1. Introduction
High voltage direct current transmission technology, based on the modular multilevel converter (MMC-HVDC), is widely used in the fields of AC grid asynchronous interconnection, the distributed new energy grid connection, urban load center, and island power supply [
1]. With the rapid development of MMC-HVDC transmission technology, the oscillation problem of the MMC-HVDC system has become increasingly prominent. Many different types of oscillation accidents have occurred worldwide, such as the blockage of the converter station and even the outage of the transmission system, resulting in serious economic losses [
2].
In recent years, the oscillation problem has gradually shown the trend of high-frequency. For example, when the North Sea wind farm in Germany is connected to the MMC-HVDC system, an intermediate frequency oscillation of between 250 and 350 Hz occurs [
3]. When the receiving-end MMC of the Luxi HVDC project is connected to the weak AC power grid, a high-frequency oscillation of approximately 1200 Hz occurs, causing a system outage [
4]. Moreover, 1600 Hz high-frequency oscillation occurred in the INELFE France–Spain grid interconnection project [
5]. High-frequency oscillation may excite the AC system to generate harmonics with large amplitudes, causing severe distortion of the AC voltage and AC current, resulting in the system facing the risk of lock-up and shutdown [
6]. Therefore, it is important to research the high-frequency oscillation mechanism and suppression method of the MMC-HVDC system for improving safety and stability. Moreover, there may also be high-frequency oscillation problems in the offshore wind power transmission system [
7,
8,
9]. The research on the high-frequency oscillation of the MMC-HVDC system is mainly based on the eigenvalue analysis method and the impedance analysis method [
10,
11].
The eigenvalue analysis method judges the stability of the system by analyzing the eigenvalues and eigenvectors of the system coefficient matrix and needs to establish a detailed state space model considering the characteristics of the time delay [
12], which is difficult to apply to the MMC-HVDC system with a complex structure.
In order to overcome the limitation of the eigenvalue analysis method, scholars have proposed the impedance analysis method. The author of [
13] compares the AC impedance models of two-level converters and MMC and points out that the internal dynamic characteristics of MMC can be ignored at high-frequency. The author of [
14] considers the delay characteristics of the MMC controller and proposes an analytical impedance model in a synchronous rotating coordinate system. The author of [
15] analyzes the influence of delay and system operating conditions changes on the system impedance characteristics and points out that increasing the delay will cause negative damping in the high-frequency band of the system, which makes the system more prone to oscillation and instability.
However, the above model involves a large number of complex matrix operations, which is difficult to meet the requirements of engineering applications. Therefore, there is currently no simplified impedance model of the MMC-HVDC system considering the characteristics of time delay.
On the other hand, for the high-frequency oscillation suppression of the MMC-HVDC system, related research generally includes optimizing the controller parameters [
16], optimizing the controller structure [
17], and additional damping control [
18]. The author of [
19] proposes a high-frequency resonance suppression strategy by adding a band-stop filter in the voltage feedforward link and designs the control parameters of the band-stop filter. The authors of [
20] proposes to add a second-order band-pass filter in the voltage feedforward control link and implement it on site to verify its effectiveness. The authors of [
21] proposes an optimization nonlinear voltage feedforward strategy with rounding function.
Adding a filter to the feedforward voltage link of the inner loop control is an effective method to improve the system damping. The author of [
19] proposed an active control strategy based on notch filter, which can effectively suppress the oscillation of the two-level converter in a specific frequency band. The author of [
20] proposed an active damping controller based on virtual flux linkage, which uses integral AC voltage as feedback to reduce the impact of control link delay on harmonic impedance. The author of [
21] proposed that the feedforward voltage is added with a notch link to compensate the phase at a specific frequency, which can reduce the impact on the performance of the control system. However, the AC model used is simple and the system resonance point is single, which cannot fully reproduce the actual situation. These active damping control methods can weaken the negative damping effect caused by control delay but cannot enhance the damping level of the system. In addition, they may cause new negative damping in other frequency bands of the system. For this reason, some scholars of [
22], on the basis of the feedforward voltage plus low-pass filtering link, propose that the control scheme where the additional voltage is superposed with the reference current after the damping controller can solve the problem of high-frequency oscillation in the full frequency band. However, due to the need to coordinate and balance multiple potential resonance points, the parameter design is complex, and the adaptability of the controller parameters needs to be investigated.
