A power grid load visualization analysis method and system based on elastic network regression

Abstract

The application provides a power grid load visualization analysis method and system based on elastic network regression, relates to the technical field of data interaction analysis, iteratively fuses historical power grid load data and multi-source fluctuation information according to a constraint strength sequence adapted to power grid load fluctuation, obtains a coefficient evolution track of load change sensitivity when the power grid load fluctuates, determines a fluctuation response order of the power grid load when a load peak fluctuation is triggered according to a non-zero jump node of the coefficient, performs layer rendering on the coefficient evolution track according to the fluctuation response order, and generates a load tracing view by superimposing a load contribution distribution in the multi-source fluctuation information, dynamically updates a display state of the load tracing view in response to an interaction instruction triggered by a graph display interface, and obtains a tracing feature graph of the power grid load fluctuation. The application can realize collaborative construction of coefficient evolution time sequence analysis and layered interactive visualization, so as to improve the accuracy of power grid load fluctuation tracing analysis.

Description

Technical Field

[0001] This application relates to the field of data interaction analysis technology, and more specifically, to a method and system for power grid load visualization analysis based on elastic network regression. Background Technology

[0002] Data interaction analysis is an interactive technology that supports human-machine collaborative judgment in power grid load visualization analysis. Based on the results of power grid load big data analysis, this technology relies on a visualization interface to carry load source analysis results generated based on elastic network regression. By receiving user interactive operation commands, it dynamically adjusts the display dimensions, analysis granularity and focus area of ​​the map, supports operations such as filtering fluctuation factors, locating peak periods, and switching contribution dimensions, and transforms static regression analysis results into a dynamic analysis carrier that can be adjusted as needed, serving the load characteristic judgment and source tracing work in power grid operation and management.

[0003] However, existing power grid load visualization analysis based on elastic network regression lacks a full-link collaborative mechanism. This leads to a disconnect between the load sensitivity calculation results and the cascading transmission timeline of fluctuation factors. Consequently, static regression coefficients cannot fully represent the evolutionary pattern of fluctuation impacts, and the source tracing view lacks time-series hierarchical expression and directional focusing capabilities. Therefore, it is difficult to simultaneously ensure the time-series accuracy of load cause tracing and the interactive flexibility of visualization analysis. Thus, how to achieve the collaborative construction of coefficient evolution time-series analysis and hierarchical interactive visualization to improve the accuracy of power grid load fluctuation source tracing analysis is a challenge facing the industry. Summary of the Invention

[0004] This application provides a method and system for visual analysis of power grid load based on elastic network regression, which can realize the collaborative construction of coefficient evolution time series analysis and hierarchical interactive visualization to improve the accuracy of power grid load fluctuation source analysis.

[0005] Firstly, this application provides a power grid load visualization analysis method based on elastic network regression, the analysis method comprising the following steps: Acquire historical power grid load data for the target area and collect multi-source fluctuation information during power grid load fluctuations; The historical power grid load data and the multi-source fluctuation information are iteratively fused according to the constraint strength sequence of power grid load fluctuation adaptation to obtain the coefficient evolution trajectory of load change sensitivity during power grid load fluctuation. Identify the non-zero coefficient jump nodes in the coefficient evolution trajectory, determine the fluctuation response order of the grid load when triggering load peak fluctuations based on the non-zero coefficient jump nodes, perform layer rendering on the coefficient evolution trajectory based on the fluctuation response order, and overlay the load contribution distribution in the multi-source fluctuation information to generate a load source tracing view. In response to the interactive command triggered by the graph display interface, the display status of the load tracing view is dynamically updated to obtain the tracing feature graph of power grid load fluctuation.

[0006] In this embodiment, the historical power grid load data and the multi-source fluctuation information are iteratively fused according to the constraint strength sequence of power grid load fluctuation adaptation to obtain the coefficient evolution trajectory of load change sensitivity during power grid load fluctuations. Specifically, this includes: Based on the historical power grid load data and the multi-source fluctuation information, a load-driven feature matrix is ​​constructed, and the initial sensitivity coefficient is obtained by regularizing the solution using the smoothing contraction term of the constraint strength sequence. By using the initial sensitivity coefficient and the non-smooth sparse constraints of the constraint strength sequence, redundant factors in the source fluctuation information are iteratively eliminated to generate a sparsified coefficient path. Based on the sparsified coefficient path and the power flow matrix, the node response at load peak is evolved and mapped to obtain the coefficient evolution trajectory of the load change sensitivity during power grid load fluctuations.

[0007] In this embodiment, constructing the load-driven feature matrix based on the historical power grid load data and the multi-source fluctuation information specifically includes: Based on the historical power grid load data and the multi-source fluctuation information, big data time series features are extracted to construct an initial high-dimensional feature tensor for dynamic correlation of power supply and demand. A load-driven feature matrix is ​​generated by cross-domain data fusion mapping using the initial high-dimensional feature tensor and the external disturbance factor in the multi-source fluctuation information.

[0008] In this embodiment, the iterative removal of redundant factors in the source fluctuation information to generate a sparsity coefficient path specifically includes: Based on the multi-source fluctuation information and the multi-dimensional feature correlation evaluation model in big data analysis, the redundant factors are subjected to collinearity dimensionality reduction to obtain a subset of fluctuation features. The sparsified coefficient path is generated by iterative regularization of the fluctuation feature subset and the non-smooth sparse constraints of the constraint strength sequence.

[0009] In this embodiment, identifying the non-zero coefficient transition nodes in the coefficient evolution trajectory specifically includes: Based on the coefficient evolution trajectory, perform big data time-series differential operations and zero-value interval marking to determine the candidate windows for jumps corresponding to the coefficient change rate sequence and the duration of zero value; The non-zero jump amplitude of the coefficient change rate sequence is quantified by the jump candidate window and the adaptive mutation detection model to obtain the jump confidence score of the candidate node. Target nodes that meet the preset significance threshold are selected from the jump confidence scores, and load disturbance event correlation verification is performed based on the historical power grid load data to generate jump nodes with non-zero coefficients.

[0010] In this embodiment, determining the order of power grid load fluctuation response when triggering load peak fluctuations based on the non-zero coefficient jump node specifically includes: Based on the increment of the jump amplitude of the non-zero jump node, the time delay disturbance intensity of the load fluctuation triggered by each node is determined. All time-delayed disturbance intensities are cascaded and sorted to generate a time-series propagation link for grid load when peak load fluctuations are triggered. The fluctuation response order is obtained by verifying the time-series transmission link and the empirical sequence of peak triggering in the historical power grid load data.

