Power grid power flow sensitivity calculation method based on data-driven mode covariance analysis

Abstract

The application discloses a power grid power flow sensitivity calculation method based on data-driven mode covariance analysis, which realizes complex dynamic power grid sensitivity online real-time batch calculation by online real-time updating of a multi-dimension database with a correlation structure, mining the sensitivity relationship between node injection power and branch power flow in the system, and visually and real-timely seeing the range, degree and change direction of the influence of the change of the node injection power on the branch power flow disturbance, thereby providing a basis for real-time security and stability analysis of the power grid. The power grid power flow sensitivity calculation method based on covariance analysis has the characteristics of real-time updating, is used for real-time operation control of the power grid, effectively mines the economic benefit potential of power grid operation, seeks fast and reasonable power grid over-limit adjustment measures, thereby reducing the work burden of operation personnel, and greatly improving the work efficiency of power grid safety analysis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power grid operation safety analysis, in particular to a power grid power flow sensitivity calculation method based on data-driven mode covariance analysis. BACKGROUND

[0002] With the development of social economy, the demand for electricity continues to increase; at the same time, large-capacity DC power transmission and large-scale intermittent new energy such as wind power are connected to the power grid, making the structure and operation mode of China's interconnected power grid extremely complex, and power flow blocking phenomenon often occurs in some lines or transmission sections in the power grid. The pre-set control plan cannot adapt to the complex and variable operation mode, and it is difficult to ensure the safe operation of the power grid under complex operation state. It becomes increasingly important to analyze the online operation of the power system and give control strategies.

[0003] Sensitivity is a method for studying and analyzing the sensitivity of the state or output change of a system or model to the change of system parameters or surrounding conditions, which uses the physical expression relationship contained in the system or model to obtain the influence of the change of independent variables on dependent variables. When the power flow blocking phenomenon occurs in the line of the power grid, quickly finding and adjusting the output of the generator with high power flow adjustment sensitivity of the line can reduce the operation risk of the power grid. The physical concept of sensitivity analysis is clear, and it is easy to achieve the optimal adjustment amount, so it has been widely used in many fields of power systems.

[0004] However, the traditional sensitivity analysis method needs to control the injection power of other nodes to be constant when analyzing the sensitivity of single node injection power to branch power, which is contrary to the dynamic characteristics of power systems, that is, the injection power of each node of the power system is in a dynamic process at all times, and the sensitivity of each node to branch power cannot be obtained online in real time. Moreover, the traditional sensitivity analysis method needs to change the output of each generator respectively and then calculate the power flow, which cannot meet the timeliness requirement of online control of the power grid for batch calculation of power flow in a large-scale power grid.

[0005] The data of the dynamic power system itself no longer stays in a single dimension and relatively independent data structure, but more in a multi-dimensional associated structure. Therefore, how to efficiently use the real-time power flow data of the power grid, mine the relationship between the node injection power and the branch power in the system, and construct a sensitivity model that can be analyzed online in real time to realize batch calculation of sensitivity is of great significance to the safety and stability analysis of the power grid. SUMMARY

[0006] The present application provides a power grid power flow sensitivity calculation method based on data-driven mode covariance analysis.

[0007] The present application is realized by the following technical scheme:

[0008] The power grid power flow sensitivity calculation method based on data-driven mode covariance analysis comprises the following steps:

[0009] Step one: build a power grid base state power flow real-time database.

[0010] Get the online operation mode of the power grid at the current time section from the power dispatch automation system, and take the voltage, injected power of the nodes, and the power flow of the branches in the power grid as the base state power flow data of the power grid in this scenario. The sensitivity analysis takes this base state power flow data as the benchmark.

[0011] Step two: build a power flow sensitivity calculation model based on covariance analysis.

[0012] To more accurately study the influence of the injected power of a node as a control variable on the power flow of a branch, the influence of the injected power of other nodes on the power flow of the branch should be removed. Through the covariance analysis method, the injected power of other nodes in the sample database is taken as a covariate, and under the condition of removing the influence of the covariate on the power flow of the branch, the sensitivity index of the injected power of the node as the control variable on the power flow of the branch is analyzed to meet the requirements of online real-time analysis.

[0013] The specific operation is as follows:

[0014] A sample database containing the output of N generator nodes (G1, G2, …, G N ) and the power flow of branch L is provided. It is assumed that the line L is power blocked, and it is necessary to quickly search for the generator set that has a greater impact on the transmission power of L.

[0015] A power flow sensitivity calculation model based on covariance analysis is built to obtain the sensitivity relationship between the output of generator node G1 and the power flow of branch L.

