阅读说明：本技术 一种变压器励磁电流特征量在线获取方法 (Online acquisition method for transformer exciting current characteristic quantity ) 是由 牛唯 刘君 曾鹏 欧阳泽宇 谈竹奎 马春雷 曾华荣 马晓红 张迅 陈沛龙 田承越 于 2019-09-23 设计创作，主要内容包括：本发明公开了一种变压器励磁电流特征量在线获取方法,该方法包括步骤：A、变压器中性点直流电流与励磁电流特征量样本数据的实验获取；B、神经网络的训练；C、变压器励磁电流特征量的在线获取。本发明能够通过在线测量变压器中性点电流的直流分量,利用神经网络算法拟合计算变压器励磁电流直流分量、最大值、最小值和总谐波畸变率四类特征量,可得到清晰的变压器励磁电流特征量与其中性点直流电流的关系,进而可为变压器直流偏磁的分析和抑制提供数据支持和指导,具有较高的工程应用价值。(The invention discloses a transformer exciting current characteristic quantity online obtaining method, which comprises the following steps: A. acquiring sample data of characteristic quantities of direct current and exciting current of a neutral point of a transformer through an experiment; B. training a neural network; C. and (4) online acquisition of the transformer exciting current characteristic quantity. The method can calculate the four characteristic quantities of the direct-current component, the maximum value, the minimum value and the total harmonic distortion rate of the exciting current of the transformer by fitting through the neural network algorithm by measuring the direct-current component of the neutral point current of the transformer on line, can obtain a clear relation between the characteristic quantities of the exciting current of the transformer and the direct-current component of the neutral point of the transformer, can further provide data support and guidance for analysis and inhibition of direct-current magnetic biasing of the transformer, and has high engineering application value.)

1. A transformer exciting current characteristic quantity online obtaining method is characterized by comprising the following steps: the method comprises the following steps:

A. obtaining sample data of characteristic quantities of direct current and exciting current of a neutral point of a transformer through an experiment:

a1, performing no-load test on the transformer, and injecting direct current i changing along with time into the neutral point of the transformer₁The duration is t seconds; real-time acquisition of transformer exciting current i in experiment₂(ii) a Wherein the injection frequency of the direct current is f₁Excitation current i of transformer₂Has a sampling frequency of f₂；

A2, mixing i₁In the amount of data m contained per 0.04 second₁＝0.04f₁Segmenting at intervals to obtain current sample data of neutral points of n segments of transformers, wherein the current sample data is i₁(k)、i₂(k)、…、i_n(k) (ii) a Wherein k is 1, 2, …, m₁The sampling point serial number of each segment of data;

a3, mixing i₂In the amount of data m contained per 0.04 second₂＝0.04f₂Segmenting at intervals to obtain n segments of transformer exciting current sample data; respectively calculating the DC component of the excitation current sample data of each section of transformer to obtain the excitation current DC component sample data i₂(1)_dc、i₂(2)_dc、…、i₂(n)_dc(ii) a Respectively calculating the maximum value of the sample data of the excitation current of each section of the transformer to obtain sample data of the maximum value of the excitation current, i₂(1)_max、i₂(2)_max、…、i₂(n)_max(ii) a Respectively calculating the minimum value of the sample data of the excitation current of each section of the transformer, and respectively obtaining sample data of the minimum value of the excitation current as i₂(1)_min、i₂(2)_min、…、i₂(n)_min(ii) a Respectively calculating the total harmonic distortion rate of the excitation current sample data of each section of transformer to obtain the sample data of the total harmonic distortion rate of the excitation current, i₂(1)_THD、i₂(2)_THD、…、i₂(n)_THD；

B. Training a neural network:

b1, taking the current sample data of the neutral point of the transformer and the DC component sample data, the maximum sample data, the minimum sample data and the total harmonic distortion sample data of the exciting current of the transformer in the same 0.04 second time period as a training sample T of the neural network_rain(j)＝[i_j(1)，i_j(2)，…，i_j(k)，i₂(j)_dc，i₂(j)_max，i₂(j)_min，i₂(j)_THD]Wherein j is 1, 2, …, n; n training samples in n periods form a training sample set T of the neural network_rain＝[T_rain(1)；T_rain(2)；…；T_rain(n)]；