[0011] In this embodiment, the process of rendering the coefficient evolution trajectory according to the fluctuation response order and overlaying the load contribution distribution from the multi-source fluctuation information to generate a load source tracing view specifically includes: Based on the order of the fluctuation response, the coefficient evolution trajectory is layered and colored and time-series transparency encoded to generate a trajectory rendering layer; The load contribution distribution of each load source is extracted from the big data clustering results of the multi-source fluctuation information to determine the load contribution heatmap; The trajectory rendering layer and the load contribution heatmap are aligned, overlaid, and color-blended to generate a load source tracing view.

[0012] In this embodiment, in response to an interactive command triggered by the graph display interface, the display state of the load tracing view is dynamically updated to obtain the tracing feature graph of power grid load fluctuations. Specifically, this includes: Based on the interactive commands triggered by the map display interface and the load evolution characteristics of the focus area extracted by big data analysis, determine the dynamic focus data of the view; Based on the dynamic focus data of the view and the underlying rendering nodes of the load tracing view, the layer state is reconstructed to generate an updated tracing display layer. By performing feature fusion mapping between the source display layer and the multidimensional correlation topology of power grid load fluctuations, a source feature map of power grid load fluctuations is obtained.

[0013] In this embodiment, the process of reconstructing the layer state based on the dynamic focus data of the view and the underlying rendering nodes of the load tracing view to generate an updated tracing display layer specifically includes: Based on the dynamic focus data of the view and the load topology characteristics of big data analysis, the view frustum is clipped on the bottom rendering node to obtain the layer reconstruction parameters; The transparency and color mapping status of the underlying rendering node are dynamically refreshed using the layer reconstruction parameters to generate an updated source traceability display layer.

[0014] Secondly, this application provides a power grid load visualization analysis system based on elastic network regression, used to execute a power grid load visualization analysis method based on elastic network regression, the analysis system comprising: The acquisition module is used to acquire historical power grid load data of the target area and collect multi-source fluctuation information when the power grid load fluctuates; An iterative module is used to iteratively fuse the historical power grid load data and the multi-source fluctuation information according to the constraint strength sequence adapted to power grid load fluctuations, so as to obtain the coefficient evolution trajectory of the load change sensitivity during power grid load fluctuations. The processing module is used to identify non-zero coefficient jump nodes in the coefficient evolution trajectory, determine the fluctuation response order of the power grid load when triggering load peak fluctuations based on the non-zero coefficient jump nodes, perform layer rendering on the coefficient evolution trajectory based on the fluctuation response order, and overlay the load contribution distribution in the multi-source fluctuation information to generate a load source tracing view. The execution module is used to dynamically update the display status of the load tracing view in response to the interactive command triggered by the graph display interface, so as to obtain the tracing feature graph of power grid load fluctuation.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: Historical power grid load data for the target area is acquired, and multi-source fluctuation information during power grid load fluctuations is collected. The historical power grid load data and the multi-source fluctuation information are iteratively fused according to the constraint strength sequence adapted to power grid load fluctuations to obtain the coefficient evolution trajectory of load change sensitivity during power grid load fluctuations. Non-zero coefficient jump nodes in the coefficient evolution trajectory are identified, and the fluctuation response order of the power grid load when triggering peak load fluctuations is determined based on these nodes. The coefficient evolution trajectory is then rendered using a layer based on the fluctuation response order, and the load contribution distribution from the multi-source fluctuation information is overlaid to generate a load tracing view. In response to interactive commands triggered by the graph display interface, the display state of the load tracing view is dynamically updated to obtain a tracing feature graph of power grid load fluctuations.

[0016] Therefore, this application demonstrates that the display status of the load tracing view can be dynamically updated to obtain a tracing feature map of power grid load fluctuations. By determining the coefficient evolution trajectory, the evolution sequence of sensitivity coefficients of each fluctuation factor as the constraint strength continuously changes, as well as the judgment basis for the fluctuation response timing corresponding to non-zero jump nodes, can be obtained. This overcomes the analytical limitations of static regression coefficients under a single regular parameter, presenting the gradual change law of the influence of different driving factors on load fluctuations, accurately locating the critical time points of each factor's intervention in the load peak formation process, and sorting out the fluctuation response sequence that conforms to the cascading transmission characteristics, effectively improving load stability. Due to the accuracy of the time-series dimension in load source tracing, determining the load source tracing view provides a hierarchical visualization basis for the evolution time-series characteristics of the fusion coefficient and the load contribution weight information. This allows the abstract elastic network regression calculation results to be transformed into a hierarchical visualization presentation. Through layered coloring and transparency encoding, the hierarchical relationship of the fluctuation response is intuitively reflected. The superimposed load contribution thermal distribution intuitively quantifies the influence weight of each driving factor, realizing the visualization integration of time-series transmission logic and contribution ratio information. This ensures the consistency between the hierarchical interactive visualization and the time-series analysis results, and collaboratively improves the overall accuracy of load fluctuation source tracing analysis.

[0017] In summary, the technical solution adopted in this application can achieve the collaborative construction of coefficient evolution time series analysis and hierarchical interactive visualization, so as to improve the accuracy of power grid load fluctuation source analysis. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this embodiment of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is an exemplary flowchart of a power grid load visualization analysis method based on resilient network regression provided in this application; Figure 2 This is a flowchart illustrating the determination of the order of fluctuation responses provided in this application; Figure 3 This is a module structure diagram of a power grid load visualization analysis system based on elastic network regression provided in this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0021] This application provides a method and system for visual analysis of power grid load based on elastic network regression. The core of this method is to acquire historical power grid load data for a target area and collect multi-source fluctuation information during power grid load fluctuations. Iteratively, the historical power grid load data and the multi-source fluctuation information are fused according to a constraint strength sequence adapted to power grid load fluctuations to obtain the coefficient evolution trajectory of load change sensitivity during power grid load fluctuations. Non-zero coefficient jump nodes in the coefficient evolution trajectory are identified, and the fluctuation response order of the power grid load when triggering load peak fluctuations is determined based on these nodes. The coefficient evolution trajectory is then rendered using a layer based on the fluctuation response order, and the load contribution distribution from the multi-source fluctuation information is overlaid to generate a load source tracing view. In response to interactive commands triggered by the graph display interface, the display state of the load source tracing view is dynamically updated to obtain a source feature graph of power grid load fluctuations.

[0022] Example 1: To better understand the above technical solution, the following will provide a detailed description of the technical solution in conjunction with the accompanying drawings and specific implementation methods. (Refer to...) Figure 1 As shown in the figure, this is an exemplary flowchart of a power grid load visualization analysis method based on resilient network regression according to this embodiment of the present application. The analysis method includes the following steps: In step S1, historical power grid load data of the target area is obtained, and multi-source fluctuation information during power grid load fluctuations is collected.