However, although the above methods can effectively suppress high-frequency oscillation in some cases, they may deteriorate the MMC impedance characteristics in other frequency bands.
In December 2018, a high-frequency oscillation of approximately 1.8 kHz occurred on the Hubei side during the open line test (OLT) at the Shizhou converter station of the South Pass of the Yu-E project, and then a low-pass filter was added to the voltage feedforward link to successfully suppress oscillation. However, during the OLT test on the Chongqing side, high-frequency oscillations of approximately 660 Hz and 700 Hz appeared in the system, respectively. At this time, the above-mentioned high-frequency oscillation suppression strategy failed, which eventually led to the blocking and tripping of the converter station [
23]. Therefore, the existing high-frequency oscillation suppression scheme does not respond to the needs of actual engineering. The suppression effect may not be obvious or invalid and may even cause oscillation of additional frequencies.
Therefore, this paper proposes a high-frequency oscillation suppression strategy based on nonlinear low-pass filtering, which can improve the full-band impedance characteristics of the flexible straight MMC-HVDC system. Firstly, the system is divided into blocks by the modular modeling methods, and the simplified high-frequency impedance model of MMC considering the time delay is derived. On the basis of this, the influence of the dynamic interaction between the control system and the harmonic disturbance on the impedance characteristics of the MMC-HVDC system is analyzed, and the essential reasons for the continuous high-frequency oscillation of the MMC-HVDC system are revealed. Finally, a high-frequency resonance suppression strategy is designed based on a nonlinear low-pass filter. The simulation model of the MMC-HVDC grid-connected system is built to verify the correctness of the simplified impedance model and high-frequency oscillation suppression strategy.
5. High-Frequency Oscillation Suppression Strategy Based on Nonlinear Low Pass Filter
From the analysis results of the high-frequency continuous oscillation mechanism, the main influencing factor of the high-frequency oscillation of the system is the control link delay, so optimizing the delay link is the best way to suppress high-frequency oscillation. However, for the system in which the delay link cannot be further optimized, additional damping control is needed to improve the negative damping characteristics of MMC-HVDC.
Therefore, in this section, based on the existing additional low-pass or band-stop filter control, an MMC-HVDC high-frequency resonance suppression strategy with nonlinear low-pass filter added to the voltage feedforward link is proposed. This method can better improve the negative damping characteristics of MMC-HVDC in the high-frequency band than the traditional additional low-pass filter control strategy.
5.1. Operation Principle of Nonlinear Low-Pass Filter
For MMC-HVDC based on MMC, the impedance characteristics of MMC can be used to replace the impedance characteristics of MMC-HVDC in the high-frequency band. Adding a low-pass filter to the voltage feed-forward link of the MMC control system can improve the impedance characteristics of MMC in the high-frequency band and block the oscillation frequency component in the grid feed-forward voltage, so as to realize the high-frequency resonance suppression of the interconnected system. The transfer function of the low-pass filter is:
where
ξ is the damping coefficient of the low-pass filter and ω
n is the center frequency of the filter.
The frequency response curve of the low-pass filter can be obtained from the transfer function of the filter, as shown in
Figure 13.
From the frequency response curve of the traditional low-pass filter, it can be seen that the low-pass filter can filter the frequency component above the cut-off frequency. By setting the cut-off frequency of the low-pass filter to less than the high-frequency oscillation frequency in MMC-HVDC, the high-frequency component in the feed-forward voltage of the AC grid can be filtered and the high-frequency resonance suppression can be realized.
