A power market risk response strategy optimization method based on treatment time effectiveness and cost

Abstract

This invention relates to the field of power market risk response technology, and particularly to a method for optimizing power market risk response strategies based on response timeliness and cost. The method includes: collecting real-time and planned electricity consumption data from each user to generate standardized time-series data; determining the real-time deviation and a first deviation sensitivity index for each user; matching user grouping strategies based on this index, extracting electricity consumption behavior characteristics, and dividing the user into homogeneous groups through clustering; evaluating the consistency of each group, calculating the user contribution level for groups that meet the conditions, selecting users according to their contribution level to form a priority adjustment queue and executing adjustment; calculating a second deviation sensitivity index after adjustment, comparing it with the first index to determine the effectiveness of the strategy, and switching the grouping strategy if ineffective. This invention improves the accuracy and adaptability of load regulation by dynamically matching grouping strategies through the deviation sensitivity index, combining response timeliness and adjustment cost for priority ranking and closed-loop feedback optimization.

Description

Technical Field

[0001] This invention relates to the field of power market risk response technology, and in particular to an optimization method for power market risk response strategies based on the timeliness and cost of handling. Background Technology

[0002] With the deepening of electricity market reforms, market players such as electricity retailers, large users, and energy storage operators face increasingly complex market risks. Among these, the risk of deviation assessment arising from the discrepancy between actual electricity consumption and the day-ahead reporting curve is particularly prominent. When actual electricity consumption deviates from the reporting curve, market players will face high deviation assessment fees, thus requiring timely load regulation measures to correct the deviation. Traditional load regulation methods typically cut off user loads in a fixed order or randomly, lacking a comprehensive consideration of user response time, regulation costs, and electricity consumption behavior characteristics, resulting in poor regulation effects or excessive impact on user experience.

[0003] In existing technologies, some methods have been used to classify and manage users through cluster analysis or to optimize the adjustment order through priority ranking. However, most methods have failed to organically combine bias sensitivity assessment, user behavior clustering, contribution calculation and closed-loop feedback of strategy effectiveness, and lack a mechanism to balance the adjustment timeliness and adjustment cost.

[0004] Chinese Patent Publication No. CN114462747A discloses a method and system for user-side adjustable resources to participate in peak-shaving ancillary services. The method includes: acquiring configuration information, electricity consumption information, and cost information of all adjustable resources for a single user; calculating the adjustment cost of each adjustable resource using a pre-determined calculation model; determining the load forecast value of each adjustable resource on the forecast day based on a clustering algorithm and historical electricity load data from the user side; and submitting a quantity and price declaration with the goal of maximizing user benefits. Advantages: This method combines historical electricity load data from the user side and a clustering algorithm to perform load forecasting for adjustable resources in groups. Simultaneously, it analyzes the operating costs of adjustable resources based on their configuration, electricity consumption information, and cost information. Based on the load forecast results and cost analysis results, and through a declaration process, it assists user-side adjustable resources in participating in peak-shaving ancillary services, which can alleviate peak-valley differences, reduce transaction risks, and improve the utilization rate of power grid assets.

[0005] Therefore, the existing technology has the following problems: The pricing and quantity declarations are made solely from the perspective of maximizing the interests of individual users, without involving collaborative adjustments among multiple users based on the degree of contribution of deviations and the timeliness of adjustments, and without forming a closed-loop feedback optimization mechanism for the effectiveness of the grouping strategy. Summary of the Invention

[0006] To address this, the present invention provides an optimization method for power market risk response strategies based on the timeliness and cost of handling, in order to overcome the problems of existing technologies that only focus on maximizing the interests of individual users in quantity and price declarations, without involving the coordinated adjustment among multiple users based on the degree of deviation contribution and adjustment timeliness, and without forming a closed-loop feedback optimization mechanism for the effectiveness of grouping strategies.

[0007] To achieve the above objectives, this invention provides a method for optimizing power market risk response strategies based on disposal timeliness and cost, comprising: Step S1: Collect real-time electricity consumption data and planned electricity consumption data of each user, and preprocess the data to generate standardized time-series data; Step S2: Based on the standardized time-series data, determine the real-time deviation between the real-time electricity consumption data and the planned electricity consumption data of each user, determine the total system deviation value based on the real-time deviation, and determine the first deviation sensitivity index based on the total system deviation value. Step S3: Determine the corresponding user grouping strategy based on the first deviation sensitivity index. Under the determined user grouping strategy, extract the electricity consumption behavior features of each user and divide the users into several homogeneous electricity consumption behavior groups based on the similarity between the electricity consumption behavior features. Step S4: Evaluate the consistency of the electricity consumption behavior characteristics of each user within each homogeneous electricity consumption behavior group. For homogeneous electricity consumption behavior groups that meet the consistency conditions, calculate the contribution of each user within the homogeneous electricity consumption behavior group to the total deviation value of the system, and select users from high to low contribution levels to form a priority adjustment queue. Step S5: Issue load adjustment instructions sequentially according to the priority adjustment queue, perform load adjustment operation, and after adjustment is completed, re-collect real-time electricity consumption data of each user. Based on the re-collected real-time electricity consumption data and the planned electricity consumption data, calculate the second deviation sensitivity index. Step S6: Compare the second deviation sensitivity index with the first deviation sensitivity index, determine whether the current user grouping strategy is effective based on the comparison result, and switch the user grouping strategy when the current user grouping strategy is ineffective.

[0008] Further, in step S2, determining the first deviation sensitivity index includes: Step S21: Based on the standardized time series data, calculate the difference between the actual power value and the planned power value of each user at each sampling time point to obtain the real-time deviation of each user; Step S22: At each sampling time point, the real-time deviations of all users are summed to obtain the total system deviation value at that sampling time point. The total system deviation time series is formed by traversing all sampling time points. Step S23: Extract the absolute value of the total system deviation value corresponding to the latest sampling time point from the total system deviation time series, and use it as the deviation intensity feature; Step S24: Identify the duration for which the total system deviation value continuously exceeds the preset dead zone boundary from the total system deviation time series, and use this as a deviation persistence feature; Step S25: The deviation intensity feature and the deviation persistence feature are weighted and summed to obtain the first deviation sensitivity index.

[0009] Further, in step S3, determining the corresponding user grouping strategy based on the first deviation sensitivity index includes: If the first deviation sensitivity index is greater than or equal to the sensitivity threshold, it is determined that the current state is highly sensitive, and a fine-grained clustering strategy based on all users is matched. If the first deviation sensitivity index is less than the sensitivity threshold, it is determined that the current state is low sensitivity, and a coarse-grained grouping strategy based on historical deviation records is matched.

[0010] Further, in step S3, users are divided into several groups with homogeneous electricity consumption behaviors, including: Step S31: Construct a feature vector from the electricity consumption behavior characteristics of each user; Step S32: Calculate the Euclidean distance between any two user feature vectors to form a distance matrix between users; Step S33: Based on the distance matrix between users, a clustering algorithm is used to classify users whose Euclidean distance between feature vectors is less than a preset clustering threshold into a homogeneous group with the same electricity consumption behavior, and users whose Euclidean distance between feature vectors is greater than or equal to the preset clustering threshold into a homogeneous group with different electricity consumption behaviors. Step S34: Output several homogeneous groups of electricity consumption behaviors after division, where users in each homogeneous group have similar electricity consumption behavior characteristics.

