Method for deducing user-transformer relation by known intelligent electric meter data

Abstract

The invention discloses a method for deducing a user-transformer relation through known intelligent electric meter data. The method specifically comprises the steps that 1, voltage and power data, measured by an intelligent electric meter, of a user side and voltage and power data, measured by the intelligent electric meter, of a distribution transformer low-voltage side are extracted from a database; 2, performing preliminary clustering on a transformer area to which a user belongs through the historical voltage measurement data, extracted in the step 1, of the distribution transformer low-voltage side and the user side; 3, clustering the historical power measurement data of the user side measured by the intelligent electric meter in the step 1, and dividing the users into photovoltaic users and non-photovoltaic users; 4, obtaining the photovoltaic output power of the photovoltaic power generation agent station by using the photovoltaic users and the non-photovoltaic users obtained in the step 3; step 5, decomposing the power of the photovoltaic user and the transformer into photovoltaic power generation power and load consumption power through the photovoltaic output power of the photovoltaic power generation agent station obtained in the step 4; and step 6, determining the court to which the user belongs. According to the method, the user-transformer relationship can be accurately determined.

Description

Technical Field

[0001] The present invention belongs to the technical field of power distribution networks, and in particular relates to a method for inferring household-transformer relationship using known smart meter data. Background Art

[0002] With the rapid development of modern power systems, the scale of power grids is expanding, and the operation mode is more flexible, which puts higher requirements on the operation and management of distribution networks. The management level of low-voltage distribution networks is backward, which is a weak link in the power network, and it is urgent to improve the management and operation level. However, the topology of the distribution network is still unclear. The main reason is that the distribution network is directly connected to the user side. There are too many user nodes. Considering the impact of cost factors, these user nodes cannot be installed with equipment that can obtain refined electrical data.

[0003] In recent years, with the rapid growth of the number of low-voltage users, the relationship between households and transformers in the area has changed frequently, which has put forward higher requirements for the refined management level of the area. However, due to problems such as low troubleshooting efficiency and untimely information update, the archive information of household-to-transformer relationship may be wrong, affecting the development of services such as area management and fault analysis. With the continuous popularization of household photovoltaic systems, the traditional method of judging the relationship between households and transformers in the area based on the similarity of voltage timing curves has become more difficult. Due to the randomness and volatility of photovoltaic power generation, the flow and voltage characteristics of the distribution network have undergone essential changes, further exacerbating the difficulty of identifying household-to-transformer relationships based on electrical characteristics.

[0004] Traditional identification methods generally do not consider the impact of photovoltaic access, and have low accuracy and economy. Existing methods that consider distributed photovoltaic access have deficiencies in usage scenarios and data acquisition, and there is little work on household-to-transformer relationships for low-voltage distribution networks involving distributed household photovoltaic systems. Summary of the invention

[0005] The purpose of the present invention is to provide a method for inferring household-to-transformer relationship based on known smart meter data, to identify household-to-transformer relationship through known smart meter data, and to extract photovoltaic output characteristics by using voltage and power data from user smart meters when a high proportion of household photovoltaic access is connected, thereby eliminating the adverse effects of photovoltaic power generation on household-to-transformer relationship identification, and ultimately obtaining accurate household-to-transformer relationship.

[0006] The technical solution adopted by the present invention is a method for inferring household-to-electricity transformer relationship based on known smart meter data, comprising the following steps:

[0007] Step 1: Extract the voltage and power data on the user side and the voltage and power data on the low-voltage side of the distribution transformer measured by the smart meter from the database;

[0008] Step 2: Perform preliminary clustering of the substations to which the users belong using the historical voltage measurement data on the low-voltage side and user side of the distribution transformer extracted in step 1;

[0009] Step 3: Cluster the historical power measurement data on the user side measured by the smart meter in step 1 to divide the users into photovoltaic users and non-photovoltaic users;

[0010] Step 4: Using the photovoltaic users and non-photovoltaic users obtained in step 3, the photovoltaic output power of the photovoltaic power generation agent site is obtained;

[0011] Step 5: Decompose the power of photovoltaic users and transformers into photovoltaic power generation power and load consumption power according to the photovoltaic output power of the photovoltaic power generation proxy site obtained in step 4;

[0012] Step 6: Based on the user load obtained in step 5 and the load consumption power of the distribution station area, according to the power conservation principle, a multivariate linear regression equation is established through 0-1 planning to determine the distribution station area to which the user belongs.

[0013] The present invention is also characterized in that:

[0014] Step 1 is specifically: extracting from the database the voltage and power historical measurement data on the user side and the power and voltage historical measurement data on the low-voltage side of the distribution transformer measured by the smart meter for a time length of T.

[0015] Step 2 is implemented according to the following steps:

[0016] Step 2.1: Using the voltage series on the low-voltage side and user side of the distribution transformer extracted in step 1, calculate the Pearson correlation coefficient ρ between the voltage time series curves of each transformer and all users ab , define the set P as the set of all Pearson correlation coefficients, where ρ ab The expression of is shown in formula (1):

[0017]

[0018] In the formula, ρ ab is the Pearson correlation coefficient between the transformer voltage time series data a and the user voltage time series data b; cov(a, b) is the covariance between the transformer voltage time series data a and the user voltage time series data b; σ a , σ b The standard deviation of transformer voltage time series data a and user voltage time series data b respectively; X a and X b are random variables representing transformer voltage time series data a and user voltage time series data b respectively; μ a and μ bare the means of transformer voltage time series data a and user voltage time series data b respectively;

[0019] Step 2.2: Pearson correlation coefficient ρ between transformer voltage time series data a and user voltage time series data b calculated according to equation (1) in step 2.1 ab , by judging the ρ in the correlation coefficient set P ab The size of is used to judge the subordinate relationship between the distribution transformer and each user. When user b belongs to area a, the Pearson correlation coefficient between user b and area a will be greater than the Pearson correlation coefficient between user b and other transformers.

[0020] Step 3 is implemented according to the following steps:

[0021] Step 3.1: Use agglomerative hierarchical clustering to find the two closest clusters in each step of the operation and merge them. This process is repeated until the preset number of clusters is reached; the Euclidean distance is used between clusters to measure the distance between two data. The calculation formula is shown in formula (2):

[0022] d(p i,t ,p j,t )=||p i,t ,p j,t ||2 (2)

[0023] Where ||·||2 represents the L2 norm, which is the square root of the sum of the squares of each element of the vector; p i,t represents the power of user i at time t, d(p i,t ,p j,t ) represents the Euclidean distance between the power of user i and user j at time t;

[0024] Step 3.2: The similarity coefficient between the powers of two users is determined by averaging the Euclidean distance between the powers of user i and user j at all times calculated in step 3.1. The similarity coefficient r(P i ,P j ) is calculated as shown in formula (3):

[0025] r(P i ,P j )=[d(p i,1 ,p j,1 )+d(p i,2 ,p j,2 )+…+d(p i,T ,p j,T )] / T(3)

[0026] In the formula, r(P i ,P j ) represents user Pi and user P j The power time series similarity coefficient between them.

[0027] Define set R as the similarity coefficient set obtained by calculating the similarity coefficients between all users using formula (3);

[0028] Step 3.3: Using the similarity coefficient set R obtained in step 3.2, select the two users with the largest similarity coefficient to merge. This step is performed in an iterative manner, merging the two closest users in each iteration until all users are divided into two categories, one of which represents photovoltaic users and the other represents non-photovoltaic users.

