Development of a PV/Battery Micro-Grid for a Data Center in Bangladesh: Resilience and Sustainability Analysis

Abstract

Energy resiliency plays an important role in the proper functioning of data centers as they heavily rely on an uninterrupted power supply to ensure smooth operation. In the case of a power outage, the data center’s operation might be hampered, which results in system downtime, data, and economic loss. This issue is severe in developing countries where power supply infrastructures are inadequate and conventional. Microgrids can be an effective solution in this regard. Although several studies developed microgrids to observe the energy resilience benefit for some critical facilities, critical facilities like data centers are often overlooked. In addition, sustainability analysis of a microgrid is also scarce in the present literature. Therefore, one new resilience and sustainability indicator has been developed and implemented in this analysis to fill this gap. For this, new indicators, such as the resilience cost index (RCI) and renewable energy penetration (REP), were used in this study. This study used HOMER version 3.13.3 and REopt software to simulate a robust photovoltaic (PV) and battery microgrid for a hypothetical data center in Bangladesh. A random (48 h) outage was assigned to witness the adaptability of the modelled micro-grid. The suitable size of PV and battery was found to be 249,219 kW and 398,547 kWh, respectively. The system’s USD 18,079,948 net present value (NPV) demonstrates the economic potential of utilizing PV and battery microgrids for data centers. The RCI of the system is found to be 35%, while the REP is 87%. The energy consumption saving of the system is USD 21,822,076. The system emits 652% less CO₂ than the grid. The result of this system is also compared with a diesel-based system. After comparison, it is found that the developed PV/battery microgrid provides better environmental and economical service than the diesel generator. During blackouts, the system keeps the data center powered up without interruption while improving energy resilience and lowering carbon emissions. The outcome of this current analysis can serve as a blueprint for other microgrid projects in Bangladesh and other developing countries. By integrating PV/battery microgrids, data centers can cut costs, reduce emissions, and optimize energy use. This will make data centers less reliant on grid services and more flexible to forthcoming development.

1. Introduction

1.1. Background

Various nations worldwide suffer badly due to the weak and inadequate grid infrastructure [1,2,3]. A weak or vulnerable grid does not have sufficient capability to manage fault currents, voltage fluctuations, harmonic distortion, and overall power stability [4,5,6,7]. Also, the weather has a notable impact on the grid infrastructure [8]. The world’s temperature has increased significantly in recent years [9]. Extreme heat waves increase the electricity demand, which puts pressure on the grid. As a result, grid outages can occur. In this context, a microgrid could be an effective solution to these problems because a microgrid provides independent and localized power generation [10,11]. Microgrids can be used to fulfill both thermal and electrical needs [12,13]. Due to these numerous benefits, microgrids have found their way into various applications [14,15,16].

The demand for energy is continuously rising rapidly due to the tremendous growth of the population [17,18,19]. This population growth is also the driving factor behind the rise of digital services. Digital services are in increasingly high demand around the world. Global internet traffic has increased 20 fold since 2010, while internet users have more than doubled. Although data infrastructure facilities and networks consume about 1–1.5% of the world’s electricity consumption, the rapid growth in energy demand from these sources has been tempered by these developments in energy efficiency. Data centres worldwide consumed approximately 220 to 320 TWh of electricity in 2021, representing 0.9 to 1.3% of global electricity usage, disregarding the additional 100–140 TWh utilized for Bitcoin mining [20]. Even while the world’s data center power usage has barely increased, several smaller nations with developing data center economies have experienced substantial development. As an illustration, the electricity consumption of data centers in Ireland has seen a notable growth since 2015, constituting 14% of the total electricity usage in the country in 2021 [20]. Denmark’s data center energy consumption is forecasted to triple by 2025, constituting about 7% of the country’s total electricity usage. An anticipated rise in global carbon emissions is expected due to increasing energy usage in data centers [21]. It is estimated that data centers across the globe contribute to approximately 1% of GHG emissions generated from energy consumption [20]. However, this amount needs to be halved to meet the targets of net zero emissions by 2030. Improving infrastructure efficiency, ensuring proper monitoring, and integrating sustainable energy resources can play a significant part in achieving this target [22]. Several studies discuss various initiatives for enhancing the energy efficiency of the data centers, such as “Dynamic Voltage and Frequency Scaling (DVFS)”, the application of energy and power management, enhancing the efficiency of servers and networks, the usage of direct current instead of alternating current, electric distribution efficiency, etc. [23,24,25,26,27]. However, the implementation of these initiatives has several shortcomings, such as a declined performance, rebound effect, and circuit breaking in the DC power supply [28,29]. Integration of renewable energy resources can be essential in reducing carbon emissions [30,31,32,33]. Several renewable energy sources such as wind, PV, biomass (biogas, biodiesel, etc.), and hydrogen have been identified and implemented worldwide [34,35,36,37]. Among these sources, wind and solar have been popular among researchers due to their economic and environmental benefits [38,39,40,41]. However, these sources are highly unpredictable [42,43]. Therefore, a battery is required to store the excess energy and supply the load when necessary [44,45].

