1 Introduction

In order to address the environmental pollution and energy crisis, many distributed generations (DGs), e.g., solar energy, wind energy, microturbines, and fuel cells are widely used. These DGs are connected to the utility grid via inverters, which are also known as the microgrid [1, 2]. The operation mode of the microgrid can be switched between grid-connected and islanded mode flexibly [3]. A necessary problem for parallel-operated DGs in a microgrid is how to avoid the circulating current among them. One of the key techniques is to use droop control [4, 5], which has been widely used in conventional synchronous generators [6] and uninterruptible power supply (UPS) [7]. Its advantage is that each DG is able to be operated independently without any critical communication.

Although the droop control is widely used as a wireless control strategy for load sharing, it has some drawbacks that limit its application. For instance, the load sharing accuracy is degraded if the per-unit impedances (the output impedance and the line impedance) of each DG are unbalanced [8]. Usually, P-f and Q-V droop control are used when the equivalent impedance which connecting DGs to the gird is inductive in a microgrid [9]. The active power can be shared accurately using P-f droop control since it is just based on the frequency of the microgrid. However, the DG voltage output in Q-V droop control is different with the voltage at the point of common coupling (PCC) because of the different line impedances. Thus the accuracy of reactive power sharing using Q-V droop control is usually substantially affected by the equivalent impedance of DG and line impedance [10].

In order to enhance the accuracy of reactive power sharing, the droop control gains are necessary to be optimized [11, 12]. However, there are still inherent tradeoff between the load sharing accuracy and voltage regulation. Therefore, these methods are only valid for the microgrid with short distribution lines where the line impedance is negligible. Recently, the virtual reactor or integral unit to adjust the equivalent impedance of DGs is added to the conventional droop control loop [13, 14]. However, the line impedance is still unknown even the equivalent impedance of DGs can be established. Thus, the line impedances are detected or calculated in some literatures [15, 16] in order to mitigate their influence. However, it is quite complicated to implement. In some cases, the external communication is still adopted [17]. However, the reliability and stability of microgrids will be affected substantially by the communication system. The complicated communication system would also increase the difficulties significantly in design and implementation of large scale microgrids.

Reference [8] proposed a robust droop control and the robustness of the voltage output and load sharing accuracy could be enhanced obviously. However, it is only suitable for the centralized microgrid with communication equipment. Besides, a virtual impedance unit added in control loop should be bigger enough so that the effect of line impedance can be omitted, which is complex and limited to its application. Nowadays, the high-voltage large scale microgrid for campus and island are most likely to be developed [18, 19]. In these kinds of microgrids, DGs are always connected to the utility grid by a step-up coupling transformer. As the equivalent impedance of the coupling transformer is usually much larger than that of the line, in this sense, the effect of line impedance may be negligible. If the load sharing error caused by the equivalent impedance of transformer can be eliminated, the performance of microgrids can also be improved obviously.

A robust droop control method is proposed in this paper which is suitable for high-voltage microgrids and easy to implement. In Section 2, the impact of line impedance on load sharing accuracy is analyzed. In Section 3, the improved robust droop control is presented. The system model for analyzing the influence of system parameters on the performance of the proposed control concept and the design consideration of robust droop control gain are presented in Section 4. The PSCAD simulation results are given in Section 5, followed by conclusions in Section 6.

2 Impacts of line impedance on load sharing in microgrids

Figure 1 shows a general scheme of a microgrid which consists of sets of microsource and energy storage, distributed loads and electric power interfaces inverter connected to the step-up coupling transformer. Besides, the microgrid can be connected to the utility grid operating in grid-connection mode if the switch STS is turned on. When the utility grid is not present, the DG units should be able to share the total power demanded together.

Fig. 1
figure 1

A high-voltage microgrid

Figure 2 shows the equivalent high-voltage side circuit of the microgrid. E i θ i (i = 1,2,⋯) is the voltage output of DG i ; P i and Q i are the active and reactive power flowing into loads generated by DG i , respectively; Vθ is the AC common bus voltage; Z i φ i is the equivalent impedance.