[0029] Step 4 is implemented according to the following steps:

[0030] Step 4.1: First, determine the estimated photovoltaic output power set PV according to the photovoltaic users and non-photovoltaic users obtained in step 3;

[0031] Step 4.2: According to the estimated photovoltaic output power set PV obtained in step 4.1, first find the photovoltaic power generation sequence whose estimated photovoltaic output power is closest to 0 at the non-photovoltaic power generation time in the estimated photovoltaic output power set PV, and the photovoltaic power generation sequence whose estimated photovoltaic output power is closest to 0 at all photovoltaic power generation times constitutes the photovoltaic power generation proxy site set Ω;

[0032] The photovoltaic power generation time of the data in the photovoltaic power generation proxy site set Ω is removed from the upper quartile and the lower quartile, and the remaining part is averaged to obtain the power at the photovoltaic power generation time; then the power at the non-photovoltaic power generation time is set to 0, and finally the photovoltaic output power PV of the photovoltaic power generation proxy site is obtained. d .

[0033] Step 4.1 is implemented as follows:

[0034] First, determine that the power of non-PV user m at time t minus the power of PV user n at time t equals the estimated PV output power of PV user n at time t, as shown in formula (4):

[0035]

[0036] In the formula, The power of non-PV user m at time t, The power of photovoltaic user n at time t, The estimated PV output power of PV user n at time t;

[0037] The power of the PV user at the corresponding time is subtracted from the power of each non-PV user at all times to obtain the estimated PV output power of the PV user at all times. The estimated PV output power set PV is defined as the set of all estimated PV output powers.

[0038] Step 5 is implemented according to the following steps:

[0039] Step 5.1: Based on the PV proxy site established in step 4, the maximum information coefficient method is used to decompose the user power into PV power generation and load power consumption, as follows:

[0040] First, the photovoltaic output power of the photovoltaic power generation agent site multiplied by the reference photovoltaic ratio is equal to the actual photovoltaic output power of the photovoltaic user, and its calculation formula is shown in formula (5); the reference photovoltaic ratio is the ratio of the actual photovoltaic output power of the user to the photovoltaic output power of the photovoltaic agent site, and its calculation formula is shown in formula (6):

[0041]

[0042]

[0043] In the formula, is the actual photovoltaic output power of user n, PV d is the photovoltaic output power of the photovoltaic power generation proxy site, ω is the reference photovoltaic ratio, which represents the ratio of the actual photovoltaic output power to the photovoltaic output power of the photovoltaic proxy site;

[0044] The load power consumption of a photovoltaic user is the sum of the user's net load and the user's photovoltaic output power. The relationship between the actual photovoltaic output power and the load power consumption of photovoltaic user n at time t is shown in formula (7):

[0045]

[0046] In the formula, The actual load power consumption of photovoltaic user n at time t is: The net load of user n measured by the smart meter at time t;

[0047] Using the maximum information coefficient to calculate the net load of user n Photovoltaic feature extraction is performed, and the original net load data mixed with photovoltaics is processed. The estimated photovoltaic output power of the photovoltaic proxy site in formula (5) is multiplied by the reference photovoltaic multiple to obtain the actual photovoltaic output power of the user. Then, the actual photovoltaic output power of the user plus the user's net load is added to the user's load consumption power using formula (7).

[0048] Utilize user load to consume power Photovoltaic output power PV of the photovoltaic power generation agent site d The minimum value of the maximum information coefficient between And the actual photovoltaic output power of the user The actual power consumption of the user's load The reference photovoltaic ratio is calculated by taking the maximum information coefficient between them as equal, and the calculation formula is shown in formula (10):

[0049]

[0050] By establishing equation (10) to calculate the reference photovoltaic ratio, the maximum information coefficient between the user's consumption load and the photovoltaic power generation of the photovoltaic proxy site is minimized. The closer ω is to the optimal solution, the less photovoltaic information is contained in the load information, and the better the decomposition effect is, that is, the smaller the maximum information coefficient of the two is;

[0051] Finally, the actual photovoltaic power generation of user n is calculated by formula (5), and the load power consumption of the user is calculated by formula (7).

[0052] Step 5.2: Based on the photovoltaic proxy site established in step 4, the transformer power is decomposed into photovoltaic power generation and load power consumption, specifically:

[0053] The actual photovoltaic power generation of the distribution area is obtained by multiplying the estimated photovoltaic power generation of the photovoltaic proxy site in step 5.1 by the reference photovoltaic ratio, and its calculation formula is shown in formula (11); the reference photovoltaic ratio is the ratio of the actual photovoltaic output power of the transformer to the photovoltaic output power of the photovoltaic proxy site, and its calculation formula is shown in formula (12);

[0054] PV l g =PV d ×ω(11)

[0055]

[0056] Where, PV l g Photovoltaic output power in the distribution area, PV d is the photovoltaic output power of the photovoltaic power generation proxy site, ω is the reference photovoltaic ratio, which represents the ratio of the actual photovoltaic output power to the photovoltaic output power of the photovoltaic proxy site.

[0057] The relationship between the actual photovoltaic output power and load power consumption in the distribution station area is shown in formula (13):

[0058]

[0059] In the formula, It represents the actual load power consumption of distribution station area l at time t, Net load power of distribution station area l at time t.

[0060] Using the maximum information coefficient to calculate the net load of distribution area l Photovoltaic feature extraction is performed to process the original net load data mixed with photovoltaics. The estimated photovoltaic output power of the photovoltaic proxy site in formula (11) is multiplied by the reference photovoltaic ratio to obtain the actual photovoltaic output power of the transformer. Then, the user's actual photovoltaic output power plus the user's net load is added to the user's load consumption power using formula (13).

[0061] Utilize user load to consume power Photovoltaic output power PV of the photovoltaic power generation agent site d The minimum value of the maximum information coefficient between and the actual photovoltaic output power PV of the transformer l g The actual load power consumption of the transformer The reference photovoltaic ratio is calculated by taking the maximum information coefficient between them as equal, and the calculation formula is shown in formula (14):

[0062]

[0063] The reference PV ratio is calculated by establishing equation (14) so ​​that the maximum information coefficient between the consumed load power of the distribution area and the PV output power of the PV proxy site is minimized. The closer ω is to the optimal solution, the less PV information is contained in the load information, and the better the decomposition effect is, that is, the smaller the maximum information coefficient of the two is;

[0064] Finally, the actual photovoltaic power generation in the distribution area is calculated by formula (11), and the load power consumption in the distribution area is calculated by formula (13).

[0065] Step 6 is implemented according to the following steps:

[0066] Step 6.1: According to the principle of power conservation, at any time, the load consumption of the distribution transformer area is equal to the total load consumption of its users, as shown in formula (15):

[0067]

[0068] In the formula, The load power consumption of distribution station area e at time t is: The load power consumption of user f at time t, θ e Represents the set of users belonging to the distribution station area e;

[0069] For the net load data of T time sections, after decomposing the photovoltaic output in step 5, the load data matrices on the distribution transformer side and the user side are obtained as shown in equations (16) and (17):

[0070]

[0071]

[0072] Where D B Distribution station area load consumption matrix, D Y User load consumption matrix, M represents the number of distribution stations, and N represents the number of users;

[0073] Step 6.2: In order to describe the connection relationship between the distribution station area and its users, the relational connection matrix Y is introduced. The expression of Y is shown in formula (18):

[0074] Y M×N =[y e,f ] M×N (18)

[0075] In the formula, Y is a matrix composed of 0 and 1, 1 indicates that the transformer has a connection relationship with the user, and 0 indicates that there is no connection relationship;

[0076] The product of matrix Y and the user-side load data matrix plus the error and loss is the distribution station area load consumption matrix, and its calculation formula is shown in formula (19):

[0077]

[0078] In the formula, ζ represents the error and loss;

[0079] Step 6.3: Determine the relationship matrix Y between the distribution station area and the user by minimizing the error and loss. The calculation formula is shown in formula (20):

[0080]

[0081] However, due to the large number of users, the preliminary clustering result obtained in step 2 is used as the initial value to iteratively solve equation (20). The iteration ends when the iterative solution has the minimum error. At this time, the matrix Y obtained is the subordinate relationship matrix between the user and the distribution station area. The 0 and 1 in this matrix can determine the connection relationship between the transformer and the user.

