Multi-time-scale user power consumption behavior portraying method based on hybrid power consumption characteristics

Abstract

The invention provides a multi-time-scale user power consumption behavior portraying method based on hybrid power consumption characteristics, and particularly relates to the technical field of power user power consumption behavior analysis and modeling. The method specifically comprises the following steps: S1, constructing a power utilization time sequence data set of power consumers under multiple time scales; s2, constructing a user electricity consumption fluctuation characteristic evaluation index system; s3, quantitatively representing continuous-discrete hybrid power utilization characteristics under different time scales; s4, constructing a K-prototypes power user division model, and determining an optimal user classification number under different time scales; s5, continuous power consumption fluctuation images under different time scales are drawn, and continuous power consumption behavior class labels of each class of users are defined; s6, drawing discrete power utilization characteristic radar maps under different time scales; and S7, forming power consumption behavior personality portraits of the power consumers under different time scales. According to the method, accurate division and power consumption behavior portraying of the power users can be realized, and a reference is provided for an electric power company to formulate personalized power supply services for users in different power consumption modes.

Description

Technical Field

[0001] This invention patent relates to the technical field of power user electricity consumption behavior analysis and modeling, and specifically relates to a method for multi-time scale user electricity consumption behavior profiling based on hybrid electricity consumption characteristics. Background Art

[0002] With the acceleration of the electrification process and the continuous improvement of the informatization level of the power grid, the information integration system of the power department has accumulated a large amount of electricity consumption data. And each power load presents the characteristics of diversity and complexity, and the electricity consumption characteristics of different industries and types of power users have differences and similarities. In order to adapt to the development trend of refined management and dispatching on the user side, it is urgent to conduct user electricity consumption behavior profiling according to electricity consumption characteristics, so as to provide reference for power companies to formulate personalized power supply services for users with different electricity consumption patterns. Summary of the Invention

[0003] To overcome the deficiencies of the above-mentioned prior art, the present invention provides a method for multi-time scale user electricity consumption behavior profiling based on hybrid electricity consumption characteristics. The specific technical solutions are as follows:

[0004] A method for multi-time scale user electricity consumption behavior profiling based on hybrid electricity consumption characteristics specifically includes the following steps: S1. Construct an electricity consumption time series data set of power users at multi-time scales; S2. Construct an evaluation index system for the electricity consumption fluctuation characteristics of power users; S3. Quantitatively characterize the continuous-discrete hybrid electricity consumption characteristics of power users at different time scales based on the electricity consumption time series data set and the evaluation index system for electricity consumption fluctuation characteristics; S4. Based on the continuous-discrete hybrid electricity consumption characteristics of power users at different time scales obtained by quantitative characterization in S3, construct a K-prototypes power user classification model to determine the optimal number of user classifications at daily, monthly, and annual time scales; S5. Based on the K-prototypes power user classification model, draw the continuous electricity consumption fluctuation images of power users at different time scales and define the category labels of the continuous electricity consumption behaviors of each type of user at different time scales according to the image fluctuation characteristics; S6. Based on the discrete hybrid electricity consumption characteristics of power users at different time scales obtained by quantitative characterization in S3, draw the discrete electricity consumption characteristic radar charts of power users at different time scales; S7. Form the electricity consumption behavior personality profiles of power users at different time scales.

[0005] Preferably, for S1 to construct the electricity consumption time series data set of power users at multi-time scales, it specifically includes the following sub-steps:

[0006] S1.1 Select N power users, divide the electricity consumption data within a typical day into 96 time periods with a time step of 15 minutes. Among them, the power meter power data set of the nth user every 15 minutes on the ath day collected in real time is:

[0007] Pn_a_day = [P n,a,1po int P n,a,2po int … P n,a,96po int ;

[0008] where 1 ≤ a ≤ 365; n is the user number, n = 1, 2, 3, … N; N is the total number of users;

[0009] S1.2 Use the elimination method and the mean filling method to process the outliers in the electricity meter power dataset obtained in S1.1, and calculate the time series dataset of the electricity consumption per 15 minutes on the ath day of the nth user's real-time collection as:

[0010] E n_a_day = P n_a_day ·t = [E n,a,1po int E n,a,2po int … E n,a,96po int ;

[0011] In the formula, t = 0.25; 1 ≤ a ≤ 365; n is the user number, n = 1, 2, 3, … N;

[0012] S1.3 Take the ath day as the typical day, then the time series dataset of the electricity consumption per 15 minutes on the ath day of N users at the daily time scale is:

[0013]

[0014] S1.4 Take the month A where the ath day is located as the typical month and assume that there are m days in month A, then the time series dataset of the daily electricity consumption of the nth user in month A is:

[0015] E n_A_month = [E n,A,1day E n,A,2day … E n,A,mday ;

[0016] In the formula, E n,A,mday is the electricity consumption of the nth user on the mth day of month A, then the time series dataset of the daily electricity consumption of N users in month A at the monthly time scale is:

[0017]

[0018] S1.5 According to 1.4, the time series dataset of the monthly electricity consumption of the nth user in a year is:

[0019] E n_year = [E n,1month E n,2month … E n,12month ;

[0020] In the formula, E n,A,1day , …, E n,12monthLet \(E_{n1}, E_{n2}, \cdots, E_{n12}\) be the electricity consumption of the \(n\)th user in January, February, \(\cdots\), December respectively. Then, the time - series dataset of the monthly electricity consumption of \(N\) users within a year on an annual time scale is:

[0021]

[0022] Preferably, in S2, constructing the evaluation index system for the electricity consumption fluctuation characteristics of electricity users specifically includes the following sub - steps:

[0023] S2.1 On a daily time scale, according to the formula for the average volatility of every 15 minutes of the \(n\)th user on the \(a\)th day The formula for the peak - valley difference of load \(K\) n \(=P\) n,max \(-P\) n,min And the formula for the maximum fluctuation amount of adjacent 1 - hour \(\Delta E\) max n \(=\max\{\Delta E\) n,1 ,\(\Delta E\) n,2 \(\cdots\Delta E\) n,i \}\), the evaluation index dataset of the \(n\)th user on a daily time scale is:

[0024]

[0025] Then, the evaluation index dataset of \(N\) users on a daily time scale is:

[0026]

[0027] In the formula, \(1\leq i\leq95\), \(\eta\) n,i is the electricity consumption volatility of every 15 minutes of the \(n\)th user on the \(a\)th day; \(P\) n,max is the maximum load power of the \(n\)th user on the \(a\)th day, \(P\) n,min is the minimum load power of the \(n\)th user on the \(a\)th day; \(\Delta E\) n,1 , \(\Delta E\) n,2 \(\cdots\Delta E\) n,i are the fluctuation amounts of the electricity consumption of the \(n\)th user between adjacent two - hour periods on the \(a\)th day;

[0028] S2.2 On a monthly time scale, according to the formula for the maximum electricity consumption fluctuation amount between adjacent two days of the \(n\)th user in the \(A\)th month \(\Delta E'\) max n \(=\max\{\Delta E'\) n,1 , \(\Delta E'\) n,2 , \(\Delta E'\) n,3 \(\cdots\Delta E'\) n,m-1 \}\), the formula for the fluctuation amount of the average electricity consumption between weekdays and non - weekdays and the formula for the daily load peak - valley difference rate The evaluation index dataset of the \(n\)th user on a monthly time scale is:

[0029] W n_month= [ΔE′ max n ΔE n μ n ;

[0030] Then the evaluation index data set of N users on a monthly time scale is:

[0031]

[0032] Where n = 1, 2, 3,... N; ΔE′ n,1 , ΔE′ n,2 …ΔE′ n,m-1 are respectively the electricity consumption fluctuation amounts of the nth user on two adjacent days in the A-th month; is the average value of the electricity consumption of the nth user on all working days in the A-th month, is the average value of the electricity consumption of the nth user on all non-working days in the A-th month; E′ n,max is the maximum daily electricity consumption of the nth user in the A-th month, E′ n,min is the minimum daily electricity consumption of the nth user in the A-th month;

[0033] S2.3 On an annual time scale, according to the annual load factor calculation formula of the nth user in that year and the maximum load utilization hours calculation formula and together with the rated capacity V n constitute the evaluation index data set of the nth user on an annual time scale as:

[0034] W n_year = [R n T max n V n ;

[0035] Then the evaluation index data set of N users on an annual time scale is:

[0036]

[0037] Where n = 1, 2, 3,... N; E n,av is the average daily electricity consumption of the nth user throughout the year, E n,max is the maximum daily electricity consumption of the nth user in that year; ΣE n is the total annual electricity consumption of the nth user, P′ n,max is the 1-hour maximum load power of the nth user.