1.2. Literature Review

Several renewable projects are underway or completed in integrating clean energy sources in data centres. For instance, tech giants like Apple, Google, and Facebook have invested in solar power plants, such as Apple’s 200 MW facility in Reno, Nevada, and Google’s combined 300 MW plants in Alabama and Tennessee, to lower their carbon footprint and promote sustainable practices [29]. In the literature, several studies can be found that deal with the simulation of integrating renewable energy (RE)-based microgrids in data centers. Guizzi and Manno (2012) analyzed the financial and energy management performance of a data center building’s natural gas reformer and fuel cell-based cogeneration system [46]. A TRNSYS model was developed by Kummert et al. to showcase a 5 MW chilled water system at a UK data center [47]. A TRNSYS thermal model for a data center’s combined cooling, heating, and power system was developed by Xu and Qu [48]. The study determined the system’s energy, economic, and environmental performance and provided recommendations to improve the energy utilization scenario. Two R&D buildings in Shanghai were utilized as case studies to simulate the energy performance and cost-saving potential in Energy Plus by Pan et al. [49]. Iverson et al. simulated different scenarios of a hybrid energy system utilizing hydrogen energy storage for data centers [50]. Zhu et al. analyzed nine energy-saving and five emission-reduction technologies for the future decarbonization of data centers. The study implemented the saving strategies for twenty case study data centers and found that about 20–40% and 15–27% energy savings can be obtained by optimizing and improving the IT and cooling equipment [51]. Hadad et al. simulated an optimal hybrid system with a 500 kW capacity, effectively incorporating PV panels, wind turbines, fuel cells, batteries, and hydrogen storage and primarily relying on PV and wind turbines with batteries as a backup [52]. Li et al. developed a simulation model for a hybrid system incorporating renewable energy sources to optimize the power consumption and thermal load of cooling and computing appliances [53]. A simulation study was undertaken by Khalaj et al. to analyze 42 hybrid systems comprising solar, wind, and battery power sources deployed in different data centers worldwide. The study found that about 80% and 50% of cooling power reduction and grid dependency can be obtained for the system [54]. Li et al. introduced a power management framework called “Greenworks”, explicitly designed for a sustainable data center that relies on a combination of renewable energy sources [55]. A standalone system comprising of PV, wind turbines, battery, and hydrogen systems were proposed for the optimal sizing of power and IT for a data center [56]. A mixed Linear Integer approach was applied to design a RE-based hybrid system to power up a data center [57]. Integration of Danish data centers and RE was performed by Monsalves et al. The study obtained outstanding results of 63% and 180% cost and emissions reduction after integrating the data centre and RE system [58]. Feasibility analysis of the design and implementation of energy efficient superconducting cables from an RE source to a 100 MW data centre was performed by Chen et al. [59]. The analysis yielded higher technical and financial benefits compared to conventional cables. Gugul et al. determined the study feasibility of the net zero data center in a bank of Kocaeli, Turkey, based on one year of monitored data [60]. The study integrated PV-based electricity generation, a free cooling system, and reusing waste heat for space heating to reduce energy consumption in the bank. The study obtained a higher reduction of 83% in cooling energy consumption and improved power usage ratings from 1.8 to 1.1. Zhao and Zhu developed a new concept of placing virtual machines for cloud data centers and introduced an algorithm to enhance renewable Energy (RE) utilization and obtain the trade-off between carbon emissions and energy consumption [61]. The study achieved a higher RE usage rate of 73.11% for the base case study compared to other simulated cases. Han et al. introduced a new concept of shared energy storage for data centers to improve the RE utilization scenario and reduce financial burdens [62]. Moreover, this study developed a novel optimization technique, “Chance-Constrained Goal Programming”, to consider the availability of renewable resources. The study’s outcomes suggest that adopting shared energy storage in the data centers will not only enhance the financial benefits but also reduce power consumption and thus improve renewable energy penetration and utilization. Duarte and Fan (2023) applied a robust optimization approach and job scheduling techniques to study the penetration of RE in several interconnected data centers. The study found that implementing energy management, renewable energy, and optimal job scheduling techniques yield lower economic costs and environmental emissions [63]. Jahangir et al. presented a comprehensive techno-economic and environmental comparison of local RE generation and free cooling technology to minimize the cooling load for a data center in Tehran, Iran [64]. Moreover, to estimate the maximum energy reduction potential, the combination of both of these technologies was also explored. The study determined that local RE generation can save 37% of cooling energy, while free cooling yields 86% of energy and 30% of net present cost (NPC) saving.

