Deep Neural Networks (DNN) for Day-Ahead Electricity Price Markets

Abstract

This work investigates the application of a multilayered Perceptron (MLP) deep neural network for the day-ahead price forecast of the Iberian electricity market (MIBEL) which serves the mainland areas of the Spain and Portugal. The 3-month and 6-month period of price and energy data are treated as a historical dataset to train and predict the price for day-ahead markets. The network structure is implemented using Google's machine learning TensorFlow platform. Activation function such as Rectifier Linear Unit (ReLU) was tested to achieve a better Mean Absolute Percentage Error (MAPE). Three different layers (2, 3, and 4) were tested to understand the behavior of the model. Three different sets of variables (17, 4, 2) were used and variable selection approaches were used to discard irrelevant variables.