KR20060112747A - Improved adaptive predictive filter for generating reference signal and active power filter using the same - Google Patents

Abstract

An adaptive predictive filter for generating a reference signal and an active power filter using the same are provided to compensate for a drift phenomenon by using a leakage factor and/or a scale coefficient compensation algorithm. An adaptive predictive filter reduces a distortion due to harmonic components. A pre-filter(420) reduces the harmonic component from an input, which is delivered to an adaptive predictive filter(430). A fifth-order butter-worth IIR(Infinite Impulse Response) filter is used for the pre-filter. A leakage factor is introduced to the adaptive predictive filter to alleviate or remove drifting of filter coefficients. When the coefficient of the adaptive predictive filter is updated, an output is determined by using an FIR(Finite Impulse Response) filter. The leakage factor is set, so that a THD(Total Harmonic Distortion) of output data from the adaptive predictive filter is minimized and the filter coefficient is not drifted.

Description

Improved Adaptive Prediction Filter for Generating Reference Signals and Active Power Filter System Using the Same {IMPROVED ADAPTIVE PREDICTIVE FILTER FOR GENERATING REFERENCE SIGNAL AND ACTIVE POWER FILTER USING THE SAME}

1 is a view for explaining the configuration and operation principle of a general active power filter.

2 is a block diagram showing the data flow of the reference signal generator 100 of the prior art.

3 is a block diagram illustrating a data processing flow of the adaptive prediction filter 30 of the prior art.

4 schematically shows a preferred embodiment of the improved reference signal generator of the present invention.

5 shows the total harmonic distortion (THD) included in the adaptive prediction filter output when the leakage factor is changed from 0.5 to 0.999.

6 (a) shows the waveform of the load current, and (b) and (c) show the frequency response of the output and the filter coefficient of the adaptive prediction filter generated when the load current passes through the reference signal generator of the prior art.

7 shows a simulation result performed for 20 seconds with respect to the conventional reference signal generator.

8 shows the frequency response of the filter coefficient and the output of the adaptive prediction filter generated when the load current passes through the reference signal generator of the present invention.

9 is a simulation result carried out to check the operating state of the reference signal generator of the present invention when the load changes.

10 schematically shows the configuration of a test apparatus for testing by implementing the reference signal generator of the present invention in hardware.

11 shows the operation of a reference signal generator based on the adaptive prediction filter of the present invention.

12 shows the operation of the active power filter by the dead bit current controller.

(A) of FIG. 13 shows the fundamental wave component which is the interpolator output of the reference signal generator and the current on the compensated power supply side when the load increases by 20% step.

<Explanation of symbols for the main parts of the drawings>

10, 410: peak detector 20, 420: prefilter

30, 430: adaptive prediction filter 40, 440: interpolator

435: fundamental wave size correction unit

The present invention relates to an active power filter system, and more particularly, to provide a reference signal generator capable of accurately extracting a reference signal without phase delay by improving the performance of the adaptive prediction filter. To provide a filter system.

Recently, as the spread of power electronic devices, which are nonlinear loads, becomes more common, a problem caused by harmonic currents is serious. Harmonic currents distort the terminal voltage of the AC system, causing overheating and vibration in the associated power equipment. Therefore, harmonic current shortens the life of power equipment and breaks down AC system.

An active power filter is a power electronic device designed to effectively remove harmonic currents generated by nonlinear loads. The function of the active power filter is to inject a current having the same magnitude and opposite phase as the harmonic included in the load current in parallel with the load so that only the current corresponding to the fundamental wave component without distortion is supplied from the power supply side. Therefore, the performance of the active power filter is determined by the characteristics of the inverter, the harmonic current control method, and the method of extracting the reference current signal.

As a method of extracting a reference current signal from a load current, a method of obtaining a load current through a Notch filter or a method of obtaining an instantaneous power theory by rectangular coordinate transformation has been proposed. However, when the notch filter is used, a phase delay is caused in the reference signal. In the instantaneous power theory method, complex coordinate transformation and inverse transformation should be performed by measuring load current, supply voltage, and injection current of active power filter, and low-pass filter should be used.