Fig. 2
figure 2

Equivalent circuit of a high-voltage microgrid connected to a common bus

The current output of each DG is

$$ \varvec{I}_{i} = \frac{{E_{i} \cos \delta_{ig} - V + {\text{j}}E_{i} \sin \delta_{ig} }}{{R_{i} + {\text{j}}X_{i} }} $$
(1)

where X i and R i are the equivalent inductive and resistive component, respectively; δ ig is the phase difference between E i and \( V,\delta_{ig} = \theta_{i} - \theta. \)

The active power and reactive power output of each DG are calculated as

$$ \begin{aligned} &P_{i} = \frac{{X_{i} }}{{Z_{i} }}\frac{{E_{i} V}}{{Z_{i} }}{ \sin }\;\delta_{ig} + \frac{{R_{i} }}{{Z_{i} }}\frac{{V(E_{i} - V{ \cos }\;\delta_{ig} )}}{{Z_{i} }} \hfill \\ &Q_{i} = - \frac{{R_{i} }}{{Z_{i} }}\frac{{E_{i} V}}{{Z_{i} }}{ \sin }\;\delta_{ig} + \frac{{X_{i} }}{{Z_{i} }}\frac{{V(E_{i} - V{ \cos }\;\delta_{ig} )}}{{Z_{i} }} \hfill \\ \end{aligned} $$
(2)

In high-voltage power systems, the inductive reactance X i is much larger than resistive component R i which means \( { \cos }\delta_{ig} \approx 0 \) and \( { \sin }\delta_{ig} \approx \delta_{ig} \), so (2) can be simplified as

$$ \begin{aligned} &P_{i} = \frac{{E_{i} V}}{{X_{i} }}\delta_{ig} \hfill \\ &Q_{i} = \frac{V}{{X_{i} }}(E_{i} - V) \hfill \\ \end{aligned} $$
(3)

Equation (3) is the theoretical basis of traditional P-f and Q-V droop control. Accordingly, the frequency and voltage output of DGs can be regulated as

$$ \begin{aligned} \omega_{i} = {\omega^{*}} - n_{i} P_{i} \hfill \\ E_{i} = {E^{*}} - m_{i} Q_{i} \hfill \\ \end{aligned} $$
(4)

where ω i and \({\omega^{*}}\) and are the frequency output and the rated frequency of DG i ; E * is the rated voltage of high-voltage microgrid; n i and m i are the frequency droop gain and voltage droop gain, respectively.

With the traditional and normal droop control strategy, when the microgrid operates in steady state, it will share the same frequency with utility grid. Thus, the frequency and active power of DG i can be described as

$$ \left\{ \begin{array} {ll}\omega_{1} = \omega_{2} = \ldots = \omega_{i} \hfill \\ n_{1} P_{1} = n_{2} P_{2} = \ldots = n_{i} P_{i} \hfill \\ \end{array} \right. $$
(5)

However, the reactive power Q i mainly depends on E i , V and the line impedance Z i [20]. V can be expressed as

$$ {\mathbf{V}} = (E_{1} \angle \theta_{1} , \ldots ,E_{i} \angle \theta_{i} ,R_{1} + {\text{j}}X_{1} , \ldots ,R_{i} + {\text{j}}X_{i} ) $$
(6)

As can be seen from (6), the accurate reactive power sharing is difficult to be implemented with the conventional droop control scheme unless the impedance of distribution lines is strictly the same as (7).

$$ \left\{ \begin{array} {ll}E_{1} = E_{2} = \ldots = E_{i} \hfill \\ \frac{{m_{1} }}{{X_{1} }} = \frac{{m_{2} }}{{X_{2} }} = \ldots = \frac{{m_{i} }}{{X_{i} }} \hfill \\ \end{array} \right. $$
(7)

Due to the unpredictable and unchangeable of the microgrid parameters, (7) is almost impossible to be satisfied, especially the per-unit impedance. Therefore, how to achieve accurate reactive power sharing is one of the key topics in microgrids.

For the inverter-interfaced high-voltage microgrid, DGs are always connected to the utility grid through a step-up coupling transformer. Thus, a simplified and general topology of a typical high-voltage microgrid system with paralleled DGs is shown in Fig. 3. L si and C si are inductor and capacitor of the LC filter, respectively. V fi is the voltage of the inverter output port. K Ti is the voltage ratio of transformer; V Ti and i Ti are the output voltage and current at the high-voltage side of transformer T i, respectively; V i and i i are the output voltage and current at the low-voltage side of transformer T i, respectively; X Li is the equivalent impedance of Line i. In additional, the fundamental frequency element of V fi is equal to E i . Therefore, the output impedances X i is

$$ X_{i} \approx \omega k_{{{\text{T}}i}}^{2} L_{{{\text{s}}i}} + X_{{{\text{T}}i}} + X_{{{\text{L}}i}} $$
(8)

where X Ti is the transformer equivalent impedance in the high-voltage side of step-up coupling transformer.

