Resilient Control Strategies for Urban Energy Transitions: A Robust HRES Sizing Typology for Nearly Zero Energy Ports

Abstract

Ports located within dense urban environments face a major challenge in achieving deep decarbonization without compromising the reliability and safety of critical maritime operations. This study develops and validates a resilience-oriented control and sizing typology for Hybrid Renewable Energy Systems (HRESs), supporting the transition of a medium-sized Mediterranean port toward a Nearly Zero Energy Port (nZEP). The framework integrates five years of measured electrical demand at 15 min resolution to capture stochastic load variability, seasonal effects, and safety-critical peak events. Thirty-five HRES configurations are simulated using HOMER Pro, assessing photovoltaic and wind generation combined with alternative Energy Storage System (ESS) technologies under two grid-interface control strategies: Net Metering (NM) and non-NM curtailment-based operation. Conventional Lead–Acid batteries are compared with inherently safer Vanadium Redox Flow Batteries (VRFBs), while autonomy constraints of 24 h and 48 h are imposed to represent operational resilience. System performance is evaluated through a multi-criteria framework encompassing economic viability (Levelized Cost of Energy), environmental impact (Lifecycle Assessment-based carbon footprint), and operational reliability. Results indicate that NM-enabled HRES architectures significantly outperform non-NM configurations by exploiting the external grid as an active balancing layer. The optimal NM configuration achieves a Levelized Cost of Energy of 0.063 €/kWh under a 24 h autonomy constraint, while reducing operational carbon intensity to approximately 70 gCO₂,eq/kWh, corresponding to a reduction exceeding 90% relative to baseline grid-dependent operation. In contrast, non-NM systems require substantial storage and generation oversizing to maintain resilience, resulting in higher curtailment losses and Levelized Cost of Energy values of 0.12–0.15 €/kWh. Across both control regimes, VRFB-based systems consistently exhibit superior robustness and safety performance compared to Lead–Acid alternatives. The proposed typology provides a transferable framework for resilient and low-carbon port microgrid design under real-world operational constraints.

1. Introduction

The decarbonization of critical maritime infrastructure is no longer a discretionary initiative but an operational imperative driven by tightening regulatory and policy regimes from the European Commission (EC) and the International Maritime Organization (IMO) [1,2]. While the primary driver remains the mitigation of greenhouse gas emissions, the maritime sector is also a major source of air pollutants; peer-reviewed assessments widely report shipping’s contribution on the order of ~15% of global nitrogen oxide and ~13% of sulphur oxide emissions [3,4]. At the same time, the rapid integration of decentralized Renewable Energy Systems (RESs) and port microgrids (with non-dispatchable generation and high-power, time-varying loads such as shore power) introduces material stochasticity and operational uncertainty into port energy grids, which must be managed through robust sizing and energy management strategies [5,6,7]. Ports, as high-density industrial nodes, therefore have to balance the “energy trilemma” (security/reliability, environmental performance, and affordability) under explicit regulatory pressure, while ensuring operational safety and resilience to disturbances and grid failures [8,9,10].

The transition from fossil-fuel reliance to Hybrid Renewable Energy Systems (HRESs) is widely recognized as the most viable pathway for achieving the Nearly Zero Energy Port (nZEP) concept, enabling ports to decouple growth in activity from emissions and energy imports [11,12]. Nevertheless, the inherent intermittency and limited predictability of photovoltaic and wind turbine generation introduce non-trivial challenges for process safety, power quality, and supply reliability in port microgrids [13,14,15]. Empirical and modeling studies consistently show that, in the absence of coordinated control, high renewable penetration can induce voltage deviations, frequency instability, and power imbalance events that directly threaten safety-critical port services such as quay crane operation, terminal automation, navigation lighting, and security systems [16,17]. As a result, the engineering focus in modern port energy systems is shifting away from simple installed generation capacity towards dynamic system stability and operational robustness. This shift necessitates the systematic integration of Energy Storage Systems and advanced dispatch and control strategies, which act as buffers against renewable variability, enable peak shaving and load following, and ensure voltage and frequency stability under stochastic operating conditions [18,19,20].

The current literature predominantly relies on techno-economic optimization tools such as HOMER Pro to size HRESs, with the objective function most commonly defined as the minimization of Net Present Cost (NPC) or LCOE [21,22]. Seminal and highly cited studies by Ahadi et al. and Caballero et al. have convincingly demonstrated the economic feasibility of photovoltaic–wind–energy-storage configurations for remote communities, islands, and relatively simple grid-connected microgrids, validating the cost competitiveness of high renewable penetration under favorable boundary conditions [23,24]. However, there remains a pronounced research gap when these approaches are transferred to the context of fully operational commercial ports, which exhibit highly heterogeneous, safety-critical, and temporally coupled energy demands [9,16,25]. Most maritime-focused studies concentrate narrowly on Cold Ironing deployment, treating shore-to-ship power as an isolated subsystem rather than as an integrated component of a port-wide microgrid [6]. Moreover, existing optimization frameworks typically adopt single-objective or weakly constrained formulations, prioritizing cost reduction while systematically neglecting operational resilience, process safety under islanded operation, and lifecycle environmental impacts, factors that are central to the viability of energy-autonomous ports [18,26,27,28].

Energy Storage Systems (ESSs) can effectively normalize the fluctuating, unpredictable, and inherently unreliable power output of renewable energy sources, acting as temporal buffers that decouple generation variability from demand requirements and enhance controllability at the system level [29,30]. In parallel, ESSs have attracted increasing research and deployment interest over the last decade due to rapid cost reductions, improved round-trip efficiencies, and maturing supply chains, which have significantly enhanced their economic feasibility within HRESs [31,32]. Consequently, HRESs incorporating Energy Storage Systems have been extensively investigated in the literature and are consistently shown to outperform non-storage configurations in terms of reliability, renewable penetration, and operational flexibility, particularly under high variability and partial islanding conditions [17,19,33].

Nevertheless, despite these operational advantages, the integration of ESSs introduces notable trade-offs. Multiple studies report that while storage substantially improves system stability and reliability metrics, it also increases total investment cost and lifecycle environmental burdens due to material intensity, manufacturing impacts, and end-of-life considerations, which can adversely affect composite sustainability indicators [34,35]. Importantly, only a limited subset of the literature moves beyond generic battery modeling to systematically compare different Energy Storage System technologies, such as lithium-ion batteries (Li-Ion), Lead–Acid batteries (LA), flow batteries (VRFBs), hydrogen-based storage, and hybrid storage architectures, within the same HRES framework [36,37]. As a result, the comparative impact of storage technology selection on overall system efficiency, environmental performance, and long-term sustainability remains insufficiently explored, particularly for complex, high-demand applications such as industrial microgrids and ports.

There is an extensive and mature body of research addressing the modeling, simulation, and optimization of HRESs across a wide spectrum of applications worldwide, including isolated communities, islands, campuses, and grid-connected microgrids [38,39]. Within this literature, single-objective techno-economic optimization remains the dominant paradigm, with most studies formulating the sizing problem as the minimization of NPC or, equivalently, the LCOE [23,24]. Indicatively, a study investigated the optimal configuration of an HRES for remote communities and demonstrated that a photovoltaic–wind–energy-storage architecture consistently yields the lowest NPC under realistic demand and resource conditions, thereby establishing its economic superiority over fossil-based or non-storage alternatives [17,40,41]. Similarly, there is a study that analyzed the optimal design of a grid-connected photovoltaic–wind HRES and concluded that such configurations significantly reduce long-term system costs while simultaneously enabling the provision of low-carbon electricity and enhanced energy autonomy [42,43]. Collectively, these studies confirm the economic viability of HRESs but also highlight their predominant focus on cost minimization, often at the expense of operational resilience, safety constraints, and multidimensional sustainability assessment.

However, the design of a port-scale HRES is inherently a multi-objective optimization problem, as techno-economic performance, reliability, power quality, and renewable penetration must be addressed simultaneously under stringent operational constraints [44]. In this context, another study investigated a domestic HRES using a multi-objective formulation that minimized Net Present Cost while maximizing system reliability, demonstrating that photovoltaic–wind–energy-storage configurations consistently outperform photovoltaic–energy-storage and wind–energy-storage alternatives in terms of cost–reliability trade-offs [45]. Their findings reinforce the systemic value of technology complementarity when intermittency and load uncertainty are explicitly considered.

For grid-connected applications, several studies have shown that renewable fraction levels exceeding 90% can be achieved through the coordinated integration of photovoltaic, wind, and energy storage subsystems, without compromising supply adequacy or operational stability [16,46]. As a result, Hybrid Renewable Energy Systems deployed in on-grid configurations are increasingly recognized as a viable pathway towards cleaner and more reliable power systems, particularly in high-demand nodes with partial autonomy requirements [47,48,49,50]. Nevertheless, despite these promising results, the existing body of literature remains insufficient to support generalized, transferable conclusions regarding HRES effectiveness and stability. Many published studies rely on highly specific testbeds, local resource conditions, and ad hoc modeling assumptions, resulting in low methodological replicability and limited external validity of the proposed solutions [18,51]. Addressing this gap requires systematic, reproducible frameworks capable of evaluating HRESs under realistic and operationally complex environments.

