Less Information, Similar Performance: Comparing Machine Learning-Based Time Series of Wind Power Generation to Renewables.ninja

Johann Baumgartner,Yves-Marie Saint-Drenan,Katharina Gruber,Sofia G Simoes,Johannes Schmidt

doi:10.3390/en13092277

Johann Baumgartner, Yves-Marie Saint-Drenan + Show 3 more

Open Access

https://doi.org/10.3390/en13092277

Copy DOI

Abstract

Driven by climatic processes, wind power generation is inherently variable. Long-term simulated wind power time series are therefore an essential component for understanding the temporal availability of wind power and its integration into future renewable energy systems. In the recent past, mainly power curve-based models such as Renewables.ninja (RN) have been used for deriving synthetic time series for wind power generation, despite their need for accurate location information and bias correction, as well as their insufficient replication of extreme events and short-term power ramps. In this paper, we assessed how time series generated by machine learning models (MLMs) compare to RN in terms of their ability to replicate the characteristics of observed nationally aggregated wind power generation for Germany. Hence, we applied neural networks to one wind speed input dataset derived from MERRA2 reanalysis with no location information and two with additional location information. The resulting time series and RN time series were compared with actual generation. All MLM time series feature an equal or even better time series quality than RN, depending on the characteristics considered. We conclude that MLM models show a similar performance to RN, even when information on turbine locations and turbine types is unavailable.

Highlights

The installed wind power capacity has increased more than six-fold within eleven years, from 93.9 GW in 2007 to 591.5.8 GW in 2018
We present results of MLM1 and MLM2 in comparison to RN
We used three datasets—one without location information, and two with implicit location information via incremental climate data grid point subsetting—as the input for machine learning models (MLMs) to assess whether location information is needed to obtain a time series quality comparable to that of RN

Summary

Introduction

The installed wind power capacity has increased more than six-fold within eleven years, from 93.9 GW in 2007 to 591.5.8 GW in 2018. 22.2 GW in 2007 to 53.2 GW in 2018 This increase resulted in a power generation of 111.5 TWh in 2018, corresponding to 21% of the electricity demand in Germany [1,2,3,4]. Due to this significant expansion, the spatial and temporal availability of climate-dependent wind resources increasingly affects the whole power system. Accurate multi-year generation time series (i.e., multiple years of temporally highly-resolved values) are used as input data for power system models (e.g., [8,9])

Methods

Results

Conclusion