Implementation of MUlti-SOurce deep neural Network (MUSONet) and SIngle-SOurce deep neural Network (SISONet) models proposed in the article "Simultaneous multi-step wind speed prediction on multiple farms using multi-task deep learning"
Title: Simultaneous multi-step wind speed prediction on multiple farms using multi-task deep learning
Authors: Rafael Ayllón-Gavilán, Antonio Manuel Gómez-Orellana, Víctor Manuel Vargas, David Guijo-Rubio, Jorge Pérez-Aracil, Luis Prieto-Godino, Sancho Salcedo-Sanz, Pedro Antonio Gutiérrez, and César Hervás-Martínez
Abstract: In this paper, we present a deep neural network architecture for multi-task wind speed prediction. Specifically, we propose a Multi-Farm Multi-Step (MFMS) model, which leverages information from different wind farms and prediction horizons to improve the wind speed prediction in all the sites considered. The main goal of this approach is improving the prediction at other wind farms and at longer prediction horizons. Thus, the proposed model is able to simultaneously predict the wind speed at three different prediction horizons (6h, 12h, and 24h), across three different wind farms located in Spain.} We also evaluate the performance of the presented methodology by considering three different activation functions for hidden neurons in the neural network: Sigmoid, relu, and elusplus2L. We will show how the proposed multi-farm approach even improves the performance of the single-farm counterpart for the longest prediction horizons (12h and 24h). In addition, the proposed multi-farm algorithm reduces by over 70% the number of parameters compared to three single-farm models (one per farm), resulting in a simpler solution for the problem addressed and requiring much lower computational resources.
Status: Currently under review in Integrated Computer-Aided Engineering journal