The high-frequency problem can be effectively solved by using the low-pass filter. However, due to the inherent characteristics of the quality factor of the low-pass filter, the harmonics near the cut-off frequency cannot be completely filtered, which may lead to the transfer of resonant points. Therefore, special strategy design must be carried out. In this paper, based on the double closed-loop controller, the nonlinear low-pass filter is used for filtering in the inner loop voltage feedforward link. According to the fundamental level of AC voltage, different cut-off frequencies of the low-pass filter is used to further improve the effect of the low-pass filter and avoid the influence of harmonics on the voltage feedforward control. After adding the nonlinear cut-off frequency filter, the simplified control block diagram of MMC is shown in
Figure 14:
Iref represents the reference voltage of the current inner loop, Iout is the AC current, Uabc is the grid voltage, Gi is the PI controller of the current inner loop, KPWM is the modulation coefficient, and 1/(sL) is the simplified model of the main circuit.
5.2. Parameter Design of Nonlinear Low-Pass Filter
According to the transfer function and frequency response characteristics of the low-pass filter, it can be concluded that the main parameters of the low-pass filter design are cut-off frequency ωn and damping coefficient ξ0. The following describes the parameter design process of the nonlinear low-pass filter:
- (1)
Determination of center frequency and cut-off frequency of the low-pass filter. After the low-pass filter is put into operation, it is still necessary to ensure the flow of the voltage components of fundamental frequency and nearby frequencies, that is, the passband of the low-pass filter is required to contain the fundamental frequency 50 Hz, and the low-pass filter can also block high-frequency oscillation, so the center frequency can be set at 50 Hz, and then the cut-off frequency can be determined according to the selection result of the cut-off frequency screening module ωn.
- (2)
Initial value of damping coefficient ξ0 of the low-pass filter. The cut-off frequency and center frequency f0 of the low-pass filter jointly define the adjustment range of the low-pass filter. The initial value of the damping coefficient can be calculated according to the cut-off frequency and center frequency of the filter.
- (3)
Final value of damping coefficient ξn. When the center frequency is given, the greater the damping coefficient, the greater the adjustment degree of the low pass filter to the amplitude and phase characteristics of the impedance, and the larger the corresponding influence range. Therefore, on the basis of the initial value of the damping coefficient, further adjust the damping coefficient according to the requirements of the stability margin of the system and the dynamic adjustment ability of the control link to obtain the final value of the damping coefficient.
The specific resonance suppression process is shown in
Figure 15:
5.3. Effect of Nonlinear Low-Pass Filter on Impedance Characteristics of MMC
5.3.1. Analysis of MMC Impedance after Adding Nonlinear Low-Pass Filter
In order to further illustrate the principle of high-frequency oscillation suppression by adding the nonlinear low-pass filter to the voltage feedforward link, the impedance characteristics of MMC before and after adding the nonlinear low-pass filter to the voltage feedforward link are compared, as shown in
Figure 16.
From the comparison of the impedance characteristics of the MMC, it can be seen that the nonlinear low-pass filter added to the voltage feed-forward link can effectively improve the impedance characteristics of the MMC in the high-frequency band. The comparison of amplitude–frequency characteristics shows that the resonant peaks of MMC impedance at 2.2 kHz and 4.5 kHz are effectively suppressed. The comparison of phase–frequency characteristics shows that the negative damping phenomenon of the MMC in the high-frequency band is greatly improved, and the stability margin of the system is improved.
5.3.2. Performance Comparison of Different Filters Added to the Feedforward Link
In order to further illustrate the advantages of the additional band-pass filter proposed in this paper, the impedance characteristics of the MMC after adding the nonlinear low-pass filter, a band-stop filter, and a band-pass filter are compared as
Figure 17:
Band-pass and band-stop filter transfer functions are described in detail below:
Among ξ is the damping coefficient of the band-stop filter, and f0 is the center frequency of the filter.
- (2)
Band-pass filter:
Among ξ is the damping coefficient of the band-pass filter, and f0 is the center frequency of the filter.
