Research and Application of Combined Reactive Power Compensation Device Based on SVG+SC in Wind Power Gathering Area

Abstract

With the increasing proportion of wind power access year by year, it brings many challenges to the voltage stability of power systems. In order to maintain the stability of the voltage in the power grid, it is impossible to take into account the regulation ability and economy when a single reactive power compensation device is installed. In this paper, a combined reactive power compensation device was installed, which is composed of a static var generator (SVG) and a parallel capacitor bank. The SVG has the characteristics of fast and smooth adjustment, and the application of the capacitor bank reduces the overall investment cost and has a great economy. The modal analysis method was used to find the optimal installation position for the reactive power compensation device. The improved particle swarm algorithm was used to optimize the capacity of the optimal reactive power compensation device to ensure the best performance of the compensation device. Finally, by formulating the control strategy of the combined reactive power compensation system, the reliable switching of the compensation device is controlled. The PSCAD simulation software was used to model the power grid in the Hami area, and six different configuration programs were set for static voltage stability simulation verification and three different configurations. The program was simulated and verified for transient voltage stability, and comparative analysis showed that the proposed method was correct, which strongly supports the voltage stability of the region and meets the demand of reactive power compensation of the power grid. This provides a good reference program for other wind power gathering areas.

1. Introduction

Due to the reverse distribution of energy and load in China, large-scale thermal power and clean energy need to be transmitted to the receiving end grid over long distances and across regions, while the ultrahigh voltage direct current (UHV DC) transmission project is responsible for the arduous mission of Xinjiang power transmission [1,2]. Since the operation of ±800 kV Tianzhong DC, the wind power in the Hami area has been continuously transported to the power load center. However, the large-scale wind power grid connection has also brought voltage fluctuations and other influences to the region, which has severely challenged the stability of the power grid [3,4,5,6]. In recent years, the rapid development of the static reactive power generator SVG device represents the latest reactive power compensation system. It has a rapid response, wide operating range, and can suppress voltage flicker, but its complex structure and high cost make it difficult to be used on a large scale. Therefore, it is of high practical significance to study a new type of combined reactive power compensation device which would have both the cost advantage of traditional reactive power compensation device SC and the continuous smooth adjustment ability of SVG device.

According to the current research status, the authors of [7] proposes using SVC to replace the conventional reactive power compensation device, and rationally configuring the capacity of SVC to further ensure the stability of bus voltage. In order to evaluate the voltage support, the dynamic reactive power support is studied by the comprehensive analysis of time domain and frequency domain. Related scholars proposed that in the process of voltage static stability research, the corresponding reactive power compensation device should be added to the weak point of voltage, which can basically meet the requirements of wind power delivery in the Hami area to a certain extent. The authors of [8] put forward the optimization measures to install the corresponding series and parallel reactive power compensation devices on the top of the network channel, and carried out relevant research, analysis, and demonstration. The authors of [9,10] conducted a comparative analysis of the effect of reactive power compensation on the medium-voltage side compensation and low-voltage side of the UHV transformer. Although the operation of the UHV DC project was guaranteed, the DC near-region voltage was not analyzed. The authors of [11] rationalized the configuration of the capacity through SVC and conventional capacitor banks in order to minimize the total investment, but did not consider the best performance of the compensation device. In [12], the reactive power capacity of the large-scale wind power grid-connected system was calculated, and the recommended and unrecommended operation modes were analyzed. The authors of [13] studied the structure, coordinate control method, and parameter design of the hybrid system proposed for active power injection and non-active power (reactive, harmonic, and unbalanced power) compensation. The proposed hybrid system consists of a static var compensator (SVC) and a capacitively coupled grid-connected inverter (CGCI) (SVC//CGCI). The authors of [14] analyzed the transient voltage stability of the wind farm connection point, and created the transient voltage margin by formulating a reactive power compensation strategy. The authors of [15] optimized the configuration of wind and fire bundling capacity and DC drop point through a quantum particle swarm algorithm. The authors of [16,17] proposed a new type of reactive power compensation device or application scenario, but the scope of the application is limited. The authors of [18,19] proposed a combined reactive power compensation device, and optimized its topology and control strategy, but did not optimize the parameters. The existing voltage control methods have shortcomings in terms of adjustment range and response speed. The authors of [20,21,22,23,24] proposed a corresponding dynamic coordinated control strategy to improve the support capability of the voltage. However, the selection of dynamic reactive power planning address based on trajectory sensitivity has been introduced in detail, and the corresponding characteristics and effects of reactive power compensation of two commonly used dynamic reactive power compensation devices, SVC and STATCOM, have been fully compared and analyzed.

