Hourly Electricity Price Prediction Applying Deep Learning for Electricity Market Management

The provision of accurate electricity price forecast is crucial for all electricity market players to encourage power generation and consumption decisionmaking. Since electricity price time sequence behavior is highly non-linear and seasonal, deep neural networks are the best model for learning non-linear data behavior for the purpose of prediction. This paper utilizes a Long-short term memory (LSTM) neural network-based model for forecasting the electricity settlement price which can successfully identity nonlinear behavior within the input data. The proposed model uses an Adam optimizer for less memory usage. ERCOT wholesale market price data along with load and temperature data have been used to illustrate the effectiveness of the proposed model. Model performance is tested, and the electricity price forecasting model is verified and validated.