Fine-grained image and management method of power users considering load and social information

Abstract

The application discloses a power user fine image and management method considering load and social information, establishes a multi-dimensional power user electricity consumption behavior evaluation system from four dimensions of energy consumption and load characteristics, adjustable potential, behavior consciousness and user nature, and carries out fine image on a single user. Facing demand response, an adjustable potential, load matching degree and behavior consciousness calculation method based on fine-grained load data and user electricity consumption preference investigation information mining is proposed to comprehensively reflect the user adjustable capacity. Based on the DBSCAN clustering algorithm, large-scale power user classification, group labeling and management are realized. The application can deeply mine the value of load data and social information, fully reflects the user load characteristics and subjective behavior characteristics, helps users understand their own electricity consumption behavior, supports power enterprises to provide personalized services and auxiliary decision for users, thus helping to integrate flexible resources on the demand side and supporting the construction of a new power system.

Description

TECHNICAL FIELD

[0001] The present application relates to the power system, and particularly to a power user fine portrait and management method considering load and social information. BACKGROUND

[0002] Under the double carbon target, China will build a new power system dominated by new energy, and the regulation capacity of the system will become a scarce resource. With the continuous development of urbanization in China, the power consumption of urban users has increased rapidly. Influenced by non-economic factors such as natural environment and living needs, there is a great space for flexible load regulation. However, it is difficult for power companies to accurately portrait the power consumption behavior of users with strong subjectivity, so it is also difficult to develop effective interaction strategies to stimulate users' willingness to participate in interactive regulation and guide users' power consumption behavior. Therefore, how to comprehensively consider the subjective power consumption behavior information based on social research and the massive fine load data collected based on Internet of Things technology to accurately portray the power consumption behavior of users and classify and manage a large number of users is a problem to be solved in the field of power supply and demand interaction.

[0003] At present, there are many studies on user power consumption behavior, but the existing research lacks consideration of user social awareness, and it is difficult to reflect the social awareness and subjective needs of users in the analysis of user power consumption characteristics. Traditional analysis based on total user power load cannot achieve in-depth mining of user fine power consumption behavior, and lacks consideration of non-data information such as social environment, family members, and consumer psychology. It is difficult for power companies to accurately assess the adjustable capacity of power users, so as to realize friendly interaction between users and power supply and demand, and support the development of new power systems. SUMMARY

[0004] The purpose of the present application is to provide a power user fine portrait and management method considering load and social information, so as to fully mine the fine power consumption behavior of users, help users and power companies fully understand the power consumption rules and adjustable capacity of users, and realize large-scale user classification and group management for demand response.

[0005] Technical scheme: The power user fine portrait and management method considering load and social information provided by the present application comprises the following steps:

[0006] (1) A multi-dimensional power user power consumption behavior evaluation system with four secondary indicators and 14 tertiary indicators is established from four dimensions of energy consumption and load characteristics, adjustable potential, behavior awareness, and user nature, to portrait a single user in detail.

[0007] (2) Through the mining of user electricity data, a user and regional power grid load matching degree index calculation method and a load adjustable potential index calculation method are proposed, and the load adjustable potential index includes transferable load potential and reducible load potential.

[0008] (3) The user electricity data is combined with the questionnaire survey results of the electricity users, and the subjective behavior consciousness of the users is comprehensively evaluated from four aspects of policy sensitivity, electricity fee sensitivity, user interaction willingness and green electricity consciousness based on the subjective and objective evaluation method combining the analytic hierarchy process and the entropy weight method.

[0009] (4) According to the quantified scores of the user load data level and the subjective consciousness level obtained in steps (2) and (3), the portrait results considering the user adjustable potential, the load matching degree and the subjective behavior consciousness can be clustered and analyzed based on the DBSCAN clustering algorithm, so as to realize the classification, grouping and labeling management of large-scale users for demand response.

[0010] The step (1) is specifically:

[0011] (1.1) Four dimensions of power user electricity behavior evaluation are established.

