With the ever-growing developments of sustainable energy sources such as fuel cells, photovoltaics, and other distributed generation, the need for a reliable power conversion system that interfaces these sources is in great demand. In order to provide the highest degree of flexibility in a truly distributed network, it is desired to not only interface multiple sources, but to also interface multiple loads. Modern multi-port converters use high frequency transformers to deliver the different power levels, which add to the size and complexity of the system. The different topological variations of the proposed multi-port dc-dc converter have the potential to solve these problems. This thesis proposes a unique dc-dc current source converter for multi-port power conversion. The presented work will explain the proposed multi-port dc-dc converter's operating characteristics, control algorithms, design and a proof of application. The converter will be evaluated to determine its functionality and applicability. Also, it will be shown that our converter has advantages over modern multi-port converters in its ease of scalability from kW to MW, low cost, high power density and adaption to countless combinations of multiple sources. Finally we will present modeling and simulation of the proposed converter using the PSIM software. This research will show that this new converter topology is unstable without feedback control. If the operating point is moved, one of the source ports of the multiport converter becomes unstable and dies off supplying very little or no power to the load while the remaining source port supplies all of the power the load demands. In order to prevent this and add stability to the converter a simple yet unique control method was implemented. This control method allowed for the load power demanded to be shared between the two sources as well as regulate the load voltage about its desired value.

This dissertation studies hybrid dc-dc power conversion systems based on multiple-input converters (MICs), or more generally, multiport converters. MICs allow for the integration of multiple distributed generation sources and loads. Thanks to the modular design, an MIC yields a scalable system with independent control in all sources. Additional characteristics of MICs include the improved reliability and reduced cost. This dissertation mainly studies three issues of MICs: efficiency improvement, modeling, and control. First, this work develops a cost-effective design of a highly-efficient non-isolated MIC without additional components. Time-multiplexing (TM) MICs, which are driven by a time-multiplexing switching control scheme, contain forward-conducting-bidirectional-blocking (FCBB) switches. TM-MICs are considered to be subject to low efficiency because of high power loss introduced by FCBB switches. In order to reduce the power loss in FCBB switches, this work adopts a modified realization of the FCBB switch and proposes a novel switching control strategy. The design and experimental verifications are motivated through a multiple-input (MI) SEPIC converter. Through the design modifications, the switching transients are improved (comparing to the switching transients in a conventional MI-SEPIC) and the power loss is significantly reduced. Moreover, this design maintains a low parts-count because of the absence of additional components. Experimental results show that for output power ranging from 1 W to 220 W, the modified MIC presents high efficiency (96 % optimally). The design can be readily extended to a general n-input SEPIC. The same modifications can be applied to an MI-Ćuk converter. Second, this dissertation examines the modeling of TM-MICs. In the dynamic equations of a TM-MIC, a state variable from one input leg is possible to be affected by state variables and switching functions associated with other input legs. In this way, inputs are coupled both topologically and in terms of control actions through switching functions. Coupling among the state variable and the time-multiplexing switching functions complicate TM-MICs’ behavior. Consequently, substantial modeling errors may occur when a classical averaging approach is used to model an MIC even with moderately high switching frequencies or small ripples. The errors may increase with incremental number of input legs. In addition to demonstrating the special features on MIC modeling, this dissertation uses the generalized averaging approach to generate a more accurate model, which is also used to derive a small-signal model. The proposed model is an important tool that yields better results when analyzing power budgeting, performing large-signal simulations, and designing controllers for TM-MICs via a more precise representation than classical averaging methods. Analyses are supported by simulations and experimental results. Third, this dissertation studies application of a decentralized controller on an MI-SEPIC. For an MIC, a multiple-input-multiple-output (MIMO) state-space representation can be derived by an averaging method. Based on the averaged MIMO model, an MIMO small-signal model can be generated. Both conventional method and modern multivariable frequency analysis are applied to the small-signal model of an MI-SEPIC to evaluate open-loop and closed-loop characteristics. In addition to verifying the nominal stability and nominal performance, this work evaluates robust stability and robust performance with the structured singular value. The robust performance test shows that a compromised performance may be expected under the decentralized control. Simulations and experimental results verify the theoretical analysis on stability and demonstrate that the decentralized PI controller could be effective to regulate the output of an MIC under uncertainties. Finally, this work studies the control of the MIMO dc-dc converter serving as an active distribution node in an intelligent dc distribution grid. The unified model of a MIMO converter is derived, enabling a systematical analysis and control design that allows this converter to control power flow in all its ports and to act as a power buffer that compensates for mismatches between power generation and consumption. Based on the derived high-order multivariable model, a robust controller is designed with disturbance-attenuation and pole-placement constraints via the linear matrix inequality (LMI) synthesis. The closed-loop robust stability and robust performance are tested through the structured singular value synthesis. Again, the desirable stability and performance are verified by simulations and experimental results. / text