A main sub-metering user multi-grid point photovoltaic accurate metering method

Abstract

The application discloses a kind of main sub-metering user multiple grid point photovoltaic accurate metering method, comprising the following steps: step 1, obtain the impedance value at photovoltaic power generation metering point, obtain H class impedance sample set;Step 2, obtain sparse impedance value using K-SVD sparse decomposition algorithm, construct sparse matrix;Step 3, update discriminant dictionary based on the sparse matrix;Step 4, using Autoformer model encoder, the sparse matrix is decomposed into trend item and periodic term;Step 5, using Autoformer model, impedance value is decomposed in real time, obtains the real-time impedance of commercial power, photovoltaic impedance, sub-metering impedance and other partial impedance under different frequency;Step 6, the impedance value of each part under power frequency is obtained, and the accurate metering of main sub-metering user multiple grid point distributed photovoltaic is realized.The application can not increase additional hardware, and the physical quantity is obtained using the carrier signal of electric energy meter to realize accurate metering.

Description

Technical Field

[0001] This invention belongs to the field of smart grid metering, specifically relating to a method for accurate metering of photovoltaic power at multiple grid connection points for main and branch metering users. Background Technology

[0002] The construction of new power systems is driving large-scale grid connection of distributed photovoltaic (PV) power. Based on the metering method and the number of grid connection points, distributed PV can be categorized into single-metering, main-and-distribution metering with a single grid connection point, and main-and-distribution metering with multiple grid connection points. The distributed PV grid connection situation of Suzhou power supply companies in 2023 is as follows: Figure 1 As shown.

[0003] Depend on Figure 1 It can be seen that, in terms of both quantity and capacity, multi-grid connection points with main and branch metering account for the majority. This is mainly because the capacity limit for a single grid connection point is 400kW, and most photovoltaic users' grid connection capacity exceeds this limit. Therefore, multi-grid connection points are adopted. A typical wiring diagram of a distributed photovoltaic primary system with multiple grid connection points for main and branch metering users is shown below. Figure 2 As shown.

[0004] For the common scenario of distributed photovoltaic systems with multiple grid-connected users and main / sub-metering, the problem of accurate metering has remained unsolved. For example... Figure 2 As shown, there are four metering points: main metering, sub-metering, PV metering 1, and PV metering 2. These four meters can measure forward and reverse active and reactive power data. However, the problem lies in the uncertainty surrounding whether the electricity generated by PV power station 2 is transmitted to the public grid, consumed by sub-metered loads, or consumed by other loads. This leads to incorrect electricity billing if the main metering price and sub-metering price are inconsistent, hindering accurate metering. The metering problem of distributed photovoltaic systems with multiple grid-connected points for main and sub-metered users has existed for many years but remains unresolved, and no relevant literature has addressed this issue. With the increasing penetration rate of distributed photovoltaic systems, the accurate metering problem of distributed photovoltaic systems with multiple grid-connected points for main and sub-metered users urgently needs to be solved.

[0005] For most 10 kV and above high-voltage users, due to the diverse nature of their internal electrical loads, power companies set up main and sub-metering points to achieve accurate metering of different load types. The most common and typical application of main and sub-metering is for large industrial users. Their main metering applies the large industrial electricity price, while sub-metering typically applies other lighting electricity prices, resulting in a price difference between the two. In scenarios without distributed power sources, electricity flows in a single phase, and the electrical energy consumed by each component is accurately measured. Multiplying this by the corresponding electricity price yields the monthly electricity bill.

[0006] Large industrial users' factory rooftops are currently the most common application scenario for distributed photovoltaic (PV) systems. As the cost of PV modules continues to decrease, the penetration rate of distributed PV is increasing. When electricity users apply for distributed PV, the power company will install an electricity meter at each grid connection point to measure PV power generation. Simultaneously, it will also measure the reverse electricity from the user's main meter, i.e., the PV power fed into the grid. For low-voltage grid-connected PV users, the maximum capacity at each grid connection point is 400kW.

[0007] Under the current distributed photovoltaic (PV) metering configuration, the amount of electricity fed into the grid and the PV power generation at each grid connection point can be accurately measured. However, for distributed PV users with multiple grid connection points, the flow of electricity generated by their PV power generation is not accurately measured. This can lead to metering discrepancies when the user is the primary metered user. Figure 2 The power transmission diagram corresponding to the wiring diagram is as follows: Figure 3 As shown.