To overcome the drawbacks of the above methods, the application of an adaptive predictive filter to generate a reference current signal without phase delay for its application as an active power filter was first published in 1998 by Valiviita and Ovaska [S. Valiviita and S. J. Ovaska, "Delayless method to generate current reference for active filters." IEEE Trans. Ind. Electron., Vol. 45, pp. 559-567, Aug. 1998.]. In this paper, we propose a reference signal generator consisting of a pre-filter, adaptive prediction filter, peak detector, and Lagrange interpolator to extract the reference signal, and describe the performance evaluation by MATLAB simulation.

Overall configuration of the active power filter system:

The basic function of an active power filter system is to mitigate the harmonic content of the supply current supplied to a nonlinear load. 1 is a view for explaining the configuration and operation principle of a general active power filter. The active power filter 300 is connected in parallel or in series with the load, but a method generally used in the power system is a method in which the active power filter 300 is coupled in parallel with the load 500 as shown in FIG. 1.

를 다음 수식과 같이 산출한다.As shown in FIG. 1, when the active power filter 300 supplies a current (i _F ) having the same phase and opposite magnitude as the harmonic current flowing through the load 500 to the connection point in parallel, the power current is based on Kirchhoff's law. Only wave components remain. To actually implement this, the reference current signal that the active power filter should inject into the system through the reference signal generator by detecting the load current.

Is calculated as in the following formula.

(1)

(One)

은 비선형 부하에 의해 왜곡된 부하전류이고

은 부하전류에 포함된 기본파 성분이다. 이렇게 산출된 기준 전류

는 인버터 출력 전류가 이를 추종하도록 스위칭 펄스를 만들어 낸다.here

Is the load current distorted by nonlinear load

Is the fundamental wave component included in the load current. So calculated reference current

Generates a switching pulse so that the inverter output current follows it.

The process of generating the reference signal represented by Equation (1) is simple in theory, but the problem is how to actually obtain the harmonic reference signal from which the fundamental wave components are removed from the load current including the harmonic current. The first proposal for this is the use of a notch filter to remove fundamental components. In this case, it is possible to extract harmonic components, but there is a problem that the extracted harmonics are delayed in phase. In order to effectively solve the problem of phase delay, in case of 3-phase nonlinear load, instantaneous coordinate transformation is used to change the fundamental wave component into direct current component, and the low-pass filter extracts the fundamental wave component. The method by the instantaneous power theory to generate the reference signal is included. The disadvantage of this method is that it requires a lot of input sensors and complex coordinate transformation.

Configuration of the reference signal generator 100:

2 is a block diagram showing the data flow of the reference signal generator 100 of the prior art. The conventional reference signal generator 100 is composed of a peak detector 10, a pre-filter 20, an adaptive predictive filter 30, and a Lagrange interpolator 40. .

First, since the adaptive prediction filter 30 is difficult to process when the magnitude of the input signal is greatly changed, the measured value is divided by the peak value and normalized. The signal passing through the adaptive prediction filter 30 is multiplied by the measured peak value to be denormalized. To remove unwanted frequency components of the normalized signal, filter with lowpass filter 20. If the original distorted signal directly enters the adaptive prediction filter 30, the least square method (LMS) algorithm is sensitive to all correlation elements, which prevents the adaptive performance of the filter coefficients.

, 10dB인 5차 Chebyshev II IIR(infinite impulse response) 필터를 사용하여 성능 시험을 수행하였다. 이때, 전치필터(20)의 샘플링 주파수는 1.67㎑ 이므로 통과영역 차단주파수 값은 142㎐가 된다. 따라서 5차, 7차 그리고 그 이상의 고조파는 전치필터(20)에 의해 제거된다. 또한 이 전치필터(20)에서 발생되는 위상지연 문제는 적응예측필터(30)에서 보상된다.The prefilter 20 has a passband cut-off frequency and a stopband ripple of 0.17, respectively.

The performance test was performed using a 5th order Chebyshev II infinite impulse response (IRR) filter, 10 dB. At this time, since the sampling frequency of the prefilter 20 is 1.67 kHz, the pass region cutoff frequency is 142 kHz. Thus, the fifth, seventh and more harmonics are removed by the prefilter 20. In addition, the phase delay problem generated by the prefilter 20 is compensated by the adaptive prediction filter 30.

The sampling frequency of 1.67 ㎑ is rather low to apply to PWM (Pulse Width Modulation) switching of the inverter in the active power filter 100, and a sampling frequency of about 10 kHz is required to remove harmonics caused by sampling. An interpolator 40 was applied.