[0016] One: preliminary work, select the output of generator node G1 as the control variable P, the output of the remaining generator nodes as the covariate X, and the power flow of branch L as the dependent variable Y. Due to the real-time nature of power grid power flow operation, the data in the sample database has certain fluctuations. To reduce the impact of data fluctuations, through clustering, set r clustering centers of the control variable P, and divide the sample database into r groups.

[0017] Two: build a covariance analysis model, let Y ij represent the dependent variable of the jth sample in the ith group, X ij be the covariate of the same sample, represent the mean of all samples, τ i is the i-th grouping effect.

[0018] The model meets the condition: random error ε ij ~ NID (0, σ2 ), the covariate and the control variable are independent of each other; X ij not affected by the grouping τ i ; there is a linear relationship between the dependent variable Y and the covariate X, and the linear regression coefficients β i of each group are equal.

[0019] The covariance analysis model is:

[0020]

[0021] The parameter estimation of the model is:

[0022]

[0023] Three: correction of data, under the condition of hypothesis testing of the covariance model, correct the branch current to eliminate the influence of the output of other generator nodes as the covariate X on the dependent variable data Y.

[0024]

[0025] After data correction, the corrected branch current Y' completely eliminates the influence of the output of the remaining generator nodes as the covariate X.

[0026] Four: establish a linear regression relationship between the control variable P as the output of the generator node G1 and the branch current Y' eliminating the influence of the covariate X, and the regression coefficient obtained is the sensitivity of the branch L current to the output of the generator node G1, which is the sensitivity index S.

[0027] Step three: construct the real-time sensitivity index of all nodes in the grid operation scenario.

[0028] In the real-time sample database of the grid, through the current sensitivity calculation model based on covariance analysis, the injection power of other nodes in the sample database which is not the control variable is treated as the covariate, and the linear regression relationship between the injection power of a single node and the branch current eliminating the influence of the covariate is established in turn, to obtain the real-time sensitivity index of the injection power of all nodes and the branch current in the online real-time operation scenario of the grid.

[0029] Step four: count the generator nodes sensitive to the branch of the node in the grid operation scenario.

[0030] When performing safety analysis on the grid, the generator nodes that have a greater impact on the branch are counted in reverse with the branch as the unit, and the sensitivity index S of the injection power of all nodes to the branch is counted, wherein the node with the largest absolute value of S is the most sensitive generator node corresponding to the branch in the grid operation scenario, which can be used to prioritize the power regulation of the line L.

[0031] Further, in the step one, the power grid flow data in the sample database takes the data of each ten minutes before and after the current base state flow scene as the analysis data sample, the sample database is refreshed every 1 min, the requirement of online real-time operation of the power grid is met, and the accuracy of the power grid sensitivity analysis is ensured.

[0032] The beneficial effects of the present application are:

[0033] Compared with the prior art, the power grid flow sensitivity calculation method based on data-driven mode covariance analysis is more in line with the dynamic real-time characteristics of the power system, can avoid batch power flow calculation on the system, and does not require accurate power grid topology data. By online real-time updating of the multi-dimensional database with associated structures, the sensitivity relationship between the node injection power and the branch flow in the system is mined, the online real-time batch calculation of the complex dynamic power grid sensitivity is realized, the range, degree and change direction of the influence of the change of the node injection power on the branch flow disturbance can be directly and intuitively seen in real time, and thus a basis for real-time safety and stability analysis of the power grid is provided. The power grid flow sensitivity calculation method based on covariance analysis has the characteristics of real-time updating, changes the traditional power grid dispatching operation mode guided by offline calculation to the real-time dispatching operation mode based on online knowledge discovery, can be used for real-time operation control of the power grid, effectively mines the economic benefit potential of the power grid operation, seeks fast and reasonable power grid out-of-limit adjustment measures, thereby reducing the work burden of the operation personnel, and greatly improving the work efficiency of the power grid safety analysis. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application is further described in detail. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0035] The power grid flow sensitivity calculation method based on data-driven mode covariance analysis comprises the following steps:

[0036] Step one: constructing a power grid base state flow real-time database.

[0037] The online operation mode of the power grid under the current time section is obtained from the power dispatching automation system, the voltage, injection power of the nodes and the flow of the branches in the power grid are taken as the base state flow data of the power grid under this scene, and the base state flow data is taken as the benchmark for sensitivity analysis.

