1. Introduction
Distributed energy storage is the key issue to solve the issue of grid-connected renewable energy generation. For example, it can improve the ability of the grid to accept wind and photovoltaic (PV) power [
1,
2,
3]. A typical DC microgrid structure is mainly composed of a distributed generation unit, an energy storage unit, a load cell, and a grid-connected converter [
4,
5,
6], as shown in
Figure 1. DC microgrid research focuses on stabilizing the DC bus voltage to ensure the power balance of the system. To stabilize bus voltage fluctuations and solve energy supply volatility issues, adding energy storage devices can improve the device’s voltage sag and the inrush issues caused by load-switching, changes in natural conditions, and instantaneous faults in DC microgrid systems; this improves the reliability and scheduling flexibility of the distributed generation grid connection. Using low bandwidth communication control reduces the long-distance stability and DC microgrid distribution of the system due to network delay packet loss and other issues. To avoid the risk caused by long-distance communication control, each unit is divided into three control layers according to the normalized voltage of DC bus and coordinated control of various units [
7]. Compared with the two-level converter, the three-level converter has only half of the switches to change state peer cycle, and the voltage stress on the switch is only half of the bus voltage [
8,
9].
As a key device connecting the energy storage unit and the DC bus, the bidirectional DC/DC converter requires a high-voltage and high-power bidirectional DC/DC converter topology. However, the traditional two-level bidirectional DC/DC converter topology is not suitable for high-voltage and high-power applications, requires a multi-level topology, and has the characteristics of simple structure, high efficiency, reliability, and easy modular expansion. The three-level bidirectional DC/DC converters suitable for accessing the DC microgrid are mainly isolated [
10,
11,
12,
13] and non-isolated [
14,
15,
16]. For energy storage side modular bidirectional DC/DC converter, when using input parallel, output parallel (IPOP), the current sharing issue between modules needs to be considered; when using input series, output series (ISOS), the voltage equalization issue needs to be considered; when using input parallel, output series (IPOS) or input side series connection, output side parallel (ISOP), both current sharing and voltage equalization issues need to be considered [
17,
18].
The DC voltage represents the state of power balance. When the renewable energy generation is greater than the load demand, the DC line-of-sight voltage rises. When the renewable energy generation is less than the load demand, the DC line-of-sight voltage decreases. The energy storage medium is inconsistent with the state of charge (SOC) of the energy storage unit due to the process and external environmental factors and exhibits randomness. Considering the randomness and slow change of the state of charge of the energy storage unit in the distributed energy storage system, an improved SOC power exponential control strategy is adopted in the charging and discharging process of the energy storage unit to find the optimal droop curve to make the system fast [
19,
20,
21,
22,
23,
24,
25] and then converge to equilibrium. When distributed energy storage is suitable for low-bandwidth communication, SOC power exponential droop control should be adopted to achieve fast and accurate distribution of load current. In the case of communication failure or unsuitable conditions for communication, switch to emergency state to continue operation.
If the energy storage system is connected to the DC microgrid to participate in the constant regulation of the bus voltage, the energy storage unit quickly reaches the maximum allowable number of charge-discharge cycles and shortens the service life. It is necessary to quantify and analyze the best solution through voltage fluctuation level and power fluctuation level quantification and compensation scheme, to achieve the lowest voltage fluctuation and power fluctuation in the DC microgrid, and to provide high quality electric energy for a large grid [
26,
27,
28,
29]. Pre-adjusting the SOC to the optimal state during the first layer control is to prepare for the full play of the second-layer control function of the energy storage unit.
