DACF

DACF: Day-ahead Carbon Intensity Forecasting of Power Grids using Machine Learning.
(Refer DACF.pdf for the paper)

We have built CarbonCast, which is an extension of DACF. CarbonCast is more up-to-date and covers more regions. Please refer to that repo for further information.

Version: 1.0
Authors: Diptyaroop Maji, Ramesh K Sitaraman, Prashant Shenoy
Affiliation: University of Massachusetts, Amherst

0. Citing DACF

If you use DACF, please consider citing our paper. The BibTex format is as follows:

       @inproceedings{maji2022dacf,
        title={DACF: Day-ahead Carbon Intensity Forecasting of Power Grids using Machine Learning},
        author={Maji, Diptyaroop and Sitaraman, Ramesh K and Shenoy, Prashant},
        booktitle={Proceedings of the Thirteenth ACM International Conference on Future Energy Systems},
        year={2022}
       }

1. Regions covered

• US:
  • California (CISO)
  • Pennsylvania-Jersey-Maryland Interconnection (PJM)
  • Texas (ERCOT)
  • New England (ISO-NE)
• Europe (European regions are monitored by ENTSOE):
  • Sweden
  • Germany

2. Data Sources

US ISO electricity generation by source: EIA hourly grid monitor

European regions electricity generation by source: ENTSOE

Weather forecasts: GFS weather forecast archive

Solar/Wind Forecasts:

• CISO: OASIS
• European regions: ENTSOE
• We currently do not have solar/wind forecasts for other regions, Hence, we generate them using ANN models along with pther source production forecasts.

3. Usage

3.1 Installing dependencies:

DACF requires Python 3, Keras & Tensorflow 2.x
Other required packages:

• Numpy, Pandas, MatplotLib, SKLearn, Pytz, Datetime

3.2 Getting source production forecasts:

For getting source production forecasts, run the following file:
                    python3 sourceProductionForecast.py <region> <source>
Example: python3 sourceProductionForecast.py CISO nat_gas
Regions: CISO, PJM, ERCO, ISNE, SE, DE
Sources: coal, nat_gas, oil, solar, wind, hydro, unknown, geothermal, biomass, nuclear

3.3 Calculating average carbon intensity:

We use the following formula for calculating both real-time and forecasted avg carbon intensity:
, where

CI_avg = Average carbon intensity (real-time or forecast) of a region
E_i = Electricity produced by source i, when we are calculating real-time avg. carbon intensity, & day-ahead predicted electricity produced by source i, when we are calculating day-ahead carbon intensity forecasts.
CR_i = Median operational (direct) carbon emission rate (also known as carbon emission factor) of source i.

To calculate carbon intensity, run the following file:
                    python3 carbonIntensityCalculator.py <region> <f/r> <num_sources>
Example: python3 carbonIntensityCalculator.py CISO r 8
Regions: CISO, PJM, ERCO, ISNE, SE, DE
f : forecast (based on source production forecasts), r : real-time (based on historical electricity production data)
No. of sources producting electricity: CISO: 8, PJM: 8, ERCO: 7, ISNE: 8, SE: 4, DE: 10

4. Developer mode

DACF is a working prototype. However, we understand that it still needs a lot of improvements. We will be updating the codebase periodically to add new things (features, regions, improved models etc.). In addition to that, we welcome users to suggest modifications to improve DACF and/or add new features or models to the existing codebase.

5. Acknowledgements

This work is part of the CarbonFirst project, supported by NSF grants 2105494, 2021693, and 2020888, and a grant from VMware.