Photovoltaic decomposition method for post-metering system based on photovoltaic output correlation

Abstract

The present invention provides a photovoltaic decomposition method of a post-meter system based on photovoltaic output correlation, which relates to the field of distribution networks. The method is as follows: pre-processing historical measurement data based on min-max normalization; defining the upper and lower envelopes of the user's net load data, and estimating the photovoltaic installed capacity of the post-meter on this basis; extracting features of a typical user load curve based on DBSCAN clustering; establishing a post-meter photovoltaic user load fitting model based on the user load characteristic curve, establishing a photovoltaic output fitting model of the post-meter photovoltaic user based on photovoltaic output correlation, and constructing a photovoltaic output correlation coefficient solution model on this basis; based on the photovoltaic output correlation coefficient obtained by solving, combined with the real-time measurement data of the net load of the post-meter photovoltaic user, real-time decomposition of the post-meter photovoltaic output is performed. The present invention realizes real-time decomposition of post-meter photovoltaic output based on limited measurement information, can enhance the observability of the post-meter photovoltaic user to the front side of the meter, and is helpful to formulate an optimal scheduling strategy.

Description

Technical Field

[0001] The present invention relates to the field of decoupling and decomposition of distribution network front-end measurement data, and in particular to a photovoltaic decomposition method of a back-end system based on photovoltaic output correlation. Background Art

[0002] The popularization of the post-meter system has made users with small-scale distributed energy (photovoltaic) no longer just electricity consumers, but gradually transformed into electricity producers and consumers, which can enable users to participate more fully in the electricity market. The promotion of the post-meter system can not only increase the flexibility of the power system, but also improve the utilization rate of distributed energy, optimize energy allocation, and reduce users' electricity costs. However, since the power generation behavior of photovoltaic power behind the meter is difficult to be recorded by the smart meter installed on the user side, the front side of the meter cannot obtain the actual power generation of the user's photovoltaic power; and since the net load of the user obtained by the front side of the meter is the difference between the actual load and photovoltaic power generation, photovoltaic power generation will interfere with the analysis and judgment of the actual load and electricity consumption behavior of the user by the front side of the meter.

[0003] At present, the research on post-meter photovoltaic decomposition of distribution networks is mainly divided into data-driven photovoltaic decomposition methods and post-meter photovoltaic decomposition methods based on photovoltaic system models. Among them, the data-driven photovoltaic decomposition method can be further divided into data-driven decomposition methods that rely on energy proxy sites and pure data-driven photovoltaic output decomposition methods. The former uses users or photovoltaic sites adjacent to the users to be decomposed as energy proxy sites, extracts features based on the data of energy proxy sites, and predicts and decomposes the photovoltaic output and load data of the users to be decomposed. The latter does not rely on energy proxy sites, but uses machine learning, deep learning and other means to establish a more complex nonlinear model for photovoltaic output decomposition. The post-meter photovoltaic decomposition method based on the photovoltaic system model builds a photoelectric conversion physical model based on meteorological data, photovoltaic system physical set data and geographic information data, and realizes the decomposition of post-meter photovoltaic output through this model.

[0004] In the above studies, the decomposition method based on the photovoltaic system model is relatively complex and requires a large amount of additional data for assistance. The data-driven decomposition method has a small demand for data types, but most of them directly fit the photovoltaic output or user load data, or train complex nonlinear models through machine learning, deep learning and other means to predict photovoltaic output, lacking certain physical principles to support it. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the present invention provides a photovoltaic decomposition method for a post-meter system based on photovoltaic output correlation.

[0006] The present invention is achieved through the following technical solutions.

