Automatic learning for the system identification. A case study in the prediction of power generation in a wind farm

Abstract

One of the greatest technical challenges of today is obtaining predictive models for complex systems. In this paper we propose using data collected during a process to identify said process by means of automatic learning algorithms. Specifically, we describe the development of a project to determine the predictive model of a system based on supervised automatic learning algorithms. As an example, we use the problem of determining the energy generated in a wind farm. We do so by studying how the data collected are transformed, the search for the best algorithm, how to determine its goodness, and finally, the training and adjustment of the selected model. This study relies on the Python programming language, which has libraries that facilitate this type of project, and the Jupyter Notebook environment to carry out the project and disseminate the results.