Within this broader context, the maritime sector constitutes a particularly critical domain. Shipping activities are estimated to contribute approximately 3% of global greenhouse gas emissions, positioning ports as pivotal actors in climate change mitigation strategies [52,53]. Consequently, many major ports worldwide have initiated structured energy-transition pathways toward the nZEP paradigm. Empirical evidence from recent port-scale demonstrators indicates substantial potential for emissions reduction, energy efficiency improvement, and enhanced resilience through integrated renewable generation, electrification, and smart energy management [54]. The nZEP concept explicitly promotes the holistic integration of all available sustainable technologies and operational practices, aiming to deliver zero-emissions port infrastructures while maintaining safety, reliability, and economic viability [55].

Consequently, ports have progressively begun deploying a broad portfolio of renewable energy technologies, including onshore and floating offshore photovoltaic systems, wind turbines, tidal and wave energy installations, and shallow geothermal systems for heating and cooling applications [56,57]. The number of ports worldwide pursuing renewable energy integration is steadily increasing, driven by regulatory pressure, rising energy costs, and corporate decarbonization commitments; however, the majority of implemented HRESs to date are primarily dimensioned to supply electricity to vessels at berth through shore-side power infrastructure rather than address the full spectrum of port energy demands [58,59,60]. A critical shortcoming of the existing literature is that most studies do not examine the optimal sizing, integrated modeling, and coordinated operation of HRESs at the scale of an entire port, encompassing terminal operations, auxiliary services, logistics facilities, and thermal loads. Instead, research efforts are largely fragmented and confined to vessel energy supply or isolated subsystems, leaving significant knowledge gaps regarding system-wide performance, interactions, and scalability [61,62,63]. This fragmentation hinders informed decision-making by port authorities, technology providers, and prospective investors, who face elevated uncertainty due to the lack of consolidated, scientifically grounded evidence. Moreover, the prevailing analytical approaches remain predominantly single objective, with optimization frameworks almost exclusively focused on minimizing NPC or electricity tariffs. Such cost-centric formulations inadequately capture the broader sustainability implications of port-scale energy transitions, systematically overlooking environmental lifecycle impacts and social dimensions such as operational safety, resilience, and local air-quality benefits. Consequently, there is a clear and pressing need for comprehensive studies that explicitly evaluate HRESs in ports across all three pillars of sustainability—economic, environmental, and social—using multi-objective, reproducible, and decision-oriented frameworks capable of supporting real-world implementation.

This research distinguishes itself from existing techno-economic studies by establishing a reliability-centered sizing framework that prioritizes operational resilience over simple cost minimization. The study introduces three critical novelties to the domain of port microgrid design. First, unlike low-fidelity models that rely on synthetic hourly averages, this methodology integrates actual 5-year, 15 min interval energy demand data. This granular load profiling is a requisite for identifying stochastic peak demands and ensuring the system is sized to withstand extreme volatility without service interruption. Second, the study explicitly evaluates inherently safer storage technologies, contrasting the thermal stability and cycle life of Vanadium Redox Flow Batteries (VRFBs) against traditional Lead–Acid chemistries, addressing the specific fire safety constraints of dense urban–maritime environments. Third, Net Metering (NM) is modeled not merely as a billing mechanism, but as a dynamic grid-stabilization control strategy that allows the port to act as a buffer for the external grid during peak loads. By concurrently optimizing economic prosperity (LCOE), environmental quality (LCA-based carbon footprint), and operational resilience (Autonomy), this work provides a validated, replicable framework for the safe integration of high-penetration renewables into critical infrastructure.

The remainder of this paper is structured as follows: A review of the state-of-the-art regarding HRES sizing and safety in maritime environments is presented. This is followed by a detailed description of the case study, data acquisition protocols, and the formulation of the control scenarios. Finally, the results are analyzed with a focus on system reliability and safety, concluding with specific recommendations for future resilient port infrastructure.

2. Materials and Methods

The methodology was split into four major phases, aiming to provide the most solid and reliable outcomes (Figure 1). Each phase is thoroughly explained later in this chapter.

2.1. Initial State: Case Study Selection

The case study is a medium-sized port in Crete, a typical Mediterranean port, which is essential due to its strategic position, operating daily routes with other Greek ports. It mainly serves domestic and foreign ships, transporting thousands of tourists yearly, impacting its economy. This port was picked as the ideal case because of its high-RES implementation potential. This port can be a paradigm for numerous other ports across the globe as a pivotal aspect for the decision process; there is high applicability of the suggested framework after modifying the input data because there are thousands of similar seaports worldwide.

2.2. Data Acquisition and Stochastic Load Profiling

The validity of any microgrid sizing optimization relies entirely on the fidelity of the input load profile. Unlike previous studies that utilize synthetic or hourly-averaged data, this framework integrates empirical, high-frequency data acquired via smart metering infrastructure installed directly on the port’s medium-voltage distribution network. This granular approach is critical for capturing the stochastic load fluctuations inherent to complex port operations, where energy consumption is a non-linear function of meteorological conditions and sporadic tourism logic.

To mitigate the uncertainty of short-term anomalies, a longitudinal 5-year dataset (15 min intervals) was aggregated to construct a representative stochastic load profile for the passenger terminal. Furthermore, to account for the unmonitored commercial sector, for which only monthly billing data existed, a conservative load scaling factor of 1.6 (an additional 60% capacity) was integrated into each timestep. This ensures that the optimization yields a robust, “worst-case” system design capable of handling total facility demand without compromising reliability.

Analysis of the load duration curve (Figure 2) reveals a base load fluctuating between 50 and 150 kWh, devoid of extreme transient spikes that typically destabilize weak grids. However, the load composition reveals a critical operational constraint: outdoor lighting accounts for >50% of the total demand. This results in a seasonal inversion where winter demand is driven by prolonged illumination requirements essential for navigation safety and security, whereas summer peaks are correlated with high tourism density. The annual aggregated demand stands at 874.6 MWh (2.4 MWh/day), with a peak load of 172.04 kW occurring during nocturnal operations in August. Currently, the facility operates as a strictly grid-dependent load with zero in situ generation, representing a significant resilience vulnerability in the event of external grid failure.

The operational stability of the proposed HRES is strictly dependent on the accuracy of the meteorological input data. To ensure simulation fidelity, renewable resource potential was retrieved from the NASA POWER database and subjected to ground-truth validation against on-site measurements conducted by the research team. The calculated deviation remained below 10%, confirming the dataset’s reliability for robust system sizing. Wind resource analysis (Figure 3) reveals a mean speed of 6.26 m/s. Crucially for structural safety and grid ramping rates, the histogram indicates a low frequency of extreme wind speed events, suggesting a favorable profile for consistent turbine operation. Regarding solar potential, the site exhibits high availability, with an average daily global horizontal irradiance (GHI) of 5.28 kWh/m²/day. Figure 4 details the monthly correlation between solar radiation and the clearness index, validating the site’s capacity to support high-penetration PV integration.

As a next rational step, all the data were coded, inserted, and analyzed in the Minitab software v20.4 to make them exploitable for this research work.

2.3. System Architecture and Mathematical Modeling of Control Strategies

This phase establishes the mathematical framework for the optimal sizing and configuration of the HRES. In collaboration with port engineering teams, the system design was constrained by a multi-objective optimization function targeting three critical performance metrics: environmental quality (carbon abatement), operational resilience (energy autonomy/safety), and economic viability (LCOE minimization).

2.3.1. Scenario Topology and Design Variables

The simulation matrix comprises 35 distinct scenarios constructed around three fundamental design variables: Renewable Energy Source (RES) mix, Energy Storage System (ESS) chemistry (Lead–Acid vs. Vanadium Redox Flow Batteries), and the grid-interface control strategy (specifically, the presence or absence of Net Metering). The optimization model operates under strict boundary conditions, including the regional legislative limits for Net Metering (NM), maximum annual capacity shortage fractions, and specific electrochemical constraints (e.g., depth of discharge, charge rates) for the ESS technologies. The scenario architecture is bifurcated into two primary control logic groups:

• Grid-Interactive Mode (Net Metering): Scenarios 2–18, where the grid acts as an infinite buffer for surplus generation.
• Non-Interactive Mode (No Net Metering): Scenarios 19–35, where excess energy is curtailed or stored, enforcing stricter reliance on internal balancing.

Within each control group, the system is stressed against three levels of operational resilience:

• No Autonomy: Baseline reliability dependent on external grid stability.
• 24 h Ride-Through: System sized to sustain critical port operations for 24 h during a blackout.
• 48 h Ride-Through: Extended autonomy for prolonged grid failures.

Storage topology is further differentiated to evaluate the safety and performance trade-offs between conventional Lead–Acid (LA) units and inherently safer Vanadium Redox Flow Batteries (VRFBs). Additionally, specific sub-scenarios integrate wind turbines (WTs) to analyze the impact of generation diversity on system reliability. Figure 5 details the technology stack for each scenario.

Operational resilience is enforced through explicit autonomy constraints of 24 h and 48 h, representing the system’s ability to sustain safety-critical port services during complete external grid outages. While resilience is inherently multidimensional, autonomy is deliberately employed here as a conservative, engineering-oriented proxy that captures the system’s disturbance absorption capacity and ensures uninterrupted operation under worst-case conditions.