[0012] Multi-dimensional power user fine portrait is performed from four dimensions of energy consumption and load characteristics, adjustable potential, behavior consciousness and user nature, which are used as secondary indexes of power user electricity behavior evaluation.

[0013] (1.2) Fourteen tertiary indexes of power user electricity behavior evaluation are established.

[0014] Four tertiary indexes of daily load curve, peak-valley characteristics, load matching degree and energy consumption level are considered under the secondary index of energy consumption and load characteristics.

[0015] Two tertiary indexes of transferable load potential and reducible load potential are considered under the secondary index of adjustable potential.

[0016] Four tertiary indexes of policy sensitivity, electricity fee sensitivity, user interaction willingness and green electricity consciousness are considered under the secondary index of behavior consciousness.

[0017] Three tertiary indexes of user electricity habit, user family attribute and new energy attribute are considered under the secondary index of user nature.

[0018] The step (2) is specifically:

[0019] (2.1) The adjustable potential of the user is calculated:

[0020] By combining the transferable load potential and the curtailable load potential, the user load power is calculated and used as an indicator of the user's adjustable potential to evaluate the adjustable potential of large-scale users. The user's adjustable potential coefficient is calculated as follows:

[0021]

[0022] in, and The number of electrical equipment representing the categories of load that can be transferred and load that can be reduced; is the transferable load potential of the i-th equipment; is the load reduction potential of the jth device; is the total load.

[0023] (2.2) Calculate the load matching degree:

[0024] The user's daily load curve is analyzed and matched with the daily load curve of the local power grid. The similarity reflects the user's potential for participating in demand response. The load matching degree uses the Fréchet distance, which comprehensively considers the position and order of curve points, to measure the similarity between curves.

[0025] make is a metric space, let and for two given curves in ; then, the Fréchet distance between A and B is defined as: and After parameterization On a plane and The infimum of the maximum distance between , where ; then the Fréchet distance yes:

[0026]

[0027] in, yes The distance function.

[0028] The step (3) is specifically as follows:

[0029] (3.1) Construct a hierarchical system for comprehensive evaluation of users’ subjective behavioral awareness:

[0030] The policy sensitivity, electricity fee sensitivity, user interaction willingness, and green power consumption awareness of the behavioral consciousness dimension are analyzed, and corresponding questions in the questionnaire are selected as the basis for weighting the four indicators. A hierarchical system is established according to the three-layer structure of the target layer, the criterion layer, and the factor layer. The answers to the questions in the questionnaire related to the four influencing factors are scored. The scoring principle mainly refers to the scoring method of the Likert scale, which consists of a set of statements. Each statement has five answers: "very agree", "agree", "not sure", "disagree", and "very disagree", which are scored as 5, 4, 3, 2, and 1, respectively. The total score of each respondent is the sum of the scores obtained from the answers to each question, which indicates the strength of their attitude or their different status on the scale.

[0031] (3.2) Consistency test:

[0032] According to the expert scoring results, the judgment matrix is constructed and , respectively, which are and , respectively. The order of the matrix elements is the number of criterion layer dimensions and the number of factors under the first dimension, respectively. The matrix elements are the values filled by experts according to the proportional scale table. The judgment matrix must pass the consistency test before the weight results calculated by it can be used. The consistency test formula is as follows:

[0033]

[0034]

[0035] wherein, is the maximum eigenvalue of the matrix; is the consistency index; is the order of the matrix; is the random consistency index, which is related to and can be obtained by looking up the table; is the consistency ratio;

[0036] When , the matrix is considered to pass the consistency test, otherwise the judgment matrix should be modified.

[0037] (3.3) Determine the comprehensive evaluation weight:

[0038] (3.3.1) Analytic hierarchy process weighting:

[0039] The eigenvalue method is used to derive the relative ordering weight of elements under a single criterion from the judgment matrix, i.e. the characteristic vector corresponding to the maximum eigenvalue of the judgment matrix. After normalization, the ordering weight vector is denoted as , record the ranking weight vector of , the ranking weight vector of the first factor, then:

[0040]

[0041] Similarly, the ranking weight vector of , the ranking weight vector of for , then:

[0042]

[0043] Multiply the weight of a certain factor with the weight of for , the weight of the factor for the first factor of the factor layer for , is recorded as .

[0044] (3.3.2) Entropy weight method:

[0045] On the basis of questionnaire scoring, list the scores of each factor for users; let the user's index be , and the normalized result be . Calculate the proportion of the user's index , according to , the information entropy of the index can be calculated, and finally the objective weight of the index is obtained, recorded as ; the formula is as follows:

[0046]

[0047]

[0048]

[0049]

[0050] (3.3.3) Combination weighting: ​​

[0051] The subjective weight and the objective weight are combined, so that the finally determined weight reflects both subjective and objective information, thereby making the final result more accurate.

[0052] The weight of the first factor obtained according to the analytic hierarchy process and the entropy weight method is and respectively, and the combined weight of the combined weight formula is :

[0053]

[0054] (3.3.4) Calculate the behavior consciousness comprehensive score:

[0055] According to the scoring situation, the score of each factor problem of each user is obtained, and the final score of the user response willingness level is obtained by weighting and summing the scores of each factor and the corresponding weight obtained by the combined weight. The score of the user response willingness level obtained by the above method after processing the natural information can reflect the high and low of the user's response willingness, thereby realizing more refined and efficient evaluation of large-scale user adjustable resources.

[0056] The step (4) is specifically:

[0057] (4.1) Initialize the core object set , initialize the cluster number , initialize the unvisited sample set , and divide the cluster .

[0058] (4.2) For , find all core objects according to the following steps.

[0059] (4.2.1) Find the neighborhood sub-sample set of the sample by distance measurement.

[0060] (4.2.2) If the number of sub-sample set samples meets , add the sample to the core object sample set: .

[0061] (4.3) If the core object set , the algorithm ends, otherwise go to step (4.4).

[0062] (4.4) In the core object set , randomly select a core object ​, initialize the current cluster core object queue , initialize the category number , initialize the current cluster sample set , update the unvisited sample set .

[0063] (4.5) if the current cluster core object queue , the current cluster is generated , update the cluster division , update the core object set , go to step (3.3); otherwise, update the core object set .

[0064] (4.6) take out a core object from the current cluster core object queue , find all neighbor sub-sample sets through the neighborhood distance threshold , let , update the current cluster sample set , update the unvisited sample set , update , go to step (4.5).

[0065] In the whole process, the input is the sample set , the neighborhood parameter and the sample distance measurement method; the output result is the cluster division .

[0066] Through the above steps, the data obtained in steps (2) and (3) can be subjected to cluster analysis, so as to obtain the large-scale user classification and grouping result.

[0067] A computer storage medium, which stores a computer program, the computer program is executed by a processor to realize the above-mentioned power user fine portrait and management method considering load and social information.

[0068] A computer device, comprising a storage, a processor and a computer program stored on the storage and executable on the processor, wherein the processor executes the computer program to realize the above-mentioned power user fine portrait and management method considering load and social information.

[0069] Advantages: compared with the prior art, the present application has the following advantages:

[0070] 1. The present application can fully exploit the value of fine-grained load data and social research information, fully evaluate the user power load characteristics and power consumption behavior consciousness, and realize multi-dimensional fine portrait of power users.

[0071] 2、The application can face the demand response scene, comprehensively evaluate the load adjustable ability of the user behavior consciousness, and realize the classification, grouping and labeling management of large-scale users by the selection of key indicators.

[0072] 3、The application can help users understand their own power consumption habits, and also provide the power enterprise with portrait theory and method, support the power enterprise to provide personalized services and auxiliary decision for users, thereby helping the integration of demand side flexible resources and assisting the construction of new energy power system flexible resource pool. BRIEF DESCRIPTION OF DRAWINGS

[0073] Figure 1 The step flow chart of the application is shown in the figure;

[0074] Figure 2 The multi-dimensional power user power consumption behavior evaluation system is shown in the figure;

[0075] Figure 3 The user adjustable potential evaluation curve example is shown in the figure;

[0076] Figure 4 The user side and grid side load matching degree example is shown in the figure;

[0077] Figure 5 The user behavior consciousness quantitative evaluation example is shown in the figure;

[0078] Figure 6 The clustering result example of a large number of users is shown in the figure, wherein, Figure 6 a is the user classification A-D group, Figure 6 b is the user classification E-H group. DETAILED DESCRIPTION

[0079] The application is a power user fine portrait and management method considering load and social information, establishes a multi-dimensional power user power consumption behavior evaluation system to realize the multi-dimensional fine portrait of a single user, proposes a load matching degree and adjustable potential calculation method based on fine-grained load data facing demand response, adopts the analytic hierarchy process and entropy weight method combined weighting method, proposes a user subjective behavior consciousness comprehensive evaluation method based on social research information, and proposes a large-scale user classification, grouping and labeling management method based on DBSCAN, realizes the comprehensive classification evaluation and management of large-scale user adjustable ability based on fine portrait.

[0080] The technical solutions of the application will be further described below in combination with the drawings and specific embodiments.

[0081] The technical solutions of the application will be further described below in combination with the drawings and specific embodiments.

[0082] As Figure 1As shown, a method for fine-grained profiling and management of power users taking into account both load and social information includes the following steps:

[0083] (1) It is proposed to establish a multi-dimensional electricity user behavior evaluation system with 4 secondary indicators and 14 tertiary indicators from the four dimensions of energy consumption and load characteristics, adjustable potential, behavioral awareness, and user nature, such as Figure 2 As shown, it is used to make a detailed portrait of a single user.

[0084] (2) By mining user electricity consumption data, a method for calculating the load matching index between users and regional power grids, as well as a method for calculating the load adjustable potential index, is proposed. The load adjustable potential index includes the transferable load potential and the curtailable load potential.

[0085] (2.1) Calculate the user's adjustable potential:

[0086] By combining the transferable load potential and the curtailable load potential, the user load power is calculated and used as an indicator of the user's adjustable potential to evaluate the adjustable potential of large-scale users. The user's adjustable potential coefficient is calculated as follows:

[0087]

[0088] in, and The number of electrical equipment representing the categories of load that can be shifted and load that can be curtailed; is the transferable load potential of the i-th equipment; is the load reduction potential of the jth device; is the total load.

[0089] (2.2) Calculate the load matching degree:

[0090] The user's daily load curve is analyzed and matched with the daily load curve of the local power grid. The similarity reflects the user's potential for participating in demand response. The load matching degree uses the Fréchet distance, which comprehensively considers the position and order of curve points, to measure the similarity between curves.

[0091] make is a metric space, let and for two given curves in ; then, the Fréchet distance between A and B is defined as: and After parameterization On a plane and The infimum of the maximum distance between , where ; then the Fréchet distance is:

[0092]

[0093] wherein, is a distance function.

[0094] (3) The user electricity consumption data is combined with the questionnaire survey results of the electricity users, from the policy sensitivity, electricity fee sensitivity, user interaction willingness, and green electricity awareness four aspects, the subjective behavior consciousness of the users is comprehensively evaluated based on the subjective and objective evaluation method combining the analytic hierarchy process and the entropy weight method.

[0095] (3.1) A hierarchical system for comprehensively evaluating the subjective behavior consciousness of the users is constructed:

[0096] Through the demand analysis of the policy sensitivity, electricity fee sensitivity, user interaction willingness, and green electricity awareness of the behavior consciousness dimensions, the corresponding problems in the survey questionnaire are selected as the trade-off basis of the four indexes, the hierarchical system is established according to the three-layer structure of the target layer, the criterion layer, and the factor layer; the answers to the questions in the survey questionnaire involving the four influencing factors are scored; the scoring principle mainly refers to the scoring method of the Likert scale, the scale is composed of a group of statements, each statement has "very agree", "agree", "not sure", "disagree", and "very disagree" five answers, which are recorded as 5, 4, 3, 2, and 1 respectively, the total score of each respondent's attitude is the sum of the scores obtained by his answers to each question, and the total score can indicate his attitude strength or his different state on the scale.

[0097] (3.2) Consistency test:

[0098] According to the expert scoring results, the and judgment matrices are constructed, which are respectively denoted as and , the order of which is the number of criterion layer dimensions and the number of factors under the first dimension respectively, and the matrix elements are the values filled by the experts according to the proportion scale table; the judgment matrix must pass the consistency test before the weight results calculated by it can be used, and the consistency test formula is as follows:

[0099]

[0100]

[0101] wherein, is the maximum eigenvalue of the matrix; is the consistency index; is the order of the matrix; ​For random consistency index, its size is related to , the specific value can be obtained by table lookup; For consistency ratio;

[0102] When , the matrix is considered to pass the consistency test, otherwise the judgment matrix should be modified.

[0103] (3.3) Determine the comprehensive evaluation weight.

[0104] (3.3.1) Analytic hierarchy process weight:

[0105] Eigenvalue method is used to derive the relative ranking weight of elements under a single criterion from the judgment matrix, that is, the eigenvector corresponding to the largest eigenvalue of the judgment matrix, after normalization, the ranking weight vector is obtained , and is the ranking weight vector of , and is the weight of the th factor under , then:

[0106]

[0107] Similarly, the ranking weight vector of can be obtained , and is the weight of , then:

[0108]

[0109] Multiplying the weight of a certain factor under with the weight of for can obtain the weight of the factor for , and the weight of the th factor in the factor layer for is denoted as .

[0110] (3.3.2) Entropy weight method:

[0111] On the basis of questionnaire scoring, the scores of each factor for users are listed; let the th user's th index be , and the normalized result be , and the proportion of the th index of the th user is calculated , according to the information entropy of the first index can be calculated , and finally the objective weight of the first index is obtained, denoted as ; The formula is as follows:

[0112]

[0113]

[0114]

[0115]

[0116] (3.3.3) Combination weighting:

[0117] The subjective weighting and objective weighting are combined, so that the finally determined weight reflects both subjective and objective information, so that the final result is more accurate.

[0118] According to the weights of the first factor obtained by the analytic hierarchy process and the entropy weight method are and , the combined weighting weight can be obtained by the combined weighting formula:

[0119]

[0120] (3.3.4) Calculate the comprehensive score of behavior consciousness:

[0121] According to the scoring situation, the score of each factor problem of each user is obtained, and combined with the weight result obtained in the combined weighting, the score of each factor is weighted and summed with the corresponding weight to obtain the final score of the user response willingness level; Through the above method, the score of the resident user response willingness level obtained after processing the natural information can reflect the user's response willingness, so as to realize more refined and efficient evaluation of large-scale user adjustable resources.

[0122] (4) According to the quantified scores of the user load data level and the subjective consciousness level obtained in steps (2) and (3), based on the DBSCAN clustering algorithm, the portrait results considering the user adjustable potential, load matching degree and subjective behavior consciousness three dimensions can be clustered and analyzed, realizing the classification and group labeling management of large-scale users for demand response.

[0123] (4.1) Initialize the core object set , initialize the cluster number , and initialize the unvisited sample set cluster division .

[0124] (4.2) For each core object , find all core objects by the following steps:

[0125] (4.2.1) Find the neighborhood sub-sample set of sample by distance metric method. .

[0126] (4.2.2) If the number of samples in the sub-sample set satisfies , add sample to the core object sample set: .

[0127] (4.3) If the core object set , the algorithm ends, otherwise go to step (4.4).

[0128] (4.4) In the core object set , randomly select a core object , initialize the current cluster core object queue , initialize the class serial number , initialize the current cluster sample set , and update the unvisited sample set .

[0129] (4.5) If the current cluster core object queue , the current cluster is generated, update the cluster division , update the core object set , and go to step (3.3); otherwise, update the core object set .

[0130] (4.6) Take a core object from the current cluster core object queue , find all neighborhood sub-sample sets by the neighborhood distance threshold , let , update the current cluster sample set , update the unvisited sample set , update , and go to step (4.5).