Fig. 3
figure 3

A simplified and general topology of a typical high-voltage microgrid system

Besides, coupling transformers are usually used for high-voltage microgrids. Hence, compared with a low-voltage microgrid, the impact of the coupling transformer in the high-voltage microgrid should be considered when deal with the influence of line impedance. Especially for a large-capacity transformer, the inductive component of the short circuit impedance is usually larger than the resistive component. The short-circuit reactance X T of the transformer is calculated [21].

$$ X_{\text{T}} = \frac{{V_{\text{s}} }}{100}\frac{{V_{\text{N}}^{2} }}{{S_{\text{N}} }} $$
(9)

where S N and V N are the apparent power and rated voltage of the transformer; V s is the percentage of short-circuit voltage.

It is noted that the inductive component X T of the transformer with large capacity in a high-voltage microgrid system is bigger enough so that the line impedance could be omitted. For example, a transformer with the parameters: \( S_{\text{N}} = 0. 1\sim 1 {\text{ MVA}} \), \( V_{\text{N}} = 10{\text{ kV}} \), \( V_{\text{s}} = 10\,\% \), then, \( X_{\text{T}} = 10\sim 100 \, \varOmega \). In addition, the line impedance X L of a high-voltage microgrid is generally about 0.5 Ω. Therefore, in most cases, \( X_{\text{T}} \gg X_{\text{L}} \) can be established for the high-voltage microgrid. Thus, it is noted that \( V_{{{\text{T}}i}} \approx V \) in a high-voltage microgrid with the step-up coupling transformer.

Reference [8] proposed a robust droop control based on the detection of AC common bus voltage V and current i Ti flowing into the AC common bus, which can not only overcome the influence of system unbalanced impedance but also pull up the AC common bus decreased voltage due to the output impedance and voltage droop coefficient. However, this robust droop control is only suitable for microgrids with communication equipments, which limits its application. Fortunately, based on the above analysis, the equivalent impedance of coupling transformer is usually much larger than that of the line. It would be highly advantageous to develop a robust droop control strategy without signals communication which can eliminate the influence of the mismatched equivalent line impedance.

3 Robust droop control strategy for high-voltage microgrid

If the robust droop method proposed in reference [8] is applied in the high-voltage microgrid as shown in Fig. 3, in addition to current flowing into AC common bus, the remote bus voltage V also should be sampled by voltage sensors as shown in Fig. 4 with K s connected to Channel 1. Compared with the traditional droop control, there is a great difference that the drop voltage E *V is fed back via an amplifier K e, and then the feedback signal is added to –m i Q i on account of Q-V droop effect. As to be explained below, it is able to maintain accurate proportional load sharing and voltage robustness in spite of mismatched line impedances.

Fig. 4
figure 4

Robust droop control for high-voltage microgrid

When the microgrid with the robust droop control operates in steady state, the condition (10) is satisfied.

$$ K_{\text{e}} \left( {E^{*} - V} \right) = m_{i} Q_{i} $$
(10)

The left-hand side of the above is always the same for all DGs connected to microgrid as long as K e is set the same. Hence, the right-hand sides of (10) are equal. Then

$$ m_{1} Q_{1} = m_{2} Q_{2} = \ldots = m_{i} Q_{i} $$
(11)

Assume that

$$ m_{1} S_{1} = m_{2} S_{2} = \ldots = m_{i} S_{i} $$
(12)

where S i is the rated power output of DG i .

Thus, the supply power of each inverter for loads is proportional to its capacity of inverter, and (13) can be derived.

$$ \frac{{Q_{1} }}{{S_{1} }} = \frac{{Q_{2} }}{{S_{2} }} = \ldots = \frac{{Q_{i} }}{{S_{i} }} $$
(13)

Accordingly, regardless of the mismatched line impedance, the reactive power output of DG is proportional to its capacity with the robust droop control, and then the high accuracy of reactive power sharing can be achieved.

As shown in Fig. 4, in order to achieve a highly load sharing accuracy with the robust droop control, the common bus voltage and power of each DG flowing into a concentrated load have to be detected. Thus, this kind of droop control concept is only suitable for the microgrid with concentrated loads. Also, the transmission process of feedback signals for a long-distance may also challenge the load sharing accuracy and reliability of microgrids. According to the analysis in Section 2, the equivalent impedance of coupling transformer is generally much larger than the line impedance for a high-voltage microgrid, namely X Ti  >>X Li , so V Ti is much close to V. Thus, it is possible to applied robust power droop control by detecting V Ti at the high-voltage side of step-up coupling transformer without communication devices and instead of AC common bus voltage V.

Accordingly, a new robust droop control is proposed based on the detection of V Ti and the current output at the high-voltage side of transformer in this paper. The control block of the new robust droop control is shown in Fig. 4 when K s is connected to Channel 2. So (10) can be rewritten as

$$ K_{\text{e}} \left( {E^{*} - V_{{{\text{T}}i}} } \right) = m_{i} Q_{i} $$
(14)

The Q-V droop expression can be expressed as

$$ V_{{{\text{T}}i}} = E^{*} - \frac{{m_{i} }}{{K_{\text{e}} }}Q_{i} = E^{*} - \frac{{m_{i} Q_{i} }}{{K_{\text{e}} E^{*} }}E^{*} $$
(15)

where \({m_{i}}Q_{i}/(K_{\text{e}} E^{*})\)is the voltage drop ratio. Note that the overall effective voltage drop coefficient m i /K e in (15) is much smaller than that m i in the conventional droop controller as long as the K e is big enough. In addition, meanwhile, it means the robust droop controller could achieve fast response despite of a smaller voltage droop coefficient and the robustness of output voltage is improved.

Assume that

$$ \left\{ \begin{aligned} &Q_{\text{desired}} = \frac{{S_{i} }}{{\sum\limits_{i = 1}^{n} {S_{i} } }}Q_{\text{load}} \hfill \\ &Q_{i} = Q_{\text{desired}} + \Delta Q_{i} \hfill \\ \end{aligned} \right. $$
(16)

where Q load and Q desired are the real reactive power load and desired reactive power of DG i , respectively; ΔQ i is the error of Q i and Q desired; n is the sum of inverter. Note that there must be a voltage V T within the scope of real field which could satisfy (17).

$$ K_{\text{e}} (E^{*} - V_{\text{T}} ) = m_{i} Q_{\text{desried}} $$
(17)

Thus, according to (14), (16) and (17), we can obtain

$$ K_{\text{e}} \left[E^{*} - V_{\text{T}} + (V_{\text{T}} - V_{{{\text{T}}i}} )\right] = m_{i} (Q_{\text{desried}} + \Delta Q_{i} ) $$
(18)

So (18) can be rewritten as

$$ K_{\text{e}} (V_{\text{T}} - V_{{{\text{T}}i}} ) = m_{i} \Delta Q_{i} $$
(19)

According to the conservation of energy theorem, it is noted that the system must satisfy

$$ \left\{ \begin{array}{l} Q_{\text{desried}} \sum\limits_{i = 1}^{n} {S_{i} } = Q_{\text{load}} S_{i} \text{ } \hfill \\ \sum\limits_{i = 1}^{n} {\Delta Q_{i} } = 0\text{ } \hfill \\ \sum\limits_{i = 1}^{n} {Q_{i} } = Q_{\text{load}} \hfill \\ \end{array} \right. $$
(20)

According to (19) and (20), (21) can be achieved through accumulating and simplifying.

$$ K_{\text{e}} \sum\limits_{i = 1}^{n} {\frac{{V_{\text{T}} - V_{{{\text{T}}i}} }}{{m_{i} }}} = \sum\limits_{i = 1}^{n} {\Delta Q_{i} } $$
(21)

According to (21), the voltage V T can be described as

$$ V_{\text{T}} = m\sum\limits_{i = 1}^{n} {\frac{{V_{{{\text{T}}i}} }}{{m_{i} }}} $$
(22)

where \( m = {1 \mathord{\left/ {\vphantom {1 {\sum\limits_{{i = 1}}^{n} {\frac{1}{{m_{i} }}} }}} \right. \kern-\nulldelimiterspace} {\sum\limits_{{i = 1}}^{n} {\frac{1}{{m_{i} }}} }} \).

