JP6119297B2 - Periodic disturbance suppression control device - Google Patents

Description

  The present invention relates to an active filter control technique for suppressing harmonics of a system bus in a power converter connected to a power system bus.

  In recent years, attention has been focused on the use of distributed power sources by new energy such as smart grids, micro grids, solar power, wind power generation, etc., and the spread and expansion are expected in the future. Along with this, the load that becomes a harmonic generation source and the number of grid interconnections of the semiconductor power conversion system also increase, and there is a concern about harmonic disturbance in the distribution system. Since harmonics adversely affect other grid-connected devices, they are managed and specified by harmonic guidelines and the like, and various countermeasures against harmonic disturbances have been made by power active filters and the like.

  However, when a large number of unspecified distributed power sources are connected everywhere, the accumulated harmonics increase, the source is complicated, the distribution system is changed, the phase-advancing capacitor is switched, and the operation status is increased. System and load impedance fluctuations and system resonance characteristics fluctuations associated with the above are considered, and conventional countermeasures are likely to be insufficient from the viewpoint of harmonic compensation effect and control stability.

  In addition, considering the application and operation of active filters, there is a need to adjust control parameters locally so that a desired suppression effect can be obtained according to the installation location, and readjustment work is also required when the operation status changes .

  Based on the above background, the generalized periodic disturbance observer compensation method using the complex vector representation already proposed by the inventors of the present application has been expanded, and the transfer characteristic of the observer's real system can be estimated from the complex vector trajectory information of the periodic disturbance. Patent Document 1 is disclosed as a method for learning and correcting the reciprocal. In Patent Literature 1, a harmonic frequency component is extracted for each order in a power system active filter control system whose system current is a control target, and a frequency transfer function (detection) from a harmonic compensation command value of each order to a detected value. This is a method that automatically learns the harmonic current, which is periodic disturbance, and estimates and suppresses harmonic current, which is a periodic disturbance, and does not require prior system identification and is robust. A high power active filter control system can be constructed.

JP 2012-55418 A

  However, in the learning function in Patent Document 1, the criterion for determining whether it is the influence of model error or the influence of disturbance is not clear, and the stability and convergence of the learning function are improved by not looking at the divergence condition. There was room for improvement. There was also room for improvement in terms of periodic disturbance suppression response.

  An object of the present invention is to improve a learning function with a more stable and responsiveness and improve a periodic disturbance suppression response in a periodic disturbance suppression control apparatus.

  The present invention has been devised in view of the above-described conventional problems. One aspect of the present invention is to superimpose a periodic disturbance suppression command value on a command value of a power converter connected to a system bus of a power supply. A periodic disturbance suppression control apparatus for suppressing waves, which is determined based on a periodic disturbance detection unit that outputs a periodic disturbance to be suppressed as a periodic disturbance detection value of a DC component, and a transfer characteristic of a control system A signal obtained by multiplying the periodic disturbance detection value by a multiplier using the inverse of the transfer characteristic from the periodic disturbance suppression command value to the input signal detection value, and a signal obtained by adding only a detection delay to the periodic disturbance suppression command value The difference between the periodic disturbance estimator that estimates the periodic disturbance, the estimated periodic disturbance estimated by the periodic disturbance observer, and the periodic disturbance command value that suppresses the disturbance And calculate the periodic disturbance suppression command value. And the reciprocal of the transfer characteristic according to the harmonic amplitude, harmonic amplitude difference, harmonic vector displacement, amplitude error estimate, and phase error estimate calculated from the harmonic vector to be suppressed. A learning control unit that corrects, the learning control unit determines that the harmonic amplitude difference is continuously greater than or equal to a threshold value, and determines that the state is a divergence, and the harmonic vector movement amount is continuously equal to or less than an arbitrary threshold value. In the case of the divergence state, the learning function is turned ON, and the amplitude error command value or the phase error command value in the reciprocal of the transfer characteristic is updated based on the amplitude error estimated value and the phase error estimated value. Then, the reciprocal of the transfer characteristic is corrected based on the amplitude error command value, the amplitude initial value, the phase error command value, and the phase initial value. If the stagnation period exceeds the threshold value, the transfer characteristic is Reciprocal of The definitive amplitude error command value, and updated based on the amplitude error estimation value, the amplitude error command value, based on the amplitude initial value, and correcting the inverse of the transfer characteristic.

  As one aspect thereof, in the learning control unit, an amplitude error estimated value is calculated based on a ratio of a moving amount of the harmonic vector on the complex plane and a distance from the origin, and the phase error estimated value is calculated on the complex plane. The calculation is based on the angle formed in the origin direction of the upper harmonic vector locus.

  Furthermore, as one aspect thereof, the learning control unit turns off the learning function when the harmonic amplitude is less than or equal to the threshold, turns on the learning function when the harmonic amplitude is decreasing or diverging, and the harmonic amplitude is increasing. If the direction of divergence is not in the divergent direction, the learning function is turned off, and if the harmonic amplitude is greater than or equal to the threshold value, the learning function is turned on, and if the stagnation period of the harmonic vector has exceeded the threshold value, When the amplitude in the reciprocal is corrected, and the variation of the amplitude error estimated value and the phase error estimated value and the period when the error amount is greater than or equal to the threshold value have exceeded the threshold time, the reciprocal of the transfer characteristic is updated and the phase in the reciprocal of the transfer characteristic During the correction of the error command value, the amplitude error command value in the reciprocal of the transfer characteristic is not corrected.