[0011] Further, in step S4, the consistency of electricity consumption behavior characteristics among users within each homogeneous electricity consumption behavior group is evaluated, including: Step S411: Calculate the center vector of all user feature vectors within the homogeneous group of electricity consumption behavior; Step S412: Calculate the Euclidean distance between each user feature vector and the center vector to form an intra-group distance vector; Step S413: Calculate the average value of the distance vectors within the group, and use this average value as the dispersion value of the homogeneous group; Step S414: Compare the dispersion value with a preset consistency threshold. If the dispersion value is less than the consistency threshold, it is determined that the electricity consumption behavior of the homogeneous group meets the consistency condition; if the dispersion value is greater than or equal to the consistency threshold, it is determined that the electricity consumption behavior of the homogeneous group does not meet the consistency condition.

[0012] Further, in step S4, the contribution of each user's electricity consumption behavior within the homogeneous group to the total deviation value of the system is calculated, including: Step S421: Obtain the real-time deviation of each user in the homogeneous electricity consumption behavior group, calculate the ratio of the absolute value of the real-time deviation of each user to the sum of the absolute values ​​of the real-time deviation of all users in the group, and use this ratio as the initial contribution of each user. Step S422: Obtain the adjustment cost index for each user. The adjustment cost index is preset according to the user load type and is used to characterize the degree of negative impact caused by cutting off the user's load. The negative impact includes the degree of deviation of user comfort or the degree of equipment wear. Step S423: Obtain the historical response success rate of each user. The historical response success rate is determined based on the ratio of the number of times the user actually completed the adjustment when executing the load adjustment command to the total number of times the adjustment command was issued. Step S424: Multiply the initial contribution rate by the historical response success rate and divide by the adjustment cost index to obtain the contribution level of each user to the total deviation value of the system.

[0013] Further, in step S4, users are selected in descending order of contribution to form a priority adjustment queue, including: Step S431: Sort users in the same electricity consumption group who meet the consistency condition in descending order of contribution, to obtain the sorted user sequence. Step S432: Obtain the total load that needs to be adjusted. The total load is determined based on the difference between the total system deviation value and the preset dead zone boundary. Step S433: Starting from the user with the highest contribution in the sorted user sequence, select users sequentially. For each user selected, add the interruptible capacity of that user to the total interruptible capacity of the selected users, until the total interruptible capacity is greater than or equal to the total load, then stop selecting. Step S434: The selected users are arranged into a priority adjustment queue according to their selection order.

[0014] Further, in step S5, the second deviation sensitivity index is calculated, including: Step S51: After adjustment, re-collect the real-time power value of each user at the latest sampling time point as the adjusted real-time power data. Step S52: Based on the adjusted real-time power data and the planned electricity consumption data, calculate the adjusted real-time deviation for each user at the latest sampling time point; Step S53: Sum the real-time deviations of all users after adjustment to obtain the total deviation value of the adjusted system at the sampling time point; Step S54: Extract the duration of time that continuously exceeds the preset dead zone boundary before the latest sampling time point from the total system deviation time series formed in step S22 of claim 2, as the deviation persistence feature before adjustment; Step S55: The absolute value of the total deviation of the adjusted system is weighted and summed with the deviation persistence feature before adjustment to obtain the second deviation sensitivity index.

[0015] Further, in step S6, determining whether the current user grouping policy is effective includes: Step S61: Obtain the first deviation sensitivity index and the second deviation sensitivity index; Step S62: Calculate the rate of decrease of the second deviation sensitivity index relative to the first deviation sensitivity index; Step S63: Compare the descent rate with a preset effective descent threshold; Step S64: If the decline rate is greater than or equal to the effective decline threshold, the current user grouping strategy is determined to be effective; if the decline rate is less than the effective decline threshold, the current user grouping strategy is determined to be invalid.

[0016] Further, in step S6, switching the user grouping policy when the current user grouping policy is invalid includes: Record the number of invalid attempts for the current user grouping strategy and compare the number of invalid attempts with the preset maximum number of attempts; if the number of invalid attempts is less than the maximum number of attempts, switch to another user grouping strategy and return to step S3; if the number of invalid attempts reaches the maximum number of attempts, trigger an alarm and terminate the current adjustment process.

[0017] Compared with existing technologies, the advantages of this invention lie in its ability to quantitatively assess the overall electricity consumption deviation of a user group by constructing a system total deviation time series, extracting features from two dimensions—deviation intensity and deviation duration—and generating a first deviation sensitivity index through weighted summation. By weighted fusion of these two features, the risk of deviation assessment can be more comprehensively evaluated, providing a quantitative basis for the adaptive selection of subsequent user grouping strategies.

[0018] Furthermore, this invention divides the deviation sensitivity index into two intervals—high sensitivity and low sensitivity—using a sensitivity threshold, and matches different granularity grouping strategies to different intervals: when the deviation risk is high, fine-grained clustering of all users is used to ensure adjustment accuracy; when the deviation risk is low, coarse-grained grouping of selected users is used to reduce computational complexity. This achieves dynamic adaptation of computational resource allocation to risk level, optimizing system computational efficiency while ensuring adjustment effectiveness.

[0019] Furthermore, this invention constructs user electricity consumption behavior feature vectors and performs clustering based on Euclidean distance, grouping users with similar electricity consumption behaviors into the same homogeneous group. This ensures that the load response characteristics of users within the group have inherent consistency. The clustering results provide a reasonable grouping basis for subsequent consistency assessment and contribution calculation, ensuring that users in the priority adjustment queue not only have large deviation contributions but also belong to the same behavioral pattern group, which facilitates the implementation of targeted load adjustment strategies.

[0020] Furthermore, this invention quantifies the consistency of electricity consumption behavior among users within a homogeneous group by calculating the Euclidean distance between the feature vectors of each user and the center vector, and its average value. The dispersion value, as a quantitative indicator, objectively reflects the similarity among users within the group. Only homogeneous groups that meet the consistency condition proceed to subsequent contribution calculations and priority queue construction, ensuring that users in the priority adjustment queue have high similarity in electricity consumption behavior, thus improving the targeting and effectiveness of load regulation strategies.

[0021] Furthermore, this invention constructs a multi-factor fusion contribution evaluation model by comprehensively calculating the initial contribution level, historical response success rate, and adjustment cost index. The weighted fusion of the initial contribution level, historical response success rate, and adjustment cost index ensures that the ranking of contribution levels considers both the targeted nature of the adjustment effect and the reliability of the adjustment execution and the minimization of the impact on users, thus achieving a balance between the timeliness of handling and the cost of adjustment.

[0022] Furthermore, this invention constructs a priority adjustment queue by combining contribution ranking and greedy selection. This ensures that users with the highest contribution are prioritized for inclusion in the queue, while meeting the total load requirement with the minimum number of users. The greedy algorithm ensures that the number of users being adjusted is minimized while meeting the adjustment requirements, reducing the overall impact on the user group. The ordered nature of the queue provides a clear execution order for subsequent instructions, enabling load adjustment operations to be executed sequentially according to contribution priority, thus improving adjustment efficiency and accuracy.

[0023] Furthermore, this invention obtains the total system deviation value after adjustment by re-acquiring and calculating real-time power data after adjustment, and constructs a second deviation sensitivity index using the deviation persistence characteristics before adjustment. This comprehensively evaluates the residual deviation degree after adjustment and the risk accumulation degree before adjustment, providing a quantitative basis for subsequent comparison with the first deviation sensitivity index. By comparing the difference between the two indices, it can be determined whether the currently implemented grouping strategy and adjustment queue have effectively reduced the system deviation risk.