Adaptive prediction filter 30:

3 is a block diagram illustrating a data processing flow of the adaptive prediction filter 30 of the prior art. In order to increase the sampling rate to 10 Hz, the adaptive prediction filter 30 must predict one step first. Thus, as shown in FIG. 3, the signal passing through the prefilter 20 is delayed by one step, and then the adaptive prediction filter processes the signal. Basically, the adaptive prediction filter 30 operates as a one-step-ahead predictor.

The adaptive prediction filter 30 is based on a finite impulse response (FIR) filter. The filter coefficient adaptive processing is based on Widrow-Hoff Least Mean Square. The algorithm is as follows.

(2)

(2)

The error involved in coefficient adaptation in this equation is defined by the following equation.

(3)

(3)

는 필터 계수벡터이고,

는 필터의 데이터 벡터이다. N은 FIR 필터의 길이이다. 식 (2)에 사용된

는 LMS 알고리즘의 안정성을 보장하기 위해 충분히 작게 선택되어진 값이다.

값의 선택은 항상 적응률과 전반적인 시스템의 안정성 사이에서 트레이드-오프(trade-off)관계에 있다. 여기에서는

=0.002,

=22로 선택되었다.here

Is the filter coefficient vector,

Is the data vector of the filter. N is the length of the FIR filter. Used in equation (2)

Is a value chosen small enough to ensure the stability of the LMS algorithm.

The choice of value is always a trade-off between adaptation rate and overall system stability. Here

= 0.002,

= 22 was chosen.

는 다음 식에 의해 계산된다.Finally, the final output of the adaptive prediction filter

Is calculated by the following equation.

(4)

(4)

Interpolator (40):

The output value of the adaptive prediction filter is 1.67 1. sampling data. The time resolution can be improved by up-sampling by the interpolator 40. The interpolation rate is determined by software and hardware conditions. The switching frequency of an active power filter is typically from a few kHz to 10 kHz.

Secondary Lagrange interpolator 40 was used as a method of interpolating arbitrary points. The advantage of this Lagrange interpolation is that it is not necessary to calculate the difference of the data points, and the calculation amount can be reduced by precomputing and tabulating the interpolation coefficients.

,

,

을 조작하는 6상의 필터로 구성되어 있다. 각각의 새로운 입력 샘플에 대해 출력 측에서 각 6상 필터로부터의 샘플이 모아지게 된다. 다상 필터의 계수,

는 식 (5) 와 같이 주어진다.The interpolator 40 de-normalizes the adaptive predictive filter and corrects the fundamental wave size.

,

,

It consists of a six-phase filter that operates. For each new input sample, samples from each six-phase filter are collected at the output side. Coefficient of polyphase filter,

Is given by equation (5).

(5)

(5)

As a result, the output signal v (m) of the interpolator 40 is determined by equation (6).

(6)

(6)

는 interpolator(40)의 데이터 벡터이다. 다상 필터 인덱스

는 출력 시간 인덱스 m이 입력 시간 인덱스 n보다 6배 빠르게 증가하기 때문에

의 나머지로서 10㎑비율로 회전 이동한다.here,

Is the data vector of the interpolator 40. Polyphase filter index

Since the output time index m increases six times faster than the input time index n

It rotates at a rate of 10 Hz as the rest of.

Peak detector (10):

Since the large change in the current magnitude is a factor that hinders the stable performance of the adaptive prediction filter, the normalization is performed by dividing by the measured peak value to alleviate this. This normalizes the signal to a signal with a peak value of nearly one. The peak detector consists of selecting the maximum value of the input sample during one period of the fundamental frequency. If a sampling frequency of 1.67 kHz is used in a 60 kHz power system, 22 samples are required for peak detection.

However, the reference signal extracted from the load current using the reference signal generator proposed in the prior art appears to operate normally for the first 2-3 seconds, but after 20 seconds or more elapsed, the extracted fundamental wave component is severely distorted. . Therefore, it is impossible to apply this to the actual active power filter as it is.

SUMMARY OF THE INVENTION The present invention has been made to solve such a problem, and to provide an improved reference signal generator and an active power filter using the same to improve the operation algorithm of the adaptive prediction filter in the reference signal generator. will be.

In addition, the present invention is to provide an adaptive prediction filter with a drift corrected by introducing a leak factor (leakage factor) and by adding an algorithm for correcting the scale factor to be applied to the reference signal generator.

