CN102520373B - Distinguishing method of direct current magnetic biasing of power transformer based on vibration analysis - Google Patents

Abstract

The invention discloses a distinguishing method of direct current magnetic biasing of a power transformer based on vibration analysis, which includes obtaining a corresponding numeric symbol sequence by collecting vibration signals on the surface of an oil tank of the transformer and embedding the vibration signals into an m-dimensional space through time delay, and constructing corresponding characteristic function through a column diagram of the sequence and then distinguishing the direct current magnetic biasing according to the characteristic quantity. Therefore, the distinguishing method can distinguish whether abnormal strong vibration of the power transformer is caused by the direct current magnetic biasing or not and the severe degree of the magnetic biasing, is not required to be electrically connected with the transformer and accordingly has no interference effect on operation of the transformer. The distinguishing method is intuitive and simple to operate after being written into software program and has no special requirements for operators, and the operators can operate after simple training. Simultaneously, the distinguishing method is high in distinguishing accuracy, easy to implement and popularize, low in implementation cost and capable of being applied to online detection and online state detection and fault diagnosis of the power transformer.

Description

A kind of method of discrimination of the power transformer DC magnetic biasing based on vibration analysis

Technical field

The invention belongs to input analysis technical field, be specifically related to a kind of method of discrimination of the power transformer DC magnetic biasing based on vibration analysis.

Background technology

Large-scale power transformer is node important in electrical network, and the safe operation of large-scale power transformer is extremely important for the reliability service of electrical network.Adopting the running status of vibration analysis method on-line monitoring transformer is a study hotspot in recent years, and its main points are to detect transformer whether abnormal or fault occur by analyzing the tank surface vibration signal of power transformer.When transformer generation abnormal vibrations and noise, vibration can significantly increase.The strong vibration meeting of transformer affects the safe operation of transformer, and the winding of transformer and iron core etc. are partly broken down, and also surrounding enviroment is produced to noise pollution simultaneously.Therefore, the reason of searching transformer exception vibration and noise contributes to " suiting the remedy to the case ", takes measures in time to safeguard the safe and stable operation of transformer.

DC magnetic biasing is the recurrent a kind of abnormal conditions of neutral ground power transformer, the vibration and the noise that when the vibration that transformer in service produces while there is DC magnetic biasing and noise normally move far above transformer, produce.The essence of DC magnetic biasing is to occur that direct current biasing obviously increases magnetic flux maximum value in iron core under the effect due to neutral point direct current of the magnetic flux of transformer fe in-core, and inspires the strong vibration of iron core.While there is DC magnetic biasing, the exciting current in winding also can significantly increase, and basket vibration is increased.Because the abnormal vibrations of operating transformer also comprised the various physical construction fault state such as transformer overexcitation process, winding and core slackness distortion, it is necessary from transformer strong vibration signal, accurately distinguishing the vibration that DC magnetic biasing causes.At present the magnetic bias research of power transformer is concentrated on relevant electrical parameter analysis substantially, but abnormal vibrations and the noise of only from electrical quantity, analyzing transformer have limitation.

Existing magnetic bias method of discrimination is generally whether to occur that from neutral point current DC component differentiates, but electric parameter and transformer device structure relevance are little, whether the vibration of electrical quantity and power transformer does not have clear and definite funtcional relationship, only from neutral point current judgement, can not provide magnetic bias to impact transformer device structure.In addition, Power Transformer in Field carries out comparatively difficulty and can cause certain influence to transformer operation of electric measurement, such as the measurement of the neutral point current fixed form (being fixed on cement wall or bar) due to neutral ground bar is often difficult for measuring.

Summary of the invention

For the existing above-mentioned technological deficiency of prior art, the invention provides a kind of method of discrimination of the power transformer DC magnetic biasing based on vibration analysis, whether the abnormal strong vibration that can differentiate power transformer is caused by DC magnetic biasing and the order of severity of magnetic bias.

A method of discrimination for the power transformer DC magnetic biasing of vibration analysis, comprises the steps:

(1) vibration signal on power transformer fuel tank surface is obtained in sampling, and described vibration signal is normalized;

(2) vibration signal after normalization is embedded in m-dimensional space by time delay, obtains the space oscillations signal of m dimension, and then build the numeric character sequence of space oscillations signal;

(3) according to described numeric character sequence, set up corresponding histogram, and then build characteristic of correspondence function according to described histogram;

(4), according to described fundamental function, differentiate in power transformer whether have DC magnetic biasing; If there is DC magnetic biasing, according to the vibration signal after described normalization, determine the intensity grade of DC magnetic biasing.