[0024] Furthermore, this invention achieves a quantitative evaluation of the effectiveness of the user grouping strategy by calculating the rate of decrease in the deviation sensitivity index before and after adjustment and comparing it with a preset effective decrease threshold. When the strategy is determined to be ineffective, the system triggers a grouping strategy switching process to try another grouping strategy to seek a better adjustment effect. When the strategy is determined to be effective, it indicates that the current grouping strategy is suitable for the current power consumption deviation state and can continue to be used. This closed-loop feedback mechanism enables the grouping strategy to be dynamically adjusted according to the actual adjustment effect, improving the system's adaptability to different deviation scenarios. Attached Figure Description

[0025] Figure 1 This is a flowchart of the power market risk response strategy optimization method based on disposal timeliness and cost in this embodiment; Figure 2 This is a flowchart illustrating the determination of the first deviation sensitivity index in the power market risk response strategy optimization method based on disposal timeliness and cost in this embodiment; Figure 3 This is a flowchart illustrating the process of determining the corresponding user grouping strategy in the power market risk response strategy optimization method based on handling timeliness and cost in this embodiment. Detailed Implementation

[0026] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0027] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0028] Please see Figures 1-3 As shown, Figure 1 This is a flowchart of the power market risk response strategy optimization method based on disposal timeliness and cost in this embodiment; Figure 2 This is a flowchart illustrating the determination of the first deviation sensitivity index in the power market risk response strategy optimization method based on disposal timeliness and cost in this embodiment; Figure 3This is a flowchart illustrating the process of determining the corresponding user grouping strategy in the power market risk response strategy optimization method based on handling timeliness and cost in this embodiment.

[0029] This embodiment provides a method for optimizing power market risk response strategies based on disposal timeliness and cost, including: Step S1: Collect real-time electricity consumption data and planned electricity consumption data of each user, and preprocess the data to generate standardized time-series data; Step S2: Based on the standardized time-series data, determine the real-time deviation between the real-time electricity consumption data and the planned electricity consumption data of each user, determine the total system deviation value based on the real-time deviation, and determine the first deviation sensitivity index based on the total system deviation value. Step S3: Determine the corresponding user grouping strategy based on the first deviation sensitivity index. Under the determined user grouping strategy, extract the electricity consumption behavior features of each user and divide the users into several homogeneous electricity consumption behavior groups based on the similarity between the electricity consumption behavior features. Step S4: Evaluate the consistency of the electricity consumption behavior characteristics of each user within each homogeneous electricity consumption behavior group. For homogeneous electricity consumption behavior groups that meet the consistency conditions, calculate the contribution of each user within the homogeneous electricity consumption behavior group to the total deviation value of the system, and select users from high to low contribution levels to form a priority adjustment queue. Step S5: Issue load adjustment instructions sequentially according to the priority adjustment queue, perform load adjustment operation, and after adjustment is completed, re-collect real-time electricity consumption data of each user. Based on the re-collected real-time electricity consumption data and the planned electricity consumption data, calculate the second deviation sensitivity index. Step S6: Compare the second deviation sensitivity index with the first deviation sensitivity index, determine whether the current user grouping strategy is effective based on the comparison result, and switch the user grouping strategy when the current user grouping strategy is ineffective.

[0030] In this embodiment of the invention, step S1 specifically includes: collecting the actual power values ​​of each user at each sampling time point on the same day through the user-side smart meter or dedicated transformer terminal at a sampling frequency of once per minute, as real-time electricity consumption data; retrieving the planned power values ​​of each user reported to the power trading center for each 15-minute window the previous day from the local database of the power sales company, and expanding the planned power values ​​to one data point per minute using linear interpolation to obtain planned electricity consumption data with the same time granularity as the real-time electricity consumption data. Data preprocessing includes: using the Raida criterion (3σ criterion) to remove outliers in the real-time electricity consumption data, i.e., removing data points whose deviation from the mean exceeds 3 times the standard deviation; filling in missing data caused by communication failures using linear interpolation; and mapping the real-time electricity consumption data and planned electricity consumption data to the [0,1] interval using the maximum-minimum normalization method to eliminate the impact of differences in load levels of different users on subsequent analysis. After the above processing, standardized time-series data is generated, which includes user identifier, sampling time point, standardized actual power value, and standardized planned power value.

[0031] Specifically, in step S2, determining the first deviation sensitivity index includes: Step S21: Based on the standardized time-series data, calculate the difference between the actual power value and the planned power value of each user at each sampling time point to obtain the real-time deviation of each user.

[0032] In this embodiment of the invention, multiple interruptible users are connected within the jurisdiction of a power sales company. The sampling period is preset to once per minute, and the current analysis period includes several consecutive sampling time points. The planned electricity consumption data is a constant value every fifteen minutes, which is extended to the same time granularity as the real-time electricity consumption data after linear interpolation. At each sampling time point, the actual power value of each user is collected by the user-side smart meter, and the corresponding planned power value is read from the power sales company's local database. The actual power value is subtracted from the planned power value to obtain the real-time deviation of each user at that sampling time point. A positive real-time deviation indicates that the user's actual electricity consumption is higher than the planned electricity consumption, and a negative value indicates that the actual electricity consumption is lower than the planned electricity consumption. The above calculation is repeated, traversing all sampling time points, to obtain the real-time deviation of each user at each sampling time point.

[0033] Step S22: At each sampling time point, the real-time deviations of all users are summed to obtain the total system deviation value at that sampling time point. The total system deviation time series is formed by traversing all sampling time points.

[0034] In this embodiment of the invention, for each sampling time point, the real-time deviations of all users at that time point are algebraically summed, and the accumulated result is the total system deviation value for that sampling time point. A positive total system deviation value indicates that the overall actual electricity consumption of all users is higher than the planned electricity consumption, while a negative value indicates that the overall actual electricity consumption is lower than the planned electricity consumption. The total system deviation value for each time point is calculated sequentially according to the sampling time order, and the calculation results are arranged in chronological order to form a time series of total system deviation. This series reflects the dynamic trend of the total system electricity consumption deviation over time.

[0035] Step S23: Extract the absolute value of the total system deviation value corresponding to the latest sampling time point from the total system deviation time series, and use it as the deviation intensity feature.

[0036] In this embodiment of the invention, the latest sampling time point is located from the system total deviation time series, and the corresponding system total deviation value is read. The absolute value of this value is taken as the deviation intensity feature. The deviation intensity feature characterizes the magnitude of the deviation between the current system total electricity consumption and the reported plan. The larger the value, the more severe the deviation at the current time.

[0037] Step S24: Identify the duration for which the total system deviation value continuously exceeds the preset dead zone boundary from the total system deviation time series, as a deviation persistence feature.

[0038] In this embodiment of the invention, the method for determining the preset dead zone boundary is as follows: During the system initialization phase of the power sales company, the total system deviation data under normal operating conditions over the past thirty days is collected, its standard deviation is calculated, and the preset dead zone boundary is set to twice the standard deviation. Starting from the beginning of the total system deviation time series, the system scans backwards, identifying time intervals where the absolute value of the total system deviation continuously exceeds the preset dead zone boundary. The time length from the first sampling time point exceeding the boundary to the last sampling time point exceeding the boundary within this time interval is recorded, and this time length is used as the deviation persistence feature. The deviation persistence feature characterizes the length of time the total system deviation remains outside the acceptable range; the larger the value, the longer the deviation problem persists and the higher the degree of risk accumulation.