[0131] In the whole process, the input is the sample set , the neighborhood parameter , and the sample distance metric method; the output result is the cluster division .​​

[0132] Through the above steps, the data obtained in steps (2) and (3) can be subjected to cluster analysis, thereby obtaining a large-scale user classification and grouping result.

[0133] Embodiment:

[0134] As shown in Figure 3 , the user adjustable potential evaluation curve can be obtained according to step (2). Taking the data of a typical family in Suzhou in summer 2019 as an example, the adjustable potential is evaluated, and the data-driven load characteristic coefficient is calculated. Figure 3 is a single-day load curve of a user, the abscissa is the total of 96 time points recorded at fifteen-minute intervals throughout the day, and the ordinate is the normalized load of the user at each time point. The calculated TPC = 86.72, and the user has a high adjustable potential among urban users.

[0135] As shown in Figure 4 , the load matching degree curve of the user side and the grid side can be obtained according to step (2). According to the user daily load curve and the Suzhou load standard, the Fréchet distance is used to reflect the similarity between the user load curve and the total grid load curve. The Fréchet distance is 1.376. Since the higher the load matching degree between the user and the grid, the greater the response potential of the user at peak and valley times, the higher the curve matching degree, and the higher the index score of the user. After de-scaling and percentile indexing, an index of 66.51 is obtained.

[0136] As shown in Figure 5 , the comprehensive score of user behavior awareness can be obtained according to step (3). In this embodiment, the analytic hierarchy process is first used to analyze the target layer, the criterion layer and the factor layer, and the weights of the 19 problems in the factor layer are obtained. Secondly, the entropy weight method is used to analyze the example, and the weights of the 19 problems can also be obtained. Thirdly, the combined weighting method is used to combine the analytic hierarchy process and the entropy weight method to obtain the final weights of the 19 problems. Combining the score of the scheme layer and the weights of the four dimensions of behavior awareness, the weighted sum is performed to obtain the index of each dimension of the user and the total index of the user's subjective awareness of electricity consumption. Finally, the result is presented through a radar chart. The index on the Figure Four axis represents the scores of the user's policy sensitivity, electricity cost sensitivity, user interaction willingness, and green electricity awareness. The middle index represents the total score of the user's subjective awareness of electricity consumption.

[0137] As shown in Figure 6As shown, the clustering result map of a large number of users can be obtained according to step (4). Taking the load of 504 households in Suzhou, Jiangsu Province as a sample, the historical fine-grained household appliance load data measured based on the Internet of Things technology and the questionnaire results of "Jiangsu Province Urban Residential User Basic Information and Participation in Demand and Supply Interaction Willingness Survey" are used to construct the user portrait. By clustering analysis on the quantification results of the load matching degree, the user interaction willingness and the adjustable potential of the users, the 504 users are classified into different user groups to help the load aggregator to take different auxiliary decisions for users of different categories according to the actual situation. Among the 504 users, there are 39 high-quality users A, accounting for 7.7% of the total number of users, with an estimated adjustable potential of 293kW·h / day; there are 57 first-level potential users B, accounting for 11.3% of the total number of users, with an estimated adjustable potential of 170kW·h / day; there are 57 first-level potential users C, accounting for 11.3% of the total number of users, with an estimated adjustable potential of 435kW·h / day; there are 86 second-level potential users D, accounting for 17.1% of the total number of users, with an estimated adjustable potential of 254kW·h / day; there are 74 second-level potential users E, accounting for 14.7% of the total number of users, with an estimated adjustable potential of 490kW·h / day; there are 62 first-level potential users G, accounting for 12.3% of the total number of users, with an estimated adjustable potential of 457kW·h / day; there are 129 poor-quality users H and F, accounting for 25.6% of the total number of users, with an estimated adjustable potential of 382kW·h / day.