According to the transfer function of the filter, the frequency response curves of the band-stop filter and the band-pass filter can be obtained, as shown in
Figure 18 and
Figure 19.
The parameters of the band-pass filter and band-stop filter are ξ as the damping coefficient and f0 as the center frequency of the filter. Among ξ, the damping coefficient is usually set to 0.707 and f0, the filter center frequency, is usually set to 50 Hz.
As can be seen from
Figure 16, the three filters can effectively improve the negative damping phenomenon of the MMC in the high-frequency band, but the performance of the three filters shows obvious differences in the middle frequency band (500~1000 Hz). After the low-pass filter and the band-stop filter, the MMC still has serious negative damping in the mid-frequency band, and the nonlinear low-pass filter can also suppress the negative damping phenomenon in the mid-frequency band well, which shows that the nonlinear low-pass filter can suppress full frequency resonance.
To further illustrate the advantages of the proposed additional nonlinear low-pass filter, the impedance characteristics of the MMC with an additional nonlinear low-pass filter and traditional low-pass filter are compared as follows:
It can be seen from
Figure 20 that both filters can effectively improve the negative damping phenomenon of the MMC in the high-frequency band, but the characteristics of the two filters show obvious differences in the frequency band from 500–1000 Hz. After the addition of the traditional band-pass filter, the MMC still has a serious negative damping phenomenon. By using the nonlinear low-pass filter, it can ensure that the phase angle of the interconnected system impedance is not more than 93°, and the phase angle difference of the system impedance is less than 180° in the whole frequency band. There is no impedance resonance matching point, indicating that the nonlinear low-pass filter has better resonance suppression ability than the traditional low-pass filter.
It is obvious from
Table 2 that the voltage feedforward damping control strategy with the nonlinear low-pass filter has a narrow negative damping band and low control complexity. Although the additional control link can also achieve full frequency resonance suppression, it has a high complexity. Therefore, the additional nonlinear low-pass filter in the voltage feedforward link has a better oscillation suppression effect than the traditional high-frequency oscillation suppression strategy.
6. Time Domain Simulation Verification
In order to verify the effectiveness of the high-frequency oscillation suppression strategy of the nonlinear low-pass filter, based on the parameters of the MMC converter station in a back-to-back MMC-HVDC project, an MMC-HVDC system is built in MATLAB/Simulink, and the parameters of the system are shown in
Table 1. The control system structure of converters are shown in
Figure 20.
The simulation results of the high-frequency resonance suppression of the MMC-HVDC system are shown in
Figure 20. When t = 1 s, the disturbance source with a frequency of 1.8 kHz is injected into the system, and the voltage amplitude of the disturbance source is 1% of the fundamental voltage amplitude. The system runs stably before the disturbance is injected. After the disturbance is injected, the three-phase voltage and three-phase current at the PCC point show different degrees of high-frequency oscillation. The Nyquist curve of the interconnected system is shown as the blue line in
Figure 20. The curve surrounds (−1,j0) points and the system is unstable, which is consistent with the simulation results.
The resonant frequency of the system is determined as high-frequency resonance. The main factor of high-frequency resonance is the delay link. In order to compensate the influence of the delay link, as shown in
Figure 21, when t = 1.1 s, the frequency resonance suppression strategy improved in this paper is put into the simulation, which adds the nonlinear low-pass filter to the voltage feedforward link. It can be seen from
Figure 20 that the voltage and current of the MMC system gradually returns to stable operation after the filter is put into operation.
The Nyquist curve of the system is shown as the orange line in
Figure 22, and the curve does not surround (−1,j0) points. The system is stable and consistent with the simulation results, which verifies the effectiveness of the proposed oscillation suppression strategy. It should be noted that the main influencing factor of high-frequency resonance is the delay link, and the suppression of high-frequency resonance can also be achieved by directly reducing the control system delay. Therefore, in practical engineering, the high-frequency resonance characteristics of the system can be improved by optimizing the control system delay.