Line parameter identification method based on non-intrusive pmu prior error analysis and correction

Abstract

The application belongs to the technical field of power system measurement and modeling, and particularly relates to a line parameter identification method based on non-intrusive PMU prior error analysis and correction, comprising the following steps: calibrating the PMU through a synchronous standard source, obtaining the measurement errors of voltage amplitude, phase angle and current phase angle, and statistically analyzing the distribution characteristics; establishing an error mapping relationship between the true value known scene and the unknown scene to quantify the influence of the operating state on the resistance, reactance and admittance identification; correcting the identified parameters accordingly, and selecting the arithmetic mean, the truncated mean, the Huber estimation or the GMM-EM weighted average method to determine the final parameters according to the error distribution form. The application can effectively compensate for the PMU measurement error, and significantly improve the accuracy and robustness of parameter identification in the short line and other scenes.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power transmission line parameter measurement, and particularly relates to a line parameter identification method based on non-invasive PMU prior error analysis and correction. BACKGROUND

[0002] Accurate power transmission line parameters are the basis for power system power flow calculation, protection setting, stability analysis and fault location. Parameter setting errors can lead to relay protection misoperation / refusal, stability control errors, new energy consumption capacity evaluation distortion, and other chain risks, affecting the safe operation of large power grids and power supply quality. With the advancement of new power system construction, the demand for model accuracy in power grid operation is changing from static design values to dynamic real values. This change has higher requirements for real-time acquisition and dynamic tracking of parameters.

[0003] The parameter acquisition method of the power transmission line is divided into offline measurement method and online measurement method. The offline measurement method requires a long preparation time for each test, high requirements for manpower and financial resources, and poor test repeatability, which has been gradually replaced by the online measurement method.

[0004] The online measurement method mainly includes line parameter identification based on the SCADA (Supervisory Control And Data Acquisition) and line parameter identification based on the PMU (Phasor Measurement Unit). SCADA lacks measurement data with common time marks, has low data accuracy and lacks phase information, which leads to the error being amplified sharply in the scenes of load fluctuation and asymmetric operation, and cannot meet the high-precision modeling requirements.

[0005] With the large-scale deployment of WAMS (Wide Area Measurement System), PMU has become a revolutionary tool for online identification with microsecond-level synchronization accuracy, millisecond-level data refresh rate and complete phasor data. However, due to the influence of the inherent measurement error of PMU, the line parameter identification method based on PMU has the problem of insufficient identification accuracy: the identification accuracy is insufficient, that is, there is a non-negligible deviation between the identification value and the true value, especially for short lines. The identification result of this method is almost unusable. SUMMARY

[0006] The application provides a line parameter identification method based on non-invasive PMU prior error analysis and correction, which solves the problem of insufficient identification accuracy of the line parameter identification method based on PMU.

[0007] The technical scheme of the application is as follows:

[0008] The line parameter identification method based on non-invasive PMU prior error analysis and correction includes the following steps:

[0009] S1, access the PMU to the synchronization standard source, take the parameter setting value of the synchronization standard source as the standard value, and record the difference between the parameter measurement value output by the PMU and the standard value as the measurement error of the PMU;

[0010] S2, calculate the average value and standard deviation of the measurement error, and obtain the distribution form of the measurement error based on the average value and the standard deviation, the distribution form being described based on kurtosis, skewness and bimodal coefficient;

[0011] S3, based on the PMU output value with the measurement error, calculate the identification error in the simulation scene and the real scene, the difference between the simulation scene and the real scene being that the identification parameter of the simulation scene is known and the identification parameter of the real scene is unknown;

[0012] S4, based on the principle that the resistance is much smaller than the reactance, simplify the resistance to reactance ratio in the identification error mapping relationship formula of the real scene and the simulation scene to be approximately 0;

[0013] Based on the law of large numbers, the random error term in the identification error mapping relationship formula of the real scene and the simulation scene is divided by processing;

[0014] Based on the inherent error short-time invariance principle, the inherent error term in the identification error mapping relationship formula of the real scene and the simulation scene is divided by processing;

[0015] S5, based on the simplified identification parameter calculation formula of the real scene, calculate the corrected identification parameter;

[0016] S6, according to the distribution form of the measurement error, select the corresponding distribution center calculation method to calculate the distribution center of the corrected identification parameter, and take the distribution center as the identification result;

[0017] Wherein:

[0018] When the distribution form is uniform distribution, the arithmetic mean method is used to calculate the distribution center;

[0019] When the distribution form is multimodal distribution, the GMM-EM weighted average method is used to calculate the distribution center;

[0020] When the distribution form is unimodal distribution and symmetric distribution, the 20% truncated mean method is used to calculate the distribution center;

[0021] When the distribution form is unimodal distribution and skew distribution, the Huber estimation method is used to calculate the distribution center.