[0007] According to one aspect of the present invention, a photovoltaic decomposition method for a post-metering system based on photovoltaic output correlation is provided, comprising:

[0008] S1. Preprocess the historical measurement data based on min-max normalization;

[0009] S2. Define the upper and lower envelopes of the user's net load data, and estimate the photovoltaic installed capacity after the meter on this basis;

[0010] S3, based on DBSCAN clustering, feature extraction of typical user load curves;

[0011] S4. Establish a post-meter photovoltaic user load fitting model based on the user load characteristic curve, establish a post-meter photovoltaic user photovoltaic output fitting model based on the photovoltaic output correlation, and build a photovoltaic output correlation coefficient solution model on this basis;

[0012] S5. Based on the solved photovoltaic output correlation coefficient and combined with the real-time measurement data of the net load of the photovoltaic users behind the meter, the real-time decomposition of the photovoltaic output behind the meter is performed.

[0013] Preferably, the method of preprocessing the historical measurement data based on min-max normalization in step S1 is specifically as follows:

[0014] The historical measurement data that can be obtained by the distribution network side based on the smart meter include: the historical measurement value of the net load of the photovoltaic users to be decomposed, the historical measurement value of the load of the pure load users in the same area, and the historical measurement value of the photovoltaic output of the reference photovoltaic measurement site. The sampling interval of the above historical measurement data is 15 minutes, corresponding to 96 points a day. The historical measurement data is normalized for subsequent clustering and output decomposition.

[0015] The historical measurement data uses the following min-max normalization formula:

[0016]

[0017] in, is the normalized historical measurement data of measurement point i; is the original historical measurement data of measurement point i; are the maximum and minimum values ​​in the original historical measurement data set of measurement point i, respectively; h is the sequence number of 96 data points in one day, and x(h) represents the hth data point in a set of data x.

[0018] Preferably, the step S2 defines the upper and lower envelopes of the user's net load data, and on this basis, performs a method for estimating the photovoltaic installed capacity after the meter, specifically as follows:

[0019] The net load curve of users with photovoltaic power behind the meter will show great differences under different lighting conditions. This feature can be used to estimate the installed capacity of photovoltaic power behind the meter.

[0020] Based on the normalized photovoltaic output data of the reference photovoltaic site, the photovoltaic output is screened to be in the range [l lower ,l upper ] in the date tag collection T pv,low , where 0≤l lower ≤l upper ≤0.2. Define the photovoltaic user i (i=1,2,...,N BTM ) on date T pv,low The minimum value of all net load curves is the upper envelope:

[0021]

[0022] in, is the upper envelope of PV user i after decomposition; is the original net load data of user i on day t.

[0023] After defining the table to be decomposed, the photovoltaic user i (i = 1, 2, ..., N BTM ) is the minimum value of all net load curves of the lower envelope:

[0024]

[0025] in, is the lower envelope of PV user i after decomposition; T is the total number of days in the historical data set.

[0026] The schematic diagram of the upper and lower envelopes of the net load of photovoltaic users behind the meter is as follows Figure 1 Based on the upper and lower envelopes, the estimated installed capacity of photovoltaic power can be obtained:

[0027]

[0028] Among them, C i is the estimated value of the photovoltaic installed capacity of photovoltaic user i after decomposition.

[0029] Preferably, the method of extracting features from typical user load curves based on DBSCAN clustering in step S3 is specifically as follows:

[0030] For the normalized dataset X norm , where the total number of elements is k, the DBSCAN clustering algorithm has the following definition:

[0031] 1)ε-neighborhood: distance from data point The area where the distance between ε and ε is less than ε is called the ε-neighborhood; the number of data points in this neighborhood is recorded as

[0032] 2) Core point: For any data point If its ε-neighborhood corresponds to Contains at least m data points, that is but It is the core point.

[0033] 3) Density direct access: If lie in In the ε-neighborhood of is the core point, then Depend on Density direct.

[0034] 4) Density can be reached: If there is a point chain in Depend on Density direct, α=1,2,...,n, then from the core point To data point It is density-attainable.

[0035] 5) Density connection: If there are data points To data point If the density is reachable, then

[0036] They are density connected.