B2 setting node number p of input layer of neural network₁Equal to the sample data quantity of current at neutral point of a section of transformer, i.e. p₁＝m＝0.04f₁(ii) a Setting the number p of nodes of an output layer₂Number of types equal to characteristic quantity of transformer exciting current, i.e. p₂4; setting the number of hidden layers as 1 and the number of nodes of the hidden layers as

b3 training sample set T_rainLeading the data into a neural network, training the neural network by using a back propagation algorithm, finishing the training when the training error is less than 0.001 or the iteration number of the training is more than 50000, and storing the trained neural network;

C. online acquisition of transformer exciting current characteristic quantity:

at a sampling frequency f₃Real-time measuring transformer neutral point current in practical engineering and extracting its DC component as i₁ ^*According to the data window N₁＝0.04f₃With N₂＝0.001f₃For spacing slidably move i₁ ^*Data i in_q ^*(z) inputting the input layer of the trained neural network to obtain the direct current component i of the transformer exciting current₂ ^*(q)_dcMaximum value i₂ ^*(q)_maxMinimum value i₂ ^*(q)_minAnd the total harmonic distortion rate i₂ ^*(q)_THD4-type feature quantity; wherein z is 1, 2, …, N₁The sampling point serial number of the data in each data window is used; q is the number of the sliding times of the data window, and q is 1, 2, ….

2. The method for acquiring the characteristic quantity of the transformer exciting current on line according to claim 1, wherein the method comprises the following steps: and D, the type and the structural parameters of the transformer used in the air-load experiment in the step A are consistent with those of the transformer of the actual engineering in the step C.

3. The method for acquiring the characteristic quantity of the transformer exciting current on line according to claim 1, wherein the method comprises the following steps: the number n of training samples used for neural network training in the step B is not less than 500.

4. The method for acquiring the characteristic quantity of the transformer exciting current on line according to claim 1, wherein the method comprises the following steps: the range of the amplitude variation of the direct current injected into the neutral point of the transformer in the step A1 is consistent with the range of the amplitude variation of the direct current component of the neutral point of the transformer in the actual engineering, and the range of the amplitude variation varies from-100A to 100A.

5. The method for acquiring the characteristic quantity of the transformer exciting current on line according to claim 1, wherein the method comprises the following steps: step A1, the injection frequency f of the DC signal injected into the neutral point of the transformer₁Exciting current i of transformer₂Sampling frequency f₂And C, sampling frequency f of the current of the neutral point of the transformer for practical engineering in the step C₃Are equal.

Technical Field

The invention belongs to the technical field of power system detection, and particularly relates to an online acquisition method for a transformer exciting current characteristic quantity.

Background

When urban rail transit operation, single-pole operation of a high-voltage direct-current transmission system or a ground magnetic storm phenomenon occurs, a neutral point of a transformer can enter direct current, so that the exciting current of the transformer is distorted, namely the transformer generates a direct-current magnetic bias phenomenon. Practical operation cases show that the direct current magnetic biasing of the transformer mainly has the following hazards:

(1) the direct current magnetic biasing can enable the transformer to work in an oversaturated state, so that deformation of an iron core of the transformer is caused, and vibration of the transformer is caused to generate noise.

(2) The output current of the transformer is distorted due to the DC magnetic bias, and further, the false operation or the failure operation of the relay protection is caused.

(3) The direct current magnetic biasing can increase the magnetic leakage of the transformer, cause the eddy current loss of the transformer, further generate abnormal heating, damage transformer parts and reduce the service life of the transformer.

The existing research indicates that 4 types of characteristic quantities, namely the direct-current component, the maximum value, the minimum value and the total harmonic distortion rate of the exciting current of the transformer are obtained, so that the severity of the direct-current magnetic biasing of the transformer can be analyzed, and further data support can be provided for analyzing and inhibiting the direct-current magnetic biasing; the method is beneficial to solving the harm caused by the direct current magnetic bias of the transformer from the source by analyzing the relation between the exciting current, particularly the exciting current characteristic quantity and the direct current of the neutral point of the transformer.