[0022] Further, in S3, the resistance true value, the reactance true value and the susceptance true value of the simulation scene are respectively R ref1 、 X ref1 ,B ref1 ;

[0023] The identification error calculation formula of the simulation scene is as follows:

[0024] ;

[0025] ;

[0026] ;

[0027] In the formula, R, X, Y respectively represent the resistance identification error, the reactance identification error, and the admittance identification error of the simulation scene at the time point t1 and t2. 、 、 respectively represent the resistance identification error, the reactance identification error, and the admittance identification error of the real scene at the time point t1 and t2. t i 、 、 respectively represent the resistance identification error, the reactance identification error, and the admittance identification error of the real scene at the time point t1 and t2. t i

[0028] Further, in S4, the identification error calculation formula of the simplified real scene is as follows:

[0029] ;

[0030] ;

[0031] ;

[0032] In the formula, R, X, Y respectively represent the resistance identification error, the reactance identification error, and the admittance identification error of the real scene at the time point t1 and t2. 、 、 respectively represent the resistance identification error, the reactance identification error, and the admittance identification error of the real scene at the time point t1 and t2. t i 、 respectively represent the line voltage phase angle difference and the current phase angle difference, f (·) represents the distribution center calculation formula. t i t represents the time point, and the subscripts 1 and 2 respectively represent the simulation scene and the real scene.

[0033] Further, in step S5, the corrected reactance identification parameter is as follows:

[0034] ;

[0035] The corrected admittance identification parameter is as follows: ​​​

[0036] ;

[0037] The corrected resistance identification parameters are as follows:

[0038] ;

[0039] In the formula, 、 、 respectively represent the corrected reactance identification parameter, the susceptance identification parameter, and the resistance identification parameter; 、 、 respectively represent the uncorrected reactance identification parameter, the susceptance identification parameter, and the resistance identification parameter.

[0040] Further, in S4, the calculation method of the distribution center in the simplified real scene identification error calculation formula is:

[0041] According to the identification error distribution form of the simulation scene, the corresponding distribution center calculation method is selected to calculate the identification error of the simulation scene;

[0042] wherein:

[0043] When the distribution form is uniform distribution, the arithmetic mean method is used to calculate the distribution center;

[0044] When the distribution form is multimodal distribution, the GMM-EM weighted average method is used to calculate the distribution center;

[0045] When the distribution form is unimodal distribution and symmetric distribution, the 20% truncated mean method is used to calculate the distribution center;

[0046] When the distribution form is unimodal distribution and skewed distribution, the Huber estimation method is used to calculate the distribution center.

[0047] Due to the adoption of the above technical solutions, the application has the following beneficial effects:

[0048] 1. The application uses high-precision standard sources in a controlled laboratory environment to non-invasively calibrate a single PMU, obtain the measured error data of the voltage amplitude, voltage phase angle and current phase angle, and comprehensively characterize the measurement error distribution characteristics based on statistical indicators (such as mean, standard deviation, skewness, kurtosis, and bimodal coefficient), thereby providing reliable prior information for subsequent parameter identification.

[0049] 2. Construct a true value known simulation scene, combined with a true value unknown real scene. The application establishes an error mapping bridge through the construction of two scenes, that is, after injecting real PMU measurement error in simulation, first quantify the influence of known operating scenes (such as voltage phase angle difference, current phase angle difference) on R / X / B identification error, then migrate the influence to the actual scene, and finally rely only on the measurable operating state quantity (no true value) - uncorrected identification parameters (not available) to obtain usable identification parameters.

[0050] 3. The application discriminates the distribution form of the corrected multi-period identification result sequence. According to the distribution form of the identification result, the optimal center estimation method is selected to ensure that the final output parameter has high robustness and statistical consistency. BRIEF DESCRIPTION OF DRAWINGS

[0051] The drawings described herein are used to provide further understanding of the application, and form a part of the application. The illustrative embodiments of the application and their descriptions serve to explain the application, and do not constitute an improper limitation on the application.

