Track: Doctoral Dissertation Competition

Abstract

With the energy transition underway, there is a consensus effort from both worldwide governments and industry, to facilitate the integration of more renewable energy into power grids. However, the main challenge with integrating renewables like wind and solar, is the intermittent nature of these sources that result in inefficiencies and a lack of reliability. One possible energy system that can allow for an effective integration with intermittent sources is a polygeneration energy system (PES). The concept of a PES has been proposed in academic literature in recent times, where a typical pathway is the use of excess power from a fossil power plant being directed towards the production of valuable liquid fuels, in addition to electricity to the grid. In this study, the concept of a polygeneration system has been modified and extended to include renewable energy generation from wind, and a chemical production pathway of methanol to act as a form of long duration storage for times when the wind generation is much higher for any load demand and therefore increasing the flexibility.  

The approach taken in this model development departs from previous polygeneration modeling studies, by using a network constrained Unit Commitment (UC) model that is well established in the domain of power systems engineering, to optimally schedule the power planning and power flow. First, the design and operation of a power generation planning model is developed to showcase how the power system responds to the intermittency of wind in the form of wind scenarios. Mixed-Integer Linear Programming (MILP) models are then developed for the chemical production of methanol and integrated with the power planning model as a multi-scale (design and operation) model of a renewable polygeneration energy system (RPES) with chemical storage. To showcase the design and operation of the proposed RPES, the model is first solved in a deterministic manner.  The total RPES system cost, based on real world wind power data and load demand data, was found to be USD 2317.93 million. The chemical production block had a cost of USD 138.51 million when integrated as a part of the RPES and the power generation planning block had a cost of USD 2179.42 million. The integrated model resulted in costs for the chemical production block that were much lower than the stand-alone plant while the RPES model also showed how excess intermittent wind power could be used for driving the chemical production. A key contribution to this work is also the implementation of machine learning methods, like K-Means clustering to help with the model’s solution tractability and representation of a full year’s hourly wind data and load demand. The MILP models have been developed using the General Algebraic Modeling System (GAMS) software and solved using state of the art optimization solvers BARON and CPLEX.