The main contributions of this paper are as follows:

• Using the advantages of fast response and strong economy of the combined reactive power compensation device, a control strategy for the reactive power compensation system based on the wind farm is formulated. This strategy not only avoids the frequent switching of capacitor banks but also guarantees the linear demand of the system for reactive power.
• The active power margin and modal numerical analysis of the wind power gathering area in the Hami area show that the Mahuanggou east wind farm, located in the Santanghu wind area, is the weak node of voltage in this area, so this area is the best place to install reactive power compensation devices.
• The improved particle swarm algorithm is used to optimize the capacity of the optimal reactive power compensation device, which ensures the best performance of the compensation device.
• Using PSCAD simulation software to model the power grid in the Hami area, six different reactive power compensation device configuration programs are set up for static voltage stability simulation verification, as well as three different reactive power compensation device combination programs for transient voltage stability simulation verification.

In summary, in order to solve the problem of weak voltage support in wind power gathering areas, considering the continuity and smoothness of regulation in reactive power compensation, and taking into account the economy of cost, this paper proposes a new combined reactive power compensation device composed of SVG and a capacitor bank. Taking the Tianzhong DC near-area power grid as an example, we analyzed and derived the voltage weakness area of the grid, used the improved particle swarm algorithm to find the optimal capacity of the compensation device, and finally verified the effectiveness of the method by setting up four configuration programs, which have guiding significance for the power grid in this area.

2. Research the Basic Characteristics and Compensation Strategy of SVG

SVG uses fully-controlled switching devices to form a self-commutation inverter, and then uses a small-capacity energy storage element as an auxiliary to perform the corresponding reactive power compensation work, so it can realize the function of instantaneous reactive power compensation [25,26,27].

The control method of SVG is mainly composed of two layers: outer loop control and inner loop control. The outer loop control mainly includes the following parts: the reactive power control part, the DC voltage control part, and the grid connection point voltage control part; its main function is to generate d, q axis current reference values i_dref, i_qref.

The DC voltage control part is used to further compare the DC side voltage U_dc and its reference value voltage U_dcref. The voltage value is also used to generate the d-axis current reference value i_dref, which is used to further stabilize the DC side voltage U_dc, and i_dref satisfies the following equation.

$$i_{\mathrm{dref}}=(K_{\mathrm{dp}}+\frac{K_{\mathrm{di}}}{s})(U_{\mathrm{dcref}}-U_{\mathrm{dc}})$$

(1)

In Formula (1), K_di and K_dp represent the integral coefficient and proportional coefficient of the d-axis PI controller, respectively.

The control method of the reactive power and grid connection point voltage is to compare the reactive power output in the grid connection point voltage U_pcc or SVG with the reactive power reference, and to obtain the q-axis current reference value i_qref. The voltage of the stable grid-connected point is determined through the reference value, or the reactive power required for the output with the reference value is directly adjusted [28,29]. The current reference value satisfies the following equation:

$$i_{\mathrm{qref}}=(K_{\mathrm{qp}}+\frac{K_{\mathrm{qi}}}{s})(U_{\mathrm{PCCref}}-U_{\mathrm{PCC}})$$

(2)

In terms of reactive power, it needs to meet:

$$i_{\mathrm{qref}}=(K_{\mathrm{qp}}+\frac{K_{\mathrm{qi}}}{s})(Q_{\mathrm{ref}}-Q)$$

(3)

In the above formula, K_qi and K_qp represent the integral coefficient and proportional coefficient in the PI controller corresponding to the q-axis, respectively.

The main purpose of the current inner loop control is to enable the output current i_d, and for the i_q of the SVG to be further controlled and regulated by U_Ad and U_Aq. However, for the dynamic reactive power compensation device SVG, there is a certain coupling relationship between the d-axis and the q-axis of each voltage and current amount. With the further change in the current amount i_d, it will directly cause the U_Aq and i_q values to change accordingly. For coupled loops, with the general treatment, it is difficult to separate them for separate numerical analysis, and decoupling is necessary when encountering this situation. To the PI output value of the inner loop circuit control, two feedforward compensations are added, namely ωLi_d and ωLi_q. The control situation is shown in Figure 1 below.

