Optimal Power Flow with Probabilistic Renewable Energy Forecasting for Multi-area Balancing Services

The increasing integration of renewable energy sources, such as wind and solar, introduces real-time uncertainties that complicate the balancing service. To deal with this dilemma, the paper presents a novel real-time Optimal Power Flow (OPF) method integrated with probabilistic renewable energy forecasting to balance the power system’s demand and supply. This study focuses on learning-based OPF methods that leverage the approximation capabilities of neural networks to map from load to OPF solutions for real-time operation settings. However, these learning-based methods may not satisfy power flow equations and inequality constraints. To enhance the feasibility of solutions, this paper introduces a method that utilizes a Recurrent Neural Network (CNN) for short-term probabilistic renewable energy forecasting, providing the upper limit of renewable energy for the next operational step. Subsequently, a neural network solver is trained to predict OPF solutions based on the load and forecasted renewable generations. Furthermore, a Graph Convolutional Network (GCN) is embedded within the neural network solver to address OPF problems with topology changes due to N-1 contingencies. The effectiveness of the proposed method is validated on the IEEE standard test system.