1.3. Problem Statement

Electricity is an indispensable part of everyday life [65]. The electric grid consists of a transmission and distribution infrastructure network connected to end users by generators and storage assets. Power outage is generally caused when there needs to be more generation than the demand, even though the majority are brought on by damage to the distribution grid and transmission line components. Most power outages are due to minor incidents, such as downed trees or animals shorting the lines, high wind, lightning, and transformers breaking. These outages affect fewer people and can be restored relatively quickly. However, large or major outages last several days and affect many people and areas. Such major or large outages occur due to natural hazards like cyclones, hurricanes, etc. [66]. According to Ericson et al., due to natural storms like cyclones and hurricanes, there is a 33% possibility that a 48 h outage will occur [66]. Considering its geographical location, Bangladesh is prone to natural disasters. From 1980–2020, 129 storms passed over Bangladesh, causing significant economic and environmental damage that is illustrated in Figure 1 [67].

In addition, Bangladesh’s current grid infrastructure is conventional and needs to be modernized [68,69]. Uninterrupted power supply is essential to guarantee the proper operation of data centres. In the case of a power outage, the data center’s operation might be hampered, which results in system downtime and data and economic loss. Approximately, 43% of outages are due to power failure and about 4 in 10 outages cause between USD 100,000 and USD 1 million. This issue is more severe for cloud-based computing and online services as it can impact multiple organizations. However, the existing literature shows that energy resiliency and sustainability analysis of a data center during a power outage have yet to be conducted. Although several studies deal with the safety resilience of the data center [70], energy resilience and sustainability analysis still need to be included. The current study aims to overcome this gap and evaluate the energy resilience of a robust photovoltaic (PV) and battery microgrid for a data centre in Bangladesh via simulation and indicators. HOMER Pro software is utilized to model the diesel-based system. At the same time, REopt software (https://reopt.nrel.gov/tool, accessed on 25 May 2023) is used to simulate the impact of grid outages on the data centre and has been successfully implemented to study the effects of power outages in many case studies [71,72,73]. PV/battery system is sized with the help of Renewable Energy Integration and Optimization (REopt) software, while the diesel generator was sized with HOMER Pro software in this analysis. This is because HOMER allows detailed generator modelling. Fuel properties and more economic parameters can be considered in HOMER. Also, it is not possible to assess the shading issue during the modelling of PV in HOMER [74]. Due to this, PV/battery was sized in REopt, not in HOMER. Indicators such as REP and RCI were also used to observe the system’s performance. Bangladesh is chosen as a case study due to the country’s recent IT infrastructure boom. Bangladesh has the world’s seventh largest data center [75]. Currently, 19 data centers are located throughout the country and the capital Dhaka hosts 16 infrastructures. This number is expected to increase more due to low operational costs, favorable Government and business policies, favorable demographic dividends, and increasing income [76]. About 25 hi-tech parks are currently under construction nationwide, hosting many data centers. Bangladesh’s power infrastructure is conventional and prone to frequent power outages [77]. Therefore, it is necessary to prioritize building a backup energy system for data centers and analyze the impact of the power outage on the microgrids. The current analysis can serve as a benchmark model for other developing countries in studying the impact of power outages on microgrids.

1.4. Key Contributions

Main contributions of this study are given below.

• Perform energy resilience of a microgrid designed for a data center. By energy resilience, it was meant that the developed microgrid could supply electricity during the assigned outage.
• Carry out sustainability analysis of a microgrid designed for a data center. Through sustainability analysis, we tried to show the proposed microgrid’s energy independence, environmental impact, sustainable development goals (SDG), and resource conservation benefits.
• Analyze a new sustainability indicator and resilience indicator.

2. Methods

In this study, both indicator- and simulation-based analysis has been carried out. First, necessary input data are collected from the literature review. Then, utilizing the input data, the microgrid and diesel-based systems have been modelled. After that, the necessary indicators have been developed to analyze the designed systems. A flowchart of the study is shown in Figure 2.

2.1. Optimization Simulation Using REopt

Various researchers have applied numerous techniques to size systems based on different constraints [78,79]. Among these techniques, mixed linear integer programming has been widely used for its wide versatility and flexibility. Therefore, this study adopts the REopt-based MILP optimization technique to determine the optimal configuration for sizing a distributed energy system. The process consists of an “optimization module (OM) and a simulation module (SM)” [80]. The primary focus of OM is to minimize the overall energy costs throughout the life cycle of a location, considering various expenses such as taxes and incentives associated with different project options. Additionally, OM guarantees that the microgrid design allows for supplying power to critical or essential loads independently during a blackout, as defined or predetermined. Financial and resilience analyses are integral parts of OM, encompassing dispatch strategy and size optimization to reduce life cycle energy costs effectively [80]. In contrast, resilience analysis serves a similar purpose. However, it offers the additional advantage of enabling a designed model to provide power to essential loads during outages without relying on the utility grid. The SM, which refers to the resilience analysis, assesses the annual resilience of the system by using the recommended sizes and technologies from the OM as inputs. The simulations are performed to test for any service disruptions every hour of the year, totalling 8760 tests. The SM distinguishes itself from the OM by employing modelling techniques. Unlike the OM, which determines the system size, dispatch strategy, and outage period as fixed values, the SM does not consider a single outage but instead simulates outages for every hour of the year. Each outage simulation employs a load-following strategy to determine the hourly dispatch. The evaluation of the optimization simulation has been performed using Equations (1)–(8) [80]. The nomenclature list for each parameter can be found before the reference.

