Proyección energética de la central eólica Huascachaca de 50 MW aplicando redes neuronales

Abstract

This paper deals with the prediction of electricity generation at the Huascachaca wind farm. The objective is to develop an energy projection model using neural networks to improve the energy dispatch planning of the wind farm. Two prediction tools are used, long term memory neural networks with short term units (LSTM) and convolutional neural networks (CNN). The method identifies and selects the most suitable energy forecasting model for the wind power plant. The Mean Absolute Percentage Error is significantly lower in the LSTM network compared to the CNN network. This study is presented as a valid tool for decision making in the management of electricity generation in wind farms.