[0039] Step S25: The deviation intensity feature and the deviation persistence feature are weighted and summed to obtain the first deviation sensitivity index.

[0040] In this embodiment of the invention, the method for determining the weights of the deviation intensity feature and the deviation persistence feature is as follows: using the entropy weight method or the analytic hierarchy process (AHP), the weights are allocated based on the contribution of the deviation intensity feature and the deviation persistence feature to the actual adjustment effect, with the sum of the two weights being 1. The preferred weight for the deviation intensity feature is 0.6, and the preferred weight for the deviation persistence feature is 0.4. The deviation intensity feature is multiplied by its corresponding weight, and the deviation persistence feature is multiplied by its corresponding weight; the weighted sum is then obtained to obtain the first deviation sensitivity index. The first deviation sensitivity index comprehensively reflects the severity of the current electricity consumption deviation and the urgency of the risk of deviation persistence; the larger the index value, the higher the deviation assessment risk faced by the system.

[0041] This invention constructs a system total deviation time series, extracts features from two dimensions—deviation intensity and deviation duration—and generates a first deviation sensitivity index using a weighted summation method, thereby achieving a quantitative assessment of the overall electricity consumption deviation of a user group. By weightedly fusing the two features, the risk of deviation assessment can be more comprehensively evaluated, providing a quantitative basis for the adaptive selection of subsequent user grouping strategies.

[0042] Specifically, in step S3, determining the corresponding user grouping strategy based on the first deviation sensitivity index includes: If the first deviation sensitivity index is greater than or equal to the sensitivity threshold, it is determined that the current state is highly sensitive, and a fine-grained clustering strategy based on all users is matched. If the first deviation sensitivity index is less than the sensitivity threshold, it is determined that the current state is low sensitivity, and a coarse-grained grouping strategy based on historical deviation records is matched.

[0043] In this embodiment of the invention, the sensitivity threshold is determined as follows: During the initialization phase of the power sales company's system, first deviation sensitivity index data under historical normal operating conditions over the past thirty days are collected, and their average value is calculated as a benchmark sensitivity value. The sensitivity threshold is then set as this benchmark sensitivity value. When the first deviation sensitivity index is greater than or equal to the sensitivity threshold, it indicates that the current total system deviation has exceeded the normal fluctuation range, the deviation is severe, and the continuous risk is high, thus determining that the system is currently in a high-sensitivity state. When the first deviation sensitivity index is less than the sensitivity threshold, it indicates that the current total system deviation is within the normal fluctuation range, the deviation is slight, or only momentary, thus determining that the system is currently in a low-sensitivity state.

[0044] The fine-grained clustering strategy based on all users involves including all interruptible users within the electricity sales company's jurisdiction in the clustering analysis. Each user's electricity consumption behavior feature vector is extracted, and a clustering algorithm is used to divide all users into multiple fine-grained homogeneous groups based on their electricity consumption behavior. Each homogeneous group contains a small number of users with highly similar characteristics. This strategy is suitable for highly sensitive situations. Fine-grained grouping can accurately identify differences in electricity consumption behavior among different user groups, providing a more refined user differentiation for subsequent contribution ranking.

[0045] The coarse-grained grouping strategy for filtering users based on historical deviation records is as follows: First, select a number of users from the historical database whose cumulative deviation contribution ranks highest over a preset period (e.g., the past 30 days) to form a set of users to be analyzed; then, cluster only the users within this set to obtain a small number of coarse-grained groups with homogeneous electricity consumption behavior. This strategy is suitable for low-sensitivity conditions. By filtering users with larger historical deviations, the analysis scope is narrowed, reducing computational overhead. Furthermore, because the degree of deviation is relatively minor, coarse-grained grouping is sufficient to meet the adjustment requirements.

[0046] This invention divides the deviation sensitivity index into two intervals, high sensitivity and low sensitivity, using a sensitivity threshold, and then matches different granularity grouping strategies to each interval: when the deviation risk is high, fine-grained clustering of all users is used to ensure adjustment accuracy; when the deviation risk is low, coarse-grained grouping of selected users is used to reduce computational complexity. This achieves dynamic adaptation of computational resource allocation to risk levels, optimizing system computational efficiency while ensuring adjustment effectiveness.

[0047] Specifically, in step S3, users are divided into several groups with homogeneous electricity consumption behaviors, including: Step S31: Construct a feature vector from the electricity consumption behavior characteristics of each user; Step S32: Calculate the Euclidean distance between any two user feature vectors to form a distance matrix between users; Step S33: Based on the distance matrix between users, a clustering algorithm is used to classify users whose Euclidean distance between feature vectors is less than a preset clustering threshold into a homogeneous group with the same electricity consumption behavior, and users whose Euclidean distance between feature vectors is greater than or equal to the preset clustering threshold into a homogeneous group with different electricity consumption behaviors. Step S34: Output several homogeneous groups of electricity consumption behaviors after division, where users in each homogeneous group have similar electricity consumption behavior characteristics.

[0048] In this embodiment of the invention, the electricity consumption behavior characteristics include the user's historical average load factor, load fluctuation variance, peak-valley electricity consumption ratio, correlation coefficient with the total system deviation, and historical response success rate. The above characteristics of each user are combined into a multi-dimensional feature vector. Each feature dimension is normalized to eliminate the influence of different units on distance calculation. The preset clustering threshold is determined by calculating the Euclidean distance between all user feature vectors, obtaining all pairwise distance values, and taking 60% of the arithmetic mean of these distance values ​​as the preset clustering threshold. This percentage can be adaptively adjusted based on the clustering effect during system operation. In step S32, the user distance matrix is ​​a symmetric matrix. The element in the i-th row and j-th column of the matrix represents the Euclidean distance between the i-th user feature vector and the j-th user feature vector, and the diagonal elements are zero. The Euclidean distance is calculated as the square root of the sum of the squares of the differences between the corresponding components of two n-dimensional feature vectors. The smaller the distance value, the more similar the electricity consumption behavior characteristics of the two users.

[0049] In step S33, users are segmented using a hierarchical clustering algorithm or a density-based clustering algorithm. Taking hierarchical clustering as an example: first, each user is considered as an independent initial cluster; then, the two clusters with the closest Euclidean distance between their feature vectors are iteratively merged. The distance between the merged new cluster and other clusters is calculated using the average distance method, which is the average distance between all user pairs in the two clusters; the above merging process is repeated until the minimum distance between any two clusters is greater than or equal to a preset clustering threshold. At this point, each final cluster is a homogeneous group of electricity consumption behavior, where the Euclidean distance between any two user feature vectors within a cluster is less than the preset clustering threshold, while the Euclidean distance between user feature vectors in different clusters is greater than or equal to the preset clustering threshold. In step S34, several homogeneous groups of electricity consumption behavior are output. Users in each homogeneous group have similar electricity consumption behavior characteristics. For example, users in a homogeneous group may all be high-load fluctuating users, or all be stable users, or all be users with significant peak-valley differences. Each homogeneous group serves as the basic unit for evaluating consistency and calculating contribution in subsequent steps.