Furthermore, the present invention improves the components of the prefilter, the peak detector, and the adaptive prediction filter, thereby simplifying the system configuration such as reducing the number of sensors, and stably extracting the reference signal even for a signal having a variable size. This is possible, to provide a reference signal generator that can ensure a very good operating characteristics when used in an active power filter.

및 상기 정규화된 전력 신호에서 m차 이상의 고조파 성분을 제거한 입력 신호

을 입력받아, 출력신호

[여기서

는 필터 계수벡터]을 출력하는 FIR 필터 및 최소자승법 기반의 적응예측필터이며, n+1 스텝의 필터 계수 벡터 H(n+1)를 결정하기 위하여,

으로 정의되는 관계를 사용하고, 여기서,

,

는 필터의 데이터 벡터이며, δ는 필터 계수의 표류를 완화하기 위해 0<δ<1 범위에서 조정되는 리키지 팩터(leakage factor) 값인 것을 특징으로 한다.In order to achieve the above object, the adaptive prediction filter according to the first aspect of the present invention is a normalized current signal.

And an input signal from which harmonic components of m order or more are removed from the normalized power signal.

Input signal, output signal

[here

Is an FIR filter and a least-squares adaptive prediction filter that outputs a filter coefficient vector], and to determine a filter coefficient vector H (n + 1) of n + 1 steps,

Using a relationship defined by

,

Is a data vector of the filter, and δ is a leakage factor value adjusted in the range of 0 <δ <1 to mitigate the drift of the filter coefficients.

If necessary, the adaptive prediction filter of the present invention preferably compensates for the loss of the fundamental wave in the output signal generated by the introduction of the above-mentioned liquid factor.

In order to compensate for the loss of the fundamental wave size, two orthogonal sinusoids of unit size are passed through a built-in FIR filter to obtain a scaling factor (SF) by taking the inverse of the vector sum of the two sinusoidal components as the inverse. The method of multiplying by the output signal can be used.

The reference signal generator according to the second aspect of the present invention is characterized by including the adaptive prediction filter as described above.

Here, the reference signal generator of the present invention further includes a peak detector for selecting the maximum value of the input sample during one period of the fundamental frequency to normalize the current signal, the peak detector further comprises a low pass filter (LPF) It may be.

In addition, the reference signal generator of the present invention may further include a pre-filter for removing harmonic components of the signal input to the adaptive prediction filter.

The active power filter according to the third aspect of the present invention is characterized by including the reference signal generator as described above.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The present inventors improve the operation algorithm of the reference signal generator, perform simulation using MATLAB to evaluate its performance, implement it with hardware using TMS320C31 DSP, and manufacture single-phase active power filter based on this reference signal generator. As a result of the actual operation test connected to the nonlinear load composed of the single-phase diode rectifier, it was verified that the following good characteristics were obtained.

이 약 20초 정도 경과 후 표류하여 결국에는 출력으로 나오는 기본파에 고조파가 함유되어 출력 파형이 왜곡되는 현상이 발생한다. 본 발명자들은 이러한 표류 문제가 지속적인 적응조건을 만족하지 못하는 환경에서 도 3과 같이 LMS 알고리즘이 사용되어질 때 나타난다는 사실을 관측하였다. 따라서 이러한 표류현상을 제거하고 출력 기본파에 고조파 함유를 최소화하기 위해 본 발명자들은 기준신호발생기를 도 4와 같이 재구성하였다.4 schematically shows a preferred embodiment of the improved reference signal generator of the present invention. The reference signal generator according to the related art as shown in FIG. 2 described above is a coefficient vector of the adaptive prediction filter.

After about 20 seconds, the drift takes place, and eventually, the fundamental wave coming out of the output contains harmonics, causing the output waveform to be distorted. The inventors observed that this drift problem occurs when the LMS algorithm is used as shown in FIG. 3 in an environment that does not satisfy the continuous adaptation condition. Therefore, in order to eliminate such drift and to minimize harmonic content in the output fundamental wave, the present inventors reconfigure the reference signal generator as shown in FIG. 4.

In the configuration of the proposed reference signal generator, the interpolator 440 is configured in the same form as the conventional reference signal generator, and the remaining components are improved as follows.