[0023] In practice, historical power grid load data for the target area can be obtained in the following way: retrieve historical operation records for a preset duration from the power grid energy management system and dispatch automation system corresponding to the target area. The time span is selected from 1 to 3 years, and the time granularity is uniformly set to 15 minutes. The extracted data includes indicators such as the total active power load of the area, the load of each administrative division, and the bus and outgoing line loads of substations at each voltage level. After retrieval, the raw data is standardized and cleaned: the 3σ criterion is used to identify abnormal data points that deviate from the normal fluctuation range, and jump values ​​caused by acquisition failures are removed; for missing data with a continuous missing duration of no more than 2 sampling points, linear interpolation is used to complete the data; and all data are aligned in time sequence according to a unified timestamp to form power grid load data.

[0024] It should be noted that, in this application, the target area refers to the designated geographical area covered by the power grid load visualization analysis; historical power grid load data refers to the operational data that records the temporal changes in power grid load in the target area over past periods.

[0025] In addition, in specific implementation, the collection of multi-source fluctuation information during power grid load fluctuations can be achieved in the following way: when the power grid load monitoring value of the target area deviates from the normal operating benchmark and the fluctuation amplitude exceeds the preset threshold, the collection of multi-source fluctuation information for the corresponding fluctuation period is initiated. In the meteorological dimension, meteorological parameters such as hourly temperature, relative humidity, wind speed, and precipitation during the fluctuation period are obtained through the public data interface of the local meteorological department; in the power grid operation dimension, operational data such as distributed new energy output, maintenance status of transmission and transformation equipment, and switching records of reactive power compensation devices are extracted from the dispatch monitoring platform and equipment management system; in the social electricity consumption dimension, industrial enterprise start-up identifiers, commercial load peak markers, and holiday attribute tags are retrieved from the electricity consumption information collection system; in the user-side dimension, electricity consumption plans of large industrial users and charging pile cluster operation data are collected. All collected information is uniformly aligned to a 15-minute time granularity consistent with historical power grid load data. Short-term missing data is filled in using the average of adjacent times. The data is matched and associated with the load data of the corresponding time period according to the time sequence index. The result of the matching and association is used as the multi-source fluctuation information during power grid load fluctuations.

[0026] It should be noted that, in this application, power grid load fluctuation refers to the dynamic change phenomenon in which the actual load of the power grid in the target area deviates from the normal operating level; multi-source fluctuation information refers to a set of quantitative information from different dimensions that reflects the status of various influencing factors of load fluctuation.

[0027] In step S2, the historical power grid load data and the multi-source fluctuation information are iteratively fused according to the constraint strength sequence of power grid load fluctuation adaptation to obtain the coefficient evolution trajectory of load change sensitivity during power grid load fluctuation.

[0028] In this embodiment, the evolution trajectory of the coefficient of load change sensitivity during grid load fluctuations can be obtained by iteratively fusing the historical grid load data and the multi-source fluctuation information according to the constraint strength sequence of grid load fluctuation adaptation using the following steps: Based on the historical power grid load data and the multi-source fluctuation information, a load-driven feature matrix is ​​constructed, and the initial sensitivity coefficient is obtained by regularizing the solution using the smoothing contraction term of the constraint strength sequence. By using the initial sensitivity coefficient and the non-smooth sparse constraints of the constraint strength sequence, redundant factors in the source fluctuation information are iteratively eliminated to generate a sparsified coefficient path. Based on the sparsified coefficient path and the power flow matrix, the node response at load peak is evolved and mapped to obtain the coefficient evolution trajectory of the load change sensitivity during power grid load fluctuations.

[0029] In specific implementation, firstly, a dependent variable sequence is constructed using the time-by-time load values ​​of historical power grid load data. The time-by-time quantized values ​​of various dimensions including meteorology, power grid operation, and social electricity consumption from multi-source fluctuation information are used as independent variable features. After time-series alignment according to a unified timestamp, these are arranged row-wise to construct a load-driven feature matrix with the number of rows equal to the number of sampling times and the number of columns equal to the feature dimensions. Initial parameters of the constraint strength sequence are selected, and a smoothing contraction term is added to the least squares loss function. The coordinate descent method is used for iteration until convergence, and the initial sensitivity coefficients corresponding to each fluctuation feature are obtained. Then, using the initial sensitivity coefficients as the starting point for iterative calculation, each set of parameters with gradient changes in the constraint strength sequence is traversed sequentially. Non-smooth sparse constraints are added to the loss function, and the coordinate descent method is used to complete the regression solution under each round of constraints. After each round of solution, the fluctuation feature whose absolute value of the coefficient converges to zero is identified as a redundant factor and removed from the current feature set. After completing the solution of all constraint strength parameters, the non-zero sensitivity coefficients obtained from each round are concatenated in descending order of constraint strength to form a continuous sparse coefficient path. Finally, based on the power grid topology and line impedance parameters of the target area, a power flow matrix describing the power distribution relationship of each node is constructed. The matrix elements correspond to the power distribution weight of each node to the total load of the area. The global sensitivity coefficient in the sparsified coefficient path is multiplied with the corresponding element of the power flow matrix at each constraint strength point to obtain the node response coefficient of each power grid node under the corresponding constraint strength. The response coefficients of each node are connected in descending order of constraint strength to form the coefficient evolution trajectory of load change sensitivity.

[0030] It should be noted that, in this application, power grid load fluctuation refers to the dynamic change phenomenon where the actual operating load of the power grid in the target area deviates from the normal baseline; constraint strength sequence refers to the parameter set composed of regularization parameters with gradient changes in sequence; iterative fusion refers to the modeling process of integrating load data and multi-source fluctuation information through multiple rounds of parameter iteration; load change sensitivity is an indicator that quantifies the magnitude of power grid load change caused by a unit change in a single type of fluctuation factor; load driving feature matrix refers to a structured matrix composed of the features of each dimension of multi-source fluctuation information arranged in time sequence; smoothing contraction term refers to the L2 norm regularization constraint term in elastic network regression; regularization solution refers to the solution process of adding constraint terms to the regression loss function to limit the coefficient size; initial sensitivity coefficient refers to the coefficient obtained after preliminary solution through smoothing regularization. The regression coefficients of various fluctuation factors are: non-smooth sparse constraints refer to the L1 norm regularization constraint term in elastic network regression; redundancy factors refer to multi-source fluctuation characteristics that have weak explanatory power for grid load fluctuations and have no significant driving effect; iterative elimination refers to the operation of gradually eliminating redundant factors through multiple rounds of constraint strength adjustment; sparsification coefficient path refers to the set of curves in which sensitivity coefficients after sparse screening are connected in sequence under different constraint strengths; the power flow matrix is ​​a matrix describing the power transmission and topology connection characteristics of each node in the power grid; node response refers to the degree of response of the load of each node in the power grid to changes in fluctuation driving factors; evolution mapping refers to the process of converting global regression coefficients into node coefficients by combining the power grid topology; coefficient evolution trajectory refers to the set of curves in which the load sensitivity coefficients of each node change continuously with the constraint strength.

