Predicción de la velocidad del viento mediante redes neuronales, para la Estación Meteorológica Chimborazo.

Abstract

The need to increase the contribution of non-conventional renewable energies, within the productive matrix in Ecuador, is increasing. This context has prompted several public and private institutions to undertake wind power projects, among which the Chimborazo Polytechnic School has undertaken the wind power generation project, installing meteorological stations in different areas of the province of Chimborazo, with the aim to analyze the viability of wind power generation. However, the boom in renewable energy sources in recent years has forced a new forecasting problem, namely, to know in sufficient time the energy that could be generated by future wind farms. In this context, the need arises to have tools that are capable of predicting the random behavior of wind speed, with a certain degree of reliability, in order to predict how much energy can be delivered to the system at a given time. The applied study responds to an exploratory research, with descriptive and explanatory elements, which includes the grouping of historical data, wind speed behavior analysis, through automatic learning techniques such as nonlinear, non-linear,

self-managing artificial neural networks, applied to a set of data measured hourly, for 12 consecutive months, of the Chimborazo meteorological station, property of ESPOCH. The results obtained showed that the use of RNA in the prediction process is very efficient with an automatic learning level of 8% and a performance level measured through the mean square error of 0.016 with a prediction horizon of seventy two hours, which will allow to predict with a high degree of reliability, the amount of wind energy to be.