  Further, as one aspect thereof, the learning control unit is characterized in that the amplitude initial value in the reciprocal of the transfer characteristic is set to a value of 1 or more and given as a gain.

  Further, as one aspect thereof, the learning control unit sets an initial amplitude value in the reciprocal of the transfer characteristic based on a phase error state.

  Further, as one aspect thereof, the periodic disturbance suppression control device is parallelized to suppress a multi-order harmonic.

  Furthermore, as one aspect thereof, the periodic disturbance suppression control device is used to suppress power source current harmonics.

  As another aspect, the periodic disturbance suppression control device is used to suppress power supply voltage harmonics.

  ADVANTAGE OF THE INVENTION According to this invention, in a periodic disturbance suppression control apparatus, it can improve to the learning function with more stability and responsiveness.

It is a conceptual diagram which shows an example of a power distribution system. It is a circuit block diagram which shows the control model of the parallel type active filter of a power supply current detection system. It is a basic lineblock diagram of an active filter control system. It is a control block diagram of the periodic disturbance observer with respect to the harmonic of an nth-order frequency component. Only n order frequency components of the d _n q _n coordinates is a diagram showing a control system of the focused periodic disturbance observer. It is a graph which shows the stable boundary condition of a model error. It is the figure which showed the behavior of the average value of n-th harmonic current detection value on the complex plane. It is the figure which showed the behavior of the average value of n-th harmonic current detection value on the complex plane. It is a figure which shows the behavior on the complex plane of a harmonic vector. 3 is a block diagram illustrating a learning control unit according to Embodiment 1. FIG. It is a logic circuit diagram showing an update flag generator. It is the graph which compared the presence or absence of the learning function with the response waveform and the vector locus. FIG. 10 is a block diagram illustrating a learning control unit according to Embodiment 2. 10 is a block diagram illustrating a learning control unit according to Embodiment 3. FIG. It is a figure which shows the periodic disturbance suppression control apparatus in Embodiment 4. It is a figure which shows the control system of the periodic disturbance observer in Embodiment 5.

  The present invention is an invention in which the control method and the learning function of current distortion suppression and voltage distortion suppression are improved, and the active filter device configuration described below is an example of the present invention. In the embodiment, the power supply current detection method in the parallel type active filter for the distribution system will be described as a representative example. However, other device configurations (for example, load current detection method, development to voltage detection method, or power supply voltage instead of current distortion) Even if the distortion is to be compensated, etc., the same control method can be used.

[Embodiment 1]
FIG. 1 is a conceptual diagram of a typical power distribution system existing in the past. In the distribution network for supplying electric power from the system power supply 1 to each of the consumers Feeder1, Feeder2,... It is installed.

  In the consumer Feeder 2, there are harmonic generation sources (harmonic loads) 2a and 2b, and fluctuations in the harmonic load current are assumed by the switch SW1, and fluctuations in the load impedance due to switching of the phase advance capacitor 3a by the switch SW2. Further, in order to consider the system impedance fluctuation of the consumer Feeder2 and the influence of other consumers when viewed from the active filter AF, it is assumed that the phase advance capacitor 3b is switched by the switch SW3 of the consumer Feeder1.

  In the first embodiment, as described above, a state change occurs in the distribution system in consideration of “harmonic load fluctuation”, “harmonic load impedance fluctuation”, and “system impedance fluctuation” that may occur in a normal distribution system. Even so, a control function and a learning function that stably realize harmonic suppression control are provided.

  FIG. 2 is a configuration example of a circuit configuration and a control device showing a control model of a power source current detection type parallel active filter AF. The actual impedance of the system 4 and the loads 2a, 2b,..., 2n (n is a natural number) is unknown, and the amount of generated harmonics also changes due to load fluctuations.

As shown in FIG. 2, the voltage at the grid connection point is v _S , the current at the grid connection point is i _S , the active filter current is i _AF , the active filter input current is i _AFi , the active filter DC voltage is V _dc , The load current (total) is defined as i _L. The inside of the active filter control system will be described later.

In addition, instead of detecting the current i _s at the grid connection point, the active filter input current i _AF and the load current (total) i _L are detected and added together in the controller so as to correspond to the power supply current i _S. Is possible.

  In the power distribution system, the order of the harmonics that are problematic beyond the guideline value is limited to, for example, the fifth order and the seventh order. Therefore, the frequency component of the specific order is extracted and compensated for each order.

  FIG. 3 shows a basic configuration diagram of an active filter control system as an example of the first embodiment.

First, an AC voltage phase θ is detected from a grid connection point voltage v _S by a PLL (Phase Locked Loop) 11. A dq axis orthogonal rotation coordinate system synchronized with the AC voltage phase θ as a reference phase is constructed, and current vector control is performed on the dq axis.