[0031] In addition, in this embodiment, the construction of the load-driven feature matrix based on the historical power grid load data and the multi-source fluctuation information can be achieved by the following steps: Based on the historical power grid load data and the multi-source fluctuation information, big data time series features are extracted to construct an initial high-dimensional feature tensor for dynamic correlation of power supply and demand. A load-driven feature matrix is ​​generated by cross-domain data fusion mapping using the initial high-dimensional feature tensor and the external disturbance factor in the multi-source fluctuation information.

[0032] In practice, firstly, historical power grid load data and multi-source fluctuation information are time-series aligned using a unified timestamp. A fixed-step sliding time window is used to extract time-series statistical features such as mean, extreme values, volatility, and first-order difference for each data sequence window by window, while retaining the original sampled values ​​as basic features. Using time step as the first dimension, feature category as the second dimension, and data domain as the third dimension, all features are categorized and arranged according to these dimensions to construct an initial high-dimensional feature tensor with a three-dimensional structure. Then, Z-score standardization is performed on the feature sequences of each data domain in the initial high-dimensional feature tensor to eliminate numerical scale differences caused by different units. The feature sequences corresponding to external disturbance factors are embedded into the feature tensor as supplementary dimensions. A linear projection method using tensor modulus expansion is then used to expand the three-dimensional high-dimensional feature tensor along the time dimension, mapping it to a two-dimensional matrix structure. After expansion, the validity of each feature is verified column by column, and invalid feature columns with all zeros or zero variance are removed. The resulting matrix is ​​used as the load-driven feature matrix.

[0033] It should be noted that, in this application, big data time series feature extraction refers to the processing stage of feature mining for load and multi-source data with time series arrangement; dynamic correlation of power supply and demand refers to the linkage and correspondence between power supply and demand as they change dynamically with various influencing factors; initial high-dimensional feature tensor refers to the initial feature data carrier organized in a multi-dimensional structure; external disturbance factor refers to the set of load fluctuation influencing factors from outside the power grid; cross-domain data fusion mapping refers to the processing process of transforming features from different data domains to the same feature space.

[0034] In addition, in this embodiment, the iterative removal of redundant factors in the source fluctuation information to generate the sparsity coefficient path can be achieved by the following steps: Based on the multi-source fluctuation information and the multi-dimensional feature correlation evaluation model in big data analysis, the redundant factors are subjected to collinearity dimensionality reduction to obtain a subset of fluctuation features. The sparsified coefficient path is generated by iterative regularization of the fluctuation feature subset and the non-smooth sparse constraints of the constraint strength sequence.

[0035] In practical implementation, firstly, taking all features contained in the multi-source fluctuation information as the analysis object, a multi-dimensional feature correlation assessment model is built using the variance inflation factor method. The variance inflation factor value corresponding to each feature is calculated one by one to measure the degree to which the feature can be linearly explained by all other features. A collinearity judgment threshold is pre-set, and features with variance inflation factors higher than the threshold are marked as redundant factors with strong collinearity. These features are then eliminated sequentially from high to low values. After each feature is eliminated, the variance inflation factor of the remaining features is recalculated. This process is repeated until all remaining features meet the threshold requirement, resulting in a subset of fluctuation features. Then, using historical power grid load data as the regression dependent variable and the selected subset of fluctuation features as the independent variable, calculations are performed round by round according to the constraint strength sequence from largest to smallest. In each round, the non-smooth sparse constraint parameters corresponding to the current constraint strength are selected and added to the least squares loss function to construct the objective function of the elastic network regression. The coordinate descent method is used for iterative calculation until numerical convergence, and the sparsity sensitivity coefficients corresponding to each feature under the current constraint are obtained. After solving for all constraint strengths, the coefficients from each round are concatenated in the constraint order to form a sparsified coefficient path.

[0036] It should be noted that, in this application, the multidimensional feature correlation assessment model refers to an analytical tool for quantifying the degree of linear correlation between features of multi-source fluctuation information; collinearity dimensionality reduction refers to the process of reducing the feature dimension by eliminating highly correlated redundant features; fluctuation feature subset refers to the set of low-redundancy fluctuation features retained after collinearity screening; iterative regularization solution refers to the process of performing regression calculation with sparse constraints round by round according to the constraint strength sequence.

[0037] The multidimensional feature correlation assessment model can be constructed as follows: using all candidate feature sequences contained in the multi-source fluctuation information as model input, firstly perform zero-mean standardization on each feature sequence to eliminate the interference of different physical dimensions on the correlation calculation, build the assessment calculation logic based on the principle of multiple linear regression, take each feature as a temporary dependent variable and all other features as independent variables, construct the corresponding linear regression equations respectively, solve for the determination coefficient of each regression equation, and calculate the variance inflation factor value corresponding to each feature according to the variance inflation factor calculation formula, that is, the variance inflation factor is equal to the reciprocal of the difference with the determination coefficient, thus completing the construction of the multidimensional feature correlation assessment model. This will not be elaborated further here.

[0038] In step S3, non-zero coefficient jump nodes in the coefficient evolution trajectory are identified, and the fluctuation response order of the power grid load when triggering load peak fluctuations is determined based on the non-zero coefficient jump nodes. The coefficient evolution trajectory is rendered in layers according to the fluctuation response order, and the load contribution distribution in the multi-source fluctuation information is superimposed to generate a load source tracing view.

[0039] In this embodiment, identifying the non-zero coefficient transition nodes in the coefficient evolution trajectory can be achieved using the following steps: Based on the coefficient evolution trajectory, perform big data time-series differential operations and zero-value interval marking to determine the candidate windows for jumps corresponding to the coefficient change rate sequence and the duration of zero value; The non-zero jump amplitude of the coefficient change rate sequence is quantified by the jump candidate window and the adaptive mutation detection model to obtain the jump confidence score of the candidate node. Target nodes that meet the preset significance threshold are selected from the jump confidence scores, and load disturbance event correlation verification is performed based on the historical power grid load data to generate jump nodes with non-zero coefficients.

