A data-driven method for calculating power system flow

By introducing auxiliary state variables and data-driven methods in the current calculation, the problem of difficult convergence of model driving methods in the smart grid is solved, efficient calculation and adaptability are achieved, and it is suitable for frequent topological changes in the smart grid.

CN113609435BActive Publication Date: 2024-11-15NORTH CHINA ELECTRIC POWER UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202110922962.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-08-12
Publication Date
2024-11-15
Estimated Expiration
2041-08-12

AI Technical Summary

Technical Problem

The existing model-driven trend calculation method is difficult to converge in smart grids, and the uncertainty of grid parameters affects the calculation results. The traditional Newton-Ravson method and the rapid decomposition method are difficult to face the output and load of intermittent and uncertainties. Adapt to topological changes.

Method used

By introducing auxiliary state variables, the nonlinear flow equation is converted into linear equations, and linear regression is performed using data-driven methods to judge the topological structure similarity and obtain the mapping matrix through offline learning, so as to realize online calculations to obtain state variables and avoid iterative calculations.

Benefits of technology

It realizes efficient computing, avoids convergence problems, adapts to different topological structures, and is especially suitable for frequent changes in smart grids, and has a much higher computing efficiency than traditional methods.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113609435B_ABST
    Figure CN113609435B_ABST
Patent Text Reader

Abstract

The present invention discloses a data-driven power system power flow calculation method. By introducing auxiliary state variables, the original nonlinear power flow equation is converted into an accurate linear power flow equation, which is further expressed as an accurate linear regression equation. Before solving, a step method is given to determine whether the current power flow data has the same topological structure as the historical / simulated data. Finally, an offline learning step method is given to solve the linear regression equation to obtain a mapping matrix, and an online calculation step method is given to obtain auxiliary state variables, and state variables are obtained through nonlinear and linear transformations. The power flow calculation method described in the present invention has no convergence problem. While ensuring similar calculation accuracy, the calculation efficiency is much higher than the traditional Newton-Raphson method and the fast decomposition method, and is suitable for online applications in large-scale networks.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of power system analysis, and specifically relates to a data-driven power flow calculation method (DDPF). Background Art

[0002] Power flow calculation is one of the most important calculation methods in power systems and is widely used in power system planning, operation and control. The earliest proposed power flow calculation methods include Gauss-Seidel method and Newton-Raphson method. Compared with Gauss-Seidel method, Newton-Raphson method has better convergence and higher computational efficiency. In order to further improve the computational efficiency of Newton-Raphson method, researchers proposed fast decomposition method. In the emerging smart grid, the existing model-driven calculation methods have the following significant disadvantages: (1) In smart grid, the intermittent and uncertain nature of output and load become more and more obvious, which will make it difficult for traditional Newton-Raphson method and fast decomposition method to converge. This is because the two algorithms usually use the power flow calculation results of the previous data as the initial value; (2) The grid parameters have certain uncertainties. The grid parameters stored in the database of the grid control center may differ from the actual values, which has a negative impact on the model-driven calculation method.

[0003] Purpose of the Invention

[0004] The purpose of the present invention is to address at least one of the above-mentioned deficiencies in the prior art and to provide at least a useful commercial option. By proposing a data-driven power flow calculation method, complex convergence problems can be avoided, the method has high computational efficiency, can handle different topologies, and can adapt to frequent changes in the topology of smart grids. Summary of the invention

[0005] The present invention provides a data-driven power system flow calculation method, comprising the following steps:

[0006] Step A. Introduce auxiliary state variables to transform the original nonlinear power flow equation into an accurate linear power flow equation, which is further expressed as an accurate linear vector matrix regression equation Y=HZ+V;

[0007] Step B. Give the steps to determine whether the current flow data and historical / simulated data have the same topological structure, and calculate Determine whether there is the same topology.

[0008] Step C. Give the offline learning step method to solve the linear regression equation and obtain the mapping matrix Given the online calculation steps, we can get the auxiliary state variables Finally, linear and nonlinear transformations are performed to obtain the state variable u i ,θ i .

