comply with the design specifications, and generate the interface parameter configuration IPC = {ID: 0.672, VD: 0.678, IAS: 0.478, AS: 0.64}.

[0085] Step Five: Optimize the dynamic layout of resources.

[0086] 5.1. Construct the layout optimization problem.

[0087] Integrate the comprehensive resource flow topology structure RFTS, the resource association strength matrix RCM, and the interface parameter configuration IPC to construct a multi-objective layout optimization problem: construct a layout evaluation function that includes a cognitive load evaluation sub-function, a visual aesthetics evaluation sub-function, and an operation efficiency evaluation sub-function: calculate the cognitive load score of the layout based on information theory CLS = (1 - VE / VE_max)×w1 + UDD×w2 + (1 - VGC)×w3; where VE is the visual entropy, calculated as -∑p(i)·log2(p(i)), p(i) is the visual element density of region i; VE_max is the maximum allowable visual entropy; UDD is the non-uniformity of the element density distribution; VGC is the clarity of the visual guidance path; w1, w2, w3 are weight coefficients, which are 0.4, 0.3, 0.3 respectively. For example, calculate the cognitive load score of a certain layout configuration: VE = 3.2 (visual entropy); VE_max = 4.5; UDD = 0.25 (non-uniformity); VGC = 0.7 (clarity); CLS = (1 - 3.2 / 4.5)×0.4 + 0.25×0.3 + (1 - 0.7)×0.3 = 0.286.

[0088] Evaluate the visual aesthetics score of the layout based on the principle of computational aesthetics VAS = BAL×w1 + SYM×w2 + REG×w3 + RHY×w4; where BAL is the balance of the layout; SYM is the symmetry; REG is the regularity; RHY is the rhythm; w1, w2, w3, w4 are weight coefficients, which are 0.3, 0.2, 0.3, 0.2 respectively. For example, calculate the visual aesthetics score of a certain layout configuration: BAL = 0.75 (balance); SYM = 0.6 (symmetry); REG = 0.8 (regularity); RHY = 0.7 (rhythm); VAS = 0.75×0.3 + 0.6×0.2 + 0.8×0.3 + 0.7×0.2 = 0.725.

[0089] Based on Fitts' law and operation path analysis, the operation efficiency score OES of the layout is evaluated as: OES = (1 - APL / APL_max) × w1 + (1 - PC) × w2 + TA × w3; where APL is the average operation path length; APL_max is the maximum allowable path length; PC is the path complexity; TA is the target reachability; w1, w2, and w3 are weight coefficients, which are 0.4, 0.3, and 0.3 respectively. For example, the calculation of the operation efficiency score for a certain layout configuration: APL = 320 (pixels); APL_max = 800 (pixels); PC = 0.3 (complexity); TA = 0.85 (reachability); OES = (1 - 320 / 800) × 0.4 + (1 - 0.3) × 0.3 + 0.85 × 0.3 = 0.395.

[0090] Based on the current user cognitive load index UCI, the weights of the three sub-functions in the multi-objective evaluation function are dynamically adjusted: when the cognitive load is at the "low" level, wCLS = 0.2, wVAS = 0.5, wOES = 0.3; when the cognitive load is at the "medium" level, wCLS = 0.4, wVAS = 0.3, wOES = 0.3; when the cognitive load is at the "high" level, wCLS = 0.6, wVAS = 0.2, wOES = 0.2; when the cognitive load is at the "extremely high" level, wCLS = 0.7, wVAS = 0.1, wOES = 0.2. For the current "medium" level of cognitive load, an adaptive weighted evaluation function AWEF = 0.4 × CLS + 0.3 × VAS + 0.3 × OES is generated. Applying this function to the current layout, the score is obtained: total score = 0.4 × 0.286 + 0.3 × 0.725 + 0.3 × 0.395 = 0.4438 Combining the topological constraints in the resource flow topology structure RFTS and the display parameters in the interface parameter configuration IPC, a set of layout constraint conditions is constructed: the minimum distance constraint between nodes MIN_DIST(i, j) ≥ α·(size(i) + size(j)); the non-crossing preference constraint of streamlines NCL(e1, e2) = w·RC(e1)·RC(e2); the cohesion constraint of resource clusters COH(C) = ∑(i, j∈C)dist(i, j) / |C| 2≤β; where size(i) is the display size of node i; α is the spacing coefficient with a value of 0.8; NCL(e1, e2) is the penalty value for the crossing of edges e1 and e2; w is the weight with a value of 10; COH(C) is the cohesion metric of cluster C; β is the maximum allowable cohesion. By integrating the comprehensive adaptive weighted evaluation function AWEF and the layout constraint model LCM, a complete multi-objective optimization problem description is constructed, including decision variables (node coordinates), objective function (maximizing the comprehensive score), and constraint conditions, to generate the layout optimization model LOM.