[0008] like Figure 3 As shown, the arrows indicate the direction of energy transmission, the black lines represent the measurable portion, and the black dashed lines represent the non-measurable portion. Although the total amount of energy transmitted from the common point to the sub-metering is measurable, it is transmitted from both the main metering direction and the photovoltaic grid connection point direction. The amount of energy transmitted to the sub-metering in each direction is unknown. The user's electricity bill can be calculated as follows:

[0009] C = (W m -W s )*p m +W s *p s

[0010] In the formula: C is the monthly electricity cost; W m Total electricity consumption; W s For the unit of electricity consumption; p m Main metered electricity price; p s This refers to the metered electricity price. Peak-valley time-of-use pricing is not considered here. When the main metering and metered electricity prices are inconsistent, multiple grid-connected distributed photovoltaic users with both main and metered pricing will cause electricity billing errors. Summary of the Invention

[0011] The purpose of this invention is to provide a method for accurate metering of photovoltaic power at multiple grid-connected points for main and branch metering users, which can achieve accurate metering by acquiring physical quantities using the carrier signal built into the electricity meter without adding extra hardware.

[0012] To achieve the above objectives, the solution of the present invention is:

[0013] A method for accurate metering of photovoltaic power at multiple grid-connected points for main and branch metering users includes the following steps:

[0014] Step 1: Obtain the impedance value at the photovoltaic power generation metering point to obtain the H-type impedance sample set Y = [Y1, Y2, ..., Y]. H ];

[0015] Step 2: Use the K-SVD sparse decomposition algorithm to obtain sparse impedance values ​​and construct a sparse matrix;

[0016] Step 3: Update the discrimination dictionary D based on the sparse matrix;

[0017] Step 4: Use the Autoformer model encoder to decompose the sparse matrix into trend and periodic terms;

[0018] Step 5: Use the Autoformer model to decompose the impedance value in real time to obtain the real-time mains impedance Z at different frequencies. m Photovoltaic impedance Z n , Metered impedance Z s Other part of the impedance Z o ;

[0019] Step 6: Calculate the impedance values ​​of each part at the power frequency to achieve accurate photovoltaic metering for multiple grid-connected users.

[0020] In step 1 above, a sweep frequency signal is transmitted from the photovoltaic power generation metering point to obtain the voltage and current data corresponding to the frequency of the sweep frequency signal, and the impedance value at the photovoltaic power generation metering point is calculated.

[0021] In step 2 above, the H-type impedance sample set Y = [Y1, Y2, ..., Y] obtained in step 1 is used... H Using these as training samples, a sub-dictionary D for each pattern class is obtained through training using the K-SVD algorithm. i Let i = 1, 2, ..., H. Concatenating sub-dictionaries yields a discriminant dictionary D = [D1, D2, ..., D]. H The objective equation of the K-SVD algorithm is expressed as follows:

[0022]

[0023] Where X is a sparse matrix; x i For the i-th sub-dictionary D i The corresponding sparsity coefficients; T0 is the sparsity threshold, F is the F-norm of the matrix; then the test sample y i The reconstruction residuals between the various sub-dictionaries are,

[0024] e i =||y i -D i x i ||2

[0025] The K-SVD algorithm can be expressed using a sparse model as follows:

[0026]

[0027] This yields the sparse impedance value x. i This forms a sparse matrix X.

[0028] In step 3 above, the i-th column d of the discriminant dictionary D is based on the sparse matrix X. i Update

[0029]

[0030] in, Let d be the i-th column of the sparse matrix X. i The corresponding i-th row;

[0031] Let E i To define the error matrix of the discrimination dictionary D for I signal samples in Y after removing the k-th atom from D, we define: ω k This indicates that no atom d was used. i Signal number; Define matrix Ω k The corrected matrix error is Transform the above expression into:

[0032]

[0033] right Perform singular value decomposition, then make If d is the first column of U, then i The updated value is Simultaneously, the product of the first column of V and Δ(1,1) is updated. And update each column using the new dictionary. The sparsity coefficients are decomposed until the stopping condition is met.