$$minLCC=\mathrm{m}\mathrm{i}\mathrm{n} (C_{Eg}+C_{OM}+C_{Dn}+C_{PVBAT})$$

(1)

$$C_{Eg}=\sum _{l\in \mathcal{L},hϵ\mathcal{H}}(F_{tlh}^{pd}\times P_{tlh}\times c_h^e)$$

(2)

$$C_{PVBAT}=\sum _{t\in \mathcal{T}}{(X}_t\times c_t)+{(B}^{kWh}\times c_{kWh}^b)+{(B}^{kW}\times c_{kW}^b)$$

(3)

$$C_{Dn}=\sum _{r\in \mathfrak{R}}{(d}_r\times c_r^d)+\sum _{m\in \mathcal{M}}(d_m\times c_m^d)$$

(4)

$$C_{OM}=\sum _{t\in \mathcal{T}}{(X}_t\times c_t^{OM})$$

(5)

These three load constraint expressions minimize the life cycle cost (LCC).

$$\sum _{t\in T}{(F}_{t{l}^{Ex}h}^{pd}\times P_{t{l}^{Ex}h}\times F_t^{dgr})\le L_{lh}, \forall h ϵ\mathcal{H}$$

(6)

$$\sum _{t\in T}{F}_{t{l}^{R}h}^{pd}\times P_{t{l}^{Ex}h}\times F_t^{dgr}+B_h^{-}\le L_{lh}, \forall h ϵ\mathcal{H}$$

(7)

$$\sum _{h\in \mathcal{H}}{F}_{PVlh}^{pd}\times P_{PVlh}\times F_{PV}^{dgr}\le \sum _{h\in \mathcal{g}}{L}_{l^{S}h},\forall l ϵ\mathcal{L}$$

(8)

The equations mentioned, such as Equation (6), confirm that the cumulative energy generated within a specific location will always be within the combined load of all energy sources. Consequently, combining photovoltaic (PV), grid, and battery power must always fulfill the site’s energy requirements. Conversely, as described in Equation (7), Equation (8) establishes an upper limit on the total solar electricity produced, ensuring it does not exceed the data center’s power demand. Additionally, Equation (9) guarantees that the solar energy generated aligns with the chosen system size for each stage, irrespective of the load being served.

$$\sum _{l\in \mathcal{L}}{P}_{tlh}\le X_{th}, \forall h ϵ\mathcal{H}$$

(9)

Battery storage technology constraints are expressed by Equations (10)–(13). This covers the battery’s charge level throughout a specific time frame and charging and discharging. When the battery is charging, the value of $Z_h^{B+}$ will be 1. The battery discharge state is denoted by the value of $Z_h^{B-},$ which is equal to 1.

$$B_h^{+}=\sum _{t\in T}(F_{t{l}^{B}h}^{pd}\ast X_t\ast F_t^{dgr}\ast {\mathrm{ɳ}}_B),\forall h ϵ\mathcal{H}$$

(10)

$$B_h^{SOC}=B_{h-1}^{SOC}+B_h^{+}-B_h^{-},\forall h ϵ\mathcal{H}$$

(11)

$$B_h^{-}\le B_{h-1}^{SOC}, \forall h ϵ\mathcal{H}$$

(12)

$$Z_h^{B+}+Z_h^{B-}\le 1,\forall h ϵ\mathcal{H}$$

(13)

Demand rate constraints are represented by Equations (14) and (15). The monthly grid electricity should not be less than or equal to the demand.

$$\sum _{h\in {\mathcal{H}}_r, l\in \mathcal{L}}{P}_{Glh}\le d_{r,} \forall r ϵ\mathcal{R}$$

(14)

$$\sum _{h\in {\mathcal{H}}_m, l\in \mathcal{L}}{P}_{Glh}\le d_{m,}\forall m ϵ\mathcal{M}$$

(15)

According to Equation (16), PV technology demonstrates similarity to or a smaller size than the net metering capacity when functioning at that specific capacity. However, it is absent or non-applicable when operating below that capacity.

$$\sum _{t\in T}{X}_t\le L_{sv}^{NEM}\ast Y_{sv}, \forall v ϵ V, s\in S$$

(16)

It should be considered that only applicable constraints are included in this analysis. This analysis does not include other technical constraints such as generators, wind, and CHP.

