Evaluating LMP Forecasting with LSTM Networks: A Deep Learning Approach to Analyzing Electricity Prices During Unpredictable Events

Abstract

The unpredictable events can significantly impact energy demand and supply in the electricity market, leading to price volatility. This study aims to evaluate the effectiveness of Long Short Term Memory (LSTM) approach in analyzing real-time data on Locational Marginal Prices (LMPs) during periods before, during, and after the COVID-19 pandemic. Open data from the Midcontinent Independent System Operator (MISO) are utilized to obtain the LMP data. To evaluate the accuracy of the model predictions, three performance metrics were utilized, namely Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and coefficient of determination (R-2). Additionally, the study assesses the ability of LSTM to forecast LMP, considering yearly fluctuations. Graphical visualizations are created to depict the trends and patterns of LMP changes and forecasts over time. The results demonstrate the promising potential of LSTM in forecasting LMP even in unpredictable situations like pandemic. Despite the challenges of accurately estimating extreme energy demands during the pandemic, the LSTM model generates reliable forecasts, as evidenced by the performance metrics. The graphical visualizations also illustrate the effectiveness of LSTM in capturing the underlying trends and patterns of LMP changes over time.