Electricity-carbon cooperative intelligent pushing method based on user portraits

Abstract

The invention discloses an electricity-carbon cooperative intelligent pushing method based on user portraits, and belongs to the field of electricity-carbon cooperative intelligent pushing, and the method comprises the following steps: S1, collecting and classifying annual historical load data and time-of-use electricity price data sets of users; s2, extracting a seasonal typical daily load curve and a typical time-of-use electricity price curve; s3, calculating daily electricity price demand elasticity of workdays and weekends; s4, constructing a user portrait; s5, elastically grading the electricity price demands of the users; and S6, respectively pushing a differentiated power purchase strategy and a carbon emission reduction supervision strategy to the price-sensitive user and the price-insensitive user. According to the electricity and carbon cooperative intelligent pushing method based on the user portraits, the user portraits are accurately depicted based on the user electricity price demand elastic coefficients, so that the platform can customize differentiated strategies for different types of users, the operation capability of the supervision platform is improved, and the method has good application value.

Description

Technical Field

[0001] The present invention relates to the field of electric-carbon collaborative intelligent push, and in particular to an electric-carbon collaborative intelligent push method based on user portraits. Background Art

[0003] In the field of electricity and carbon, the research perspective of intelligent push technology for user electricity consumption behavior is mostly to minimize user costs, ignoring the load characteristics of the user itself. Considering the characteristics of the user, if the user is price sensitive, then the time-of-use electricity price can effectively reduce the customer's electricity costs and reduce the pressure on grid load scheduling by guiding the user to change the electricity purchasing behavior in different time periods, which is conducive to the realization of carbon emission reduction goals; if the user is price insensitive, the role of the time-of-use electricity price is limited.

[0004] Moreover, under current research, the study of user portraits only stops at studying user electricity consumption behavior by season, and does not go deep into distinguishing between weekdays and weekends. Summary of the invention

[0005] To solve the above problems, the present invention provides an intelligent push method for electricity-carbon collaboration based on user portraits, which accurately portrays user portraits based on the user's electricity price demand elasticity coefficient, so that the platform can customize differentiated strategies for different types of users, thereby improving the operational capabilities of the supervision platform and having good application value.

[0006] To achieve the above object, the present invention provides an electric-carbon collaborative intelligent push method based on user portrait, comprising the following steps:

[0007] S1. Collect and classify the user's annual historical load data and time-of-use electricity price data set according to the season, and perform data preprocessing;

[0008] S2. Classify the daily load curves and time-of-use electricity price curves of the same season, classify the typical load curves and time-of-use electricity price curves of weekdays and weekends, and use the particle swarm optimization algorithm to extract the typical daily load curve and typical time-of-use electricity price curve of the season;

[0009] S3. In the same season, based on the typical daily load curve and time-of-use electricity price curve, calculate the daily electricity price demand elasticity on weekdays and weekends;

[0010] S4. Use the GMM algorithm to perform cluster analysis on electricity price demand elasticity, build user portraits, and form a user tag library;

[0011] S5. Classify the elasticity of users’ electricity price demand: divide users into price-sensitive users and insensitive users based on the constructed user profiles;

[0012] S6. Push differentiated electricity purchasing strategies and carbon emission reduction supervision strategies to price-sensitive users and insensitive users respectively.

[0013] Preferably, the data preprocessing described in step S1 includes missing value processing and malformed value processing;

[0014] Among them, missing value processing: use the daily average data of load or electricity price in the corresponding season to supplement the missing values, that is, the mean interpolation method, the formula is:

[0015]

[0016] In the formula, is the daily average data of load or electricity price in the corresponding season; n is the number of days in the corresponding season; x i is the load or electricity price data for the i-th day;

[0017] The steps for processing abnormal values ​​are as follows: Using the Laida criterion, first calculate the average value of the load data X i is the load data of the i-th day, and the standard deviation of the load data V i =X i -X(i=1,2,3,...n);

[0018] Then use the Bessel formula to calculate the standard deviation δ and set the measured value x b The standard deviation V b (1≤b≤n);

[0019] When: |V b |=|x b -X|>3δ, determine x b For outliers, filter them and replace them with X;

[0020] After the outlier, recalculate X and V i , conduct a new round of screening to check other abnormal data in the remaining data until all data meet the requirements.

