Comparing the carbon costs and benefits of low-resource solar nowcasting

TL;DR

This paper evaluates low-resource solar PV yield nowcasting methods using UK satellite data, highlighting their potential for significant carbon savings despite small model sizes, and analyzing model errors and CNN activations.

Contribution

It compares multiple low-resource nowcasting approaches and estimates their carbon impact, demonstrating that small models can provide substantial environmental benefits.

Findings

Small models can save more carbon than their own emissions

Low-resource approaches perform effectively over 1-4 hour forecasts

Analysis of CNN activations reveals insights into model predictions

Abstract

Solar PV yield nowcasting is used to help anticipate peaks and troughs in demand to support grid integration. This paper compares multiple low-resource approaches to nowcasting solar PV yield, using a dataset of UK satellite imagery and solar PV energy readings over a 1 to 4-hour time range. The paper also estimates the carbon emissions generated and averted by deploying models, and finds that even small models that could be deployable in low-resource settings may have a benefit several orders of magnitude greater than its carbon cost. The paper also examines prediction errors and the activations in a CNN.