GTC: Diffusion on the Clouds
Here you will find all the relevant information to get you started on training a diffusion model for solar energy forecasting
Created on March 9|Last edited on December 6
Comment
The massive deployment of Renewable Energy Sources (RES) is crucial in achieving carbon neutrality by 2050 and limiting global warming to +1.5°C. These low-carbon energy sources are becoming more cost-competitive with fossil fuels that emit greenhouse gases. However, their reliance on weather conditions presents significant challenges across the energy value chain.
TLDR: The code to train the model is here 👈
The GTC session
Steadysun: The solar energy forecasting company
Founded in 2013 as a spin-off of INES (French National Solar Energy Institute), Steadysun is a French company that offers SaaS solutions for delivering reliable and actionable information via API to foster the integration of RES into electrical systems while reducing the risks and the costs associated with weather variability. The team comprises 20 highly skilled PhDs and engineers delivering expertise in Weather/Climate, Renewable Energy, AI, and Web Services.
Steadysun is a member of NVIDIA Inception, a global program designed to help startups evolve faster through access to cutting-edge technology and NVIDIA experts.
💡
Know more about Steadysun
The Dataset
You can grab the dataset for training the Cloud Diffusion model with the following script:
"Downloads dataset from wandb artifact"DATASET_ARTIFACT_SMALL = 'capecape/gtc/np_dataset:v0'DATASET_ARTIFACT_LARGE = 'capecape/gtc/np_dataset:V1'. #twice the sequences by combining 2 bands.with wandb.init(job_type="download_dataset"):artifact = wandb.use_artifact(DATASET_ARTIFACT, type='dataset')artifact_dir = artifact.download()
Understanding Cloud Diffusion
Diffusion on the Clouds: Short-term solar energy forecasting with Diffusion Models
Using diffusion models to predict cloud movement on satellite imagery to forecast solar energy production
Next Frame Prediction Using Diffusion: The fastai Approach
In this article, we look at how to use diffusion models to predict the next frame on a sequence of images, and we iterate fast over the MovingMNIST dataset.
Using Stable Diffusion VAE to encode satellite images
We take the pretrained Stable Diffusion Variable Auto Encoder (VAE) to represent satellite imagery into latent space.
How To Train a Conditional Diffusion Model From Scratch
In this article, we look at how to train a conditional diffusion model and find out what you can learn by doing so, using W&B to log and track our experiments.
Code
The Speakers
- Thomas Capelle is a machine learning engineer at Weights & Biases, working on the Growth Team. He’s a contributor to fastai library and a maintainer of wandb/examples repository. He focuses on MLOps, wandb applications in industry, and fun deep learning in general. Previously, Thomas used deep learning to solve short-term forecasting for solar energy at SteadySun. He has a background in urban planning, combinatorial optimization, transportation economics, and applied math.
- Having graduated from ISAE-Supaero, the French engineering school for aeronautics and aerospace, I wanted to help protect the environment with the Earth observation knowledge I gathered. That is why I have been working for two years at Steadysun as an R&D engineer for the satellite segment of the forecasting system. Here at Steadysun, we develop products and offer services to facilitate the integration of solar energy into electrical systems while mitigating the costs and risks associated with weather variability. Thus, we support all electricity value chain players throughout their projects' life cycles. The company has been designing solutions and developing innovative products to accelerate the energy transition since 2013.
Add a comment