The space oscillations signal of described m dimension is comprised of n element, and each element consists of m sampled point, and the element expression in space oscillations signal is as follows:

x _s(i)＝{x(i)，x(i+u)，x(i+2u)，...，x[i+(m-1)u]} (1)

Wherein: x _s(i) be the i element in space oscillations signal, n is the sampled point number in vibration signal, and x (i) is the accekeration of i sampled point in the vibration signal after normalization, and u is that time delay is counted and is practical experience value.

In described step (2), by following formula, build the numeric character sequence of space oscillations signal;

R(τ)＝Θ[ε-||x _s(i)-x _s(i+τ)||] (2)

Wherein: R (τ) is numeric character sequence, Θ is Heaviside function, x _s(i) be the i element in space oscillations signal, ε is distance threshold and for practical experience value, τ is that time delay is counted.

In described step (3), according to histogram, by following formula, build characteristic of correspondence function;

$L\left(\mathrm{\τ}\right)=\frac{{\mathrm{\Σ}}_{k=\min }^{n-\mathrm{\τ}}k{P}_{\mathrm{\τ}}\left(k\right)}{{\mathrm{\Σ}}_{k=1}^{n-\mathrm{\τ}}k{P}_{\mathrm{\τ}}\left(k\right)}---\left(3\right)$

Wherein: L (τ) is fundamental function, P _τ(k) in numeric character sequence by the number of continuous k 1 array forming, n is the sampled point number in vibration signal, τ is that time delay is counted, min is minimum combination coefficient and is practical experience value.

In described step (4), differentiate in power transformer and whether exist the method for DC magnetic biasing to be:

1) according to formula τ=f/F, calculative determination τ; Wherein, the sample frequency that f is vibration signal, F is time delay frequency and for practical experience value, τ is that time delay is counted;

2) τ, τ-1 and τ+1 respectively in substitution fundamental function, are asked for to three corresponding eigenwerts, and three eigenwerts are averaging, if characteristic mean is less than discrimination threshold, show to exist in power transformer DC magnetic biasing.

In described step (4), the method of determining the intensity grade of DC magnetic biasing according to the vibration signal after normalization is: first the vibration signal after normalization is carried out to high-pass filtering and obtain filtering vibration signal, then by following formula, solve the degree value of DC magnetic biasing, and then according to described degree value, determine the intensity grade of DC magnetic biasing;

$B\left(i\right)=\frac{\left|\right|x_1\left(i\right)\left|\right|}{\left|\right|x\left(i\right)\left|\right|}---\left(4\right)$

$B=\frac{1}{n}{\mathrm{\Σ}}_{i=1}^{n}B\left(i\right)---\left(5\right)$

Wherein: x (i) is the accekeration of i sampled point in the vibration signal after normalization, x _l(i) be the accekeration of i sampled point in filtering vibration signal, n is the sampled point number in vibration signal, the degree value that B is DC magnetic biasing.

Whether the abnormal strong vibration that the present invention can differentiate power transformer is caused by DC magnetic biasing and the order of severity of magnetic bias, without being electrically connected with transformer, therefore on the noiseless impact of the operation of transformer; After the inventive method is compiled into software program, intuitively simply, to operating personnel, without specific (special) requirements, training can operate a little in operation; The inventive method differentiation accuracy rate is high simultaneously, easy practice and extension, and implementation cost is low, can be applicable to online detection and on-line condition monitoring and the fault diagnosis of power transformer.

Accompanying drawing explanation

Fig. 1 is steps flow chart schematic diagram of the present invention.

Fig. 2 is for adopting the field conduct diagnosis schematic diagram of the inventive method.

Fig. 3 is the vibration signal of two measuring points of the power transformer time dependent trend map of characteristic mean in a magnetic bias process.

Fig. 4 is that cutoff frequency is the amplitude-frequency response figure of the 200 rank FIR Hi-pass filters of 700Hz.

Fig. 5 is the vibration signal of four measuring points of the power transformer time dependent trend map of B value in a magnetic bias process.

The oscillogram of the vibration signal of measuring point after normalization when Fig. 6 (a) is the serious magnetic bias of power transformer.

Fig. 6 (b) is that the vibration signal of Fig. 6 (a) is through the oscillogram of filtered filtering vibration signal.

Embodiment

In order more specifically to describe the present invention, below in conjunction with the drawings and the specific embodiments, method of discrimination of the present invention is elaborated.