[0138] It can be seen that the present application can comprehensively depict the user's electricity consumption behavior from the aspects of load data and social information, and evaluate the users with deep potential, high interaction willingness and objective interaction conditions. Analyzing the user's electricity consumption behavior helps the power enterprise to understand the interaction willingness and behavior consciousness of the users in the jurisdiction, and clustering analysis on a large number of users helps the load aggregator to more intuitively and quickly distinguish different types of users, which is of great significance for the power enterprise to formulate interactive user invitation set and demand response strategy, and to a certain extent, can improve the flexibility of the new power system.

Claims

1. A method for fine profiling and management of power users considering load and social information, characterized by, Comprise the following steps: (1) Propose to establish a multi-dimensional power user electricity consumption behavior evaluation system from four dimensions of energy consumption and load characteristics, adjustable potential, behavior awareness, and user nature, four secondary indicators, and 14 tertiary indicators, to finely profile individual users; the step (1) is specifically: (1.1) Establish four dimensions of power user electricity consumption behavior evaluation: Multi-dimensional power user fine profiling is performed from four dimensions of energy consumption and load characteristics, adjustable potential, behavior awareness, and user nature, as secondary indicators of power user electricity consumption behavior evaluation; (1.2) Establish 14 tertiary indicators of power user electricity consumption behavior evaluation: Four tertiary indicators of daily load curve, peak-valley characteristics, load matching degree, and energy consumption level are considered under the secondary indicator of energy consumption and load characteristics; Two tertiary indicators of transferable load potential and reducible load potential are considered under the secondary indicator of adjustable potential; Four tertiary indicators of policy sensitivity, electricity fee sensitivity, user interaction willingness, and green electricity awareness are considered under the secondary indicator of behavior awareness; Three tertiary indicators of user electricity consumption habit, user family attribute, and new energy attribute are considered under the secondary indicator of user nature; (2) Through mining of user electricity consumption data, propose a user and regional power grid load matching degree index calculation method, and a load adjustable potential index calculation method, the load adjustable potential index includes transferable load potential and reducible load potential; the step (2) is specifically: (2.1) Calculate the adjustable potential of the user: By combining the transferable load potential and the reducible load potential, the user load power is calculated, and it is used as an indicator of user adjustable potential to evaluate the adjustable potential of large-scale users; the user adjustable potential coefficient is calculated as follows: ， wherein, and respectively represent the number of electrical devices in the transferable load and the curta- ble load categories; is the transferable load potential for the i-th device; is the curta-ble load potential for the j-th device; is the total load; (2.2) Calculate the load matching degree: The user daily load curve is analyzed and matched with the daily load curve of the local power grid, and the similarity degree is used to reflect the potential of the user to participate in demand response; the load matching degree uses the Fréchet distance considering the curve point position and order to measure the similarity between curves; make is a metric space, let and for two given curves in ; then, the Fréchet distance between A and B is defined as: and After parameterization On a plane and The infimum of the maximum distance between , where ; then the Fréchet distance yes: ， wherein is a distance function; (3) Combine the user electricity consumption data with the questionnaire survey results of the electricity user, and comprehensively evaluate the user's subjective behavior awareness from four aspects of policy sensitivity, electricity fee sensitivity, user interaction willingness, and green electricity awareness based on the subjective and objective evaluation method combining the analytic hierarchy process and entropy weight method; (4) Based on the quantitative scores of the user load data level and the subjective awareness level obtained in steps (2) and (3), the clustering algorithm based on DBSCAN can be used to cluster the profile results considering the user adjustable potential, load matching degree, and subjective behavior awareness three dimensions, to realize the classification and group labeling management of large-scale users for demand response. 