[0091] 5.2. Apply the improved simulated annealing algorithm.

[0092] Apply the improved simulated annealing algorithm to the layout optimization model LOM to optimize the positions of resource nodes: the initial temperature T0 = 100, the cooling coefficient c = 0.95, and the termination temperature Tf = 0.1; at each temperature T, perform n = 50 random perturbations on the layout and calculate the score change ΔS; if ΔS > 0, accept the new layout; if ΔS < 0, accept the new layout with a probability of exp(ΔS / T); update the temperature T = c·T. During the execution process, as the temperature decreases, the algorithm gradually converges to the local optimal solution. Finally, through multiple runs with different initializations, select the optimal result to generate the optimized coordinates of resource nodes ONOC. For example, for the position optimization of node R001: the initial coordinates: (230, 221.67); the final optimized coordinates: (235.8, 205.4).

[0093] 5.3. Integrate the visual description objects.

[0094] Integrate the optimized coordinates of resource nodes ONOC and the visual performance attributes into complete visual description objects: for each resource node i, construct the visual description VD(i) = {position: ONOC(i), size: size(i), color: color(i), opacity: opacity(i),...}; for each flow bundle j, construct the visual description VD(j) = {path: path(j), width: width(j), color: color(j),...}; where size(i) is determined based on the resource importance weight IW(i) and the information density in the interface parameter configuration IPC; color(i) is determined based on the resource type and visual distinctiveness; width(j) is determined based on the resource association strength. Finally, generate the resource flow visualization layout RFVL, which contains the complete visual descriptions of 10 resource nodes and 7 main flow bundles, realizing the visual display and interactive response on the interface.

[0095] Compared with the prior art, this embodiment shows obvious advantages in the following aspects: reduced cognitive load: in the test scenario where 20 resources are orchestrated simultaneously, the average cognitive load score of users decreased from 0.68 to 0.41, a reduction of approximately 40%; improved operation efficiency: the average time required to complete the same orchestration task decreased from 185 seconds to 115 seconds, an efficiency improvement of approximately 38%; reduced error rate: the operation error rate of users during the strategy orchestration process decreased from 18% to 6%, a reduction of approximately 67%; improved user satisfaction: according to the subjective user rating (on a scale of 1 - 10), the satisfaction level increased from an average of 5.8 points to 8.2 points, an increase of approximately 41%. From the above comparison data, it can be seen that this embodiment effectively solves problems such as difficult visual recognition and low operation accuracy faced by users during the demand response resource orchestration process through dynamic cognitive load assessment and interface adaptive adjustment, improving the usability of the system and the user experience.

[0096] This embodiment details a method for dynamically evolving the analysis of user - side demand response strategies applicable to graphical representation. Through key technologies such as constructing a dynamic resource flow structure model, real - time monitoring and evaluation of cognitive load, adaptive adjustment of interface elements, and optimizing the dynamic layout of resources, it realizes the precise perception of user cognitive load and the intelligent adjustment of the interface, effectively solving the cognitive and operation challenges during the orchestration of a large number of resources. The implementation results show that this embodiment can reduce the user's cognitive load, improve operation efficiency, reduce the error rate, and enhance user satisfaction, providing effective technical support for the visual orchestration of user - side demand response strategies.