In the above figure, the voltage transfer function output by SVG is given as follows.

$$\{\begin{array}{l}{U}_{\mathrm{Ad}}=(K_{\mathrm{p}}+\frac{K_1}{s})(i_{\mathrm{dref}}-i_{\mathrm{d}})-\omega L{i}_{\mathrm{q}}+U_{\mathrm{s}}\\ U_{\mathrm{Aq}}=(K_{\mathrm{p}}+\frac{K_1}{s})(i_{\mathrm{qref}}-i_{\mathrm{q}})+\omega L{i}_{\mathrm{d}}\end{array}$$

(4)

In Formula (4), K₁ and K_p represent the integral coefficient and proportional coefficient of the PI controller, respectively.

Formula (1) is substituted into Formula (4), and the following equation is obtained after simplification:

$$L\frac{d}{dt}\left[\begin{array}{l}{i}_{\mathrm{d}}\\ i_{\mathrm{q}}\end{array}\right]=-\left[\begin{array}{cc}R+K_{\mathrm{p}}+\frac{K_1}{s}& 0\\ 0& R+K_{\mathrm{p}}+\frac{K_1}{s}\end{array}\right]\left[\begin{array}{l}{i}_{\mathrm{d}}\\ i_{\mathrm{q}}\end{array}\right]+(K_{\mathrm{p}}+\frac{K_1}{s})\left[\begin{array}{l}{i}_{\mathrm{dref}}\\ i_{\mathrm{qref}}\end{array}\right]$$

(5)

In the above Formula (5), the non-diagonal elements of the matrix are all set to zero values so that before the PI controller output in the current inner-loop control, and after adding feedforward compensation, the d, q axes have basically achieved complete decoupling control. The overall double-loop control of SVG is shown in Figure 2 below.

3. Control Strategy of Hybrid Reactive Power Compensation System

In order to be able to fully consider the response speed of reactive power compensation equipment, and, at the same time, to achieve its economy, a combined reactive power compensation program is proposed. This solution not only has the characteristics of SVG continuous and fast adjustment, but also has the advantages of low-cost and large-capacity reactive power compensation, such as parallel capacitors. Its corresponding system structure is shown in Figure 3 below.

We assume that the capacity of a single group of switchable parallel capacitors is Q_c, and the reactive power capacity required in the wind farm is Q. When kQ_c < Q < (k + 1) Q_c, the control system issues a command to switch in the k groups of capacitor banks. At the same time, the system detects the gap in reactive power compensation and issues the input SVG command, in order to further perform reactive power compensation on the system. In this way, the smooth regulation of reactive power and the voltage stability of the grid connection point are always maintained.

Conventional capacitor banks generally choose the SC type. Usually, to avoid frequent switching of SC capacitor banks, a compensation margin (Y) will be set within the reactive power range. If the power grid system has been put into k groups of capacitor banks as reactive power compensation, then: (1) when Q > (k + 1)Q_c + Y, put in the k + 1 group of capacitor banks; (2) when Q < kQ_c − Y, exit the k-th capacitor bank. Setting the compensation margin not only avoids the frequent switching of capacitor banks, but also ensures the linear demand of the system for reactive power. The control system flow of the hybrid reactive power compensation is shown in Figure 4 below.

The cost advantage of the device itself, compared with the SVC type, is that the SVG type reactive power compensation device adopts digital control technology, which has high system reliability, basically does not need maintenance, and can save a lot of maintenance costs. SVG adopts static operation, with no large rotating equipment, no wear, and no mechanical noise. This greatly improves the device’s life, and also improves the environmental impact. With the same compensation capacity, the footprint of SVG is reduced by 1/2 to 2/3 compared with that of SVC. Since SVG uses fewer reactors and capacitors than SVC, it greatly reduces the volume and footprint of the device. The reactors in SVC not only have a large volume, but also occupy a large area, considering the installation interval between them.

In terms of economic analysis of assessment, taking a station as an example, the average monthly assessment was RMB 20,000 before the installation of the reactive power compensation device. After the installation, the assessment was not only exempted, but the monthly reward was stable at more than RMB 5000, and good economic benefits were achieved, as shown in

Table 1. Summary table of assessment indicators.