In the field of excitation characteristic analysis, a Chinese patent with the application number of 201510214689.2 provides an excitation characteristic change method for analyzing the influence of direct current magnetic biasing of a transformer, which obtains the excitation characteristic of the transformer under the direct current magnetic biasing based on the magnetic coupling principle and Fourier transform, but the method cannot obtain the excitation current or the excitation current characteristic quantity of the transformer on line and further cannot obtain the relation between the excitation current or the excitation current characteristic quantity and the direct current at the center point of the transformer. Chinese patent application No. 201510184067.X proposes a method for determining hysteresis characteristics and loss characteristics of a transformer in a dc magnetic bias state, which mainly determines the hysteresis characteristics and loss characteristics of the transformer by establishing a finite element model of a laminated core of the transformer, but the method cannot acquire the excitation current or excitation current characteristic quantity of the transformer on line. In the field of obtaining exciting current, chinese patent application No. 2015103125304 proposes a transformer exciting current simulation method based on a J-a hysteresis model, and a document "simulation and experimental study of transformer exciting current under the condition of direct-current magnetic bias based on a J-a model (written in S2 of 2013 by the journal of electrotechnical science, zhao, dao, chen de zhi wang jian li bao peng, am) proposes a transformer exciting current experimental method based on a J-a model", but these two methods can only obtain transformer exciting current through simulation or experiment, cannot obtain transformer exciting current or characteristic quantity thereof in actual engineering on line, and do not relate to obtaining relation between transformer exciting current or exciting current characteristic quantity and transformer neutral point direct current. The chinese patent with application number 2016108556946 proposes a real-time calculation method for extra-high voltage transformer no-load dc magnetic bias exciting current, but it can only obtain exciting current under the dc magnetic bias condition during the transformer no-load experiment, and still can not obtain the exciting current or its characteristic quantity of the transformer in the actual engineering on line.

In summary, in the aspect of obtaining and researching the transformer exciting current or the characteristic quantity thereof, two main problems mainly exist in the existing research or technology, one is that the exciting current of the transformer can only be obtained when the transformer performs a no-load experiment, that is, when the transformer is connected to a power system to operate, the exciting current cannot be obtained; secondly, the existing research or technology is still in the blank stage in the relation between the transformer exciting current characteristic quantity and the transformer neutral point direct current.

Disclosure of Invention

The technical problem to be solved by the invention is as follows: the method for acquiring the characteristic quantity of the transformer exciting current on line is provided to solve the problems in the prior art.

The technical scheme adopted by the invention is as follows: a transformer exciting current characteristic quantity online obtaining method comprises the following steps:

A. obtaining sample data of characteristic quantities of direct current and exciting current of a neutral point of a transformer through an experiment:

a1, performing no-load test on the transformer, and injecting direct current i changing along with time into the neutral point of the transformer₁The duration is t seconds; real-time acquisition of transformer exciting current i in experiment₂(ii) a Wherein the injection frequency of the direct current is f₁Excitation current i of transformer₂Has a sampling frequency of f₂；

A2, mixing i₁Data contained every 0.04 secondsQuantity m₁＝0.04f₁Segmenting at intervals to obtain current sample data of neutral points of n segments of transformers, wherein the current sample data is i₁(k)、i₂(k)、…、i_n(k) (ii) a Wherein k is 1, 2, …, m₁The sampling point serial number of each segment of data;

a3, mixing i₂In the amount of data m contained per 0.04 second₂＝0.04f₂Segmenting at intervals to obtain n segments of transformer exciting current sample data; respectively calculating the DC component of the excitation current sample data of each section of transformer to obtain the excitation current DC component sample data i₂(1)_dc、i₂(2)_dc、…、i₂(n)_dc(ii) a Respectively calculating the maximum value of the sample data of the excitation current of each section of the transformer to obtain sample data of the maximum value of the excitation current, i₂(1)_max、i₂(2)_max、…、i₂(n)_max(ii) a Respectively calculating the minimum value of the sample data of the excitation current of each section of the transformer, and respectively obtaining sample data of the minimum value of the excitation current as i₂(1)_min、i₂(2)_min、…、i₂(n)_min(ii) a Respectively calculating the total harmonic distortion rate of the excitation current sample data of each section of transformer to obtain the sample data of the total harmonic distortion rate of the excitation current, i₂(1)_THD、i₂(2)_THD、…、i₂(n)_THD；

