Failures are a major problem for wind turbine electric systems, due to their harsh working environment and dispersed locations. Problem mitigation can be a possible method of failure prediction. Traditional failure prediction involves complex analysis of specific measurements which can be predictive of future failures. For no more than a decade, a methodology born in the field of artificial intelligence could have the potential for predictions that go beyond the analysis of the root causes of failures. It is called “deep learning” and it is the evolution of artificial neural networks, with an arbitrary number of layers of neurons. In our study, we developed a deep neural model that predicts wind turbine failures from limited SCADA data. The model is organized with two symmetric branches, one of which is always fed with a time series of wind speed, and the other is an operating variable of the wind turbine, such as active power. Each branch of the model has a stack of feed forward layers, converging into a single softmax layer which returns the probability of failure. We tested the model in a one-year simulation with 1000 machines, with promising results.

A Deep Neural Model for Wind Turbine Failure Prediction

Lucia Guerrisi
2021-01-01

Abstract

Failures are a major problem for wind turbine electric systems, due to their harsh working environment and dispersed locations. Problem mitigation can be a possible method of failure prediction. Traditional failure prediction involves complex analysis of specific measurements which can be predictive of future failures. For no more than a decade, a methodology born in the field of artificial intelligence could have the potential for predictions that go beyond the analysis of the root causes of failures. It is called “deep learning” and it is the evolution of artificial neural networks, with an arbitrary number of layers of neurons. In our study, we developed a deep neural model that predicts wind turbine failures from limited SCADA data. The model is organized with two symmetric branches, one of which is always fed with a time series of wind speed, and the other is an operating variable of the wind turbine, such as active power. Each branch of the model has a stack of feed forward layers, converging into a single softmax layer which returns the probability of failure. We tested the model in a one-year simulation with 1000 machines, with promising results.
File in questo prodotto:
Non ci sono file associati a questo prodotto.
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11570/3230579
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact