Designing power scheduling algorithms for electric vehicles and energy storage systems in bi-directional markets using mixed-integer programming

Abstract

Optimal battery scheduling for electric vehicles and energy storage systems when cooperating with renewable energy generation in behind-the-meter applications is studied in this thesis in the framework of Mixed-Integer Programming (MIP). High capability in obtaining global optima in optimization problems has made MIP a popular tool in smart grid research and particularly for battery scheduling problems. One important issue with regards to the battery cycling, is the battery degradation which could complicate the Unit Commitment (UC) models. This is because the battery wear model can be nonlinear and difficult to be incorporated into a UC problem. To address the existing research gap, in this thesis a battery degradation model has been introduced to be incorporated into a short-term MIP battery scheduling model to estimate the capacity loss of a Liion battery caused by irregular charging and discharging events.

Hence, a MIP UC model is developed in this work which incorporates the introduced battery wear model. Based on that, the UC problem for a home-based microgrid is investigated and different UC strategies have been presented to minimize the operation cost as well as the capacity loss of batteries and the carbon footprint for a home equipped with a smart residential microgrid. The impact of the resolution of a home UC model on the capacity loss of batteries is another studied subject in this work, where hourly and intra-hourly granularities are compared in terms of the battery aging. A Controller-Hardware-in-the-Loop (C-HIL) setup is developed to measure the performance of the UC strategies as well as the battery degradation rates. Optimal battery scheduling in applications with multiple beneficiaries such as workplaces with electric vehicle (EV) charging stations is also investigated in this work by studying different UC strategies that take into account the interests of system operators and EV users to different extents. The results of this work show that the presented MIP UC models, that incorporate the introduced battery wear model, can be solved for real optimums in different smart grid applications.