[0021] Preferably, the steps of extracting the seasonal typical working day load curve and the seasonal typical weekend load curve in step S2 are as follows:

[0022] Step 1: Assume that it is known that the user has m+1 working days or weekends in the jth season. Then the working day or weekend load set Y of the user in this season is j ∈(m+1)×24:

[0023]

[0024] In the formula, is the load value of the user at the tth hour on the nth day during the jth season working day or weekend, n = 0, 1, 2, ... m, t = 0, 1, 2, ... 23;

[0025] Step 2: Assume the typical load X of the user on the jth seasonal weekday or weekend j for:

[0026] X j =[x 0 , x 1 , x 2 ,...x t ,...x 23 ] (3) In the formula, x t is the typical load value of the user at hour t on a weekday or weekend in the jth season;

[0027] The third step is to take the shortest distance between all coordinates of the curves as the objective function:

[0028]

[0029] In the formula, S j is the objective function value; Y j [n] is the user's load value on the nth day of the jth season; X j [t] is the load value at hour t in the jth season;

[0030] Among them, X j Pay attention to satisfying the load constraint value and ensure that it is within the load range, that is:

[0031] min{(Y j [n][t]}≤x j [t]≤max{(Y j [n][t]} (5)

[0032] In the formula, min{(Y j [n][t]} is the minimum value of the load interval; max{(Y j [n][t]} is the maximum value of the load interval; and t=0, 1, 2, ..., 23; n=0, 1, 2, ..., m;

[0033] Step 4: Solve the objective function described in formula (4) by particle swarm optimization algorithm, and obtain the optimal X through iterative solution. j As the typical load curve in the corresponding season;

[0034] The steps for extracting seasonal typical weekday time-of-use electricity price curves and seasonal typical weekend time-of-use electricity price curves are as follows:

[0035] Step 1: Assume that the user has m+1 working days or weekends in the jth season. Then the time-of-use electricity price set Z for the user on working days or weekends in this season is j ∈(m+1)×24:

[0036]

[0037] In the formula, The electricity price for the user at the tth hour on the nth day during the jth season working day or weekend;

[0038] Step 2: Assume that the typical time-of-use electricity price curve for the user’s jth seasonal working day or weekend is

[0039] A j =[a 0 , a 1 , a 2 , ...a t , ...a 23 ] (7)

[0040] Where: a t is the electricity price for the user at hour t on a weekday or weekend in the jth season;

[0041] The third step is to take the shortest distance between all coordinates of the curves as the objective function:

[0042]

[0043] Where: P j is the objective function value; m is the number of working days or weekend days in the jth season; Z j [n] is the electricity price for the user on the nth day of the jth season; a j [t] is the typical time-of-use electricity price for the user at time t in season j;

[0044] Among them, a j The value at each moment cannot exceed the upper and lower limits of the corresponding moment:

[0045] min{Z j [n][t]}≤a j [t]≤max{(Z j [n][t]} (9)

[0046] In the formula, min{Z j [n][t]} is the lower limit of time t; it is max{(Z j [n][t]} is the upper limit value at time t;

[0047] Step 4: Solve the objective function described in formula (8) by particle swarm optimization algorithm, and obtain the optimal a through iterative solution.j As the typical time-of-use electricity price curve in the corresponding season.

[0048] Preferably, in step S3, the typical load curve and the time-of-use electricity price curve of the same season and corresponding date are combined to form a typical load-electricity price curve, and then the user's daily electricity price demand elasticity is calculated by the electricity price elasticity formula to form an electricity price demand elasticity curve;

[0049] The elasticity coefficient E of the power load in the electricity price elasticity formula represents the percentage of the change ΔL in the user's electricity demand caused by the change ΔP in the electricity price within a certain period of time. The expression is as follows:

[0050]

[0051] Where P 0 , L 0 are the typical electricity price and electricity demand respectively.

