Combining Machine Learning and Optimization for Efficient Price Forecasting
Original version
10.1109/EEM49802.2020.9221968Abstract
We present a framework based on machine learning for reducing the problem size of a short-term hydrothermal scheduling optimization model applied for price forecasting. The general idea is to reduce the optimization problem dimensions by finding patterns in input data, and without compromising the solution quality. The framework was tested on a data description of the Northern European power system, demonstrating significant reductions in computation times.