[0038] Preferably, in S3, based on the electricity consumption time series data set and the evaluation index system of electricity consumption fluctuation characteristics, quantitatively characterize the continuous-discrete hybrid electricity consumption characteristics of power users at different time scales, specifically including the following sub-steps:

[0039] S3.1 N-user evaluation index dataset based on daily time scale Use the GMM clustering algorithm to divide W day The average volatility of the N users at every 15 minutes on the a-th day in the first column into a1 interval levels from small to large, the load peak-valley difference of the N users on the a-th day in the second column into b1 interval levels from small to large, and the maximum fluctuation amount between adjacent 1 hour of the N users on the a-th day in the third column into c1 interval levels from small to large. W day The interval level corresponding to the column where each value in it is located constitutes the discrete electricity consumption feature dataset on the daily time scale as follows:

[0040]

[0041] In the formula, x n_day is the interval level of the n-th user's in the first column of data, and x n_day ranges from 1 to a1; y n_day is the interval level of the n-th user's K n in the second column of data, and y n_day ranges from 1 to b1; z n_day is the interval level of the n-th user's ΔE max n in the third column of data, and z n_day ranges from 1 to c1;

[0042] Merge the discrete electricity consumption feature dataset W' day on the daily time scale with the electricity consumption time series dataset E of the N users at every 15 minutes on the a-th day day to obtain the continuous-discrete hybrid electricity consumption feature dataset on the daily time scale as follows:

[0043]

[0044] S3.2 N-user evaluation index dataset based on monthly time scale Use the GMM clustering algorithm to divide W month The maximum electricity consumption fluctuation between adjacent two days of the N users in the A-th month in the first column into a2 interval levels from small to large, the fluctuation amount of the average electricity consumption of the N users on working days and non-working days in the A-th month in the second column into b2 interval levels from small to large, and the daily load peak-valley difference rate of the N users in the A-th month in the third column into c2 interval levels from small to large. W month The interval level corresponding to the column where each value in it is located constitutes the discrete electricity consumption feature dataset on the monthly time scale as follows:

[0045]

[0046] In the formula, x n_month is the ΔE' of the n-th usermax n The interval level in the first column of data, x n_month The value range is from 1 to a2; y n_month Is the ΔE of the nth user n The interval level in the second column of data, y n_month The value range is from 1 to b2; z n_month Is the μ of the nth user n The interval level in the third column of data, z n_month The value range is from 1 to c2;

[0047] The discrete electricity consumption feature dataset W′ at the monthly time scale month And the time series dataset E of the daily electricity consumption of N users in the A-th month month Merging them can obtain the continuous-discrete hybrid electricity consumption feature dataset at the monthly time scale as follows:

[0048]

[0049] S3.3 Based on the evaluation index dataset of N users at the annual time scale Using the GMM clustering algorithm, divide the annual load factor of N users in the first column of W year from small to large into a3 interval levels, divide the maximum load utilization hours of N users in the second column from small to large into b3 interval levels, divide the rated capacity of N users in the third column from small to large into c3 interval levels. The interval level corresponding to each value in the column of W year constitutes the discrete electricity consumption feature dataset at the annual time scale as follows:

[0050]

[0051] In the formula, x n_year Is the R of the nth user n The interval level in the first column of data, x n_year The value range is from 1 to a3; y n_year Is the T of the nth user max n The interval level in the second column of data, y n_year The value range is from 1 to b3; z n_year Is the V of the nth user n The interval level in the third column of data, z n_year The value range is from 1 to c3;

[0052] The discrete electricity consumption feature dataset W′ at the annual time scale year And the time series dataset E of the monthly electricity consumption of N users within one year year Merging them can obtain the continuous-discrete hybrid electricity consumption feature dataset at the annual time scale as follows:

[0053]

[0054] Preferably, in S4, based on the continuous-discrete hybrid electricity consumption characteristics at different time scales obtained after quantization in S3, a K-prototypes power user classification model is constructed to determine the optimal number of user classifications at daily, monthly, and annual time scales, specifically including the following sub-steps:

[0055] S4.1 Construct a K-prototypes power user classification model and input the continuous-discrete hybrid electricity consumption characteristic datasets W″ day 、W″ month and W″ year ;

[0056] S4.2 Perform Z-score standardization on the continuous electricity consumption characteristics E day at the daily time scale, adjust the weights between continuous and discrete electricity consumption characteristics, and use the elbow method to determine the optimal number of user classifications k day ;

[0057] Among them, the calculation formula for the weight factor at the daily time scale is:

[0058]

[0059] In the formula, σ day,n is the standard deviation of the continuous electricity consumption data of the nth user at the daily time scale after standardization;

[0060] S4.3 Perform Z-score standardization on the continuous electricity consumption characteristics E month at the monthly time scale, adjust the weights between continuous and discrete electricity consumption characteristics, and use the elbow method to determine the optimal number of user classifications k month ;

[0061] Among them, the calculation formula for the weight factor at the monthly time scale is:

[0062]

[0063] In the formula, σ month,n is the standard deviation of the continuous electricity consumption data of the nth user at the monthly time scale after standardization;

[0064] S4.4 Perform Z-score standardization on the continuous electricity consumption characteristics E year at the annual time scale, adjust the weights between continuous and discrete electricity consumption characteristics, and use the elbow method to determine the optimal number of user classifications k year ;

[0065] Among them, the calculation formula for the weight factor at the annual time scale is:

[0066]

[0067] wherein, σ year,n is the standard deviation of the continuous electricity consumption data of the nth user under the annual time scale after standardization.

[0068] Further preferably, in S5, based on the partitioning model, draw the continuous electricity consumption fluctuation images of electricity users at different time scales and define the category labels of continuous electricity consumption behaviors at different time scales according to the image fluctuation characteristics, which specifically include the following sub-steps:

[0069] S5.1 Based on the K-prototypes user partitioning model, construct the horizontal axis with a time step of 15 minutes, and construct the vertical axis with the standard deviation of the continuous electricity consumption data after standardization at the daily time scale, and draw the continuous electricity consumption fluctuation images of the first category, the second category,..., the kth day category of users at the daily time scale, and summarize the category labels of the continuous electricity consumption behaviors of each category of users at the daily time scale from the aspects of periodicity, increase and decrease, peak value, valley value, and electricity consumption characteristics in the early, middle, and late periods of the day;

[0070] S5.2 Construct the horizontal axis with a time step of one day, and construct the vertical axis with the standard deviation of the continuous electricity consumption data after standardization at the monthly time scale, and draw the continuous electricity consumption fluctuation images of the first category, the second category,..., the kth month category of users at the monthly time scale, and summarize the category labels of the continuous electricity consumption behaviors of each category of users at the monthly time scale from the aspects of periodicity, increase and decrease, peak value, valley value, and electricity consumption characteristics at the beginning, middle, and end of the month;

[0071] S5.3 Construct the horizontal axis with a time step of one month, and use the standardized continuous electricity consumption characteristics at the annual time scale as the vertical axis, and draw the continuous electricity consumption fluctuation images of the first category, the second category,..., the kth year category of users at the annual time scale, and summarize the category labels of the continuous electricity consumption behaviors of each category of users at the annual time scale from the aspects of periodicity, increase and decrease, peak value, valley value, and electricity consumption characteristics in different seasons of spring, summer, autumn, and winter within the year.

