Research on multiple calibration methods for measuring error of indicator value of withstand voltage tester

Voltage resistance testing often needs to be carried out under specific environmental conditions, such as high pressure and high temperature. These complex environments may interfere with the test equipment and calibration process, thus affecting the accuracy of calibration results. Therefore, it is necessary to ensure that the tester can adapt to various environmental conditions and make calibration adjustments accordingly. Therefore, a multi-calibration method based on the BP neural network is proposed to calibrate the error of the indicator value of the voltage tester. The power input and output characteristic function of the withstand voltage tester is established, and the output of each layer is described as the input of the BP neural network. Based on the adjustment results, multiple calibration steps for the measuring error of the voltage tester are designed. The experimental results show that the method can calibrate the measuring error of the voltage tester at different positions at the same time, and the application effect is good.


Introduction
Dielectric strength is an important index to evaluate the electrical safety of electrical equipment.The key parameters to evaluate the dielectric properties are the level of the test voltage applied to the electrical equipment, the size of the leakage current, and the length of the test voltage holding (duration).The voltage tester is a special instrument for dielectric strength tests [1].When conducting dielectric strength testing, the output voltage of the tester should rise steadily from zero to a specified value at a certain rate, and then maintain it for a specified period.In power systems, high-voltage equipment and electrical equipment often need to be tested for voltage resistance to ensure their quality and reliability.The voltage resistance tester is widely used to test and evaluate the insulation performance of electrical equipment.However, to ensure real and reliable test results, the voltage tester needs to provide accurate indication value measurement.The measurement error of the indicator value of the withstand voltage tester is directly related to the determination of the insulation state and the evaluation of the quality of electrical equipment [2].Therefore, to meet the needs of practical engineering applications, it is necessary to calibrate the error of the voltage tester to improve its measurement accuracy.
With the continuous progress of science and technology and the development of power systems, the accuracy requirements of voltage-resistant instruments are also constantly improving.Therefore, the study of the error and calibration of the voltage resistance tester can provide a more accurate and reliable insulation performance evaluation tool for the power industry.

Research on multiple calibration methods for measuring error of voltage tester indication value
We set the power input and output characteristics of the voltage tester: ( ) where  is the error compensation parameter affecting the measurement of the instrument.Although the power input and output of the instrument are expected to be linear [3], F in Formula (1) is a nonlinear function, and to compensate for it, the measured output of the withstand voltage tester is taken as input to the following functions: According to the formula, BP neural network is a widely used multi-layer feed-forward network.Its learning algorithm is a weight optimization process built on the premise of error gradient decline, which is composed of forward propagation and back propagation [4][5].The structure of BP neural network is shown in Figure 1.Based on the basic principle of neural network [6][7], the output values of each layer are described as follows: ( ) Several training samples are used to train the network.In each training time, the weight coefficient will change to different degrees.The correction formula for the weight coefficient from the middle layer to the output layer is as follows: where  represents the learning rate, which determines the weight change obtained in each cycle of training.Too high a learning rate is not conducive to the stability of the network, and too small a learning rate will increase the training time [9][10].We design an adaptive learning rate and check the weight correction value to determine the learning rate.The formula for adjusting the indication value of the instrument error is: Based on the adjustment of the adaptive learning rate based on the BP neural network, the detailed steps of multiple calibration of error in the static environment of the design voltage tester are as follows: ① The distance between the full components of the voltage tester is equal; ② The points are sorted from large to small, increasing the input standard value in turn, and recording the output value of each point; then the input standard value is reduced successively, and the output value of each point is recorded.
③ According to process (2), the sensor is repeatedly carried out forward and reverse stroke, and the input and output data are automatically drawn into curves; ④ The data obtained from (3) is calculated by Formula (9) to complete the calibration of the indicated value.

Experimental process and result analysis
To verify the error multiple calibration performance of the proposed method, the experiment was carried out in the Windows 10 system.The programming language was SQL, and three voltage resistance testers installed at different positions in the circuit were selected as the research objects.Table 1 shows the ideal values of "S1", "S2", and "S3", three voltage tester components.From the analysis of Table 1, it can be seen that in the whole experiment process, the indicator level of "S2" voltage withstand tester is slightly higher than that of "S1" voltage withstand tester, but the combination of "S1" voltage withstand tester indicator value and "S2" voltage withstand tester indicator value cannot be completely equal to "S3" voltage withstand tester indicator number, which means that in the grid platform environment, there are always some uncontrollable power consumption behaviors; the indicator level of "S3" voltage tester is relatively high.The result of "S1" withstand voltage tester is shown in Figure 2. It can be seen from the analysis of Figure 2 that the "S1" voltage resistance tester of the experimental 200, 300, and 500 is equal to the ideal value of the "S1" voltage resistance tester.Although the numerical results of Group 1, Group 4, and Group 6 are lower than the ideal numerical values, the difference between them is small.The value of the control group "S1" voltage tester was always lower than the ideal value of the "S1" voltage tester, and the difference between the two was always relatively large.The result of "S2" withstand voltage tester is shown in Figure 3.The analysis of Figure 3 shows that although the indicative value level of the "S2" withstand voltage tester in the experimental group is always lower than the ideal indicative value level of the "S2" withstand voltage tester, the difference between the two is very small.In the whole process of the experiment, the indicator level of the "S2" voltage resistance tester in the control group was always lower than the ideal indicator level of the "S2" voltage resistance tester, but the difference between the two was significantly greater.
Figure 4 reflects the values of "S1", "S2", and "S3" of the three voltage resistance tester components under the action of the calibration methods of the experimental group and the control group.According to the analysis of Figure 4, except for the experimental results of the fourth group, the indicating value of the "S3" voltage resistance tester in the control group was always lower than the ideal indicating value of the "S3" voltage resistance tester; the indicative values of the test group "S3" voltage resistance tester are exactly equal to the ideal indicative values of the "S3" voltage resistance tester.

Conclusion
To optimize the application effect of the voltage resistance tester, a multi-calibration method based on the BP neural network is proposed.By using the multilevel computing power of the BP neural network, the output of each layer is used as the input of the next layer.The adjustment formula of instrument error indication value is established.In this process, the nonlinear modeling ability of the neural network can be used to describe the error characteristics of the voltage tolerance tester more accurately, and further improve the calibration accuracy.In this method, according to the results of the adjustment formula, a series of calibration steps for the measurement errors of indicated values are designed.These steps use different calibration parameters and adjustment strategies to gradually optimize the accuracy of the indication measurement of the test equipment.Through the combined application of multiple calibration steps, the measurement error of indication value can be effectively reduced and the calibration effect can be improved.

Figure 1 .
Figure 1.Structure of BP neural network.

X
represents the input value of the i node of the input layer, j Y is the output value of the j node of the middle layer, n Z is the output value of the n node of the input layer, ij  is the weight of the i node of the input layer to the j node of the middle layer, jk  is the weight coefficient of the j node of the middle layer to the n node of the output layer, and j  represents the internal critical value of the j node of the middle layer.n  is the internal critical value of the n -th node of the output layer[8].If the expected value of the output from the n -th node of the output layer is n d , then the network error is:

Table 1 .
Ideal values of voltage resistance tester components.