[0038] The power grid flow data in the sample database takes the data of each ten minutes before and after the current base state flow scene as the analysis data sample, the sample database is refreshed every 1 min, the requirement of online real-time operation of the power grid is met, and the accuracy of the power grid sensitivity analysis is ensured.

[0039] Step two: build the power flow sensitivity calculation model based on covariance analysis.

[0040] In order to more accurately study the influence of the power injection of a certain node as a control variable on the branch power flow, the influence of the power injection of other nodes on the branch power flow should be eliminated. The power injection of other nodes in the sample database is taken as a covariant by the covariance analysis method, and the sensitivity index of the power injection of a node as a control variable on the branch power flow is analyzed under the condition that the influence of the covariant on the branch power flow is eliminated, so as to meet the requirement of online real-time analysis.

[0041] The specific operation is as follows:

[0042] A sample database containing the output of N generator nodes (G1, G2, …, G N ) and the branch L power flow is provided. It is assumed that the line L is power blocked, and the generator group that has a greater influence on the transmission power of L needs to be quickly searched.

[0043] A power flow sensitivity calculation model based on covariance analysis is built to obtain the sensitivity relationship between the output of the generator node G1 and the branch L power flow.

[0044] One: preliminary work, the output of the generator node G1 is selected as the control variable P, the output of the remaining generator nodes is selected as the covariant X, and the power flow of the branch L is selected as the dependent variable Y. In view of the real-time operation of the power grid power flow, the data in the sample database has a certain fluctuation. In order to reduce the influence of data fluctuation, the sample database is divided into r groups by clustering and setting r clustering centers of the control variable P.

[0045] Two: build the covariance analysis model, let Y ij represent the dependent variable of the jth sample in the ith group, X ij represent the covariant of the same sample, represent the mean value of all samples, τ i is the i-th grouping effect.

[0046] The model meets the following conditions: the random error ε ij ~NID(0, σ 2 ), the covariant and the control variable are independent of each other; X ij is not affected by the grouping τ i ; the dependent variable Y and the covariant X have a linear relationship, and the linear regression coefficients β i of each group are equal.

[0047] The covariance analysis model is:

[0048]

[0049] The parameter estimation of the model is:

[0050]

[0051] Three: correction of data, under the condition of hypothesis test of covariance model, correct the branch flow to eliminate the influence of other unit node output as the covariant X on the dependent variable data Y.

[0052]

[0053] After data correction, the corrected branch flow Y' completely eliminates the influence of the rest of the unit node output as the covariant X.

[0054] Four: the control variable P as the unit node G1 output and the branch flow Y' eliminating the influence of the covariant X are established linear regression relationship, and the regression coefficient obtained is the sensitivity of the branch L flow to the unit node G1 output eliminating the rest of the unit node output, which is the sensitivity index S.

[0055] Step three: construct the real-time sensitivity index of all nodes in the grid operation scenario.

[0056] In the grid real-time sample database, through the power flow sensitivity calculation model based on covariance analysis, the sample database other node injection power not as a control variable is treated as a covariant, and the linear regression relationship of single node injection power as a control variable and branch flow eliminating the influence of the covariant is established, and the real-time sensitivity index of all node injection power and branch flow in the grid online real-time operation scenario is obtained.

[0057] Step four: count the sensitive generator node of the branch to the node in the grid operation scenario.

[0058] When the grid is analyzed, the generator node with greater influence on the branch is counted in the reverse direction, and the sensitivity index S of all node injection power to the branch is counted, wherein the node with the largest absolute value of S is the most sensitive generator node corresponding to the branch in the grid operation scenario, which can be preferentially used for power regulation of the line L.

[0059] The beneficial effects of the present application are:

[0060] Compared with the prior art, the power grid power flow sensitivity calculation method based on data-driven mode covariance analysis is more in line with the dynamic real-time characteristics of the power system, can avoid batch power flow calculation on the system, and does not require accurate power grid topology data. By online real-time updating of the multi-dimensional database with associated structures, the sensitivity relationship between node injection power and branch power flow in the system is mined, online real-time batch calculation of complex dynamic power grid sensitivity is realized, the range, degree and change direction of the influence of the change of node injection power on branch power flow disturbance can be directly and intuitively seen in real time, and thus a basis for real-time safety and stability analysis of the power grid is provided. The power grid power flow sensitivity calculation method based on covariance analysis has the characteristics of real-time updating, changes the traditional power grid dispatching operation mode guided by offline calculation into a real-time dispatching operation mode based on online knowledge discovery, can be used for real-time operation control of the power grid, effectively mines the economic benefit potential of power grid operation, seeks fast and reasonable power grid out-of-limit adjustment measures, thereby reducing the work burden of operation personnel, and greatly improving the work efficiency of power grid safety analysis.