Based on the above analysis, the error degree of reactive power sharing ρ i can be expressed as

$$ \rho_{i} = \frac{{\Delta Q_{i} }}{{Q_{\text{desired}} }} = \frac{{K_{\text{e}} }}{{m_{i} Q_{\text{desired}} }}(V_{\text{T}} - V_{{{\text{T}}i}} ) $$
(23)

Substitute (16) and (22) to (23), it can be obtained that

$$ \rho_{i} = \frac{{K_{\text{e}} S}}{{m_{i} S_{i} Q_{\text{load}} }}\left(m\sum\limits_{j = 1}^{n} {\frac{{V_{{{\text{T}}j}} }}{{m_{j} }}} - V_{{{\text{T}}i}} \right) $$
(24)

where \( S = \sum\limits_{i = 1}^{n} {S_{i} } \). It is seen that ρ i mainly depends on V Ti , Q load and K e in a special high-voltage microgrid with the proposed local robust droop controller scheme. However, the parameters in (24) may be interactional, such as V Ti with Q load. Accordingly, the following Section will analyze influences of K e, Q load and X Li on ρ i in detail.

4 Relationship between load sharing and system parameters

In order to accurately show the influence of line impedance and robust coefficient K e on the load sharing accuracy, the accurate relationship model between the load sharing and system parameters is presented in this section. It is known that V T i and V satisfy

$$ {\mathbf{V}} = V_{{{\text{T}}i}} \angle \delta_{{{\text{T}}i}} - \Delta {\mathbf{V}}_{{{\text{L}}i}} $$
(25)

where δ Ti is phase angle of V Ti ; ΔV Li is voltage drop due to the line impedance.

Since the phase angle difference δ Ti between V Ti and V is small and negligible, so (26) can be derived.

$$ V_{Ti} = V + \Delta V_{Li} $$
(26)

Substitute (26) into (24), it can be obtained

$$ \rho_{i} = \frac{{K_{\text{e}} S}}{{m_{i} S_{i} Q_{\text{load}} }}\left( {m\sum\limits_{j = 1}^{n} {\frac{{\Delta V_{{{\text{L}}j}} }}{{m_{j} }}} - \Delta V_{{{\text{L}}i}} } \right) $$
(27)

Based on the power flow calculation, voltage drop ΔV Li can be calculated as

$$ \Delta V_{{{\text{L}}i}} = \frac{{P_{i} R_{{{\text{L}}i}} + Q_{i} X_{{{\text{L}}i}} }}{V} $$
(28)

where R Li and X Li are the resistive and inductive part of line impedance, respectively.

In the high-voltage microgrid, X Li is the main part of line impedance, and V is approximately equal to E *, so (28) can be converted to

$$ \Delta V_{{{\text{L}}i}} = \frac{{Q_{i} X_{{{\text{L}}i}} }}{{E^{*} }} $$
(29)

Substitute (29) into (27), (30) is derived.

$$ \rho_{i} = \frac{{K_{\text{e}} S}}{{m_{i} S_{i} E^{*} Q_{\text{load}} }}\left( {m\sum\limits_{j = 1}^{n} {\frac{{Q_{j} X_{{{\text{L}}j}} }}{{m_{j} }}} - Q_{i} X_{{{\text{L}}i}} } \right) $$
(30)

Based on (15), (26) and (29), (31) can be obtained.

$$ V = E^{*} - \frac{{m_{i} }}{{K_{\text{e}} }}Q_{i} - \Delta V_{{{\text{L}}i}} = E^{*} - m_{{{\text{eq}}i}} Q_{i} $$
(31)

where \( m_{{{\text{eq}}i}} = \frac{{m_{i} }}{{K_{\text{e}} }} + \frac{{X_{{{\text{L}}i}} }}{{E^{*} }} \).

Assume that

$$ V = E^{*} - m_{\text{eq}} Q_{\text{load}} $$
(32)

where m eq is the equivalent droop coefficient of V versus reactive power Q load.

