Code and dataset for "Seasonal hydropower forecast framework using a combination of deep learning and physics-based methods"

Dataset Description

This dataset contains plant-level, monthly seasonal hydropower generation estimates for the contiguous United States produced by the final (hydropower) step of an end-to-end forecasting workflow that combines deep learning and physics-based components. The workflow translates coarse-resolution NMME climate forecasts into 0.1° forcings, generates probabilistic runoff fields, routes and manages water with MOSART-WM, and converts regulated flows and storage into ensemble hydropower estimates using a plant-level autoregressive model. Forecasts are provided for 1,432 hydropower plants. 

What’s included

Primary product: Monthly ensemble hydropower generation estimates at plant level across CONUS.

Context fields: plant eia_id, year–month timestamp, ensemble member index

Scripts : 

Project Documentation=====================

This documentation describes the workflow for generating a CSV file that lists the power output from 1432 dams located in the USA. The process leverages sequentially executed Python scripts and utilizes North American Multi-Model Ensemble (NMME) forecast data alongside MOSARTwm for reservoir inflow generation.

1. Prerequisites-----------------

       Python Environment:      Ensure you have a working Python environment with the necessary dependencies installed. See the [Environment and Dependency Setup](environment-and-dependency-setup) section for details.       Workflow Sequence:      The workflow files are sequentially numbered from 01 to 011. They must be executed in order. For example:           First, run 01NMMEDownloadForecast.py and wait until it completes.       Then, run 02NMMEDownscaleprepapration.py, and so on.   File Descriptions:      Each file begins with a header that describes its purpose and main functions. Please review these comments within the files to better understand what each script accomplishes.    

2. Workflow Execution----------------------

 2.1 NMME Data Processing

   Downloading Forecast Data:      01NMMEDownloadForecast.py downloads forecast data using 10 ensemble members from the North American Multi-Model Ensemble (NMME).           Customization: You can adjust the number of ensemble members by modifying the configuration within this file.   Downscaling Preparation:      02NMMEDownscaleprepapration.py (and subsequent scripts) process and prepare the downloaded data sequentially.    

 2.2 Generating Reservoir Inflows with MOSARTwm

   Required Files:      The generation of reservoir inflows involves three key files:           08arunMosartwm.py       08brunmosartconfig.yaml       08crunMosartWMpy.sh   Workflow Process:      The provided example shows how to run MosartWM for a single ensemble member.           Note: To compute reservoir inflows for all ensemble members, ensure you update the configuration to include the appropriate ensemble settings.

 2.3 Final Output

   CSV Results:      Upon completion of the workflow, a CSV file is generated that lists the power output from the 1979 dams located in the USA.

3. Environment and Dependency Setup------------------------------------

 3.1 Generating requirements.txt from Your Python Environment

Use the requirements.txt to replicate the environment

    pip install -e requirements.txt