CN110967661A

CN110967661A - Electrical data calibration method based on curve fitting

Info

Publication number: CN110967661A
Application number: CN201911327267.0A
Authority: CN
Inventors: 张衍奎; 王川; 朱建华; 鲍伟; 谢秉旺; 闫亮; 马伟伟
Original assignee: Ningxia Kaichen Electric Group Co ltd
Current assignee: Ningxia Kaichen Electric Group Co ltd
Priority date: 2019-12-20
Filing date: 2019-12-20
Publication date: 2020-04-07

Abstract

The invention relates to an electrical data calibration method based on curve fitting. The method is characterized by comprising the following steps: (1) collecting a measured value and an actual value of the electrical measurement data, and calculating an error value; (2) performing visualization processing on the obtained measurement value and the error value, and defining an x axis as the measurement value and a y axis as the error value; (3) selecting a proper model according to the obtained image trend; (4) calculating various parameters of the model according to the error minimum principle; (5) calculating a residual error according to the original data brought by the model, analyzing whether the residual error meets the requirement, and returning to the step (3) to reselect a higher-order model if the residual error does not meet the requirement; (6) and if the residual error meets the requirement, calibrating the measured value according to the model, and taking the original measured value of the model as the input quantity of the model, wherein the output quantity is the calibrated value. The method can solve the precision problem caused by the nonlinear error caused by the difference of sensor hardware.

Description

Electrical data calibration method based on curve fitting

Technical Field

The invention relates to an electrical data calibration method based on curve fitting.

Background

High-voltage complete equipment (high-voltage distribution cabinet) refers to indoor and outdoor alternating-current switch equipment which operates in a power system with voltage of 3kV or above and frequency of 50Hz or below. The power system protection device is mainly used for controlling and protecting power systems (including users of power plants, substations, power transmission and distribution lines, industrial and mining enterprises and the like), can be used for not only putting a part of power equipment or lines into or out of operation according to the operation requirements of a power grid, but also quickly cutting off the fault part from the power grid when the power equipment or lines have faults, thereby ensuring the normal operation of the fault-free part in the power grid and the safety of equipment and operation maintenance personnel. In the whole link, voltage and current are required to be collected from high-voltage electricity (3KV and above) for judgment, the voltage and the current on the high-voltage side are called primary data, the data is subjected to transformation ratio through a mutual inductor, the transformation ratio of the current (CT transformation ratio) is generally 2000 times, and the transformation ratio of the voltage (PT transformation ratio) is generally 100 times. The transformed data is referred to as secondary data.

The error value of the secondary data is amplified by the transformation ratio of the transformer, so that the precision of the secondary value is particularly important, and the comprehensive protection algorithm and the measurement algorithm are all based on the secondary value data, so the quality and the safety of the electrical equipment and the comprehensive protection equipment are influenced by the precision of the electrical data, and the quality of the power grid is also concerned.

Disclosure of Invention

The invention aims to provide an electrical data calibration method based on curve fitting, which can improve the accuracy of secondary values of electrical data as much as possible.

A method for calibrating electrical data based on curve fitting is characterized by comprising the following steps:

(1) collecting a measured value and an actual value of the electrical measurement data, and calculating an error value;

(2) performing visualization processing on the obtained measurement value and the error value, and defining an x axis as the measurement value and a y axis as the error value;

(3) selecting a proper model according to the obtained image trend;

(4) calculating various parameters of the model according to the error minimum principle;

(5) calculating a residual error according to the original data brought by the model, analyzing whether the residual error meets the requirement, and returning to the step (3) to reselect a higher-order model if the residual error does not meet the requirement;

(6) and if the residual error meets the requirement, calibrating the measured value according to the model, taking the original measured value of the model as the input quantity of the model, wherein the output quantity is the calibrated value, and ending the whole calibration process of the electrical data.

The electrical measurement data is specifically current, voltage, electrical power, electrical energy, phase, frequency, resistance, capacitance or inductance.

The method has the beneficial effects that: 1. the accuracy problem caused by non-linear errors due to differences in sensor hardware can be solved. 2. The accuracy of the secondary values of the electrical data can be improved as much as possible.

Drawings

FIG. 1 is a graph of the error of the A-phase voltage versus the measured voltage;

FIG. 2 is a fitting curve and an error curve of the A-phase voltage;

FIG. 3 is a graph of the fitting of the A-phase voltage and the residual after the fitting of the A-phase voltage.