Based on (20), (31) and (32), (33) can be obtained.

$$ Q_{i} = \frac{{m_{\text{eq}} }}{{m_{{{\text{eq}}i}} }}Q_{\text{load}} $$
(33)

where \( m_{{{\text{eq}}}} = {1 \mathord{\left/ {\vphantom {1 {\sum\limits_{{i = 1}}^{n} {\frac{1}{{m_{{{\text{eq}}i}} }}} }}} \right. \kern-\nulldelimiterspace} {\sum\limits_{{i = 1}}^{n} {\frac{1}{{m_{{{\text{eq}}i}} }}} }} \).

Substitute (33) into (30), then

$$ \rho_{i} = \frac{{K_{\text{e}} S}}{{m_{i} S_{i} E^{*} }}\left( {m\sum\limits_{j = 1}^{n} {\frac{1}{{m_{j} }}\frac{{m_{\text{eq}} X_{{{\text{L}}j}} }}{{m_{{{\text{eq}}j}} }} - \frac{{m_{\text{eq}} X_{{{\text{L}}i}} }}{{m_{{{\text{eq}}i}} }}} } \right) $$
(34)

Naturally, (34) can be utilized as the mathematic model to analyze the power sharing performance of the proposed control strategy. Noted that ρ i only depends on the line impedance X Li and K e, and it has no difference with reactive power output Q load or active power output P load when parameters m i , S i , S and E * are constant.

Assume that the sum of inverter n = 2 in the high-voltage microgrid system, and then the ratio coefficient of capacity of DG2 to that of DG1 is k. At the same time, the parameter m 1 is set to k times as much as m 2. Accordingly, S i /S is equal to k i−1 /(1 + k), and (34) can be simplified as

$$ \rho_{i} \approx \left( { - 1} \right)^{i} \frac{{k^{3 - 2i} K_{\text{e}} (X_{{{\text{L}}1}} - kX_{{{\text{L}}2}} )}}{{(1 + k)m_{i} E^{*} }}\quad i = 1, 2 $$
(35)

Equation (35) shows that the larger difference between kX L2 and X L1 is, the larger |ρ i | will be, and |ρ i | will be zero when X L1 = kX L2. In addition, it should also be noted that the value of ρ x may be positive or negative, e.g., ρ 1 > 0 and ρ 2 < 0 when X L1 > kX L2, ρ 2 > 0 and ρ 1 < 0 when X L1 > kX L2. ρ i  > 0 represents that the reactive power output of inverter i exceeds the desired power output, i.e., Q i  > Q desired, and the DG is on the risk of overload. Whereas, it indicates that the reactive power output of DG i is less than the desired output, e.g., Q i  < Q desired, and the DG i is on the condition of under-load.

Taking the absolute value of the both sides of (35) and (36) can be obtained.

$$ \left| {\rho_{i} } \right| \approx \frac{{k^{3 - 2i} K_{\text{e}} \left| {X_{{{\text{L}}1}} - kX_{{{\text{L}}2}} } \right|}}{{(1 + k)m_{i} E^{*} }} $$
(36)

According to (35), K e can be represented as

$$ K_{\text{e}} \approx \frac{{(1 + k)m_{i} E^{*} \left| {\rho_{i} } \right|}}{{k^{3 - 2i} \left| {X_{{{\text{L}}1}} - kX_{{{\text{L}}2}} } \right|}} $$
(37)

Thus, to ensure |ρ i | ≤ 0.1, K e has to satisfy

$$ K_{\text{e}} \le \frac{{(1 + k)m_{i} E^{*} }}{{10k^{3 - 2i} \left| {X_{{{\text{L}}1}} - kX_{{{\text{L}}2}} } \right|}} $$
(38)

In addition, it is necessary to consider the voltage robustness at the moment of designing K e. As shown in (15), the robustness of voltage can be guaranteed only if K e is larger than 1. If K e is less than 1, the equivalent voltage droop coefficient m i /K e will be larger than the setting droop coefficient m i which indicates that the voltage will drop more not less than the reactive power drops. For example, if i = 1, k = 2, E * = 8.156 kV, m 1 = 3 × 10−3 and \( X_{\text{L1}} - 2X_{\text{L2}} = 1 \), then 1 < K e ≤ 3.67. Thus, to enhance the robustness of voltage and improve the accurateness of sharing load at the same time, comprehensive consideration should be taken to select the most suitable K e.