2.3.2. Stochastic Wind Power Generation Model

To accurately simulate the intermittency of the wind resource, a critical factor for microgrid stability, the WT power output PWT is modeled using a deterministic piecewise function based on the instantaneous wind velocity (V). This ensures that cut-in and cut-out behaviors, which introduce sudden power ramps, are accurately captured in the reliability analysis. The power curve is defined by Equation (1):

$$P_{WT}=\left\{\begin{array}{c} 0 V<V_{cut-in} and V\ge V_{cut-out}\\ {\displaystyle \frac{P_r\left(V-V_{cut-in}\right)}{\left(V_r-V_{cut-in}\right)}} V_{cut-in\mathrm{ }}\le V\le V_r\\ {\mathrm{ }\mathrm{ }P}_r V_r\le V<V_{cut-out} \end{array}\right.$$

(1)

where P_r represents the turbine’s rated capacity, v_r is the rated wind speed, and ucut-in and ucut-out denote the operational safety limits for startup and shutdown, respectively. Since wind velocity is height-dependent due to surface boundary layer friction, the anemometer data ubase measured at reference height H_base is extrapolated to the turbine hub height H_hub using the Power Law (Equation (2)):

$$V=V_{base}{\left({\displaystyle \frac{H}{H_{base}}}\right)}^{\alpha }$$

(2)

The Hellman exponent (α) quantifies the shear caused by surface roughness and thermal stability. For this coastal port environment, a coefficient of α = 0.143 is applied, corresponding to open terrain under neutral stability conditions. Finally, to prevent the overestimation of generation capabilities during high-temperature events, the theoretical power output P_WT,STP is dynamically corrected for air density fluctuations (Equation (3)):

$$P_{WT}=\left({\displaystyle \frac{\rho }{{\rho }_0}}\right)P_{WT,STP}$$

(3)

2.3.3. Photovoltaic Panel Mathematical Models

To account for the impact of ambient temperature on generation efficiency, a critical stability factor in the high-temperature Mediterranean climate, the instantaneous power output (P_PV) of the photovoltaic array is modeled as a function of solar irradiance and cell temperature. The governing relationship (Equation (4)) explicitly integrates a thermal degradation coefficient to prevent the overestimation of power availability during summer peak load events [3].

$$P_{PV}=Y_{PV}{f}_{PV}\left({\displaystyle \frac{G_T}{G_{T,STC}}}\right)[1+a_{\rho }\left(T_c-T_{C,STC}\right)]$$

(4)

where Y_PV is the rated peak capacity of the PV array (kW) under STC, and f_PV represents the solar derating factor. The derating factor is a composite reliability metric that accounts for systemic losses due to panel soiling, wiring resistance, shading, and module aging over the project lifecycle. Regarding environmental inputs, G_T is the solar irradiance incident on the PV array at the current timestep (kW/m²), while G_T,STC denotes the standard incident radiation (1 kW/m²). The thermal behavior is defined by α_P, the temperature coefficient of power (%/°C), which quantifies the efficiency drop as the cell temperature (T_c) exceeds the standard reference temperature (T_c,STC = 25 °C).

2.3.4. Energy Storage Dynamics and Kinetic Modeling

To accurately simulate the dynamic availability of stored energy, a prerequisite for reliable grid-forming operations, both the Lead–Acid (LA) and Vanadium Redox Flow Battery (VRFB) systems are modeled using the Kinetic Battery Model (KiBaM). While the electrochemical mechanisms differ, the KiBaM framework provides a robust phenomenological approximation for both technologies by treating the storage as two coupled “tanks” (available charge vs. bound charge). This allows the control system to predict capacity shortages during rapid discharge events. The instantaneous energy state (Q_ESS) is derived by integrating the power flow over time, as defined in Equation (5) [3,4,5]:

$$Q_{ESS}=Q_{ESS,0}+{\int }_0^t{V}_{ESS}{I}_{ESS}dt$$

(5)

where Q_ESS,0 is the initial state of charge, V_ESS is the nominal DC bus voltage, and I_ESS is the current flux. The operational state of charge (B_SOC), which serves as the primary control variable for the Battery Management System (BMS), is calculated as the ratio of instantaneous energy to maximum capacity (Equation (6)) [3,4,5]:

$$B_{SOC}={\displaystyle \frac{Q_{ESS}}{Q_{ESS,max}}} \times 100$$

(6)

Crucially for process safety, the maximum permissible charging power P_ESS,max is not a static value, but a dynamic constraint determined by the most limiting of three physical boundaries: the kinetic rate of charge transfer P_kbm, the maximum charge rate P_mcr, and the maximum current limit P_mcc. The control algorithm enforces the strictest of these limits (Equation (7)) to prevent thermal stress and over-current conditions:

$$Q_{ESS,max}={\displaystyle \frac{\mathrm{m}\mathrm{i}\mathrm{n}\left(P_{ESS,max,kbm}{P}_{ESS,max,mcr}{P}_{ESS,max,mcc}\right)}{{\eta }_{ESS,c}}}$$

(7)

The specific values for these safety constraints are calculated via Equations (8), (9) and (10), respectively:

$$P_{ESS,max,kbm}={\displaystyle \frac{k{Q}_1{e}^{-k\Delta t}+Qkc\left(1-e^{-k\Delta t}\right)}{1-e^{k\Delta t}+c\left(k\Delta t-1+e^{-k\Delta t}\right)}}$$

(8)

$$P_{ESS,max,mcr}={\displaystyle \frac{1-e^{-{\alpha }_c\Delta t} {(Q}_{ESS,max}-Q_{ESS})}{\Delta t}}$$

(9)

$$P_{ESS,max,mcc}={\displaystyle \frac{N_{ESS}{I}_{max}{V}_{nom}}{1000}}$$

(10)

where η_ESS,c represents the round-trip coulombic efficiency. In the kinetic model (Equation (8)), k denotes the charge rate constant h⁻¹, c is the capacity ratio (available vs. total), and Δt is the simulation timestep. Q₁ represents the available energy at the start of the timestep. For the charge rate limit (Equation (9)), α_c is the maximum charge rate A/Ah. Finally, the maximum current constraint (Equation (10)) is a function of the number of strings N_ESS, the maximum rated current I_MAX, and the nominal voltage V_NOM.

Battery lifetime is not treated as a fixed replacement event but is instead evaluated dynamically through cumulative energy throughput and equivalent full-cycle tracking. The model continuously monitors storage utilization against technology-specific cycle life limits reported in the literature, allowing the identification of stress-intensive operating regimes. This formulation enables the framework to function as a predictive maintenance proxy, supporting proactive planning of maintenance or replacement actions before critical degradation occurs, thereby minimizing unplanned outages and operational losses in safety-critical port environments.

2.3.5. Bidirectional Power Conversion Interface

The coupling between the DC microgrid (PV generation, ESS) and the AC distribution network is mediated by a bidirectional inverter, which serves as the primary actuator for voltage regulation and power dispatch. To account for thermal dissipation and switching losses, factors that influence the thermal management of the control room, the power transfer capability is modeled linearly. The AC output power (P_out) delivered to the port load or grid is governed by Equation (11) [6]:

$$P_{out}=P_{in}{\eta }_{inv}$$

(11)

where P_in represents the aggregate DC input power and η_inv denotes the conversion efficiency. For this simulation, a conservative constant efficiency of η_inv = 0.95 is applied, consistent with industrial standards for high-capacity maritime converters.

2.3.6. Economic Feasibility and Lifecycle Cost Modeling

To ensure that the proposed resilience upgrades (VRFB integration, Net Metering controls) are financially sustainable, the system is evaluated using the Net Present Cost (NPC) and Levelized Cost of Energy (LCOE) metrics. The total initial capital expenditure C_system acts as the baseline for the investment model and is defined as the aggregate procurement cost of all generation and balancing assets, such as the sum of the PV costs (C_pv), the WT costs (C_WP), the electricity grid’s cost (C_grid), the battery’s cost (C_battery), and the converter’s cost (C_conv) (Equation (12)) [3,4,6]:

$$C_{system}=C_{PV}+C_{WT}+C_{grid}+C_{battery}+C_{conv}$$

(12)

The project’s lifecycle viability is determined by the NPC, calculated by discounting the total annualized cost C_ann,total, which includes capital amortization, O&M, fuel, and grid purchases minus salvage values, by the Capital Recovery Factor (CRF), as shown in Equation (13):

$$C_{NPC}={\displaystyle \frac{C_{ann,total}}{CRF(i,n_{pro})}}$$

(13)

To compare the oper. efficiency against standard utility rates, the LCOE is derived by normalizing the NPC against the total electric load served (E_served) over the project lifetime (Equation (14)):

$$LCOE={\displaystyle \frac{NPC}{E_{served}}}$$

(14)

The economic boundary conditions are defined by the real interest rate i and the project lifetime N. The CRF is calculated by Equation (15), while the real interest rate is adjusted for the inflation rate f = 2.0% and the nominal discount rate i′ = 8.0% using Equation (16):

$$CRF\left(i,N\right)={\displaystyle \frac{i{\left(1+i\right)}^N}{{\left(1+i\right)}^N-1}}$$

(15)

$$i={\displaystyle \frac{i^{\prime }-f}{1+f}}$$

(16)

Finally, the investment attractiveness is quantified using three standard financial indicators: the simple payback period (PP) (Equation (17)), the Internal Rate of Return (IRR) (Equation (18)), and the return on investment (ROI) (Equation (19)). The IRR is solved iteratively by finding the discount rate that results in a Net Present Value (NPV) of zero.

$$PP={\displaystyle \frac{C_{init}}{C_{af}}}$$

(17)

$$NPV=0=>NPC==\sum _{t=1}^T{\displaystyle \frac{C_t}{{\left(1+IRR\right)}^t}}-C_0$$

(18)

$$ROI={\displaystyle \frac{{\sum }_{I=0}^{n_{pro}}{C}_{af,ref}-C_{af}}{n_{pro}(C_{cap}-C_{cap,ref})}}$$

(19)

where C_init is the initial investment, C_af is the annual cash inflow, C_af,ref is the nominal annual cash flow for the baseline, n_pro is the project lifetime in years, C_cap is the capital cost of the initial-current system, and C_cap,ref is the baseline’s capital cost.