[0097] According to another aspect of the present application, a method for dynamically evolving the analysis of user - side demand response strategies applicable to graphical representation includes: Measuring the demand response potential: obtaining the historical data of demand response resources in the past 30 days and performing data pre - processing; calculating the response potential of the resources through an algorithm; Pre - grouping and pre - sorting of resources: users pre - group the demand response resources according to production characteristics, and through an algorithm, a recommended sorting of the control priorities of the resources is provided. Users then perform a secondary precise sorting; Orchestrating the execution strategy: the model provides resources that meet the requirements for interface manifestation, and then adjusts the participation time period and response volume of the resources through graphical means such as dragging.

[0098] Measuring the estimated response load and compensation amount: during the orchestration process, the server calculates the estimated response load of the current demand response based on the participation time and response volume of the resources, and calculates the estimated compensation amount according to the algorithm in the demand response implementation rules.

[0099] Persistent execution resources: When the user determines the participating resources, the model will verify the superior-subordinate relationship of the resource files for validity verification. After passing the verification, the regulation amount, duration, overlap order, and axis information of each resource will be persisted for echo display when the user views them; Calculate the baseline: For same-day and real-time demand response, calculate the baseline of the execution resources after persistence.

[0100] Specifically, obtain all the resource file information of the user from the database, loop through the files. For non-regulatable resources, directly set the response potential to 0; obtain the 96-point load curve data of the demand response resources in the past 30 days, and fit the daily curve data into a single-day average; reverse the single-day average of a single demand response resource, remove invalid data such as null values, and retain the top 7 largest averages. Then remove the largest two values to exclude abnormal electricity consumption situations. If there are less than 4 days, no removal is done, and the remaining daily averages are averaged again to be used as the potential base of the resources; process the potential base according to the adjustment method of the resources. If the method is adjustable, the resource potential = potential base * adjustment ratio. If it is non-adjustable, the resource potential = potential base.

[0101] Pre-group and pre-sort the demand response resources to generate a regulation priority grouping. Provide filtering of demand response resources according to specific resource characteristics, and then incorporate the filtered resources into the grouping; provide three pre-sorting methods: fewer participating resources first, fewer participation times first, and larger current load first for the user to perform regulation priority sorting; Fewer participating resources first: The model sorts according to the calculated resource potential from large to small. Fewer participation times first: The model counts the number of times each resource participates in the demand response in the current year and sorts from few to many. Larger current load first: The model obtains the current load of each resource and sorts from large to small. The sorting recommended by the model often cannot fully meet the user's needs due to the imperfection of the dataset and other characteristics of the resources. At this time, the user can drag the resources based on the recommended sorting results for secondary precise sorting to generate a regulation priority grouping.

[0102] Perform strategy orchestration. The model filters the resources based on the advance notice time of the demand response event and in combination with the load ramp-up duration of the resources. For example, if it takes 2 hours for a certain device's load to reach the target regulation amount, and this demand response event is notified 1 hour in advance, then this device does not have the regulation potential. The user adjusts the participation time period and response amount of the resources through graphical means such as dragging.

[0103] Calculate the estimated response load and the estimated compensation amount in real time. Calculate the average of the potential of the participating resources at each time point during the current event period to obtain the estimated response load; obtain the current power generation plan, add it to the estimated response load to get the event response volume for this time; then correspond the event demand volume and the event response volume to the scope of the local demand response implementation rules to obtain the baseline price, the recognized response volume, and the response volume price coefficient; the estimated compensation amount = event response volume * baseline price * recognized response volume * response volume price coefficient.

[0104] Perform validity verification on the execution resources and persist them. For the execution strategy orchestrated by the user, based on the security principle that the client is untrusted, the model will perform validity verification on the execution information of the resources in the strategy. The verification content includes: the integrity and non-repeatability of the execution information; integrity includes: the execution start time, end time, expected response volume of the resources, as well as the graphical coordinates and the overlapping sequence numbers. Non-repeatability includes: the non-repeatability of the resources themselves and the non-repeatability of the superior-subordinate relationships. Those that pass the verification are persisted into the database for the user to view again.