Peak detector 440:

The function of the peak detector 440 to reduce distortion due to harmonics is the same as that of the peak detector 40 of the conventional reference signal generator. However, the conventional peak value detection was performed at a sampling rate of 1.67 kHz, and a signal having such a low sampling rate was interrupted by harmonics of 835 Hz or higher, which is good for the input of the adaptive prediction filter 430. This affects the output stability of the adaptive prediction filter 430 as a result. To minimize this effect, a second LPF with a 180 Hz cut-off frequency was placed at the input stage to filter the input signal. In addition, the filtered signal was sampled at a rate of 10 kHz, which significantly reduces the effects of higher-order harmonics.

Prefilter 420:

In the above-described reference signal generator 100 of FIG. 2, the fifth order Chebyshev II is a pre-filter 20 for the input transmitted to the adaptive predictive filter 30 to be a fundamental wave component with little contamination by harmonics. An IIR filter is used, which has good skirt characteristics and low phase delay, while harmonics of 835 Hz or higher remain after the prefilter 20 due to an aliasing-sensitive disadvantage, so that the operation of the adaptive prediction filter 30 is performed. It will exist as an obstacle to Therefore, in the reference signal generator 400 of the present invention, a fifth-order butter-worth IIR filter, which is lagging in phase characteristics and skirt characteristics but less sensitive to aliasing, is used as the prefilter 420 than the Chebyshev II IIR filter.

Adaptive prediction filter 430:

를 도입하여 관계식 (2)를 다음과 같이 수정하였다.Leakage factor to mitigate and eliminate drift in filter coefficients

By introducing the equation (2) was modified as follows.

(7)

(7)

,

그리고 적응률

등은 개선 전의 형태와 동일한 값을 취하였다. 그리고 계수가 업데이트 되면 식 (4)와 같이 FIR 필터에 의해 출력이 결정된다.The error calculation function is the same as Equation (3), and the coefficient vector

,

And adaptation rate

Etc. took the same value as the form before improvement. When the coefficient is updated, the output is determined by the FIR filter as shown in Equation (4).

=0.999로 리키지 팩터(leakage factor)가 설정되었다. 리키지 팩터는 0보다 크고 1보다 작은 범위 내에서 적절히 설정 가능하며, 리키지 팩터가 작으면 적응예측필터 계수의 업데이트가 빠르나 기본파 손실이 커져 신호의 안정성이 결여되고, 1에 가까울수록 기본파 손실이 적으며 고조파 함유율이 작아지는 특성을 가지나 필터계수의 표류가 발생하기 쉬움이 관측되었다. 따라서 이러한 범위 중 필터 계수의 표류가 발생하지 않는 값으로 설정하면 된다.5 shows the total harmonic distortion (THD) included in the adaptive prediction filter output when the leakage factor is changed from 0.5 to 0.999. In this way, the THD of the adaptive predictive filter output data is the smallest and the filter coefficient does not drift.

A leakage factor was set to = 0.999. The liquidity factor can be appropriately set within the range of greater than 0 and less than 1, and the smaller the factor, the faster the update of the adaptive predictive filter coefficient, but the greater the fundamental wave loss, resulting in a lack of stability of the signal. It is observed that the loss is small and the harmonic content is small, but the drift of the filter coefficient is likely to occur. Therefore, what is necessary is just to set it to the value which does not produce the drift of a filter coefficient in this range.

의 표류가 발생하 지 않도록 하는 개선이 가능하지만, 적응예측필터의 주파수 크기특성에 있어서 원하는 주파수대역에 손실이 발생한다. 이러한 손실을 보상하기 위해서는 기본파의 크기를 보정해야 한다.Introducing such a liquidity factor, filter coefficient vectors

Although it is possible to improve such that the drifting does not occur, a loss occurs in a desired frequency band in the frequency magnitude characteristic of the adaptive prediction filter. To compensate for this loss, the magnitude of the fundamental wave must be corrected.

Fundamental wave size corrector 435:

The magnitude loss of the fundamental wave according to the introduction of the leakage factor into the adaptive predictive filter 430 of the present invention is compensated by calculating the magnitude correction factor SF and multiplying the output value. The fundamental wave size loss can be found by directly measuring the fundamental wave size loss of the FIR filter. To measure this, two sinusoids of magnitude 1 and orthogonal are passed through the FIR filter as shown in Equation (8) (see FIG. 4). Two sinusoids are defined as vectors in each axis of the Cartesian coordinate system, and the magnitude of the vector sum of the two components is one. After passing the filter, SF is obtained by taking the inverse of the vector sum of two components, which is multiplied by the adaptive predictive filter output value when denormalizing.