[0040] In specific implementation, firstly, for the time series of a single coefficient evolution trajectory, a large-scale time series differential operation is performed using the first-order backward difference method. The ratio of the coefficient difference between adjacent sampling points to the constraint strength step size is calculated point by point to obtain the coefficient change rate at the corresponding position. The coefficient change rate sequence is formed in sequence. At the same time, all sampling points of the coefficient evolution trajectory are traversed, and all continuous segments with absolute values ​​of coefficients less than the numerical tolerance are marked as zero value intervals. The number of sampling points in each zero value interval is counted as the duration of the zero value. The segment adjacent to the end of the zero value interval is selected as the jump candidate window. The window covers the end of the zero value and the subsequent preset number of non-zero sampling points. Then, using the coefficient change rate sequence within each candidate window as input, an adaptive mutation detection model based on the nonparametric rank-sum test is employed for calculation. The adaptive mutation detection model first statistically analyzes the mean and variance characteristics of the sequence within the window, automatically adapts to generate a mutation detection benchmark value for the corresponding window, performs point-by-point mutation testing on each sampling point within the window, locates the candidate node position, and calculates the amplitude of its coefficient jumping from zero to non-zero, thus quantifying the non-zero jump amplitude. Based on the ratio of the jump amplitude to the background fluctuation level of the sequence, a jump confidence score corresponding to each candidate node is calculated. Finally, a preset significance threshold that meets statistical significance requirements is pre-set, and the jump confidence scores of all candidate nodes are compared with the threshold one by one. Nodes with scores higher than the threshold are selected as target nodes. Historical power grid load data associated with the corresponding constraint strength is retrieved, and recorded load disturbance events are matched to verify whether the coefficient jump corresponding to the target node corresponds to the actual load fluctuation event. False nodes without actual load disturbances are eliminated, thus obtaining the non-zero coefficient jump nodes.

[0041] It should be noted that in this application, big data time-series differential operation refers to the calculation operation of differential differentiation on the coefficient evolution sequence arranged in time series; zero-value interval marking refers to the processing operation of marking continuous segments in the coefficient change rate sequence that take values ​​close to zero; coefficient change rate sequence refers to the time series sequence composed of the instantaneous change rates of coefficients at each sampling point in sequence; zero-value duration refers to the sampling point span covered by the continuous zero-value segments in the coefficient change rate sequence; jump candidate window refers to the local time series segment where the possibility of a non-zero coefficient jump is initially determined; adaptive mutation detection model refers to a mutation identification tool that can automatically adjust the detection benchmark according to the sequence statistical characteristics; non-zero jump amplitude quantization refers to the process of... The process of numerically calculating the magnitude of a coefficient jump from zero to non-zero within a candidate window; a candidate node refers to a potential jump location point initially identified within the jump candidate window; a jump confidence score is a quantitative indicator that quantifies the credibility of a candidate node as a true non-zero jump; a preset significance threshold refers to a pre-set critical value used to determine whether a jump has statistical significance; a target node refers to a high-confidence jump location point retained after screening by the significance threshold; load disturbance event correlation verification refers to the step of matching and verifying jump nodes by combining actual disturbance events in historical load data; a coefficient non-zero jump node refers to a critical location point in the coefficient evolution trajectory where the sensitivity coefficient first changes from zero to non-zero.

[0042] It should also be noted that in this application, the adaptive mutation detection model can be determined in the following way: The adaptive mutation detection model takes the coefficient change rate sequence within the candidate window of the jump as the processing object, and is built using a sliding double window architecture combined with nonparametric rank sum test logic. First, the candidate window is divided into a reference sub-window and a detection sub-window. The reference sub-window is taken from the stationary segment at the end of the zero value interval, which is used to characterize the background fluctuation level when the coefficient has not jumped. The detection sub-window covers the transition segment of the suspected jump. The adaptive mutation detection model can automatically adjust the width of the two sub-windows according to the total sequence length and background variance. The larger the background fluctuation amplitude, the larger the corresponding window width, thereby reducing the interference of random noise on the detection results. Then, the Man-Whitney rank sum test is performed on the sequences within the two sub-windows to calculate the test statistic and the corresponding significance probability value. Then, the non-zero jump amplitude is calculated by combining the coefficient difference before and after the jump, and the candidate node pair is obtained. The corresponding jump confidence score is used to complete the quantitative identification of jump characteristics. The significance threshold can be set in the following way: based on the principle of statistical hypothesis testing and combined with the historical sample calibration method, the historical operation dataset of labeled real load disturbance events in the target area is selected as the calibration sample set. The complete coefficient evolution trajectory calculation and jump candidate node extraction are performed on all samples in the sample set. The distribution law of jump confidence scores corresponding to real jump nodes and false jump nodes are statistically analyzed. Based on the general statistical significance level as the judgment benchmark, combined with the business fault tolerance requirements of power grid load fluctuation analysis, the critical score value that can cover the vast majority of real jump nodes and control the false node false detection ratio within a reasonable range is selected as the initial threshold. Then, cross-validation is carried out through multiple sets of samples from different seasons and different load conditions. The threshold value is fine-tuned until the false detection rate and the missed detection rate reach a balance. Finally, the preset significance threshold is determined.

[0043] Preferably, in this embodiment, the order of the grid load fluctuation response when triggering load peak fluctuations is determined based on the non-zero coefficient jump node, with reference to... Figure 2 As shown in the figure, this is a flowchart illustrating the process of determining the order of fluctuation responses in some embodiments of this application. In this embodiment, the determination of the order of fluctuation responses can be achieved using the following steps: In step S31, the time delay disturbance intensity that triggers load fluctuations at each node is determined based on the jump amplitude increment of the non-zero coefficient jump node. In step S32, all time delay disturbance intensities are cascaded and sorted to generate a time-series propagation link for the grid load when triggering load peak fluctuations; In step S33, the fluctuation response order is obtained by verifying the empirical sequence of peak triggering in the time-series transmission link and the historical power grid load data.

[0044] In practice, firstly, the increment of the jump amplitude corresponding to each non-zero jump node is extracted, that is, the difference between the non-zero sensitivity coefficient at the node and the mean of the coefficients in the preceding zero-value interval. A mapping relationship is established based on the known laws of power grid disturbance transmission. The larger the increment of the jump amplitude, the faster the disturbance transmission speed of the corresponding fluctuation factor and the shorter the delay in triggering the load response. The increment of the jump amplitude is weighted with the significance weight of the coefficient under the corresponding constraint strength to obtain the time delay disturbance intensity of each node triggering load fluctuations. Then, the time delay disturbance intensity corresponding to all fluctuation factors is initially sorted from high to low. The factor with the higher the time delay disturbance intensity, the earlier it intervenes in the transmission process of load peak fluctuations. Based on the cascading propagation logic, the transmission dependency relationship of adjacent sorted factors is checked one by one in combination with the physical properties of each fluctuation factor. If the triggering of the subsequent factor depends on the disturbance output of the preceding factor, the original level is maintained; if there is no dependency, it is adjusted to the same transmission level. The fluctuation factors are connected in series according to the order of the levels to form a hierarchical time-series transmission link. Finally, multiple sets of typical load peak events with labeled causes are extracted from historical power grid load data. The actual response sequence of various fluctuation factors is statistically analyzed according to the corresponding dispatch operation records, and an empirical sequence of peak triggering is formed. The derived time-series transmission link is compared with the empirical sequence one by one, and the order overlap between the two is calculated. Factor nodes with low overlap are corrected in combination with corresponding historical operating conditions, and their positions in the transmission link are adjusted. After correction, the order of each fluctuation factor is determined, which is the fluctuation response sequence corresponding to the load peak fluctuation.