[0052] Preferably, step S4 specifically includes the following steps:

[0053] S41. Determine the cluster number k of the electricity price elasticity on weekdays and weekends in the same season, and use the sum of squared errors SSE as the evaluation index for determining the cluster number:

[0054]

[0055] In the formula, C i is a cluster in the cluster; u i C i The mean of each data point in; x is a point in the cluster;

[0056] S42, assuming that the actual number of clusters is k*, comparing the actual number of clusters to k* and k, and determining the optimal number of clusters based on the comparison result and the change of SSE;

[0057] S43. Use GMM to fit the data set to a mixture of multiple Gaussian distributions to discover the underlying structure of the data:

[0058] S431. Assume that each type of data obeys normal distribution, and the probability model of normal distribution is:

[0059]

[0060] In the formula, k is the number of clusters; α k is the weight; is the probability density of the normal distribution; θ k is the parameter of probability density; y is the sample;

[0061] The probability density expression is:

[0062]

[0063] Where: k is the standard deviation of sample y; μ k is the mean of sample y;

[0064] The probability density function of k types of normal distribution and the weight α of each type are estimated through the normal distribution probability model training data. k ;

[0065] S432, calculate the probability of each data appearing in k types of normal distribution, that is, substitute the data into k normal distributions to find the probability P(y i ) k :

[0066]

[0067] Where: y i is a data in the data set; k is the kth normal distribution;

[0068] S433, by comparing the probability P(y i ) k The value of is used to classify the sample into the cluster with the highest probability value.

[0069] Preferably, in step S5, a user electricity consumption behavior profile is generated based on the clustering results of the user price demand elasticity, and the users are divided into price-sensitive users and price-insensitive users.

[0070] Preferably, in step S5, the user price demand elasticity curve is updated regularly in combination with the user's real-time feedback on the time-of-use electricity price, and the user type is iterated to improve the push accuracy.

[0071] The present invention has the following beneficial effects:

[0072] 1. Through the GMM algorithm, user clustering analysis is carried out to build user portraits, which can dig out the demand characteristics of power users, thereby customizing differentiated strategies and improving the operational capabilities of the supervision platform, which has good application value;

[0073] 2. Real-time monitoring solutions with high cost-effectiveness and refined management help users clearly identify the current status of emissions, evaluate the economic feasibility of carbon reduction measures, and promote the development and application of low-carbon technologies;

[0074] 3. Help users accurately grasp the characteristics of their own electricity purchasing behavior, reduce electricity costs, and assist users in formulating more reasonable electricity purchasing and emission reduction measures.

[0075] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 This is a flow chart of the electric-carbon collaborative intelligent push method based on user portrait of the present invention. DETAILED DESCRIPTION

[0077] In order to make the purpose, technical scheme and advantages disclosed in the embodiments of the present invention clearer, the embodiments of the present invention are further described in detail in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not intended to limit the embodiments of the present invention. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions.

[0078] It should be noted that the terms "include", "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or devices.

[0079] Like reference numerals and letters denote similar items in the following drawings, and thus, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0080] like Figure 1 As shown, a method for intelligent push of electric-carbon collaboration based on user portraits includes the following steps:

[0081] S1. Collect and classify the user's annual historical load data and time-of-use electricity price data set according to the season, and perform data preprocessing;

[0082] The data preprocessing described in step S1 includes missing value processing and malformed value processing;

[0083] Among them, missing value processing: Since it has been classified by season and the fluctuations caused by seasonal factors are excluded, the distribution of electricity consumption data in each season is similar. Therefore, the daily average data of load or electricity price in the corresponding season is used to supplement the missing values, that is, the mean interpolation method, the formula is:

[0084]

