Consensus Control of Distributed Battery Energy Storage Devices in Smart Grids

1. Introduction

Integration of highly distributed renewable energy sources has introduced significant challenges to the resiliency and efficiency of the smart grid systems. This is mainly due to uncertain behaviors of these renewable energy sources and their dependency to weather conditions [1, 2]. Battery energy storage has been introduced as a solution to solve the intermittency and uncertain behaviors of renewable energy sources in smart grids. The energy storage is normally connected to the electric power system through a voltage source converter and by controlling the charge/discharge rate of the storage units, output power regulation of renewable energy sources can be achieved [3, 4].

Currently, energy storage units are distributed throughout the grid. Given the centralized structure of supervisory control and data acquisition (SCADA) system, it cannot meet the requirements of highly distributed renewable energy and storage devices of future smart grid systems. In addition, the centralized controllers suffer from single point of failure and are not a suitable choice for energy storage control when the grid resiliency is significantly important [5, 6, 7]. As an example, an energy storage device which is permanently out of energy is not able to use its power capability to support the load until a charge event is scheduled. Moreover, any charge discrepancies will result in battery degradation due to the increased depth of discharge.

Distributed control of distributed energy storage units has recently been introduced [8, 9, 10, 11]. For example, a coordinated control is proposed for low voltage distribution networks in [8], which mitigates voltage fluctuations in a distribution feeder using distributed storage units. As another example, the voltage regulation issue of distribution feeders is resolved using a droop-based distributed controller to cooperatively charge/discharge the storage units to regulate the feeder voltage considering the state of charge of the batteries [9]. In Hammad et al. [10], a virtual inertia-based distributed controller is designed for transient stability of a power system using distributed storage units. The authors then used a feedback linearization control algorithm to evaluate the proposed virtual inertia and its effectiveness in stability of the system.

Consensus control of energy storage units has also recently been proposed as an emerging technique for synchronization of distributed storage devices [12, 13, 14, 15, 16, 17, 18, 19, 20]. In Khazaei and Miao [13], the authors introduced a state of charge balancing algorithm for distributed storage devices in AC microgrids using consensus theory and validated the results using a real-time simulator. In Guan et al. [15], a dynamic consensus approach was introduced to balance the discharge rate of energy storage devices in AC microgrids. The proposed model achieved power regulation by adjusting the virtual resistance of voltage-controlled inverters. A novel distributed controller was designed for load management in distribution networks using distributed battery storage systems. The proposed methodology used limited communications to coordinate multiple storage units with solar power energy penetration. The authors also have significantly studied the consensus design for storage devices for power sharing and energy synchronization [13], power sharing of heterogeneous storage units with droop control [14], voltage and frequency regulation of storage devices in smart grids [16], output power regulation of double-fed induction generator-based wind farms [19], and power sharing of storage devices with different droop schemes [20].

This chapter summarizes the findings of the authors in the distributed control design of energy storage devices in smart grids to provide ancillary services including: (1) voltage regulation, (2) primary frequency support, (3) equal active power sharing between storage units based on their capacities, (4) equal reactive power sharing based on storage capacity, and finally, and (5) controlling the load in both islanded and grid-connected modes. Time-domain simulations on a modified IEEE 14-bus system are performed to validate the effectiveness of the proposed designs.

The rest of the paper is as follows: Section 2 covers the battery energy storage model. Consensus design for heterogeneous storage units is considered in Section 3. Section 4 includes multiple case studies to validate the designs, and Section 5 concludes the chapter.

2. Battery energy storage model

A basic schematic of an energy storage device, which is connected to the grid through a DC/AC converter and an LCL filter is illustrated in Figure 1. The energy storage control uses the well-known synchronous reference frame control, where three cascaded control loops are adopted. The main objective is to control the active and reactive demand using a droop control method in dqreference frame. The droop control receives active and reactive power measurements from a sensor installed at the point of common coupling (PCC) as

where Pmiand Qmiare measured active and reactive powers, respectively. In addition, vod,voq,iod,and ioqare measured converter voltages and currents at the point of common coupling, as illustrated in Figure 1. The measurements will then pass through a low-pass filter,

where αpand αqare the bandwidths of the low-pass filters. The AC-side dynamics of the ithenergy storage system in dqframe is expressed as:

where vodand voqare dqvoltages at the point of common coupling (PCC), vidand viqare the dqframe converter output voltages, and iidand iiqare the dqreference frame currents flowing from the PCC to the converter.