[0034] In step 4 above, the sparse matrix X is decomposed into two parts: a trend term and a periodic term.

[0035] X trend =AvgPool(Padding(X))

[0036] X s =XX trend

[0037] Among them, X trend Indicates the trend term, AvgPool represents the average pooling layer, and Padding represents padding; X s Indicates a periodic term.

[0038] In step 5 above, the Autoformer model calculates the autocorrelation coefficient of the sparse matrix according to the following formula:

[0039]

[0040] Among them, X t The initial timing sequence is the input; F represents the Fourier transform, F * It is a conjugate operation; f represents frequency, multiplied by 2π to obtain the angular frequency result; F and F * Multiply the results obtained by integrating the trend term separately, and convert the time series to the frequency domain S. XX .

[0041] In step 6 above, the real-time current value I of the photovoltaic power generation is measured. s for,

[0042]

[0043] Among them, I n This represents the real-time current value of photovoltaic power generation.

[0044] By adopting the above scheme, this invention utilizes the carrier signal built into the electricity meter to acquire physical quantities and achieve accurate metering. The electricity meter's carrier module transmits data via power lines through a high-frequency, small-amplitude carrier wave, enabling data communication. Similarly, input impedance frequency sweeping can be achieved by measuring the voltage-to-current ratio of the carrier wave signal at different frequencies. Based on the power output of the power system, the power distribution of each power source is related to the impedance parameters within the system. By obtaining the real-time impedance through the carrier signal and combining it with relevant algorithms, the specific distribution of impedance within the photovoltaic system can be obtained. Thus, the flow of electricity generated by the photovoltaic grid-connected point under non-distributed metering can be determined, achieving accurate metering. However, it is necessary to consider that since the electricity meter and the back-end master station use a wireless power remote meter reading system (GPRS) for transmission, the data transmission bandwidth is relatively small. Therefore, it is necessary to limit the amount of data for input impedance frequency sweeping to facilitate subsequent transmission and calculation.

[0045] Compared with the prior art, the beneficial effects of the present invention are reflected in:

[0046] 1) A precise metering method for photovoltaic (PV) systems with multiple grid-connected points and multiple metering users is proposed. The metering problem of PV systems with multiple grid-connected points and multiple metering users has existed for many years. Despite the increasing penetration rate of distributed PV, it has not been solved and no relevant literature has studied this problem.

[0047] 2) Considering that the transmission direction of photovoltaic power generation at non-metering points is related to the real-time impedance of each branch, the impedance of each branch is obtained through frequency sweep response analysis, so as to realize the real-time accurate measurement of photovoltaic power generation transmission energy at non-metering points.

[0048] 3) A data processing method based on K-SVD sparse classifier is proposed, which effectively reduces the computational and storage requirements of the proposed method.

[0049] 4) A method for fitting impedance values ​​of various parts based on the Autoformer model is proposed. Compared with the traditional Transformer method, the relative error R of the proposed method is significantly reduced. E Root mean square error R MSE Both are relatively small. Attached Figure Description

[0050] Figure 1 This is the grid connection status of distributed photovoltaic power generation by Suzhou power supply companies in 2023;

[0051] Figure 2 This is a wiring diagram of a distributed photovoltaic primary system with multiple grid connection points for main and branch metering users;

[0052] Figure 3 This is a diagram of distributed photovoltaic energy transmission with multiple grid connection points for main and branch metering users;

[0053] Figure 4 It is a distributed photovoltaic impedance model with multiple grid connection points for main and sub-metering users;

[0054] Figure 5 This is a structural diagram of the Autoformer model used in this invention;

[0055] Figure 6 This is the autocorrelation module in this invention;

[0056] Figure 7 This is a flowchart of the present invention;

[0057] Figure 8 This is the predicted result of photovoltaic power transmission in commercial entities in this invention;

[0058] Figure 9 This is the prediction result of photovoltaic power transmission in the electronics industry in this invention;

[0059] Figure 10 This is the prediction result of photovoltaic power transmission in the chemical industry in this invention;

[0060] Figure 11 This is the predicted result of photovoltaic power transmission in the industrial park in this invention;

[0061] Figure 12 This is the predicted result of commercial photovoltaic power transmission in this invention;

[0062] Figure 13 This is the prediction result of photovoltaic power transmission in the electronics industry in this invention;

[0063] Figure 14This is the prediction result of photovoltaic power transmission in the chemical industry in this invention;

[0064] Figure 15 This is the predicted result of photovoltaic power transmission in the industrial park in this invention. Detailed Implementation

[0065] The technical solution and beneficial effects of the present invention will be described in detail below with reference to the accompanying drawings.