[0082] The beneficial effects of the present invention are:

[0083] (1) The method of the present invention identifies the household-to-electricity transformer relationship through known smart meter data. When a high proportion of household photovoltaic access is used, the voltage and power data from the user's smart meter are used to extract photovoltaic output characteristics, eliminate the adverse effects of photovoltaic power generation on household-to-electricity transformer relationship identification, and finally obtain accurate household-to-electricity transformer relationship.

[0084] (2) The method of the present invention does not require data such as photovoltaic power generation and photovoltaic equipment physical models on the user side. When only the user net load measured by the smart meter, the user voltage, and the load and power of the transformer are known, the connection relationship between the household transformers in the low-voltage distribution network under a high proportion of household photovoltaic access is obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 The photovoltaic power generation data of the photovoltaic proxy site in Example 1;

[0086] Figure 2 The photovoltaic power generation decomposition result of a certain user in Example 1;

[0087] Figure 3 The household change relationship identification result in a certain area in Example 1;

[0088] Figure 4 Flowchart of the method of the present invention. DETAILED DESCRIPTION

[0089] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0090] The present invention provides a method for inferring the relationship between users and transformers based on the data of smart meters. When the data of smart meters are known, the low-voltage distribution station area to which the user belongs can be inferred. Figure 4 As shown, the specific steps include:

[0091] Step 1: Extract the voltage and power data on the user side and the voltage and power data on the low-voltage side of the distribution transformer measured by the smart meter from the database;

[0092] Step 1 is specifically: extracting from the database the voltage and power historical measurement data on the user side and the power and voltage historical measurement data on the low-voltage side of the distribution transformer measured by the smart meter for a time length of T.

[0093] Step 2: Perform preliminary clustering of the substations to which the users belong using the historical voltage measurement data on the low-voltage side and user side of the distribution transformer extracted in step 1;

[0094] Step 2 is implemented according to the following steps:

[0095] Step 2.1: Using the voltage series on the low-voltage side and user side of the distribution transformer extracted in step 1, calculate the Pearson correlation coefficient ρ between the voltage time series curves of each transformer and all usersab , define the set P as the set of all Pearson correlation coefficients, where ρ ab The expression of is shown in formula (1):

[0096]

[0097] In the formula, ρ ab is the Pearson correlation coefficient between the transformer voltage time series data a and the user voltage time series data b; cov(a, b) is the covariance between the transformer voltage time series data a and the user voltage time series data b; σ a , σ b The standard deviation of transformer voltage time series data a and user voltage time series data b respectively; X a and X b are random variables representing transformer voltage time series data a and user voltage time series data b respectively; μ a and μ b are the means of transformer voltage time series data a and user voltage time series data b respectively;

[0098] Step 2.2: Pearson correlation coefficient ρ between transformer voltage time series data a and user voltage time series data b calculated according to equation (1) in step 2.1 ab , by judging the ρ in the correlation coefficient set P ab The size of is used to judge the subordinate relationship between the distribution transformer and each user. When user b belongs to area a, the Pearson correlation coefficient between user b and area a will be greater than the Pearson correlation coefficient between user b and other transformers.

[0099] This can be used to preliminarily determine the subordinate relationship between the distribution station area and each user.

[0100] Step 3: Cluster the historical power measurement data on the user side measured by the smart meter in step 1 to divide the users into photovoltaic users and non-photovoltaic users;

[0101] Step 3 is implemented according to the following steps:

[0102] Step 3.1: Use agglomerative hierarchical clustering to divide user loads into two categories, namely users with photovoltaic power generation and users without photovoltaic power generation. Specifically, agglomerative hierarchical clustering is used to find the two closest clusters in each step of the operation and merge them. This process is repeated until the preset number of clusters is reached. The Euclidean distance is used between clusters to measure the distance between two data. The calculation formula is shown in formula (2):

[0103]

[0104] Where ||·||2 represents the L2 norm, which is the square root of the sum of the squares of each element of the vector; p i,t represents the power of user i at time t, d(p i,t ,p j,t ) represents the Euclidean distance between the power of user i and user j at time t;

[0105] Step 3.2: The similarity coefficient between the powers of two users is determined by averaging the Euclidean distance between the powers of user i and user j at all times calculated in step 3.1. The similarity coefficient r(P i ,P j ) is calculated as shown in formula (3):

[0106] r(P i ,P j )=[d(p i,1 ,p j,1 )+d(p i,2 ,p j,2 )+…+d(p i,T ,p j,T )] / T(3)

[0107] In the formula, r(P i ,P j ) represents user P i and user P j The power time series similarity coefficient between them.

[0108] Define set R as the similarity coefficient set obtained by calculating the similarity coefficients between all users using formula (3);

[0109] Step 3.3: Using the similarity coefficient set R obtained in step 3.2, select the two users with the largest similarity coefficient to merge. This step can be performed in an iterative manner, merging the two closest users in each iteration until all users are divided into two categories, one of which represents photovoltaic users and the other represents non-photovoltaic users.

[0110] Step 4: Using the photovoltaic users and non-photovoltaic users obtained in step 3, the photovoltaic output power of the photovoltaic power generation agent site is obtained;

[0111] Step 4 is implemented according to the following steps:

[0112] Step 4.1: First, determine the estimated photovoltaic output power set PV based on the photovoltaic users and non-photovoltaic users obtained in step 3, specifically:

[0113] Due to the same geographical location, the weather conditions in the same community are similar, such as solar radiation, ambient temperature, etc.; therefore, the irradiance per unit area and the operating temperature received by the solar photovoltaic panels should be similar, which means that the PV power generation curves of different photovoltaic panels in the entire community are similar in shape, but there are differences in amplitude due to different installed capacities. In addition, users in the same area have great similarity in electricity usage habits and behaviors. Therefore, it is determined that the power of non-PV user m at time t minus the power of PV user n at time t is equal to the estimated PV output power of PV user n at time t, as shown in formula (4):

[0114]

[0115] In the formula, The power of non-PV user m at time t, The power of photovoltaic user n at time t, Estimated PV output power of PV user n at time t.

[0116] The power of the PV user at the corresponding time is subtracted from the power of each non-PV user at all times to obtain the estimated PV output power of the PV user at all times. The estimated PV output power set PV is defined as the set of all estimated PV output powers.