Month	Active Capacity (MW)	Reactive Capacity (MVar)	Power Factor	Assessment Reward/RMB
June	1252	565	0.91	1247.83
July	1009	462	0.91	1039.80
August	1175	531	0.91	1173.31
September	1444	532	0.94	5727.96
October	1238	368	0.96	6369.16
November	1089	234	0.98	5492.54

below.

4. Parameter Configuration of Reactive Power Compensation System

The optimal capacity configuration of the reactive power compensation device is crucial to improving the efficiency of regulating reactive power. This section aims to optimize the capacity of the reactive power compensation, and uses an improved particle swarm optimization algorithm (chaotic particle swarm optimization with adaptive mutation operator) to optimize the reactive power capacity.

4.1. Improved Particle Swarm Optimization

At present, most scholars use random initialization to operate the particle position when using particle swarm optimization, but this will directly lead to the problem of the particle distribution becoming more and more uneven in the iterative process [30,31,32]. The chaotic motion has the characteristics of going through all states without repetition according to its laws within a certain range. Therefore, using chaos theory to initialize the particles in the algorithm can further enhance the diversity of the population, and fundamentally prevent the solution set from falling into a local optimum situation, which generates the corresponding chaotic sequence, as shown in Equation (6):

$$\{\begin{array}{c}y(h+1)=4y{(h)}^3-3y(h)\\ -1\le y(h)\le 1,\hspace{1em}h=0,1,2,\dots \end{array}$$

(6)

For particle i in W-dimensional space, the first step is to generate a W-dimensional vector y_i(1), and each component must be in the range of (−1, 1). Then, each dimension of the vector is entered into the above formula, and H_set-1 iterations are performed. After the iteration, the generated H_set-1 chaotic sequences enter the search space of the corresponding particle swarm.

It can be obtained from the above that q^t_best represents the optimal position of the particle after one iteration; for the q^t_best generated after each iteration process, it is given a relatively reasonable weight: s_t (0 < s_t < 1, Σs_t = 1). In order to further weigh all the particles in the algorithm, the following can be obtained: ${\overline{q}}_{\mathrm{best}}={\displaystyle \sum _{t=1}^h{s}_t{q}^t{}_{\mathrm{best}}}$. The speed update formula of the basic algorithm is replaced by ${\overline{q}}_{\mathrm{best}}$, in which q^t_best_iw is replaced. Thus, after the improvement, the new particle speed update formula is as follows:

$$v_{iw}^{t+1}=s_t{v}_{iw}^t+g_1{f}_1({\stackrel{\_}{q}}_{\mathrm{best}}-x_{iw}^t)+g_2{f}_2(d_{\mathrm{best}\_w}^t-x_{dw}^t)$$

(7)

$$x_{iw}^{t+1}=x_{iw}^t+v_{iw}^{t+1}$$

(8)

In order to further ensure the diversity of the population, to avoid local optima at the beginning of iterative evolution, and to ensure the stability of the population and improve the ability of local searches at the later stage of iterative evolution, this paper adopts a strategy to improve the particle initialization by adaptive operators. This strategy updates the particle positions variably during the iterative evolution of the algorithm by setting the variation probability, using a relatively large variation size at the beginning of the iteration, and causing it to gradually decrease during the evolution process.

$$\{\begin{array}{l}\begin{array}{l}\eta (x)=1-a^{{(1-\frac{x}{\max gen})}^b},U(0,1)>p_x\\ x^{t+1}=x^t+\eta \cdot x^t\cdot U(0,1),r⩾0.5\end{array}\hfill \\ x^{t+1}=x^t-\eta \cdot x^t\cdot U(0,1),r⩽0.5\hfill \end{array}$$

(9)

In this formula, P_x: represents the probability of occurrence of mutation, which is 0.5, and the random number is U(0,1). If the value is greater than P_x, the corresponding population particle will mutate. If the value is smaller than P_x, the corresponding population will not mutate. Additionally, η(x) ∈ [0, 1] is the variation scale of the population in the iterative evolution process; a ∈ (0, 1); max gen is the total number of evolutionary iterations; the value of b is 2, and r is a random value.