The d-axis current command value I ^* _d and the q-axis current command value I ^* _q of the periodic disturbance suppression command value are generated by a periodic disturbance observer (periodic disturbance suppression control device) 12 described later, and the d-axis current command value I ^* _d is added with a current I ^* _dc for controlling the active filter DC voltage _Vdc constant. The constant control of the active filter DC voltage V _dc is realized by the PI controller 13 to follow the active filter DC voltage command value V ^* _dc .

The detection value of the active filter input current i _AFi is converted into the d-axis current detection value Id and the q-axis current detection value Iq by the dq coordinate conversion unit 14, and the d-axis current command value I ^* _d and the q-axis current command value I, respectively. ^* Deviation from _q is taken, and the d-axis voltage command value V ^* _d and q-axis voltage command value V ^* _q are generated by the PI controllers 15d and 15q. These are converted into a three-phase voltage command value v ^* _AFi based on the AC voltage phase θ by the dq coordinate inverse conversion unit 16, and a PWM control gate control signal is generated by the triangular wave comparison PWM 17.

FIG. 4 is a control configuration diagram of the periodic disturbance observer 12 for the harmonics of the nth-order frequency component. In dq coordinate converter 21, to construct a suppression control system by extracting the n-th harmonic at d _n q _n coordinate axes synchronized with the n-th order frequency components, systems with d _n q _n coordinate conversion shown in (1) The interconnection point three-phase current i _S (= [i _Su i _Sv i _Sw ] ^T ) is converted into the _dn- axis component i _Sdn and q _n- axis component i _Sqn of the harmonic current vector.

d _n axis component i _Sdn, the q _n axis component i _Sqn, since the frequency components other than the n-th extracted as specified order, as a variation, LPF shown in (2) (Low-Pass Filter) 22d, a 22q Then, the d _n -axis current value I _Sdn and the q _n -axis current value I _Sqn which are the periodic disturbance detection values of the DC component synchronized with the d _n q _n rotation coordinates are extracted. Since the expression (2) is an example of the simplest LPF that prioritizes the simplicity of calculation, the order and type can be changed according to the situation such as noise at the time of pulsating component extraction. Alternatively, Fourier transform or the like can be used. Note that the periodic disturbance detection unit 30 is configured by the dq coordinate conversion unit 21 and the LPFs 22d and 22q.

Periodic disturbance observer 12, since it is constructed that contribute control system only in a specific frequency component, d _n q _n is the transfer characteristic P _n of the real system of coordinates, (3) d _n-axis component P as formula _It can be expressed by a one-dimensional complex vector where _dn is a real part and q _n- axis component P _qn is an imaginary part.

The transfer characteristic P _n of this actual system is determined from the harmonic suppression command value I ^* _n (= I ^* _dn + jI ^* _qn ) of the periodic disturbance observer 12 to the current detection value (input signal detection value) i _Sn (= i _Sdn + ji _Sqn ) N-th order frequency transfer characteristics, and not only the impedance characteristics to be controlled but also the peripherals related to control devices and controls other than disturbances of circuit characteristics such as inverter control characteristics, calculation delay, dead time, and current detection delay It becomes generalized including the transfer characteristics of the equipment (detector, etc.).

Here, in order to briefly describe the operation of the periodic disturbance observer 12, a control system focusing on only the n-th order frequency component of the d _n q _n coordinates is expressed as shown in FIG. The vector notation in FIG. 5 means a complex vector, and G _F (s) shown in Equation (2) is an LPF that functions for each of the real part and imaginary part components.

The basic operation is a model Q ^ _n (=) of the reciprocal Q _n of the transfer characteristic of the real system shown in the equation (4) from the current detection value (periodic disturbance detection value) I _Sn (= I _Sdn + jI _Sqn ) through the LPF 22. The actual system input current I ^ _n (= I ^ _dn + jI ^ _qn ) is estimated by the integrator 23 using Equation (5) using P ^ _n- ¹ ).

The actual system input current I _n, because it contains periodic disturbance current dI _n, (6) as shown in the expression, from the estimated value I ^ _n of the real system input current (5) in the adder 26, The periodic disturbance current estimated value dI ^ _n (= dI ^ _dn + jdI ^ _qn ) is estimated by subtracting the current command value I ^* _n (= I ^* _dn + jI ^* _qn ) via G _F (s) 25.

In the adder 27, the periodic disturbance current estimated value dI ^ _n in the equation (6) is subtracted from the periodic disturbance current command value dI ^* _n (= dI ^* _dn + jdI ^* _qn ) (usually 0) to obtain the periodic disturbance current dI. You can cancel _n .

FIG. 4 shows the basic configuration of FIG. 5 expanded to d _n q _n coordinate active filter control. FIG. 4 shows a harmonic suppression current command that can suppress periodic disturbances by a learning function using the reciprocal Q _{n of the} transfer characteristic. d _n-axis current command value I ^* _dn value, q _n-axis current command value I ^* _qn is obtained. Specifically, first, in the integrators 23da, 23db, 23qa, and 23qb, the product of the periodic disturbance current detection values I _sdn and I _sqn and the reciprocal Q _dn + jQ _qn of the transfer characteristic of the actual system is taken, and the adder 24d and 24q are added to calculate the actual system input current estimated values I ^ _dn and I ^ _qn .