[0037] The load data set of pure load users in the same area after normalization Perform DBSCAN clustering (clustering effect diagram as shown in Figure 2 As shown), extract the typical user load curve features: For the data set Apply ε-neighborhood to traverse all elements, and group the data points with mutually accessible density into one cluster to obtain the clustering result. and the corresponding cluster center Where N C The total number of clusters obtained by the DBSCAN clustering algorithm; the remaining elements are outliers and are not included in any cluster.

[0038] Preferably, the step S4 establishes a post-meter photovoltaic user load fitting model based on the user load characteristic curve, establishes a post-meter photovoltaic user photovoltaic output fitting model based on the photovoltaic output correlation, and constructs a photovoltaic output correlation coefficient solution model based on this method, specifically:

[0039] Based on the normalized reference PV site dataset in the same region Establish the historical photovoltaic output fitting model of the photovoltaic users behind the meter:

[0040]

[0041] Among them, N BTM is the total number of photovoltaic users after the table; The photovoltaic output fitting of photovoltaic user i on the tth day after the table; is the historical measured value of photovoltaic output of reference photovoltaic site j on day t; r ij for and The correlation coefficient is independent of t; N PV is the total number of reference PV sites.

[0042] Based on the normalized load data set of pure load users in the same area Cluster center of clustering results Establish the historical load fitting model of photovoltaic users behind the meter:

[0043]

[0044] in, is the load fitting value of PV user i on day t after the table; is the load clustering center; w ij,t for and The correlation coefficient of .

[0045] For off-meter photovoltaic users, the following relationship is satisfied between their photovoltaic output, load and net load:

[0046]

[0047] in, is the net load of PV user i on day t after the meter; is the load of user i on day t; is the photovoltaic output of user i on day t.

[0048] Based on the historical photovoltaic output fitting model of the photovoltaic users behind the table, the historical load fitting model and the relationship between photovoltaic output, load and net load, the following model for solving the correlation coefficient between photovoltaic and reference photovoltaic sites is constructed:

[0049]

[0050]

[0051] r ij ≥0,

[0052]

[0053]

[0054] w ij,t ≥0,

[0055] i=1,2,...,N BTM

[0056] Since photovoltaic output is strongly correlated with irradiance, and the irradiance in the same region is approximately the same, the normalized photovoltaic output in the same region is approximately equal, with only a small difference caused by factors such as cloud cover. That is, on the premise of ensuring the correct overall trend of the photovoltaic output behind the meter, the output curve characteristics of different reference photovoltaic sites can be used to estimate and characterize the details of the photovoltaic output behind the meter.

[0057] Preferably, the step S5 is based on the obtained photovoltaic output correlation coefficient, combined with the real-time measurement data of the net load of the photovoltaic users behind the meter, to perform a method for real-time decomposition of the photovoltaic output behind the meter, specifically:

[0058] Solve to get the correlation coefficient r between the photovoltaic and reference photovoltaic sites in the table ij Finally, based on the photovoltaic output prediction model of photovoltaic users in the following table, combined with the real-time measurement values ​​of reference photovoltaic sites in the same region The photovoltaic output of the photovoltaic users behind the meter can be predicted in real time.

[0059]

[0060] in, is the real-time forecast value of photovoltaic output of photovoltaic user i behind the meter; is the real-time measurement value of the photovoltaic output of the reference photovoltaic site j.

[0061] Combined with the real-time measurement value of the normalized net load of photovoltaic users behind the meter The output of photovoltaic users behind the meter can be decomposed:

[0062]

[0063] in, is the real-time load prediction value of photovoltaic user i behind the meter; The real-time measured value of the normalized net load of photovoltaic user i after the meter. The decomposition result of the real-time photovoltaic output after the meter is as follows: Figure 4 shown.