B. Training a neural network:

b1, taking the current sample data of the neutral point of the transformer and the DC component sample data, the maximum sample data, the minimum sample data and the total harmonic distortion sample data of the exciting current of the transformer in the same 0.04 second time period as a training sample T of the neural network_rain(j)＝[i_j(1)，i_j(2)，…，i_j(k)，i₂(j)_dc，i₂(j)_max，i₂(j)_min，i₂(j)_THD]Wherein j is 1, 2, …, n; n training samples in n periods form a training sample set T of the neural network_rain＝[T_rain(1)；T_rain(2)；…；T_rain(n)]；

B2 setting node number p of input layer of neural network₁Equal to the sample data quantity of current at neutral point of a section of transformer, i.e. p₁＝m＝0.04f₁(ii) a Setting the number p of nodes of an output layer₂Number of types equal to characteristic quantity of transformer exciting current, i.e. p₂4; setting the number of hidden layers as 1 and the number of nodes of the hidden layers as

Setting a stimulus function as a sigmoid function;

b3 training sample set T_rainLeading the data into a neural network, training the neural network by using a back propagation algorithm, finishing the training when the training error is less than 0.001 or the iteration number of the training is more than 50000, and storing the trained neural network;

C. online acquisition of transformer exciting current characteristic quantity:

at a sampling frequency f₃Real-time measuring transformer neutral point current in practical engineering and extracting its DC component as i₁ ^*According to the data window N₁＝0.04f₃With N₂＝0.001f₃For spacing slidably move i₁ ^*Data i in_q ^*(z) inputting the input layer of the trained neural network to obtain the direct current component i of the transformer exciting current₂ ^*(q)_dcMaximum value i₂ ^*(q)_maxMinimum value i₂ ^*(q)_minAnd the total harmonic distortion rate i₂ ^*(q)_THD4-type feature quantity; wherein z is 1, 2, …, N₁The sampling point serial number of the data in each data window is used; q is the number of the sliding times of the data window, and q is 1, 2, ….

And D, the type and the structural parameters of the transformer used in the air-load experiment in the step A are consistent with those of the transformer of the actual engineering in the step C.

The number n of training samples used for neural network training in the step B is not less than 500.

The range of the amplitude variation of the direct current injected into the neutral point of the transformer in the step A1 is consistent with the range of the amplitude variation of the direct current component of the neutral point of the transformer in the actual engineering, and the range of the amplitude variation varies from-100A to 100A.

Step A1, the injection frequency f of the DC signal injected into the neutral point of the transformer₁Exciting current i of transformer₂Sampling frequency f₂And C, sampling frequency f of the current of the neutral point of the transformer for practical engineering in the step C₃Are equal.

The invention has the beneficial effects that: compared with the prior art, the invention has the following effects:

(1) according to the method, the direct-current component of the neutral point current of the transformer is measured on line, four types of characteristic quantities, namely the direct-current component, the maximum value, the minimum value and the total harmonic distortion rate of the exciting current of the transformer are calculated by fitting through a neural network algorithm, so that a clear relation between the characteristic quantity of the exciting current of the transformer and the direct-current component of the neutral point of the transformer can be obtained, data support and guidance can be further provided for analysis and inhibition of direct-current magnetic biasing of the transformer, and the method has high engineering significance;

(2) only the current of the neutral point of the transformer needs to be collected, the characteristic quantity of the exciting current of the transformer is obtained through the fitting of the trained neural network model, the electric quantity needing to be collected is less, and the fitting calculation of the characteristic quantity of the exciting current of the transformer is simpler and easy to implement.

Drawings

FIG. 1 is a flow chart of the operation of the present invention.

Detailed Description

The invention is further described with reference to the accompanying drawings and specific embodiments.