Inverse Q _dn, Q _qn transfer characteristics of the actual system is the inverse of the transfer characteristic of the periodic disturbance suppression command value I ^* _d, the I ^* _q to the system interconnection point current (input signal detected value) i _s. Thereby, transfer characteristics such as phase delay can be canceled. The reciprocal of the transfer characteristic is corrected by the learning control unit 29 based on the harmonic current vectors i _Sdn and i _Sqn . The learning control unit 29 will be described later.

Next, disturbance is estimated. Disturbance is obtained by taking the deviation of two signals.
(1) those harmonic suppression current command value I ^* _dn, I ^* _qn passes the real system, the inverse Q _dn transfer characteristics of the real system, canceling the transfer characteristics of the real system over the product of the Q _qn ( I ^ _dn , I ^ _qn ).
(2) Harmonic suppression current command values I ^* _dn and I ^* _qn do not pass through the actual system and only LPFs 25d and 25q are applied.

The above (1) is a signal on which a disturbance on the actual system is superimposed, and the above (2) is just applying LPFs 25d and 25q to the harmonic suppression current command values I ^* _dn and I ^* _qn , and does not include the disturbance. Signal. By taking the difference between the two signals by the adders 26d and 26q, the periodic disturbance current detection values dI ^ _dn and dI ^ _qn can be obtained.

The adders 27d and 27q take deviations between the estimated periodic disturbance current values dI ^ _dn and dI ^ _qn and the periodic disturbance current command values dI ^* _dn and dI ^* _qn . Usually, the periodic disturbance current command value is set to 0 in order to suppress the disturbance to “0”.

The harmonic suppression current command values I ^* _dn and I ^* _qn are obtained by this calculation. Further, the harmonic suppression current command values I ^* _dn and I ^* _qn are subjected to LPF processing by LPFs 25d and 25q, and compared with the input current estimated values I ^ _dn and I ^ _{qn of} the actual system, and the periodic disturbance current estimated values Used to estimate dI ^ _dn and dI ^ _qn .

d _n q _n coordinates harmonic suppression current command value I ^* _dn, I ^* _qn is (7) as shown in the expression in the dq inverse coordinate transformation unit 28, the periodical disturbance suppression command value dq coordinates I ^* _d , I ^* _q to obtain a command value for the dq-axis current vector control system in FIG.

  Here, in the dq coordinate inverse transform unit 28, (n−1) θ is input because when the inverse transform is performed on the (n−1) th order, the signal of n times the fundamental wave becomes a direct current 2 This is because the axial component can be obtained.

[Influence of model error]
Next, the effect of model error is considered.

Since an actual power system has impedance variations and the like, the transfer characteristic P _n of the actual system is a time-varying parameter. Therefore, the influence of the model error Q ^ _n ≠ Q _{n which} is the reciprocal of the transfer characteristic P _n of the actual system on the stability of the periodic disturbance observer 12 will be considered.

In the control system of the d _n q _n coordinate shown in FIG. 5, the disturbance response transfer function from the periodic disturbance to the detected value is expressed by equation (8). However, an ideal condition in which other frequency components are removed by the LPF of the expression (2) is used.

Here, the amplitude error in the model Q ^ _n of the reciprocal Qn (= P ⁻¹ _n ) of the transfer characteristic of the real system is represented by _An (Amplitude error) (A _n > 0), and the phase error is represented by φ _n (phase error) ( as -π <φ _n ≦ π), to define the model Q ^ _n as (9), the model Q ^ is true value when the amplitude error a _n = 1, the phase error φ _n = 0.

Substituting equation (9) into equation (8) and rearranging results in equation (10). Here, the periodic disturbance response transfer function is C _n (s).

Since −w _f <0, −w _{f An} cos φ _n <0 in which the real part of the pole of C _n (s) is negative, the robust stability condition for the phase error φ _{n is} expressed by equation (11).

The amplitude error is A _n> 0, but does not affect the stability condition of a continuous system, the poles become stable direction greater the value of the true value A _n = 1 rather than amplitude error A _n. Although the representative pole in A _n > 1 is determined by the cutoff frequency w _f of the LPF, it can be controlled by A _n > 1 as a kind of observer gain for the purpose of improving the quick response. However, since it is necessary to avoid computation dead time and algebraic loop associated with digital control, the robust stability condition is considered in the transfer function (12) in the discrete system.

When G _F (s) in the equation (2) is bilinearly transformed as in the equation (13) at the operation cycle T _S and substituted into the equation (12), the characteristic equation is solved to obtain the equation (14).

If all the poles of the discrete system of equation (14) are arranged in the unit circle, it becomes stable, but since it is difficult to obtain an algebraic solution, for example, the cutoff frequency w _f = 2π [rad / s], the calculation cycle When numerically determine the robust stability as T _S = 100 [μs] for amplitude error a _n and the phase error phi _n, stable boundary conditions is obtained as shown in FIG.

If there is no phase error phi _n, stability margin is maximized with respect to the amplitude error A _n. As the phase error φ _n increases, the stability decreases and becomes unstable outside the range of equation (11). In the stable region stable region, the amplitude error _An can be increased to improve the speed response of periodic disturbance suppression. However, in actual operation, the model error variation due to system impedance variation etc. is taken into account, and a sufficient stability margin is provided. It is also necessary to secure and set.

