Multivariate Multi-step Forecasting for Cable Pooling Applications

The proliferation of renewable energy sources (RES) poses a significant challenge with respect to grid connection due to the existing infrastructure’s inability to keep pace with the rapid expansion of renewable energy generators. This has emerged as a bottleneck in the energy transition process. To overcome this challenge, cable pooling solutions have been proposed, which allow multiple RES plants with varying features to share a single connection and leverage existing grid infrastructure to expand their capacity rapidly. In such systems, coordination between energy supply and demand is critical, requiring consideration of power production forecasts to prevent overproduction and facilitate storage capacity planning. In this study, we propose the selection of an appropriate LSTM variant for multivariate multistep forecasting models that can be leveraged in cable pooling applications. Our results indicate that the bidirectional LSTM model outperforms other variants in multivariate multistep forecasting. However, this model’s duration of execution is comparatively longer than that of other models.