[0009] Step A includes:

[0010] Step A1: Convert the original nonlinear power flow equation into an accurate linear power flow equation, where P i and Q i is the active and reactive injection power of node i, k(i) is all nodes directly connected to node i (including i), and the original nonlinear power flow equation is Transformed into an accurate linear power flow equation: Where R ij =v i v j cosθ ij , I ij =v i v j sinθ ij and Auxiliary state variables.

[0011] After introducing the auxiliary state variables, the original nonlinear power flow equation is transformed into the corresponding linear equation. For different PQ and PV nodes, the corresponding linear power flow equations are established, the total number is 2(N-1), and the total number of auxiliary state variables is N-1+2b, where b is the number of branches.

[0012] Step A2: Based on the accurate linear power flow equation obtained in step A1, it is further expressed as in is the auxiliary boundary vector, is the auxiliary state vector, is a constant matrix whose elements are determined by the topological structure and parameters of the network. Further, the accurate linear regression equation is obtained in It is an unknown constant mapping matrix that depends on the network topology and network parameters. Assuming that there are s historical / simulated power flow data with the same topology, and each data contains given boundary conditions and power flow calculation results, the auxiliary boundary vectors and auxiliary state vectors of all these s historical / simulated power flow data are aggregated, and the estimated value of H is obtained through the offline learning stage. is the error that may exist in the historical / simulated data when the expected value E(v)=0.

[0013] Preferably, step B includes processing different topological structures and determining whether the current flow data and the historical / simulated data have the same topological structure.

[0014] Preferably, step B further comprises the following sub-steps:

[0015] Sub-step B1: Take any two historical data and calculate their corresponding auxiliary boundary vectors based on step A and if and If there is a strong linear correlation, they can be considered to have the same topological structure.

[0016] Sub-step B2: Determine its linear correlation. If It is considered that they have symphonic linear correlation, and their corresponding topological structures are the same. threshold The value of is very important, but in a large number of simulation experiments, it is found that even if R threshold When the value is 0.9, the above method can still correctly identify historical measurement data with exactly the same topological structure. RCC is used instead of Pearson correlation coefficient (PCC) because RCC does not require the auxiliary measurement vector to follow a normal distribution and RCC is robust when there are significant errors in the historical data.

[0017] Preferably, in step C, based on steps A and B, an offline learning step is given to obtain a mapping matrix And the auxiliary state variables are obtained through the online calculation stage, and finally the state variables are obtained by nonlinear and linear transformation.

[0018] Preferably, the step C further comprises the following sub-steps:

[0019] Sub-step C1: Based on step A, the historical / simulated power flow data with the same topological structure are aggregated, the collinearity effect is ignored, and the mapping matrix H is directly estimated using the weighted least squares (WLS) method. The estimated mapping matrix can be obtained The ratio of the number of historical / simulated power flow data s to the number of auxiliary state variables 2(N-1) affects the mapping matrix The accuracy is guaranteed to be available when s ≥ 2(N-1);

[0020] Sub-step C2: Based on step B, quickly find the historical / simulated data with the same topological structure as the current power flow data and obtain the corresponding Matrix, through Get the auxiliary state vector of the current data. and are the auxiliary boundary vectors of the current power flow data and the estimated auxiliary state vector;

[0021] Step C3: The solution of the power flow calculation is the auxiliary state variable, and the original state variable is obtained through nonlinear and linear transformation.

[0022] More preferably, the sub-step B1 further comprises:

[0023] Step B11: Calculation

[0024] More preferably, the sub-step C1 further comprises:

[0025] Step C11: Calculation

[0026] More preferably, the step C2 further comprises:

[0027] Step C21: Calculation BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 The present invention is a flowchart of the data-driven power flow calculation method.

[0029] Figure 2 It is the RCC correlation matrix of 6 groups of auxiliary boundary vectors in the IEEE300 node system selected in the specific embodiment of the present invention.

[0030] Figure 3 Ratio and |dv| of different IEEE systems in the specific embodiment of the present invention. m relation.

[0031] Figure 4 Ratio and |dθ| of different IEEE systems in the specific embodiment of the present invention m relation. DETAILED DESCRIPTION

[0032] The present invention provides a data-driven power flow calculation method (DDPF), which has no convergence problem. While ensuring similar calculation accuracy, the calculation efficiency is much higher than that of the traditional Newton-Raphson method and the fast decomposition method. The simulation example verifies that the proposed method has no convergence problem and high calculation efficiency, and is suitable for online application in large-scale networks.