  From the above, it can be seen that the periodic disturbance observer 12 is a control system having robustness against model errors. Also, unlike the approach of improving the harmonic current detection and control calculation delay to increase the stability margin, a robust control system is built by modeling a series of transfer characteristics of the control system with a generalized one-dimensional complex vector. There are advantages you can do.

However, for example, under severe conditions in which the amplitude / phase characteristics change greatly due to fluctuations in the system resonance / antiresonance point, the model error may fall into the unstable region shown in FIG. Further, (10) as shown in formula, because the range of the amplitude error A _n <1 representative pole is dependent on the characteristics of the amplitude error A _n, has a problem that leads to the deterioration of the cycle disturbance suppressing response performance . The purpose of the first embodiment is to propose a generalized periodic disturbance observer compensation method including means for learning and correcting a model error that becomes an unstable region as a countermeasure against these problems.

[Model learning function]
n-th harmonic current detection value i mean value of _{Sn i¯ Sn} behavior (described later) shows an example drawn in the complex plane in FIG. In Figure 7, the left side I _Sdn, the time response waveform of I _Sqn, right indicates the complex vector locus of i¯ _Sn. Suppression control was started at time 0.5 [s], and points were plotted every 0.1 [s] for easy understanding of the passage of time. When this harmonic current vector converges to the origin, it means that the harmonic current is suppressed.

When there is no model error, the harmonic current vector linearly converges to the origin as shown in FIG. When there is a phase error φ _n , the stable region converges while drawing a circle at an angle corresponding to the phase error amount with respect to the origin as shown in FIG. 7B, and is an unstable region outside the range of the expression (11). Then, it diverges while drawing a circle as shown in FIG.

Next, vector trajectories for amplitude errors of 0.5 and 2 with reference to the phase error of 60 [deg] in FIG. 7B are shown in FIGS. 8A and 8B, respectively. Figure 7 compared to the case where there is no amplitude error of A _n = 1 in (b), A _n <convergence to 1 in the origin is delayed, A _n> in 1 convergence can be seen that quickened.

As described above, the model error and vector locus are related, for example, the convergence of the amplitude error A _n is the origin, the phase error phi _n is related to the angle between the origin direction. Focusing on this point, means for detecting the model error from the geometric information of the vector locus will be described below.

The learning control in the first embodiment is an auxiliary function for the purpose of preventing deterioration and destabilization of the control performance of the periodic disturbance observer 12, and therefore does not require high-speed computation as compared with the harmonic suppression control period. . For example, considering the calculation load of the harmonic current vector locus, the learning control period T _L = 20 [ms] (50 Hz system fundamental wave period) is set with respect to the calculation period T _S = 100 [μs] of the periodic disturbance observer. . Further, the average value _Sn harmonic current vector i _Sn in the periodic T _L as (15).

In order to examine the time response of the harmonic current vector on the complex plane, the amplitude R _n and the phase ψ _n on the polar coordinates are defined as in the equation (16). Further, (17), (18) and the origin direction vector R _n, harmonic movement vector L _n, and the amplitude R _n the angle [psi _n harmonic vector L _n is defined as in FIG. 9, the short time Assuming that the harmonic vector movement amount ΔL _n , the amplitude change amount ΔR _n , and the phase change amount ΔΨ _n in ΔT, the equations (19) and (20) are established from the geometric relationship on the polar coordinates.

Next, when the inverse Laplace transform of the periodic disturbance response to I _Sn when fed a step disturbance input to the dI _n in FIG. 5, is obtained time response equation (21) below. However, as the minute time [Delta] T = learning control period T _L, and T _L a dead time according to the average calculation i¯ _Sn.

Here, when the equation (21) is compared with the equation (16), the origin direction vector R _n and the phase ψ _n on the polar coordinates can be expressed as the equations (22) and (23).

By substituting the equations (22) and (23) and the differences (24) and (25) in the learning control period T _L into the equations (19) and (20), equations (26) and (27) are obtained.

Further, (26) by dividing equation in formula (22) and rearranging terms amplitude error A _n (28) where the relationships are obtained.

From the above analysis results, the following relationship is established between the locus of the harmonic current vector on the complex plane and the model error.
From equation (27), the phase error φ _n is equal to the angle ψ _n that the harmonic current vector locus on the complex plane makes with the origin direction.
- from (28), the amplitude error A _n can be detected from the ratio of the distance R _n from the amount of movement [Delta] L _n and the origin of the harmonic current vector.

  However, the actual model error is detected by performing arbitrary filter processing, paying attention to the influence of harmonic characteristic fluctuations other than step disturbance and observation noise.

In the first embodiment, by using the estimated value _Sn harmonic current vector, (28) the amplitude error estimation value A ^ _n from the equation (20), the phase error estimate phi ^ _n from (27) , Respectively, as estimated by the equations (29) and (30). Here, the suffix [N] means the Nth sampling, and [N-1] means the value one sample before. Further, R _n [N], ΔR _n [N], ΔL _n [N] are obtained by equations (31) to (33).