[0117] Step 4.2: According to the estimated photovoltaic output power set PV obtained in step 4.1, first find the photovoltaic power generation sequence whose estimated photovoltaic output power is closest to 0 at the non-photovoltaic power generation time in the estimated photovoltaic output power set PV, and the photovoltaic power generation sequence whose estimated photovoltaic output power is closest to 0 at all photovoltaic power generation times constitutes the photovoltaic power generation proxy site set Ω;

[0118] The photovoltaic power generation time of the data in the photovoltaic power generation proxy site set Ω is removed from the upper quartile and the lower quartile, and the remaining part is averaged to obtain the power at the photovoltaic power generation time; then the power at the non-photovoltaic power generation time is set to 0, and finally the photovoltaic output power PV of the photovoltaic power generation proxy site is obtained. d .

[0119] Step 5: The photovoltaic output power PV of the photovoltaic power generation proxy site obtained in step 4 d The power of photovoltaic users and transformers is decomposed into photovoltaic power generation and load consumption power.

[0120] Step 5 is implemented according to the following steps:

[0121] Step 5.1: Based on the PV proxy site established in step 4, the maximum information coefficient method is used to decompose the user power into PV power generation and load power consumption, as follows:

[0122] First, the photovoltaic output power of the photovoltaic power generation agent site multiplied by the reference photovoltaic ratio is equal to the actual photovoltaic output power of the photovoltaic user, and its calculation formula is shown in formula (5); the reference photovoltaic ratio is the ratio of the actual photovoltaic output power of the user to the photovoltaic output power of the photovoltaic agent site, and its calculation formula is shown in formula (6):

[0123]

[0124]

[0125] In the formula, is the actual photovoltaic output power of user n, PV d is the photovoltaic output power of the photovoltaic power generation proxy site, ω is the reference photovoltaic ratio, which represents the ratio of the actual photovoltaic output power to the photovoltaic output power of the photovoltaic proxy site;

[0126] The load power consumption of a photovoltaic user is the sum of the user's net load and the user's photovoltaic output power. The relationship between the actual photovoltaic output power and the load power consumption of photovoltaic user n at time t is shown in formula (7):

[0127]

[0128] In the formula, The actual load power consumption of photovoltaic user n at time t is: The net load of user n measured by the smart meter at time t;

[0129] The maximum information coefficient (MIC) is a statistic used to measure the strength of the relationship between two variables. The scatter plot consisting of two variables is gridded with different numbers of columns and rows, that is, the scatter plot is covered with grids of different sizes. Under each gridding scheme, the probability of joint distribution and marginal distribution is estimated according to the frequency of points falling in each grid, and then the mutual information value is calculated. Among all possible gridding schemes, find the scheme that maximizes the mutual information value. The maximum mutual information value is divided by the minimum value of the logarithmic function and normalized. In gridding of different scales, the maximum value of the normalized mutual information is selected as the MIC value, which ranges from 0 to 1. The larger the value, the stronger the correlation between the two variables. The mutual information I(x, y) between x and y and the maximum information coefficient between x and y are expressed as shown in equations (8)-(9):

[0130]

[0131] Where I(x;y) represents the mutual information between x and y, p(x,y) represents the joint probability density between x and y, p(x) and p(y) represent the probability distribution of x and y respectively, a represents the number of rows of the grid, b represents the number of columns of the grid, log2min(a,b) is the normalization factor, and B is an upper limit value used to limit the size of the grid to avoid overfitting, which is usually 0.6 power of the number of data points.

[0132] Using the Maximum Information Coefficient (MIC) to calculate the net load of user n Photovoltaic feature extraction is performed, and the original net load data mixed with photovoltaics is processed. The estimated photovoltaic output power of the photovoltaic proxy site in formula (5) is multiplied by the reference photovoltaic multiple to obtain the actual photovoltaic output power of the user. Then, the actual photovoltaic output power of the user plus the user's net load is added to the user's load consumption power using formula (7).

[0133] Utilize user load to consume power Photovoltaic output power PV of the photovoltaic power generation agent site d The minimum value of the maximum information coefficient between And the actual photovoltaic output power of the user The actual power consumption of the user's load The reference photovoltaic ratio is calculated by taking the maximum information coefficient between them as equal, and the calculation formula is shown in formula (10):

[0134]

[0135] By establishing equation (10) to calculate the reference photovoltaic ratio, the maximum information coefficient between the user's consumption load and the photovoltaic power generation of the photovoltaic proxy site is minimized. The closer ω is to the optimal solution, the less photovoltaic information is contained in the load information, and the better the decomposition effect is, that is, the smaller the maximum information coefficient of the two is;

[0136] Finally, the actual photovoltaic power generation of user n is calculated by formula (5), and the load power consumption of the user is calculated by formula (7).

[0137] Step 5.2: Based on the photovoltaic proxy site established in step 4, the transformer power is decomposed into photovoltaic power generation and load power consumption, specifically:

[0138] Similarly, the actual photovoltaic power generation of the distribution station area can be obtained by multiplying the estimated photovoltaic power generation of the photovoltaic proxy site in step 5.1 by the reference photovoltaic ratio, and its calculation formula is shown in formula (11); the reference photovoltaic ratio is the ratio of the actual photovoltaic output power of the transformer to the photovoltaic output power of the photovoltaic proxy site, and its calculation formula is shown in formula (12);

[0139] PV lg =PV d ×ω(11)

[0140]

[0141] Where, PV l g Photovoltaic output power in the distribution area, PV d is the photovoltaic output power of the photovoltaic power generation proxy site, ω is the reference photovoltaic ratio, which represents the ratio of the actual photovoltaic output power to the photovoltaic output power of the photovoltaic proxy site.

[0142] The relationship between the actual photovoltaic output power and load power consumption in the distribution station area is shown in formula (13):

[0143]

[0144] In the formula, It represents the actual load power consumption of distribution station area l at time t, Net load power of distribution station area l at time t.

[0145] Using the maximum information coefficient to calculate the net load of distribution area l Photovoltaic feature extraction is performed to process the original net load data mixed with photovoltaics. The estimated photovoltaic output power of the photovoltaic proxy site in formula (11) is multiplied by the reference photovoltaic ratio to obtain the actual photovoltaic output power of the transformer. Then, the user's actual photovoltaic output power plus the user's net load is added to the user's load consumption power using formula (13).

[0146] Utilize user load to consume power Photovoltaic output power PV of the photovoltaic power generation agent site d The minimum value of the maximum information coefficient between and the actual photovoltaic output power PV of the transformer l g The actual load power consumption of the transformer The reference photovoltaic ratio is calculated by taking the maximum information coefficient between them as equal, and the calculation formula is shown in formula (14):

[0147]

[0148] The reference PV ratio is calculated by establishing equation (14) so ​​that the maximum information coefficient between the consumed load power of the distribution station area and the PV output power of the PV proxy station is minimized. The closer ω is to the optimal solution, the less PV information is contained in the load information and the better the decomposition effect is, that is, the smaller the maximum information coefficient of the two is.

[0149] Finally, the actual photovoltaic power generation in the distribution area is calculated by formula (11), and the load power consumption in the distribution area is calculated by formula (13).

[0150] Step 6: Based on the user load obtained in step 5 and the load consumption power of the distribution station area, according to the power conservation principle, a multivariate linear regression equation is established through 0-1 planning to determine the distribution station area to which the user belongs.