[0064] The normalized PV output obtained by decomposing the real-time output is multiplied by the PV capacity to obtain the original PV output; adding it to the net load gives the original load:

[0065]

[0066]

[0067]

[0068] i=1,2,...,N BTM ,h=1,2,...,96

[0069] in, is the original photovoltaic output of user i obtained by decomposition; is the original load of user i obtained by decomposition; is the real-time measurement value of the original net load of user i; and The maximum and minimum values ​​of the historical measurements used for normalization.

[0070] Based on the above technical scheme, the present invention has the following beneficial effects: the photovoltaic decomposition method of the post-meter system based on the correlation of photovoltaic output provided by the present invention, based on the correlation of photovoltaic output in the same region, can rely on limited measurement information to realize the real-time decomposition of the photovoltaic output behind the meter, without the need for additional information such as meteorological data and geographic information data; the established model has low complexity, effectively improves the solution efficiency, and ensures the real-time decomposition of the photovoltaic output behind the meter; by realizing the real-time decomposition of the photovoltaic output behind the meter, the observability of the photovoltaic users behind the meter to the front side of the meter can be enhanced, which is helpful for the front side of the meter to formulate the optimal scheduling strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a schematic flow diagram of the method of the present invention;

[0072] Figure 2 It is a schematic diagram of the upper and lower envelopes of the net load of photovoltaic users behind the meter in a preferred embodiment of the present invention;

[0073] Figure 3 It is a historical data decomposition result diagram based on a correlation coefficient solution model in a preferred embodiment of the present invention;

[0074] Figure 4 It is a real-time data decomposition result diagram based on a correlation coefficient solution model in a preferred embodiment of the present invention; DETAILED DESCRIPTION

[0075] The following is a detailed description of the embodiments of the present invention: This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and a specific operation process are given. It should be pointed out that for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

[0076] The embodiment of the present invention provides a photovoltaic decomposition method of a post-meter system based on photovoltaic output correlation, which involves the field of decoupling and decomposition of distribution network front-meter measurement data, pre-processes historical measurement data based on min-max normalization; defines the upper and lower envelopes of the user's net load data, and estimates the photovoltaic installed capacity of the post-meter based on this; extracts features of typical user load curves based on DBSCAN clustering; establishes a post-meter photovoltaic user load fitting model based on the user load characteristic curve, establishes a photovoltaic output fitting model of the post-meter photovoltaic user based on the photovoltaic output correlation, and constructs a photovoltaic output correlation coefficient solution model on this basis; based on the photovoltaic output correlation coefficient obtained by solving, combined with the net load real-time measurement data of the post-meter photovoltaic user, the real-time decomposition of the post-meter photovoltaic output is performed. Starting from the correlation of photovoltaic output in the same region, the present invention realizes the real-time decomposition of the post-meter photovoltaic output based on limited measurement information, which can enhance the observability of the post-meter photovoltaic user to the front side of the meter, and help the front side to formulate the optimal scheduling strategy. The method has practical theoretical significance and promotion value for the decoupling and decomposition of the front-meter measurement data in the actual distribution network operation.

[0077] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0078] Please read first Figure 1 , Figure 1 1 is a flow chart of a photovoltaic decomposition method for a post-meter system based on photovoltaic output correlation provided by an embodiment of the present invention. As shown in the figure, the photovoltaic decomposition method for a post-meter system based on photovoltaic output correlation implemented by the present invention includes the following steps:

[0079] Step 1), preprocess the historical measurement data based on min-max normalization.

[0080] As a preferred embodiment, the method of step 1) is:

[0081] The historical measurement data that can be obtained by the distribution network side based on the smart meter include: the historical measurement value of the net load of the photovoltaic users to be decomposed, the historical measurement value of the load of the pure load users in the same area, and the historical measurement value of the photovoltaic output of the reference photovoltaic measurement site. The sampling interval of the above historical measurement data is 15 minutes, corresponding to 96 points a day. The historical measurement data is normalized for subsequent clustering and output decomposition.