Replacement costs for major power electronics, including bidirectional inverters, are internalized within the Net Present Cost formulation using conservative service life assumptions. Operational dispatch strategies are simultaneously designed to avoid excessive cycling and thermal stress, ensuring that component replacement reflects realistic long-term operation rather than premature failure driven by suboptimal control.

2.3.7. Lifecycle Assessment and Environmental Impact Modeling

To quantify the environmental efficacy of the proposed control strategies, the system is evaluated using a rigorous Lifecycle Assessment (LCA) framework. The total annual greenhouse gas (GHG) emissions are a function of the local grid’s fossil-heavy generation mix and the embodied carbon of the installed HRES components. The baseline grid emissions are calculated based on the island’s specific generation profile, which is dominated by heavy fuel oil (66.58%) and diesel (11.95%), with only 21.47% contributed by existing renewables. The annual emissions inventory is derived using Equation (20):

$$Emissions\left({\displaystyle \frac{\mathrm{t}\mathrm{n}}{\mathrm{y}}}\right)={\displaystyle \frac{E_{grid}\left(\frac{\mathrm{k}\mathrm{W}\mathrm{h}}{\mathrm{y}}\right) \times Pollutant Index \left(\mathrm{g}\frac{\mathrm{C}{\mathrm{O}}_{2\mathrm{e}\mathrm{q}}}{\mathrm{k}\mathrm{W}\mathrm{h}}\right)}{{10}^6}}$$

(20)

For the proposed HRES technologies, the environmental impact is quantified using the CO_2,eq index, integrating the 100-year Global Warming Potential (GWP) factors defined by the Intergovernmental Panel on Climate Change (IPCC). This accounts for the cumulative impact of Carbon Dioxide (CO₂), Nitrous Oxide (N₂O), Methane (CH₄), and Chlorofluorocarbons (CFCs). The specific carbon footprint (CF_S) for each technology S (e.g., PV, WT, ESS) is calculated by normalizing the total lifecycle emissions over the total energy produced (E_P), as shown in Equation (21):

$$C{F}_S={\displaystyle \frac{{\sum }_{y\in GHG}{GWP}_{y}E{M}_y}{E_P}}$$

(21)

where GWP_y is the warming potential of pollutant y, and EM_y represents the mass of direct and indirect emissions over the technology’s lifespan. The CF is measured in gCO_2,eq/kWh. The specific carbon intensity of the local electricity grid (CF_DS) serves as the benchmark for abatement and is calculated in Equation (22). Given the dominance of liquid fossil fuels, the factor combines the specific emissions of diesel generators (F_D = 0.989 kgCO_2,eq/kWh) with the fossil-fuel fraction (F_FF = 78.5%) of the island’s mix.

The calculation of the electricity grid’s CF is based on Equation (22) [7,8]:

$$C{F}_{DS}=F_D \times F_{FF}=0.989 \times 0.785=0.78 \mathrm{kg} {\displaystyle \frac{\mathrm{C}{\mathrm{O}}_{2,\mathrm{e}\mathrm{q}}}{\mathrm{k}\mathrm{W}\mathrm{h}}}$$

(22)

This results in a high baseline grid intensity of 0.785 kgCO_2,eq/kWh. In contrast, the PV systems utilized in this study are modeled with a mean European emission factor of 37.3 gCO_2,eq/kWh. Crucially for the storage comparison, the LA batteries exhibit a high embodied carbon footprint of 24,250 gCO_2,eq per kWh of installed capacity, whereas the VRFBs are modeled at 38.2 gCO_2,eq per kWh of throughput, reflecting their lower degradation rates and longer cycle life (Figure 6). Finally, the system’s net energy balance, used to determine the true autonomous capability, is defined by Equation (23):

$$Net Energy=E_{grid}-E_{PV}-E_{WT}$$

(23)

The Lifecycle Assessment adopted in this study is based on technology-specific carbon footprint factors derived from peer-reviewed cradle-to-grave LCA studies. These factors inherently account for upstream material extraction, manufacturing, transportation, operational lifetime, and end-of-life treatment phases. Balance-of-system elements, including cabling, mounting structures, foundations, and auxiliary equipment, are not modeled as separate inventory items; instead, their environmental burdens are implicitly embedded within the aggregated literature-based emission factors used for each technology. This approach ensures methodological consistency across all simulated configurations and avoids introducing site-specific assumptions that would reduce the transferability of the proposed HRES sizing typology.

2.3.8. Study’s Assumptions

To ensure that the simulation yields a robust system design capable of withstanding real-world operational stresses, the optimization model is constrained by the following techno-economic parameters:

• Regional capital and operational expenditure data for mature technologies, specifically PV and Lead–Acid storage, were derived from verified quotations provided by domestic Greek suppliers. Costs for Vanadium Redox Flow Batteries were based on the international literature averages due to limited local commercial availability.
• The electricity cost is modeled as a flat aggregate tariff of 0.16 €/kWh, encompassing the energy generation charge and regulated network distribution fees, but not any applicable taxes. Transactional costs for the taxes and the NM administrative interface are excluded, as preliminary sensitivity analysis indicated a negligible impact on the global NPC.
• The capital cost of the bidirectional inverter is aggregated within the PV procurement model to reflect standard turnkey installation practices.
• To mitigate the risk of stochastic power drops, the control logic enforces a strict operating reserve requirement. The system must maintain spinning reserves equal to 25% of the instantaneous solar PV output and 50% of the instantaneous wind turbine output to buffer against rapid meteorological fluctuations.
• The lifecycle durability of the Lead–Acid ESS is modeled using a non-linear throughput curve, defined as 3000 cycles at 50% depth of discharge (DoD) and 2000 cycles at 80% DoD. To simulate realistic grid-interactive operation and prevent premature failure, the model constrains the annual throughput to fewer than 120 full equivalent cycles.

The imposed reserve margin reflects conservative microgrid design practice for safety-critical infrastructures, ensuring adequate buffering against short-term renewable intermittency and forecasting errors.

The adopted derating factors aggregate realistic performance losses due to wiring resistance, inverter inefficiencies, temperature effects, soiling, and long-term component aging. The selected values are consistent with established practice reported in the microgrid and hybrid energy systems literature and are intentionally conservative to avoid overestimation of system performance.

2.4. Results and Discussion: Strategic Business Case for Resilient Infrastructure

The transition to a Nearly Zero Energy Port (nZEP) is evaluated here not merely as an environmental compliance exercise, but as a critical infrastructure upgrade designed to mitigate operational risk and stabilize long-term energy costs. Following the simulation of 35 control strategies, the results are filtered to construct a robust business case that balances capital efficiency against the imperative for process safety. The selection of the “optimal” configuration is driven by a hierarchy of constraints: first, the system must guarantee the requisite operational resilience (24 h or 48 h autonomy) to secure critical port services against external grid failures; second, it must minimize the Levelized Cost of Energy (LCOE) below the current utility baseline to ensure financial viability; and third, it must maximize Environmental Abatement through high-fidelity renewable integration. Consequently, the discussion below isolates the most viable architectures, comparing the grid-stabilizing potential of Net Metering against the strict autonomy of off-grid storage topologies, to provide a validated roadmap for investment.

3. Results

The simulation results confirm that the transition to nZEP status is a technically feasible operational model, enhancing both grid stability and infrastructure resilience. The post-processing of the 35 scenarios demonstrates that high-penetration HRES integration yields robust techno-economic outcomes, effectively decoupling critical port loads from external grid volatility. Significantly, scenarios configured for 48 h autonomy, a critical safety benchmark for maintaining essential services during grid failures, maintain economic viability while ensuring unhindered operation. To validate system robustness against market and meteorological volatility, a sensitivity analysis was conducted on four stochastic variables: renewable energy (RE) potential, discount rate, inflation rate, and average base load (

Table 1. Sensitivity analysis outcomes: Impact of stochastic variables on LCOE robustness.

