Energy Management System for DC Microgrids Considering Battery Degradation

DC microgrids require sources that maintain dc microgrid voltages, with Battery Energy Storage Systems (BESSs) being a good option for this task, given their multiple control alternatives. However, BESS cycling is an issue, and thus, this paper proposes an Energy Management System (EMS) for dc microgrids that considers battery degradation. Therefore, an EMS model is proposed and discussed, demonstrating its application for a practical dc microgrid of a building in Xiamen University. The simulation results show that the proposed EMS, which accounts for BES degradation costs, is an effective tool that avoids frequent battery charging and discharging, while maintaining the required dc bus voltage.


INTRODUCTION
DC microgrids are being considered as an efficient structure for the integration of Renewable Energy Sources (RES), such as solar and wind power, providing more adequate and simpler system controllability than conventional ac microgrids [1]. Thus, dc microgrids do not need to deal with frequency issues, and, although voltage control is required, reactive power management is not necessary. In terms of Energy Management Systems (EMSs), dc microgrids are simpler than ac systems, since there is no need to deal with the reactive power nonlinearities. Furthermore, ac microgrids are three-phase power systems, leading to complex and computationally expensive EMS problems based on linear and nonlinear programming models [2], which is not necessary in dc microgrids, as these are simple two-wire systems.
EMSs are part of the top control levels from a hierarchical control perspective in dc microgrids [3], [4]. A typical dc microgrid and its control scheme is shown in Fig. 1, where the interactions of the generating sources are coordinated by a central controller (CC). Previous research works have focused on the primary and secondary control levels of dc microgrids, as for example in the case of droop controls [5]. On the other hand, various EMS papers have mainly focused on energy regulation at lower control levels. Thus, for example, in [6], dc bus voltage levels are used to indicate the system state, so that energy storage can be regulated through the dc bus, focusing on system dynamic techniques as opposed to optimization approaches for optimal system operation. A rule-based EMS is proposed in [7] for dc microgrid operation, as opposed to applying optimization-based techniques for energy management. To the best of the authors' knowledge, there are not many papers on dc microgrid EMS based on optimization techniques similar to those that have been proposed for ac microgrids.
DC microgrids require a generating source to maintain the microgrid bus voltage. Usually, grid-connected bidirectional converters are used for these purposes, since the utility grid has sufficient power to adjust to the dynamic power changes of the microgrid. However, when the microgrid is operated in islanded mode, a local generating source must maintain the system dc bus voltage, using typically Battery Energy Storage Systems (BESSs) for these purposes, given their bidirectional power flow capability, and their ability to regulate voltage through interface converters. In fact, the need for a seamless switch for on-grid and off-grid operating modes can be avoided by using BESS to regulate bus voltages in dc microgrids directly [8]. 978-1-7281-5508-1/20/$31.00 ©2020 IEEE [9]. Therefore, to increase a BESS lifespan, frequent charging and discharging needs to be avoided by maintaining the battery's State of Charge (SoC) at certain healthy levels, thus minimizing battery degradation.
There are several research works considering battery degradation costs in ac microgrid EMS. Thus, for example, a cooperative distributed energy scheduling algorithm is proposed in [9], studying the impact of DoD on battery lifespan. In [10], the authors present an EMS based on a stochastic dual dynamic programming model that considers battery degradation costs for optimal microgrids operation. The authors in [11], study battery degradation costs compared with utility electricity costs in ac microgrid operation. These and other papers focus on battery degradation in the context of ac microgrid EMS; however, the relevance of battery degradation in dc microgrid EMS has not yet been considered, to the authors' best knowledge.
Based on the aforementioned shortcomings in the existing technical literature, the current paper focuses on an EMS design for dc microgrids. The battery storage is used to maintain the dc bus voltage in the microgrid, and a battery degradation model is integrated into an EMS system model, proposing a degradation cost model for the EMS objective function. The proposed EMS model is simulated on a model of an existing grid-connected dc microgrid of a building in Xiamen University. Hence, the main contributions of this paper are the proposed function to represent battery degradation costs in dc microgrids, and an EMS model for optimal operation of BESSbased dc microgrids.
The rest of this paper is organized as follows: Section II reviews the basic control approach for dc microgrids, discussing the possible working modes of these microgrids. Section III describes the proposed EMS and battery degradation models, and a realistic case study and associated simulation results to demonstrate of the application of the proposed EMS are presented and explained in Section IV. Finally, relevant concluding remarks are provided in Section V.

A. EMS Overview of DC Microgrids
The hierarchical control scheme shown in Fig. 2 is normally applied in dc microgrids [3]. Primary controls deal with the direct control of power electronic devices, and secondary controls are associated with dc bus voltage regulation, typically through droop controls. An EMS would be normally part of a tertiary control level, with either centralized or distributed control approaches, in a time scale of seconds to days. In this context, the EMS aims to schedule dispatchable microgrid generation in the most efficient manner to, for example, minimize running costs and power losses.