(8)

(8)

In order to compare the performance of the reference signal generator 400 of the present invention with the above-described improvement with the conventional reference signal generator 100, the waveform of the steady state and the filter coefficient at that time through simulation by MATLAB Frequency characteristics were analyzed. The simulation considered 20 seconds as steady state. Table 1 shows the fundamental and harmonic content (%) of the load current used as an input signal for performance comparison.

<Table 1. Frequency components of load current hypothesis signal>

Degree size[%] A fundamental wave 100 5 Harmonics 22.6 7harmonics 10.5 11harmonics 7.3 13 Harmonics 4.7

The waveform of the load current composed of the fundamental and harmonics of Table 1 is shown in FIG. Based on this signal, the fundamental wave of the reference signal generator of the prior art and the present invention was simulated for 20 seconds and analyzed.

6 (b) and 6 (c) show the frequency response of the output of the adaptive predictive filter and the filter coefficient generated when the load current passes through the conventional reference signal generator. For about one second of simulation time, as shown in Figs. 6B and 6C, it is seen that the harmonic content generates a signal with a very low rate and the reference signal generator works well.

FIG. 7 shows a simulation result of 20 seconds with respect to the reference signal generator of the prior art, and it can be seen that severe distortion has been performed, which means that the reference signal generator does not operate properly anymore. This can be seen more clearly by the filter coefficient frequency response of.

8 shows the frequency response of the filter coefficient and the output of the adaptive prediction filter generated when the load current passes through the reference signal generator of the present invention. Table 2 shows the ratio of each frequency component of the output signal. The simulation was run for 20 seconds for equivalent comparison. Through these results, it can be seen that the adaptive predictive filter output signal of the present invention has a very low harmonic content and operates stably after a lapse of time. Finally, the adaptive predictive filter is optimized to operate as a bandpass filter that compensates for phase delay. Can be.

<Table 2 Frequency Components of Output Signals>

Degree size[%] A fundamental wave 100 5 Harmonics 0.53 7harmonics 0.245 11harmonics 0.1 13 Harmonics 0.075

9 is a simulation result carried out to check the operating state of the reference signal generator of the present invention when the load changes. It shows the output of the adaptive prediction filter inside the reference signal generator when the harmonic-containing input signal is input to the reference signal generator to increase the 20% or 40% step in 0.303 seconds and then decrease the step to its original size in 0.395 seconds. It can be seen that the reference signal generator of the present invention has a transient state of one cycle due to the delay element of the peak detector when the load step changes, but operates stably after one cycle.

10 schematically shows the configuration of a test apparatus for testing by implementing the reference signal generator of the present invention in hardware. For the purpose of evaluating the performance of the reference signal generator of the present invention and examining its practical applicability, a reduced model of a single phase active power filter having a nonlinear load composed of an RL load and a diode rectifier was fabricated and tested as shown in FIG. 10. The capacity of the active power filter was 1KVA, and the output stage was equipped with an LRC passive filter to reduce switching ripple. In addition, the RL load used a variable resistor to test the load variation. The circuit parameters of the whole system are shown in Table 3.

Table 3 System Circuit Constants

Circuit constant Power supply voltage 110 [V], 60 [Hz] inverter Filter L 6 [mH] Filter C 20 [㎌] Filter R 5 [Ω] DC-Link Capacitor 2200 [㎌] DC-Link Voltage 200 [V] Load condition Diode rectifier load R (Variable) 20 [Ω] ⇒ 16.6 [Ω] 16.6 [Ω] ⇒ 20 [Ω] L 50 [mH]

The algorithm constituting each element of the proposed reference signal generator is implemented in TMS320C31 DSP with floating-point arithmetic, and the current control is based on the deadbit current controller so that the inverter output current follows the reference current. Deadbit current control has been introduced in current control of power electronics since the 1990s and is a control method to obtain the deadbeat response mentioned in the control theory. The control variables used in this controller are line voltage, coupling reactor value, current inverter output current, and reference current. The delay of control is a fundamental disadvantage of this method. The dead-bit current controller used for the test in the present invention predicts an error at time n caused by the output of n-1 and compensates the error at time n and the error at time n-1 at the time n + 1. The control was carried out in a manner to improve the delay of the controller.

