Doctor of Philosophy (PhD)
Division of Electrical Engineering
In recent years, large scale deployments of electrical energy generation using renewable sources (RES) such as wind, solar and ocean wave power, along with more sustainable means of transformation have emerged in response to different initiatives oriented toward reducing greenhouse gas emissions. Strategies facilitating the integration of renewable generation into the grid and electric propulsion in transportation systems are proposed in this work.
Chapter 2 investigates the grid-connected operation of a wave energy converter (WEC) along with a hybrid supercapacitor/undersea energy storage system (HESS). A combined sizing and energy management strategy (EMS) based on reinforcement learning (RL) is proposed. Comparisons in terms of power and energy capacity between the HESS sized with the proposed approach, and SC-only and UESS-only cases are performed. To facilitate fair comparisons a similar WEC output power profile is employed, and it is assumed that the storage components, as hybrid or individually, counteract the power variations. The adaptability of the RL-based EMS is verified using different power profiles in the learning and testing phases. Real-time simulation results corroborate that the capacity of the HESS components is notably reduced when EMS is considered in the sizing stage. Furthermore, RL-based EMS is able to regulate WEC output power, even in presence of serious imbalances between harvested and dispatched wave energy.
In the marine sector, new shipboard power system architectures are been proposed in response to the increasing use of electric propulsion, e.g. medium-voltage dc (MVDC) topologies. Due to interactions of the ship and the propeller with sea waves, large thrust/torque variations are translated into steep power fluctuations on the MVDC bus of the ship, affecting the stability and quality of the onboard power grid. A method to joint sizing/EMS a HESS comprising battery and supercapacitor to mitigate power fluctuations on the medium-voltage dc bus associated to propulsion system thrust/torque variations is studied in chapter 3. A deep reinforcement learning framework is employed for the joint sizing/EMS problem. The proposed strategy avoids the requirement for knowledge of the ship propulsion power profile, and it features adaptability to varying sea states and feasibility of real-time implementation. A comparative analysis between the HESS designed with the proposed methodology and the cases where battery-only, and SC-only, mitigate power fluctuations caused by propulsion system variations demonstrates the efficacy of the joint sizing/EMS on reducing the size of the energy storage system. Furthermore, real-time implementation feasibility and adaptability to different ship propulsion power profiles is validated through real-time simulations.
In chapter 4, a control method for grid-side power electronic converters in grid-connected renewable energy generators (REG) is presented. The scheme, known as hybrid data-model predictive direct power control (HD-MPDPC), employs long-prediction horizons to provide more reliable REG output power dispatch. Computational load of classical MPDPC is mitigated by reducing the number of candidate voltage vectors to be examined in the cost function. Candidate voltage vector reduction is accomplished by using data-driven forecast of REG output power and the principle of direct power control (DPC). REG power is forecasted using recurrent neural networks. In DPC, active and reactive power hysteresis controllers along with the sector in which grid voltage vector lies are used to determine the switching states of the grid-side power electronic converter. Thanks to reduced computational burden offered by the hybrid structure, the proposed strategy is able to dispatch REG power more reliably over long horizons. This in turn enables REG as a regulating reserves service provider in power systems. Real-time simulation studies of a grid-connected wave energy conversion system demonstrate the reduced computational toll of HD-MPDPC and its effectiveness in regulating REG output power over long horizons.
Nunez Forestieri, Juan Rafael, "Intelligent Data-Driven Energy Flow Controllers for Renewable Energy and Electrified Transportation Systems" (2020). LSU Doctoral Dissertations. 5394.
Available for download on Sunday, October 24, 2021