Abstract Summary
We present a smart charging bidding framework that submits an optimal bid to the day-ahead electricity market by harnessing the flexibility of electric vehicles. Its optimization objective is to minimize the charging costs of the total EV fleet. The bidding framework consists of a forecasting module for the day-ahead price and for the EV flexibility, and of a linear optimization module to compose the most optimal bid. We explore several regression and machine learning models as possibility for the forecast modules. In order to assess the composition of a most optimal operational bidding framework, we compare the outcome of the bidding module when fed with each of these models. We show that the combination of forecasting modules that creates on average the lowest charging costs, lead to savings of around 26%. These costs are calculated by by considering the day-ahead price and the imbalance price as a penalty for forecasting errors. We find that the most accurate self-contained forecasting modules, do not deliver the best over-all results when included in the bidding framework.