[0085] In the formula, is the daily average data of load or electricity price in the corresponding season; n is the number of days in the corresponding season; x i is the load or electricity price data for the i-th day;

[0086] The steps for processing abnormal values ​​are as follows: the data set contains a large amount of user load data that conforms to the normal distribution characteristics, so the Laida criterion is used to first calculate the average value of the load data X i is the load data of the i-th day, and the standard deviation of the load data V i =X i -X(i=1, 2, 3, ... n);

[0087] Then use the Bessel formula to calculate the standard deviation δ and set the measured value x b The standard deviation V b (1≤b≤n);

[0088] When: |V b |=|x b -X|>3δ, determine x b For outliers, filter them and replace them with X;

[0089] After the outlier, recalculate X and V i , conduct a new round of screening to check other abnormal data in the remaining data until all data meet the requirements.

[0090] S2. Classify the daily load curves and time-of-use electricity price curves of the same season, and classify the typical load curves and time-of-use electricity price curves of weekdays and weekends. Use particle swarm optimization algorithm to extract seasonal typical daily load curves and typical time-of-use electricity price curves to ensure the objectivity of basic load extraction;

[0091] The steps for extracting the seasonal typical working day load curve and the seasonal typical weekend load curve in step S2 are as follows:

[0092] Step 1: Assume that it is known that the user has m+1 working days or weekends in the jth season. Then the working day or weekend load set Y of the user in this season is j ∈(m+1)×24:

[0093]

[0094] In the formula, is the load value of the user at the tth hour on the nth day during the jth season working day or weekend, n = 0, 1, 2, ... m, t = 0, 1, 2, ... 23;

[0095] Step 2: Assume the typical load X of the user on the jth seasonal weekday or weekend j for:

[0096] X j =[x 0 , x 1 , x 2 ,...x t ,...x 23 ] (3) In the formula, x t is the typical load value of the user at hour t on a weekday or weekend in the jth season;

[0097] The third step is to take the shortest distance between all coordinates of the curves as the objective function:

[0098]

[0099] In the formula, S j is the objective function value; Y j [n] is the user's load value on the nth day of the jth season; X j [t] is the load value at hour t in the jth season;

[0100] Among them, X j Pay attention to satisfying the load constraint value and ensure that it is within the load range, that is:

[0101] min{(Y j [n][t]}≤x j [t]≤max{(Y j [n][t]} (5)

[0102] In the formula, min{(Y j [n][t]} is the minimum value of the load interval; max{(Y j [n][t]} is the maximum value of the load interval; and t = 0, 1, 2, ..., 23; n = 0, 1, 2, ..., m;

[0103] Step 4: Solve the objective function described in formula (4) by particle swarm optimization algorithm, and obtain the optimal X through iterative solution. j As the typical load curve in the corresponding season;

[0104] The steps for extracting seasonal typical weekday time-of-use electricity price curves and seasonal typical weekend time-of-use electricity price curves are as follows:

[0105] Step 1: Assume that the user has m+1 working days or weekends in the jth season. Then the time-of-use electricity price set Z for the user on working days or weekends in this season is j ∈(m+1)×24:

[0106]

[0107] In the formula, The electricity price for the user at the tth hour on the nth day during the jth season working day or weekend;

[0108] Step 2: Assume that the typical time-of-use electricity price curve for the user’s jth seasonal working day or weekend is

[0109] A j =[a 0 , a 1 , a 2 , ...a t , ...a 23 ] (7)

[0110] Where: a t is the electricity price for the user at hour t on a weekday or weekend in the jth season;

[0111] The third step is to take the shortest distance between all coordinates of the curves as the objective function:

[0112]

[0113] Where: P j is the objective function value; m is the number of working days or weekend days in the jth season; Z j [n] is the electricity price for the user on the nth day of the jth season; a j [t] is the typical time-of-use electricity price for the user at time t in season j;

[0114] Among them, a j The value at each moment cannot exceed the upper and lower limits of the corresponding moment:

[0115] min{Z j [n][t]}≤a j [t]≤max{(Z j [n][t]} (9)

[0116] In the formula, min{Z j [n][t]} is the lower limit of time t; it is max{(Z j [n][t]} is the upper limit value at time t;

[0117] Step 4: Solve the objective function described in formula (8) by particle swarm optimization algorithm, and obtain the optimal a through iterative solution. j As the typical time-of-use electricity price curve in the corresponding season.

