Robust wind speed estimation with modified fuzzy regression functions with a noise cluster

Chakravarty S., Demirhan H., BAŞER F.

Energy Conversion and Management, cilt.266, 2022 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 266
  • Basım Tarihi: 2022
  • Doi Numarası: 10.1016/j.enconman.2022.115815
  • Dergi Adı: Energy Conversion and Management
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, PASCAL, Aerospace Database, Applied Science & Technology Source, CAB Abstracts, Communication Abstracts, Computer & Applied Sciences, Environment Index, INSPEC, Pollution Abstracts, Veterinary Science Database, Civil Engineering Abstracts
  • Anahtar Kelimeler: Deep neural networks, Fuzzy regression functions, Noise cluster, Outlier, Possibilistic clustering, Robustness, Support vector machines, Wind energy, DETERMINING WEIBULL PARAMETERS, NUMERICAL-METHODS, NORTHEAST REGION, MODELS, DISTRIBUTIONS, PREDICTION, SITES
  • Ankara Üniversitesi Adresli: Evet

Özet

© 2022 Elsevier LtdThe wind is gaining prominence as a renewable energy source as the climate change burden worsens. One of the main components of wind is wind speed, which has many outlier observations that differ significantly from the other measurements. Robust methods are needed to reliably analyse wind speed data by minimising the impact of outliers. The modified fuzzy regression functions framework (MFRFN) using a noise cluster is shown to be robust to the presence of outliers. The current study employs the MFRFN framework for robust wind speed estimation. It compares the estimation accuracy of the MFRFN framework to deep neural networks (DNNs) and support vector machines (SVMs) for wind speed estimation with an extensive dataset including more than 12,000 on-shore and off-shore locations for each month and at the annual level. Then, the MFRFN framework is implemented to produce wind speed estimates at 50 m, and annual, and monthly robust global wind speed potential maps are created. The best monthly and annual MFRFN models are also validated using out-of-sample ground measurements from 572 locations. Consequently, the MFRFN framework performs better than DNNs and SVMs for wind speed estimation and provides us with a robust estimation method for wind speed potential at a given location over the globe.