[0072] Even further preferably, in S6, based on the discrete mixed electricity consumption characteristics of electricity users at different time scales obtained after quantization in S3, draw the radar charts of the discrete electricity consumption characteristics of electricity users at different time scales, which specifically include the following sub-steps:

[0073] S6.1 Based on the discrete electricity consumption characteristic dataset W' at the daily time scale in S3.1 day, draw the radar chart of the discrete electricity consumption characteristics of each user on the daily time scale, and through the shape and size of the radar chart on the daily time scale, the strengths of the three discrete electricity consumption characteristics of the average volatility, load peak-valley difference, and maximum fluctuation volume in the adjacent 1 hour of a single user on the same day can be intuitively reflected;

[0074] S6.2 Based on the discrete electricity consumption characteristic dataset W' on the monthly time scale in S3.2 month , draw the radar chart of the discrete electricity consumption characteristics of each user on the monthly time scale, and through the shape and size of the radar chart on the monthly time scale, the strengths of the three discrete electricity consumption characteristics of the maximum electricity consumption fluctuation volume between adjacent two days, the fluctuation volume of the average electricity consumption on working days and non-working days, and the daily load peak-valley difference rate of a single user can be intuitively reflected;

[0075] S6.3 Based on the discrete electricity consumption characteristic dataset W' on the annual time scale in S3.3 year , draw the radar chart of the discrete electricity consumption characteristics of each user on the annual time scale, and through the shape and size of the radar chart on the annual time scale, the strengths of the three discrete electricity consumption characteristics of the annual load rate, maximum load utilization hours, and rated capacity of a single user in the current year can be intuitively reflected.

[0076] Even more preferably, S7 forms the personalized portrait of the electricity consumption behavior of power users based on the radar charts of the discrete electricity consumption characteristics at different time scales obtained from S6.1-S6.3 and the category labels of the continuous electricity consumption behavior at the time scales defined in S5.1-S5.3, combined with the industry type of the power users themselves.

[0077] The beneficial effects of the present invention are:

[0078] The method for drawing the portrait of the electricity consumption behavior of power users with multiple time scales provided by the present invention comprehensively considers the continuous fluctuation characteristics, discrete electricity consumption characteristics of the load data itself at multiple time scales, and multiple load characteristic evaluation indexes that can represent the electricity consumption behavior of users, and combines the user's basic attribute of the industry type to draw a multi-faceted portrait of power users, providing an effective reference for power companies to improve the power supply service quality more accurately and effectively. Brief Description of the Drawings

[0079] The attached drawings constituting the specification of the present invention are used to provide a further understanding of the present application and do not constitute an improper limitation to the present application.

[0080] Figure 1 is the flow chart of the method of the present invention;

[0081] Figures 2(a) - 2(d) is the continuous electricity consumption fluctuation image of each type of user on the daily time scale in this embodiment;

[0082] Figures 3(a) - 3(d)For each type of user, it is the continuous power consumption fluctuation image under the monthly time scale in this embodiment;

[0083] Figures 4(a) - 4(e) For each type of user, it is the continuous power consumption fluctuation image under the annual time scale in this embodiment;

[0084] Figure 5 For user 1 in this embodiment, it is the radar chart of discrete power consumption characteristics under the daily time scale;

[0085] Figure 6 For user 1 in this embodiment, it is the radar chart of discrete power consumption characteristics under the monthly time scale;

[0086] Figure 7 For user 1 in this embodiment, it is the radar chart of discrete power consumption characteristics under the annual time scale;

[0087] Figure 8 For user 1 in this embodiment, it is the personality portrait under different time scales. Specific implementation manners

[0088] The specific implementation manners of the method for multi-time-scale user power consumption behavior portrait based on hybrid power consumption characteristics provided by the present invention will be further described in conjunction with the accompanying drawings and embodiments.

[0089] As Figure 1 shown, a method for multi-time-scale user power consumption behavior portrait based on hybrid power consumption characteristics, which is characterized in that it specifically includes the following steps:

[0090] S1. Construct a power consumption time series data set of power users under multiple time scales, which specifically includes the following sub-steps:

[0091] S1.1 Select N power users, and divide the typical daily power consumption data into 96 time periods with a 15-minute time step. The power meter power data set for every 15 minutes of the a-th day collected in real time by the n-th user is:

[0092] P n_a_day = [P n,a,1po int P n,a,2po int … P n,a,96po int ];

[0093] Wherein, 1 ≤ a ≤ 365; n is the user number, n = 1, 2, 3, … N; N is the total number of users;

[0094] S1.2 Use the elimination method and the mean filling method to process the outliers of the power meter power data set obtained in S1.1, and calculate the power consumption time series data set for every 15 minutes of the a-th day collected in real time by the n-th user as:

[0095] E n_a_day = P n_a_day · t = [En,a,1po int E n,a,2po int … E n,a,96po int ;

[0096] Where t = 0.25; 1 ≤ a ≤ 365; n is the user number, n = 1, 2, 3, … N;

[0097] S1.3 Taking the ath day as the typical day, the time series dataset of the electricity consumption of N users every 15 minutes on the ath day at the daily time scale is:

[0098]

[0099] S1.4 Taking the month A where the ath day is located as the typical month and assuming that the month A has m days, the time series dataset of the daily electricity consumption of the nth user in the month A is:

[0100] E n_A_month =[E n,A,1day E n,A,2day … E n,A,mday ;

[0101] Where E n,A,mday is the electricity consumption of the nth user on the mth day of the month A, then the time series dataset of the daily electricity consumption of N users in the month A at the monthly time scale is:

[0102]

[0103] S1.5 According to 1.4, the time series dataset of the monthly electricity consumption of the nth user in a year is:

[0104] E n_year =[E n,1month E n,2month … E n,12month ;

[0105] Where E n,A,1day , …, E n,12month are the electricity consumptions of the nth user in January, February, … December respectively, then the time series dataset of the monthly electricity consumption of N users in a year at the annual time scale is:

[0106]

[0107] S2. Construct an evaluation index system for the electricity consumption fluctuation characteristics of power users, which specifically includes the following sub-steps:

[0108] S2.1 At the daily time scale, according to the calculation formula of the average volatility of every 15 minutes of the nth user on the ath day The calculation formula of the load peak-valley difference K n =P n,max -P n,minAnd the calculation formula for the maximum fluctuation amount within adjacent 1 hour, ΔE max n = max{ΔE n,1 , ΔE n,2 … ΔE n,i}, the evaluation index data set of the nth user on the daily time scale is:

[0109]

[0110] Then the evaluation index data set of N users on the daily time scale is:

[0111]

