Applying an Encoder-Decoder LSTM Model for Short-Term Photovoltaic Power Forecasting in Subtropical Region

One of the pillars of technological development in modern society is electricity, with an ever growing demand f motivated by technological advances. Much of the power generation comes from fossil fuels; however, renewable sources have gained prominence due to the impacts of climate change and the global commitments aimed at reducing greenhouse gas emissions to control the rise in the planet’s temperature and its competitive economics. Solar energy is an essential alternative for the transition to renewable sources, but the intermittency of photovoltaic generation, associated with weather conditions, is one of the main challenges. For this reason, this work applied a methodology for forecasting photovoltaic generation for a horizon of 1 to 3 hours, based on neural networks using an architecture known as Encoder-Decoder Long Short-Term Memory (EDLSTM). This model is highly efficient for problems known as seq2seq, where one data sequence is used as input, and another is generated as output. Meteorological data and the historical power series of the PV power plant at the University of São Paulo were used for training. Data from 2018 to 2021 was used for training and validation, while data from 2022 was reserved for testing. Four groups of input attributes were considered, two with atmospheric variables and two with the historical power series. The model was evaluated using the Root Mean Square Error (RMSE), Mean Bias Error (MBE), and Mean Absolute Error (MAE). In addition, the Skill Score (SS) was used, which compares performance with a persistence model based on the clear sky index. As a result, the models that used power attributes with solar angles as input performed best, with SS ranging from 15.16% to 48.67%.