Research code for the paper "Unsupervised Deep Learning Approach for Scenario Forecasts" accepted to Power Systems Computation Conference (PSCC) 2018, which is also free to view on ArXiv: https://arxiv.org/pdf/1711.02247.pdf
Authors: Yize Chen, Xiyu Wang and Baosen Zhang
The integration of high penetration of renewable generation into power systems calls for a growing need to model the uncertain and intermittent characteristics of these resources. An important method used in characterizing the behavior of renewable resources is scenario generation, where a set of possible future realizations are provided for the system operator. Compared to deterministic point forecasts or probabilistic forecasts, scenario forecasts could not only inform users of the uncertainty about the future, but also reflect the temporal dependence of renewable power generation. This is a practical setting for applying GAN to scenario generaion.
We combine the Generative Adversarial Networks (GAN) for scenario generation, as well as an optimization step based on specific-designed loss to make forecasts based on historical observations.
For more information about code and methods, please feel free to contact Yize Chen: yizechen@uw.edu
There is a Github limit on the size of data. I haven't uploaded the forecast data. If you have any inquires on that particular dataset, please feel free to ask me.
(Update December, 2019) I have included my processed original dataset and forecast data compatible with the code in this repo in google drive. You can download the original data and forecast data.