[0118] S3. In the same season, based on the typical daily load curve and time-of-use electricity price curve, calculate the daily electricity price demand elasticity on weekdays and weekends;

[0119] In step S3, the typical load curve and the time-of-use electricity price curve of the same season and corresponding date are combined to form a typical load-electricity price curve, and then the user's daily electricity price demand elasticity is calculated by the electricity price elasticity formula to form an electricity price demand elasticity curve;

[0120] The elasticity coefficient E of the power load in the electricity price elasticity formula represents the percentage of the change ΔL in the user's electricity demand caused by the change in electricity price AP in a certain period of time. Its expression is as follows:

[0121]

[0122] Where P 0 , L 0 are the typical electricity price and electricity demand respectively.

[0123] S4. Use the GMM algorithm to perform cluster analysis on electricity price demand elasticity, build user portraits, and form a user tag library;

[0124] Step S4 specifically includes the following steps:

[0125] S41. Determine the cluster number k of the electricity price elasticity on weekdays and weekends in the same season, and use the sum of squared errors SSE as the evaluation index for determining the cluster number:

[0126]

[0127] In the formula, C i is a cluster in the cluster; u i C i The mean of each data point in; x is a point in the cluster;

[0128] S42, assuming the actual number of clusters is k * , compared with the actual number of clusters k * and k, based on the comparison results, determine the optimal number of clusters according to the change of SSE;

[0129] S43. Use GMM to fit the data set to a mixture of multiple Gaussian distributions to discover the underlying structure of the data:

[0130] S431. Assume that each type of data obeys normal distribution, and the probability model of normal distribution is:

[0131]

[0132] In the formula, k is the number of clusters; α k is the weight; is the probability density of the normal distribution; θ k is the parameter of probability density; y is the sample;

[0133] The probability density expression is:

[0134]

[0135] Where: k is the standard deviation of sample y; μ k is the mean of sample y;

[0136] The probability density function of k types of normal distribution and the weight α of each type are estimated through the normal distribution probability model training data. k ;

[0137] S432, calculate the probability of each data appearing in k types of normal distribution, that is, substitute the data into k normal distributions to find the probability P(y i ) k :

[0138]

[0139] Where: y i is a data in the data set; k is the kth normal distribution;

[0140] S433, by comparing the probability P(y i ) k The value of is used to classify the sample into the cluster with the highest probability value.

[0141] S5. Classify the elasticity of users’ electricity price demand: divide users into price-sensitive users and insensitive users based on the constructed user profiles;

[0142] In step S5, a user electricity consumption behavior profile is generated based on the clustering results of the user price demand elasticity, and the users are divided into price-sensitive users and price-insensitive users, so as to facilitate the subsequent analysis of the user's inherent electricity consumption behavior attributes in this season.

[0143] In step S5, the user price demand elasticity curve is updated regularly in combination with the user's real-time feedback on the time-of-use electricity price, and the user type is iterated to improve the push accuracy.

[0144] S6. Push differentiated electricity purchasing strategies and carbon emission reduction supervision strategies to price-sensitive users and insensitive users respectively.

[0145] For example, for price-sensitive users, preferential electricity price policies can be provided to guide users to use electricity at off-peak times, that is, reduce electricity consumption during peak hours and increase electricity purchases during off-peak hours, so as to reduce their electricity costs and also reduce carbon emissions during this period. For price-insensitive users, the regulatory effect of time-of-use electricity prices is relatively weak. We can generate a model to minimize the user's electricity purchase cost and carbon emission cost, guide users to adjust their energy structure, reduce the proportion of thermal power purchases, increase the proportion of green power purchases, and help users reduce carbon emissions.