As shown in Figure 1, a kind of method of discrimination of the power transformer DC magnetic biasing based on vibration analysis, comprises the steps:

(1) vibration signal on power transformer fuel tank surface is obtained in sampling, and vibration signal is normalized.

As shown in Figure 2, present embodiment utilizes piezoelectric acceleration vibration transducer 2 to be arranged on power transformer 1 tank surface (positions such as the weld seam, reinforcement, tank wall edge of fuel tank are avoided in fixed position as far as possible) of load running, utilize the vibration signal on signal pickup assembly 3 sampling power transformer fuel tank surfaces, for computing machine 4 analyzing and diagnosings; Sample frequency is 8000Hz, and the continuous sampling time is 1 second, therefore the sampled point number n=8000 in vibration signal.

(2) vibration signal after normalization is embedded in m-dimensional space by time delay, obtains the space oscillations signal of m dimension;

The space oscillations signal of m dimension is comprised of n element, and each element consists of m sampled point, and the element expression in space oscillations signal is as follows:

x _s(i)＝{x(i)，x(i+u)，x(i+2u)，...，x[i+(m-1)u]} (1)

Wherein: x _s(i) be the i element in space oscillations signal, x (i) is the accekeration of i sampled point in the vibration signal after normalization, and u is that time delay is counted; U=20 in present embodiment, m=3.

By following formula, build the numeric character sequence of space oscillations signal;

R(τ)＝Θ[ε-||x _s(i)-x _s(i+τ)||] (2)

Wherein: R (τ) is numeric character sequence, Θ is Heaviside function, and ε is distance threshold, and τ is that time delay is counted; ε=0.25 in present embodiment, the norm calculation in formula 2 adopts 2-norm calculation mode.

(3) according to numeric character sequence, set up corresponding histogram, and then by following formula, build characteristic of correspondence function according to histogram;

$L\left(\mathrm{\τ}\right)=\frac{{\mathrm{\Σ}}_{k=\min }^{n-\mathrm{\τ}}k{P}_{\mathrm{\τ}}\left(k\right)}{{\mathrm{\Σ}}_{k=1}^{n-\mathrm{\τ}}k{P}_{\mathrm{\τ}}\left(k\right)}---\left(3\right)$

Wherein: L (τ) is fundamental function, P _τ(k) in numeric character sequence by the number of continuous k 1 array forming, min is minimum combination coefficient; Min=3 in present embodiment.

(4) according to formula τ=f/F, calculative determination τ; Wherein, the sample frequency that f is vibration signal, F is time delay frequency; F=8000 in present embodiment, F=100, calculates to obtain τ=80.

τ, τ-1 and τ+1 respectively in the fundamental function of substitution formula 3, are asked for to three corresponding eigenwerts, and three eigenwerts are averaging, if characteristic mean is less than discrimination threshold, show to exist in power transformer DC magnetic biasing; In present embodiment, discrimination threshold is 0.2.Fig. 3 be 500KV power transformer when a DC magnetic biasing occurs gradually, according to two vibration detection, put in continuous two hours the characteristic mean L that the data (at interval of 1 second vibration data of 60 seconds continuous acquisition) of record calculate _atime dependent trend map; In Fig. 3, provided the DC component that flows into power transformer neutral point current, neutral point current DC component is not 0 to be magnetic bias while occurring, as can be seen from the figure L simultaneously _aprovided exactly transformer DC magnetic bias situation.

If there is DC magnetic biasing in power transformer, first utilize the FDAtool kit in Matlab software to design a FIR Hi-pass filter.Fig. 4 is for adopting the amplitude-frequency response figure of the 200 rank FIR Hi-pass filters that the cutoff frequency of FDAtool kit design is 700Hz.

Vibration signal after first utilizing FIR Hi-pass filter to normalization carries out filtering and obtains filtering vibration signal, then by following formula, solves the degree value of DC magnetic biasing, and then according to degree value, determines the intensity grade of DC magnetic biasing;

$B\left(i\right)=\frac{\left|\right|x_1\left(i\right)\left|\right|}{\left|\right|x\left(i\right)\left|\right|}---\left(4\right)$

$B=\frac{1}{n}{\mathrm{\Σ}}_{i=1}^{n}B\left(i\right)---\left(5\right)$

Wherein: x (i) is the accekeration of i sampled point in the vibration signal after normalization, x _l(i) be the accekeration of i sampled point in filtering vibration signal, n is the sampled point number in vibration signal, the degree value that B is DC magnetic biasing; Norm calculation in formula 4 adopts 1-norm calculation mode.