[0045] It should be noted that, in this application, load peak fluctuation refers to the significant upward fluctuation state of the power grid load operating to a local extreme value; jump amplitude increment refers to the jump amplitude value of the sensitivity coefficient at the non-zero jump node; time delay disturbance intensity refers to the quantitative indicator that measures the timeliness and impact of the disturbance transmission of fluctuation factors; cascade propagation sorting refers to the process of sorting factors according to the hierarchical dependence of fluctuation transmission; time-series transmission link refers to the hierarchical chain that connects various fluctuation factors in sequence according to the transmission order; empirical sequence refers to the sequence of fluctuation factor response order obtained based on the summary of historical peak events; verification refers to the step of matching and verifying the derived transmission link with the historical empirical sequence; fluctuation response order refers to the order in which each fluctuation factor begins to exert a significant influence during the formation of load peak.

[0046] In this embodiment, the following steps can be used to generate a load source tracing view by rendering the coefficient evolution trajectory according to the fluctuation response order and overlaying the load contribution distribution in the multi-source fluctuation information: Based on the order of the fluctuation response, the coefficient evolution trajectory is layered and colored and time-series transparency encoded to generate a trajectory rendering layer; The load contribution distribution of each load source is extracted from the big data clustering results of the multi-source fluctuation information to determine the load contribution heatmap; The trajectory rendering layer and the load contribution heatmap are aligned, overlaid, and color-blended to generate a load source tracing view.

[0047] In practice, firstly, according to the order of the fluctuation response, a unique color is assigned to the coefficient evolution trajectory corresponding to each fluctuation factor. Factors that respond first use a low-saturation base color, while factors that respond later use a high-saturation eye-catching color, thus completing the layered coloring. Then, according to the time sequence, time-series transparency encoding is performed. The trajectory layer that responds earlier in the order is set with higher opacity, and the layer that responds later has its transparency increased accordingly, forming a progressive visual hierarchy. A unified drawing canvas is established with constraint strength as the horizontal axis and sensitivity coefficient as the vertical axis. The trajectory curves of each factor are drawn sequentially from the bottom layer to the top layer according to the response order, generating a complete trajectory rendering layer. Then, the K-means clustering algorithm can be used to cluster the feature sequences of multi-source fluctuation information. Different fluctuation characteristics are divided into corresponding load source categories according to their causal attributes, resulting in big data clustering results. The actual fluctuation amplitude of each load source during the period corresponding to the load peak is selected and multiplied by the corresponding sensitivity coefficient to obtain the contribution of a single load source. After normalization, the load contribution distribution of each load source is obtained. With the load source category as the horizontal axis and the contribution ratio as the numerical benchmark, a color gradient mapping rule is used to convert the contribution ratio into corresponding light and dark heatmap color blocks, generating a load contribution heatmap. Finally, using a unified coordinate system as the benchmark, the trajectory rendering layer and the load contribution heatmap are aligned and overlaid, ensuring that the load source categories in the heatmap match the corresponding fluctuation factors in the trajectory layer. An alpha channel blending algorithm is used for color fusion, and the final color parameters after fusion are calculated based on the transparency values ​​of corresponding pixels in the two layers, avoiding color mixing and information occlusion problems after layer overlay. Coordinate axis labels and legends are added to the overlaid view to generate a load source tracing view.

[0048] It should be noted that, in this application, layer rendering refers to the processing step of converting the coefficient evolution trajectory into a layered visualization graphic according to set rules; load contribution distribution refers to the quantitative distribution result of the contribution ratio of each load source to the target load peak; layered coloring refers to the rendering rule of assigning differentiated colors to different factor trajectories according to the fluctuation response level; time-series transparency encoding refers to the encoding method of adjusting the transparency of the trajectory layer according to the response time sequence; trajectory rendering layer refers to the coefficient evolution trajectory visualization layer generated after coloring and transparency encoding; big data clustering result refers to the grouping result obtained after clustering multi-source fluctuation information according to the load cause category; load source refers to the collection of various driving sources that cause power grid load fluctuations; load contribution heat map refers to the heat visualization graphic that uses color depth to represent the contribution ratio of each load source; alignment and overlay refers to the operation of matching and overlaying different visualization layers according to a unified coordinate reference; color fusion refers to the process of mixing the color and transparency of the overlay layers; load source tracing view refers to the static visualization result of integrating trajectory evolution and contribution distribution.

[0049] In step S4, in response to the interactive command triggered by the graph display interface, the display status of the load tracing view is dynamically updated to obtain the tracing feature graph of power grid load fluctuation.

[0050] In this embodiment, the display state of the load tracing view is dynamically updated in response to an interactive command triggered by the graph display interface to obtain the tracing feature graph of power grid load fluctuations. This can be achieved through the following steps: Based on the interactive commands triggered by the map display interface and the load evolution characteristics of the focus area extracted by big data analysis, determine the dynamic focus data of the view; Based on the dynamic focus data of the view and the underlying rendering nodes of the load tracing view, the layer state is reconstructed to generate an updated tracing display layer. By performing feature fusion mapping between the source display layer and the multidimensional correlation topology of power grid load fluctuations, a source feature map of power grid load fluctuations is obtained.

[0051] In practice, the process begins by parsing the interactive commands triggered by the map display interface, identifying the operation type and corresponding view coordinate parameters, and determining the data coverage area corresponding to the user-selected focus region through a pre-defined mapping relationship between view coordinates and data space. From the pre-stored coefficient evolution trajectory and load contribution dataset, load evolution characteristics corresponding to the constraint intensity segment and time series segment covered by the focus region are extracted, including information such as the amplitude of local coefficient changes, the contribution ratio of each factor, and the location of transition nodes. The command parameters and extracted feature data are then structurally integrated to generate dynamic focus data for the view. Next, all underlying rendering nodes corresponding to the load tracing view are retrieved. The layer, coordinate range, and rendering attributes of each node are pre-defined. Using the dynamic focus data as the adjustment basis, the coordinates of each node are matched to see if they are within the focus region. Nodes within the region have their rendering resolution increased and highlighted, while nodes outside the region are adjusted to a semi-transparent or hidden state. The coordinate transformation matrix of each layer is updated synchronously to adapt to the scaling ratio. The layer stacking order is recombined to complete the layer state reconstruction and generate the tracing display layer. Finally, based on the causal relationship between the physical topology of the power grid and fluctuation factors in the target area, a multi-dimensional correlation topology covering power grid nodes, load sources, and fluctuation factors is constructed in advance. The corresponding correlation logic between each element is clarified, and each visual element in the traceability display layer is mapped and matched with the nodes of the multi-dimensional correlation topology one by one. Topological correlation lines and node attribute labels are added to the layer. The coefficient evolution characteristics of the time dimension and the topological correlation characteristics of the spatial dimension are fused and mapped. The legend and interactive control configuration are improved, and a traceability feature map is generated.