[0082] The historical measurement data uses the following min-max normalization formula:

[0083]

[0084] in, is the normalized historical measurement data of measurement point i; is the original historical measurement data of measurement point i; are the maximum and minimum values ​​in the original historical measurement data set of measurement point i, respectively; h is the sequence number of 96 data points in one day, and x(h) represents the hth data point in a set of data x.

[0085] Step 2), define the upper and lower envelopes of the user's net load data, and on this basis, estimate the post-meter photovoltaic installed capacity.

[0086] As a preferred embodiment, the method of step 2) is:

[0087] The net load curve of users with photovoltaic power behind the meter will show great differences under different lighting conditions. This feature can be used to estimate the installed capacity of photovoltaic power behind the meter.

[0088] Based on the normalized photovoltaic output data of the reference photovoltaic site, the photovoltaic output is screened to be in the range [l lower ,l upper ] in the date tag collection T pv,low , where 0≤l lower ≤l upper ≤0.2. Define the photovoltaic user i (i=1,2,...,N BTM ) on date T pv,low The minimum value of all net load curves is the upper envelope:

[0089]

[0090] in, is the upper envelope of PV user i after decomposition; is the original net load data of user i on day t.

[0091] After defining the table to be decomposed, the photovoltaic user i (i = 1, 2, ..., N BTM ) is the minimum value of all net load curves of the lower envelope:

[0092]

[0093] in, is the lower envelope of PV user i after decomposition; T is the total number of days in the historical data set.

[0094] The schematic diagram of the upper and lower envelopes of the net load of photovoltaic users behind the meter is as follows Figure 1 Based on the upper and lower envelopes, the estimated installed capacity of photovoltaic power can be obtained:

[0095]

[0096] Among them, C i is the estimated value of the photovoltaic installed capacity of photovoltaic user i after decomposition.

[0097] Step 3), based on DBSCAN clustering, feature extraction is performed on typical user load curves.

[0098] As a preferred embodiment, the method of step 3) is:

[0099] For the normalized dataset X norm , where the total number of elements is k, the DBSCAN clustering algorithm has the following definition:

[0100] 1)ε-neighborhood: distance from data point The area where the distance between ε and ε is less than ε is called the ε-neighborhood; the number of data points in this neighborhood is recorded as

[0101] 2) Core point: For any data point If its ε-neighborhood corresponds to Contains at least m data points, that is but It is the core point.

[0102] 3) Density direct access: If lie in In the ε-neighborhood of is the core point, then Depend on Density direct.

[0103] 4) Density can be reached: If there is a point chain in Depend on Density direct, α=1,2,...,n, then from the core point To data point It is density-attainable.

[0104] 5) Density connection: If there are data points To data point If the density is reachable, then They are density connected.

[0105] The load data set of pure load users in the same area after normalization Perform DBSCAN clustering (clustering effect diagram as shown in Figure 2 As shown), extract the typical user load curve features: For the data set Apply ε-neighborhood to traverse all elements, and group the data points with mutually accessible density into one cluster to obtain the clustering result. and the corresponding cluster center Where N C The total number of clusters obtained by the DBSCAN clustering algorithm; the remaining elements are outliers and are not included in any cluster.

[0106] Step 4) A post-meter photovoltaic user load fitting model is established based on the user load characteristic curve, a post-meter photovoltaic user photovoltaic output fitting model is established based on the photovoltaic output correlation, and a photovoltaic output correlation coefficient solution model is constructed on this basis.

[0107] As a preferred embodiment, the method of step 4) is:

[0108] Based on the normalized reference PV site dataset in the same region Establish the historical photovoltaic output fitting model of the photovoltaic users behind the meter:

[0109]

[0110] Among them, N BTM is the total number of photovoltaic users after the table; The photovoltaic output fitting of photovoltaic user i on the tth day after the table; is the historical measured value of photovoltaic output of reference photovoltaic site j on day t; r ij for and The correlation coefficient is independent of t; N PV is the total number of reference PV sites.