[0105] Calculate the baseline. Select the effective typical days for the baseline: obtain the configuration of the number of typical days. If the configuration is empty, if the current day is a weekend, one typical day is selected by default; if the current day is a working day, five typical days are selected by default. Select the corresponding weekends or working days in the previous 30 days, excluding the event day, until the required number of days is selected; calculate the resource baseline: query the execution resource information, loop through the resources, calculate the baseline of a single resource, loop through the typical days, obtain the resource load curve, assemble the load curves into a two-dimensional array, and then average and fit them into a single-day curve, and persist the curve into the database; calculate the total user baseline: loop through the typical days, obtain the user load curve, assemble the load curves into a two-dimensional array, and then average and fit them into a single-day curve, and persist the curve into the database.

[0106] The present invention is particularly applicable to the situation where the number of demand response resources is relatively large (>20), and can effectively solve the problem of excessive cognitive load faced by users when performing resource drag-and-drop orchestration. It realizes a paradigm shift from passive adaptation to active perception. In traditional methods, the interface is a passive information presentation tool, while the interface of the present invention becomes an interactive system that can actively perceive the user's cognitive state, predict the user's intention, and provide intelligent assistance. The system does not simply wait for the user to complete the operation and then make a response, but can monitor the change of the user's cognitive load in real time during the operation process, prevent cognitive overload through progressive parameter adjustment, and even predict the possible next operation and provide assistance. It transforms the static space concept into a time-sequence-based flow structure, upgrades the independent focus adjustment mechanism to a globally adaptive system based on cognitive load, and introduces multimodal interaction to reduce the visual cognitive burden. This integration is not a simple superposition, but a systematic reconstruction under a new conceptual model, forming an organic whole, which jointly solves the complex cognitive challenges in demand response resource orchestration.

[0107] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all belong to the protection scope of the present invention.

Claims

1. A method for dynamically evolving analysis of graphical user-side demand response strategies, characterized in that Including: Obtain the original data of demand response resources and construct an initial resource model, generate resource basic characterization data, and based on this, construct a dynamic resource flow structure model, generate a resource flow topological structure and a resource association strength matrix; Realize real-time monitoring and evaluation of cognitive load according to user real-time interaction data, calculate user cognitive load indicators and perform adaptive adjustment of interface elements based on them, and generate interface parameter configurations; Optimize the dynamic layout of resource flow based on the resource flow topological structure, resource association strength matrix and interface parameter configuration, and generate a visual layout of resource flow.

2. The method according to claim 1, wherein The steps of constructing an initial resource model and generating resource basic characterization data include: Read the original data of demand response resources, including resource identifiers, resource types, historical load curves, adjustment ranges, response durations, and superior-subordinate relationships. After standardization and cleaning, obtain the cleaned load data; calculate the response potential indicators of each resource in different time periods based on this, and generate a resource response potential characteristic matrix; Integrate the cleaned load data and the resource response potential characteristic matrix, construct a resource logical association graph, calculate the importance weights of each resource, and generate resource basic characterization data.

3. The method according to claim 2, characterized in that, The steps of generating a resource response potential characteristic matrix include: Divide the cleaned load data into typical daily load patterns, classify them based on date types and seasonal characteristics, and generate a classified load data set; statistically analyze the load characteristics based on this and extract basic response potential indicators through a basic response potential evaluation model; Conduct time sensitivity analysis on the basic response potential indicators, calculate the response potential changes of resources in different time periods, and generate a time-varying response potential curve; Analyze the actual response performance of resources in historical demand response events, calculate the deviation rate between the expected response and the actual response, and generate a response reliability indicator; Integrate the time-varying response potential curve and the response reliability indicator, construct a multi-dimensional response potential evaluation matrix, and generate a resource response potential characteristic matrix.

4. The method according to claim 1, characterized in that, The steps of generating a resource flow topological structure and a resource association strength matrix include: Read the resource basic characterization data, construct a directed graph of resource relationships based on resource time series correlation and operation relevance, and generate a resource backbone structure diagram; perform time series mapping on it, and map resources to the time axis according to resource time attributes to generate a time-ordered resource structure; Based on the time-ordered resource structure, calculate the association strength between resource pairs, and generate a resource association strength matrix; combine it with the time-ordered resource structure, and apply an improved Sankey diagram algorithm to construct a resource flow model, and generate a resource flow topological structure.