[0146] Therefore, the present invention adopts the above-mentioned user-profile-based electricity-carbon collaborative intelligent push method to accurately characterize user portraits based on the user's electricity price demand elasticity coefficient, so that the platform can customize differentiated strategies for different types of users, improve the operational capabilities of the regulatory platform, and has good application value.

[0147] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for intelligent push of electricity and carbon collaboration based on user portrait, characterized in that: The following steps are involved: S1. Collect and classify the user's annual historical load data and time-of-use electricity price data set according to the season, and perform data preprocessing; S2. Classify the daily load curves and time-of-use electricity price curves of the same season, classify the typical load curves and time-of-use electricity price curves of weekdays and weekends, and use the particle swarm optimization algorithm to extract the typical daily load curve and typical time-of-use electricity price curve of the season; S3. In the same season, based on the typical daily load curve and time-of-use electricity price curve, calculate the daily electricity price demand elasticity on weekdays and weekends; S4. Use the GMM algorithm to perform cluster analysis on electricity price demand elasticity, build user portraits, and form a user tag library; S5. Classify the elasticity of users’ electricity price demand: divide users into price-sensitive users and insensitive users based on the constructed user profiles; S6. Push differentiated electricity purchasing strategies and carbon emission reduction supervision strategies to price-sensitive users and insensitive users respectively.

2. According to the user portrait-based electric-carbon collaborative intelligent push method of claim 1, it is characterized by: The data preprocessing described in step S1 includes missing value processing and malformed value processing; Among them, missing value processing: use the daily average data of load or electricity price in the corresponding season to supplement the missing values, that is, the mean interpolation method, the formula is: In the formula, is the daily average data of load or electricity price in the corresponding season; n is the number of days in the corresponding season; x i is the load or electricity price data for the i-th day; The steps for processing abnormal values ​​are as follows: Using the Laida criterion, first calculate the average value of the load data X i is the load data of the i-th day, and the standard deviation of the load data V i =X i -X(i=1,2,3,…n); Then use the Bessel formula to calculate the standard deviation δ and set the measured value x b The standard deviation V b (1≤b≤n); When: |V b |=|x b -X|>3δ, determine x b For outliers, filter them and replace them with X; After the outlier, recalculate X and V i , conduct a new round of screening to check other abnormal data in the remaining data until all data meet the requirements.

3. The method for intelligently pushing electricity and carbon based on user portrait according to claim 1 is characterized by: The steps for extracting the seasonal typical working day load curve and the seasonal typical weekend load curve in step S2 are as follows: Step 1: Assume that it is known that the user has m+1 working days or weekends in the jth season. Then the working day or weekend load set Y of the user in this season is j ∈(m+1)×24: In the formula, is the load value of the user at the tth hour on the nth day during the jth season working day or weekend, n = 0, 1, 2, ... m, t = 0, 1, 2, ... 23; Step 2: Assume the typical load X of the user on the jth seasonal weekday or weekend j for: X j =[x0,x1,x2,…x t ,…x 23 ] (3) In the formula, x t is the typical load value of the user at hour t on a weekday or weekend in the jth season; The third step is to take the shortest distance between all coordinates of the curves as the objective function: In the formula, S j is the objective function value; Y j [n] is the user's load value on the nth day of the jth season; X j [t] is the load value at hour t in the jth season; Among them, X j Pay attention to satisfying the load constraint value and ensure that it is within the load range, that is: min{(Y j [n][t]}≤x j [t]≤max{(Y j [n][t]} (5) In the formula, min{(Y j [n][t]} is the minimum value of the load interval; max[(Y j [n][t] is the maximum value of the load range; and t=0,1,2,…,23; n=0,1,2,…,m; Step 4: Solve the objective function described in formula (4) by particle swarm optimization algorithm, and obtain the optimal X through iterative solution. j As the typical load curve in the corresponding season; The steps for extracting seasonal typical weekday time-of-use electricity price curves and seasonal typical weekend time-of-use electricity price curves are as follows: Step 1: Assume that the user has m+1 working days or weekends in the jth season. Then the time-of-use electricity price set Z for the user on working days or weekends in this season is j ∈(m+1)×24: In the formula, The electricity price for the user at the tth hour on the nth day during the jth season working day or weekend; Step 2: Assume that the typical time-of-use electricity price curve for the user’s jth seasonal working day or weekend is A j =[a0,a1,a2,…a t ,…a 23 ] (7) Where: a t is the electricity price for the user at hour t on a weekday or weekend in the jth season; The third step is to take the shortest distance between all coordinates of the curves as the objective function: Where: P j is the objective function value; m is the number of working days or weekend days in the jth season; Z j [n] is the electricity price for the user on the nth day of the jth season; a j [t] is the typical time-of-use electricity price for the user at time t in season j; Among them, a j The value at each moment cannot exceed the upper and lower limits of the corresponding moment: min{Z j [n][t]}≤a j [t]≤max{(Z j [n][t]} (9) In the formula, min{Z j [n][t]} is the lower limit of time t; it is max{(Z j [n][t]} is the upper limit value at time t; Step 4: Solve the objective function described in formula (8) by particle swarm optimization algorithm, and obtain the optimal a through iterative solution. j As the typical time-of-use electricity price curve in the corresponding season.