Detailed Description

The invention relates to a solution for electric data requiring high precision, which is based on the minimum error theory, and different models can be selected to match the error of the whole precision, thereby establishing a set of voltage and current precision fitting compensation method system suitable for the electric industry. The device is suitable for microcomputer comprehensive protection and enterprise units needing high-precision current and voltage measurement.

The invention aims to provide a method for compensating the accuracy of the original value of the electrical data, which is mainly based on the two ideas that the sum of all errors of the created model and the electrical data value is minimum or the sum of the squares of the errors is minimum. The sample regression function has the smallest total error of the values of the explained variables from the known sample data point values, i.e. in the coordinate graph the perpendicular distance of the points of all the explained variables on the fitting function from the known sample data points is smallest with the same value of the explained variable.

The principle of the method is as follows:

for a given set of data points { (x)_i，y_i): if the curve model is fitted, i is 1, 2, 3, …, n

The error distance at the i-th position is

Then the fitted curve model value F (x) at the i-th position_i) With the actual given data point Y_iThe sum of the squares of the differences is

Further, find out

To obtain a fitted curve

The method firstly determines a model for fitting a curve, such as a linear model, a quadratic model, a cubic model, a fourth-order polynomial, a fifth-order polynomial and the like. After the model is determined, the parameters of the selected model are determined according to the concept of minimum error. After obtaining parameters of the model, the original data to be fitted is taken as a variable of the model to be input, and the obtained output is the data after fitting.

Regarding the principle of linear fitting, for a given data point (x)_i，y_i): 1, 2, 3, …, n is obtained as a linear equation of y ═ a + bx, with the total error being calculated

To minimize, the parameters a, b that make Q extremely good should be satisfied

This is true.

The parameters of the linear model are then found from the set of equations above.

With respect to the principle of polynomial fitting, sometimes it is not appropriate to fit a given data point with a straight line, and the use of polynomial fitting is considered. For a given data point (x)_i，y_i): 1, 2, 3, …, n is obtained as a polynomial of degree m (m < n)

Make the total error

Is minimal. Since Q can be considered as relating to a_j(j is 0, 1, …, m), so the problem of fitting a polynomial can be reduced to the problem of the extremum of the multivariate function.

To obtain

Namely, the coefficient a is obtained_jThe system of linear equations is commonly referred to as a system of regular equations.

Example 1:

in a first step, measured and actual values of the electrical data are collected and an error value is calculated.

The first step of the method is to collect electrical data meeting the requirements, and the mode of collecting the electrical data can refer to the following scheme that high voltage and high current, namely voltage and current of a primary side are transmitted to a secondary transformer after being transformed by the transformer, 3kv and above are converted into 110V, 10kA is converted into a range of 5A, the alternating current data is sampled by a high-precision AD sampler, the effective value of the sampled data is calculated by a fourier transform or root mean square method, and a data interval required to be subjected to precision compensation is subjected to value taking, wherein the smaller the value step length of the interval is, the higher the compensation precision is. The error values obtained by subtracting the measured values from the actual values are shown in the following table.

And secondly, performing visualization processing on the data: the data is visualized, the horizontal axis (x) is measurement data, the vertical axis (y) is an error value, and x (after compensation) is x + y; and f (x), we need to find f (x) in the following, see fig. 1 for details.

Thirdly, selecting a proper function model and calculating model parameters:

an appropriate functional model is selected based on the error and the graphical trend of the measurement data, where we select a linear model, and we obtain the parameters of the model based on the principle of error minimization and based on the inventive principles described above. The linear model is y ═ 0.089 x-0.18. This is visualized as shown in the following figure, which is described in detail in figure 2.

Fourthly, calculating a residual error, and judging whether the requirement is met:

and calculating a residual error between the compensation curve and the error curve, wherein the residual error is the error fluctuation of the compensation curve, and the error fluctuation after compensation is basically within 0.2V, if the error fluctuation does not meet the requirement, performing high-order model compensation, and particularly referring to an attached figure 3.

Claims

1. A method for calibrating electrical data based on curve fitting is characterized by comprising the following steps:

(3) selecting a proper model according to the obtained image trend;

2. A method of electrical data calibration based on curve fitting according to claim 1, wherein: the electrical measurement data is specifically current, voltage, electrical power, electrical energy, phase, frequency, resistance, capacitance or inductance.