3. Distributed SOC Droop Control
Distributed droop control with low bandwidth communication has high reliability, high redundancy, and meets the demand of distributed generation. Therefore, it has been widely used and tested in recent years [
19]. In [
20], authors use voltage droop control to obtain the current reference value of each converter based on the reference power divided by voltage, thus achieving power allocation between converters. Average current control is added to the droop control to realize the average load distribution [
21]. In [
22], the authors consider voltage fluctuations and power-sharing while a converter is removed in the system. Papers [
23,
24,
25] propose droop control power by state of charge (SOC) distribution; thus relating the droop coefficient with the storage module SOC. When discharging the droop coefficient
mp, it is proportional to 1/
socn. When charging the droop coefficient
mp, it is proportional to
socn; thus, the larger n, the higher the SOC equilibrium velocity and the lower the average precision. In [
26], the equivalent capacity storage unit adopts a two-layer control strategy. The energy storage unit of SOC tends to average A
soc at discharging mode. The droop coefficient
kd is proportional to
exp[-
p(
soc-A
soc)] at charging mode, the droop coefficient
kd is proportional to
exp[
p(
soc-A
soc)]. Therefore, as
p increases, the average speed will increase, and the accuracy of equalization will reduce. Consequently, the estimation accuracy of SOC will impact the accuracy of load distribution. The energy storage unit is mainly connected to the DC microgrid during the control of the second layer. The energy storage unit allocates the output power rationally, according to the SOC information, so that the energy storage unit can reach a consistent state rapidly and accurately. Due to the adoption of low bandwidth communication control, the communication failure conditions must be considered as that will affect the system stability.
To solve the SOC equilibrium speed and the accuracy of the energy storage unit, we propose to improve the SOC power exponential droop coefficient. The method can transmit the load quickly by only transmitting the SOC and udc. Without the impedance information of the line, the output voltage can be stabilized by restoring the average output voltage of the port. Experimental results show that the proposed SOC power exponential droop control can improve the SOC balancing speed and accuracy of the energy storage unit.
Distributed storage systems require an energy storage unit with a high SOC to emit more power when discharging, but to have less power absorption when charging. This can be realized by fast equalization of the SOC and has no effect on the stability of the system; thus, it can take the effect of line impedance because the SOC information is realized by indefinite integration and the rate of change is low. Thus, the response coefficient is not sensitive enough according to the SOC coefficient, which affects the equipartition effect. In this paper, an improved SOC power exponent droop is proposed to improve the SOC resolution, and the steady DC link voltage is stabilized by the SOC power exponential droop control, so that the system converges to the SOC equilibrium state quickly.
where,
.
Droop control can cause a drop in the DC bus voltage, so we should select a normal droop coefficient in a suitable range of the DC bus voltage drop. When the virtual droop coefficient
R(soc) is large enough and satisfies requirement (2), the line impedance can be ignored. Equation (1) shows that soc
n tends to the average value of A
socn and
R(
soc) is equal to
kD; thus, to overcome the line impedance, the
kD>
r condition needs to be satisfied. While the energy storage unit is charging,
idc< 0, the SOC is above its average value,
R(
soc) >
kD, and the energy storage units absorb less electricity. When the SOC is below the average value,
R(
soc) <
kD and the energy storage units absorb more electricity. When the energy storage unit discharges,
idc > 0, the SOC is above the average value,
R(
soc) <
kD, and the energy storage units absorb less electricity. When the SOC under the average value,
R(
soc) >
kD and the energy storage units absorb more electricity. In the whole process of charge and discharge, it meets
R(
soc) +
RLoad >>
r and should overcome the line impedance effect of
r in terms of the load distribution. The parameters of
R(
soc) have great influence on the droop coefficient, which is shown in
Figure 8.
Figure 8 shows that the
R(
soc) surface is relatively flat, which is conducive to the design parameters. If soc
n is different exactly from average A
socn, mainly affected by
p and
n. The output current differences Δ
idc and the remaining battery difference Δ
soc have a relation:
The change curve of the output current difference Δidc(Δsoc) can be obtained by Equation (2). The n is fixed, the larger the p is, the faster the output current difference changes. However, when Δsoc approaches 0, the current’s difference is not obvious; thus, the output current’s regulation is weak, which will affect the equilibrium rate of the SOC and needs to regulate R(soc) with n fit. The output current difference Δidc can be adjusted quickly in the whole range of SOC change. At the same time, considering the actual use of the lithium iron phosphate battery, the probability of Δsoc > 0.5 is low when |Δsoc| is close to 1. The output current difference of convergence is a constant when |Δsoc| = 0, idc tends to 0, and the variation curve of the output current difference Δidc is relatively flat.