[0050] This invention constructs a user electricity consumption behavior feature vector and performs clustering based on Euclidean distance, grouping users with similar electricity consumption behaviors into the same homogeneous group. This ensures that the load response characteristics of users within the group are inherently consistent. The clustering results provide a reasonable grouping basis for subsequent consistency assessment and contribution calculation, ensuring that users in the priority adjustment queue not only have large deviation contributions but also belong to the same behavioral pattern group, which facilitates the implementation of targeted load adjustment strategies.

[0051] Specifically, in step S4, the consistency of electricity consumption behavior characteristics among users within each homogeneous electricity consumption behavior group is evaluated, including: Step S411: Calculate the center vector of all user feature vectors within the homogeneous group of electricity consumption behavior; Step S412: Calculate the Euclidean distance between each user feature vector and the center vector to form an intra-group distance vector; Step S413: Calculate the average value of the distance vectors within the group, and use this average value as the dispersion value of the homogeneous group; Step S414: Compare the dispersion value with a preset consistency threshold. If the dispersion value is less than the consistency threshold, it is determined that the electricity consumption behavior of the homogeneous group meets the consistency condition; if the dispersion value is greater than or equal to the consistency threshold, it is determined that the electricity consumption behavior of the homogeneous group does not meet the consistency condition.

[0052] In this embodiment of the invention, the method for calculating the center vector is as follows: For all users within a homogeneous electricity consumption behavior group, the arithmetic mean of each user's electricity consumption behavior feature vector is taken on the same feature dimension to obtain a new feature vector, which is the center vector of the homogeneous group. The center vector represents the "average" electricity consumption behavior pattern of all users within the homogeneous group. In step S412, the method for constructing the intra-group distance vector is as follows: The Euclidean distance between the feature vector of each user within the homogeneous group and the center vector is calculated respectively. Each user corresponds to a distance value. The distance values ​​of all users are arranged in user order to form the intra-group distance vector. The length of this vector is equal to the number of users within the homogeneous group. In step S413, the method for calculating the dispersion value is as follows: The arithmetic mean of all elements in the intra-group distance vector is taken, that is, the sum is divided by the number of users, and the average value is taken as the dispersion value of the homogeneous group. The smaller the dispersion value, the more concentrated the feature vectors of each user within the group are around the center vector, that is, the more consistent the electricity consumption behavior characteristics of the users; the larger the dispersion value, the more dispersed the feature vectors of the users within the group are, and the greater the difference in electricity consumption behavior.

[0053] The method for determining the preset consistency threshold is as follows: During the initialization phase of the electricity sales company's system, for multiple homogeneous groups of electricity consumption behavior already divided under historical normal operating conditions, the dispersion value of each homogeneous group is calculated. The arithmetic mean of all dispersion values ​​is taken, and this mean is multiplied by a redundancy coefficient to obtain the preset consistency threshold. The redundancy coefficient ranges from 1.2 to 1.5, which is used to appropriately relax the judgment criteria based on the historical average level to avoid overly stringent consistency judgment due to small fluctuations. This preset consistency threshold can be dynamically corrected during system operation based on the actual adjustment effect. In step S414, the dispersion value of the current homogeneous group is compared with the preset consistency threshold: if the dispersion value is less than the consistency threshold, it indicates that the user feature vectors within the group are closely distributed around the central vector, and the electricity consumption behaviors among users are highly similar, and the homogeneous group is judged to meet the consistency condition; if the dispersion value is greater than or equal to the consistency threshold, it indicates that the user feature vectors within the group are dispersed, and there are users with significantly different electricity consumption behaviors, and the homogeneous group is judged not to meet the consistency condition, and the users within the homogeneous group should not be prioritized as the same adjustment unit.

[0054] This invention quantifies the consistency of electricity consumption behavior among users within a homogeneous group by calculating the Euclidean distance and its average value between the feature vectors of each user and the center vector. The dispersion value, as a quantitative indicator, objectively reflects the similarity among users within the group. Only homogeneous groups that meet the consistency condition proceed to subsequent contribution calculations and priority queue construction, ensuring that users in the priority adjustment queue exhibit high similarity in electricity consumption behavior and improving the targeting and effectiveness of load regulation strategies.

[0055] Specifically, in step S4, the contribution of each user within the homogeneous electricity consumption group to the total deviation value of the system is calculated, including: Step S421: Obtain the real-time deviation of each user in the homogeneous electricity consumption behavior group, calculate the ratio of the absolute value of the real-time deviation of each user to the sum of the absolute values ​​of the real-time deviation of all users in the group, and use this ratio as the initial contribution of each user. Step S422: Obtain the adjustment cost index for each user. The adjustment cost index is preset according to the user load type and is used to characterize the degree of negative impact caused by cutting off the user's load. The negative impact includes the degree of deviation of user comfort or the degree of equipment wear. Step S423: Obtain the historical response success rate of each user. The historical response success rate is determined based on the ratio of the number of times the user actually completed the adjustment when executing the load adjustment command to the total number of times the adjustment command was issued. Step S424: Multiply the initial contribution rate by the historical response success rate and divide by the adjustment cost index to obtain the contribution level of each user to the total deviation value of the system.

[0056] In this embodiment of the invention, the calculation process for the contribution of each user within a homogeneous electricity consumption group to the total system deviation is as follows: In step S421, the preliminary contribution is calculated as follows: For each user within the homogeneous electricity consumption group, the absolute value of the user's real-time deviation at the current sampling time is obtained. Simultaneously, the sum of the absolute values ​​of the real-time deviations of all users within the group is calculated. The user's absolute value of real-time deviation is divided by this sum, and the resulting ratio is the user's preliminary contribution. The preliminary contribution reflects the relative contribution of the user to the total deviation within the group at the current moment, and its value ranges from 0 to 1. The sum of the preliminary contributions of all users within the group equals 1. Users with larger absolute values ​​of real-time deviation and higher proportions in the total deviation within the group have larger preliminary contributions.

[0057] In step S422, the method for setting the adjustment cost index is as follows: users are divided into three categories—residential users, commercial users, and industrial users—based on their load type, and different adjustment cost index values ​​are preset for each category. Residential users' loads are usually directly related to their living comfort; cutting off their load has a significant impact on user comfort, therefore, the adjustment cost index is preset to a higher value. Commercial users' loads involve lighting, air conditioning, etc., in their business premises; cutting off their load will have a certain impact on their business, so the adjustment cost index is preset to a medium value. Industrial users' loads are mostly interruptible production lines or auxiliary equipment; cutting off their load has a relatively controllable impact on production and usually has a pre-agreed response protocol, so the adjustment cost index is preset to a lower value. The higher the value of the adjustment cost index, the greater the negative impact of cutting off the user's load, and the later that user should be selected for adjustment.

[0058] In step S423, the historical response success rate is determined as follows: In the electricity sales company's historical database, the number of times each user has received load adjustment commands and the number of times the user has actually completed load adjustment commands are counted. The actual number of completed commands is divided by the total number of commands issued; the resulting ratio is the user's historical response success rate. The historical response success rate ranges from 0 to 1. A higher success rate indicates better reliability in responding to load adjustment commands, making the user more suitable for priority selection for adjustment operations. In step S424, the final calculation of the contribution level is as follows: The preliminary contribution level obtained in step S421 is multiplied by the historical response success rate obtained in step S423, and then divided by the adjustment cost index obtained in step S422. The result is the user's contribution level to the total system deviation. This contribution level comprehensively considers three factors: the user's current deviation contribution magnitude, the user's historical response reliability, and the degree of negative impact of the user being adjusted. A user with a higher contribution level indicates that the user has advantages in deviation contribution and response reliability, while also having a lower adjustment cost, and should be prioritized for load adjustment.