Variable	Value	NM (24 h)	NM (48 h)	Non-NM (24 h)	Non-NM (48 h)
		LCOE (€)	Δ (%)	LCOE (€)	Δ (%)
Discount Rate	4%	0.0608	−24.5%	0.0968	−20.0%
	6%	0.0722	−10.3%	0.108	−10.7%
	8% (Base)	0.0805	0%	0.121	0%
Inflation Rate	1%	0.0922	14.5%	0.129	6.6%
	1.50%	0.0885	9.9%	0.125	3.3%
	2% (Base)	0.0805	0%	0.121	0%
Daily Load	1497.6 kWh	0.0988	22.7%	0.0859	−29.0%
	2246.5 kWh	0.0933	15.9%	0.0986	−18.5%
	2995.3 kWh	0.0805	0%	0.121	0%
	3744.1 kWh	0.137	70%	0.15	36%
	4492.9 kWh	0.174	116%	0.174	66%
Solar Radiation	2.64 kWh/m2	0.298	270.2%	0.241	99.2%
	4.75 kWh/m2	0.198	146.0%	0.167	38.0%
	5.28 kWh/m2	0.0805	0%	0.121	0%
	5.81 kWh/m2	0.0761	−5%	0.102	−24%
	6.60 kWh/m2	0.0778	−3%	0.101	−25%

). Figure 7 illustrates the system’s stability boundaries under these varying constraints.

The smart microgrid controller (Figure 7) was evaluated under two distinct dispatch logics (Figure 8) to determine the optimal balance between grid stability and storage saturation: (a) Cycle Charging (CC) and (b) Load Following (LF). Under the CC protocol (which is picked), the controller decouples the generator/grid output from the immediate load, operating at maximum capacity to satisfy demand while simultaneously prioritizing the charging of the ESS. This strategy relies on real-time arbitrage between fixed and marginal costs to determine the most stable power mix. Conversely, the LF strategy creates a tight coupling between generation and demand, prioritizing ESS charging only up to a specified setpoint and dispatching surplus energy to the grid only after local loads are saturated. Following a rigorous comparative evaluation of the optimization matrix (

Table A3. Main techno-economic outcomes of the optimal NM systems per scenario.

			RES		Technology				Economy						Energy	Environment
Grid NM	a/a	Port Operations Autonomy	Grid (%)	RF (%)	PV (kWp)	WT (qty)	Inverter (kWp)	ESS (Model-qty)	NPC (M€)	Init. Cap. (k€)	ROI (%)	IRR (%)	PP (y)	LCOE (€/kWh)	Net Energy (kWh/y)	Emissions During Operation (tnCO2,eq/y)		gCO2,eq/kWh
																NO CF	CF
Net Metering	0	Baseline	100	0					5.09					0.360	1,093,280	2459.9		2250.0
	1	No autonomy	39.8	56.3	626		911	x	0.49	413.2	89.7	93.7	1.07	0.020	36	0.1	43.2	39.5
	2		36.6	58.1	592	1	400	x	0.71	606.0	59.6	64	1.56	0.029	449	1.0	45.0	41.2
	3		32.9	66.7		9		x	1.84	1590.0	19.6	24.3	4.07	0.088	−6665	0.0	28.5	26.0
	4	24 h autonomy	12.5	79.4	763		1000	FB250 (2)	1.15	1026.9	33.4	37.4	2.67	0.065	150	0.3	73.5	67.2
	5		12.3	77.4	794		906	FB30 (24)	1.16	1024.8	33.4	37.4	2.67	0.063	37	0.1	76.1	69.6
	6		13.8	83.0	644		769	OPzS4620 (598)	1.44	1207.4	27.1	31	3.22	0.085	43	0.1	50.6	46.3
	7		13.8	83.0	644		448	OPzS1905 (1488)	1.71	1240.1	24.8	28.8	3.47	0.100	−10	0.0	50.3	46.0
	8		13.8	82.9	644		931	OPzS4245 (638)	1.67	1418.2	22.4	26.3	3.79	0.098	36	0.1	50.7	46.4
	9		13.8	82.9	644		485	OPzS1875 (1481)	1.85	1381.9	21.9	25.8	3.87	0.108	95	0.2	50.3	46.0
	10		10.3	83.1	709	1	947	FB250 (2)	1.36	1210.2	27.6	31.7	3.15	0.078	28,042	63.1	123.0	112.5
	11	48 h autonomy	13.9	86.1	787		952	FB250 (4)	1.74	1585	20.1	24	4.15	0.106	−22	0.0	74.1	67.8
	12		14.2	85.8	836		762	FB30 (48)	1.74	1574	20.2	24.1	4.14	0.106	−6	0.0	77.9	71.3
	13		13.0	87.0	644		679	OPzS4620 (1196)	2.38	1993	14.2	17.9	5.49	0.147	−28	0.0	56.7	51.9
	14		13.0	87.0	644		400	OPzS1905 (2976)	2.91	2058	11.9	15.5	6.28	0.179	4	0.0	56.8	52.0
	15		13.0	87.0	644		546	OPzS4245 (1276)	2.83	2415	11	14.5	6.68	0.174	−20	0.0	56.5	51.6
	16		13.0	87.0	644		410	OPzS1875 (2962)	3.19	2342	10	13.4	7.14	0.196	−27	0.0	56.7	51.9
	17		12.3	87.7	762	1	617	FB30 (48)	1.94	1747	17.7	21.6	4.58	0.119	−8397	0.0	69.4	63.5

and

Table A4. Main techno-economic outcomes of the optimal non-NM systems per scenario.

			RES		Technology				Economy						Energy	Environment
Grid NM	a/a	Port Operations Autonomy	Grid (%)	RF (%)	PV (kWp)	WT (qty)	Inverter (kWp)	ESS (Model-qty)	NPC (M€)	Init. Cap. (k€)	ROI (%)	IRR (%)	PP (y)	LCOE (€/kWh)	Net Energy (kWh/y)	Emissions During Operation (tnCO2,eq/y)		gCO2,eq/kWh
																NO CF	CF
Without Net Metering	0	Baseline	100	0					1.62		0	0	0	0.360	1,093,280	2459.9		2250.0
	18	No autonomy	62.7	33.1	258.233		185.0	x	4.30	183.8	37.0	41.0	2.4	0.254	659,848	1484.7	14.8	456.9
	19		57.6	51.1	120.81	4	211.4	x	4.02	816.8	13.9	18.1	5.3	0.229	399,505	898.9	35.7	438.0
	20		73.0	53.3		6		x	4.18	1073	10.4	14.2	6.5	0.238	368,870	830.0	9.7	572.7
	21	24 h autonomy	12.4	79.8	850		215.3	FB250 (2)	2.45	1071	22.8	26.7	3.7	0.143	35,969	80.9	62.0	117.5
	22		13.1	75.5	850		215.3	FB30 (24)	2.64	1053	21.8	25.7	3.9	0.161	136,334	306.8	63.1	151.8
	23		12.0	86.7	850		215.3	OPzS4620 (598)	2.41	1310	19.5	23.4	4.3	0.138	−80,475	0.0	43.4	49.6
	24		12.0	86.7	850		215.3	OPzS1905 (1488)	2.67	1343	17.6	21.5	4.6	0.153	−80,557	0.0	43.7	50.0
	25		12.1	86.5	850		215.3	OPzS4245 (638)	2.65	1521	16.1	19.9	5.0	0.151	−78,341	0.0	43.4	49.7
	26		12.1	86.5	850		215.3	OPzS1875 (1481)	2.83	1485	15.5	19.2	5.1	0.162	−78,124	0.0	43.5	49.7
	27		10.4	88.9	770	2	954.7	FB250 (2)	2.37	1413	18.6	22.7	4.4	0.121	−252,906	0.0	47.5	54.3
	28	48 h autonomy	8.8	91.2	1050		215.3	FB250 (4)	2.43	1716	15.7	19.5	5.07	0.147	−71,496	0.0	66.1	75.6
	29		10.1	89.9	1050		238.9	FB30 (48)	2.48	1681	15.7	19.5	5.07	0.150	−57,430	0.0	66.9	76.5
	30		6.0	94.0	964		221.7	OPzS4620 (1196)	2.96	2153	11.4	14.9	6.51	0.168	−189,778	0.0	54.2	62.0
	31		6.0	94.0	964		218.8	OPzS1905 (2976)	3.49	2218	9.3	12.6	7.52	0.198	−185,543	0.0	54.7	62.5
	32		5.8	94.2	964		252.5	OPzS4245 (1276)	3.41	2575	8.8	12	7.82	0.187	−235,121	0.0	54.2	61.9
	33		5.5	94.5	964		312.9	OPzS1875 (2962)	3.77	2502	7.8	10.9	8.47	0.195	−317,088	0.0	55.4	63.3
	34		5.8	94.2	1031	1	268.1	FB250 (4)	2.50	1917	14.2	17.9	5.47	0.141	−195,919	0.0	58.3	66.7

), four superior system configurations were isolated for detailed analysis (Scenarios 4 and 7 for the Net Metering control architecture; Scenarios 23 and 27 for the non-NM case). Selection was driven by a hierarchical objective function targeting the minimization of LCOE and CF (Figure 8).

The selection of Cycle Charging as the primary dispatch strategy is not motivated solely by economic performance. From an operational perspective, Cycle Charging prioritizes the rapid recovery and preservation of the storage state of charge during favorable generation windows, thereby enhancing system readiness under disturbance conditions. This behavior is particularly relevant for safety-critical port infrastructures, where maintaining adequate energy reserves is more important than short-term load-tracking efficiency. By decoupling instantaneous load variations from generation dispatch, Cycle Charging reduces deep discharge events, stabilizes inverter operation, and improves overall system robustness under high renewable penetration and partial islanding conditions.

In this study, social acceptance is explicitly redefined as operational resilience, enforced via minimum autonomy thresholds (24 h and 48 h), ensuring uninterrupted supply to safety-critical port services during grid disturbances while simultaneously reducing peak stress on the surrounding urban distribution network.