[0112] In the formula, 1 ≤ i ≤ 95; η n,i is the electricity consumption volatility per 15 minutes on the ath day of the nth user; P n,max is the maximum load power on the ath day of the nth user, P n,min is the minimum load power on the ath day of the nth user; ΔE n,1 , ΔE n,2 ... ΔE n,i are the fluctuation amounts of electricity consumption between adjacent two hours on the ath day of the nth user respectively;

[0113] S2.2 On the monthly time scale, according to the calculation formula for the maximum electricity consumption fluctuation amount between adjacent two days of the nth user in the Ath month, ΔE′ max n = max{ΔE′ n,1 , ΔE′ n,2 , ΔE′ n,3 … ΔE′ n,m-1}, the calculation formula for the fluctuation amount of the average electricity consumption on weekdays and non - weekdays and the daily load peak - valley difference rate formula the evaluation index data set of the nth user on the monthly time scale is:

[0114] W n_month = [ΔE′ max n ΔE n μ n ;

[0115] Then the evaluation index data set of N users on the monthly time scale is:

[0116]

[0117] In the formula, n = 1, 2, 3, … N; ΔE′ n,1 , ΔE′ n,2 … ΔE′ n,m-1 are the electricity consumption fluctuation amounts between adjacent two days of the nth user in the Ath month respectively; is the average value of electricity consumption within all weekdays of the nth user in the Ath month, is the average power consumption of the nth user on all non-working days in the A-th month; E' n,max is the maximum daily power consumption of the nth user in the A-th month, E' n,min is the minimum daily power consumption of the nth user in the A-th month;

[0118] S2.3 On the annual time scale, according to the annual load factor calculation formula of the nth user in the current year and the maximum load utilization hours calculation formula and together with the rated capacity V n constitute the evaluation index data set of the nth user on the annual time scale as:

[0119] W n_year =[R n T max n V n ;

[0120] Then the evaluation index data set of N users on the annual time scale is:

[0121]

[0122] In the formula, n = 1, 2, 3,... N; E n,av is the average daily power consumption of the nth user throughout the year, E n,max is the maximum daily power consumption of the nth user in the current year; ∑E n is the total annual power consumption of the nth user, P' n,max is the maximum load power of the nth user in 1 hour.

[0123] S3. Based on the power consumption time series data set and the power consumption fluctuation characteristic evaluation index system, quantitatively characterize the continuous-discrete hybrid power consumption characteristics of power users at different time scales, specifically including the following sub-steps:

[0124] S3.1 Based on the evaluation index data set of N users on the daily time scale Use the GMM clustering algorithm to divide the average volatility of each 15 minutes of the N users on the a-th day in the first column of W day from small to large into a1 interval levels, divide the load peak-valley difference of the N users on the a-th day in the second column from small to large into b1 interval levels, divide the maximum fluctuation amount between adjacent 1 hour of the N users on the a-th day in the third column from small to large into c1 interval levels, and the interval level corresponding to the column where each value in W day is located constitutes the discrete power consumption characteristic data set on the daily time scale as:

[0125]

[0126] In the formula, x n_day is of the nth user The interval level in the first column of data, x n_day The value range is from 1 to a1; y n_day is K of the nth user n The interval level in the second column of data, y n_day The value range is from 1 to b1; z n_day is ΔE of the nth user max n The interval level in the third column of data, z n_day The value range of... is from 1 to c1;

[0127] The discrete electricity consumption feature dataset W' on the daily time scale day is merged with the electricity consumption time series dataset E of N users every 15 minutes on the ath day day to obtain the continuous-discrete hybrid electricity consumption feature dataset on the daily time scale as follows:

[0128]

[0129] S3.2 Based on the evaluation index dataset of N users on the monthly time scale Apply the GMM clustering algorithm to W month In the first column, the maximum electricity consumption fluctuations of N users on two adjacent days in the Ath month are divided into a2 interval levels from small to large. In the second column, the fluctuations of the average electricity consumption of N users on weekdays and non-weekdays in the Ath month are divided into b2 interval levels from small to large. In the third column, the daily load peak-valley difference rates of N users in the Ath month are divided into c2 interval levels from small to large. The interval level corresponding to each value in the column of W month constitutes the discrete electricity consumption feature dataset on the monthly time scale as follows:

[0130]

[0131] In the formula, x n_month is ΔE' of the nth user max n The interval level in the first column of data, x n_month The value range is from 1 to a2; y n_month is ΔE of the nth user n The interval level in the second column of data, y n_month The value range is from 1 to b2; z n_month is μ of the nth user n The interval level in the third column of data, z n_month The value range is from 1 to c2;

[0132] The discrete electricity consumption feature dataset W' on the monthly time scale month is merged with the daily electricity consumption time series dataset E of N users in the Ath month monthThe combined continuous-discrete hybrid electricity consumption feature dataset at the monthly time scale is as follows:

[0133]

[0134] S3.3 Evaluation index dataset of N users at the annual time scale Using the GMM clustering algorithm, sort the annual load rates of the N users in the first column of W year from smallest to largest into a3 interval levels, sort the maximum load utilization hours of the N users in the second column of the same year from smallest to largest into b3 interval levels, and sort the rated capacities of the N users in the third column from smallest to largest into c3 interval levels. The interval levels corresponding to each numerical value in the column of W year constitute the discrete electricity consumption feature dataset at the annual time scale as follows:

[0135]

[0136] where x n_year is the interval level of the R of the nth user n in the first column of data, and the value range of x n_year is from 1 to a3; y n_year is the interval level of the T of the nth user max n in the second column of data, and the value range of y n_year is from 1 to b3; z n_year is the interval level of the V of the nth user n in the third column of data, and the value range of z n_year is from 1 to c3;

[0137] Combining the discrete electricity consumption feature dataset W' year at the annual time scale with the time series dataset E of the monthly electricity consumption of N users within a year year yields the continuous-discrete hybrid electricity consumption feature dataset at the annual time scale as follows:

[0138]

[0139] S4. Based on the continuous-discrete hybrid electricity consumption characteristics of power users obtained after quantization in S3 at different time scales, construct a K-prototypes power user classification model to determine the optimal number of user classifications at the daily, monthly, and annual time scales, specifically including the following sub-steps:

[0140] S4.1 Construct a K-prototypes power user classification model and input the continuous-discrete hybrid electricity consumption feature datasets W″ day 、W″ month and W″ year at the daily, monthly, and annual time scales respectively;

[0141] S4.2 Continuous electricity consumption characteristics E on a daily time scale day Perform Z-score standardization, adjust the weights between continuous and discrete electricity consumption characteristics, and use the elbow method to determine the optimal number of user classifications k day ;

[0142] Among them, the calculation formula for the weight factor on a daily time scale is:

[0143]

[0144] In the formula, σ day,n is the standard deviation of the continuous electricity consumption data of the nth user on a daily time scale after standardization;

[0145] S4.3 Continuous electricity consumption characteristics E on a monthly time scale month Perform Z-score standardization, adjust the weights between continuous and discrete electricity consumption characteristics, and use the elbow method to determine the optimal number of user classifications k month ;

[0146] Among them, the calculation formula for the weight factor on a monthly time scale is:

[0147]

[0148] In the formula, σ month,n is the standard deviation of the continuous electricity consumption data of the nth user on a monthly time scale after standardization;

[0149] S4.4 Continuous electricity consumption characteristics E on an annual time scale year Perform Z-score standardization, adjust the weights between continuous and discrete electricity consumption characteristics, and use the elbow method to determine the optimal number of user classifications k year ;

[0150] Among them, the calculation formula for the weight factor on an annual time scale is:

[0151]

[0152] In the formula, σ year,n is the standard deviation of the continuous electricity consumption data of the nth user on an annual time scale after standardization;

[0153] It should be noted here that the larger Gamma is, the greater the proportion of continuous electricity consumption characteristics in the classification process.