4.2. Parameter Optimization of Reactive Power Compensation System

By improving the particle swarm algorithm, the search range of the particle swarm can be guaranteed to a certain extent, and the diversity of the population can be further ensured. The algorithm can ensure the optimal value of reactive power capacity required by the reactive power compensation system, and the optimization process is shown in Figure 5.

5. Simulation and Verification of Hybrid Reactive Power Compensation Program

This section takes the Tianzhong DC wind power gathering area, the main channel of Xinjiang power transmission, as the main analysis object. It also uses PSCAD electromagnetic simulation software to build models of the Santanghu area, the central area of Hami, and the southeast wind area of Hami. It is connected to the wind power collection station by the 35 kV voltage level, and is then uniformly connected to the main grid of the entire power grid by the 220 kV bus. The connection method is shown in Figure 6. In view of the regional power grid in Figure 6, firstly, the active power margin is analyzed in three regions to find out the high permeability area of wind power; then, the modal analysis method is used to find out the best location for installing reactive power compensation device in this region; and, finally, the simulation method is used to verify the voltage support effect of installing the reactive power compensation device in this region.

5.1. Analysis of the Best Installation Location

Through the simulation analysis of Santanghu, central Hami, and Southeast wind areas, the distribution of tidal currents in this area is obtained. The active power of the Santanghu wind power gathering area is the maximum of 2582.3 MW in the three areas. The reactive power of the Southeast wind power gathering area is the largest of 3401.2 MVar in the three areas. The active power margin is used to judge whether the wind power gathering area is an area with high wind power penetration. As shown in

Table 2. Busbar margin analysis of the wind farm group area.

Area	Active Capacity (MW)	Reactive Capacity (MVar)	Active Power Margin KP (%)
Santanghu wind area	2582.3	1324.5	5.7
Southeast wind area	1791.3	3401.2	15.8
Central Hami	521.8	1759.6	65.4

below, the K_P of the central Hami wind area is 65.4%, and the margin is the largest; the K_P of the Southeast wind area is 15.8%, and the margin is the second; the K_P of the Santanghu wind area is 5.7%, with the smallest margin. The Santanghu area has a large wind power transmission capacity and the lowest active power margin, which belongs to the high penetration area of wind power.

The modal analysis is carried out on the 220 kV collective bus of each wind farm in the Hami area [33,34,35], and the specific voltage stability modal data in each wind farm in the Hami area can be obtained, as shown in

Table 3. The 220 kV aggregate busbar value of the wind farm in the Hami area.

Wind Area	Wind Farm Bus (220 kV)	Voltage Stability Modal Values/p.u.
Southeast wind area	Kushui West wind farm	0.027562346
	Kushui East wind farm	0.025684795
	Yandun South wind farm	0.021016546
	Yandun West wind farm	0.020150621
	Yandun North wind farm	0.016542154
Central Hami wind area	Naomaohu wind farm	0.043562145
	Guanghui Naomaohu wind farm	0.040561425
	Shisanjianfang wind farm	0.045462154
Santanghu wind area	Red Star wind farm	0.032456218
	Wangyangtai wind farm	0.027658974
	Wangyangtai east wind farm	0.028562016
	Wangyangtai west wind farm	0.027146484
	Mahuanggou west wind farm	0.015025153
	Mahuanggou east wind farm	0.014876213

.

From the margin analysis in Table 2 and the modal numerical analysis in Table 3, it is not difficult to find that the Yandun north wind farm in the southeast wind area and the Mahuanggou east wind farm in the Santanghu wind area are weak links in the voltage in this area. Therefore, the installation of reactive power compensation devices has the best effect in this area, which is the best installation position for combined reactive power equipment.

In Figure 7, the reactive power control method for the combined flow chart is shown. The control method first uses active power margin analysis and modal analysis to confirm the optimal installation position of the reactive power compensation device. Secondly, it uses the improved particle swarm optimization algorithm to optimize the best parameters. Finally, a six-team comparison program is set in the static voltage control and a three-team comparison program is set in the transient voltage control. The correctness of the proposed program is verified by the simulation results.

5.2. Static Voltage Stability Simulation Verification

By improving the particle swarm algorithm, as discussed in Section 4 of this paper, when the fixed compensation capacity of the parallel capacitor is set to 10 MVar, the best compensation effect should be obtained when the SVG compensation capacity is set to 9.2 MVar.

