Predicting Locational Based Marginal Pricing (LBMP) Using Ridge Regression

This paper evaluates the performance of the Ridge Regression Model for prediction of Locational Based Marginal Pricing (LBMP) of New York City. 2020–2021 New York Independent System Operator (NYISO) Time Weighted/Integrated Real Time LBMP Zonal data was used to train and test the Ridge Regression model. Two sets of hourly LBMP predictions were generated (same day and day ahead) using hourly Marginal Cost Losses (MCL) and Marginal Cost Congestion (MCC) values. Predictions are compared using Day-Ahead Market LBMP Zonal data. Notably, only Scikit-learn preprocessing Standard Scaler was used to normalize the features and the Ridge Regression alpha parameter was set to 4.0. The overall test set results indicate that the Ridge Regression model referred to as “RRMP”, achieved a training Mean Absolute Percentage Error (MAPE) of 3.92, Mean Absolute Error (MAE) of 1.11, and Mean Squared Error (MSE) of 2.02, demonstrating effectiveness in predicting LBMP. This study contributes to the understanding of machine learning models in energy market forecasting, in this case the context of LBMP prediction for New York City.