[0151] Step 6 is implemented according to the following steps:

[0152] Step 6.1: According to the principle of power conservation, at any time, the load consumption of the distribution transformer area is equal to the total load consumption of its users, as shown in formula (15):

[0153]

[0154] In the formula, The load power consumption of distribution station area e at time t is: The load power consumption of user f at time t, θ e Represents the set of users belonging to the distribution station area e;

[0155] For the net load data of T time sections, after decomposing the photovoltaic output in step 5, the load data matrices on the distribution transformer side and the user side are obtained as shown in equations (16) and (17):

[0156]

[0157] Where D B Distribution station area load consumption matrix, D Y User load consumption matrix, M represents the number of distribution stations, and N represents the number of users;

[0158] Step 6.2: In order to describe the connection relationship between the distribution station area and its users, the relational connection matrix Y is introduced. The expression of Y is shown in formula (18):

[0159] Y M×N =[y e,f ] M×N (18)

[0160] In the formula, Y is a matrix composed of 0 and 1, 1 indicates that the transformer has a connection relationship with the user, and 0 indicates that there is no connection relationship;

[0161] The product of matrix Y and the user-side load data matrix plus the error and loss is the distribution station area load consumption matrix, and its calculation formula is shown in formula (19):

[0162] D B =Y*D Y +ζ(19)

[0163] In the formula, ζ represents the error and loss;

[0164] Step 6.3: Determine the relationship matrix Y between the distribution station area and the user by minimizing the error and loss. The calculation formula is shown in formula (20):

[0165]

[0166] However, due to the large number of users, the preliminary clustering result obtained in step 2 is used as the initial value to iteratively solve equation (20). The iteration ends when the iterative solution has the minimum error. At this time, the matrix Y obtained is the subordinate relationship matrix between the user and the distribution station area. The 0 and 1 in this matrix can determine the connection relationship between the transformer and the user.

[0167] Example 1

[0168] This article uses data from a certain area, including 3 distribution substations and 184 end users, and takes the actual measurement data of smart meters at 30-minute intervals as an example. Among them, No. 1 distribution transformer has 64 households, No. 2 distribution transformer has 57 households, and No. 3 distribution transformer has 63 households. Among them, there are 39 users who have installed photovoltaic power generation.

[0169] The photovoltaic power generation data of a photovoltaic agent site on a certain day is as follows: Figure 1 As shown, from Figure 1 From this, we can see the changing trend of the photovoltaic power generation in the area on a certain day. This curve also represents the changing trend of the photovoltaic power generation of all local users. The different installed capacity of users will only cause the peak value of the curve to change.

[0170] The decomposition results of a user's photovoltaic power generation are as follows: Figure 2 As shown;

[0171] The results of household change relationship identification in a certain area are as follows: Figure 3 As shown, from Figure 3 It can be seen that the present invention has a higher accuracy rate.

[0172] Conclusion: The method of the present invention identifies household-to-variable relationship in a distribution network containing a high proportion of distributed photovoltaics. By clustering user voltages, establishing photovoltaic proxy sites, extracting photovoltaic output characteristics through the maximum information coefficient, and establishing a multivariate linear regression equation to solve the household-to-variable relationship, a relatively accurate household-to-variable relationship can be obtained. The present invention has a high accuracy rate.

[0173] Example 2

[0174] The method for inferring household-to-electricity transformer relationship from known smart meter data includes the following steps:

[0175] Step 1: Extract the voltage and power data on the user side and the voltage and power data on the low-voltage side of the distribution transformer measured by the smart meter from the database;

[0176] Step 2: Perform preliminary clustering of the substations to which the users belong using the historical voltage measurement data on the low-voltage side and user side of the distribution transformer extracted in step 1;

[0177] Step 3: Cluster the historical power measurement data on the user side measured by the smart meter in step 1 to divide the users into photovoltaic users and non-photovoltaic users;

[0178] Step 4: Using the photovoltaic users and non-photovoltaic users obtained in step 3, the photovoltaic output power of the photovoltaic power generation agent site is obtained;

[0179] Step 5: Decompose the power of photovoltaic users and transformers into photovoltaic power generation power and load consumption power according to the photovoltaic output power of the photovoltaic power generation proxy site obtained in step 4;

[0180] Step 6: Based on the user load obtained in step 5 and the load consumption power of the distribution station area, according to the power conservation principle, a multivariate linear regression equation is established through 0-1 planning to determine the distribution station area to which the user belongs.

[0181] Example 3

[0182] The method for inferring household-to-electricity transformer relationship from known smart meter data includes the following steps:

[0183] Step 1: Extract the voltage and power data on the user side and the voltage and power data on the low-voltage side of the distribution transformer measured by the smart meter from the database;

[0184] Step 1 is specifically: extracting from the database the voltage and power historical measurement data on the user side and the power and voltage historical measurement data on the low-voltage side of the distribution transformer measured by the smart meter for a time length of T.

[0185] Step 2: Perform preliminary clustering of the substations to which the users belong using the historical voltage measurement data on the low-voltage side and user side of the distribution transformer extracted in step 1;

[0186] Step 3: Cluster the historical power measurement data on the user side measured by the smart meter in step 1 to divide the users into photovoltaic users and non-photovoltaic users;

[0187] Step 4: Using the photovoltaic users and non-photovoltaic users obtained in step 3, the photovoltaic output power of the photovoltaic power generation agent site is obtained;

[0188] Step 5: Decompose the power of photovoltaic users and transformers into photovoltaic power generation power and load consumption power according to the photovoltaic output power of the photovoltaic power generation proxy site obtained in step 4;

[0189] Step 6: Based on the user load obtained in step 5 and the load consumption power of the distribution station area, according to the power conservation principle, a multivariate linear regression equation is established through 0-1 planning to determine the distribution station area to which the user belongs.

[0190] Example 4

[0191] The method for inferring household-to-electricity transformer relationship from known smart meter data includes the following steps:

[0192] Step 1: Extract the voltage and power data on the user side and the voltage and power data on the low-voltage side of the distribution transformer measured by the smart meter from the database;

[0193] Step 1 is specifically: extracting from the database the voltage and power historical measurement data on the user side and the power and voltage historical measurement data on the low-voltage side of the distribution transformer measured by the smart meter for a time length of T.

[0194] Step 2: Perform preliminary clustering of the substations to which the users belong using the historical voltage measurement data on the low-voltage side and user side of the distribution transformer extracted in step 1;

[0195] Step 2 is implemented according to the following steps:

[0196] Step 2.1: Using the voltage series on the low-voltage side and user side of the distribution transformer extracted in step 1, calculate the Pearson correlation coefficient ρ between the voltage time series curves of each transformer and all users ab , define the set P as the set of all Pearson correlation coefficients,; where, ρ ab The expression of is shown in formula (1):

[0197]

[0198] In the formula, ρ ab is the Pearson correlation coefficient between the transformer voltage time series data a and the user voltage time series data b; cov(a, b) is the covariance between the transformer voltage time series data a and the user voltage time series data b; σ a , σ b The standard deviation of transformer voltage time series data a and user voltage time series data b respectively; X a and X b are random variables representing transformer voltage time series data a and user voltage time series data b respectively; μ a and μ b are the means of transformer voltage time series data a and user voltage time series data b respectively;

[0199] Step 2.2: Pearson correlation coefficient ρ between transformer voltage time series data a and user voltage time series data b calculated according to equation (1) in step 2.1 ab, by judging the ρ in the correlation coefficient set P ab The size of is used to judge the subordinate relationship between the distribution transformer and each user. When user b belongs to area a, the Pearson correlation coefficient between user b and area a will be greater than the Pearson correlation coefficient between user b and other transformers.