3.1. Analysis of HRES with Net Metering Control Strategies

Under NM operation, the port microgrid is configured as a fully grid-interactive system, where the external electricity grid acts as a bidirectional balancing layer rather than a passive backup source. In this control regime, surplus renewable generation is systematically exported to the grid instead of being curtailed, while grid imports are used strategically to complement local generation and storage during periods of low renewable availability. This interaction fundamentally alters both the economic and operational behavior of the HRES, enabling high renewable penetration without compromising service continuity. From an operational standpoint, NM allows the ESS to operate in coordination with the grid rather than in isolation. As a result, system resilience is not achieved through excessive storage oversizing but through controlled grid interaction combined with a hard autonomy constraint. Figure 9 presents the financial and environmental performance envelope of all NM-enabled scenarios. The baseline case (Scenario 0) confirms the port’s current vulnerability, with total grid dependence, an LCOE of 0.360 €/kWh, and an operational carbon intensity exceeding 2250 gCO₂,eq/kWh. All NM scenarios demonstrate drastic improvements relative to this reference, validating the systemic impact of grid-interactive renewable integration.

Among the NM scenarios satisfying the 24 h autonomy requirement (Scenarios 4–10), a clear techno-economic hierarchy emerges. The best-performing configuration in economic terms is Scenario 5, which integrates 794 kWp of photovoltaic capacity, a 906 kWp inverter, and 24 FB30 VRFB modules. This system achieves the lowest LCOE of 0.063 €/kWh, corresponding to a payback period of 2.67 years, an IRR of 37.4%, and a return on investment of 33.4%. Compared to the baseline, this represents an LCOE reduction of more than 80%, driven by a high renewable fraction (77.4%), aggressive surplus exports, and the high cycling capability of the VRFB under NM dispatch. Scenario 4, based on two FB250 VRFB units, exhibits very similar financial performance (LCOE: 0.065 €/kWh) but a slightly lower renewable fraction (79.4%) due to its higher inverter sizing and different storage kinetics. Both VRFB-based solutions maintain low annual emissions, at 67.2–69.6 gCO₂,eq/kWh, confirming their strong environmental performance alongside economic competitiveness.

Lead–Acid-based NM systems (Scenarios 6–9) demonstrate consistently higher LCOE values (0.085–0.108 €/kWh) and longer payback periods, despite achieving comparable renewable fractions (~83%). This performance gap is primarily attributed to lower charge acceptance rates, stricter cycling limits, and higher effective capital intensity when scaled to meet autonomy requirements. Nevertheless, these systems still reduce operational carbon intensity to approximately 46 gCO₂,eq/kWh, highlighting their strong environmental performance under NM conditions. Scenario 10 represents a structurally different case, combining PV and wind generation with VRFB storage. Although it achieves a high renewable fraction (83.1%), its emissions increase substantially (112.5 gCO₂,eq/kWh) due to increased grid interaction and non-optimal temporal alignment between wind production and port load. This confirms that generation diversity alone does not guarantee environmental optimality under NM unless properly matched to demand dynamics.

For 48 h autonomy (Scenarios 11–17), the economic penalty of increased resilience becomes evident. All configurations in this group exhibit higher capital expenditure and LCOE values, reflecting the substantial storage oversizing required to maintain extended ride-through capability. The best-performing 48 h NM system is Scenario 12, utilizing 836 kWp of PV, a 762 kWp inverter, and 48 FB30 VRFB units, achieving an LCOE of 0.106 €/kWh and emissions of 71.3 gCO₂,eq/kWh. Scenario 11 (FB250-based) shows nearly identical performance, confirming that VRFB technology scales more efficiently than Lead–Acid alternatives when long-duration autonomy is required. Lead–Acid-based 48 h NM systems (Scenarios 13–16) suffer from sharply increasing LCOE values (0.147–0.196 €/kWh) and extended payback periods exceeding 5.5 years, despite achieving high renewable fractions (~87%). These results clearly indicate that Lead–Acid storage becomes economically and operationally suboptimal when deep autonomy constraints are imposed under NM operation. Scenario 17, which combines PV, wind, and VRFB storage, does not offset the resilience premium sufficiently, exhibiting a higher LCOE (0.119 €/kWh) without a proportional reduction in emissions. This reinforces the conclusion that storage technology selection and control strategy dominate system performance, rather than RES diversity alone.

The monthly stacked energy balance (Figure 10) highlights the seasonal redistribution of energy flows under Net Metering operation for the two examined scenarios. In both cases, grid purchases dominate during winter months, reflecting reduced photovoltaic availability and increased lighting-driven demand, while spring and summer periods are characterized by a strong shift toward locally consumed renewable energy. The progressive increase in the renewable fraction from late winter to early summer confirms effective photovoltaic utilization and limited curtailment under NM control, with peak renewable fractions exceeding 90%. Excess energy volumes remain moderate and temporally aligned with high renewable fractions, indicating that surplus production is largely absorbed through grid exports rather than storage saturation. The symmetry observed between the two scenarios suggests that differences in storage configuration primarily affect short-term dispatch rather than monthly energy balance aggregates.

The hourly heatmaps (Figure 11) provide a high-resolution view of intra-annual dispatch dynamics, clearly illustrating the diurnal and seasonal structure of photovoltaic generation. Renewable output is strictly confined to daylight hours, with maximum density observed between late spring and early autumn, coinciding with extended daylight duration and higher irradiance. This production profile is directly mirrored in the state-of-charge (SoC) heatmaps. In Scenario WNM_4, frequent saturation events are observed during summer midday hours, indicating aggressive charging and high storage utilization. In contrast, Scenario WNM_7 exhibits a smoother SoC distribution with fewer extreme saturation and depletion events, reflecting a more gradual charge–discharge behavior. The presence of low-SoC regions during winter nights in both scenarios highlights periods where storage alone is insufficient, underscoring the critical role of grid interaction in maintaining service continuity under NM operation.

The daily time-series plots (Figure 12) emphasize the complementary roles of grid purchases and grid sales within the NM framework. Grid purchases peak during winter, tracking seasonal demand increases and reduced renewable availability, while grid sales intensify during spring and summer, when photovoltaic production exceeds local consumption. The energy load profile remains comparatively smooth throughout the year, indicating that variations in grid interaction are driven primarily by supply-side dynamics rather than demand volatility. Scenario WNM_4 demonstrates higher amplitude fluctuations in both grid purchases and sales, suggesting more aggressive cycling and stronger grid coupling. Scenario WNM_7 exhibits more moderated grid exchanges, consistent with its storage configuration and dispatch characteristics. In both cases, the coexistence of substantial purchases and sales over the year confirms that the port operates as an active, bidirectional energy node rather than a passive consumer under Net Metering control.

Across all NM scenarios, the port consistently operates as a net energy exporter, with annual net energy balances remaining close to zero or negative. Grid exports are therefore transformed from curtailed losses into active economic and environmental assets. The lowest-emission NM configurations achieve operational carbon intensities below 50 gCO₂,eq/kWh, while the most economically optimized solutions remain below 0.065 €/kWh, values that are significantly lower than typical benchmarks reported for port-scale microgrids. The NM control strategy enables the port to transition from a passive grid-dependent consumer into a resilient, grid-supportive energy hub, delivering high renewable penetration, rapid investment recovery, and substantial carbon abatement while maintaining strict operational reliability constraints. Unlike the non-NM cases, grid exports (sales) here are active revenue streams rather than curtailed waste. Scenario 4 exhibits higher surplus injection than Scenario 7, correlating with higher daily sales peaks. Grid purchases are largely symmetrical, spiking primarily during winter irradiance deficits. High sales volumes during spring are attributed to lower base loads from lighting infrastructure (which accounts for >50% of total demand), coinciding with rising solar insolation. The annual net energy position for both systems is negative, confirming that the port functions as a net generator. The calculated LCOE falls well below comparable benchmarks in the literature, a discrepancy attributed to the use of high-fidelity, real-world cost data rather than generic estimates. The satisfaction of the “autonomy” constraint confirms that the port can transition from a passive load to a resilient energy hub, ensuring unhampered service delivery while stabilizing the local urban grid.

3.2. Analysis of HRES Without Net Metering Control Strategies

In the absence of NM arbitrage, the role of the external grid shifts fundamentally from a bidirectional balancing mechanism to a unidirectional supply source. Surplus renewable energy cannot be exported and monetized; instead, it is curtailed whenever instantaneous generation exceeds both the port demand and the charge acceptance limits of the ESS. Consequently, the system architecture prioritizes autonomous operation and reliability over economic arbitrage, and the ESS becomes a critical infrastructure component rather than a cost-optimization tool. Under this control paradigm, resilience is achieved exclusively through local generation and storage adequacy. Figure 13 summarizes the financial and environmental performance of all non-NM configurations. The baseline case (Scenario 0) confirms the extreme inefficiency of grid-only operation, exhibiting an LCOE of 0.360 €/kWh and an operational carbon intensity exceeding 2250 gCO₂,eq/kWh. All non-NM HRES scenarios yield substantial improvements relative to this reference; however, their performance is strongly constrained by unavoidable curtailment losses and limited flexibility in grid interaction.