In order to verify the correctness of the method proposed in this paper, through the P-U characteristic curve of the wind farm in the eastern area of Mahuangou in the Santanghu wind area, the static voltage stability under the control of different types of reactive power compensation devices was analyzed by setting different configuration programs combined with the methods in the literature [5,11,13,17,18,26,27]. Taking the Shengtian and Huaneng wind farms in weak areas as examples, the following six programs are simulated and compared.

Program 1: no reactive power compensation device.

Program 2: set 20MVar SC capacitor banks.

Program 3: set 20MVar SVC as wind farm reactive power compensation equipment.

Program 4: set the combined reactive power compensation device proposed in this paper; the parallel capacitor capacity is 10MVar, and the SVC compensation capacity is 9.2 MVar.

Program 5: set 20 MVar SVG as the reactive power compensation equipment of the wind farm.

Program 6: set the combined reactive power compensation device proposed in this paper; the parallel capacitor capacity is 10 MVar, and the SVG compensation capacity is 9.2 MVar.

The six programs are shown in Figure 8, Figure 9 and Figure 10 below.

Program 1: In the case that there is no reactive power compensation device in either wind farm when the active power is about 385 MW, the busbar voltage drops rapidly and quickly reaches the limit instability point.

Program 2: When the SC-type capacitor bank is put in, it leads to a large oscillation of the wind turbine terminal voltage. The active power is about 500 MW when the bus voltage shows a sudden drop, but the static voltage stability limit of the wind farm convergence area is improved compared to option 1.

Program 3: When the static reactive power compensation device SVC is put in, compared to the first two programs, the ultimate destabilizing active power output of the wind farm voltage can reach 543 MW before the voltage destabilization phenomenon finally occurs. This raises the voltage destabilization threshold and can further enhance the static voltage stability limit of the wind farm cluster to a certain extent.

Program 4: When combined SVC and shunt capacitor reactive power compensation equipment are used, the SVC can smooth the reactive power output and further maintain a constant voltage at the controlled bus within the reactive power regulation capacity. Compared with the aforementioned programs, it can more significantly improve the static voltage stability limit, for which the wind farm group’s active output reaches 570 MW when voltage instability eventually occurs.

Program 5: When the dynamic reactive power compensation device SVG is put in, the ultimate destabilized active power output of the wind farm voltage can reach 600 MW compared to the first two programs, which raises the voltage destabilization critical point.

Program 6: When the combined reactive power compensation device of SVG and shunt capacitor is used, because SVG also can smoothly regulate the reactive power, when a serious fault occurs in the power system, it will further cause the bus voltage to drop significantly. For the combined reactive power compensation device of the shunt capacitor plus SVG proposed in this paper, its reactive power compensation performance is better than that of SVG, which can further maintain the voltage stability. The active power output of its wind farm group reaches 610 MW before the voltage instability finally occurs.

In summary, the voltage support ability of the above six reactive power compensation configuration programs is enhanced in turn. The minimum is when the active power of program 1 is about 385 MW, and the bus voltage drops rapidly. The maximum is when the active power output of program 6 reaches 610 MW, and the voltage instability finally occurs. Program 6 adopts the combined reactive power compensation of SVG and capacitor, which has more advantages in maintaining voltage stability and improving voltage stability limit. Program 6 also takes into account the advantages of rapid adjustment ability and the small size of SVG, and has the advantages of the low cost and large capacity of shunt capacitors.

5.3. Transient Voltage Stability Simulation Verification

In this section, the characteristics related to the transient voltage stability of the power grid in the Hami region of Xinjiang are further analyzed and studied by the time-domain simulation method. Based on the previous section, it can be known that the static voltage stability of the power grid near the Santanghu wind area in the Hami region is lower compared to the other two wind areas, and is the weak point of the voltage in the whole Hami region. Thus, the Santanghu wind area is the focus of this section, and the model built is shown in Figure 11.

By improving the particle swarm algorithm, as discussed in Section 4 of this paper, when the fixed compensation capacity of the parallel capacitor is set to 10 MVar, the best compensation effect should be obtained when the SVG compensation capacity is 9.2 MVar.

Program 1: The program of the reactive power compensation device, SVC, and a combination of SVC and parallel capacitors are used. The system has a three-phase short-circuit fault in the middle part of the transmission line between the busbar and the grid-connected busbar at 2.5 s, and the fault is eliminated after 0.3 s resection.