[0052] Figure 1 A line parameter identification method based on non-intrusive PMU prior error analysis and correction is provided for the application;

[0053] Figure 2 A comparison chart of the identification result before and after correction with the true value is provided. DETAILED DESCRIPTION

[0054] The traditional PMU is directly connected to the secondary side output of the CT through the current input channel. When installing, repairing and replacing the PMU, the CT secondary loop terminal must be short-circuited. In order to ensure safety, the primary equipment must be powered off, but this affects the power supply reliability and user power supply, so the PMU cannot be checked regularly. Different PMU error characteristics differ, and the error characteristics of the same PMU will change after long-term operation, so only the error can be assumed to satisfy the normal distribution characteristics. A single distribution characteristic cannot accurately reflect the characteristics of each device, and the error distribution should be analyzed. Based on this, as shown in the accompanying Figure 1 The application provides a line parameter identification method based on non-intrusive PMU prior error analysis and correction, which includes the following steps:

[0055] S1, connect the synchronous standard source to the PMU, and set the parameter value of the synchronous standard source as the standard value. The difference between the parameter measurement value output by the PMU and the standard value is recorded as the measurement error of the PMU.

[0056] High-precision synchronous standard source is a high-precision signal generation and measurement standard source widely used in power systems. In the embodiment, the Omicron company's relay protection tester (CMC series) is connected to the PMU. The high-precision synchronous standard source directly provides a known standard electrical signal (voltage and current) for the PMU to calibrate the measurement error of the PMU itself. In the measurement or calibration scene, the output of the synchronous standard source is regarded as the "theoretical true value", that is, the standard value of the application. The difference between the output result of the PMU measuring the standard value and the standard value is regarded as the measurement error. The measurement error includes inherent error and random error.

[0057] S2, calculate the average value and standard deviation of the measurement error, and obtain the distribution form of the measurement error based on the average value and the standard deviation, the distribution form being described based on kurtosis, skewness, and bimodal coefficient.

[0058] Before calculating the measurement error, outliers should be removed, which refers to observation values significantly deviating from the subject part of the data set. The average value and the standard deviation of the measurement error are used to describe the central tendency and dispersion degree of the error, and the distribution form is described based on kurtosis (Kurt), skewness (Skew), and bimodal coefficient (BC). Specifically,

[0059] ;

[0060] ;

[0061] ;

[0062] In the formula, E [ ] represents a mathematical expectation operator, X represents a value error random variable, μ represents the mean of the error distribution; represents the standard deviation of the error distribution.

[0063] The kurtosis parameter represents the peak state characteristic of the error distribution form, and the greater the value indicates that the heavy-tailed distribution characteristic of the error data is more significant. The skewness parameter represents the skewness characteristic of the error distribution form, and the greater the absolute value indicates that the asymmetry of the error data is stronger. The bimodal coefficient is a multi-peak distribution criterion index, and it is generally considered that when BC≥0.56, it can be determined that the error source has a multi-peak distribution characteristic; when BC<0.56, it can be determined that the error source is a unimodal distribution.

[0064] The measurement error distribution characteristics of the PMU will be different under different operating scenarios. Since the voltage and current change constantly in the actual operation of the power grid, the true values of the voltage and current cannot be determined for a long period of time, and the distribution characteristics of the measurement error cannot be obtained. Therefore, the method of directly correcting the measurement will ignore the influence of random error, and if the random error is large, the accuracy will be significantly reduced. The premise of the implementation of the present application is that the line parameters are generally considered to remain unchanged for a short period of time, and the identification results at all time points within a short period of time can reflect the distribution of identification error.

[0065] S3, calculate the identification error in the simulation scenario and the real scenario, the difference between the simulation scenario and the real scenario being that the identification parameters of the simulation scenario are known, and the identification parameters of the real scenario are unknown.

[0066] The identification parameters in the simulation scenario are known quantities, and the identification parameters in the real scenario are unknown quantities. The resistance identification parameter, the reactance identification parameter, and the admittance identification parameter are respectively denoted as R est1 、 X est1 、 B est1 In implementation, the identification parameters are calculated by the PMU output values with measurement error:

[0067] ;

[0068] Z = R + jX ;

[0069] Y = jB ;

[0070] In the formula, 、 are the line m terminal voltage and current phasors, 、 are the line n terminal voltage and current phasors, m and n represent the two ends of the line; Z is the impedance, Y is the admittance, X is the reactance, R is the resistance, j is the imaginary unit.

[0071] The voltage phasor contains both the amplitude and the phase angle .

[0072] In the simulation scenario, the true value of the resistance, the true value of the reactance, and the true value of the admittance of the simulation scenario are respectively denoted as R ref1 ,X ref1 , B ref1 ;

[0073] The formula for calculating the identification error of the simulation scene is as follows:

[0074] ;

[0075] ;

[0076] ;

[0077] In the formula, , , These represent the simulation scenarios at the [number]th [year]. t i The resistance identification error, reactance identification error, and susceptance identification error at time points; , , These represent the values ​​after adding measurement error on the [number]th [day]. t i Resistance identification parameters, reactance identification parameters, and susceptance identification parameters at specific time points.