Scenarios 18–20 represent non-autonomous systems optimized primarily for partial renewable penetration without storage-driven resilience. These configurations achieve moderate renewable fractions (33–53%) but remain heavily dependent on grid imports, with grid contribution exceeding 57% in all cases. Although capital requirements are relatively low, their LCOE values remain high (0.229–0.254 €/kWh), reflecting inefficient utilization of renewable generation and the absence of storage-enabled load shifting. Environmentally, these scenarios reduce emissions compared to the baseline but still exhibit carbon intensities above 430 gCO₂,eq/kWh, rendering them unsuitable for safety-critical port applications.

The monthly energy balance (Figure 14) clearly illustrates the structural impact of operating without Net Metering. In both scenarios, renewable penetration increases sharply from spring to early autumn, with renewable fractions exceeding 85–90% during peak months. However, unlike NM cases, a substantial portion of this renewable generation manifests as excess energy, particularly between April and September, indicating systematic curtailment once storage charge limits are reached. Scen. NM_27 exhibits lower grid purchases throughout most of the year compared to NM_23, reflecting its higher renewable fraction and improved temporal alignment enabled by generation diversity. Winter months remain grid-dominated in both cases, driven by reduced solar availability and increased demand, but the seasonal contrast is markedly stronger in NM_23, where PV-only production leads to steeper transitions between surplus and deficit periods.

The introduction of a 24 h autonomy constraint (Scenarios 21–27) marks a structural transition in system behavior. These configurations maintain grid participation at approximately 10–13%, while renewable fractions exceed 75%, confirming effective decoupling of generation from demand through storage. From an economic perspective, Scenario 27 emerges as the optimal non-NM configuration. By combining 770 kWp of photovoltaic capacity, two wind turbines, and two FB250 VRFB units, this system achieves the lowest LCOE of 0.121 €/kWh within the non-NM class, alongside a payback period of 4.4 years. The inclusion of wind generation introduces temporal diversity, reducing reliance on diurnal solar production and lowering curtailment volumes. Environmentally, Scenario 27 achieves a carbon intensity of 54.3 gCO₂,eq/kWh, representing a reduction exceeding 97% relative to the baseline. Among PV-only configurations, Scenario 21 (FB250-based) represents the most balanced solution, achieving an LCOE of 0.143 €/kWh and emissions of 117.5 gCO₂,eq/kWh. VRFB-based solutions consistently outperform Lead–Acid alternatives in economic terms due to higher cycle efficiency and superior depth-of-discharge tolerance. Lead–Acid systems (Scenarios 23–26) achieve lower operational emissions (≈50 gCO₂,eq/kWh) but at the expense of higher capital intensity, longer payback periods, and reduced economic attractiveness.

The heatmaps (Figure 15) reveal the intrinsic limitations of non-NM operation at high renewable penetration. Renewable output is strictly diurnal, with dense production bands during midday hours that intensify from late spring to early autumn. In Scenario NM_23, the SoC heatmap shows frequent and prolonged saturation during daylight hours, indicating that the ESS reaches full capacity early and remains unable to absorb additional energy, directly triggering curtailment. Conversely, Scenario NM_27 displays a more distributed SoC profile, with fewer saturation plateaus and improved overnight charge retention. This behavior reflects the contribution of wind generation, which provides partial nocturnal charging and reduces the depth of discharge during pre-dawn hours. The contrast highlights how generation diversity acts as a stabilizing factor under non-NM conditions, mitigating extreme cycling and reducing storage stress.

Scenarios 28–34 extend the autonomy requirement to 48 h, substantially increasing storage capacity and capital expenditure. While renewable fractions exceed 90% in all cases, the economic penalty of long-duration autonomy becomes evident. LCOE values rise to the 0.147–0.198 €/kWh range, with payback periods extending beyond 5 years. Scenario 28, based on four FB250 VRFB units, represents the most economically efficient 48 h autonomous system, achieving an LCOE of 0.147 €/kWh and emissions of 75.6 gCO₂,eq/kWh. Lead–Acid configurations (Scenarios 30–33) achieve slightly lower carbon intensities (≈62 gCO₂,eq/kWh) but exhibit sharply diminishing returns due to excessive storage oversizing and increased curtailment under high renewable penetration. Scenario 34 demonstrates that reintroducing limited wind generation can partially offset this penalty but cannot fully compensate for the structural inefficiencies imposed by the absence of Net Metering.

Figure 15 illustrates daily grid interaction and curtailment dynamics. In non-NM operation, grid sales are structurally absent; any surplus generation beyond ESS charge acceptance is curtailed to maintain frequency stability. Curtailment therefore represents a necessary control action, not a system malfunction. PV-dominant configurations exhibit pronounced midday curtailment peaks, particularly during summer months, driven by the mismatch between concentrated solar output and relatively flat port demand profiles. The inclusion of wind generation (Scenario 27) redistributes energy production across nocturnal periods, reducing curtailment and flattening residual load curves. Actually, the daily time-series plots (Figure 16) emphasize the unidirectional nature of grid interaction in non-NM configurations. Grid purchases dominate during winter and shoulder seasons, while summer periods are characterized by near-zero imports despite high renewable availability, confirming that surplus energy is curtailed rather than exported. Scenario NM_23 exhibits sharper and more frequent grid purchase spikes during winter, corresponding to rapid ESS depletion following extended low-irradiance periods. In contrast, Scenario NM_27 demonstrates smoother grid interaction, with reduced purchase volatility and more consistent residual load coverage. The energy load profile remains relatively stable across the year, confirming that observed variations are supply-driven rather than demand-induced. Overall, the plots demonstrate that without NM, the grid functions purely as a fallback supply, while the ESS bears the full burden of balancing intermittency, an operational regime that inherently increases curtailment and capital intensity.

The non-NM operation enforces a fundamentally different optimization logic: resilience is achieved through storage and generation oversizing rather than grid interaction. While high renewable penetration and deep decarbonization are technically feasible, they are accompanied by higher system costs and reduced economic efficiency. These results confirm that Net Metering is not merely a financial incentive mechanism but a structural enabler of efficient, resilient port microgrids, whereas its absence necessitates conservative, capital-intensive design choices to ensure operational safety.

To assess the robustness of the proposed system configurations under varying economic and operational conditions, a sensitivity analysis was performed. Table 1 summarizes the influence of four key parameters, discount rate, inflation rate, baseline electrical demand, and solar irradiance, on the resulting LCOE.

The sensitivity analysis highlights the extent to which the techno-economic performance of the optimal system configurations depends on both financial assumptions and key operational parameters (Figure 17). As expected for capital-intensive energy infrastructures, variations in the discount rate exert a strong influence on the resulting LCOE for both Net Metering and non-Net Metering configurations. A reduction in the discount rate from the base value of 8% to 4% leads to a substantial decrease in LCOE, which is translated to 24.5% for the NM (24 h autonomy) case and to 20.0% for the non-NM (24 h autonomy) case, whereas higher discounting progressively undermines economic viability. Changes in the inflation rate follow a similar, albeit less pronounced, pattern, confirming the sensitivity of long-term cost performance to macroeconomic conditions.

Beyond financial parameters, the analysis clearly identifies solar resource availability as the dominant operational driver of cost variability. Reductions in mean daily solar irradiation result in a rapid escalation of LCOE, particularly in configurations with higher autonomy requirements. When solar radiation decreases from the base level of 5.28 kWh/m²/day to 2.64 kWh/m²/day, the LCOE increases by 270.2% in the NM configuration and by 99.2% in the non-NM configuration. These results indicate that, while the system is structurally capable of accommodating fluctuations in demand, its economic performance remains highly exposed to prolonged reductions in solar input. Conversely, increases in irradiation above the base case yield comparatively limited cost reductions, suggesting diminishing marginal returns once the PV subsystem approaches optimal utilization.

Variations in mean daily load reveal a non-linear response. Demand levels below the design point improve economic performance, particularly for the non-NM configuration, where reduced grid dependence translates directly into lower LCOE. In contrast, demand growth beyond the nominal capacity leads to a sharp deterioration in both configurations, with LCOE increases reaching 116% for the NM case and 66% for the non-NM case at the highest load level examined. This behavior reflects the combined effects of storage oversizing, increased cycling losses, and greater reliance on grid electricity during peak conditions. Finally, the comparison between 24 h and 48 h autonomy scenarios allows the economic implications of enhanced resilience to be quantified. Across all sensitivity dimensions, extending autonomy to 48 h systematically increases LCOE due to the additional storage capacity required. This resilience premium is substantial and highlights a central design trade-off: while higher autonomy improves operational robustness and security of supply, it does so at a non-negligible economic cost.

The sensitivity analysis suggests that system optimization efforts should prioritize accurate assessment of local solar resources and realistic financial assumptions, while autonomy levels should be selected with careful consideration of their marginal contribution to resilience relative to their cost impact. The sensitivity analysis conducted in this study is intentionally confined to parameters that directly affect the physical and operational performance of the HRES, including renewable resource availability, electrical demand, discount rate, and inflation. Policy-driven variables such as electricity tariff evolution and Net Metering regulatory persistence are treated as boundary conditions rather than stochastic inputs, as their future trajectories are jurisdiction-specific, non-technical, and outside the control of system design. Incorporating speculative assumptions on regulatory persistence would not improve the robustness of the engineering conclusions and could compromise the methodological transparency of the framework.