3. Heterogeneous consensus design

The main objective of this section is to supplement a secondary controller to the droop controller of the storage devices. The controller receives information from neighboring storage units and shares the power between storage devices to regulate the voltage and frequency of the system at the point of common coupling. Furthermore, a virtual leader is considered, which can be assigned to one energy storage in the system, or a few storage devices. The leader will have the setpoints of the voltage and frequency in the system and will share the information with its neighboring storage units. To develop such a control design, the battery energy storage model needs to be developed. In our recent work, a simplified battery energy storage model was developed. The model accounts for the dynamics of the droop controller and active power/energy relationship of the battery. Such model can accurately incorporate the dynamics of energy storage devices in smart grids. Dynamics of the energy storage devices can be represented by [14]:

To develop such simplified model, it is assumed that dynamics of voltage controller and current controller are much faster than the droop controller, therefore, their dynamics can be ignored. In the above model, uiPis the input for distributed active power sharing, and DiPreflects the heterogeneity of batteries. To achieve equal power sharing, DiPPishould be regulated among batteries so that a battery with higher capacity (lower droop gain, DiP) contributes more to the power sharing. To minimize the number of communication links between the storage devices, this paper regulates the nominal voltage and frequency of neighboring storage units. In this method, there will be no need to receive measured voltage and frequency signals from neighboring storage units and the control design only requires the nominal frequencies ωjnomand nominal voltages ∣Vjnom∣of its neighboring storage devices.

To provide voltage and frequency regulation as well as active/reactive power sharing, new distributed inputs can be designed using consensus theories. These inputs include ∣V̇inom∣=uiV, Q̇i=uiQ, and ω̇inom=uiω. The overall dynamics of the ithenergy storage device is then formulated as [14]:

In the next section, control design to develop these new inputs will be elaborated in detail.

3.2 Consensus control design

Let ωrefand Vrefbe the reference frequency and voltage magnitude of energy storage devices. These references serve as external commands to force the frequency and voltage magnitude of storage devices to converge precisely to the expected values. In other words, they are virtual leaders while the frequency and voltage magnitude of batteries are the followers. In this sense, the consensus design is proposed as follows,

uiP=−C2DiP∑j∈NiDiPPi−DjPPj,uiV=−C3∑j∈NiVinom−Vjnom−C0Va0iVinom−DiQQi−Vref,uiQ=−C3DiQ∑j∈NiDiQQi−DjQQj,uiω=−C2∑j∈Niωinom−ωjnom−C0ωa0iωinom−DiPPi−ωref.E18

Note here that the controller gains for the active power and frequency are the same, and similarly the controller gains for the reactive power and voltage magnitude are also the same. The structure of the proposed distributed primary and secondary voltage/frequency controller is illustrated in Figure 2.

4. Case studies

To validate the proposed designs, IEEE 14-bus benchmark is used. The system represents an approximation of U.S. utility system around 1962. The benchmark includes five generation units and 11 loads. Parameters of the test system were adopted from [21]. The system was modified for the current study. The generator dynamics were replaced by the battery energy storage dynamics. The system was modeled in MATLAB Simulink and a combination of MatPower and MatDyn toolboxes are used for dynamic simulation of the proposed control algorithms [22]. The MatPower toolbox was used for power flow and initial conditions of the system, where MatDyn was used for dynamic simulations and control design. The authors have extensively studied integration of battery energy storage units to IEEE benchmark cases using MatDyn toolbox in their previous publications [9, 10]. The consensus controllers then are supplemented to the model as inputs. The schematic of the modified IEEE 14-bus system used for the simulations in this study is illustrated in Figure 3. Parameters of the storage units are included in the Appendix section.

