Hidden Markov Models for wind farm power output

Debarati Bhaumik, Daan Crommelin, Stella Kapodistria, Bert Zwart

Research output: Contribution to journalArticleAcademicpeer-review

14 Citations (Scopus)
314 Downloads (Pure)


The reliability of the transmission grid is challenged by the integration of intermittent renewable energy sources into the grid. For model-based reliability studies, it is important to have suitable models available of renewable energy sources like wind and solar power. In this study, we investigate to what extent the power output of wind farms can be modeled with discrete Hidden Markov Models (HMMs). The parameters of the HMMs are inferred from measurement data from multiple turbines in a wind farm. We use these models both for individual turbine output and for total aggregated power output of multiple turbines. When modeling individual turbine output, the hidden process in the HMM is instrumental in capturing the dependencies between the output of the different turbines. It is important to account for these dependencies in order to correctly capture the upper quantiles (90%, 95%, 99%) of the distribution of the wind farm aggregated power output. We show that despite their simple structure, HMMs are able to reproduce important features of the power output of wind farms. This opens up possibilities to model and analyze these features with methods and techniques stemming from the field of Markov models and stochastic processes.

Original languageEnglish
Article number8356139
Pages (from-to)533-539
Number of pages7
JournalIEEE Transactions on Sustainable Energy
Issue number2
Publication statusPublished - Apr 2019


  • data models
  • Hidden Markov models
  • hidden Markov models
  • power system modeling
  • Power system reliability
  • Stochastic processes
  • time series analysis
  • Wind farm power generation
  • Wind farms
  • Wind power generation
  • Wind speed
  • Wind turbines


Dive into the research topics of 'Hidden Markov Models for wind farm power output'. Together they form a unique fingerprint.

Cite this