Active Fault Diagnosis Method Based on Deep Forest and Its Industrial Application

Wind turbine equipment, as a key component in the process production control of thermal power plants, directly affects the power generation efficiency of the plant. Due to the non-linear and non-stationary nature of the signals acquired from wind turbine sensors, this paper presents a novel active f...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:2024 8th Asian Conference on Artificial Intelligence Technology (ACAIT) s. 241 - 246
Hlavní autoři: Li, Dong, He, Kaixun, Tao, Lili, Liu, Hanlong
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 08.11.2024
Témata:
active fault diagnosis
Borutashap
deep forest
Deep learning
Fault diagnosis
Feature extraction
Machine learning algorithms
pearson
Power generation
Random forests
Regression analysis
Tuning
Whale optimization algorithms
wind turbine equipment
Wind turbines
WOA
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Wind turbine equipment, as a key component in the process production control of thermal power plants, directly affects the power generation efficiency of the plant. Due to the non-linear and non-stationary nature of the signals acquired from wind turbine sensors, this paper presents a novel active fault diagnosis approach for wind turbine systems based on deep forest techniques. Initially, multi-dimensional time-domain features are extracted from the raw fault data. Simultaneously, important features are selected from multiple angles by integrating Pearson correlation coefficient and Borutashap algorithm, following the principles of active fault diagnosis. Additionally, the whale optimization algorithm (WOA) is utilised to search the optimal sinusoidal auxiliary signal parameter, thereby highlighting specific fault manifestations and improving the model's diagnostic capability. The deep forest, with its strong classification ability and noise resistance, overcomes the overfitting issue in shallow machine learning algorithms and the complex parameter tuning in deep learning. Then, the designed sinusoidal auxiliary signal is used as an additional input to the deep forest model, enabling the final fault classification outcomes. Finally, experiments are conducted using a wind turbine power transmission system fault simulator. The experimental results demonstrate an improvement in fault diagnosis accuracy, confirming the effectiveness of the proposed method.
DOI:10.1109/ACAIT63902.2024.11022011