The energy storage medium of the distributed energy storage system is a lithium battery, which has the characteristics of high density and high energy density, and thus has a long charge-discharge cycle. In order to study the SOC equalization speed, precision, and voltage drop of the DC bus with the change of parameters in distributed droop control, lithium battery capacity 0.5 Ah, rated voltage 200 V, and charge-discharge cycle 20 C were set. MATLAB/Simulink comparative analysis of SOC droop control and bus voltage secondary regulation was done to find a suitable strategy for distributed energy storage control in the DC microgrid. The system structure consisting of two energy storage units, load, and adjustable power supply is shown in
Figure 9.
Simulation results of the effect of droop on parameter changes are shown in
Figure 10.
Figure 10a–d include A
socn, which indicates that the SOC power exponential droop control needs to be created by low bandwidth communications. When
n = 1, the average load current sharing rate is slower (Δ
idcmax = 5.2 A), and the SOC convergence accuracy is lower (Δ
soc = 2 ~ 5% at 100 s). When
n = 2, the load current sharing rate is faster (Δ
idcmax = 8 A), and the SOC convergence precision is high (Δ
soc = 2% at 70 s and Δ
soc = 1% at 100 s). Increasing
kD increases the SOC equalization but also increases the DC bus voltage drop.
Figure 10e–f, excluding the A
socn terms, suggests a communication loss. After the communication is lost,
Figure 10e
kD remains the same. When discharging, the
R(
soc) is very small, and it is too weak to adjust the load of the current. When charging,
R(
soc) is larger (beyond the allowable range) and turns into the traditional
U-
I droop control. After the communication is lost,
Figure 10f
kD is changed according to the charge-discharge mode (
kD = 10 at discharging and
kD = 0.1 at charging). The systems can still work stably, however, the SOC equilibrating time is longer and the accuracy is lower. As seen in
Figure 10g–h, the SOC equalization accuracy and speed can be adjusted by
n. However, the SOC equalization accuracy and speed are lower than
Figure 10a–d and the DC bus voltage drops further. Droop control leads to voltage drops at the access point, reducing the DC bus voltage quality. To realize rapid and accurate distribution of the load current and compensate a bus voltage drop at the same time, the DC bus voltage is adjusted for a second time and the output characteristic curve is translated to achieve the rated value to enhance the stability of the bus voltage. If the
upcc is used as the feedback value at the common point, the DC bus voltage deviation is within 1 V, as shown in
Figure 10k.
If the average value of the outlet voltage of the energy storage converter (
udc1 +
udc2)/2 is used as the feedback value, the DC bus voltage deviation is within 10 V, as shown in
Figure 10l. The bus voltage secondary control is added based on the droop control of the SOC power exponent. As the discharge-current of the energy storage unit increases, the charging current of the energy storage unit decreases, which has little effect on the SOC convergence accuracy or the load current distribution accuracy. Comprehensive analysis of
Figure 10k in the fast and accurate distribution of load current DC bus voltage shows that the deviation is smaller; once the communication line fails, it automatically switches to
Figure 10f mode to continue running. Simulation analysis can achieve the design requirements, but also needs further experimental verification.
The experimental comparison verification of SOC droop control is shown in
Figure 11. The soc
n droop control experiment is shown in
Figure 11a–b. The droop rate is faster at the start time and the voltage shift is larger, which is consistent with the simulation
Figure 10h. The disadvantage is that when soc approaches 0, the voltage drops by a larger amount. The exp[p (soc
n-Asoc
n)] droop control experiment is shown in
Figure 11c–d, SOC can quickly and effectively adjust the load current distribution in the [0 1] interval and the voltage offset is small, which is convenient for the DC bus voltage secondary recovery control.
