ExARNN: An Environment-Driven Adaptive RNN for Learning Non-Stationary Power Dynamics
Haoran Li, Muhao Guo, Yang Weng, Marija Ilic, Guangchun Ruan
TL;DR
The paper addresses the challenge of forecasting non-stationary power system dynamics shaped by renewable variability, weather, and demand. It introduces ExARNN, an environment-driven adaptive RNN that uses a hypernetwork to generate environment-conditioned parameters and Neural Controlled Differential Equations (NCDE) to fuse external data into a continuous feature flow, enabling parameter adaptation at irregular timestamps. By treating external information as meta-knowledge and employing NCDE for continuous integration, ExARNN achieves high sample efficiency and generalization, demonstrated on Spain and Texas datasets where it outperforms baselines such as Vanilla RNN, RNN-$\Delta t$, ODE-RNN, and NCDE. The results show ExARNN delivers superior forecasting accuracy with competitive computational cost, and the authors plan to extend the framework to probabilistic settings for broader dynamic learning tasks in power systems.
Abstract
Non-stationary power system dynamics, influenced by renewable energy variability, evolving demand patterns, and climate change, are becoming increasingly complex. Accurately capturing these dynamics requires a model capable of adapting to environmental factors. Traditional models, including Recurrent Neural Networks (RNNs), lack efficient mechanisms to encode external factors, such as time or environmental data, for dynamic adaptation. To address this, we propose the External Adaptive RNN (ExARNN), a novel framework that integrates external data (e.g., weather, time) to continuously adjust the parameters of a base RNN. ExARNN achieves this through a hierarchical hypernetwork design, using Neural Controlled Differential Equations (NCDE) to process external data and generate RNN parameters adaptively. This approach enables ExARNN to handle inconsistent timestamps between power and external measurements, ensuring continuous adaptation. Extensive forecasting tests demonstrate ExARNN's superiority over established baseline models.