Program 2: The reactive power compensation device, SVG, SVG, and shunt capacitor combination program are adopted, and other conditions are the same as program 1.

Program 3: The combination of SVC and the shunt capacitor as well as the combination of SVG and the shunt capacitor are adopted, and other conditions are the same as program 1.

The three programs are shown in Figure 12, Figure 13 and Figure 14 below.

Program 1: When there is no reactive power compensation device in the system, the voltage drops to 0.102 P.U. at the time of the fault. At only 1.6 s after the fault, the wind farm outlet voltage bus finally reaches a new stabilization curve, though the voltage value at the moment of stabilization is only 0.937 P.U., which is much smaller than the normal voltage value. However, after adding the dynamic reactive power compensation device SVC to the system, although the fall position was basically the same as above without the reactive power compensation device, the short circuit fault was removed. This further enabled the bus to recover stability quickly, with the voltage at the controlled bus returning to 1.001 P.U. within 1.4 s after the fault was removed. When the combined SVC and shunt capacitor reactive power compensation program is added, it is obvious that the bus voltage drops to 0.132 P.U., which shows some improvement compared with the first two methods, and the bus voltage is restored to 1.00 P.U. by 0.7 s after the fault removal.

Program 2: When the dynamic reactive power compensation device SVG is added, the voltage dip is only 0.134 P.U. when the fault voltage dip occurs in the system. This is nearly 0.032 P.U. higher than without the reactive power compensation device, and the bus voltage recovery speed after the fault is removed is also much faster than without the reactive power compensation device. The bus voltage recovered to 1.001 P.U. within 0.76 s after the fault. However, when SVG is used for reactive power compensation, the tracking of bus voltage oscillation occurs, which directly causes the corresponding bus voltage to oscillate to a large extent. When the proposed combined reactive power compensation program of SVG and the shunt capacitor is used, it is similar to the dynamic reactive power compensation device SVG alone in terms of suppressing voltage dips, but after fault removal, especially in the process of bus voltage stabilization, it is obvious that the oscillation voltage is much smaller when the combined reactive power compensation program is used than when the SVG alone is used. The oscillation voltage is obviously much smaller than that of SVG alone.

Program 3: When a fault occurs in the system, both combined reactive power compensation programs perform relatively well in suppressing voltage dips, with essentially the same effect. However, when the fault is completely removed, especially in the process of bus voltage recovery, the combined reactive power compensation program of SVG and the shunt capacitor makes the oscillation of the bus voltage smaller. Therefore, its reactive power compensation performance is better than that of the combined program of SVC and the shunt capacitor.

According to the above three comparison programs, when the system fails, both the SVC + SC and the SVG + SC combination can effectively suppress voltage sags, but when the fault is removed, the SVG + SC combination performs better in the voltage recovery stage. Therefore, compared with other reactive power compensation programs, the SVG + SC combination has obvious advantages in transient voltage stability. It is also more suitable to adopt the proposed reactive power compensation combination program where the voltage is weaker in the wind power gathering area.

6. Conclusions

Firstly, this paper finds out the weak points of the power grid through the modal analysis of the gathering bus of the wind farm. Secondly, the improved particle swarm optimization algorithm is used to find the optimal reactive power installed capacity. Finally, six static voltage stability simulation verification programs were configured based on the power grid in the Hami region, and the analysis and comparison of the program simulation results verified that the combined reactive power compensation device of SVG and a capacitor bank has the best effect on static voltage stability support. Based on the power grid in the Santanghu area, three types of transient voltage stability simulation verification programs were configured, and through the analysis and comparison of the program simulation results, it was verified that the combined reactive power compensation device of SVG and a capacitor bank has the best effect on supporting transient voltage stability. The proposed combined reactive power compensation device, based on SVG and a capacitor bank, not only achieves the effect of reactive power smooth regulation, but also achieves economy of equipment. The combined reactive power compensation device not only enhances the static and transient voltage stability of the grid, but also improves the capability of wind power transmission via UHV Tianzhong DC. This study conducts a detailed study on the wind power gathering area in only the Hami region, and lacks research data from other wind power gathering areas. In the future, detailed studies on other areas will be conducted in order to verify the popularization of the proposed method.