5. The method according to claim 4, wherein The steps of applying an improved Sankey diagram algorithm to construct a resource flow model and generate a resource flow topological structure include: Read the time-ordered resource structure and the resource association strength matrix, perform association strength threshold filtering and weighted network modularity analysis, and generate resource function clusters; Combine the resource function clusters and the resource time attributes, construct an initial structure of a directed flow graph, arrange resources in chronological order and maintain the internal aggregation of clusters, and generate an initial flow graph skeleton; Merge the connections with parallel orientations and similar targets in the initial flow graph skeleton into weighted flow bundles to generate an optimized flow connection structure; combine it with the resource association strength matrix, calculate the flow bundle width, generate a visually weighted flow graph, and perform fine-tuning of node positions and edge path optimization to obtain the resource flow topology structure.

6. The method according to claim 4, characterized in that The steps for performing temporal mapping to generate a temporalized resource structure include: Read the resource backbone structure diagram and resource time attribute data, construct a resource time axis coordinate system, and generate a time axis mapping framework; normalize the time attributes of all resources therein to generate normalized time attributes; Based on the normalized time attributes, calculate the horizontal coordinate values to generate the initial horizontal positions of the resources; Combine the hierarchical relationship and functional classification information of the resources to calculate the vertical coordinate values and generate the initial vertical positions of the resources; Combine the initial horizontal and vertical positions of the resources to construct an initial two-dimensional layout, solve the resource overlap problem through force-directed local adjustment, generate the adjusted resource positions, and maintain time correlation to obtain the temporalized resource structure.

7. The method according to claim 1, wherein The steps for calculating the user cognitive load index include: Capture the real-time interaction data of the user on the interface, including the mouse movement trajectory, dwell time, click frequency, and drag operation characteristics, to generate the original interaction data stream; Apply sliding window analysis to the original interaction data stream, extract the operation speed change rate, operation accuracy, and repeated operation frequency to generate an operation feature vector; Based on the operation feature vector, perform real-time cognitive load assessment to generate the user cognitive load index.

8. The method according to claim 7, characterized in that, The steps for performing real-time cognitive load assessment to generate the user cognitive load index include: Decompose the operation feature vector into operation trajectory, operation time, and operation frequency features; analyze the operation trajectory features, calculate the trajectory smoothness, direction change frequency, and acceleration volatility to generate a trajectory uncertainty index; Analyze the operation time features, identify the operation pause points and acceleration points, calculate the operation rhythm stability, and generate an attention dispersion index; Analyze the operation frequency features, identify the repeated operation sequences and error-correction patterns, calculate the operation redundancy, and generate an operation efficiency index; Integrate the trajectory uncertainty, attention dispersion, and operation efficiency indexes, perform non-linear weighted mapping and temporal smoothing processing according to mental load, time pressure, and frustration level, generate a smoothed cognitive load curve and map it to discrete cognitive load levels, and at the same time calculate the current load trend to form the user cognitive load index.

9. The method according to claim 1, wherein The steps for performing adaptive adjustment of interface elements to generate interface parameter configurations include: Receive the user cognitive load index, determine the priority of the interface parameters to be adjusted to form a list of adjustment parameter priorities; combine it with the preset interface parameter adjustment rules to calculate the target values of the interface parameters, including information density, visual discrimination, interaction assistance intensity, and animation smoothness; Based on the target values of the interface parameters, adopt a progressive parameter adjustment method to generate the interface parameter configuration.

10. The method according to claim 9, wherein, The steps for adopting a progressive parameter adjustment method to generate the interface parameter configuration include: Read the target values of the interface parameters and the current interface parameter status, calculate the parameter differences to generate a parameter adjustment requirement vector; assign weights to it according to the influence degree of different parameters on the cognitive load to generate a weighted adjustment requirement vector; Generate an adaptive adjustment rate table according to the weighted adjustment demand vector and the user cognitive load index; and combine it with the weighted adjustment demand vector to calculate the parameter increment value; Add the current parameter value to the parameter increment value, and perform parameter constraint and parameter coordination check to generate the interface parameter configuration.

Images