11 shows the operation of a reference signal generator based on the adaptive prediction filter of the present invention. Here, (a) of FIG. 11 is a load current, (b) is a fundamental wave component that is an interpolator output, and (c) is an output of a reference current signal generator. It can be seen that a harmonic reference signal with no phase delay is generated as an input.

12 illustrates an operation of an active power filter by a dead bit current controller. Figure 12 (a) shows the supply current, (b) shows the injection current of the active power filter, and (c) shows the load current. Total Harmonic Distortion (THD) of the load current is 25.3%, whereas THD of the compensated current is 5.93%.

The active power filter using the reference signal generator and the deadbeat current controller of the present invention can compensate for harmonics caused by nonlinear loads, although there is a slight transient in the power supply current. The slight transient that exists here occurs because the step rise is too steep to follow the control at the moment the load input current steps up. The steepness of the load current can be mitigated by reducing the inductance of the dc reactor.

In general, the load changes with time, so it is very important to verify that the proposed reference signal generator operates stably against sudden load changes. 13 shows the results of experiments when the load increased by 20% step and decreased by 20% step. (A) of FIG. 13 shows the fundamental wave component which is the interpolator output of the reference signal generator and the current on the compensated power supply side when the load increases by 20% step. As expected, it can be seen that the reference signal generator of the present invention operates well without serious transients even when the step load is varied.

The reference signal generator and the active power filter using the same according to the present invention can be modified and applied in various forms within the scope of the technical idea of the present invention and are not limited to the above preferred embodiment. In addition, the embodiments and drawings are merely for the purpose of describing the contents of the invention in detail, and are not intended to limit the scope of the technical idea of the invention, the present invention described above is common knowledge in the technical field to which the present invention belongs As those skilled in the art can have various substitutions, modifications, and changes without departing from the spirit and scope of the present invention, it is not limited to the embodiments and the accompanying drawings. And should be judged to include equality.

According to the present invention, it is possible to implement an improved reference signal generator and an active power filter using the same to improve the operation algorithm of the adaptive prediction filter in the reference signal generator.

In addition, it is possible to provide an adaptive prediction filter in which the drift phenomenon is corrected by introducing a leakage factor and / or adding a scale factor correction algorithm of the present invention, which can be applied to a reference signal generator.

Furthermore, according to the present invention, the system configuration is simplified, such as reducing the number of sensors by improving components such as a prefilter, a peak detector, and an adaptive prediction filter, and stably applying a reference signal without a phase delay even for a signal having a variable size. By extracting, it is possible to provide a reference signal generator that can ensure very good operating characteristics when used in an active power filter.

Through simulation and experiment, it was verified that the reference signal generator of the present invention can extract the reference signal without phase delay and use it in the active power filter, and its performance is also excellent. The proposed reference current signal generator can reduce the number of sensors by extracting the reference signal only by detecting the load current, and has the advantage of stable operation without phase delay even for the signal of varying size.

Claims (7)

Normalized Current Signal

And an input signal from which harmonic components of m order or more are removed from the normalized power signal.

Input signal, output signal

[here

In the FIR filter and the least-squares adaptive prediction filter for outputting the filter coefficient vector], To determine the filter coefficient vector H (n + 1) of n + 1 steps,

Use the relationship defined by here,

,

Is a data vector of the filter, and δ is a leakage factor value adjusted in the range of 0 <δ <1 to mitigate the drift of the filter coefficients. The method of claim 1, Adaptive prediction filter, characterized in that for compensating for the loss of the fundamental wave in the output signal generated by the introduction of the leakage factor. The method of claim 2, In order to compensate for the fundamental wave size loss, two orthogonal sinusoids of unit size are passed through a built-in FIR filter to obtain a scaling factor (SF) by taking the inverse of the vector sum of the two sinusoidal components as the inverse, and outputting the SF to the output. Adaptive prediction filter, characterized in that to multiply the signal. A reference signal generator comprising the adaptive prediction filter of any one of claims 1 to 3. The method of claim 4, wherein And a peak detector for selecting a maximum value of the input sample during one period of the fundamental frequency to normalize the current signal, wherein the peak detector comprises a low pass filter (LPF). The method of claim 4, wherein And a prefilter for removing harmonic components of the signal input to the adaptive prediction filter. An active power filter system comprising the reference signal generator of claim 4.

Images