Control-oriented modeling, validation, and analysis of a natural gas engine architecture

In order to improve performance and meet increasingly tight emissions regulations, engine manufacturers must improve algorithms used to control the engine. One possible strategy is to utilize centralized control algorithms that take into account the coupled interactions between inputs and outputs. Implementing a centralized control strategy often requires some kind of dynamic model of the system, which is a primary motivation for modeling efforts in this thesis. In a methodical fashion, this thesis derives a control model for a natural gas engine architecture and validates this control model against reference data in simulation. Additionally, this thesis performs control-oriented analysis on a state-space model provided by Caterpillar to determine the engine’s suitability to decentralized control. Based on the results of the control-oriented analysis, the thesis demonstrates how a decentralized control framework can be implemented. The first study declares a set of seven state variables that characterize the operation of the engine. The engine of interest runs on natural gas and is used in power generation applications. Additionally, this study models all mass flow rates and power terms as functions of the selected state variables. These models are then validated against truth-reference data. This study also explicitly states the assumptions and simplifications that correspond to each of the models. The second study derives dynamic equations for each of the seven state variables via a first-principles approach. The dynamic state equations contain expressions for mass flow rates and power that were modeled in the first study. This study then numerically validates the entire state-space model by exercising control inputs from reference data on it. Together, the seven state equations effectively serve as a control model that can be used for controller synthesis. The goal of the first two studies is to demonstrate a procedure for obtaining a control model for an engine architecture, not to obtain a high-fidelity simulation model. The third study demonstrates control-oriented analysis on a state-space model provided by Caterpillar. The relative gain array (RGA) is used to show that the system is well-suited is for decentralized control. This study implements a decentralized control structure on the state-space model provided by Caterpillar and validates, in simulation, its ability to achieve reference tracking for desired outputs