Three financial parameters, such as the “Payback period (PB), Net Present Value (NPV), and Internal rate of return (IRR)”, are considered here to model the system in REopt. PB shows the duration required to retrieve the assets for one project. NPV represents the present value of all future cash flows when adjusted for a specific discount rate. On the other hand, IRR signifies the rate of discount at which the NPV becomes zero. The NPV, PB, and IRR were calculated by Equations (17), (18), and (19), respectively.

$$NPV={\sum }_{z=0}^N\frac{c_{1,z}}{{(1+r)}^z}$$

(17)

$$0={\sum }_{z=0}^N\frac{C_z}{{(1+IRR)}^z}$$

(18)

$$PB=\frac{I}{P}$$

(19)

2.2. Grid Simulation

Equation (20) was utilized to quantify the electricity supplied by the utility grid, considering that the grid functions as a dependable power source without additional maintenance, capital investment, or operation costs. The utility company’s tariff only covers the costs associated with the electricity provided through the power network [73].

$$P_g\left(t\right)=P_l\left(t\right)-\sum (P_{PV,}{P}_{batt})$$

(20)

To accurately model the electricity system, it is necessary to consider the charging rate, which is assumed to be a flat rate, for this study. The demand rate and grid power price are 0.577 USD/kW/month and 0.094 USD/kWh. The sell back rate is considered 0.060 USD/kWh [73,81].

2.3. Simulation for Photovoltaic Module

The power generated from solar resources can be harnessed and converted into electricity via a PV module [81,82]. Equation (21) was employed to determine the available power derived from the solar energy [73].

$$P_{PV}={C_{P_{PV}}D}_{PV}(\frac{I_r}{I_{r_{STC}}})[1+{\mathsf{\alpha }}_{\mathrm{P}}(T_C-T_{C,STC})]$$

(21)

Equation (22) can be used to evaluate the efficiency of a photovoltaic module, considering the topmost power generated under standard test conditions [73].

$${\mathsf{\eta }}_{STC}=\frac{C_{P_{PV}}}{A_{PV}{I}_{r_{STC}}}$$

(22)

According to this study, the projected capital cost for the solar panel system is USD 310 per kilowatt (kW) and it is anticipated to operate for 25 years. The estimated operational and maintenance expenses total USD 17 per kW per year [73]. It is essential to mention that the system’s installation will take place on the surface instead of the rooftop and will not incorporate any tracking device. Additionally, any surplus energy produced by the PV system that goes beyond the required load demand and battery charging will not be utilized in this investigation and will be deemed a loss. The assessment of the electricity output from the PV system that is set up is carried out by utilizing the “PVWatts” program that was created by the “National Renewable Energy Laboratory (NREL)” simultaneously with the REopt software. It is believed that every 2.42811 hectares of space that is accessible will be utilized to install one megawatt of photovoltaic capability, quantified in direct current (MW-DC). The installed system will have a “DC-to-AC size ratio” and a system loss of 1.2 and 14%, respectively [73].

2.4. Simulation for Battery

Due to weather variability, the output of PV is unstable. The quality of the distribution system may be impacted when this power is added to the electrical network. So, to accommodate the irregular availability of renewable energy sources, a battery system must be used [83,84,85,86]. If the amount of electricity generated by photovoltaic (PV) systems is higher than required, the excess power is stored in the battery for later use [87]. And, when the power generation is less than the consumption, the battery is recharged. This stored energy can be further used in the absence of solar power. Equation (23) can be employed to determine the critical battery size [73].

$$C_{{ap}_{bat}}=\frac{E_{load}DA}{{\mathsf{\eta }}_{con}{n}_{batt}DOD}$$

(23)

Several types of batteries are available in the market [88,89,90]. However, lithium-ion batteries have several advantages, like a longer life span and high energy density [91,92]. Therefore, this study considers a lithium-ion battery to have a lifetime of 10 years. The battery system is projected to have a capital cost of USD 419 per kWh, while the power capacity cost will be USD 775 per kW [73]. Assumptions have been made regarding charging the battery by the grid, with the battery starting at a SoC of 50% and a minimum value of 20% SoC. Furthermore, it is anticipated that the inverter will need to be replaced once during the system’s lifetime and the replacement cost will be incorporated into the yearly operational and maintenance costs. The inverter’s efficiency is estimated to be 96%.