[0033] Additional advantages of the present invention will be partly presented in the following description and will become more apparent from the following description.

[0034] The present invention will be further described in detail below in conjunction with the accompanying drawings.

[0035] like Figure 1 As shown, the data-driven power flow calculation method of the present invention comprises the following steps:

[0036] Step A. Introduce auxiliary state variables, transform the original nonlinear power flow equation into an accurate linear power flow equation, further expressed as an accurate linear vector matrix regression equation Y=HZ+V, propose a data-driven power flow solution method, establish a linear solution equation, and make it free of computational convergence problems. Step A further includes the following sub-steps:

[0037] Step A1: Introduce auxiliary state variable R ij ,I ij 、u i , the original nonlinear power flow equation is converted into an accurate linear power flow equation. The original nonlinear power flow equation is:

[0038] P i =v i ∑ j∈k(i) v j (G ij cosθ ij +B ij sinθ ij ) (1)

[0039] Q i =v i ∑ j∈k(i) v j (G ij sinθ ij -B ij cosθ ij ) (2)

[0040] Where: P i and Q i is the active and reactive injection power of node i, k(i) is all nodes directly connected to node i (including i), G ij +jB ij is the node admittance matrix element. In order to ensure generality, it is assumed that node i is a relaxed node and the voltage amplitude phase angle is v i ∠θ i (i=2,3,…,N). Transformed into an accurate linear power flow equation:

[0041] P i =G ii u i +∑ j∈k(i),j≠i (G ij R ij +B ij I ij ) (3)

[0042] Q i =-B ii u i +∑ j∈k(i),j≠i (G ij Iij -B ij R ij ) (4)

[0043] Where: R ij =v i v j cosθ ij , I ij =v i v j sinθ ij and Auxiliary state variables.

[0044] For each PV node, establish the equation:

[0045] u i.PV =u i (5)

[0046] Where: And v i.PV is the voltage amplitude of the given ith PV node.

[0047] After introducing the auxiliary state variables, the original nonlinear power flow equation is transformed into the corresponding linear equation. For different PQ and PV nodes, the corresponding linear power flow equations are established, the total number is 2(N-1), and the total number of auxiliary state variables is N-1+2b, where b is the number of branches.

[0048] Step A2: A data-driven power flow solution method is proposed, an accurate linear regression equation is established, and a mapping constant matrix H to be solved is formed.

[0049] Based on the accurate linear power flow equation obtained in step A1, it is further expressed as:

[0050]

[0051] Where: is the auxiliary boundary vector, is the auxiliary state vector, is a constant matrix whose elements are determined by the topological structure and parameters of the network. According to formula (4), the exact linear regression equation is obtained:

[0052]

[0053] Where: It is an unknown constant mapping matrix that depends on the network topology and network parameters and is obtained through the offline learning phase. is the error that may exist in the historical / simulated data when the expected value E(v)=0.

[0054] Use a large amount of historical / simulated power flow data to learn H. Assume that there are s historical / simulated power flow data with the same topology, and each data contains given boundary conditions and power flow calculation results. Aggregate the auxiliary boundary vectors and auxiliary state vectors of all these s historical / simulated power flow data into:

[0055]

[0056]

[0057] Where: and are the auxiliary boundary vector and auxiliary state vector in the i-th historical / simulated flow data, respectively. Y and Z can be obtained by the accurate linear regression equation Y = HZ + V, where It is the error matrix when the expected value E(v)=0.

[0058] Step B: Process different topological structures and determine whether the current flow data and the historical / simulated data have the same topological structure.

[0059] Step B1: Take any two historical data and calculate their corresponding auxiliary boundary vectors based on step A and if and If there is a strong linear correlation, then they can be considered to have the same topological structure. In statistics, the Spearman rank correlation coefficient (RCC) can be used to measure the correlation between them:

[0060]

[0061] Where: is the rank difference between the two auxiliary boundary vectors, rg(*) represents the continuous sorting number from small (starting from 1) to large, and yes and The i-th element of m represents the number of auxiliary boundary vector elements.

