Offshore CO2 storage can support industrial decarbonization by reusing existing oil and gas infrastructure. However, repurposed offshore platforms require a reliable, low-emission power supply within limited space. This thesis develops a generalizable mixed-integer linear programming model for the techno-economic sizing optimization and dispatch of space-constrained offshore microgrids for CO2 storage platforms. This work combines hourly renewable generation, short-duration and long-duration battery storage, a diesel generator, and physical platform installation constraints in one model. Spatial feasibility is represented through platform zone dimensions, including battery room layouts, technology eligibility, and component dimensions. From these parameters, the proposed spatial preprocessing method derives component upper bounds in addition to aggregated area constraints. We investigate the integration of long-duration energy storage (LDES) as a replacement for diesel backup and an enabler for carbon-neutral operation. The model is evaluated using a benchmark configuration of the P18 platform in the Dutch North Sea and with scenario groups that vary area, layout, battery technology, generator availability, demand, and location. Results show that extra installation space mainly changes the photovoltaic (PV) and wind mix while the generator is operated close to its emission limit as long as it remains part of the installed energy resources. PV is preferred when space is not binding, while under generator-free operation, platform deck space becomes binding and wind and LDES become critical to the operation. LDES based on sodium metal chloride and aqueous zinc storage makes carbon-neutral operation feasible, reduces costs by 23%, and absorbs prolonged energy surpluses. The model can be applied to platforms that vary in layout, components, location, and load profile. It can be used to identify the physical and economic thresholds that determine whether an offshore platform can move from generator-supported operation to carbon-neutral operation.
... Read more

DC microgrids are increasingly relevant for integrating distributed energy resources, storage systems, and power-electronic loads. However, their dynamic behavior is strongly influenced by converter control loops, line impedances, and load interactions. As these systems become increasingly converter-dominated, stability must be assessed using models that capture both electrical and control dynamics.

This thesis develops a modular small-signal modeling framework for converter-dominated DC microgrids using linearized state-space representations. Rather than deriving a single monolithic model, individual subsystems, including converters, controllers, lines, loads, buses, and interconnection elements, are formulated as separate linear time-invariant modules. These modules are interconnected in MATLAB using the Robust Control Toolbox. The resulting system-level model enables eigenvalue analysis, participation factor analysis, and parameter-sweep-based sensitivity analysis.

The framework is applied to a case study consisting of three parallel interleaved boost converters connected to a common DC bus. The system includes line impedances, cascaded PI control, droop control, and a resistive load. A nonlinear Simulink model is used to validate the linearized model under source-voltage, voltage-reference, and load-current perturbations. The comparison shows that the modular model captures the dominant system-level behavior, steady-state response, and current-sharing trends sufficiently well for small-signal stability analysis. Differences remain during larger transients, as expected from a model that is locally valid around a selected operating point.

The results show that eigenvalue analysis can determine the local stability of the operating point, while participation factors link the relevant modes to physical system states. In the case study, the fastest modes are primarily associated with line-capacitance dynamics, whereas the modes governing the stability margin are mainly associated with the voltage-controller states. The framework therefore identifies both whether the system is stable and which components or controllers dominate its critical dynamics.

Sensitivity analysis further shows that droop coefficients significantly influence the system eigenstructure. Increasing a droop coefficient may improve the stability margin of one mode while reducing the damping of other oscillatory modes. Droop tuning should therefore be treated as a coupled system-level design problem rather than as an independent adjustment of individual converters.

The main conclusion is that modular small-signal state-space modeling provides a practical and physically interpretable method for analysing converter-dominated DC microgrids. By preserving access to individual subsystem states while enabling system-level modal and sensitivity analyses, the framework can support stability-oriented design and controller tuning. Future work should include more detailed source, storage, and load models, digital control effects, reduced-order modeling, systematic optimization, and experimental validation. ... Read more

The recent rise in power demand in data centers, driven largely by AI workloads, has exposed the limitations of conventional AC and radial DC distribution architectures, particularly the losses and added complexity introduced when each stage of voltage conversion is handled by a separate, single-purpose
converter. This thesis investigates whether a single multiport converter, designed around the operating conditions of a meshed DC microgrid, can replace this stack of point-of-load converters while preserving efficiency, redundancy, and standards compliance.

A Tier IV, 2N redundant DC data center was modeled in DCIDE as a meshed system under power-voltage droop control. Lifetime-weighted simulation across normal, degraded, and fault scenarios showed that each port of the converter spends the majority of its operating life well below rated power. This helped determine the right value that the converter ports should be sized for, rather than sizing them around their rated maximum, as is conventional. Building on this, a port-specific earthing strategy was derived against the Current/OS distribution standard, and candidate multiport converter topologies were compared using a weighted scoring framework using cost, complexity, and efficiency.

The selected topology, built around the dual active bridge (DAB) as its isolated DC-DC stage, was validated first in MATLAB Simulink using an ideal-component model with closed-loop voltage control for the two voltage-regulated ports and open-loop, fixed-phase-shift control for the current-controlled battery-interfacing port, and then in PLECS using manufacturer-derived switching-loss and thermal
models. The PLECS results showed sub-converter efficiencies above 97.5%, junction temperatures within design limits, and an overall simulated system efficiency of 98.29% against a theoretical estimate of 98.00%, validating the functionality and efficiency of the selected topology.

Beyond the data center case study, this thesis develops a transferable methodology for sizing, earthing, and selecting multiport converter topologies in meshed DC microgrids more broadly. ... Read more