[0066] To address the issue of accurate metering for distributed photovoltaic (PV) systems with multiple grid-connected points and multiple metered users, this invention provides a method for accurate metering of PV systems with multiple grid-connected points and multiple metered users based on sparse frequency sweep response analysis. The method involves obtaining the user-side wiring diagram during the PV project expansion application stage, acquiring the spectrum diagram through frequency sweep response analysis, and utilizing a K-SVD sparse classifier to reduce transmission and computational complexity. An Autoformer model is introduced to calculate the impedance values ​​of each component at the mains frequency, and a relative error R is selected. E Root mean square error R MSE The flowchart for evaluating the prediction effect is as follows: Figure 7 As shown, the specific steps include the following:

[0067] Step 1: Using the electricity meter installed at the photovoltaic power generation metering point, emit a sinusoidal signal with an amplitude of 5V, a frequency sweep range of 1kHz-1MHz, and a frequency interval of 1000Hz (a total of 100 sweep sampling data points), obtain the voltage and current data at the corresponding frequencies, and divide the two to obtain the impedance value at the corresponding photovoltaic power generation metering point.

[0068]

[0069] This provides a real-time sweep impedance, laying the foundation for subsequent calculations of the electricity transmitted from photovoltaic power generation to sub-metering.

[0070] Step 2: Considering that existing smart meters transmit input impedance data to the backend master station via the wireless power remote meter reading system (GPRS), which makes large-scale data storage and transmission difficult, a K-SVD sparse classifier is introduced based on frequency sweep response analysis. The K-SVD sparse decomposition algorithm can adaptively update dictionary atoms according to a given dataset to find a globally optimal dictionary of a custom size.

[0071] Assume the impedance matrix obtained in step 1 is the H-class impedance sample set Y = [Y1, Y2, ..., Y]. H Using a set of training samples, the K-SVD algorithm is used to train and derive a sub-dictionary D for each pattern class. i (i = 1, 2, ..., H) concatenate sub-dictionaries to obtain a discriminant dictionary D = [D1, D2, ..., D...]. H The objective equation of the K-SVD algorithm is expressed as:

[0072]

[0073] In the formula: X is the sparsity coefficient x i The sparse matrix formed; x i = (i = 1, 2, ..., H) is the sub-dictionary D of the i-th class. i The corresponding sparsity coefficients; T0 is the sparsity threshold, and F is the F-norm of the matrix. Then the test sample y... i The reconstruction residuals between the various sub-dictionaries are:

[0074] e i =||y i -D i x i ||2 (3)

[0075] The K-SVD algorithm, expressed using a sparse model, is as follows:

[0076]

[0077] This yields impedance data x that is sparse, satisfies transmission constraints, and retains as much characteristic information as possible. i .

[0078] Step 3, Dictionary D Update: Using the sparse matrix X obtained through sparse coding, update the i-th column d of the dictionary. i To update, let d in the sparse coefficient matrix X... i The corresponding i-th row Then the following relationship is satisfied:

[0079]

[0080] In the formula above, DX will be decomposed into the sum of I matrices of rank 1. The I-1 terms are known, so the remaining term is what needs to be solved. E i This represents the error matrix of dictionary D for the I signal samples in Y after removing the k-th atom from dictionary D. Equation (5) is divided into two parts: one part is the error matrix term E i The other part is a matrix with rank 1. We need to find d i and From the residual matrix term E i The best fit for E i Singular value decomposition (SVD) can be performed, but there's one issue to consider: not all sample sets of signals can be decomposed using d... i This atom, then after removing d i After that, some signals were not received by d i The impact of this, therefore, requires E. iApply conditional restrictions. Define. ω k This indicates that no atom d was used. i The signal sequence number is then used to define a matrix Ω. k The corrected matrix error is same This was also corrected by removing unused vectors from the trained dictionary D, thus reducing the space of the dictionary and the number of vectors. All have been reduced, and equation (5) is transformed into the following formula:

[0081]

[0082] Equation (6) is obtained, and Perform singular value decomposition, then make If d is the first column of U, then i The updated value is Simultaneously, the product of the first column of V and Δ(1,1) is updated. After updating each column one by one, use the new dictionary. Perform sparse coefficient decomposition to determine whether the stopping condition has been met.

