Spatially Explicit Prediction of Wholesale Electricity Prices

Abstract

Transmission constraints often limit the flow of electricity in a regional transmission network leading to strong interaction effects across different geographically distributed points within the system. In modern wholesale electricity markets, these transmission constraints lead to spatial patterns within the nodal electricity spot prices. This study exploits these spatial patterns to better predict spot prices within a wholesale electricity market. More specifically, we use the latest spatial panel data econometric models to compare within-sample and out-of-sample forecasts against nonspatial panel data models. The spatial panel data approach is explained by demonstrating a simple network optimization model. We find that a dynamic, spatial panel data model provides the best predictions within a forecasting error context. Our results may suggest that the spatial autocorrelation between node prices extends beyond the current market-defined zonal boundaries, which calls into question whether the zonal boundaries accurately reflect the congestion boundaries within the system.