[0078] The formula for calculating the recognition error in real-world scenarios is as follows:

[0079] ;

[0080] ;

[0081] ;

[0082] In the formula, , , These represent the measurement errors of voltage amplitude, voltage phase angle, and current phase angle, respectively. This indicates the voltage phase angle difference between the two ends of the line. This indicates the phase angle difference of the currents at both ends of the line; the superscripts inh and rnd in the parameters represent inherent error and random error, respectively, and the subscripts... t i Indicates a point in time.

[0083] Based on the principle that resistance is much smaller than reactance, the ratio of resistance to reactance in the formula for mapping the identification error between real and simulated scenarios is simplified to approximately 0; that is, in the above formula... , For both items, when the resistance is much smaller than the reactance, the calculated value is 0.

[0084] Considering that the distribution of two times of identification will tend to be stable when the number of identification is enough, so as to offset the influence of random error; the inherent error of PMU has the characteristic of short-time invariance, so the influence on the identification result is also short-time invariance. Therefore, based on the law of large numbers, the random error term in the mapping relationship formula of identification error between real scene and simulation scene is divided; based on the short-time invariance principle of inherent error, the inherent error term in the mapping relationship formula of identification error between real scene and simulation scene is divided.

[0085] Based on the above, the identification error calculation formula of real scene is as follows:

[0086]

[0087]

[0088]

[0089] In the formula, and respectively represent the resistance identification error, the reactance identification error and the susceptance identification error of real scene at the time point; respectively represent the voltage phase angle difference and the current phase angle difference of line two ends, t i f (·) represents the distribution center calculation formula; t i represents the time point, and the subscripts 1 and 2 respectively represent the simulation scene and the real scene.

[0090] In the above formula, the distribution center of identification error in simulation scene is used to replace the specific value of identification error, and the calculation method of distribution center is the same as that of measurement error distribution center in S6.

[0091] In S4, the calculation method of distribution center in the simplified identification error calculation formula of real scene is:

[0092] According to the distribution form of identification error in simulation scene, the corresponding distribution center calculation method is selected to calculate the identification error in simulation scene;

[0093] In the formula, and respectively represent the resistance identification error, the reactance identification error and the susceptance identification error of real scene at the time point;

[0094] When the distribution form is uniform distribution, the arithmetic mean method is used to calculate the distribution center;

[0095] When the distribution form is multi-peak distribution, the GMM-EM weighted average method is used to calculate the distribution center;

[0096] ​​​​​​​​When the distribution form is a single-peak distribution and is a symmetric distribution, the distribution center is calculated by using a 20% truncated mean method;

[0097] When the distribution form is a single-peak distribution and is a skewed distribution, the distribution center is calculated by using a Huber estimation method.

[0098] S5, based on the simplified real scene recognition parameter calculation formula, calculate the corrected recognition parameter;

[0099] In step S5, the corrected reactance recognition parameter is as follows:

[0100] ;

[0101] The corrected susceptance recognition parameter is as follows:

[0102] ;

[0103] The corrected resistance recognition parameter is as follows:

[0104] ;

[0105] In the formula, 、 、 respectively represent the corrected reactance recognition parameter, the corrected susceptance recognition parameter, and the corrected resistance recognition parameter; 、 、 respectively represent the uncorrected reactance recognition parameter, the uncorrected susceptance recognition parameter, and the uncorrected resistance recognition parameter. The uncorrected recognition parameter is the recognition parameter output by the actual PMU without error compensation. After the PMU outputs the recognition parameter, each recognition parameter is corrected according to the above formula.

[0106] S6, according to the distribution form of the measurement error, select the corresponding distribution center calculation method to calculate the distribution center of the corrected recognition parameter, and take the distribution center as the recognition result;

[0107] Wherein:

[0108] When the distribution form is a uniform distribution, the arithmetic mean method is used to calculate the distribution center;

[0109] When the distribution form is a multi-peak distribution, the GMM-EM weighted average method is used to calculate the distribution center;

[0110] When the distribution form is a single-peak distribution and is a symmetric distribution, the distribution center is calculated by using a 20% truncated mean method;

[0111] When the distribution form is a single-peak distribution and is a skewed distribution, the distribution center is calculated by using a Huber estimation method.

[0112] The application is verified on a 20km short line, and the identification values before and after correction are shown in Table 1.

[0113] Table 1 Comparison of identification values before and after correction of 20km short line

[0114]

[0115] As shown in Table 1 and the attached Figure 2 simulation results show that after error correction, the identification accuracy is significantly improved, and all errors are stably controlled below 5%, which meets the strict requirements of power system planning, fault analysis, operation simulation and other practical engineering applications on line parameters.