[0059] This invention constructs a multi-factor fusion contribution evaluation model by comprehensively calculating the initial contribution level, historical response success rate, and adjustment cost index. The weighted fusion of the initial contribution level, historical response success rate, and adjustment cost index ensures that the ranking of contribution levels considers both the targeted nature of the adjustment effect and the reliability of the adjustment execution and the minimization of the impact on users, thus achieving a balance between the timeliness of handling and the cost of adjustment.

[0060] Specifically, in step S4, users are selected in descending order of contribution to form a priority adjustment queue, including: Step S431: Sort users in the same electricity consumption group who meet the consistency condition in descending order of contribution, to obtain the sorted user sequence. Step S432: Obtain the total load that needs to be adjusted. The total load is determined based on the difference between the total system deviation value and the preset dead zone boundary. Step S433: Starting from the user with the highest contribution in the sorted user sequence, select users sequentially. For each user selected, add the interruptible capacity of that user to the total interruptible capacity of the selected users, until the total interruptible capacity is greater than or equal to the total load, then stop selecting. Step S434: The selected users are arranged into a priority adjustment queue according to their selection order.

[0061] In this embodiment of the invention, in step S431, the contribution value of each user is extracted from the homogeneous group of electricity consumption behaviors that meet the consistency condition. All users are sorted in descending order of contribution, with the user with the highest contribution at the top, the user with the second highest contribution at the bottom, and so on, forming a sorted user sequence. The earlier a user appears in this sequence, the better their overall performance in terms of deviation contribution, response reliability, and adjustment cost, and the higher they should be selected for load adjustment. In step S432, the calculation method for the total load that needs to be adjusted is as follows: the total system deviation value at the latest sampling time point is extracted from the total system deviation time series obtained in step S22, its absolute value is taken, and then the preset dead zone boundary value in step S24 is subtracted. The difference is the total load that needs to be adjusted. This total load represents the total load reduction required to bring the total system deviation back to within the preset dead zone boundary. If the total system deviation is positive, it indicates that the actual electricity consumption is higher than the planned electricity consumption, and user load needs to be cut off to reduce the actual electricity consumption. If the total system deviation is negative, it indicates that the actual electricity consumption is lower than the planned electricity consumption, and user load needs to be increased. This method focuses on scenarios where load is cut off. When the absolute value of the total system deviation is less than or equal to the dead zone boundary, the total load is zero or negative, and no load adjustment is required.

[0062] In step S433, each user is sequentially traversed from the sorted user sequence. For each user, the pre-set interruptible capacity is first obtained. Then, the user's interruptible capacity is accumulated to a cumulative variable with an initial value of zero, resulting in the total interruptible capacity of the selected users. After each accumulation, the accumulated result is compared with the total load calculated in step S432: if the accumulated result is less than the total load, the next user is selected; if the accumulated result is greater than or equal to the total load, selection stops. This selection process employs a greedy strategy, prioritizing users with high contribution levels until the total interruptible capacity of the selected users is sufficient to cover the total load requiring adjustment. In step S434, all users selected in step S433 are arranged sequentially according to their selection order, forming a priority adjustment queue. The first user in this queue is the user with the highest contribution level and should execute the load adjustment command first; the second user is the user with the second highest contribution level, and if there is still a remaining deviation after the first user's execution, the second user's command is executed; and so on.

[0063] This invention constructs a priority adjustment queue by combining contribution ranking and greedy selection. This ensures that users with the highest contribution are prioritized for inclusion in the queue, while meeting the total load demand with the minimum number of users. The greedy algorithm ensures that the number of users being adjusted is minimized while meeting the adjustment requirements, reducing the overall impact on the user group. The ordered nature of the queue provides a clear execution order for subsequent instructions, enabling load adjustment operations to be executed sequentially according to contribution priority, thus improving adjustment efficiency and accuracy.

[0064] Specifically, in step S5, the second deviation sensitivity index is calculated, including: Step S51: After adjustment, re-collect the real-time power value of each user at the latest sampling time point as the adjusted real-time power data. Step S52: Based on the adjusted real-time power data and the planned electricity consumption data, calculate the adjusted real-time deviation for each user at the latest sampling time point; Step S53: Sum the real-time deviations of all users after adjustment to obtain the total deviation value of the adjusted system at the sampling time point; Step S54: Extract the duration of time that continuously exceeds the preset dead zone boundary before the latest sampling time point from the total system deviation time series formed in step S22 of claim 2, as the deviation persistence feature before adjustment; Step S55: The absolute value of the total deviation of the adjusted system is weighted and summed with the deviation persistence feature before adjustment to obtain the second deviation sensitivity index.

[0065] In this embodiment of the invention, in step S51, load adjustment commands are sequentially issued to users in the priority adjustment queue formed by steps S431 to S434. Each user receiving the command executes a load interruption or reduction operation according to its agreed response method. After the adjustment command is executed, a short stabilization period (e.g., one sampling period, i.e., one minute) is waited for the user load to stabilize in the new operating state. Then, the real-time power value of each user at the latest sampling time point is re-collected through the smart meter on the user side as the adjusted real-time power data. In step S52, based on the adjusted real-time power data re-collected in step S51 and the original planned power consumption data (the same as the planned power consumption data used in step S21), the adjusted real-time deviation of each user at the latest sampling time point is calculated using the same method as in step S21. The adjusted real-time deviation is equal to the adjusted actual power value minus the planned power value.

[0066] In step S53, following the same method as in step S22, the real-time deviations of all users after adjustment are algebraically summed, and the accumulated result is the total system deviation value after adjustment at the latest sampling time point. This value reflects the residual deviation between the overall power consumption of the system and the planned power consumption after the load adjustment operation. In step S54, from the total system deviation time series formed in step S22 of claim 2, the duration for which the total system deviation value continuously exceeds the preset dead zone boundary before the latest sampling time point before the adjustment operation is executed is extracted. Specifically, taking the last sampling time point before the adjustment operation is executed as the endpoint, the starting time point of the first continuous exceedance of the dead zone boundary is traced back, and the duration from the starting time point to the endpoint is calculated. This duration is used as the deviation persistence feature before adjustment. This feature characterizes the length of time that the system deviation has been continuously exceeding the tolerance before the adjustment operation is executed, reflecting the urgency and cumulative risk of the deviation problem. In step S55, the same weighted summation rule as in step S25 is adopted, that is, the same weight coefficients are used: deviation intensity feature weight and deviation persistence feature weight. The absolute value of the total deviation of the system after adjustment obtained in step S53 is used as the deviation intensity feature after adjustment, and multiplied by the deviation intensity feature weight; the deviation persistence feature before adjustment obtained in step S54 is multiplied by the deviation persistence feature weight; the two products are added together to obtain the second deviation sensitivity index.

[0067] This invention obtains the total system deviation value after adjustment by re-acquiring and calculating real-time power data. It then constructs a second deviation sensitivity index using the deviation persistence characteristics before adjustment. This comprehensively evaluates the residual deviation after adjustment and the risk accumulation before adjustment, providing a quantitative basis for subsequent comparison with the first deviation sensitivity index. By comparing the differences between the two indices, it can be determined whether the currently implemented grouping strategy and adjustment queue have effectively reduced the system deviation risk.

