Modeling power distribution system of an electric ship for design and control

• Main Author: Ayu Aaron Alexander
• Other Authors: Jaspreet Singh Dhupia
• Format: Theses and Dissertations
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Mechanical engineering::Mechatronics; DRNTU::Engineering::Mathematics and analysis::Simulations; DRNTU::Engineering::Systems engineering; DRNTU::Engineering::Mechanical engineering::Power resources; DRNTU::Engineering::Mechanical engineering::Control engineering
• Online Access: https://hdl.handle.net/10356/65669
• Institution: Nanyang Technological University

This thesis addresses the design and control scheme optimization problem faced in the marine industry in particular, hybrid electric vessels. The performance and the achievable fuel efficiency of a marine hybrid vessel such as an electric tugboat depends on its design, i.e., the installed capacity of diesel engine-generators and batteries. This thesis presents a formulation to determine the optimal number of diesel engine-generators along with their corresponding power ratings and the total batteries' capacity that achieves an optimal trade-off between the design and operating costs for a marine vessel having a given operating characteristics. Design optimization begins with a modeling of electric tugboat power distribution system model. The power demand is met by a set of diesel generators and batteries whose switching on/off and power output is regulated by a rule-based controller. This power distribution model along with the rule-based controller is programmed in MATLAB/ Simulink, which is optimized to determine the optimal installed capacity of diesel engine-generators and batteries. The optimal solution accounts for the cost trade-offs among the fuel, equipment and design space. The cost saving achieved from the evaluated optimal configuration through application of hybrid power plant is compared against mechanical ships over a finite horizon period. Simulation of the optimized tugboat configuration subject to an industry referenced operational load profile shows a $10.2\%$ of fuel savings over a period of 10 years with the additional investment in equipment for hybrid configuration recoverable after 2 years. Since the current high investment in batteries is considered the main barrier in hybrid marine vessel technology, the robustness of the achievable fuel efficiency is considered as the battery size is varied. It is shown that a large battery size does not give significant improvements to the achievable efficiency of the system. Finally, the sensitivity of design optimization results with respect to variation in fuel price over different return on investment horizon periods is also investigated. This thesis next presents an optimal power management scheme for an electro-mechanical powertrain system of a marine vessel. To optimally split the power supply from engines and batteries in response to the load demand, while minimizing the engine fuel consumption and maintaining the batteries life, an optimization problem is formulated, in which the cost function associates penalties corresponding to the error in load tracking, the engine fuel consumption and the change in batteries' SOC. Utilizing the mixed-integer programming approach, an optimal scheduling for the power output of the engines and optimal charging/ discharging rate of the batteries are determined, while accounting for the constraints to rated power limits of engine/ batteries and batteries' SOC limits. It should be noted that the proposed optimal algorithm can schedule the operation, i.e. starting time and stopping time, for a multi-engine configuration, which is a key difference from the previously developed optimal power management algorithms for land-based hybrid electric vehicles. It is shown that when the load profile is known a priori, an optimum solution regarding the engine/ batteries power output and engine operation schedule can be obtained. Numerical illustration is presented on an industry-consulted harbor tugboat model to show the feasibility and effectiveness of the proposed algorithms. The simulation results demonstrate that the optimal cost for electric tugboat operation is $9.31\%$ lower than the baseline rule-based controller. In case of load uncertainty, the prediction based algorithm yields a cost $8.90\%$ lower than the baseline rule-based controller. The original contribution of the thesis include: (1) Formulation of the design optimization problem for the power generation. (2) Formulation of control optimization problem for power generation. (3) Modeling of the power distribution system in electrical ships. Formulation and modelling done in this dissertation lay the groundwork for better conceptualization and measurement of cost problems associated with the complex problems in controls and design of marine vehicle. In this research, an application of modern optimization technique on industrial design is done, giving designers an opportunity to observe and potentially benefit from such application. The limitations and future works of this dissertation can be described as: (1) Limited parameter data for the design optimization selection. A more robust selection can be achieved with more input parametric data given. (2) Rule based controller in design optimization simulated model may not be the most optimal choice. Implementation of an optimized adaptive controller, that can respond to changes in size of diesel generators and batteries and still give optimal control solution, could give a better design selection. (3) Load profile considered in this dissertation have only taken account of an average data from a single source. Dependent on geography, ship operating profile differs location to location. Load profile studies are highly invaluable data and collection of such data requires government/industrial/academic collaboration. (4) Weightage in costs functions have been arbitrarily chosen. The selection of weight values for the costs function in this dissertation are largely dependent on the designer's choice of importance. Additional studies can explore the pareto surface search approach and could result in better weightage parameter selection. (5) Training of neural network with more data. Strength of neural network is highly dependent on sample size of the training data. Future works could focus on improving practical implementation of control scheme using neural network with a more rigorous training methods. (6) Lastly, the design and control optimization methodology proposed in this dissertation is highly versatile and could be explored further by implementation in other power distribution systems.