2.5. Diesel-Based System

Diesel-based generators are commonly used in Bangladesh to provide electricity during an outage [93,94]. Therefore, a diesel-based generator is considered here to supply electricity to the data centre during an outage. Different sizes were considered to determine the optimum system. The generator’s capital, replacement, operational, and maintenance costs are considered as USD 370/kW, USD 296/kW, and USD 0.05/h, respectively [88]. The fuel cost of diesel is also USD 0.91 per liter. The following equation can be used to determine the fuel consumption rate of the diesel generator [95].

$$L=L_{0,dg}{Y}_{dg}+L_{1,dg}{P}_{dg}$$

(24)

The designed microgrid can be found in Figure 3.

2.6. Indicators

One new resilience indicator (RCI) and one sustainability indicator (REP) are developed in this study to determine the resilience and sustainability of the microgrid.

2.6.1. Resilience Cost Index (RCI)

RCI is the extra expense incurred for incorporating the battery storage system to improve the resilience of the microgrid. RCI can be found from the following Formula (25).

$$RCI=\frac{Cost of the resilience scenario-Cost of the PV scenario}{Cost of the PV scenario}\times 100$$

(25)

A higher value of RCI means that the resilient system is more expensive because integrating the battery for resilience purposes will require additional costs.

2.6.2. Renewable Energy Penetration (REP)

REP can be defined as the proportion of sustainable energy employed in the microgrid. Equation (26) can be utilized to determine the REP value of a system.

$$REP=\frac{Energy generation from renewable sources}{Total energy consumption}\times 100$$

(26)

A higher REP value in one scenario means the system is more sustainable than in other scenarios.

3. Result and Discussion

3.1. PV/Battery Microgrid

Since no energy consumption data were available for Bangladesh’s data center, the analysis considered a hypothetical model based on Bangladesh’s climatic condition for energy consumption data, as illustrated in Figure 4 [96]. The annual average load of the data center is around 30,822 kW. It requires servers, storage devices, networking equipment, and hardware to process and store data. For IT loads, these things are considered in this study. Moreover, to keep the proper temperature for networking equipment and servers, data centers also require extensive cooling. This study considers cooling towers, air conditioning units, and chillers as cooling loads. For lighting loads, LED lights are considered in the data center. This study also considers uninterruptible power supplies (UPS), transformers, and switchgear as power system loads since these equipment ensure power quality and supply power during interruptions.

After that, a microgrid consisting of PV/battery was modelled using REopt. Three scenarios, including base, financial, and resilience, are considered here. The critical load is considered 60% so that the system can endure any unexpected variations in energy demand. The base case system only consists of the grid. The financial represents the PV system supplying electricity to the load without storage, while the resilience scenario defines a PV system with storage. Based on the simulation, the suitable size of PV and battery was found to be 249,219 kW and 398,547 kWh, respectively. Implementing the system will save USD 18,079,948 over the project duration. Similar findings are also reported by [97,98]. These studies designed hospital microgrids and observed life cycle savings of USD 440,191 and USD 65,628,837, respectively. Implementing the proposed system in our study will also result in savings of USD 21,822,076 regarding energy consumption (compared to the base case). The IRR and PB of the system were found to be 6% and 14 years. The capital cost to implement the system was found to be 280 million. It is also estimated that the economic benefit of avoiding this outage is USD 1276 million (considering the avoided cost of USD 100 per kWh). It means preventing this 48 h outage, which will result in economic savings of USD 1276 million. Please note that PV will be installed on the ground, not the roof. PV in the resilience scenario generates 67% more energy than in the financial scenario. This is because in the financial scenario, the grid is responsible for supplying 137,245,147 kWh of energy, which is 286% more than the resilience scenario. The proposed system is also environmentally friendlier than the base case system. The resilient system emits 652% less CO₂ than the grid. In the resilience scenario, the REP value is 87%, while it is only 49% in the financial scenario. So, the resilient system is more sustainable than the financial system. It also means that the system is more dependent on renewable energy sources. Less import and consumption of fossil fuels to generate electricity happens; as a result, the system does not become affected by the supply disruption of fuel in the global market. It also ensures the promotion of clean energy for data centers, which is aligned with SDG 7. The problem with biomass-based systems is the availability of fuels and their higher cost than wind and solar [99]. Also, the cost of biomass will be a significant factor to be considered. The data center authority must also consider the land issue to cultivate the biomass. In addition, wind-based hybrid systems have higher initial costs than solar-based ones. These systems also need more maintenance than solar ones [100]. The environmental impact, such as noise and bird collisions, also worsens the situation.

