A model based approach to the control of diesel generators

© 2016 Dr. Timothy James BroomheadDiesel generators are found in a wide range of applications, including series hybrid electric vehicles and submarines. Diesel generators operate with application specific objectives, constraints and environmental disturbances. In a series hybrid electric vehicle, the diesel generator must fulfil power requests whilst also minimising fuel consumption. This objective is made challenging due to legislated emissions constraints, requiring advanced actuation of the diesel engine's air-path. In a submarine, the diesel generator must maximise power output, whilst keeping key system outputs constant to minimise detectability. Significant periodic disturbances on the engine makes this a challenging control problem. Economic Model Predictive Control (EMPC) represents a promising control solution for many diesel generator applications, given its ability to directly address system objectives and explicitly handle constraints. Despite EMPC's promise, practical techniques for robustness, which are critical for any control application, have not yet been integrated into EMPC theory. Further, existing EMPC theory does not consider periodic disturbances, necessary for a submarine application. This thesis investigates the development and application of EMPC for power tracking and submarine diesel generator applications. Throughout the investigation, a number of theoretical and practical contributions are made to the field of applied control. A physics-based model of a diesel generator is first presented, necessary to simulate and develop model based control strategies. Following calibration, the model is demonstrated to represent a diesel generator under both power tracking and submarine operating conditions. While the model is capable of representing a diesel generator over a wide range of operating conditions, it is too computationally complex for direct use in online model based control. A model reduction framework is therefore proposed, based on singular and regular perturbation techniques. The model reduction framework is applied to the diesel engine model, preparing it for use in online model based control. Subsequently, a robust EMPC scheme is proposed. The control scheme utilises a nonlinear cost function, chosen to directly address a plant's objectives, and a linear prediction model. The formulation permits known periodic disturbances to be included in the prediction model, making it suitable for the submarine application. Through the use of constraint tightening techniques, the EMPC guarantees continuous constraint adherence, despite bounded unknown disturbances. In a novel approach, constraint tightening is also employed to guarantee stability of the closed loop system, which is proven using Lyapunov techniques. Both stability and constraint adherence guarantees are made under assumptions which are consistent with the diesel generator applications considered. The reduced order model is incorporated into the proposed EMPC scheme, which is applied to both diesel generator applications. The developed control system is implemented on rapid prototyping hardware, and experimentally demonstrated on a diesel generator test bed. In the power tracking application, the diesel generator is shown to efficiently follow the requested power trajectory, while continuously satisfying emissions and safety constraints. In the submarine application, the developed control scheme is demonstrated to significantly reduce the impact of periodic disturbances in key outputs when compared to an industry standard controller. The effectiveness of the control strategy motivates an experimental study into the impact of different actuator capabilities and operational constraints on a submarine diesel generator