阅读说明：本技术 一种高电压试验中大气校正因数的计算方法及装置 (Method and device for calculating atmospheric correction factor in high-voltage test ) 是由 刘宸 李强 申萌 苏春强 于 2019-01-17 设计创作，主要内容包括：本发明公开了一种高电压试验中大气校正因数的计算方法及装置,使用空气流注传播速度代替了IEC 60060-1：2010标准中试验电压类型这一参数,实现了将试验电压类型由非连续变量转换为连续变量的空气流注传播速度,使得影响大气校正因数的所有参数均为连续变量,从而可以采用神经网络模型对大气校正因数的计算过程进行拟合,得到大气校正因数确定模型,通过将高电压试验中的空气流注传播速度、试验电压值、最短绝缘距离、大气压力、大气温度和相对空气湿度输入至大气校正因数确定模型,得到对应的大气校正因数。采用本发明无需考虑空气湿度与空气密度的比值的取值范围,因此可以计算当h/δ超出标准给定范围时的大气校正因数。(The invention discloses a method and a device for calculating an atmospheric correction factor in a high-voltage test, which use the air stream propagation speed to replace IEC60060-1:2010 standard test voltage type is a parameter, so that the test voltage type is converted from a discontinuous variable into a continuous variable of the air stream propagation speed, all parameters influencing the atmospheric correction factor are continuous variables, a neural network model can be adopted to fit the calculation process of the atmospheric correction factor to obtain an atmospheric correction factor determination model, and the corresponding atmospheric correction factor is obtained by inputting the air stream propagation speed, the test voltage value, the shortest insulation distance, the atmospheric pressure, the atmospheric temperature and the relative air humidity in a high-voltage test into the atmospheric correction factor determination model. By adopting the method and the device, the value range of the ratio of the air humidity to the air density does not need to be considered, so that the atmospheric correction factor when h/exceeds the standard given range can be calculated.)

1. A method for calculating an atmospheric correction factor in a high voltage test is characterized by comprising the following steps:

acquiring a first test parameter in a high voltage test, wherein the first test parameter comprises: the test voltage type, the test voltage value, the shortest insulation distance, the atmospheric pressure, the atmospheric temperature and the relative air humidity;

converting the test voltage type into an air stream propagation speed under the test voltage type;

determining the air stream propagation speed, the test voltage value, the shortest insulation distance, the atmospheric pressure, the atmospheric temperature and the relative air humidity as second test parameters, and inputting the second test parameters as input values of an atmospheric correction factor determination model into the atmospheric correction factor determination model to obtain an atmospheric correction factor;

the atmosphere correction factor determination model is obtained by training with the second test parameters of different high-voltage electrical appliance products as training samples and the atmosphere correction factors corresponding to the second test parameters as sample labels in high-voltage tests with different voltage grades.

2. The calculation method according to claim 1, wherein the converting the test voltage type into the air flow propagation velocity under the test voltage type specifically comprises:

and obtaining the air flow propagation speed under the test voltage type through an air gap discharge test.

3. The calculation method according to claim 1, wherein the process of training the atmospheric correction factor determination model specifically comprises:

in the high-voltage tests with different voltage grades, the first test parameters of different high-voltage electrical products adopt IEC60060-1:2010 calculating atmospheric correction factors under different voltage levels;

converting the test voltage type in the first test parameter of each high-voltage electrical appliance product into a corresponding air flow propagation speed to obtain a second test parameter of each high-voltage electrical appliance product;

establishing a neural network model containing two hidden layers as an initial atmospheric correction factor determination model, wherein the activation function of neurons of the hidden layers adopts a Re L U activation function;

and in the high-voltage tests with different voltage grades, the second test parameters of different high-voltage electrical appliance products are used as training samples, the atmospheric correction factors corresponding to the second test parameters are used as sample labels, and the initial atmospheric correction factor determination model is trained by utilizing a gradient descent method to obtain the atmospheric correction factor determination model.

4. The computing method of claim 3, further comprising:

and correcting the atmospheric correction factor determination model by adopting a loss function to obtain a corrected atmospheric correction factor determination model.

5. The computing method of claim 4, wherein the loss function comprises: mean square error function, the expression of the loss function COST is as follows:

in the formula, n is the total number of training samples,for atmospheric correction factor, Y, calculated using said atmospheric correction factor-determining model_t ⁱFor sampling the IEC60060-1:2010 standard calculated atmospheric correction factor, i ith training sample.

6. An apparatus for calculating an atmospheric correction factor in a high voltage test, comprising:

the acquisition unit is used for acquiring a first test parameter in a high voltage test, and the first test parameter comprises: the test voltage type, the test voltage value, the shortest insulation distance, the atmospheric pressure, the atmospheric temperature and the relative air humidity;

the conversion unit is used for converting the test voltage type into the air stream propagation speed under the test voltage type;

the input unit is used for determining the air stream propagation speed, the test voltage value, the shortest insulation distance, the atmospheric pressure, the atmospheric temperature and the relative air humidity as second test parameters, inputting the second test parameters as input values of an atmospheric correction factor determination model to the atmospheric correction factor determination model, and obtaining an atmospheric correction factor;

the atmosphere correction factor determination model is obtained by training with the second test parameters of different high-voltage electrical appliance products as training samples and the atmosphere correction factors corresponding to the second test parameters as sample labels in high-voltage tests with different voltage grades.

7. The computing device of claim 6, wherein the conversion unit is specifically configured to:

and obtaining the air flow propagation speed under the test voltage type through an air gap discharge test.

8. The computing device of claim 6, further comprising: the model training unit is used for training to obtain the atmosphere correction factor determination model, and specifically comprises:

and the calculating subunit is used for calculating the first test parameters of different high-voltage electrical products in the high-voltage tests with different voltage grades by adopting IEC60060-1:2010 calculating atmospheric correction factors under different voltage levels;

the conversion subunit is used for converting the test voltage type in the first test parameter of each high-voltage electrical product into a corresponding air stream propagation speed to obtain the second test parameter of each high-voltage electrical product;

the model establishing subunit is used for establishing a neural network model containing two hidden layers as an initial atmosphere correction factor determining model, and the activation function of the neurons of the hidden layers adopts a Re L U activation function;

and the training subunit is used for taking the second test parameters of different high-voltage electrical appliance products as training samples in high-voltage tests with different voltage grades, taking the atmospheric correction factors corresponding to the second test parameters as sample labels, and training the initial atmospheric correction factor determination model by using a gradient descent method to obtain the atmospheric correction factor determination model.

9. The computing device of claim 8, wherein the conversion unit further comprises:

and the corrector subunit is used for correcting the atmospheric correction factor determination model by adopting a loss function to obtain a corrected atmospheric correction factor determination model.

10. The computing device of claim 9, wherein the loss function comprises: mean square error function, the expression of the loss function COST is as follows:

in the formula, n is the total number of training samples,for atmospheric correction factor, Y, calculated using said atmospheric correction factor-determining model_t ⁱFor sampling the IEC60060-1:2010 standard calculated atmospheric correction factor, i ith training sample.