A method for automatically correcting the model error based on the model error estimated values A ^ _n and φ ^ _n obtained by the equations (29) and (30) will be described. FIG. 10 is a block diagram of the learning function control unit 29. In the average value calculating unit 41 from equation (15) obtains the harmonic current vector average value I _Sn learning control period T _L on the basis of the harmonic current vector i _Sn, the vector parameter model error calculating unit 42, ( 29) to (33), the estimated amplitude error value A ^ _{n [N]} , the phase error estimated value φ ^ _{n [N]} , the polar coordinate amplitude Rn _[N] , the amplitude change amount (harmonic amplitude difference) ΔR _{n [N]} , the amount of movement of the harmonic current vector ΔL _{n [N]} is calculated. Next, as shown in Table 1, the flag determination units 43a to 43c generate flags indicating the states of the amplitude error estimated value A ^ _{n [N]} and the phase error estimated value φ ^ _{n [N]} .

Flags f _A and f _φ for updating the reciprocal model Q ^ n of the transfer characteristic of the actual system are generated in the update flag generator 44 as shown in FIG. 11 using the above flags.

  As shown in FIG. 11, the update flag generation unit 44 includes AND circuits 51 to 54, 56, 60, OR circuits 55, 59, a NOT circuit 62, and elapsed time determination units 57, 58, 61.

  In order to correctly perform model error estimation and model update, it is necessary to distinguish whether the change in the harmonic current vector locus is due to system fluctuation or disturbance fluctuation. In other words, it is desirable not to update the model carelessly for changes in the vector trajectory due to disturbance fluctuations.

Therefore, in the first embodiment, the learning algorithm is generated mainly paying attention to the following points.
(1) When the harmonic amplitude R _n is equal to or less than the arbitrary threshold value, it is determined that the periodic disturbance observer 12 maintains robustness and is normally suppressed, and the learning function is turned off.
(2) While the harmonic amplitude R _n is decreasing or diverging, it is determined that the influence of the model error can be estimated, and the learning function is turned on.
(3) If the harmonic amplitude R _n is increasing but not in the divergence direction, it is determined that the suppression of the increase in disturbance is effective, and the learning function is turned off.
(4) When the harmonic amplitude R _n becomes _equal to or greater than an arbitrary threshold (for example, 5%), it is determined that the delay in detection of the divergence state due to the model error is a main factor, and the learning function is turned on.
(5) If the harmonic vector is stagnant, the error cannot be estimated. In this case, it is considered that either (1) the model amplitude is extremely small with respect to the true value, or (2) the suppression is effective against the ramp-like disturbance change, and the stagnation period has passed T _st or more. It is determined that (1), and the amplitude error is corrected.
(6) Monitor the variation of the error estimation value and the error amount in consideration of the influence of disturbance fluctuation. The model is updated when the amplitude and phase errors can be detected stably while T _A and T _{φ are} greater than, respectively.
(7) Prioritize phase correction important for stabilization. Amplitude correction during phase correction generates an estimation error and is not applied at the same time.

(1) is determined by the flag f _ut and the AND circuit 51 in the update flag generator 44 shown in FIG. Similarly, (2) is determined by the flags f _ΔR , f _div , AND circuits 53, 54, and OR circuit 55, and (3) is determined by the flags f _ΔR , f _div , AND circuits 53, 54, OR circuit 55. , (4) is determined by the flag f _ot , AND circuit 52, (5) is determined by the flag f _st , duration determination unit 58, and (6) is the flag f _ssA , f _ssφ , f _leA , f _leφ , The determination is made by the AND circuits 56 and 60 and the duration determination units 57 and 61, and (7) is determined by the NOT circuit 62.

In the phase error update unit 45a and the amplitude error update unit 45b, when the outputs of the update flags f _A and f _φ are “1”, the amplitude error command value A ^* _n and the phase error command value φ ^* _n of the model are estimated as amplitude errors. The value A ^ _n and the phase error estimated value φ ^ _n are updated and output. Then, the amplitude error estimated value A ^* _n and the phase error estimated value φ ^* _n are respectively divided from the amplitude initial value A ^* _on and subtracted from the phase initial value φ ^* _on . Based on the corrected model amplitude and phase, as shown in FIG. 10, reciprocals Q ^ _dn and Q ^ _qn of the transfer characteristics of the actual system are generated and updated.

12, the amplitude error A _n = 0.5, the phase error φ _n = 5π / 9 (out of robust stability range of the periodic disturbance observer), is an example of comparison of the presence or absence of learning in response waveform and vector locus. In FIG. 12A, the model learning function is OFF, and diverges in an unstable region of the periodic disturbance observer 12. On the other hand, in FIG. 12B, the model learning function is ON, and the model phase error φ _n is first detected and corrected to be correctly suppressed in the convergence direction. Subsequently, it can be confirmed that the amplitude error _An is corrected and the convergence speed is improved. It has been confirmed that the learning function works effectively even under various operating conditions such as system fluctuation and disturbance fluctuation.

As described above, according to the first embodiment, the suppression control is stabilized even in the region beyond the robust stability range of the periodic disturbance observer 12 by the automatic learning function of the reciprocal Q _n of the transfer characteristic of the real system. it can.