[0062] Step B2: Determine its linear correlation. If It is considered that they have symphonic linear correlation, and their corresponding topological structures are the same. threshold The value of is very important, but in a large number of simulation experiments, it is found that even if R threshold When the value is 0.9, the above method can still correctly identify historical measurement data with exactly the same topological structure. RCC is used instead of Pearson correlation coefficient (PCC) because RCC does not require the auxiliary measurement vector to follow a normal distribution and RCC is robust when there are significant errors in the historical data.

[0063] Step C. Give the offline learning step method to solve the linear regression equation and obtain the mapping matrix Given the online calculation steps, we can get the auxiliary state variables Finally, nonlinear and linear transformations are performed to obtain the state variable u i ,θ i .

[0064] Step C1: Based on step A, the historical / simulated power flow data with the same topological structure are aggregated, the collinearity effect is ignored, and the mapping matrix H is directly estimated using the weighted least squares (WLS) method.

[0065]

[0066] The estimated mapping matrix can be obtained The ratio of the number of historical / simulated power flow data s to the number of auxiliary state variables 2(N-1) affects the mapping matrix The accuracy is guaranteed to be available when s ≥ 2(N-1).

[0067] Step C2: Based on step B, quickly find the historical / simulated data with the same topological structure as the current power flow data and obtain the corresponding Matrix, through Get the auxiliary state vector of the current data. and are the auxiliary boundary vectors of the current power flow data and the estimated auxiliary state vector.

[0068] Step C3: The solution of the power flow calculation is the auxiliary state variable, which is transformed into the intermediate variable y I and defined as

[0069] y I =[u i,I θ ij,I ] (12)

[0070] There is the following nonlinear transformation relationship between the intermediate state variables and the required auxiliary state variables:

[0071]

[0072] The final state variable can be obtained by formula (14):

[0073]

[0074] In the formula, A eis the reduced-order node-branch correlation matrix in the power system (excluding slack nodes). The main calculation in the online calculation stage comes from matrix multiplication, which is simpler than the traditional Newton-Raphson method and fast decomposition method.

[0075] The step B1 comprises:

[0076] Step B11: Calculation

[0077] The step C1 comprises:

[0078] Step C11: Calculation

[0079] The step C2 comprises:

[0080] Step C21: Calculation

[0081] In order to enable those skilled in the art to better understand the present invention and appreciate the advantages of the present invention over the prior art, further explanation is given below in conjunction with specific embodiments.

[0082] like Figure 2 As shown in the figure, firstly, six groups of auxiliary boundary vectors in the IEEE300 node system are selected to verify the accuracy of RCC. Secondly, Ratio=s / (2(N-1)) is set to test the influence of the amount of data used in the offline learning phase on the accuracy of the proposed DDPF method. The proposed method is compared with the two existing DDPF algorithms to verify the calculation accuracy. Finally, the IEEE14, 30, 33, 57, 118, 123, and 300 node systems are used to test and compare the performance with the traditional NRPF and FDPF algorithms under the conditions of uncertain output and load and frequent changes in topology. The calculation of NRPF and FDPF is performed using Matpower (convergence accuracy is 10 -8 ) to illustrate the applicability of the proposed algorithm in large-scale network power flow calculations.

[0083] 1. Accuracy analysis

[0084] In order to verify the accuracy of the RCC-based topology identification method, six groups of auxiliary boundary vectors are selected from the IEEE-300 node system, corresponding to three different topological structures, namely and and and Each group has the same topological structure. Calculate the RCC correlation matrix:

[0085]

[0086] Where: is the rank difference between the two auxiliary boundary vectors, rg(*) represents the continuous sorting number from small (starting from 1) to large, and yes and The ith element of .

[0087] like Figure 2 As shown, the RCC method can accurately identify the same topology structure, and the topology identification results of other IEEE systems are also correct.

[0088] 2. Accuracy analysis

[0089] The algorithm proposed in this invention is an improvement on the data driven power flow algorithm. To illustrate the accuracy of the algorithm, two other data driven algorithms DDPF are respectively used. (1) and DDPF (2) The simulation analysis is carried out with the algorithm proposed in the present invention.