The communication structure of the system under investigation is illustrated in Figure 4. As it can be observed, the communication graph of the system is undirected and minimum number of communication links is needed to ensure the convergence of the proposed algorithms. Furthermore, the leader is incorporated in battery energy storage number 2. The leader can be designed in any storage number as one leader is sufficient to ensure the functionality of the proposed control design.

4.1 Constant power control

In the first case, the performance of the designed voltage and frequency controller is tested when a constant load is applied to the system. The secondary controller ensures sharing the active and reactive powers between the storage units based on their capacities as well as voltage/frequency regulation. Simulation results for this case study are illustrated in Figure 5. For this case study, the leader is activated by setting C0ω=1and C0V=1. The overall load in the system (summation of loads in all busses) is SD=0.5+j0.5p.u. It can be shown that the active/reactive power sharing (first subplot) is achieved after 15 s, and the voltage and frequency are regulated to their reference setpoints (1 p.u.) after 15 s. The designed controller can synchronize the operation of storage units in the system very fast.

4.2 Primary frequency response

In the second case, the primary active and reactive power sharing is the main objective. The batteries should share the load power equally using the primary droop concept. The leader will be deactivated in this case enforcing consensus gains C0ωand C0Vto zero. This will result in a primary voltage and frequency response, where the battery storage units will share the load active and reactive power demand, but the voltage and frequency will deviate from the nominal value. Similar to the first case study, the system starts with a 0.5+j0.5p.u. load and a load change event is scheduled to increase the demand to 1+j1p.u. after 20 s. Simulation results are illustrated in Figure 6. It is shown that the batteries can equally share the active and reactive power of the load even after the load event at 20 s. To support the active power increase in the demand, the frequency will drop and the batteries will settle in a lower frequency (0.91 p.u.). Furthermore, since the load active power has increased after 20 s, the voltage will also drop and settle to new synchronized value (0.94 p.u.).

4.3 Secondary frequency response

The third case, the performance of the secondary voltage and frequency controller during a load change event is studied. The system starts with 0.5+j0.5p.u.load and the load increases to 1+j1p.u.after 20 s. The secondary distributed controller is reactivated by tuning C0ωand C0Vto 1, as illustrated in Figure 7. It is observed that the load sharing is successfully achieved among the storage units. In addition, the voltage and frequency are regulated to 1 p.u. in less than 3 s after the load change.

4.4 Severe load change

In the last case, the performance of the proposed power sharing and voltage/frequency restoration algorithms in handling a severe load change is examined. The system initiates with 0.5+j0.5p.u.load and a load change of 1 p.u. is applied after 20 s. This means the total load increases to 1.5+j1.5p.u.after 20 s. As it can be observed in Figure 8, the proposed distributed controller can equally share the active and reactive load change while regulating the voltage and frequency. It should be noted that the frequency drops to 0.75 p.u. after the load change, but it is quickly recovered within a few seconds. Similarly, the voltage drops to 0.6 p.u. after the load change, but it is recovered within 2 s. This case study showed that the proposed controller can successfully operate under severe load changes.

5. Conclusion

This chapter proposed a novel distributed controller that can synchronize the operation of distributed energy storage units in smart grids. By minimizing the communication links between the neighboring storage units, the storage units share information (voltage/current readings) with their neighbors to share the active/reactive demand based on their capacities. Furthermore, a virtual leader is designed and supplemented to one storage unit to regulate the voltage and frequency and to provide secondary frequency response to the system. Results showed the effectiveness of the proposed algorithms in equally sharing the active power and reactive power of load during constant power load, and load change events. Furthermore, the secondary controller could successfully regulate the voltage and frequency of the system during constant load, load change, and severe load change events.

6. Future work

Future studies will focus on: (1) hardware validation of proposed approaches and (2) expansion of the developed controllers to solar and wind energy applications.

Conflict of interest

The authors declare no conflict of interest.

Appendices