[0052] It should be noted that, in this application, the map display interface refers to the human-computer interaction carrier that displays the visualization results of power grid load tracing and receives user input; the interaction command refers to the view adjustment operation signal initiated by the user through the display interface; the display status refers to the set of visualization presentation states of the current layer visibility, coordinate range, and highlight attributes of the load tracing view; the focus area refers to the local area of ​​key analysis selected by the user through the interaction command; the load evolution characteristics refer to the load-related characteristic data of the corresponding coefficient change law, contribution ratio, and jump position within the focus area; the view dynamic focus data refers to the structured dataset that integrates the interaction parameters and the load characteristics of the focus area; the bottom rendering node refers to the smallest visualization rendering unit that constitutes the tracing view; the layer state reconstruction refers to the process of adjusting the rendering node attributes and layer levels according to the focus data; the tracing display layer refers to the updated visualization layer generated after state reconstruction; the multidimensional correlation topology is the topological structure that describes the correlation between power grid nodes, load sources, and fluctuation factors; the feature fusion mapping refers to the process of integrating the visualization layer features with the corresponding topological structure features; and the tracing feature map refers to the interactive visualization result that integrates time-series evolution and topological correlation.

[0053] In addition, in this embodiment, the layer state reconstruction based on the view dynamic focus data and the underlying rendering nodes of the load tracing view to generate the updated tracing display layer can be achieved by the following steps: Based on the dynamic focus data of the view and the load topology characteristics of big data analysis, the view frustum is clipped on the bottom rendering node to obtain the layer reconstruction parameters; The transparency and color mapping status of the underlying rendering node are dynamically refreshed using the layer reconstruction parameters to generate an updated source traceability display layer.

[0054] In practice, firstly, by combining the visible range parameters corresponding to the dynamic focus data of the view, the rectangular view frustum boundary under the 2D view is determined. Simultaneously, the load topology features are retrieved to clarify the load relationships corresponding to each underlying rendering node. All underlying rendering nodes are traversed, and the positional relationship between the node coordinates and the view frustum boundary is compared one by one. Nodes completely outside the view frustum are marked as culled. For nodes at the view frustum boundary, whether to retain them is determined based on whether the associated topology nodes are within the visible range to avoid topology link breaks. The coordinate scaling ratio, visibility identifier, and layer priority parameters of the retained nodes are compiled to form layer reconstruction parameters. Then, based on the layer reconstruction parameters, the visual attributes of the underlying rendering nodes are dynamically refreshed one by one. For core rendering nodes within the focus area, they are set to a completely opaque state, and the color mapping rules are updated synchronously, increasing the color saturation of the corresponding nodes according to the load contribution ratio gradient. For transition nodes at the edge of the view frustum, the transparency is reduced according to the distance from the focus center gradient to form a smooth visual transition. Nodes marked as culled are set to a completely transparent state. After all node attributes are updated, the layers are reassembled according to the original stacking order to generate the updated source traceability display layer.

[0055] It should be noted that, in this application, the load topology feature is a topological attribute feature that describes the relationship between the load elements corresponding to each rendering node; view frustum clipping is a graphics processing operation that filters rendering nodes based on the visible range; layer reconstruction parameters are a set of parameters that carry the rules for adjusting the visibility, coordinate scaling, and visual attributes of rendering nodes; transparency and color mapping status refer to two types of visual presentation attributes of the underlying rendering nodes; dynamic refresh refers to the process of batch updating the visual attributes of rendering nodes based on the reconstruction parameters.

[0056] Therefore, this application demonstrates that the display status of the load tracing view can be dynamically updated to obtain a tracing feature map of power grid load fluctuations. By determining the coefficient evolution trajectory, the evolution sequence of sensitivity coefficients of each fluctuation factor as the constraint strength continuously changes, as well as the judgment basis for the fluctuation response timing corresponding to non-zero jump nodes, can be obtained. This overcomes the analytical limitations of static regression coefficients under a single regular parameter, presenting the gradual change law of the influence of different driving factors on load fluctuations, accurately locating the critical time points of each factor's intervention in the load peak formation process, and sorting out the fluctuation response sequence that conforms to the cascading transmission characteristics, effectively improving load stability. Due to the accuracy of the time-series dimension in load source tracing, determining the load source tracing view provides a hierarchical visualization basis for the evolution time-series characteristics of the fusion coefficient and the load contribution weight information. This allows the abstract elastic network regression calculation results to be transformed into a hierarchical visualization presentation. Through layered coloring and transparency encoding, the hierarchical relationship of the fluctuation response is intuitively reflected. The superimposed load contribution thermal distribution intuitively quantifies the influence weight of each driving factor, realizing the visualization integration of time-series transmission logic and contribution ratio information. This ensures the consistency between the hierarchical interactive visualization and the time-series analysis results, and collaboratively improves the overall accuracy of load fluctuation source tracing analysis.

[0057] In summary, the technical solution adopted in this application can achieve the collaborative construction of coefficient evolution time series analysis and hierarchical interactive visualization, so as to improve the accuracy of power grid load fluctuation source analysis.

[0058] Example 2: This application provides a power grid load visualization analysis system based on elastic network regression, referencing... Figure 3 As shown in the figure, this is a module structure diagram of a power grid load visualization analysis system based on elastic network regression according to this embodiment of the present application. The analysis system includes: The acquisition module 100 is used to acquire historical power grid load data of the target area and collect multi-source fluctuation information when the power grid load fluctuates; The iterative module 200 is used to iteratively fuse the historical power grid load data and the multi-source fluctuation information according to the constraint strength sequence adapted to the power grid load fluctuation, so as to obtain the coefficient evolution trajectory of the load change sensitivity during power grid load fluctuation. The processing module 300 is used to identify non-zero coefficient jump nodes in the coefficient evolution trajectory, determine the fluctuation response order of the power grid load when triggering load peak fluctuations based on the non-zero coefficient jump nodes, perform layer rendering on the coefficient evolution trajectory based on the fluctuation response order, and overlay the load contribution distribution in the multi-source fluctuation information to generate a load source tracing view. The execution module 400 is used to dynamically update the display status of the load tracing view in response to the interactive command triggered by the graph display interface, so as to obtain the tracing feature graph of power grid load fluctuation.

