A double deep reinforcement learning-based adaptive framework for decision-optimal wind power interval prediction

Prediction intervals (PI) effectively quantify forecasting uncertainty and serve as inputs for subsequent decision-making tasks. While it is traditionally assumed that reducing prediction errors will correspondingly reduce decision errors, this assumption is not invariably valid. This paper introduces an adaptive decision-optimal framework for optimal interval forecasting in wind power, designed to alleviate the economic dispatch challenges posed by wind power uncertainty within power systems. Methodologically, this framework employs a close-loop method based on the Double Deep Q-Network algorithm, where the forecasting module leverages a pre-trained model with Bi-Directional Long Short-Term Attention to enhance extracting features of historical data and increase quantile forecast precision. Then, Double Deep Q-Network can select decision-optimal quantile proportions. The validity of the framework is demonstrated through experiments utilizing real-world wind power data from the Belgian Elia company, validated across IEEE 6-bus and 30-bus cases. The method decreased average operation cost and risk by 0.36%/4.38% in the IEEE 6-bus and 0.76%/6.29% in the IEEE 30-bus compared to benchmarks. This framework offers a robust solution for wind power probability forecasting and the decision of power systems, thereby enhancing the stability and economic efficiency of power system operations.