[0090] Figure 3-4 The error between the true value and the proposed DDPF method is given for different ratios Ratio=s / (2(N-1))(0.1-1.2), including the mean absolute error of the voltage amplitude |dv| m and the mean absolute error of the phase angle |dθ| m . The results show that as the ratio increases, |dv| m and |dθ| m All of them decrease accordingly. In addition, it can be seen from Table 1 that the |dv| m With |dθ| m Both are smaller than DDPF (1) and DDPF (2) The results given in .

[0091] Table 1 |dv| obtained by different DDPF methods when Ratio=1 m With |dθ| m

[0092]

[0093] This proves that the DDPF method proposed in this paper has high calculation accuracy, and when Ratio=0.2, |dv| m and |dθ| m All less than 10 -4 , which means that the proposed method can still be used when there is less historical / simulation data.

[0094] 3. Convergence and computational efficiency analysis

[0095] Assuming that the uncertainty of output and load makes the voltage amplitude vary randomly between 0.95-1.05 and the phase angle varies randomly between -15°-15°, the state variables of the previous data are used as the initial values ​​of the traditional Newton-Raphson method and the fast decomposition method. Table 2 gives the number of times η that different methods do not converge in 100 tests.

[0096] Table 2 Calculation results of different power flow algorithms under uncertainty in 100 tests

[0097]

[0098] From the data in Table 2, it can be seen that when there is uncertainty in output and load, the NRPF and FDPF algorithms are difficult to converge in most tests, while the DDPF algorithm proposed in this paper can obtain quite accurate results in all cases. Since the online calculation stage of the DDPF algorithm does not require iterative calculation, the number of iterations is 0. Among the listed power flow algorithms, the online calculation efficiency of the DDPF method proposed in this paper is much higher than other methods, and it is more suitable for large-scale networks.

[0099] 4. Adaptability analysis

[0100] In 100 continuous power flow sample data, assuming that the topology changes continuously, for each sample data, the NRPF, FDPF and DDPF algorithms of this paper are tested respectively. In all tests, the state variables of the previous data are used as the initial values ​​of the traditional Newton-Raphson method and the fast decomposition method. Table 3 gives the number of times η that NRPF and FDPF do not converge in the test.

[0101]

[0102] It can be seen from Table 3 that when the topology changes frequently, the traditional NRPF and FDPF algorithms are difficult to converge, while in all tests, the DDPF algorithm proposed in this paper can still find historical / simulated data that is consistent with the current data topology and obtain the power flow solution with the same accuracy as Table 1. Table 3 gives the average number of iterations and calculation time, indicating that even when the topology changes frequently, a high-precision power flow solution can be obtained and a certain calculation efficiency can be guaranteed.

[0103] It can be seen from the above simulation analysis that the calculation method of the present invention has advantages in large-scale network power flow calculation, which are specifically manifested in: there is no convergence problem, high calculation efficiency, and the ability to handle different topological structures and adapt to frequent changes in smart grid topological structures.

[0104] The above embodiments describe the technical solutions of the present invention in detail. Obviously, the present invention is not limited to the described embodiments. Based on the embodiments of the present invention, people familiar with the technical field can also make various changes accordingly, but any changes that are equivalent or similar to the present invention belong to the scope of protection of the present invention.