[0111] Based on the normalized load data set of pure load users in the same area Cluster center of clustering results Establish the historical load fitting model of photovoltaic users behind the meter:

[0112]

[0113] in, is the load fitting value of PV user i on day t after the table; is the load clustering center; w ij,t for and The correlation coefficient of .

[0114] For off-meter photovoltaic users, the following relationship is satisfied between their photovoltaic output, load and net load:

[0115]

[0116] in, is the net load of PV user i on day t after the meter; is the load of user i on day t; is the photovoltaic output of user i on day t.

[0117] Based on the historical photovoltaic output fitting model of the photovoltaic users behind the table, the historical load fitting model and the relationship between photovoltaic output, load and net load, the following model for solving the correlation coefficient between photovoltaic and reference photovoltaic sites is constructed:

[0118]

[0119]

[0120] r ij ≥0,

[0121]

[0122]

[0123] w ij,t ≥0,

[0124] i=1,2,...,N BTM

[0125] Since photovoltaic output is strongly correlated with irradiance, and the irradiance in the same region is approximately the same, the normalized photovoltaic output in the same region is approximately equal, with only a small difference caused by factors such as cloud cover. That is, on the premise of ensuring the correct overall trend of the photovoltaic output behind the meter, the output curve characteristics of different reference photovoltaic sites can be used to estimate and characterize the details of the photovoltaic output behind the meter.

[0126] Step 5), based on the obtained photovoltaic output correlation coefficient, combined with the real-time measurement data of the net load of the photovoltaic users behind the meter, the real-time decomposition of the photovoltaic output behind the meter is performed.

[0127] As a preferred embodiment, the method of step 5) is:

[0128] Solve to get the correlation coefficient r between the photovoltaic and reference photovoltaic sites in the table ij Finally, based on the photovoltaic output prediction model of photovoltaic users in the following table, combined with the real-time measurement values ​​of reference photovoltaic sites in the same region The photovoltaic output of the photovoltaic users behind the meter can be predicted in real time.

[0129]

[0130] in, is the real-time forecast value of photovoltaic output of photovoltaic user i behind the meter; is the real-time measurement value of the photovoltaic output of the reference photovoltaic site j.

[0131] Combined with the real-time measurement value of the normalized net load of photovoltaic users behind the meter The output of photovoltaic users behind the meter can be decomposed:

[0132]

[0133] in, is the real-time load prediction value of photovoltaic user i behind the meter; The real-time measured value of the normalized net load of photovoltaic user i after the meter. The decomposition result of the real-time photovoltaic output after the meter is as follows: Figure 4 shown.

[0134] The normalized PV output obtained by decomposing the real-time output is multiplied by the PV capacity to obtain the original PV output; adding it to the net load gives the original load:

[0135]

[0136]

[0137]

[0138] i=1,2,...,N BTM ,h=1,2,...,96

[0139] in, is the original photovoltaic output of user i obtained by decomposition; is the original load of user i obtained by decomposition; is the real-time measurement value of the original net load of user i; and The maximum and minimum values ​​of the historical measurements used for normalization.

[0140] Figure 2 It is a schematic diagram of the upper and lower envelopes of the net load of the photovoltaic users behind the meter, showing the correlation between the upper and lower envelopes and the net load curve of the photovoltaic users behind the meter; Figure 3 It is a historical data decomposition result diagram based on the correlation coefficient solution model in the embodiment of the present invention, which shows the fitting effect of the photovoltaic output, load and net load of the photovoltaic users behind the meter when the photovoltaic correlation coefficient is solved by using historical data; Figure 4 This is a real-time data decomposition result diagram based on the correlation coefficient solution model, which shows the real-time decomposition effect of the net load of photovoltaic users behind the meter.

