Optimising large-scale generation and transmission systems via neural network-based wind energy forecasting: an integrated approach applied to Argentina

Abstract

The global energy landscape is shifting towards cleaner sources, gradually reducing reliance on non-renewable fuels like coal, oil, and natural gas. Over the past two decades, the integration of renewable energy into power systems has grown significantly. However, the variability of these sources poses considerable challenges for their efficient incorporation. In response, this paper presents a hybrid model for scheduling and forecasting large-scale electricity generation in complex power systems. The proposed approach combines an optimization model for managing electricity generation from diverse sources (fossil and renewable) with a neural network-based wind power forecasting model trained on real historical data. The optimization framework builds on a pre-existing model, enhanced to integrate renewable energy predictions. This integration enables effective scheduling and management of variable renewable energy sources. The hybrid model is validated through a case study on the Argentine Electric Interconnected System (SADI). Results demonstrate that electricity generation can be forecast and scheduled with approximately 10% error over a monthly planning horizon, offering a robust and innovative solution for managing large-scale power systems with significant renewable energy penetration.