[0116] The places not mentioned in the application can be realized by using or referring to the existing technology.

[0117] The above only describes the embodiments of the application and is not used to limit the application. For those skilled in the art, the application can have various changes and variations. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the application shall be included in the scope of claims of the application.

Claims

1. A method for identifying line parameters based on non-intrusive PMU prior error analysis and correction, characterized in that, Includes the following steps: S1. Connect the synchronization standard source to the PMU, use the parameter setting value of the synchronization standard source as the standard value, and record the difference between the parameter measurement value output by the PMU and the standard value as the measurement error of the PMU. S2. Calculate the mean and standard deviation of the measurement error, and obtain the distribution pattern of the measurement error based on the mean and standard deviation. The distribution pattern is described based on kurtosis, skewness, and bimodal coefficient. S3. Based on the PMU output value with measurement error, calculate the identification error in the simulation scene and the real scene. The difference between the simulation scene and the real scene is that the identification parameters of the simulation scene are known, while the identification parameters of the real scene are unknown. S4. Based on the principle that resistance is much smaller than reactance, the ratio of resistance to reactance in the formula for the identification error mapping relationship between real and simulated scenarios is simplified to approximately 0. Based on the law of large numbers, the random error term in the formula for the mapping relationship between the identification error of real scene and simulation scene is simplified. Based on the principle that the inherent error remains unchanged in the short time, the inherent error term in the formula for the mapping relationship between the identification error of the real scene and the simulation scene is simplified. S5. Calculate the corrected identification parameters based on the simplified formula for calculating identification parameters in the real scene; S6. Based on the distribution pattern of the measurement error, select the corresponding distribution center calculation method to calculate the distribution center of the corrected identification parameters, and use the distribution center as the identification result. in: When the distribution is uniform, the arithmetic mean method is used to calculate the distribution center; When the distribution pattern is multimodal, the GMM-EM weighted average method is used to calculate the distribution center; When the distribution is unimodal and symmetrical, the distribution center is calculated using the 20% truncated mean method. When the distribution is unimodal and skewed, the Huber estimation method is used to calculate the distribution center.

2. The line parameter identification method based on non-intrusive PMU prior error analysis and correction as described in claim 1, characterized in that, In S3, let the true values ​​of resistance, reactance, and susceptance in the simulation scenario be denoted as follows: R ref1 , X ref1 , B ref1 ; The formula for calculating the identification error of the simulation scene is as follows: ； ； ； In the formula, , , These represent the simulation scenarios at the [number]th [year]. t i The resistance identification error, reactance identification error, and susceptance identification error at time points; , , These represent the values ​​after adding measurement error on the [number]th [day]. t i Resistance identification parameters, reactance identification parameters, and susceptance identification parameters at specific time points.

3. The line parameter identification method based on non-intrusive PMU prior error analysis and correction as described in claim 2, characterized in that, In S4, the simplified formula for calculating the recognition error of the real scene is as follows: ； ； ； In the formula, , , These represent the real-world scenarios at the [number]th [time]. t i The resistance identification error, reactance identification error, and susceptance identification error at time points; , These represent the voltage phase angle difference and the current phase angle difference at both ends of the line, respectively. f (·) represents the formula for calculating the distribution center; t i The subscripts 1 and 2 represent the simulation scene and the real scene, respectively.

4. The line parameter identification method based on non-intrusive PMU prior error analysis and correction as described in claim 3, characterized in that, In step S5, the corrected reactance identification parameters are as follows: ； The corrected susceptance identification parameters are as follows: ； The corrected resistor identification parameters are as follows: ； In the formula, , , These represent the corrected reactance identification parameters, susceptance identification parameters, and resistance identification parameters, respectively. , , These represent the uncorrected reactance identification parameters, susceptance identification parameters, and resistance identification parameters, respectively.

5. The line parameter identification method based on non-intrusive PMU prior error analysis and correction according to claim 3, characterized in that, In S4, the calculation method for the distribution center in the simplified formula for calculating the recognition error of the real scene is as follows: Based on the distribution pattern of the identification error in the simulation scene, the corresponding distribution center calculation method is selected to calculate the identification error of the simulation scene; in: When the distribution is uniform, the arithmetic mean method is used to calculate the distribution center; When the distribution pattern is multimodal, the GMM-EM weighted average method is used to calculate the distribution center; When the distribution is unimodal and symmetrical, the distribution center is calculated using the 20% truncated mean method. When the distribution is unimodal and skewed, the Huber estimation method is used to calculate the distribution center.

Images