[0083] Step 4, for the sparsely treated impedance value x i (Sparse coefficients) require the use of an Autoformer model to predict and decompose the impedance values ​​of each component. Considering the regular changes in impedance with frequency, an Autoformer model is introduced to utilize an attention mechanism to ensure that important features play a greater role in the prediction of impedance values, thereby improving prediction accuracy. The structure of the Autoformer model is as follows... Figure 5 As shown, the basic architecture of the model is the same as the Transformer model, consisting of two main parts: an encoder and a decoder. However, the internal structure of the model has been modified. The encoder first replaces the attention mechanism of the original Transformer model with an autocorrelation mechanism, and then adds a time series decomposition unit after the autocorrelation module. The most significant feature of the decoder is that it has two inputs: the trend term and the periodic term of the time series are input separately. It uses the autocorrelation module and the time series decomposition unit to continuously decompose the trend component, which is then superimposed on the original trend term to achieve progressive decomposition.

[0084] The encoder will input the time series signal x i The terms are decomposed into two parts: a trend term and a period term, and the calculations are shown in equations (7) and (8).

[0085] X trend =AvgPool(Padding(X)) (7)

[0086] X s =XX trend (8)

[0087] In the formula: X trend The trend term is represented by AvgPool, which represents the average pooling layer, and Padding represents padding; both are mathematical formulas built into the PyTorch programming language. X s Indicates a periodic term.

[0088] Step 5: The most significant feature of the decoder is that it has two inputs. The trend term and period term of the time series are input separately. The autocorrelation module shown in Equations (9) and (10) and the time series decomposition unit shown in Equations (7) and (8) are used to continuously decompose the trend component and superimpose it with the original trend term to achieve progressive decomposition.

[0089] The autocorrelation mechanism is based on the Wiener-Khinchin theory, such as Figure 6 As shown, the autocorrelation coefficient of the sequence is calculated using the fast Fourier transform and inverse transform, and the calculation formulas are shown in equations (9) and (10).

[0090]

[0091] In the formula: X t The initial timing sequence is the input; F represents the Fourier transform, F * This is a conjugate operation; f represents frequency, multiplied by 2π to obtain the angular frequency result; F and F * Multiplying the results obtained by integrating the trend term separately can convert the time series into the frequency domain S. XX .

[0092]

[0093] In the formula: F -1 The autocorrelation coefficient is obtained by performing an inverse Fourier transform on the result obtained from equation (9). This method can reduce the computational complexity of solving the autocorrelation problem.

[0094] The Autoformer model yields the real-time impedance decomposition of each component, i.e., the real-time mains impedance Z at different frequencies. m Photovoltaic impedance Z n , Metered impedance Z s Other part of the impedance Z o .

[0095] The Autoformer model implements an autocorrelation mechanism for sequence-level connections. This mechanism breaks through the bottleneck of information utilization and decomposes the time series into more predictable components from complex time patterns, which is beneficial for mining the complex mapping relationship between features and remaining lifetime.

[0096] Step 6: Use the real-time mains impedance Z at different frequencies obtained in Step 5. m Photovoltaic impedance Z n , Metered impedance Z s Other part of the impedance Z o The impedance values ​​of each part under power frequency are solved, thereby calculating the real-time electrical energy transmitted from the photovoltaic metering point to the sub-metering.

[0097] According to the formula:

[0098] Z=R+jωL (11)

[0099] By solving for the impedance value at different frequencies, the impedance value at the power frequency is obtained, thereby realizing accurate metering of distributed photovoltaic systems with multiple grid connection points for main and sub-metering users based on sparse sweep frequency response analysis.