B. DC Microgrids Operation
In dc microgrids, there must be at least one dc voltage source to regulate the bus voltage [8], with the configuration of a microgrid determining how power is dispatched. For instance, if grid interface converters are used to maintain the bus voltage, the grid would take care of any power shortage or surplus from the microgrid. However, when the system is operated in islanded mode, other converters must be assigned to maintain dc bus voltage. In most cases, BESSs are used for these purposes in isolated mode, given its bidirectional power flow capabilities. For grid-connected systems, if a BESS is tasked with constantly maintaining the microgrid dc bus voltage, a seamless switch is not required for proper system operation. An equivalent control model of a generating source in a dc microgrid is illustrated in Fig. 3, where is the controller reference voltage; is the converter terminal voltage before considering the line resistance; is the bus voltage; is the current injected to the common dc bus from the distributed energy sources; is the virtual resistance; and is the line resistance. A simplified model of a dc microgrid is shown in Fig. 4 only one in the system, droop control is not required; however, if there is more than one, droop control is needed due to large circulating currents, because the output impedance at low frequencies tends to zero in tightly regulated dc-dc converters.
On the other hand, if the line resistances are considered, the power injected from current and power sources is not the same, since for the power source, some of the power is consumed in the line resistance, increasing the terminal voltage. Without droop control for these sources, the system will lack controllability, since the line resistance is non-controllable, as opposed to the converter virtual resistance. For microgrid EMS, line resistances may be neglected, given the small impact that the grid has in a microgrid for dispatch purposes, due to the relatively short length and high amperage of wires, as demonstrated for ac microgrids in [2].

A. Battery Models
For BESS, the charging/discharging effect on SoC can be represented by the following constraint [12]: where , is the battery SoC at time ; is the charging efficiency; is the discharging efficiency; , −1 is the charged power at time − 1; , −1 is the discharged power at time − 1; and is the time step. The BESS minimum and maximum charging and discharging power constraints at time can be written as: , Based on the battery degradation model in [9], the degradation cost in $/kWh can be represented as follows: where is the battery's total replacement (capital) cost (assumed here to be 200 $/kWh [13]); is the total energy throughput in a lifecycle; is installed battery power; ℒ ( ) is a mapping function of battery lifecycles to ; and is the battery unit cost in $/kWh.
The ℒ ( ) for a lead-acid battery in this paper is shown in Fig. 5. Note that large DoDs reduce the cycles to failure, which indicates that over-discharging will reduce the battery lifespan, even though the total throughput would not change. The total battery degradation cost can then be written as: Observe that to reduce degradation costs, the BESS power usage, i.e., charging and discharging, should be reduced. This limits the BESS contributions to the grid.

B. Operational Constraints
The battery storage and grid interconnection are both bidirectional but cannot both charge and discharge at the same time. Thus, the following constraints reflect this constraint: In dc microgrids, there is no reactive power, and the power injected to the grid should be balanced with the power demand. This power balance constraint can be written as: , + , + , = , + , + , + , + , (9) where , is electrical vehicle power, and , represents the LED light power.

C. Objetive Function
For a grid-connect dc microgrid, the following multiobjective function can be defined: where , is the local electricity price, and represents a weighting factor, assuming that the battery degradation and grid electricity costs are complementary. This factor can be seen as a competition index of the power throughput between battery storage and utility grid, and thus can be manually adjusted to define the BESS contributions to the dc microgrid.

A. Formulation of the DC System
The test dc microgrid used here is based on a dc building with classrooms and labs at Xiamen University [14], which can be represented as in Fig. 6. This system consists of a set of 150kW PV panels; a 200Ah, 336V lead-acid battery bank, with its parameters depicted in Table I, where the value is chosen based on Fig. 5 and the desired battery lifecycles; and a 160kW bidirectional ac-dc converter for grid connection. On the load side, there are 20kW of LED lights, a 40kW Electrical Vehicle (EV) charging station, and a 30kW Air Conditioning (AC) system. In this system, the battery storage is used to maintain the bus voltage, and line resistances are neglected given the small-scale dc grid, as previously argued.
The PV generation profile is shown in Fig. 7 [15], corresponding to a cloudy day in the summer. The load profiles are depicted in Fig. 8, where the LED load was extracted from [16], the EV charging of the building's fleet vehicles is assumed to take place during off-work hours, and the air conditioning is an intermittent load operating only during the building's operating hours. The local daytime electricity price is 0.077$/kWh from 8am to 10pm, while for the remaining time is 0.044$/kWh. The dispatch time interval is assumed to be 15 minutes.

B. Simulation Results
Applying the optimization model described in Section III, the resulting BESS SoC and power consumption are depicted in Fig. 9 and Fig. 10 for different values of . Observe that when the degradation cost component increases as decreases, the BESS SoC flattens out, while the charging and discharging power tends to be zero, as expected. Note also that for values of ≤ 0.8, the battery is not dispatched, which illustrates the sensitivity with respect to of the battery degradation costs versus grid costs in (10     Two representative optimal dispatch solutions are shown in Fig. 11 and Fig. 12. In Fig. 11, note that some of the excess PV power is used to charge the BESS. However, when decreases, i.e., as the battery degradation cost share increases, this is not the case, resulting in a reduced use of the battery in the dc microgrid, and thus a longer lifespan. Based on the presented variations of the dispatch solutions, costs, and battery use with respect to , users should be able to choose the value of that best suits their operating strategy, i.e., minimization of costs and/or battery degradation. For example, as the battery ages, battery degradation could be given priority over costs by decreasing the value of .

V. CONCLUSIONS
An EMS model for dc microgrids was proposed in this paper, considering basic control strategies in these types of systems, as well as battery degradation, which reduces the battery storage participation in dc microgrids to increase its the lifespan. The effectiveness of the proposed EMS was demonstrated through simulation results on a practical dc building in Xiamen University, where battery storage maintenance is an issue. The air conditioning and EV loads were modeled as non-dispatchable; however, in future EMS model enhancements, these should be considered as controllable loads.