[0059] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0060] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compactdisc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0061] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

Claims

1. A method for power grid load visualization analysis based on elastic net regression, characterized in that, The analytical method includes the following steps: Acquire historical power grid load data for the target area and collect multi-source fluctuation information during power grid load fluctuations; The historical power grid load data and the multi-source fluctuation information are iteratively fused according to the constraint strength sequence of power grid load fluctuation adaptation to obtain the coefficient evolution trajectory of load change sensitivity during power grid load fluctuation. Identify the non-zero coefficient jump nodes in the coefficient evolution trajectory, determine the fluctuation response order of the grid load when triggering load peak fluctuations based on the non-zero coefficient jump nodes, perform layer rendering on the coefficient evolution trajectory based on the fluctuation response order, and overlay the load contribution distribution in the multi-source fluctuation information to generate a load source tracing view. In response to the interactive command triggered by the graph display interface, the display status of the load tracing view is dynamically updated to obtain the tracing feature graph of power grid load fluctuation.

2. The power grid load visualization analysis method based on elastic network regression as described in claim 1, characterized in that, The historical power grid load data and the multi-source fluctuation information are iteratively fused based on the constraint strength sequence of power grid load fluctuation adaptation to obtain the coefficient evolution trajectory of load change sensitivity during power grid load fluctuations. Specifically, this includes: Based on the historical power grid load data and the multi-source fluctuation information, a load-driven feature matrix is ​​constructed, and the initial sensitivity coefficient is obtained by regularizing the solution using the smoothing contraction term of the constraint strength sequence. By using the initial sensitivity coefficient and the non-smooth sparse constraints of the constraint strength sequence, redundant factors in the source fluctuation information are iteratively eliminated to generate a sparsified coefficient path. Based on the sparsified coefficient path and the power flow matrix, the node response at load peak is evolved and mapped to obtain the coefficient evolution trajectory of the load change sensitivity during power grid load fluctuations.

3. The power grid load visualization analysis method based on elastic network regression as described in claim 2, characterized in that, Constructing a load-driven feature matrix based on the historical power grid load data and the multi-source fluctuation information specifically includes: Based on the historical power grid load data and the multi-source fluctuation information, big data time series features are extracted to construct an initial high-dimensional feature tensor for dynamic correlation of power supply and demand. A load-driven feature matrix is ​​generated by cross-domain data fusion mapping using the initial high-dimensional feature tensor and the external disturbance factor in the multi-source fluctuation information.

4. The power grid load visualization analysis method based on elastic network regression as described in claim 2, characterized in that, The iterative removal of redundant factors in the source fluctuation information to generate sparsity coefficients specifically includes: Based on the multi-source fluctuation information and the multi-dimensional feature correlation evaluation model in big data analysis, the redundant factors are subjected to collinearity dimensionality reduction to obtain a subset of fluctuation features. The sparsified coefficient path is generated by iterative regularization of the fluctuation feature subset and the non-smooth sparse constraints of the constraint strength sequence.

5. The power grid load visualization analysis method based on elastic network regression as described in claim 1, characterized in that, Identifying the non-zero coefficient transition nodes in the coefficient evolution trajectory specifically includes: Based on the coefficient evolution trajectory, perform big data time-series differential operations and zero-value interval marking to determine the candidate windows for jumps corresponding to the coefficient change rate sequence and the duration of zero value; The non-zero jump amplitude of the coefficient change rate sequence is quantified by the jump candidate window and the adaptive mutation detection model to obtain the jump confidence score of the candidate node. Target nodes that meet the preset significance threshold are selected from the jump confidence scores, and load disturbance event correlation verification is performed based on the historical power grid load data to generate jump nodes with non-zero coefficients.

6. The power grid load visualization analysis method based on elastic network regression as described in claim 1, characterized in that, Determining the order of power grid load fluctuation response when triggering peak load fluctuations based on the non-zero coefficient jump nodes specifically includes: Based on the increment of the jump amplitude of the non-zero jump node, the time delay disturbance intensity of the load fluctuation triggered by each node is determined. All time-delayed disturbance intensities are cascaded and sorted to generate a time-series propagation link for grid load when peak load fluctuations are triggered. The fluctuation response order is obtained by verifying the time-series transmission link and the empirical sequence of peak triggering in the historical power grid load data.

7. The power grid load visualization analysis method based on elastic network regression as described in claim 1, characterized in that, The process of rendering the coefficient evolution trajectory according to the fluctuation response order and overlaying the load contribution distribution from the multi-source fluctuation information to generate a load source tracing view specifically includes: Based on the order of the fluctuation response, the coefficient evolution trajectory is layered and colored and time-series transparency encoded to generate a trajectory rendering layer; The load contribution distribution of each load source is extracted from the big data clustering results of the multi-source fluctuation information to determine the load contribution heatmap; The trajectory rendering layer and the load contribution heatmap are aligned, overlaid, and color-blended to generate a load source tracing view.

8. The power grid load visualization analysis method based on elastic network regression as described in claim 1, characterized in that, In response to interactive commands triggered by the graph display interface, the display state of the load tracing view is dynamically updated to obtain the source feature graph of power grid load fluctuations, specifically including: Based on the interactive commands triggered by the map display interface and the load evolution characteristics of the focus area extracted by big data analysis, determine the dynamic focus data of the view; Based on the dynamic focus data of the view and the underlying rendering nodes of the load tracing view, the layer state is reconstructed to generate an updated tracing display layer. By performing feature fusion mapping between the source display layer and the multidimensional correlation topology of power grid load fluctuations, a source feature map of power grid load fluctuations is obtained.

9. The power grid load visualization analysis method based on elastic network regression as described in claim 8, characterized in that, Based on the dynamic focus data of the view and the underlying rendering nodes of the load tracing view, the layer state is reconstructed to generate an updated tracing display layer, specifically including: Based on the dynamic focus data of the view and the load topology characteristics of big data analysis, the view frustum is clipped on the bottom rendering node to obtain the layer reconstruction parameters; The transparency and color mapping status of the underlying rendering node are dynamically refreshed using the layer reconstruction parameters to generate an updated source traceability display layer.

10. A power grid load visualization analysis system based on elastic network regression, used to execute a power grid load visualization analysis method based on elastic network regression as described in any one of claims 1 to 9, characterized in that, The analysis system includes: The acquisition module is used to acquire historical power grid load data of the target area and collect multi-source fluctuation information when the power grid load fluctuates; An iterative module is used to iteratively fuse the historical power grid load data and the multi-source fluctuation information according to the constraint strength sequence adapted to power grid load fluctuations, so as to obtain the coefficient evolution trajectory of the load change sensitivity during power grid load fluctuations. The processing module is used to identify non-zero coefficient jump nodes in the coefficient evolution trajectory, determine the fluctuation response order of the power grid load when triggering load peak fluctuations based on the non-zero coefficient jump nodes, perform layer rendering on the coefficient evolution trajectory based on the fluctuation response order, and overlay the load contribution distribution in the multi-source fluctuation information to generate a load source tracing view. The execution module is used to dynamically update the display status of the load tracing view in response to the interactive command triggered by the graph display interface, so as to obtain the tracing feature graph of power grid load fluctuation.

Images