[0154] S5. Based on the K-prototypes power user partitioning model, draw the continuous electricity consumption fluctuation images of power users at different time scales and define the category labels of the continuous electricity consumption behaviors of each type of user at different time scales according to the image fluctuation characteristics, which specifically include the following sub-steps:

[0155] S5.1 Based on the K-prototypes user partitioning model, with a 15-minute time step to construct the horizontal axis and the standard deviation of the normalized continuous power consumption data on a daily time scale to construct the vertical axis, plot the continuous power consumption fluctuation images of the first type, the second type, …, the k-th type of users on a daily time scale, and summarize the category labels of the continuous power consumption behaviors of each type of user on a daily time scale from the aspects of periodicity, increase and decrease, peak value, valley value, and the power consumption characteristics in the early, middle, and late periods of the day; for example, if a certain type of user operates at a high load continuously during the day and at a low load continuously at night, then the label of this type of user is high load during the day type; day Based on the K-prototypes user partitioning model, with a 15-minute time step to construct the horizontal axis and the standard deviation of the normalized continuous power consumption data on a daily time scale to construct the vertical axis, plot the continuous power consumption fluctuation images of the first type, the second type, …, the k-th type of users on a daily time scale, and summarize the category labels of the continuous power consumption behaviors of each type of user on a daily time scale from the aspects of periodicity, increase and decrease, peak value, valley value, and the power consumption characteristics in the early, middle, and late periods of the day; for example, if a certain type of user operates at a high load continuously during the day and at a low load continuously at night, then the label of this type of user is high load during the day type;

[0156] S5.2 With a daily time step to construct the horizontal axis and the standard deviation of the normalized continuous power consumption data on a monthly time scale to construct the vertical axis, plot the continuous power consumption fluctuation images of the first type, the second type, …, the k-th type of users on a monthly time scale, and summarize the category labels of the continuous power consumption behaviors of each type of user on a monthly time scale from the aspects of periodicity, increase and decrease, peak value, valley value, and the power consumption characteristics at the beginning, middle, and end of the month; month With a daily time step to construct the horizontal axis and the standard deviation of the normalized continuous power consumption data on a monthly time scale to construct the vertical axis, plot the continuous power consumption fluctuation images of the first type, the second type, …, the k-th type of users on a monthly time scale, and summarize the category labels of the continuous power consumption behaviors of each type of user on a monthly time scale from the aspects of periodicity, increase and decrease, peak value, valley value, and the power consumption characteristics at the beginning, middle, and end of the month;

[0157] S5.3 With a monthly time step to construct the horizontal axis and the normalized continuous power consumption characteristics on an annual time scale as the vertical axis, plot the continuous power consumption fluctuation images of the first type, the second type, …, the k-th type of users on an annual time scale, and summarize the category labels of the continuous power consumption behaviors of each type of user on an annual time scale from the aspects of periodicity, increase and decrease, peak value, valley value, and the power consumption characteristics in different seasons of spring, summer, autumn, and winter within a year. year With a monthly time step to construct the horizontal axis and the normalized continuous power consumption characteristics on an annual time scale as the vertical axis, plot the continuous power consumption fluctuation images of the first type, the second type, …, the k-th type of users on an annual time scale, and summarize the category labels of the continuous power consumption behaviors of each type of user on an annual time scale from the aspects of periodicity, increase and decrease, peak value, valley value, and the power consumption characteristics in different seasons of spring, summer, autumn, and winter within a year.

[0158] S6. Based on the discrete mixed power consumption characteristics of power users obtained after quantization and characterization in S3 at different time scales, draw the radar charts of the discrete power consumption characteristics of power users at different time scales, which specifically include the following sub-steps:

[0159] S6.1 Based on the discrete power consumption characteristic data set W′ on a daily time scale in S3.1 day draw the radar charts of the discrete power consumption characteristics of each user on a daily time scale, and through the shape and size of the radar charts on a daily time scale, the strengths of the three discrete power consumption characteristics of the average volatility, the peak-to-valley difference of the load, and the maximum fluctuation amount in the adjacent 1 hour of a single user on the same day can be intuitively reflected;

[0160] S6.2 Based on the discrete power consumption characteristic data set W′ on a monthly time scale in S3.2 month, draw the radar chart of the discrete electricity consumption characteristics of each user on a monthly time scale, and through the shape and size of the radar chart on the monthly time scale, the strength of the three discrete electricity consumption characteristics, namely, the maximum electricity consumption fluctuation amount between two adjacent days of a single user, the fluctuation amount of the average electricity consumption on working days and non-working days, and the daily load peak-valley difference rate, can be intuitively reflected;

[0161] S6.3 Based on the discrete electricity consumption characteristic dataset W' on the annual time scale in S3.3 year , draw the radar chart of the discrete electricity consumption characteristics of each user on an annual time scale, and through the shape and size of the radar chart on the annual time scale, the strength of the three discrete electricity consumption characteristics, namely, the annual load rate, the maximum load utilization hours, and the rated capacity of a single user in the current year, can be intuitively reflected.

[0162] S7. Based on the radar charts of discrete electricity consumption characteristics at different time scales obtained in S6.1 - S6.3 and the category labels of continuous electricity consumption behaviors at the time scales defined in S5.1 - S5.3, combined with the industry types of power users themselves, form the personalized portraits of the electricity consumption behaviors of power users.

[0163] Example:

[0164] S1. Construct the electricity consumption time series dataset of power users on multiple time scales:

[0165] S1.1 - S1.2 In this case, the experimental data used comes from 85 power users in a certain area in 2023. Outlier processing is performed on the daily electricity meter power datasets of all users, and the daily electricity consumption time series datasets are calculated;

[0166] S1.3 Select May 31st as the typical day. Then, the electricity consumption time series dataset of 85 users every 15 minutes on May 31st on a daily time scale is:

[0167]

[0168] S1.4 Take May, in which May 31st is located, as the typical month. Then, the daily electricity consumption time series datasets of 85 users in May on a monthly time scale are:

[0169]

[0170] S1.5 The monthly electricity consumption time series datasets of 85 users within one year on an annual time scale are:

[0171]

[0172] S2. Construct the evaluation index system for the electricity consumption fluctuation characteristics of power users:

[0173] S2.1 According to the calculation formula (There are 96 data points with a step of 15 minutes per day, and there are 95 volatilities between the 96 data points), calculation formula K n =P n,max -P n,min and calculation formula ΔE max n = max{ΔE n,1 , ΔE n,2 …ΔE n,i}, the evaluation index data set of 85 users on the daily time scale is obtained as follows:

[0174]

[0175] S2.2 According to the calculation formula ΔE′ max n = max{ΔE′ n,1 , ΔE′ n,2 , ΔE′ n,3 …ΔE′ n,m-1}, formula and formula The evaluation index data set of 85 users on the monthly time scale is obtained as follows:

[0176]

[0177] S2.3 According to the calculation formula formula and the rated capacity V n , the evaluation index data set of 85 users on the annual time scale is obtained as follows:

[0178]

[0179] S3. Based on the electricity consumption time series data set and the evaluation index system of electricity consumption volatility characteristics, quantitatively characterize the continuous-discrete hybrid electricity consumption characteristics of electricity users at different time scales:

[0180] S3.1 Based on the evaluation index data set W of N users on the daily time scale day , use the GMM clustering algorithm to divide the average volatility in the first column of W day from small to large into 3 interval levels, namely [0, 0.084], [0.1, 0.48], [0.79, 0.92]; divide the load peak-valley difference in the second column from small to large into 4 interval levels, [0, 0.012], [0.017, 0.114], [0.12, 0.4], [0.81, 1.11]; divide the maximum fluctuation amount within adjacent 1 hour in the third column from small to large into 3 interval levels, namely [0, 0.123], [0.13, 0.7], [0.85, 1.1]; then the discrete electricity consumption characteristic data set on the daily time scale is:

[0181]

[0182] The discrete electricity consumption characteristic dataset W′ on a daily time scale day The electricity consumption time series dataset E of 85 users every 15 minutes on May 31st day The continuous-discrete mixed electricity consumption characteristic data set on a daily time scale can be obtained by merging:

[0183]

[0184] S3.2 Evaluation index dataset W based on N users on a monthly time scale month , use GMM clustering algorithm to cluster W month In the first column, the maximum power consumption fluctuation between two consecutive days is divided into three intervals from small to large, namely [0, 9.94], [10.4, 46.11], [64, 81.6]; the fluctuation of the average power consumption between working days and non-working days in the second column is divided into four intervals from small to large, namely [0.0.278], [0.287, 1.52], [2.24, 5.1], [7.3, 8.97]; the peak-to-valley difference rate of daily load in the third column is divided into three intervals from small to large, namely [0, 0.362], [0.366, 0.93], [0.96, 1]; the discrete power consumption characteristic data set at the monthly time scale is:

[0185]

[0186] The discrete electricity consumption characteristic dataset W′ on a monthly time scale month The electricity consumption time series dataset E of 85 users in May month The continuous-discrete mixed electricity consumption characteristic data set on a daily time scale can be obtained by merging:

[0187]

[0188] S3.3 Evaluation index dataset W based on N users on a yearly time scale year , use GMM clustering algorithm to cluster W yearThe annual load factor in the first column is divided into 3 interval levels from small to large, which are [0, 0.368], [0.378, 0.689], [0.728, 0.867], respectively; the maximum load utilization hours in the second column are divided into 3 interval levels from small to large, which are [1062.47, 3599.63], [3718.16, 5901.3], [6037.56, 7595.94], respectively; the rated capacity in the third column is divided into 4 interval levels from small to large, which are [9000, 135600], [151500, 353980], [506000, 538000], [840000, 1230150], respectively; then the discrete electricity consumption characteristic dataset on the annual time scale is constructed as follows:

[0189]

[0190] The discrete electricity consumption characteristic dataset W' on the annual time scale year and the monthly electricity consumption time series dataset E of 85 users in the current year year are merged to obtain the continuous-discrete hybrid electricity consumption characteristic dataset on the annual time scale as follows:

[0191]

[0192] S4. Based on the continuous-discrete hybrid electricity consumption characteristics of power users obtained after quantization and characterization in S3 at different time scales, construct a K-prototypes power user classification model to determine the optimal number of user classifications at daily, monthly, and annual time scales:

[0193] S4.1 Construct a K-prototypes power user classification model and input the continuous-discrete hybrid electricity consumption characteristic datasets W″ day 、W″ month and W″ year at daily, monthly, and annual time scales respectively;

[0194] S4.2 Adjust the weight between continuous and discrete electricity consumption characteristics at the daily time scale to 0.483, and use the elbow method to determine the optimal number of user classifications k day = 4;

[0195] S4.3 Adjust the weight between continuous and discrete electricity consumption characteristics at the monthly time scale to 0.492, and use the elbow method to determine the optimal number of user classifications k month = 4;

[0196] S4.4 Adjust the weight between continuous and discrete electricity consumption characteristics at the annual time scale to 0.453, and use the elbow method to determine the optimal number of user classifications k year = 5.

[0197] S5. Based on the K-prototypes power user classification model, draw the continuous power consumption fluctuation images of power users at different time scales and define the category labels of the continuous power consumption behaviors of each type of user at different time scales according to the image fluctuation characteristics:

[0198] S5.1 Define the category labels of the continuous power consumption behaviors of each type of user on a daily time scale according to the characteristics of the continuous power consumption fluctuation image in Figure 2 as follows: The first type of user belongs to the daytime stable type, the second type of user belongs to the slow growth with periodic troughs type, the third type of user belongs to the morning high-load type, and the fourth type of user belongs to the slow decline with periodic peaks type;

[0199] S5.2 Define the category labels of the continuous power consumption behaviors of each type of user on a monthly time scale according to the characteristics of the continuous power consumption fluctuation image in Figure 3 as follows: The first type of user belongs to the type of first decreasing and then increasing, the second type of user belongs to the type of small-amplitude fluctuating growth, the third type of user belongs to the type of small-amplitude fluctuating decrease, and the fourth type of user belongs to the stable load type;

[0200] S5.3 Define the category labels of the continuous power consumption behaviors of each type of user on an annual time scale according to the characteristics of the continuous power consumption fluctuation image in Figure 4 as follows: The first type of user belongs to the low-load type in August, the second type of user belongs to the low-load type in summer, the third type of user belongs to the type of fluctuating growth, the fourth type of user belongs to the type of first increasing and then decreasing, and the fifth type of user belongs to the high-load type in spring.

[0201] S6. Based on the discrete mixed power consumption characteristics of power users at different time scales obtained after quantization and characterization in S3, draw the radar charts of the discrete power consumption characteristics of power users at different time scales;

[0202] S7. Taking user 1, the first user among 85 users, as an example, based on the radar charts of the discrete power consumption characteristics of this user at different time scales obtained in S6.1 - S6.3, as Figures 5 - 7 shown, and the category labels of the continuous power consumption behaviors of user 1 at the time scales defined in S5.1 - S5.3, combined with the industry type to which user 1 itself belongs, the personalized portraits of the power consumption behaviors of this user 1 at different time scales can be formed, as Figure 8 shown.

[0203] The present invention comprehensively considers the continuous fluctuation characteristics, discrete power consumption characteristics of the load data itself at multiple time scales, and multiple load characteristic evaluation indicators that can represent the power consumption behaviors of users, and combines the user's basic attribute of the user's industry type to conduct multi-faceted portraits of power users, providing an effective reference for power companies to improve the quality of power supply services more accurately and effectively.

[0204] In the present invention, orientation or positional relationships indicated by terms such as "upper", "lower", "bottom", "top", etc. are based on the orientation or positional relationships shown in the drawings, and are only relational terms determined for the convenience of describing the structural relationships of various components or elements of the present invention, rather than specifically referring to any component or element in the present invention, and should not be construed as a limitation on the present invention. Terms such as "connected" and "joined" should be understood in a broad sense, indicating that it can be a fixed connection, an integral connection or a detachable connection; it can be directly connected or indirectly connected through an intermediate medium. For those skilled in relevant scientific research or technology in this field, the specific meanings of the above terms in the present invention can be determined according to specific circumstances, and should not be construed as a limitation on the present invention.

[0205] Certainly, the above description is not a limitation on the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by those skilled in the art within the scope of the essence of the present invention should also fall within the protection scope of the present invention.