[0200] Step 3: Cluster the historical power measurement data on the user side measured by the smart meter in step 1 to divide the users into photovoltaic users and non-photovoltaic users;

[0201] Step 4: Using the photovoltaic users and non-photovoltaic users obtained in step 3, the photovoltaic output power of the photovoltaic power generation agent site is obtained;

[0202] Step 5: Decompose the power of photovoltaic users and transformers into photovoltaic power generation power and load consumption power according to the photovoltaic output power of the photovoltaic power generation proxy site obtained in step 4;

[0203] Step 6: Based on the user load obtained in step 5 and the load consumption power of the distribution station area, according to the power conservation principle, a multivariate linear regression equation is established through 0-1 planning to determine the distribution station area to which the user belongs.

[0204] Example 5

[0205] The method for inferring household-to-electricity transformer relationship from known smart meter data includes the following steps:

[0206] Step 1: Extract the voltage and power data on the user side and the voltage and power data on the low-voltage side of the distribution transformer measured by the smart meter from the database;

[0207] Step 1 is specifically: extracting from the database the voltage and power historical measurement data on the user side and the power and voltage historical measurement data on the low-voltage side of the distribution transformer measured by the smart meter for a time length of T.

[0208] Step 2: Perform preliminary clustering of the substations to which the users belong using the historical voltage measurement data on the low-voltage side and user side of the distribution transformer extracted in step 1;

[0209] Step 2 is implemented according to the following steps:

[0210] Step 2.1: Using the voltage series on the low-voltage side and user side of the distribution transformer extracted in step 1, calculate the Pearson correlation coefficient ρ between the voltage time series curves of each transformer and all users ab , define the set P as the set of all Pearson correlation coefficients, where ρ ab The expression of is shown in formula (1):

[0211]

[0212] In the formula, ρ ab is the Pearson correlation coefficient between the transformer voltage time series data a and the user voltage time series data b; cov(a, b) is the covariance between the transformer voltage time series data a and the user voltage time series data b; σ a , σ b The standard deviation of transformer voltage time series data a and user voltage time series data b respectively; X a and X b are random variables representing transformer voltage time series data a and user voltage time series data b respectively; μ a and μ b are the means of transformer voltage time series data a and user voltage time series data b respectively;

[0213] Step 2.2: Pearson correlation coefficient ρ between transformer voltage time series data a and user voltage time series data b calculated according to equation (1) in step 2.1 ab , by judging the ρ in the correlation coefficient set P ab The size of is used to judge the subordinate relationship between the distribution transformer and each user. When user b belongs to area a, the Pearson correlation coefficient between user b and area a will be greater than the Pearson correlation coefficient between user b and other transformers.

[0214] Step 3: Cluster the historical power measurement data on the user side measured by the smart meter in step 1 to divide the users into photovoltaic users and non-photovoltaic users;

[0215] Step 3 is implemented according to the following steps:

[0216] Step 3.1: Use agglomerative hierarchical clustering to find the two closest clusters in each step of the operation and merge them. This process is repeated until the preset number of clusters is reached; the Euclidean distance is used between clusters to measure the distance between two data. The calculation formula is shown in formula (2):

[0217] d(p i,t ,p j,t )=||p i,t ,p j,t ||2 (2)

[0218] Where ||·||2 represents the L2 norm, which is the square root of the sum of the squares of each element of the vector; p i,t represents the power of user i at time t, d(p i,t ,p j,t ) represents the Euclidean distance between the power of user i and user j at time t;

[0219] Step 3.2: The similarity coefficient between the powers of two users is determined by averaging the Euclidean distance between the powers of user i and user j at all times calculated in step 3.1. The similarity coefficient r(P i ,P j ) is calculated as shown in formula (3):

[0220] r(P i ,P j )=[d(p i,1 ,p j,1 )+d(p i,2 ,p j,2 )+…+d(p i,T ,p j,T )] / T (3)

[0221] In the formula, r(P i ,P j ) represents user P i and user P j The power time series similarity coefficient between them.

[0222] Define set R as the similarity coefficient set obtained by calculating the similarity coefficients between all users using formula (3);

[0223] Step 3.3: Using the similarity coefficient set R obtained in step 3.2, select the two users with the largest similarity coefficient to merge. This step is performed in an iterative manner, merging the two closest users in each iteration until all users are divided into two categories, one of which represents photovoltaic users and the other represents non-photovoltaic users.

[0224] Step 4: Using the photovoltaic users and non-photovoltaic users obtained in step 3, the photovoltaic output power of the photovoltaic power generation agent site is obtained;

[0225] Step 5: Decompose the power of photovoltaic users and transformers into photovoltaic power generation power and load consumption power according to the photovoltaic output power of the photovoltaic power generation proxy site obtained in step 4;

[0226] Step 6: Based on the user load obtained in step 5 and the load consumption power of the distribution station area, according to the power conservation principle, a multivariate linear regression equation is established through 0-1 planning to determine the distribution station area to which the user belongs.

[0227] Example 6

[0228] The method for inferring household-to-electricity transformer relationship from known smart meter data includes the following steps:

[0229] Step 1: Extract the voltage and power data on the user side and the voltage and power data on the low-voltage side of the distribution transformer measured by the smart meter from the database;

[0230] Step 1 is specifically: extracting from the database the voltage and power historical measurement data on the user side and the power and voltage historical measurement data on the low-voltage side of the distribution transformer measured by the smart meter for a time length of T.

[0231] Step 2: Perform preliminary clustering of the substations to which the users belong using the historical voltage measurement data on the low-voltage side and user side of the distribution transformer extracted in step 1;

[0232] Step 2 is implemented according to the following steps:

[0233] Step 2.1: Using the voltage series on the low-voltage side and user side of the distribution transformer extracted in step 1, calculate the Pearson correlation coefficient ρ between the voltage time series curves of each transformer and all users ab , define the set P as the set of all Pearson correlation coefficients, where ρ ab The expression of is shown in formula (1):

[0234]

[0235] In the formula, ρ ab is the Pearson correlation coefficient between the transformer voltage time series data a and the user voltage time series data b; cov(a, b) is the covariance between the transformer voltage time series data a and the user voltage time series data b; σ a , σ b The standard deviation of transformer voltage time series data a and user voltage time series data b respectively; X a and X b are random variables representing transformer voltage time series data a and user voltage time series data b respectively; μ a and μ b are the means of transformer voltage time series data a and user voltage time series data b respectively;

[0236] Step 2.2: Pearson correlation coefficient ρ between transformer voltage time series data a and user voltage time series data b calculated according to equation (1) in step 2.1 ab , by judging the ρ in the correlation coefficient set P ab The size of is used to judge the subordinate relationship between the distribution transformer and each user. When user b belongs to area a, the Pearson correlation coefficient between user b and area a will be greater than the Pearson correlation coefficient between user b and other transformers.

[0237] Step 3: Cluster the historical power measurement data on the user side measured by the smart meter in step 1 to divide the users into photovoltaic users and non-photovoltaic users;

[0238] Step 3 is implemented according to the following steps:

[0239] Step 3.1: Use agglomerative hierarchical clustering to find the two closest clusters in each step of the operation and merge them. This process is repeated until the preset number of clusters is reached; the Euclidean distance is used between clusters to measure the distance between two data. The calculation formula is shown in formula (2):

[0240] d(p i,t ,p j,t )=||p i,t ,p j,t ||2 (2)

[0241] Where ||·||2 represents the L2 norm, which is the square root of the sum of the squares of each element of the vector; p i,t represents the power of user i at time t, d(p i,t ,p j,t ) represents the Euclidean distance between the power of user i and user j at time t;

[0242] Step 3.2: The similarity coefficient between the powers of two users is determined by averaging the Euclidean distance between the powers of user i and user j at all times calculated in step 3.1. The similarity coefficient r(P i ,P j ) is calculated as shown in formula (3):

[0243] r(P i ,P j )=[d(p i,1 ,p j,1 )+d(p i,2 ,p j,2 )+…+d(p i,T ,p j,T )] / T (3)

[0244] In the formula, r(P i ,P j ) represents user P i and user P j The power time series similarity coefficient between them.