The RCI value is found to be 35%. It means that the system in the resilience scenario is 35% more expensive than the financial scenario. But, it should be remembered that a battery is added to withstand the outage in the resilience scenario, resulting in an expensive system. However, the system cannot endure the specified outage without the battery in the financial scenario. So, it means that the system’s ability to withstand the outage is also increased by 35%. Therefore, the system’s ability to endure outages and improve its resilience justifies the higher RCI value. A system’s capacity to adjust or react to the changing situations is constrained when it is designed to consistently meet a specific net side load requirement without any flexibility or autonomy. The system cannot sustain itself or operate independently when unforeseen events or disturbances occur. The system becomes more fragile and less able to tolerate problems due to this lack of self-operation, which lowers the system’s overall resilience. Lightner et al. used community, energy, and critical infrastructure resilience (CIR) to assess the benefits of microgrids [101]. To achieve overall system resilience, CIR maximizes resilience and resource allocation across critical infrastructure sectors, including energy. At the same time, RCI evaluates the economic effectiveness of boosting microgrid resilience. Both ideas aim to balance cost-effective cost management with robustness. Moreover, by facilitating resource diversity, reliability, lessened susceptibility, local energy production, sustainability, and energy independence, REP and RCI help to ensure the CIR, community, and energy resilience. Communities may better allocate funds to strengthen vital infrastructure, enhance disaster preparedness, and improve overall resilience by making cost-effective investments in resilience.

After utilizing a resilient strategy and control, it is obtained that our proposed system offers two significant advantages. The proposed system reduces peak demand and cumulative hours (Figure 5). The cumulative hours (CH) are reduced up to 40,000 h, which as a result, increases the resilience of the system. Furthermore, peak demand costs are reduced, which has a significant economic benefit. A study by Rosales-Asensio et al. validated this finding [97]. This study found that implementing a PV/battery/CHP/TES (thermal energy storage) water tanks/absorption chillers microgrid can reduce CH up to 1000 h.

Before the outage, the grid used to supply electricity to fulfill the demand. At 9 AM on 9 August, the grid stopped functioning for 48 h due to a problem. Due to this, a 48 h outage is considered and assigned in the simulation of this study. During this time, the PV fulfills the demand until 5 PM. After this, the battery meets the load due to the lack of solar resources. From 5 PM to 6 AM till the next day, the battery becomes discharged to supply the electricity. At 6 AM on August 10, PV started to produce electricity. PV, along with the battery, meets the demand till 8 AM. After that, PV alone meets the load till 5 PM. The optimized system can provide sufficient energy to critical loads without the grid connection. Additionally, it ensures that the battery’s SoC is restored to the required levels after the outage, guaranteeing system stability for several hours. Figure 6 illustrates the microgrid’s performance during the outage, while Figure 7 outlines the likelihood of the system surviving the blackout. Based on Figure 7, the probability of an optimal system sustaining a critical load during a power outage decreases as the outage duration increases. An abrupt failure example is when the possibility of failure is 25% while the likelihood of sustaining a critical load is 75%. Abrupt failures are sudden and unexpected and can occur even when the probability of success is relatively high.

To verify the system’s capacity to accommodate the essential load, the program performs 8760 power disruption simulations (for every single hour in the year). These simulations determine the typical length of time the system can maintain the essential load. Following the conclusion of the simulations, the software computes the mean, minimum, and maximum durability based on the respective typical, minimum, and maximum durations withstood during the simulated power outages. The economically ideal dispatch plan determines the battery’s SoC at the starting of each outage. This indicates that the battery may be in a low SoC if it is being utilized for peak shaving before the blackout. Remember that the microgrid will function in the islanded mode for this resiliency. From this figure, the system in the financial scenario has the limited chance of surviving the predefined outage. The system’s maximum duration survivability is 12 h. But, the system in the resilience scenario can survive the predefined outage of 48 h. According to the assessments, the lowest resilience is 3 h, while the highest is 2527 h. The finding of our study is also like [102,103], in which the PV/battery microgrid successfully withstands an outage of 41 and 72 h, respectively.

3.2. Diesel System

After simulation, the optimum generator size was found to be 120,000 kW. The electricity production of the generator is 357,920,811 kWh per year (Figure 8). To produce this electricity, the fuel consumption by the generator is 99,880,422 L. Excess electricity after fulfilling the load is 87,920,426 kWh per year. LCOE of the system was found to be USD 0.6139. The initial capital cost of the system was found to be USD 44.4 M (Figure 9). The cost of the diesel-based system can be found in Figure 9.

The environmental performance of the diesel generator can be found in

Table 1. Emission from diesel generator.

Quantity	Value (kg/Year)
CO2	261,448,480
CO	1,648,027
SO2	640,226
UHC	71,914
PM	9988
NOx	1,548,147

. From Table 1, carbon dioxide (CO₂) comprises 98.52% of the total emission, followed by carbon monoxide (CO, 0.62%), nitrogen oxide (NOx, 0.58%), sulfur dioxide (SO₂, 0.24%), unburned hydrocarbon (UH), and particulate matter, (PM). So, from this economic and environmental perspective, the designed PV/battery system performs better than diesel generators.