[0141] The photovoltaic decomposition method of the post-meter system based on photovoltaic output correlation provided by the above-mentioned embodiment of the present invention pre-processes the historical measurement data based on min-max normalization; defines the upper and lower envelopes of the user's net load data, and estimates the post-meter photovoltaic installed capacity on this basis; extracts features of the typical user load curve based on DBSCAN clustering; establishes a post-meter photovoltaic user load fitting model based on the user load characteristic curve, establishes a photovoltaic output fitting model of the post-meter photovoltaic user based on photovoltaic output correlation, and constructs a photovoltaic output correlation coefficient solution model on this basis; based on the photovoltaic output correlation coefficient obtained by solution, combined with the real-time measurement data of the net load of the post-meter photovoltaic user, performs real-time decomposition of the post-meter photovoltaic output.

[0142] Compared with the traditional post-meter photovoltaic output decomposition method, it can rely on limited measurement information to achieve real-time decomposition of post-meter photovoltaic output without the need for additional information such as meteorological data and geographic information data; the established model has low complexity, which effectively improves the solution efficiency and ensures the real-time decomposition of post-meter photovoltaic output; by realizing real-time decomposition of post-meter photovoltaic output, it can enhance the visibility of post-meter photovoltaic users on the front side of the meter, which is helpful for the front side of the meter to formulate the optimal scheduling strategy.

[0143] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A photovoltaic decomposition method for a post-meter system based on photovoltaic output correlation, characterized in that: The steps include: S1. Preprocess the historical measurement data based on min-max normalization; S2. Define the upper and lower envelopes of the user's net load data, and estimate the photovoltaic installed capacity after the meter on this basis; S3, based on DBSCAN clustering, feature extraction of typical user load curves; S4. Establish a post-meter photovoltaic user load fitting model based on the user load characteristic curve, establish a post-meter photovoltaic user photovoltaic output fitting model based on the photovoltaic output correlation, and build a photovoltaic output correlation coefficient solution model on this basis; S5. Based on the obtained photovoltaic output correlation coefficient, combined with the real-time measurement data of the net load of photovoltaic users behind the meter, the real-time decomposition of photovoltaic output behind the meter is performed; The step S4 is specifically as follows: Based on the normalized reference PV site dataset in the same region Establish the historical photovoltaic output fitting model of the photovoltaic users behind the meter: Among them, N BTM is the total number of photovoltaic users after the table; The photovoltaic output fitting of photovoltaic user i on the tth day after the table; is the historical measured value of photovoltaic output of reference photovoltaic site j on day t; r ij for and The correlation coefficient is independent of t; N PV is the total number of reference PV sites; Based on the normalized load data set of pure load users in the same area Cluster center of clustering results Establish the historical load fitting model of photovoltaic users behind the meter: in, is the load fitting value of PV user i on day t after the table; is the load clustering center; w ij,t for and The correlation coefficient of For off-meter photovoltaic users, the following relationship is satisfied between their photovoltaic output, load and net load: in, is the net load of PV user i on day t after the meter; is the load of user i on day t; The photovoltaic output of user i on day t; Based on the historical photovoltaic output fitting model of the photovoltaic users behind the table, the historical load fitting model and the relationship between photovoltaic output, load and net load, the following model for solving the correlation coefficient between photovoltaic and reference photovoltaic sites is constructed: Utilizing constraints On the premise that the overall trend of post-meter photovoltaic output is correct, the output curve characteristics of different reference photovoltaic sites are used to estimate and characterize the details of post-meter photovoltaic output.

2. The photovoltaic decomposition method of the post-meter system based on photovoltaic output correlation according to claim 1 is characterized in that: The step S1 is specifically as follows: The historical measurement data that can be obtained by the smart meter on the distribution network side include: the historical measurement value of the net load of the photovoltaic users after the decomposition table, the historical measurement value of the load of the pure load users in the same area, and the historical measurement value of the photovoltaic output of the reference photovoltaic measurement site. The sampling interval of the above historical measurement data is 15 minutes, corresponding to 96 points a day, and the historical measurement data is normalized; The historical measurement data uses the following min-max normalization formula: in, is the normalized historical measurement data of measurement point i; is the original historical measurement data of measurement point i; are the maximum and minimum values ​​in the original historical measurement data set of measurement point i, respectively; h is the sequence number of 96 data points in one day, and x(h) represents the hth data point in a set of data x.