It is acknowledged that autonomy alone does not fully encapsulate all dimensions of resilience, such as recovery speed, adaptive capacity, or cyber-physical robustness. However, within the context of port energy systems, enforced autonomy thresholds provide a transparent, quantifiable, and regulator-relevant benchmark that enables systematic comparison across control strategies and HRES architectures while maintaining a strong focus on operational safety.

4. Discussion

The results of this study confirm that the transition of a medium-sized Mediterranean port towards an nZEP constitutes not merely an environmental compliance exercise, but a structurally superior operational strategy when evaluated through a combined techno-economic, environmental, and resilience lens. The systematic simulation of 35 HRES configurations using high-resolution, 15 min real load data demonstrates that deep decarbonization can be achieved without compromising the stringent reliability requirements of port infrastructure.

The findings show that the feasibility and performance of this transition are governed primarily by the grid-interface control strategy rather than by renewable capacity alone. The term “validated” in this study refers to internal validation through high-resolution empirical load data and systematic scenario stress-testing, rather than external statistical generalization. Replicability is understood in an engineering sense, indicating that the proposed modeling workflow, control logic, and decision framework can be transferred to other ports with analogous operational characteristics after site-specific data substitution. Numerical results are therefore not claimed to be universally generalizable, whereas the methodological structure and design logic are.

4.1. System-Level Techno-Economic and Environmental Performance Under Alternative Grid-Interface Strategies

The optimization results reveal a clear and robust hierarchy driven by the adopted control logic. NM configurations consistently outperform non-NM systems by leveraging the external grid as an active balancing layer rather than a passive energy source. The optimal NM configuration achieves an LCOE as low as 0.063 €/kWh under a strict 24 h autonomy constraint, representing an order-of-magnitude improvement relative to the baseline grid-dependent operation. Even when extended to higher resilience levels, NM-enabled systems maintain LCOE values below 0.11 €/kWh, confirming that resilience does not necessarily imply prohibitive cost escalation when bidirectional grid interaction is permitted.

In contrast, non-NM configurations rely exclusively on local generation and storage to achieve autonomy, leading to systematic renewable curtailment and higher capital intensity. Although deep decarbonization remains technically achievable, operational carbon intensities fall below 55 gCO₂,eq/kWh in optimized non-NM cases, and the associated LCOE increases to the 0.12–0.15 €/kWh range for comparable autonomy levels. This clearly exposes a resilience premium intrinsic to curtailment-based operation, where reliability is purchased through oversizing rather than intelligent energy exchange.

4.2. Role of Grid-Interface Control in Dispatch Flexibility and Network Interaction

A key contribution of this study lies in its explicit treatment of the utility grid as a controllable system component. Under NM operation, the grid effectively functions as a virtual, infinite-capacity buffer that absorbs surplus generation during periods of high solar availability and supplies deficit energy during prolonged low-irradiance events. This interaction significantly reduces curtailment and smooths residual load profiles, allowing the ESS to operate within favorable cycling regimes.

However, the results also highlight a structural dependency on solar resource availability. Sensitivity analysis indicates that reductions in solar irradiance lead to non-linear increases in LCOE, particularly in PV-dominant configurations. This finding underscores the importance of generation diversity as a stabilizing mechanism, especially in ports exposed to seasonal variability. Wind integration, even at moderate capacities, demonstrably improves temporal alignment between generation and demand, reducing both curtailment and grid dependency in non-NM systems and enhancing robustness in NM configurations.

4.3. Energy Storage Technology Selection: Operational Safety, Cycling Robustness, and Resilience Implications

The comparative evaluation of storage technologies yields critical insights for urban–maritime energy planning. While Lead–Acid systems occasionally achieve marginally lower capital costs in high-autonomy scenarios, their operational limitations, restricted depth of discharge, reduced cycle life, and lower charge acceptance become increasingly evident at high renewable penetration levels. VRFBs, by contrast, consistently demonstrate superior performance under both NM and non-NM control strategies. Beyond economic metrics, VRFBs offer decisive advantages from a process safety and resilience perspective. Their inherent non-flammability, thermal stability, and decoupling of power and energy capacity make them particularly suitable for ports, where energy infrastructure coexists with hazardous cargo, dense passenger flows, and safety-critical systems. The results suggest that, for maritime critical infrastructure, chemical stability and operational robustness should be treated as non-negotiable design constraints rather than secondary optimization criteria.

4.4. Methodological Implications of High-Resolution Load Modeling in HRES Design

Compared to existing HRES sizing studies, which predominantly rely on hourly or synthetic demand profiles, the use of 15 min measured load data reveals dynamics and peak events that coarser models systematically obscure. This higher temporal resolution exposes storage saturation, curtailment thresholds, and short-duration demand spikes that directly influence both system sizing and economic outcomes. Consequently, the resulting designs are more conservative but significantly more robust from an operational standpoint. The achieved LCOE and carbon intensity values are competitive with, and in several cases superior to, those reported in recent studies on islanded and port microgrids. This performance is attributed not only to favorable solar resources but primarily to the explicit modeling of control strategies and real operational constraints, rather than idealized dispatch assumptions.

4.5. Model Boundaries, Regulatory Assumptions, and Directions for Extended System Integration

Despite its robustness, the proposed framework is subject to two main limitations. First, wave and tidal energy systems were excluded due to the absence of validated, site-specific hydrokinetic data, limiting the exploration of full marine energy integration. Second, the Net Metering regulatory framework was assumed to remain static throughout the project lifetime, whereas real-world tariff structures are dynamic and policy dependent.

Future research should therefore focus on:

• Integrating green hydrogen as a long-duration seasonal storage vector to mitigate winter solar deficits;
• Implementing advanced control strategies, such as Model Predictive Control, to exploit dynamic pricing and demand response; and
• Expanding the system boundary to include thermal loads and the electrification of port logistics, enabling fully sector-coupled smart port ecosystems.

4.6. Engineering Implications for the Design of Resilient Nearly Zero Energy Ports

The implications of this research work extend beyond the examined case study. The validated typology provides a transferable, engineering-driven framework for transforming ports from passive energy consumers into resilient, grid-supportive energy hubs. By prioritizing control strategy selection, operational autonomy, and storage safety, port authorities can simultaneously enhance energy security, reduce exposure to external shocks, and accelerate urban decarbonization. In this sense, the nZEP paradigm emerges not as a policy-driven abstraction, but as a technically grounded and economically rational pathway for the future of sustainable port infrastructure.

5. Conclusions

This study developed and validated a resilience-oriented control and sizing framework for the transition of a medium-sized Mediterranean port towards an nZEP. By exploiting high-fidelity operational data at 15 min resolution and systematically evaluating 35 HRES configurations, the proposed approach explicitly integrates technical performance, operational resilience, and process safety into the system design. The resulting typology demonstrates that port decarbonization can be achieved as a technically robust and economically rational engineering solution, rather than a purely regulatory or environmental exercise.

The results clearly indicate that the grid-interface control strategy is the dominant determinant of system performance. NM-enabled architectures consistently outperform non-NM configurations by leveraging the external grid as an active balancing layer rather than a passive backup source. The optimal NM configuration, satisfying a strict 24 h autonomy constraint, achieves a minimum LCOE of 0.063 €/kWh, while reducing operational CF to approximately 70 gCO₂,eq/kWh, corresponding to a reduction exceeding 90% relative to baseline grid-dependent operation. These results confirm that grid-interactive control strategies are structurally enabling mechanisms for resilient and cost-effective decarbonization of urban–maritime infrastructure.

In the absence of NM, autonomy can still be achieved through local generation and storage, albeit at a higher cost. Optimized non-NM configurations demonstrate that full operational autonomy is technically feasible, with carbon intensities below 55 gCO₂,eq/kWh; however, this is accompanied by a pronounced resilience premium, as Levelized Cost of Energy increases to the 0.12–0.15 €/kWh range due to renewable curtailment and storage oversizing. The inclusion of generation diversity, particularly wind energy, partially mitigates this penalty by improving temporal alignment between production and demand and reducing deep storage discharge during nocturnal periods. From a technological and safety perspective, the comparative analysis strongly supports the deployment of VRFBs over Lead–Acid alternatives for port applications. Despite higher initial capital expenditure, VRFBs exhibit superior cycling robustness, operational flexibility, and inherent safety characteristics, namely, non-flammability and thermal stability, which are critical in dense, hazardous, and safety-critical port environments. These attributes justify their selection as a cornerstone technology for resilient maritime microgrids. Sensitivity analysis further reveals that the primary risk to nZEP stability is not demand variability, which can be effectively managed through sizing and storage, but rather stochastic solar resource variability. Under low-irradiance conditions, the cost of maintaining autonomy increases non-linearly, reinforcing the need for generation diversification and advanced control strategies. Future developments should therefore focus on integrating additional renewable vectors (e.g., wind, wave, and tidal energy), incorporating Model Predictive Control (MPC) for dynamic dispatch and pricing response, and evaluating green hydrogen as a long-duration seasonal storage option to address winter production deficits.

This research work clearly demonstrates that the transition to nZEP status is a feasible, scalable, and executable engineering pathway. The proposed framework provides a reproducible and transferable methodology enabling port authorities to transform from passive energy consumers into resilient, grid-supportive nodes within the smart city energy ecosystem, while ensuring uninterrupted operation under increasingly volatile energy and climate conditions.