DEVELOPMENT OF A WIND FORECAST MODEL TO THE BRAZILIAN NUCLEAR POWER PLANTS SITE BASED ON LSTM NEURAL NETWORK

Resumo

Meteorological conditions are key to estimate atmospheric dispersion of fission products during nuclear accidents involving nuclear material release. This work presents the latest investigations, by IEN researchers, on the development of a wind forecast model for the Brazilian nuclear power plants (NPP) site. Here we present the first investigation on the use of Long Short-Term Memory (LSTM) neural networks (NN) architecture to forecast wind velocity in a meteorological station at plant site. Results indicate that LSTM is promising in forecasting meteorological parameters in a realistic scenario.