[0068] Specifically, in step S6, determining whether the current user grouping policy is effective includes: Step S61: Obtain the first deviation sensitivity index and the second deviation sensitivity index; Step S62: Calculate the rate of decrease of the second deviation sensitivity index relative to the first deviation sensitivity index; Step S63: Compare the descent rate with a preset effective descent threshold; Step S64: If the decline rate is greater than or equal to the effective decline threshold, the current user grouping strategy is determined to be effective; if the decline rate is less than the effective decline threshold, the current user grouping strategy is determined to be invalid.

[0069] In this embodiment of the invention, in step S61, a first deviation sensitivity index is obtained from the calculation result of step S25, and a second deviation sensitivity index is obtained from the calculation result of step S55. The first deviation sensitivity index characterizes the initial deviation risk level before the load adjustment operation is executed, and the second deviation sensitivity index characterizes the residual deviation risk level after the load adjustment operation is executed. In step S62, the decrease rate is calculated as follows: subtract the second deviation sensitivity index from the first deviation sensitivity index to obtain the absolute decrease in the deviation sensitivity index; divide the absolute decrease by the first deviation sensitivity index, and the resulting ratio is the decrease rate of the second deviation sensitivity index relative to the first deviation sensitivity index. The formula for calculating the decrease rate is: decrease rate = (first deviation sensitivity index - second deviation sensitivity index) / first deviation sensitivity index. The decrease rate ranges from 0 to 1. The closer the decrease rate is to 1, the better the adjustment effect and the more significant the reduction in deviation risk; the closer the decrease rate is to 0, the worse the adjustment effect and the almost no reduction in deviation risk.

[0070] In step S63, the method for determining the preset effective decline threshold is as follows: During the initialization phase of the power sales company's system, the deviation sensitivity index before and after each adjustment in historical successful adjustment cases is collected, the decline rate of each adjustment is calculated, and the arithmetic mean of the decline rates of all successful cases is taken as the benchmark decline rate. The preset effective decline threshold is set as this benchmark decline rate. This preset effective decline threshold can be dynamically corrected during system operation based on the statistical distribution of the actual adjustment effect. In step S64, the actual decline rate calculated in step S62 is compared with the preset effective decline threshold in step S63: If the actual decline rate is greater than or equal to the preset effective decline threshold, it indicates that the load adjustment operation guided by the current user grouping strategy has reduced the system deviation risk by a sufficient margin, achieving the expected adjustment effect. Therefore, the current user grouping strategy is determined to be effective. If the actual decline rate is less than the preset effective decline threshold, it indicates that the reduction in deviation risk after the adjustment operation is insufficient and has not achieved the expected effect. The current user grouping strategy has failed to effectively guide load adjustment. Therefore, the current user grouping strategy is determined to be invalid.

[0071] This invention achieves a quantitative evaluation of the effectiveness of user grouping strategies by calculating the rate of decrease in the sensitivity index before and after adjustment and comparing it with a preset effective decrease threshold. When a strategy is determined to be ineffective, the system triggers a grouping strategy switching process to try another grouping strategy to seek a better adjustment effect. When a strategy is determined to be effective, it indicates that the current grouping strategy is suitable for the current power consumption deviation state and can continue to be used. This closed-loop feedback mechanism enables the grouping strategy to be dynamically adjusted according to the actual adjustment effect, improving the system's adaptability to different deviation scenarios.

[0072] Specifically, in step S6, switching the user grouping policy when the current user grouping policy is invalid includes: Record the number of invalid attempts for the current user grouping strategy and compare the number of invalid attempts with the preset maximum number of attempts; if the number of invalid attempts is less than the maximum number of attempts, switch to another user grouping strategy and return to step S3; if the number of invalid attempts reaches the maximum number of attempts, trigger an alarm and terminate the current adjustment process.

[0073] In this embodiment of the invention, the specific method for recording the number of invalid attempts for the current user grouping strategy is as follows: In each adjustment process of the power sales company system, an invalid attempt counter is set for the current user grouping strategy, with an initial value of zero. Each time step S64 is executed and the current user grouping strategy is determined to be invalid, the value of the invalid attempt counter is incremented by one, thus accumulating and recording the number of invalid attempts. If the strategy is determined to be valid in a subsequent switching attempt, the counter is reset to zero. The method for determining the preset maximum number of attempts is as follows: it is comprehensively determined based on the number of users within the power sales company's jurisdiction, the computational complexity of the clustering algorithm, and the time window requirements for load adjustment. The more users and the longer the clustering calculation takes, the smaller the maximum number of attempts should be set to ensure that the adjustment operation can be completed within the specified assessment time window. The preset maximum number of attempts ranges from two to five times, and in this embodiment, three times is preferred. This value can be dynamically adjusted during system operation according to the actual adjustment time requirements.

[0074] The specific method for switching to another user grouping strategy is as follows: if the currently used user grouping strategy is a fine-grained clustering strategy based on all users, then switch to a coarse-grained grouping strategy based on historical deviation records; if the currently used user grouping strategy is a coarse-grained grouping strategy based on historical deviation records, then switch to a fine-grained clustering strategy based on all users. The two strategies are alternatives to each other and are switched alternately when ineffective.

[0075] The specific method for returning to step S3 is as follows: after completing the grouping strategy switch, jump to the execution entry point of step S3, and re-execute steps S3 to S6 using the switched user grouping strategy, that is, re-perform user clustering, consistency assessment, contribution calculation, priority adjustment queue construction, load adjustment execution, and strategy effectiveness determination.

[0076] The specific method for triggering the alarm and terminating the current adjustment process is as follows: When the number of invalid attempts reaches the preset maximum number of attempts, it indicates that both user grouping strategies have been tried and neither has reached the effective reduction threshold, and the system cannot achieve effective load adjustment through automatic strategy switching. At this time, the system sends an alarm message to the power sales company's operation and maintenance platform. The alarm message includes the identifier of this adjustment process, the first deviation sensitivity index, the second deviation sensitivity index, the reduction rate of the two attempts, and the total system deviation value before and after adjustment. Simultaneously, the current adjustment process is terminated, and the load adjustment operation is no longer executed, awaiting manual intervention to analyze the cause.

[0077] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for optimizing power market risk response strategies based on disposal timeliness and cost, characterized in that, include: Step S1: Collect real-time electricity consumption data and planned electricity consumption data of each user, and preprocess the data to generate standardized time-series data; Step S2: Based on the standardized time-series data, determine the real-time deviation between the real-time electricity consumption data and the planned electricity consumption data of each user, determine the total system deviation value based on the real-time deviation, and determine the first deviation sensitivity index based on the total system deviation value. Step S3: Determine the corresponding user grouping strategy based on the first deviation sensitivity index. Under the determined user grouping strategy, extract the electricity consumption behavior features of each user and divide the users into several homogeneous electricity consumption behavior groups based on the similarity between the electricity consumption behavior features. Step S4: Evaluate the consistency of the electricity consumption behavior characteristics of each user within each homogeneous electricity consumption behavior group. For homogeneous electricity consumption behavior groups that meet the consistency conditions, calculate the contribution of each user within the homogeneous electricity consumption behavior group to the total deviation value of the system, and select users from high to low contribution levels to form a priority adjustment queue. Step S5: Issue load adjustment instructions sequentially according to the priority adjustment queue, perform load adjustment operation, and after adjustment is completed, re-collect real-time electricity consumption data of each user. Based on the re-collected real-time electricity consumption data and the planned electricity consumption data, calculate the second deviation sensitivity index. Step S6: Compare the second deviation sensitivity index with the first deviation sensitivity index, determine whether the current user grouping strategy is effective based on the comparison result, and switch the user grouping strategy when the current user grouping strategy is ineffective.