4. Hierarchical Coordinated Control Strategy
Considering the high cost of the energy storage unit, it should be connected to the DC microgrid in layers to achieve a reasonable allocation of resources in practical applications. In order to provide high-quality power to the large power grid, the quantification standards of the DC bus fluctuation range and the working range of each converter are further discussed to maximize the stability of the DC bus voltage and grid-connected power fluctuation. The hierarchical coordination control structure of the DC microgrid is shown in
Figure 12 and
Figure 13. In the first layer, the bus voltage is controlled by the grid-connected inverter, and the grid-connected inverter is equivalent to a resistive load. In the second layer, the energy storage unit stabilizes the bus voltage, and the grid-connected inverter is equivalent to a constant power load. The DC bus voltage fluctuation is positively correlated with power variation.
In
Figure 12, the first layer |Δ
Udc| ≤ 0.02, corresponds to the network-free mode, the second layer 0.02 <| Δ
Udc | ≤ 0.05 corresponds to the network current limiting mode, and the third layer 0.05 <| Δ
Udc | ≤ 0.1 corresponds to the current limiting or islanding mode. The first layer bus voltage | Δ
Udc | ≤ 0.02 corresponds to the microgrid-free mode. As the permanent magnet synchronous motor (PMSG) runs in maximum power point tracking (MPPT) mode, the grid inverter is used to stabilize the DC bus voltage. The energy storage unit is SOC-preconditioned and is ready to stabilize the DC link voltage in the second layer. After the power expended from the interaction between the large grid and the microgrid has reached a maximum and the bus voltage exceeds the range of layer 1, it enters layer 2 or 3. This can also occur if the output power of the PMSG or load suddenly changes.
The DC bus voltage 0.02 <| ΔUdc | ≤ 0.05 corresponds to a current limiting mode. As the PMSG runs in MPPT mode, the energy storage unit stabilizes the DC bus voltage and the grid inverter loses its ability to stabilize the bus voltage. When large grid and microgrid interactions reach the maximum power and the grid converter enters the current limiting mode, the energy storage unit switches from layer 1 to layer 2 to stabilize the DC bus voltage, and each energy storage unit performs load current distribution according to the SOC droop. The SOC droop ensures an efficient return to layer 1. If the DC bus voltage has not reached a new stabilization point under layer 2, the DC microgrid enters layer 3 and needs to reduce fan output power or enable load shedding.
The distributed energy storage unit is mainly connected to the DC bus at the second layer to keep the bus voltage stable and balance the bus power fluctuation. At present, the price of the energy storage system is relatively high, the maintenance cost is high in the later stage, and the number of charge and discharge cycles is limited. If the time difference is used, the energy storage system will be frequently charged and discharged, and the service life of the energy storage system will be reduced. If the adjustment is too slow, the wind power fluctuation will not be stabilized, and the system will be safely and stably operated.
In
Figure 13, this paper proposes three kinds of compensation schemes, based on the average power. Case 1, the output of the wind power is higher than the average power, the energy storage unit absorbs the wind power output. When the power is lower than the average power, the energy storage unit compensates for the wind power output. Case 2, the wind turbine output is more than two times the average power, the energy storage unit absorbs the wind power output. When the power is lower than the average power, the energy storage unit compensates for the wind power output. Case 3, the wind power output is more than three times the average power, the energy storage unit absorbs the wind power output. When the power is lower than the average power, the energy storage unit compensates. The wind power output is insufficient, and the system returns to the first layer control state under the control of the second layer energy storage unit. This section combines the actual operating data of four DC buses in the same area to compare and analyze the energy storage unit access schemes, assuming that the energy storage unit SOC has sufficient adjustment capability when controlling the second layer.
