Optimal scheduling of energy systems incorporating load management schemes

The advent of enabling smart grid technologies has resulted in the proliferation of heterogeneous power generation networks. In this context, the concept of microgrids has gained popularity in recent years due to their ability to integrate renewable energy sources with the power system. As such, many industrial units are increasingly displaying characteristics similar to grid-connected microgrids. Consequently, the traditional day-ahead scheduling (unit commitment) problem solved in power systems needs to account for the increasingly heterogeneous nature of the generators. Furthermore, deregulated electricity market concepts such as load management need to be incorporated in the scheduling problem. As such, there exists a need to formulate optimization models for modern energy systems which can account for the heterogeneity in the generation and the flexibility in the load. This thesis is broadly divided into four parts. The first part develops accurate scheduling models of the components which constitute the energy systems considered in the later chapters of the thesis. The mixed logical dynamical modelling framework is used to develop scheduling models of the gas turbines, steam turbines, boilers, diesel generators, battery energy storage systems, thermal energy storage systems and interruptible loads. The scheduling models of the gas turbines, the steam turbines and the boilers include the power trajectories followed by these components while undergoing the hot, warm and cold start-up processes. A detailed treatment of the modelling of an exemplar conventional fossil fuel based generating unit using the mixed logical dynamical framework is also provided. The second part of this thesis proposes a shipyard energy management system (SEMS) to optimize the cost of operating a typical shipyard drydock. The SEMS comprises three modules load forecasting, contracted capacity optimization and optimal scheduling. The load forecasting module uses artificial neural networks (ANN) to generate short term and medium term load forecasts. Historical load demand data and ship arrival schedules are provided as inputs to the ANN. The inclusion of the ship arrival schedule as an input to the ANN enhances the accuracy ix