3.3. Sensitivity Analysis

An observation of the effects of outages on the system size and NPV was carried out using a sensitivity analysis. Figure 10 represents the sensitivity analysis. The analysis clearly shows that the PV and battery sizes increase as the outage duration increases. As the outage duration increases, large PV systems are required to generate more electricity and recharge the batteries. As a result, larger batteries are necessary to manage and store the increased energy output, guaranteeing a steady stream of power during the system downtime. Another thing is that, as the outage duration increases, NPV becomes reduced. Several factors, such as the cost of larger battery and PV systems during the extended outage periods, are responsible for this decline.

3.4. Challenges and Solutions

Implementing PV/battery microgrids for data centers can face several difficulties in Bangladesh. First, consistent generation will be impacted by the variable sunlight. Batteries could be deployed to solve or reduce this problem. But, it will increase the upfront cost and payback period of the system. Also, integrating the microgrid with the national grid may require regulatory change. Moreover, there also needs to be more financial opportunities. In addition, technical experts for designing, installing, and maintaining the system must be sufficiently available in the country. There is no disposal plan for these technologies available in Bangladesh [69]. Moreover, data centers in Bangladesh are also facing several issues like storage, cooling, real-time monitoring, security, uncertainty of solar resources and load management, etc. [104]. To remove the skill obstacles, the government should introduce training programs. The government should formulate policies promoting renewable energy sources, providing financial incentives, and simplifying administrative procedures. The government should also develop proper disposal policies for these technologies. Affordable financing sources and microfinance programs should be introduced to make the projects financially viable. Hybrid systems to tackle the intermittency problem could be designed. Research into low-cost energy storage options is essential to address battery costs [105]. Novel techniques could be applied to solve the security issues [106]. For real-time monitoring, energy prediction, and load management, several new algorithms and methods can be utilized by data centers [107,108,109]. For example, scaling-basis chirplet transform (SBCT) can be used to determine peak energy demands, which will help data centers implement load-shifting techniques and reduce high electricity costs during peak demand hours. Similarly, an adaptive control system can be put into place to lessen the impact of time delays and unpredictable dynamics in data center teleoperation systems.

To achieve Bangladesh’s ambitious renewable energy adoption goals, policy tools, and funding structures capable of resolving existing impediments must be considered and designed. For instance, the current net metering rules are fundamental in encouraging the installation of rooftop solar power systems in both industrial and commercial structures. Given the competitiveness of solar and wind technologies and their continued cost-reduction tendencies, a feed-in tariff may not be necessary. Instead, Bangladesh may consider encouraging competition in grid-tied projects. Creating a time-bound action plan is also essential for encouraging the growth of renewable energy sources. Another strategic requirement for the growth of renewable energy is a thorough, time-bound action plan. An effective monitoring system would be established, together with annual targets for deploying various types of renewable energy, a commitment from the government to provide the required resources, roles for various government agencies, and definitions of their responsibilities. This strategic roadmap expedites green energy projects and conveys to financiers, technology providers, and financial institutions the government’s ambition for renewable energy. Developing an integrated energy and power master plan constitutes a substantial departure from separate energy and power sector strategies. This integrated approach is crucial because it comprehensively evaluates the advantages of developing renewable energy versus the nation’s reliance on imported fossil fuels. The need to lessen reliance on such fuels is urgently underscored by their vulnerability to price volatility on the global market and their high import cost. The timely implementation of projects on the ground is critical to meeting the 40% renewable energy objective [110]. This calls for thorough regulations that cover all facets of the implementation of renewable energy projects, such as required authorizations, schedules, power evacuation, and grid integration. Financial incentives can play a major role in adopting the renewable-based technology in Bangladesh. The Bangladesh Bank’s green banking policy requires financial institutions to devote at least 5% of their granted loans to green initiatives, but ensuring compliance remains a concern. Strict oversight of financial institutions’ green finance operations can help them meet this goal and encourage investment in renewable energy projects. Apart from these phasing out the thermal power plants, waiving import duties on renewable energy equipment and ensuring resilient grid infrastructure for integrating renewable-based technologies can be viable options [111].

4. Conclusions

This study designed an energy-resilient and sustainable PV and battery microgrid for a data center located in Bangladesh. The energy consumption of the data center was 2,700,000 kWh. The configuration included 262,767 kW of PV and 427,289 kWh of battery designed for a 48 h outage. Using PV and battery microgrids for data centers is economically viable, as evidenced by the system’s NPV of USD 12,746,169. The designed system can withstand simulated power outages and is environmentally benign due to its lower emissions. Based on the outcome of the analysis, policymakers can take the necessary initiatives to improve the current energy infrastructure. For instance, climate resilience can be integrated into national plans and fields.

To better understand the advantages of minimizing life cycle expenses, further investigation may be conducted, such as exploring incentives and a demand-focused pricing model that adjusts to tariffs. Also, hybrid systems combining wind, biomass, and solar systems with batteries can be designed for future studies to observe their impact on data centers.