(29), the amplitude error A _n model Q ^ _n periodic disturbance observer 12 on the basis of the (30) equation, estimates the phase error phi _n, 10, the state of the harmonic vector changes by 11 and Table 1 Taking the above into consideration, the above model error is corrected under conditions and timings suitable for those states.

In addition, according to the first embodiment, by adding a “divergence condition” (f _div ) that is not considered in Patent Document 1 to the learning condition, the divergent behavior in the case of control instability due to a model error can be quickly obtained. It can be detected and corrected. Similarly, by adding “stagnation state” (f _st ) not sufficiently considered in Patent Document 1 to the learning condition, the reciprocal Q _n of the transfer characteristic of the real system in the stagnation state is caused by an excessive error. It is possible to determine whether it is stagnant due to fluctuations in periodic disturbances or not, and the model can be corrected only in the case of a stagnant state due to a model error.

  Furthermore, in Patent Document 1, the correction of the model amplitude error is performed in a stepwise correction such as X times or 1 / X times, whereas in the first embodiment, the model amplitude error is estimated based on the model amplitude error estimated value of Equation (29). Since the correction is made to a value close to the true value of the actual harmonic impedance, the amplitude error of the model can be corrected more quickly. As a result, when the phase error is small, the responsiveness of harmonic suppression is improved, and when the phase error is large, the divergence rate is reduced.

[Embodiment 2]
FIG. 13 shows the configuration of the learning control unit that is the reciprocal of the transfer characteristic of the real system in the second embodiment. Other basic control configurations are the same as those in the first embodiment, and thus the description thereof is omitted.

(8) - (14) as determined by the equation and Figure 6, within the scope of the robust stability condition periodic disturbance observer 12 has, dare overall direction the amplitude error A _n of the inverse Q transfer characteristics of the real system (A _By setting _n > 1), harmonic response is improved.

The initial amplitude A ^* _on is normally set to true value A ^* _on = 1 in order to eliminate model errors, but by setting A ^* _on > 1, the inverse number Q ^ _dn , Q of the transfer characteristic of the real system is set. ^ Given as a gain for _qn , the response speed can be improved within the range of robust stability conditions.

Further, with respect to the result of the amplitude error estimated value A ^ _n estimated by the equation (29), the amplitude initial value A ^* _on is used as a command value as shown in FIG. 11 in order to improve the harmonic suppression control response performance. It gives to the flag determination part 43c. The flag determination unit 43c evaluates an error between the initial amplitude value A ^* _on and the estimated amplitude error value A ^ _n to generate an amplitude error evaluation flag _flleA .

  According to the second embodiment, in addition to the operational effects of the first embodiment, the suppression control responsiveness can be improved by setting the initial amplitude value to be larger than the true value.

[Embodiment 3]
In the second embodiment, the initial amplitude value A ^* _on is set to A ^* _on > 1, but as shown in FIG. 6, the robust stability condition of the periodic disturbance observer 12 increases as the phase error φ _n increases. the stability margin of the amplitude error a _n decreases Te. Therefore, if the amplitude initial value A ^* _0n is set large without considering the phase error φ _n , there is a risk of impairing control stability.

Therefore, in the third embodiment, as shown in FIG. 4, the initial amplitude value A ^* _on is determined using the state of the phase error command value φ ^* _n .

According to the third embodiment, the model amplitude command value generation unit 46, based on a robust stability condition as shown in FIG. 6 determined in advance by numerical analysis, receives an initial amplitude value A with the phase error command value φ ^* _n as an input. ^* _{Turn on} into a table. Further, it may be implemented as a simple mathematical expression / function. Thereby, in addition to the effects of the first and second embodiments, the initial amplitude value A ^* _on becomes variable according to the state of the phase error command value φ ^* _n , and the suppression control response can be effectively suppressed without impairing the control stability. Can be improved.

[Embodiment 4]
In the fourth embodiment, the control configuration of the periodic disturbance observer 12 with a learning function is parallelized as shown in FIG. 15, and a plurality of harmonic components (for example, −5th order, 7th order,..., Nth order). Component).

The inside of the periodic disturbance observer 12 of each order is the same as that shown in FIG. The d-axis and q-axis components I ^* _d5 , I ^* _d7 , ..., I ^* _dn , I ^* _q5 , I ^* _q7 , ..., I ^* _qn of each harmonic compensation value are added together to suppress periodic disturbance. The command value is d-axis current command value I ^* _d and q-axis current command I ^* _q .

  According to the fourth embodiment, in addition to the functions and effects of the first to third embodiments, it is possible to simultaneously suppress the harmonic components of multiple orders.

[Embodiment 5]
In the first to fourth embodiments so far, the purpose was to suppress power supply current harmonics, but in the fifth embodiment, as shown in FIG. 16, power supply voltage harmonics are detected and the same suppression control is performed. Is.

For example, increases as the frequency transfer characteristics of the harmonic impedance from the active filter compensation current i _AF until the power supply current i _s detected fall into anti-resonance state can not be compensated with sufficient accuracy, unnecessary a compensation current value Resulting in. In such a state, the voltage detection method may be able to detect with higher accuracy.