Claims

1. A multi-time-scale user electricity consumption behavior profiling method based on hybrid electricity consumption characteristics, characterized in that Specifically, it includes the following steps: S1. Construct the electricity consumption time series dataset of power users at multiple time scales; S2. Construct the evaluation index system for the electricity consumption fluctuation characteristics of power users; S3. Based on the electricity consumption time series dataset and the evaluation index system for the electricity consumption fluctuation characteristics, quantitatively characterize the continuous-discrete hybrid electricity consumption characteristics of power users at different time scales; S4. Based on the continuous-discrete hybrid electricity consumption characteristics of power users at different time scales obtained by quantitative characterization in S3, construct the K-prototypes power user classification model to determine the optimal number of user classifications at daily, monthly, and annual time scales; S5. Based on the K-prototypes power user classification model, draw the continuous electricity consumption fluctuation images of power users at different time scales and define the category labels of the continuous electricity consumption behaviors of each type of user at different time scales according to the image fluctuation characteristics; S6. Based on the discrete hybrid electricity consumption characteristics of power users at different time scales obtained by quantitative characterization in S3, draw the radar charts of the discrete electricity consumption characteristics of power users at different time scales; S7. Form the personalized portraits of the electricity consumption behaviors of power users at different time scales.

2. The method for multi-time scale user power consumption behavior profiling based on hybrid power consumption characteristics according to claim 1, wherein S1 constructs the electricity consumption time series dataset of power users at multiple time scales, specifically including the following sub-steps: S1.1 Select N power users, divide the electricity consumption data within a typical day into 96 time periods with a time step of 15 minutes. The power meter power dataset for every 15 minutes of the nth user on the a-th day collected in real time is: P n_a_day = [P n,a,1point P n,a,2point …P n,a,96point ; where 1 ≤ a ≤ 365; n is the user number, n = 1, 2, 3,..., N, and N is the total number of users; S1.2 Use the elimination method and the mean filling method to process the outliers in the power meter power dataset obtained in S1.1, and calculate the electricity consumption time series dataset for every 15 minutes of the nth user on the a-th day collected in real time as: E n_a_day = P n_a_day ·t = [E n,a,1point E n,a,2point …E n,a,96point ; where t = 0.25; 1 ≤ a ≤ 365; n is the user number, n = 1, 2, 3,..., N; S1.3 Taking the a-th day as the typical day, the electricity consumption time series dataset for every 15 minutes of N users on the a-th day at the daily time scale is: S1.4 Taking the month A where the a-th day is located as the typical month and assuming that month A has m days, the daily electricity consumption time series dataset of the nth user in month A is: E n_A_month = [E n,A,1day E n,A,2day …E n,A,mday ; where E n,A,mday is the electricity consumption of the nth user on the mth day of the A-th month. Then, the time series dataset of the daily electricity consumption of N users in the A-th month on a monthly time scale is as follows: S1.5 According to 1.4, the monthly electricity consumption time series dataset of the nth user in a year is: E n_year = [E n,1month E n,2month …E n,12month ; where, E n,A,1day , …, E n,12month are the electricity consumption of the nth user in January, February, …, December respectively. Then, the time series dataset of monthly electricity consumption of N users in a year under the annual time scale is:

3. The method for multi-time scale user power consumption behavior portrait based on hybrid power consumption characteristics according to claim 2, wherein S2 constructs the evaluation index system for the electricity consumption fluctuation characteristics of power users, specifically including the following sub-steps: S2.1 On a daily time scale, according to the average volatility calculation formula for every 15 minutes of the nth user on the a-th day Load peak-valley difference calculation formula K n = P n,max - P n,min And the calculation formula for the maximum fluctuation amount ΔE of adjacent 1 hour maxn = max{ΔE n,1 , ΔE n,2 … ΔE n,i}, the evaluation index data set of the nth user on the daily time scale is obtained as follows: Then the evaluation index dataset of N users at the daily time scale is: where 1 ≤ i ≤ 95; η n,i is the electricity consumption volatility of the nth user every 15 minutes on the ath day; P n,max is the maximum load power of the nth user on the ath day, and P n,min is the minimum load power of the nth user on the ath day; ΔE n,1 , ΔE n,2 …ΔE n,i are respectively the fluctuation amounts of the electricity consumption of the nth user in two adjacent hours on the ath day; S2.2 On a monthly time scale, according to the calculation formula for the maximum power consumption fluctuation between two adjacent days in the A-th month of the n-th user, ΔE′ maxn = max{ΔE′ n,1 , ΔE′ n,2 , ΔE′ n,3 …ΔE′ n,m-1}, the calculation formula for the fluctuation of the average power consumption on weekdays and non-working days and the formula for the daily load peak-valley difference rate The evaluation index data set of the n-th user on a monthly time scale is obtained as follows: W n_month = [ΔE′ maxn ΔE n μ n ; Then the evaluation index dataset of N users at the monthly time scale is: where n = 1, 2, 3, … N; ΔE′ n,1 , ΔE′ n,2 … ΔE′ n,m-1 are the power consumption fluctuation amounts of the nth user on two adjacent days in the A-th month respectively; is the average value of the power consumption of the nth user on all working days in the A-th month, is the average value of the power consumption of the nth user on all non-working days in the A-th month; E′ n,max is the maximum daily power consumption of the nth user in the A-th month, E′ n,min is the minimum daily power consumption of the nth user in the A-th month; S2.3 On an annual time scale, according to the annual load rate calculation formula of the nth user in the current year and the calculation formula of the maximum load utilization hours and together with the rated capacity V n jointly constitute the evaluation index data set of the nth user on the annual time scale as: W n_year = [R n T maxn V n ; Then the evaluation index dataset of N users at the annual time scale is: where n = 1, 2, 3, … N; E n,av is the average daily electricity consumption of the nth user throughout the year, E n,max is the maximum daily electricity consumption of the nth user within the year; ∑E n is the total annual electricity consumption of the nth user, P′ n,max is the maximum load power of the nth user for 1 hour.

4. The method for multi-time-scale user power consumption behavior portrait based on hybrid power consumption characteristics according to claim 3, wherein In S3, based on the electricity consumption time series dataset and the evaluation index system for the electricity consumption fluctuation characteristics, quantitatively characterize the continuous-discrete hybrid electricity consumption characteristics of power users at different time scales, specifically including the following sub-steps: S3.1 Evaluation index data set of N users based on daily time scale Use the GMM clustering algorithm to divide W day The average volatility of N users every 15 minutes on the a-th day in the first column is divided into a1 interval levels from small to large, the load peak-valley difference of N users on the a-th day in the second column is divided into b1 interval levels from small to large, and the maximum fluctuation volume of N users in adjacent 1 hour on the a-th day in the third column is divided into c1 interval levels from small to large. W day The interval level corresponding to the column where each value in W is located constitutes the discrete electricity consumption characteristic data set on the daily time scale as follows: where x n_day is the interval level of the η of the nth user n in the first column of data, and the value range of x n_day is from 1 to a1; y n_day is the interval level of the K of the nth user n in the second column of data, and the value range of y n_day is from 1 to b1; z n_day is the interval level of the ΔE of the nth user maxn in the third column of data, and the value range of z n_day is from 1 to c1; The discrete power consumption feature dataset W' at the daily time scale day is merged with the time series dataset E of the power consumption of N users every 15 minutes on the a-th day day to obtain the continuous-discrete hybrid power consumption feature dataset at the daily time scale as follows: S3.2 Evaluation index data set of N users based on monthly time scale Use the GMM clustering algorithm to divide W month The N maximum power consumption fluctuations between two adjacent days in the A-th month in the first column are divided into a2 interval levels from small to large, the N power consumption fluctuations between weekdays and non-weekdays of the A-th month of N users in the second column are divided into b2 interval levels from small to large, and the N daily load peak-valley difference rates of N users in the A-th month in the third column are divided into c2 interval levels from small to large. W month The interval level corresponding to the column where each value in W is located constitutes the discrete power consumption feature data set on the monthly time scale as follows: where x n_month is the interval level of ΔE′ of the nth user maxn in the first column of data, and the value range of x n_month is from 1 to a2; y n_month is the interval level of ΔE of the nth user n in the second column of data, and the value range of y n_month is from 1 to b2; z n_month is the interval level of μ of the nth user n in the third column of data, and the value range of z n_month is from 1 to c2; The discrete electricity consumption feature dataset W' at the monthly time scale month is merged with the time series dataset E of the daily electricity consumption of N users in the A-th month month to obtain the continuous-discrete hybrid electricity consumption feature dataset at the monthly time scale as follows: S3.3 Evaluation index data set of N users based on the annual time scale Use the GMM clustering algorithm to divide the N users' annual load rates in the first column of W year into a3 interval levels from small to large in the current year, divide the maximum load utilization hours of N users in the second column into b3 interval levels from small to large in the current year, and divide the rated capacities of N users in the third column into c3 interval levels from small to large. For W year The interval levels corresponding to the columns where each value is located in it constitute the discrete electricity consumption feature data set on the annual time scale as follows: where x n_year is the interval level of the R of the nth user in the first column of data, and the value range of x n is from 1 to a3; y n_year is the interval level of the T of the nth user in the second column of data, and the value range of y n_year is from 1 to b3; z maxn is the interval level of the V of the nth user in the third column of data, and the value range of z n_year is from 1 to c3; n_year n n_year ​​​ The discrete electricity consumption feature dataset W' on an annual time scale year is merged with the time series dataset E of monthly electricity consumption of N users in a year year to obtain a continuous-discrete hybrid electricity consumption feature dataset on an annual time scale as follows:

5. The multi-time-scale user power consumption behavior profiling method based on hybrid power consumption characteristics according to claim 4, wherein In S4, based on the continuous-discrete hybrid electricity consumption characteristics at different time scales obtained by quantitative characterization in S3, construct the K-prototypes power user classification model to determine the optimal number of user classifications at daily, monthly, and annual time scales, specifically including the following sub-steps: S4.1 Construct a K-prototypes power user division model and input the continuous-discrete hybrid electricity consumption feature datasets W″ day , W″ month and W″ year ; S4.2 Continuous electricity consumption characteristics on a daily time scale E day Perform Z-score normalization, adjust the weights between continuous and discrete electricity consumption characteristics, and use the elbow method to determine the optimal number of user classifications k day ; Among them, the calculation formula for the weight factor at the daily time scale is as follows: where, σ day,n is the standard deviation of the continuous electricity consumption data of the nth user under the standardized daily time scale; S4.3 Continuous electricity consumption characteristics E on a monthly time scale month Perform Z-score normalization, adjust the weights between continuous and discrete electricity consumption characteristics, and use the elbow method to determine the optimal number of user categories k month ; Among them, the calculation formula for the weight factor at the monthly time scale is as follows: where σ month,n is the standard deviation of the continuous electricity consumption data of the nth user under the standardized monthly time scale; S4.4 Continuous electricity consumption characteristics E on an annual time scale year Perform Z-score standardization, adjust the weights between continuous and discrete electricity consumption characteristics, and use the elbow method to determine the optimal number of user classifications k year ; Among them, the calculation formula for the weight factor at the annual time scale is as follows: where σ year,n is the standard deviation of the continuous electricity consumption data of the nth user under the annual time scale after standardization.

6. The method for multi-time scale user electricity consumption behavior portrait based on hybrid electricity consumption characteristics according to claim 5, wherein Based on the partitioning model described in S5, draw the continuous electricity consumption fluctuation images of power users at different time scales and define the category labels of continuous electricity consumption behaviors at different time scales according to the image fluctuation characteristics, specifically including the following sub-steps: S5.1 Based on the K-prototypes user partitioning model, with a 15-minute time step to construct the horizontal axis, and the standard deviation of the continuous electricity consumption data standardized on a daily time scale to construct the vertical axis, plot the continuous electricity consumption fluctuation images of users in Class 1, Class 2, … Class k on a daily time scale, and summarize the category labels of the continuous electricity consumption behavior of each class of users on a daily time scale from the aspects of periodicity, increase and decrease, peak value, valley value, and the electricity consumption characteristics in the early, middle, and late periods within a day. day Class users, and summarize the category labels of the continuous electricity consumption behavior of each class of users on a daily time scale from the periodicity, increase and decrease, peak value, valley value, and the electricity consumption characteristics in the early, middle, and late periods within a day. S5.2 Construct the horizontal axis with the day as the time step, and construct the vertical axis with the standard deviation of the continuous electricity consumption data standardized on a monthly time scale, and plot the continuous electricity consumption fluctuation images of the first type, the second type, … the k month th type of users on a monthly time scale, and summarize the category labels of the continuous electricity consumption behavior of each type of user on a monthly time scale from the aspects of periodicity, increase and decrease, peak value, valley value, and electricity consumption characteristics in the early, middle, and late periods of the month; S5.3 Construct the horizontal axis with a monthly time step and the vertical axis with the normalized continuous electricity consumption characteristics on an annual time scale, and plot the continuous electricity consumption fluctuation images of users of the first type, the second type, …, the k year th type on an annual time scale. Then, summarize the category labels of the continuous electricity consumption behavior of each type of user on an annual time scale from aspects such as periodicity, increase and decrease, peak value, valley value, and electricity consumption characteristics in different seasons of spring, summer, autumn, and winter within a year.

7. The multi-time scale user electricity consumption behavior profiling method based on hybrid electricity consumption characteristics according to claim 6, wherein Based on the discrete mixed electricity consumption characteristics of power users at different time scales obtained after quantization and characterization in S3, draw the discrete electricity consumption characteristic radar charts of power users at different time scales in S6, specifically including the following sub-steps: S6.1 Based on the discrete electricity consumption feature dataset W in the Sino-Japanese time scale of S3.1 d ′ ay , draw the radar chart of the discrete electricity consumption characteristics of each user on the daily time scale, and through the shape and size of the radar chart on the daily time scale, the strengths of the three discrete electricity consumption characteristics of the average volatility, load peak-valley difference, and maximum fluctuation volume in the adjacent 1 hour of a single user on the same day can be intuitively reflected; S6.2 Based on the discrete electricity consumption feature dataset W at the monthly time scale in S3.2 m ′ onth , draw a radar chart of the discrete electricity consumption features of each user at the monthly time scale, and through the shape and size of the radar chart at the monthly time scale, the strength of three discrete electricity consumption features, namely, the maximum electricity consumption fluctuation amount between two adjacent days of a single user, the fluctuation amount of the average electricity consumption on working days and non - working days, and the daily load peak - valley difference rate, can be intuitively reflected; S6.3 Based on the discrete electricity consumption feature dataset W at the annual time scale in S3.3 y ′ ear , draw the radar chart of the discrete electricity consumption features of each user at the annual time scale, and through the shape and size of the radar chart at the annual time scale, the strengths of the three discrete electricity consumption features of the annual load factor, the maximum load utilization hours, and the rated capacity of a single user in the current year can be intuitively reflected.

8. The multi-time-scale user power consumption behavior profiling method based on hybrid power consumption characteristics according to claim 7, wherein Based on the discrete electricity consumption characteristic radar charts at different time scales obtained in S6.1 - S6.3 and the category labels of continuous electricity consumption behaviors at the time scales defined in S5.1 - S5.3, combined with the industry type of the power users themselves, form the personalized portraits of the electricity consumption behaviors of power users.

Images