4.1. Case 1
If the wind power output is higher than the average power, the energy storage unit is charged; when the wind power output is lower than the average power, the energy storage unit is discharged. A comparison of the simulation results is shown in
Figure 14. After the compensation, the standard deviations of the four buses are 13.67 × 10
4 kWh, 12.52 × 10
4 kWh, 12.75 × 10
4 kWh, and 12.99 × 10
4 kWh, which are 9.58 × 10
4 kWh, 9.04 × 10
4 kWh, 5.64 × 10
4 kWh, and 7.99 × 10
4 kWh, respectively, before the compensation. The average outputs after compensation are 38.92 × 10
4 kWh, 31.47 × 10
4 kWh, 26.19 × 10
4 kWh, and 29.32 × 10
4 kWh, up 12.74 × 10
4 kWh, 7.08 × 10
4 kWh, 3.94 × 10
4 kWh, and 6.06 × 10
4 kWh, respectively, before compensation.
4.2. Case 2
If the wind power output is more than two times the average power, the energy storage unit is charged; when the wind power output is lower than the average power, the energy storage unit is discharged. A comparison of the simulation results is shown in
Figure 15. After compensating, the standard deviations of the four buses are 8.03 × 10
4 kWh, 7.35 × 10
4 kWh, 6.82 × 10
4 kWh, and 7.69 × 10
4 kWh, which are reduced by 15.22 × 10
4 kWh, 14.21 × 10
4 kWh, 11.57 × 10
4 kWh, and 12.99 × 10
4 kWh, respectively, before compensation. The average outputs after compensation are 38.92 × 10
4 kWh, 36.17 × 10
4 kWh, 32.49 × 10
4 kWh, and 34.5900 × 10
4 kWh, up 12.74 × 10
4 kWh, 11.78 × 10
4 kWh, 10.24 × 10
4 kWh, and 11.33 × 10
4 kWh, respectively, before compensation. The compensation effect of Case 2 is better than Case 1, and the energy storage unit is connected to the DC microgrid for moderate adjustment frequency.
4.3. Case 3
If the wind power output is more than three times the average power, the energy storage unit is charged; when the wind power output is lower than the average power, the energy storage unit is discharged. A comparison of the simulation results is shown in
Figure 16. After the compensation, the standard deviations of the four buses are 11.67 × 10
4 kWh, 11.55 × 10
4 kWh, 9.82 × 10
4 kWh, and 11.19 × 10
4 kWh, which are lower than the compensation by 11.58 × 10
4 kWh, 10.01 × 10
4 kWh, 8.57 × 10
4 kWh, and 9.79 × 10
4 kWh, respectively. The average outputs after compensation are 41.79 ×10
4 kWh, 39.39 × 10
4 kWh, 34.67 × 10
4 kWh, and 37.05 × 10
4 kWh, respectively 15.61 × 10
4 kWh, 15 × 10
4 kWh, 12.42 × 10
4 kWh, 13.79 × 10
4 kWh before compensation. The compensation effect of Case 3 is worse than that of Case 2, and the energy storage unit is moderately active.
Comprehensive comparison of the results of
Table 2: Case 1 can ensure that the SOC state of the energy storage unit is higher, but the energy storage unit frequently operates; Case 2 can ensure that the system power fluctuation is minimal and the energy storage unit action frequency is moderate; Power fluctuations are large. Comprehensive comparison shows that the second option is optimal. In order to give full play to the adjustment effect of the distributed energy storage unit on the microgrid, it is also necessary to adjust the SOC of the energy storage unit in advance in conjunction with the wind power forecast data to cope with the fluctuation of the bus voltage. In order to prevent off-site wind power accidents, the forecast and actual operational data should be uploaded to provide a decision-making basis for the superior supervisory unit.
The distributed energy storage unit is mainly connected to the DC bus at the second layer to keep the bus voltage stable and to suppress the bus line power fluctuation. If the time difference is used, the energy storage system will be frequently charged and discharged to reduce the service life of the energy storage system. If the adjustment is too slow, the wind power fluctuation will not be stabilized, and the system will be safely and stably operated.