The basic control configuration of the fifth embodiment is the same as that of the first to fourth embodiments so far, but the period disturbance observer 12 is changed from the power supply current detection i _S to the power supply voltage detection v _S. Therefore, the transfer characteristic P _n of the actual system used by the periodic disturbance observer 12 corresponds to the frequency transfer characteristic from the compensation current command values I ^* _d and I ^* _q to the voltage detection v _S.

  According to the fifth embodiment, in addition to the effects of the first to fourth embodiments, voltage harmonics at the grid connection point can be suppressed.

12 ... Periodic disturbance observer (periodic disturbance suppression control device)
23da, 23qa, 23qb, 23db ... integrator 29 ... learning control section 30 ... periodic disturbance detection unit Q _a, the inverse of the transfer characteristic of the Q _b ... real system _{i Sn (= i Sdn + ji} Sqn) ... current detection value (input Signal detection value)
I ^* _d, I ^* _q ... periodic disturbance suppression command value I _Sdn, I _Sqn ... periodic disturbance detection value _{dI ^ n (= dI ^ dn} , dI ^ qn) ... periodic disturbance current estimated value R _n ... harmonics Amplitude ΔR _n ... Harmonic amplitude difference ΔL _n ... Moving amount of harmonic vector A ^ _n ... Amplitude error estimated value φ ^ _n ... Phase error estimated value

Claims (8)

A periodic disturbance suppression control device that suppresses harmonics by superimposing a periodic disturbance suppression command value on the command value of the power converter connected to the power system bus,
A periodic disturbance detection unit that outputs the periodic disturbance to be suppressed as a periodic disturbance detection value of a DC component;
A signal obtained by multiplying the periodic disturbance detection value by an integrator using the reciprocal of the transmission characteristic from the periodic disturbance suppression command value determined based on the transmission characteristic of the control system to the input signal detection value, and the periodic disturbance suppression A periodic disturbance estimator that estimates a periodic disturbance by taking a difference between the command value and a signal obtained by adding only a detection delay;
An adder that calculates the periodic disturbance suppression command value by taking a deviation between the periodic disturbance estimated value estimated by the periodic disturbance observer and the periodic disturbance command value that suppresses the disturbance;
A learning control unit that corrects the reciprocal of the transfer characteristic according to the harmonic amplitude, the harmonic amplitude difference, the amount of movement of the harmonic vector, the amplitude error estimated value, and the phase error estimated value calculated from the harmonic vector to be suppressed And comprising
The learning control unit
When the harmonic amplitude difference is continuously greater than or equal to a threshold, it is determined as a divergence state, and when the harmonic vector movement amount is continuously equal to or less than an arbitrary threshold, it is determined as a stagnation state,
In the case of the divergence state, the learning function is turned ON and the amplitude error command value or the phase error command value in the reciprocal of the transfer characteristic is updated based on the amplitude error estimated value and the phase error estimated value. Based on the initial amplitude value, phase error command value, and initial phase value, the inverse of the transfer characteristic is corrected,
In the case of the stagnation state, when the stagnation period has exceeded the threshold value, the amplitude error command value in the reciprocal of the transfer characteristic is updated based on the amplitude error estimated value, and based on the amplitude error command value and the amplitude initial value, A periodic disturbance suppression control device that corrects the reciprocal of the transfer characteristic. In the learning control unit,
An amplitude error estimate is calculated based on the ratio of the distance from the origin to the amount of movement of the harmonic vector on the complex plane,
2. The periodic disturbance suppression control device according to claim 1, wherein the phase error estimation value is calculated based on an angle formed in the origin direction of the harmonic vector locus on the complex plane. The learning control unit
If the harmonic amplitude is below the threshold, turn off the learning function,
While the harmonic amplitude is decreasing or diverging, turn on the learning function,
If the harmonic amplitude is increasing but not in the direction of divergence, turn off the learning function,
If the harmonic amplitude exceeds the threshold, turn on the learning function,
When the stagnation period of the harmonic vector has exceeded the threshold, the amplitude in the reciprocal of the transfer function is corrected,
When the amplitude error estimated value, phase error estimated value variation or error amount exceeds the threshold time, and the threshold time has elapsed, the reciprocal of the transfer characteristic is updated.
3. The periodic disturbance suppression control device according to claim 1, wherein the amplitude error command value in the reciprocal of the transfer characteristic is not corrected during the correction of the phase error command value in the reciprocal of the transfer characteristic. The learning control unit
The periodic disturbance suppression control apparatus according to any one of claims 1 to 3, wherein an initial amplitude value in the reciprocal of the transfer characteristic is set to 1 or more. The learning control unit
5. The periodic disturbance suppression control device according to claim 4, wherein an amplitude initial value in an inverse number of the transfer characteristic is set based on a state of a phase error command value.   6. The periodic disturbance suppression control device according to claim 1, wherein the periodic disturbance suppression control device is parallelized to suppress harmonics of a plurality of orders.   The periodic disturbance suppression control apparatus according to any one of claims 1 to 6, wherein the periodic disturbance suppression control apparatus is used to suppress power supply current harmonics.   The periodic disturbance suppression control apparatus according to any one of claims 1 to 6, wherein a power supply voltage harmonic is suppressed using the periodic disturbance suppression control apparatus.

Images