[0100] To address the issue of distributed photovoltaic metering with multiple grid connection points for main and branch metering users, such as Figure 3 As shown, the distributed photovoltaic energy transmission system with multiple grid connection points for main and sub-metered users can be simplified to a system with power from the mains (G) m and its impedance Z m Photovoltaic power generation G n and its impedance Z n , Metered impedance Z s Other part of the impedance Z o Parallel connection, such as Figure 4 As shown.

[0101] Assuming the mains frequency is constant, the mains power G m and its impedance Z m It is a constant value, but photovoltaic power generation G n and its impedance Z n As illumination conditions change in real time, and as load changes lead to variations in the metering impedance Z... s Other part of the impedance Z o It is also time-varying. To calculate the real-time power transmission, it can be transformed into solving for the real-time impedance. To solve for the impedance of each part at the real-time mains frequency, swept frequency response analysis (SFRA) is introduced. By sweeping the frequency response, the impedance values ​​of each part at different frequencies are solved, and thus the resistance and inductance values ​​of each part are obtained. The real-time current value I from photovoltaic power generation to metering is... s for:

[0102]

[0103] Among them, I n This represents the real-time current value of photovoltaic power generation.

[0104] By solving for the real-time current value of photovoltaic power generation to the sub-metering, the real-time active power value can be obtained, realizing accurate metering of distributed photovoltaic power generation at multiple grid connection points for main and sub-metering users.

[0105] To verify the effectiveness and scientific validity of the proposed accurate metering method for distributed photovoltaic systems with multiple grid connection points for main and sub-metered users based on sparse sweep frequency response analysis, four multi-metered photovoltaic users connected to the grid by Suzhou Power Supply Company in 2023 were taken as typical cases. The specific parameters and annual billing errors are shown in Table 1.

[0106] Table 1

[0107]

[0108] By comparing the proposed sparse frequency sweep response method with the traditional frequency sweep response method, the metrological accuracy is compared, and the effectiveness of different impedance fitting methods is discussed. Specific parameters are shown in Table 2.

[0109] Table 2

[0110]

[0111] The hardware in this embodiment is a Linux Ubuntu 22.04 server with kernel version 5.19.0, an Intel Core i9-12900k 3.2GHz CPU, a single NVIDIA GeForce GTX 3080 GPU, and 32GB of RAM. The algorithm is implemented using the PyTorch 2.0 framework and Python 3.9.

[0112] To comprehensively evaluate the overall performance of the proposed method, this embodiment selects the relative error R. E Root mean square error R MSE The specific calculation method for evaluating the prediction effect is as follows:

[0113]

[0114] In the formula: e xy p represents the prediction error value in the x-th row and y-th column of the error matrix; xy Let X be the quasi-measured load value in the x-th row and y-th column of the error matrix; X is the number of rows in the error matrix; and Y is the number of columns in the error matrix.

[0115] Examples 1-4 show the actual and predicted power curves (using different methods) of power transmitted from non-metered systems to metered systems between 8:00 AM and 6:00 PM on a certain day, as shown below. Figure 8-11 As shown.

[0116] Examples 1-4 show the actual and predicted electricity transmission curves from non-metered to metered data for a given month, as well as the electricity transmission curves predicted by different methods. Figure 12-15 As shown.

[0117] Monthly electricity consumption R E R MSE As shown in Tables 3 and 4.

[0118] Table 3

[0119]

[0120] Table 4

[0121]

[0122] The relative errors R of different methods are shown in Tables 3 and 4. E Root mean square error R MSE The values ​​show that the first three prediction methods have relatively small errors. The computational and storage requirements of the four prediction methods are compared in Table 5.

[0123] Table 5

[0124]

[0125] As shown in Table 5, compared with the other three methods, the photovoltaic metering method for multiple grid-connected users using the Autoformer model based on sparse sweep frequency response analysis has lower computational and storage requirements. Considering the relative error R... E Root mean square error R MSE The proposed method has overall superiority.