So the following section will validate the incidence of impedance X Li and robust coefficient K e on the error degree of reactive power sharing ρ i by simulation and its results.

5 Simulation validation

In order to show the reactive power sharing accuracy with the proposed robust droop control, simulations are carried out within PSCAD/EMTDC, and the topology of the studied high-voltage microgrid is shown in Fig. 2 with i = 2. The simulation parameters are summarized in Table 1. In addition, the simulation starts at t = 0 s. At t = 1 s, DG1 and DG2 are put into operation to supply Z load connected the AC common bus simultaneously. Simulation process finishes at t = 6 s.

Table 1 Simulation parameters

5.1 Comparison between proposed robust droop control and conventional droop control

The simulations are carried out with the conventional voltage droop coefficients divided by K e = 5 so that the equivalent voltage droop coefficient is the same as the one in the proposed controller. Expecting that the voltage droop could be comparable with the one achieved with the proposed robust droop scheme. The line impedances Z L1 = j0.9425 Ω, Z L2 = j0.314 Ω.

Figure 5 shows the simulation results of both the proposed new robust droop control and the conventional droop control. P i_2 and Q i_2 are the active and reactive power of DG i and V g_2 is the AC common bus voltage when the new robust droop control is applied. P i and Q i are the active power and reactive power of DG i and V g is the AC common bus voltage when the conventional droop control is applied.

Fig. 5
figure 5

Comparison between new robust droop control and conventional droop control

In Fig. 5a, when the simulation is in steady state at about t = 2 s, the active power output of DGs with different droop controls are that P 2_2 = 2P 1_2 = 3.9 × 102 kW and P 2 = 2P 1 = 2.5 × 102 kW, which indicates that the active power sharing among DGs could be achieved either through the proposed droop control or the conventional droop control. However, the proposed robust droop control can supply higher power.

In Fig. 5b, it can be seen that Q 1 = 1.9 × 102 kVar, Q 2 = 2.5 × 102 kVar, the reactive power sharing is unable to be achieved with the conventional droop control due to the mismatched line impedances. However, in the proposed robust droop control system, Q 2_2 = 1.22 × 105 kVar is nearly equal to 2Q 1_2 = 2.66 × 105 kVar, which means that reactive power sharing has be improved much when the proposed control scheme is adopted.

As shown in Fig. 5c, the AC common bus voltage with the proposed droop control can be maintained around the rated voltage, but the voltage with traditional droop control is only about 3/4 of the rated voltage, which agrees with the analysis of (5) in Section 3. In a word, the tradeoff between the load sharing accuracy and voltage regulation can be restrained successfully as that in [8], but it is easier to be implemented with the proposed droop control.

5.2 Performance of proposed robust droop control with load variations

In the section, the influence of load changes on load sharing is analyzed. The line impedances are assumed as Z L1 = 0.9425 Ω and Z L2 = 0.314 Ω, and the robust coefficient K e is set to 5. The simulation results are shown in Fig. 6.

Fig. 6
figure 6

Performance of new robust droop control with load variation

At t = 1 s, the load with 3/4 rated power is connected to the microgrid. At t = 3 s, the load decreases to 1/4 rated power. At t = 4.5 s, the load increases to the rated power. From Fig. 6a, b, it can be seen that the ratio of active or reactive power output of DG1 to that of DG2 is about 1/2 whatever the active power or reactive power load changes. It can be seen from Fig. 6c, the AC common bus voltage is always maintained at the rated value. Thus, it can be concluded that the variation of active or reactive load has little influence on the load sharing performance with the proposed droop control.

5.3 Influence of line impedance on load sharing accuracy

According to the analysis in Section 4, the influence of line impedance on the degree error of reactive power sharing ρ x is shown in (33). And it also illustrates that ρ x mainly depends on \( X_{\text{L1}} - kX_{\text{L2}} \) with a constant K e. So we can explore how line impedance impacts load sharing accuracy based on (34). Figure 7 shows the load sharing performance of the proposed robust droop control with different line impedances.