2.The method of claim 1, wherein, The step (3) is specifically: (3.1) Build a hierarchical system for comprehensive evaluation of user subjective behavior awareness: The evaluation demand analysis of the policy sensitivity of the behavior consciousness dimension, the electricity fee sensitivity, the user interaction willingness and the green power consumption consciousness is selected as the trade-off basis of the four indexes in the questionnaire, and a hierarchical system is established according to the three-layer structure of the target layer, the criterion layer and the factor layer; the answers to the questions in the questionnaire related to the four influencing factors are scored; the scoring principle mainly refers to the scoring method of the Likert scale, which is composed of a group of statements, each statement has "very agree", "agree", "not sure", "disagree" and "very disagree", which are marked as 5, 4, 3, 2 and 1 respectively, and the total score of each respondent is the sum of the scores obtained by answering each question, which can indicate the strength of his attitude or his different state on this scale; (3.2) Consistency test: According to the expert scoring results to build and Judgment matrix, respectively, are and The order of the criterion layer dimension number and the first The number of factors under the dimension, the matrix elements are filled by experts according to the scale table value; The consistency test formula of the judgment matrix must pass the consistency test before the weight result can be used, and the consistency test formula is as follows: ， ， wherein, is the largest eigenvalue of the matrix; is the uniformity index; is the order of the matrix; is the stochastic uniformity index, which is related to and can be obtained by looking up a table for the specific value; is the uniformity ratio; When the matrix is considered to pass the consistency test, otherwise the judgment matrix should be revised; (3.3) Determine the comprehensive evaluation weight: (3.3.1) Analytic hierarchy process weight: The eigenvalue method is used to derive the relative ranking weight of the elements under a single criterion from the judgment matrix, i.e. the eigenvector corresponding to the largest eigenvalue of the judgment matrix , which, after normalization, gives the ranking weight vector , denoted as , and , the ranking weight vector of the th factor, , the weight of the th factor, then: ， Likewise, it is possible to obtain a ranking vector , for the weights of , then: ， Will The weight of a factor is for Multiplying the weight of the factor can give the effect of The weight of the factor layer Factors for The weight of ; (3.3.2) Entropy weight: Based on the questionnaire scores, The score of each factor for each user; User's The indicator is , the normalized result is , use the normalized results to calculate the User's The proportion of indicators ,according to It can also calculate the Information entropy of an indicator , and finally get The objective weight of an indicator is recorded as ; The formula is as follows: ， ， ， ， (3.3.3) Combination weighting: The subjective weighting and the objective weighting are combined to make the final determined weight reflect both subjective and objective information, so that the final result is more accurate; According to the analytic hierarchy process and the entropy weight method, the weights of the first factors are respectively , and , and the combined weight of the combined weight formula is : ， (3.3.4) Calculate the comprehensive score of behavior consciousness: According to the scoring, the score of each factor question of each user is obtained, and combined with the weight result obtained in the combination weighting, the final score of the user response willingness level can be obtained by weighting and summing the scores of each factor and the corresponding weight; the score of the user response willingness level obtained by the above method after processing the natural information can reflect the user's response willingness, so as to realize more detailed and efficient evaluation of large-scale user adjustable resources. 3.The method of claim 1, wherein, The step (4) is specifically: (4.1) initializing a set of core objects , initializing a number of clustering clusters , initializing a set of unvisited samples , cluster partitioning ; (4.2) For , find all core objects in the following steps: (4.2.1) finding the sample of neighborhood sub-sample set by distance metric (4.2.2) If the number of samples in the sub-sample set satisfies , the sample is added to the core object sample set: ; (4.3) If the set of core objects then the algorithm ends, otherwise go to step (4.4); (4.4) In the core object set , randomly select one core object , initialize the current cluster core object queue , initialize the category number , initialize the current cluster sample set , update the unvisited sample set ; (4.5) If the current cluster core object queue is empty, the current cluster is generated, the cluster division is updated , the core object set is updated , and the process goes to step (3.3); otherwise, the core object set is updated ; (4.6) Take out one core object from the current cluster core object queue Go to step (4.5);​​​​​​​​ Throughout the process, the input is a set of samples , neighborhood parameters and a sample distance metric; the output is a cluster partition ​ Through the above steps, the data obtained in steps (2) and (3) can be clustered and analyzed to obtain the classification and grouping results of large-scale users.

4. A computer storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to realize the method for fine portrait and management of power users considering load and social information according to any one of claims 1-3.

5. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the method for fine portrait and management of power users considering load and social information according to any one of claims 1-3.

Images