3. The photovoltaic decomposition method of the post-meter system based on photovoltaic output correlation according to claim 1, characterized in that: The step S2 is specifically as follows: Based on the normalized photovoltaic output data of the reference photovoltaic site, the photovoltaic output is screened to be in the range [l lower ,l upper ] in the date tag collection T pv,low , where 0≤l lower ≤l upper ≤0.2, define the photovoltaic user i (i=1,2,...,N BTM ) on date T pv,low The minimum value of all net load curves is the upper envelope: in, is the upper envelope of PV user i after decomposition; is the original net load data of user i on day t; After defining the table to be decomposed, the photovoltaic user i (i = 1, 2, K, N BTM ) is the minimum value of all net load curves of the lower envelope: in, is the lower envelope of PV user i after decomposition; T is the total number of days in the historical data set; Based on the upper and lower envelopes, the estimated installed capacity of photovoltaic power after the table is obtained: Among them, C i is the estimated value of the photovoltaic installed capacity of photovoltaic user i after decomposition.

4. The photovoltaic decomposition method of the post-meter system based on photovoltaic output correlation according to claim 1, characterized in that: The step S3 is specifically as follows: For the normalized dataset X norm , where the total number of elements is k, the DBSCAN clustering algorithm has the following definition: 1)ε-neighborhood: distance from data point The area where the distance between ε and ε is less than ε is called the ε-neighborhood; the number of data points in this neighborhood is recorded as 2) Core point: For any data point If its ε-neighborhood corresponds to Contains at least m data points, that is but It is the core point; 3) Density direct access: If lie in In the ε-neighborhood of is the core point, then Depend on Density direct access; 4) Density can be reached: If there is a point chain in Depend on Density direct, α=1,2,K,n, then from the core point To data point is density-attainable; 5) Density connection: If there are data points To data point If the density is reachable, then They are connected by density; The load data set of pure load users in the same area after normalization Perform DBSCAN clustering to extract typical user load curve features: For the data set Apply ε-neighborhood to traverse all elements, and group the data points with mutually accessible density into one cluster to obtain the clustering result. and the corresponding cluster center Where N C The total number of clusters obtained by the DBSCAN clustering algorithm; the remaining elements are outliers and are not included in any cluster.

5. The photovoltaic decomposition method of the post-meter system based on photovoltaic output correlation according to claim 1, characterized in that: The step S5 is specifically as follows: Solve to get the correlation coefficient r between the photovoltaic and reference photovoltaic sites in the table ij Finally, the photovoltaic output prediction model for photovoltaic users is based on the following table: Combined with the real-time measurement values ​​of reference photovoltaic sites in the same region The photovoltaic output of the photovoltaic users behind the meter can be predicted in real time; in, is the real-time forecast value of photovoltaic output of photovoltaic user i behind the meter; is the real-time measurement value of the photovoltaic output of the reference photovoltaic site j; Combined with the real-time measurement value of the normalized net load of photovoltaic users behind the meter The output of photovoltaic users behind the meter can be decomposed: in, is the real-time load prediction value of photovoltaic user i behind the meter; The real-time measured value of the normalized net load of photovoltaic user i behind the meter; The normalized PV output obtained by decomposing the real-time output is multiplied by the PV capacity to obtain the original PV output; adding it to the net load gives the original load: in, is the original photovoltaic output of user i obtained by decomposition; is the original load of user i obtained by decomposition; is the real-time measurement value of the original net load of user i; and The maximum and minimum values ​​of the historical measurements used for normalization.

Images