Claims

1. A data-driven power system flow calculation method, characterized in that: The following steps are involved: Step A. Introduce auxiliary state variables to transform the original nonlinear power flow equation into an accurate linear power flow equation, expressed as an accurate linear vector matrix regression equation Y=HZ+V, where Y represents the auxiliary state vector aggregation matrix, H is the unknown constant mapping matrix that depends on the network topology and network parameters, Z represents the auxiliary boundary vector aggregation matrix, and V is the error in the historical / simulated data when the expected value E(v)=0; Step B. Determine whether the current power system flow data and historical / simulated data have the same topological structure, that is, by calculating To determine whether there is the same topology, where represents the rank correlation coefficient of the two auxiliary boundary vectors, m represents the number of auxiliary boundary vector elements, d i Represents the level difference between two auxiliary boundary vectors; specifically includes the following sub-steps: Sub-step B1: Take any two historical data and calculate their corresponding auxiliary boundary vectors based on step A and if and If there is a strong linear correlation, they can be considered to have the same topological structure; Sub-step B2: Determine its linear correlation. If It is considered that they have symphonic linear correlation, and their corresponding topological structures are the same; RCC is used instead of Pearson correlation coefficient PCC; Step C. Give the offline learning step method to solve the linear regression equation and obtain the mapping matrix Among them, s represents the number of historical / simulated flow data, y i represents the auxiliary state vector in the i-th historical / simulated power flow data, H is a constant mapping matrix that depends on the network topology and network parameters, represents the auxiliary boundary vector in the i-th historical / simulated flow data, Y represents the auxiliary state vector aggregation matrix, and Z represents the auxiliary boundary vector aggregation matrix; the online calculation step method is given to obtain the auxiliary state variable Finally, linear and nonlinear transformations are performed to obtain the state variable u i ,θ i ; In step C, based on steps A and B, an offline learning step is given to obtain a mapping matrix And the auxiliary state variables are obtained through the online calculation stage, and finally the state variables are obtained by nonlinear and linear transformation; The step C further comprises the following sub-steps: Sub-step C1: Based on step A, aggregate the historical / simulated power flow data with the same topological structure, ignore the collinearity effect, and use the weighted least squares (WLS) method to directly estimate the mapping matrix H to obtain the estimated mapping matrix As shown below: The ratio of the number of historical / simulated power flow data s to the number of auxiliary state variables 2(N-1) affects the mapping matrix The accuracy is guaranteed to be available when s ≥ 2(N-1); Sub-step C2: Based on step B, quickly find the historical / simulated data with the same topological structure as the current power flow data and obtain the corresponding Matrix, through Get the auxiliary state vector of the current data, where and are the auxiliary boundary vectors of the current power flow data and the estimated auxiliary state vector; Step C3: The solution of the power flow calculation is the auxiliary state variable, and the original state variable is obtained through nonlinear and linear transformation; Transform auxiliary state variables into intermediate variables y I and defined as yes I =[u i,I ;θ ij,I ]; There is the following nonlinear transformation relationship between the intermediate state variables and the required auxiliary state variables: θ ij,I =arctan(I ij / R ij ) The final state variable is u i =u i,I and get, Where Ae is the reduced-order node-branch association matrix in the power system, excluding slack nodes.

2. The power system flow calculation method according to claim 1, characterized in that: The step A further comprises the following steps: Step A1: Convert the original nonlinear power flow equation into an accurate linear power flow equation, where P i and Q i is the active and reactive injection power of node i, k(i) is all nodes directly connected to node i including node i; the original nonlinear power flow equation is expressed as: P i =v i ∑ j∈k(i) v j (G ij cosθ ij +B ij sinθ ij ), Q i =v i ∑ j∈k(i) v j (G ij sinθ ij -B ij cosθ ij ); It is transformed into an accurate linear power flow equation as follows: P i =G ii u i +∑ j∈k(i),j≠i (G ij R ij +B ij I ij ), Q i =-B ii u i +∑ j∈k(i),j≠i (G ij I ij -B ij R ij ); Among them, R ij =v i v j cosθ ij , I ij =v i v j sinθ ij and is the auxiliary state variable; After introducing the auxiliary state variables, the original nonlinear power flow equation is transformed into the corresponding linear equation; for different PQ and PV nodes, the corresponding linear power flow equations are established, the total number is 2(N-1), and the total number of auxiliary state variables is N-1+2b, where b is the number of branches; Step A2: Based on the accurate linear power flow equation obtained in step A1, it is further expressed as Among them, the auxiliary boundary vector Auxiliary state vector is a constant matrix whose elements are determined by the topological structure and parameters of the network; furthermore, the accurate linear regression equation is obtained in It is an unknown constant mapping matrix that depends on the network topology and network parameters. Assuming that there are s historical / simulated power flow data with the same topology, and each data contains given boundary conditions and power flow calculation results, the auxiliary boundary vectors and auxiliary state vectors of all these s historical / simulated power flow data are aggregated, and the estimated value of H is obtained through the offline learning stage; is the error that may exist in the historical / simulated data when the expected value E(v)=0.

3. The power system flow calculation method according to claim 1, characterized in that: The sub-step B1 further comprises: Step B11: Calculation 4. The power system flow calculation method according to claim 1, characterized in that: The sub-step C1 further comprises: Step C11: Calculation 5. The power system flow calculation method according to claim 1, characterized in that: The step C2 further comprises: Step C21: Calculation