2. The method for optimizing power market risk response strategies based on disposal timeliness and cost according to claim 1, characterized in that, In step S2, the first bias sensitivity index is determined, including: Step S21: Based on the standardized time series data, calculate the difference between the actual power value and the planned power value of each user at each sampling time point to obtain the real-time deviation of each user; Step S22: At each sampling time point, the real-time deviations of all users are summed to obtain the total system deviation value at that sampling time point. The total system deviation time series is formed by traversing all sampling time points. Step S23: Extract the absolute value of the total system deviation value corresponding to the latest sampling time point from the total system deviation time series, and use it as the deviation intensity feature; Step S24: Identify the duration for which the total system deviation value continuously exceeds the preset dead zone boundary from the total system deviation time series, and use this as a deviation persistence feature; Step S25: The deviation intensity feature and the deviation persistence feature are weighted and summed to obtain the first deviation sensitivity index.

3. The method for optimizing power market risk response strategies based on disposal timeliness and cost according to claim 1, characterized in that, In step S3, the corresponding user grouping strategy is determined based on the first deviation sensitivity index, including: If the first deviation sensitivity index is greater than or equal to the sensitivity threshold, it is determined that the current state is highly sensitive, and a fine-grained clustering strategy based on all users is matched. If the first deviation sensitivity index is less than the sensitivity threshold, it is determined that the current state is low sensitivity, and a coarse-grained grouping strategy based on historical deviation records is matched.

4. The method for optimizing power market risk response strategies based on disposal timeliness and cost according to claim 1, characterized in that, In step S3, users are divided into several groups with homogeneous electricity consumption behaviors, including: Step S31: Construct a feature vector from the electricity consumption behavior characteristics of each user; Step S32: Calculate the Euclidean distance between any two user feature vectors to form a distance matrix between users; Step S33: Based on the distance matrix between users, a clustering algorithm is used to classify users whose Euclidean distance between feature vectors is less than a preset clustering threshold into a homogeneous group with the same electricity consumption behavior, and users whose Euclidean distance between feature vectors is greater than or equal to the preset clustering threshold into a homogeneous group with different electricity consumption behaviors. Step S34: Output several homogeneous groups of electricity consumption behaviors after division, where users in each homogeneous group have similar electricity consumption behavior characteristics.

5. The method for optimizing power market risk response strategies based on disposal timeliness and cost according to claim 1, characterized in that, In step S4, the consistency of electricity consumption behavior characteristics among users within each homogeneous electricity consumption behavior group is evaluated, including: Step S411: Calculate the center vector of all user feature vectors within the homogeneous group of electricity consumption behavior; Step S412: Calculate the Euclidean distance between each user feature vector and the center vector to form an intra-group distance vector; Step S413: Calculate the average value of the distance vectors within the group, and use this average value as the dispersion value of the homogeneous group; Step S414: Compare the dispersion value with a preset consistency threshold. If the dispersion value is less than the consistency threshold, it is determined that the electricity consumption behavior of the homogeneous group meets the consistency condition; if the dispersion value is greater than or equal to the consistency threshold, it is determined that the electricity consumption behavior of the homogeneous group does not meet the consistency condition.

6. The method for optimizing power market risk response strategies based on disposal timeliness and cost according to claim 1, characterized in that, In step S4, the contribution of each user's electricity consumption behavior within the homogeneous group to the total deviation value of the system is calculated, including: Step S421: Obtain the real-time deviation of each user in the homogeneous electricity consumption behavior group, calculate the ratio of the absolute value of the real-time deviation of each user to the sum of the absolute values ​​of the real-time deviation of all users in the group, and use this ratio as the initial contribution of each user. Step S422: Obtain the adjustment cost index for each user. The adjustment cost index is preset according to the user load type and is used to characterize the degree of negative impact caused by cutting off the user's load. The negative impact includes the degree of deviation of user comfort or the degree of equipment wear. Step S423: Obtain the historical response success rate of each user. The historical response success rate is determined based on the ratio of the number of times the user actually completed the adjustment when executing the load adjustment command to the total number of times the adjustment command was issued. Step S424: Multiply the initial contribution rate by the historical response success rate and divide by the adjustment cost index to obtain the contribution level of each user to the total deviation value of the system.

7. The method for optimizing power market risk response strategies based on disposal timeliness and cost according to claim 1, characterized in that, In step S4, users are selected in descending order of contribution to form a priority adjustment queue, including: Step S431: Sort users in the same electricity consumption group who meet the consistency condition in descending order of contribution, to obtain the sorted user sequence. Step S432: Obtain the total load that needs to be adjusted. The total load is determined based on the difference between the total system deviation value and the preset dead zone boundary. Step S433: Starting from the user with the highest contribution in the sorted user sequence, select users sequentially. For each user selected, add the interruptible capacity of that user to the total interruptible capacity of the selected users, until the total interruptible capacity is greater than or equal to the total load, then stop selecting. Step S434: The selected users are arranged into a priority adjustment queue according to their selection order.

8. The method for optimizing power market risk response strategies based on disposal timeliness and cost according to claim 1, characterized in that, In step S5, the second deviation sensitivity index is calculated, including: Step S51: After adjustment, re-collect the real-time power value of each user at the latest sampling time point as the adjusted real-time power data. Step S52: Based on the adjusted real-time power data and the planned electricity consumption data, calculate the adjusted real-time deviation for each user at the latest sampling time point; Step S53: Sum the real-time deviations of all users after adjustment to obtain the total deviation value of the adjusted system at the sampling time point; Step S54: Extract the duration of time that continuously exceeds the preset dead zone boundary before the latest sampling time point from the total system deviation time series formed in step S22 of claim 2, as the deviation persistence feature before adjustment; Step S55: The absolute value of the total deviation of the adjusted system is weighted and summed with the deviation persistence feature before adjustment to obtain the second deviation sensitivity index.

9. The method for optimizing power market risk response strategies based on disposal timeliness and cost according to claim 1, characterized in that, In step S6, it is determined whether the current user grouping policy is effective, including: Step S61: Obtain the first deviation sensitivity index and the second deviation sensitivity index; Step S62: Calculate the rate of decrease of the second deviation sensitivity index relative to the first deviation sensitivity index; Step S63: Compare the descent rate with a preset effective descent threshold; Step S64: If the decline rate is greater than or equal to the effective decline threshold, the current user grouping strategy is determined to be effective; if the decline rate is less than the effective decline threshold, the current user grouping strategy is determined to be invalid.

10. The method for optimizing power market risk response strategies based on disposal timeliness and cost according to claim 9, characterized in that, In step S6, switching the user grouping policy when the current user grouping policy is invalid includes: Record the number of invalid attempts for the current user grouping strategy and compare the number of invalid attempts with the preset maximum number of attempts; if the number of invalid attempts is less than the maximum number of attempts, switch to another user grouping strategy and return to step S3; if the number of invalid attempts reaches the maximum number of attempts, trigger an alarm and terminate the current adjustment process.

Images