[0245] Define set R as the similarity coefficient set obtained by calculating the similarity coefficients between all users using formula (3);

[0246] Step 3.3: Using the similarity coefficient set R obtained in step 3.2, select the two users with the largest similarity coefficient to merge. This step is performed in an iterative manner, merging the two closest users in each iteration until all users are divided into two categories, one of which represents photovoltaic users and the other represents non-photovoltaic users.

[0247] Step 4: Using the photovoltaic users and non-photovoltaic users obtained in step 3, the photovoltaic output power of the photovoltaic power generation agent site is obtained;

[0248] Step 4 is implemented according to the following steps:

[0249] Step 4.1: First, determine the estimated photovoltaic output power set PV according to the photovoltaic users and non-photovoltaic users obtained in step 3;

[0250] Step 4.2: According to the estimated photovoltaic output power set PV obtained in step 4.1, first find the photovoltaic power generation sequence whose estimated photovoltaic output power is closest to 0 at the non-photovoltaic power generation time in the estimated photovoltaic output power set PV, and the photovoltaic power generation sequence whose estimated photovoltaic output power is closest to 0 at all photovoltaic power generation times constitutes the photovoltaic power generation proxy site set Ω;

[0251] The photovoltaic power generation time of the data in the photovoltaic power generation proxy site set Ω is removed from the upper quartile and the lower quartile, and the remaining part is averaged to obtain the power at the photovoltaic power generation time; then the power at the non-photovoltaic power generation time is set to 0, and finally the photovoltaic output power PV of the photovoltaic power generation proxy site is obtained. d .

[0252] Step 5: Decompose the power of photovoltaic users and transformers into photovoltaic power generation power and load consumption power according to the photovoltaic output power of the photovoltaic power generation proxy site obtained in step 4;

[0253] Step 6: Based on the user load obtained in step 5 and the load consumption power of the distribution station area, according to the power conservation principle, a multivariate linear regression equation is established through 0-1 planning to determine the distribution station area to which the user belongs.

Claims

1. A method for inferring household-to-electricity transformer relationship based on known smart meter data, characterized in that: The following steps are involved: Step 1: Extract the voltage and power data on the user side and the voltage and power data on the low-voltage side of the distribution transformer measured by the smart meter from the database; Step 2: Perform preliminary clustering of the substations to which the users belong using the historical voltage measurement data on the low-voltage side and user side of the distribution transformer extracted in step 1; Step 3: Cluster the historical power measurement data on the user side measured by the smart meter in step 1 to divide the users into photovoltaic users and non-photovoltaic users; Step 4: Using the photovoltaic users and non-photovoltaic users obtained in step 3, the photovoltaic output power of the photovoltaic power generation agent site is obtained; Step 5: Decompose the power of photovoltaic users and transformers into photovoltaic power generation power and load consumption power according to the photovoltaic output power of the photovoltaic power generation proxy site obtained in step 4; Step 6: Based on the user load obtained in step 5 and the load consumption power of the distribution station area, according to the power conservation principle, a multivariate linear regression equation is established through 0-1 planning to determine the distribution station area to which the user belongs.

2. The method for inferring household-to-electricity transformer relationship from known smart meter data according to claim 1, characterized in that: Step 1 is specifically: extracting from the database the voltage and power historical measurement data on the user side and the power and voltage historical measurement data on the low-voltage side of the distribution transformer measured by the smart meter for a time length of T.

3. The method for inferring household-to-electricity transformer relationship from known smart meter data according to claim 2, characterized in that: Step 2 is implemented according to the following steps: Step 2.1: Using the voltage series on the low-voltage side and user side of the distribution transformer extracted in step 1, calculate the Pearson correlation coefficient ρ between the voltage time series curves of each transformer and all users ab , define the set P as the set of all Pearson correlation coefficients, where ρ ab The expression of is shown in formula (1): In the formula, ρa b is the Pearson correlation coefficient between the transformer voltage time series data a and the user voltage time series data b; cov(a, b) is the covariance between the transformer voltage time series data a and the user voltage time series data b; σ a , σ b The standard deviation of transformer voltage time series data a and user voltage time series data b respectively; X a and X b are random variables representing transformer voltage time series data a and user voltage time series data b respectively; μ a and μ b are the means of transformer voltage time series data a and user voltage time series data b respectively; Step 2.2: Pearson correlation coefficient ρ between transformer voltage time series data a and user voltage time series data b calculated according to equation (1) in step 2.1 ab , by judging the ρ in the correlation coefficient set P ab The size of is used to judge the subordinate relationship between the distribution transformer and each user. When user b belongs to area a, the Pearson correlation coefficient between user b and area a will be greater than the Pearson correlation coefficient between user b and other transformers.

4. The method for inferring household-to-electricity transformer relationship from known smart meter data according to claim 3, characterized in that: Step 3 is implemented according to the following steps: Step 3.1: Use agglomerative hierarchical clustering to find the two closest clusters in each step of the operation and merge them. This process is repeated until the preset number of clusters is reached; the Euclidean distance is used between clusters to measure the distance between two data. The calculation formula is shown in formula (2): d(p i，t ,p j，t )=||p i，t ,p j，t ||2 (2) Where ∥·∥2 represents the L2 norm, which is the square root of the sum of the squares of each element of the vector; p i,t represents the power of user i at time t, d(p i,t ,p j,t ) represents the Euclidean distance between the power of user i and user j at time t; Step 3.2: The similarity coefficient between the powers of two users is determined by averaging the Euclidean distance between the powers of user i and user j at all times calculated in step 3.

1. The similarity coefficient r(P i ,P j ) is calculated as shown in formula (3): r(P i ,P j )=[d(p i,1 ,p j,1 )+d(p i,2 ,p j,2 )+…+d(p i,T ,p j,T )] / T (3) In the formula, r(P i ,P j ) represents user P i and user P j The power time series similarity coefficient between them; Define set R as the similarity coefficient set obtained by calculating the similarity coefficients between all users using formula (3); Step 3.3: Using the similarity coefficient set R obtained in step 3.2, select the two users with the largest similarity coefficient to merge. This step is performed in an iterative manner, merging the two closest users in each iteration until all users are divided into two categories, one of which represents photovoltaic users and the other represents non-photovoltaic users.

5. The method for inferring household-to-electricity transformer relationship from known smart meter data according to claim 4, characterized in that: Step 4 is implemented according to the following steps: Step 4.1: First, determine the estimated photovoltaic output power set PV according to the photovoltaic users and non-photovoltaic users obtained in step 3; Step 4.2: According to the estimated photovoltaic output power set PV obtained in step 4.1, first find the photovoltaic power generation sequence whose estimated photovoltaic output power is closest to 0 at the non-photovoltaic power generation time in the estimated photovoltaic output power set PV, and the photovoltaic power generation sequence whose estimated photovoltaic output power is closest to 0 at all photovoltaic power generation times constitutes the photovoltaic power generation proxy site set Ω; The photovoltaic power generation time of the data in the photovoltaic power generation proxy site set Ω is removed from the upper quartile and the lower quartile, and the remaining part is averaged to obtain the power at the photovoltaic power generation time; then the power at the non-photovoltaic power generation time is set to 0, and finally the photovoltaic output power PV of the photovoltaic power generation proxy site is obtained. d .