For there is corresponding to the 500KV power transformer of Fig. 3 the time dependent trend map of B value that DC magnetic biasing process calculates from 4 measuring points in Fig. 5.As can be seen from the figure the reasonable degree that reflects magnetic bias of B value.During beginning, DC current is little, and transformer bias degree is lighter, and B value is generally less than 0.25; And at a period of time B > 0.25 subsequently, show that transformer has entered moderate magnetic bias state; In the geomagnetic storm serious or transformer power up, can observe the state that serious magnetic bias appears in transformer, corresponding B value is often greater than 0.4; According to magnetic bias degree, can take corresponding measure to safeguard the safe and stable operation of transformer.Fig. 6 has provided 220KV the power transformer vibration signal (a) after measuring point normalization and comparison of wave shape of filtering vibration signal (b) when there is serious magnetic bias, and now B value is about 0.518.

Claims (4)

1. a method of discrimination for the power transformer DC magnetic biasing based on vibration analysis, comprises the steps:

(1) vibration signal on power transformer fuel tank surface is obtained in sampling, and described vibration signal is normalized;

(2) vibration signal after normalization is embedded in m-dimensional space by time delay, obtains the space oscillations signal of m dimension, and then the numeric character sequence of structure space oscillations signal is as follows:

R(τ)＝Θ[ε-||x _s(i)-x _s(i+τ)||]

Wherein: R (τ) is numeric character sequence, ε is distance threshold, and Θ is Heaviside function, and τ is that time delay is counted;

(3) according to described numeric character sequence, set up corresponding histogram, and then it is as follows according to described histogram, to build characteristic of correspondence function:

$L\left(\mathrm{\τ}\right)=\frac{{\mathrm{\Σ}}_{k=\min }^{n-\mathrm{\τ}}k{P}_{\mathrm{\τ}}\left(k\right)}{{\mathrm{\Σ}}_{k=1}^{n-\mathrm{\τ}}k{P}_{\mathrm{\τ}}\left(k\right)}$

Wherein: L (τ) is fundamental function, P _τ(k) in numeric character sequence by the number of continuous k 1 array forming, min is minimum combination coefficient;

(4), according to described fundamental function, differentiate in power transformer whether have DC magnetic biasing; If there is DC magnetic biasing, according to the vibration signal after described normalization, determine the intensity grade of DC magnetic biasing.

2. the method for discrimination of the power transformer DC magnetic biasing based on vibration analysis according to claim 1, it is characterized in that: the space oscillations signal of described m dimension is comprised of n element, each element consists of m sampled point, and the element expression in space oscillations signal is as follows:

x _s(i)={x(i)，x(i+u)，x(i+2u)，…，x[i+(m-1)u]}

Wherein: x _s(i) be the i element in space oscillations signal, x (i) is the accekeration of i sampled point in the vibration signal after normalization, and u is that time delay is counted.

3. whether the method for discrimination of the power transformer DC magnetic biasing based on vibration analysis according to claim 1, is characterized in that: in described step (4), differentiate in power transformer and exist the method for DC magnetic biasing to be:

1) according to formula τ=f/F, calculative determination τ; Wherein, the sample frequency that f is vibration signal, F is time delay frequency, τ is that time delay is counted;

2) τ, τ-1 and τ+1 respectively in substitution fundamental function, are asked for to three corresponding eigenwerts, and three eigenwerts are averaging, if characteristic mean is less than discrimination threshold, show to exist in power transformer DC magnetic biasing.

4. the method for discrimination of the power transformer DC magnetic biasing based on vibration analysis according to claim 1, it is characterized in that: in described step (4), the method of determining the intensity grade of DC magnetic biasing according to the vibration signal after normalization is: first the vibration signal after normalization is carried out to high-pass filtering and obtain filtering vibration signal, then by following formula, solve the degree value of DC magnetic biasing, and then according to described degree value, determine the intensity grade of DC magnetic biasing;

$B\left(i\right)=\frac{\left|\right|x_1\left(i\right)\left|\right|}{\left|\right|x\left(i\right)\left|\right|}$

$B=\frac{1}{n}{\mathrm{\Σ}}_{i=1}^{n}B\left(i\right)$

Wherein: x (i) is the accekeration of i sampled point in the vibration signal after normalization, x ₁(i) be the accekeration of i sampled point in filtering vibration signal, the degree value that B is DC magnetic biasing.

Images