[0126] In summary, this invention provides a method for accurate metering of photovoltaic power generation at multiple grid-connected points for users with both main and sub-metering. First, the input impedance spectrum is obtained by transmitting a carrier signal through the carrier module built into the photovoltaic energy meter under non-sub-metering conditions. To reduce the difficulty of data transmission, calculation, and storage, a sparse frequency sweep method is adopted, with a frequency sweep interval of 10kHz and a time interval of 15min. Second, the input impedance data is transmitted to the backend master station daily via a wireless power remote meter reading system (GPRS). An Autoformer model is used to predict and decouple and calculate the impedance of different parts, thereby obtaining the amount of electricity generated by photovoltaic power generation under non-sub-metering conditions and transmitted to the sub-metering points, achieving accurate metering of distributed photovoltaic power generation at multiple grid-connected points for users with both main and sub-metering conditions. Finally, multiple sets of relevant field experiments verify the effectiveness of the proposed method, achieving accurate metering of distributed photovoltaic power generation at multiple grid-connected points for users with both main and sub-metering conditions.

[0127] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0128] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0129] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0130] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0131] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0132] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for accurate metering of photovoltaic power at multiple grid connection points for main and branch metering users, characterized in that... Includes the following steps: Step 1: Obtain the impedance value at the photovoltaic power generation metering point to obtain the H-type impedance sample set. ; Step 2: Use the K-SVD sparse decomposition algorithm to obtain sparse impedance values ​​and construct a sparse matrix; Step 3: Update the discrimination dictionary D based on the sparse matrix; Step 4: Use the Autoformer model encoder to decompose the sparse matrix into trend and periodic terms; Step 5: Use the Autoformer model to decompose the impedance value in real time to obtain the real-time mains impedance at different frequencies. Photovoltaic impedance , Metered impedance Other part impedance ; Step 6: Calculate the impedance values ​​of each part under power frequency to achieve accurate metering of distributed photovoltaic systems with multiple grid connection points for main and sub-metering users; The Autoformer model consists of an encoder and a decoder. The encoder decomposes the sparse matrix X into two parts: a trend term and a periodic term. (7) (8) Among them, X trend Indicates the trend term, AvgPool represents the average pooling layer, and Padding represents padding; X s Indicates a periodic term; The decoder has two inputs: the trend term and the period term of the time series. It continuously decomposes the trend component using the autocorrelation module shown in equations (9) and (10) and the time series decomposition unit shown in equations (7) and (8), and superimposes it with the original trend term to achieve progressive decomposition. The Autoformer model calculates the autocorrelation coefficient of the sparse matrix according to the following formula. (9) (10) in, The initial timing sequence is the input. Represents Fourier transform, It is a conjugate operation; f represents the frequency, and is related to 2. Multiplying them yields the angular frequency result; and Multiply the results obtained by integrating the trend term separately, and convert the time series to the frequency domain. ; In step 6, the real-time current value of photovoltaic power generation is measured. for, ， Among them, I n This represents the real-time current value of photovoltaic power generation.

2. The method as described in claim 1, characterized in that: In step 1, a sweep frequency signal is transmitted from the photovoltaic power generation metering point to obtain the voltage and current data corresponding to the frequency of the sweep frequency signal, and the impedance value at the photovoltaic power generation metering point is calculated.

3. The method as described in claim 1, characterized in that: In step 2, the H-type impedance sample set obtained in step 1 is... As training samples, a sub-dictionary of each pattern class is obtained through training using the K-SVD algorithm. , The cascaded sub-dictionaries yield the discriminant dictionary. The objective equation of the K-SVD algorithm is expressed as follows: ， Where X is a sparse matrix; For the i-th sub-dictionary D i The corresponding sparsity coefficients; T0 is the sparsity threshold, F is the F-norm of the matrix; then the test sample y i The reconstruction residuals between the various sub-dictionaries are, ， The K-SVD algorithm can be expressed using a sparse model as follows: ， This yields sparse impedance values. This forms a sparse matrix X.

4. The method as described in claim 1, characterized in that: In step 3, the i-th column d of the discrimination dictionary D is based on the sparse matrix X. i Update ， in, Let d be the i-th column of the sparse matrix X. i The corresponding i-th row; Let E i To define the error matrix of the discrimination dictionary D for I signal samples in Y after removing the k-th atom from D, we define: , This indicates that no atoms were used. Signal sequence number; Define matrix The corrected matrix error is Transform the above expression into: ， right Perform singular value decomposition, then ,make If the first column of U is... The updated value is Meanwhile, the product of the first column of V and Δ(1,1) is updated. And update each column using the new dictionary. The sparsity coefficients are decomposed until the stopping condition is met.

Images