6. The method for inferring household-to-electricity transformer relationship from known smart meter data according to claim 5, characterized in that: Step 4.1 is implemented as follows: First, determine that the power of non-PV user m at time t minus the power of PV user n at time t equals the estimated PV output power of PV user n at time t, as shown in formula (4): In the formula, The power of non-PV user m at time t, The power of photovoltaic user n at time t, The estimated PV output power of PV user n at time t; The power of the PV user at the corresponding time is subtracted from the power of each non-PV user at all times to obtain the estimated PV output power of the PV user at all times. The estimated PV output power set PV is defined as the set of all estimated PV output powers.

7. The method for inferring household-to-electricity transformer relationship from known smart meter data according to claim 5, characterized in that: Step 5 is implemented according to the following steps: Step 5.1: Based on the PV proxy site established in step 4, the maximum information coefficient method is used to decompose the user power into PV power generation and load power consumption, as follows: First, the photovoltaic output power of the photovoltaic power generation agent site multiplied by the reference photovoltaic ratio is equal to the actual photovoltaic output power of the photovoltaic user, and its calculation formula is shown in formula (5); the reference photovoltaic ratio is the ratio of the actual photovoltaic output power of the user to the photovoltaic output power of the photovoltaic agent site, and its calculation formula is shown in formula (6): In the formula, is the actual photovoltaic output power of user n, PV d is the photovoltaic output power of the photovoltaic power generation proxy site, ω is the reference photovoltaic ratio, which represents the ratio of the actual photovoltaic output power to the photovoltaic output power of the photovoltaic proxy site; The load power consumption of a photovoltaic user is the sum of the user's net load and the user's photovoltaic output power. The relationship between the actual photovoltaic output power and the load power consumption of photovoltaic user n at time t is shown in formula (7): In the formula, The actual load power consumption of photovoltaic user n at time t is: The net load of user n measured by the smart meter at time t; Using the maximum information coefficient to calculate the net load of user n Photovoltaic feature extraction is performed, and the original net load data mixed with photovoltaics is processed. The estimated photovoltaic output power of the photovoltaic proxy site in formula (5) is multiplied by the reference photovoltaic multiple to obtain the actual photovoltaic output power of the user. Then, the actual photovoltaic output power of the user plus the user's net load is added to the user's load consumption power using formula (7). Utilize user load to consume power Photovoltaic output power PV of the photovoltaic power generation agent site d The minimum value of the maximum information coefficient between And the actual photovoltaic output power of the user The actual power consumption of the user's load The reference photovoltaic ratio is calculated by taking the maximum information coefficient between them as equal, and the calculation formula is shown in formula (10): By establishing equation (10) to calculate the reference photovoltaic ratio, the maximum information coefficient between the user's consumption load and the photovoltaic power generation of the photovoltaic proxy site is minimized. The closer ω is to the optimal solution, the less photovoltaic information is contained in the load information, and the better the decomposition effect is, that is, the smaller the maximum information coefficient of the two is; Finally, the actual photovoltaic power generation of user n is calculated by formula (5), and the load power consumption of the user is calculated by formula (7); Step 5.2: Based on the photovoltaic proxy site established in step 4, the transformer power is decomposed into photovoltaic power generation and load power consumption, specifically: The actual photovoltaic power generation of the distribution area is obtained by multiplying the estimated photovoltaic power generation of the photovoltaic proxy site in step 5.1 by the reference photovoltaic ratio, and its calculation formula is shown in formula (11); the reference photovoltaic ratio is the ratio of the actual photovoltaic output power of the transformer to the photovoltaic output power of the photovoltaic proxy site, and its calculation formula is shown in formula (12); PV l g =PV d ×ω (11) In the formula, PV l g Photovoltaic output power in the distribution area, PV d is the photovoltaic output power of the photovoltaic power generation proxy site, ω is the reference photovoltaic ratio, which represents the ratio of the actual photovoltaic output power to the photovoltaic output power of the photovoltaic proxy site; The relationship between the actual photovoltaic output power and load power consumption in the distribution station area is shown in formula (13): In the formula, It represents the actual load power consumption of distribution station area l at time t, Net load power of distribution station area l at time t; Using the maximum information coefficient to calculate the net load of distribution area l Photovoltaic feature extraction is performed to process the original net load data mixed with photovoltaics. The estimated photovoltaic output power of the photovoltaic proxy site in formula (11) is multiplied by the reference photovoltaic ratio to obtain the actual photovoltaic output power of the transformer. Then, the user's actual photovoltaic output power plus the user's net load is added to the user's load consumption power using formula (13). Utilize user load to consume power Photovoltaic output power PV of the photovoltaic power generation agent site d The minimum value of the maximum information coefficient between and the actual photovoltaic output power PV of the transformer l g The actual load power consumption of the transformer The reference photovoltaic ratio is calculated by taking the maximum information coefficient between them as equal, and the calculation formula is shown in formula (14): The reference PV ratio is calculated by establishing equation (14) so ​​that the maximum information coefficient between the consumed load power of the distribution area and the PV output power of the PV proxy site is minimized. The closer ω is to the optimal solution, the less PV information is contained in the load information, and the better the decomposition effect is, that is, the smaller the maximum information coefficient of the two is; Finally, the actual photovoltaic power generation in the distribution area is calculated by formula (11), and the load power consumption in the distribution area is calculated by formula (13).

8. The method for inferring household-to-electricity transformer relationship from known smart meter data according to claim 7, characterized in that: Step 6 is implemented according to the following steps: Step 6.1: According to the principle of power conservation, at any time, the load consumption of the distribution transformer area is equal to the total load consumption of its users, as shown in formula (15): In the formula, The load power consumption of distribution station area e at time t is: The load power consumption of user f at time t, θ e Represents the set of users belonging to the distribution station area e; For the net load data of T time sections, after decomposing the photovoltaic output in step 5, the load data matrices on the distribution transformer side and the user side are obtained as shown in equations (16) and (17): Where D B Distribution station area load consumption matrix, D Y User load consumption matrix, M represents the number of distribution stations, N represents the number of users; Step 6.2: In order to describe the connection relationship between the distribution station area and its users, the relational connection matrix Y is introduced. The expression of Y is shown in formula (18): AND M×N =[and e,f ] M×N (18) In the formula, Y is a matrix composed of 0 and 1, 1 indicates that the transformer has a connection relationship with the user, and 0 indicates that there is no connection relationship; The product of matrix Y and the user-side load data matrix plus the error and loss is the distribution station area load consumption matrix, and its calculation formula is shown in formula (19): D B =Y*D Y +g (19) In the formula, ζ represents the error and loss; Step 6.3: Determine the relationship matrix Y between the distribution station area and the user by minimizing the error and loss. The calculation formula is shown in formula (20): However, due to the large number of users, the preliminary clustering result obtained in step 2 is used as the initial value to iteratively solve equation (20). The iteration ends when the iterative solution has the minimum error. At this time, the matrix Y obtained is the subordinate relationship matrix between the user and the distribution station area. The 0 and 1 in this matrix can determine the connection relationship between the transformer and the user.

Images