[0061] The basic principles, main features and advantages of the present application are shown and described above. It should be understood by those skilled in the art that the present application is not limited by the above examples, the above examples and descriptions in the specification are only preferred examples of the present application, and are not intended to limit the present application, various changes and improvements can be made to the present application without departing from the spirit and scope of the present application, and these changes and improvements all fall within the scope of the present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for power grid power flow sensitivity calculation based on data-driven pattern covariance analysis, characterized in that, The method comprises the following steps: Step 1: Constructing the real-time database of the base state power flow of the power grid: Obtain the online operation mode of the power grid at the current time section from the power dispatch automation system, take the voltage, injected power of nodes and the power flow of branches in the power grid as the base state power flow data of the power grid in this scenario, and take the base state power flow data as the benchmark for sensitivity analysis; Step 2: Constructing the power flow sensitivity calculation model based on covariance analysis: In order to more accurately study the influence of the injected power of a node as a control variable on the power flow of a branch, the influence of the injected power of other nodes on the power flow of the branch should be removed, the injected power of other nodes in the sample database is treated as a covariate through the covariance analysis method, and the sensitivity index of the injected power of a node as a control variable on the power flow of a branch is analyzed under the condition of removing the influence of the covariate on the power flow, so as to meet the requirements of online real-time analysis; The specific operation is as follows: A sample database of power output of N generator nodes (G1, G2, …, G N ) and branch L power flow is provided, and when power blockage occurs in line L, a generator set that has a greater impact on transmission power of L needs to be quickly searched. Construct the power flow sensitivity calculation model based on covariance analysis, and obtain the sensitivity relationship between the output of the unit node G1 and the power flow of the branch L: I: Preliminary work, select the output of the unit node G1 as the control variable P, the output of the remaining unit nodes as the covariate X, and the power flow of the branch L as the dependent variable Y. Due to the real-time nature of the power flow operation of the power grid, the data in the sample database has a certain fluctuation. In order to reduce the influence of data fluctuation, the sample database is divided into r groups through clustering by setting r clustering centers of the control variable P; Two: Construct the covariance analysis model, set Y ij Yij represents the dependent variable of the jth sample of the ith group, X ij is the covariate of the same sample, Y represents the mean of all samples, τ i is the ith grouping effect; The model meets the conditions: random error ε ij ~NID(0, σ 2 ), the covariate and the control variable are independent of each other; X ij Not affected by the grouping τ i ; There is a linear relationship between the dependent variable Y and the covariate X, and the linear regression coefficients β i of each group are equal; The covariance analysis model is: The parameter estimation of the model is: III: Data correction, under the condition of meeting the hypothesis test of the covariance model, correct the power flow of the branch to remove the influence of the output of the remaining unit nodes as the covariate X on the dependent variable data Y: After data correction, the corrected power flow Y' of the branch completely removes the influence of the output of the remaining unit nodes as the covariate X; IV: Establish a linear regression relationship between the control variable P of the output of the unit node G1 and the power flow Y' of the branch which removes the influence of the covariate X, and the regression coefficient obtained is the sensitivity of the power flow of the branch L to the output of the unit node G1, which is the sensitivity index S; Step 3: Constructing the real-time sensitivity index of all nodes in the operation scenario of the power grid: In the real-time sample database of the power grid, through the power flow sensitivity calculation model based on covariance analysis, the injected power of other nodes in the sample database which is not a control variable is treated as a covariate, and a linear regression relationship between the injected power of a single node as a control variable and the power flow of a branch which removes the influence of the covariate is established in turn, so as to obtain the real-time sensitivity index of the injected power of all nodes in the online real-time operation scenario of the power grid to the power flow of the branch; Step 4: Statistics of the generator nodes sensitive to the nodes in the operation scenario of the power grid: When the power grid is analyzed for safety, the generator nodes which have a greater impact on the branch are statistically analyzed in reverse with the branch as the unit, the sensitivity index S of the injected power of all nodes to the branch is obtained, the node with the largest absolute value of S is the most sensitive generator node corresponding to the branch in the operation scenario of the power grid, which can be preferentially used for power control of the line L.

2. The method of claim 1, wherein, In the step one, the power grid flow data in the sample database takes the data of each ten minutes before and after the current ground state flow scene as the analysis data sample, the sample database is refreshed every 1 min, the requirement of online real-time operation of the power grid is met, and the accuracy of the power grid sensitivity analysis is ensured.

Images