4. The method for intelligently pushing electricity and carbon based on user portrait according to claim 3 is characterized by: In step S3, the typical load curve and the time-of-use electricity price curve of the same season and corresponding date are combined to form a typical load-electricity price curve, and then the user's daily electricity price demand elasticity is calculated by the electricity price elasticity formula to form an electricity price demand elasticity curve; The elasticity coefficient E of the power load in the electricity price elasticity formula represents the percentage of the change ΔL in the user's electricity demand caused by the change ΔP in the electricity price within a certain period of time. The expression is as follows: Where P0 and L0 are the typical electricity price and electricity demand respectively.

5. The method for intelligently pushing electricity and carbon synergies based on user portraits according to claim 4 is characterized by: Step S4 specifically includes the following steps: S41. Determine the cluster number k of the electricity price elasticity on weekdays and weekends in the same season, and use the sum of squared errors SSE as the evaluation index for determining the cluster number: In the formula, C i is a cluster in the cluster; u i C i The mean of each data point in; x is a point in the cluster; S42, assuming the actual number of clusters is k * , compared with the actual number of clusters k * and k, based on the comparison results, determine the optimal number of clusters according to the change of SSE; S43. Use GMM to fit the data set to a mixture of multiple Gaussian distributions to discover the underlying structure of the data: S431. Assume that each type of data obeys normal distribution, and the probability model of normal distribution is: In the formula, k is the number of clusters; α k is the weight; is the probability density of the normal distribution; θ k is the parameter of probability density; y is the sample; The probability density expression is: Where: k is the standard deviation of sample y; μ k is the mean of sample y; The probability density function of k types of normal distribution and the weight α of each type are estimated through the normal distribution probability model training data. k ; S432, calculate the probability of each data appearing in k types of normal distribution, that is, substitute the data into k normal distributions to find the probability P(y i ) k : Where: y i is a data in the data set; k is the kth normal distribution; S433, by comparing the probability P(y i ) k The value of is used to classify the sample into the cluster with the highest probability value.

6. The method for intelligently pushing electricity and carbon synergy based on user portrait according to claim 1, characterized in that: In step S5, a user electricity consumption behavior profile is generated based on the clustering results of the user price demand elasticity, and the users are divided into price-sensitive users and price-insensitive users.

7. The method for intelligently pushing electricity and carbon synergies based on user portraits according to claim 6 is characterized by: In step S5, the user price demand elasticity curve is updated regularly in combination with the user's real-time feedback on the time-of-use electricity price, and the user type is iterated to improve the push accuracy.

Images