Fig. 7
figure 7

Reactive power sharing of DGs with different line impedance

Assume that the basic value Z LB is equal to the maximum value of line impedances, i.e., Z LB = Z Lmax. Thus, Z LB is the level of line impedance, which restrains the range interval of \( X_{\text{L1}} - kX_{\text{L2}} \). Z LB is set to 0.5, 1 and 1.5 Ω in this paper. In addition, K e is equal to 5 and other parameters are the same as in Section 5.1. The system parameters and reactive power supplied by each DG are shown in Table 2. Q 1 and Q 2 are the reactive power output of DG1 and DG2 at the high-voltage side of coupling transformer. In Fig. 7, \( X_{\text{L1}} - 2X_{\text{L2}} \) is the parameter of line impedance. According to Fig. 7 and Table 2, it can be observed that

Table 2 Systems parameters and reactive power supplied by DG with different line impedance
  1. 1)

    ρ 1 > 0 and ρ 2 < 0 when \( X_{\text{L1}} - 2X_{\text{L2}} < 0 \), and the reactive power output of DG1 is larger than its desired power and the reactive power output of DG2 is less than its desired power. On the contrary, ρ 1 < 0 and ρ 2 > 0 when \( X_{\text{L1}} - 2X_{\text{L2}} > 0 \), and the reactive power output of DG1 is less than its desired power and the reactive power output of DG2 is larger than its desired power.

  2. 2)

    The smaller the absolute value of \( X_{\text{L1}} - 2X_{\text{L2}} \) is, the more accurate the load sharing among DGs will be. In another word, the closer to zero \( X_{\text{L1}} - 2X_{\text{L2}} \) is, the smaller ρ 1

and ρ 2 are, and the better of the load sharing performance among DGs would be as result.

5.4 Influence of robust coefficient K e on load sharing accuracy

Based the simulations in Section 5.1, assume that the line impedance Z L1 = 0.9425 Ω, Z L2 = −0.314 and \( X_{\text{L1}} - 2X_{\text{L2}} > 0 \), the simulation results are shown in Fig. 8 with the robust coefficient K e equaling to 10, 5, 1 and 0.2, respectively.

Fig. 8
figure 8

Power sharing and AC common bus voltage

In Fig. 8a, b, P i and Q i are the reactive and active power output of DG i with different K e. Figure 8c shows the AC common bus voltage V with different K e.

The following can be observed from Fig. 8 and Table 3.

Table 3 Voltage, reactive power and ρ i with different K e
  1. 1)

    The change of K e has no effects on the active power sharing. The active power sharing can be always achieved among DGs when P 2 = 2P 1.

  2. 2)

    The absolute value of ρ i , i.e., degree error of the reactive power sharing of DG i , increases gradually along with the increase of K e.

  3. 3)

    In addition, Q 2 increases gradually along with the increase of K e, but Q 1 decreases except that when K e = 0.2. At this time, the reactive power output of DG1 is second largest in Fig. 8b and Table 3 because the rate of drop of voltage is larger than rate of increment of K e. What is more, the reactive power output of DG2 is always larger than the desired power, and reactive power output of DG1 is less than the desired power. ρ 1 > 0 and ρ 2 < 0 when \( X_{\text{L1}} - 2X_{\text{L2}} > 0 \), which are consistent with (33) and the simulation results in Section 5.1.

  4. 4)

    However, the AC common bus voltage drops slightly with the decrease of K e. Meanwhile, it is a pity that the total output powers of the two DGs decreases slightly due to the drop of ac common bus voltage. So choosing the suitable K e is necessary to ensure voltage robustness and power sharing.

6 Conclusions

In this paper, a new robust droop control method for high-voltage microgrid is proposed. Based on the theoretical analysis and simulation studies in PSCAD/EMTDC, the following conclusions are drawn.

  1. 1)

    As the dominated part of grid impedance, the impact of transformer impedance on the reactive power sharing accuracy is mitigated significantly. Because the local signals at the high voltage side of transformer are detected as the feedback signals within the proposed robust droop control method, so the proposed method is easy and effectively to apply in such high-voltage microgrid.

  2. 2)

    The drawback caused by the line impedance difference can be eliminated with the optimal selection of robust coefficient K e to a large extent. Besides, the ac common bus voltage can be maintained in a certain range with the proposed robust droop control.

  3. 3)

    The proposed robust droop control can not only be used in multi-microsource and high-voltage microgrid with centralized loads but also with distributed loads. Besides, the proposed robust droop control is easy to be implemented.