<p>Accurate oil temperature prediction is essential for real-time monitoring and maintenance of wind turbine gearboxes. Given the randomness and chaotic behavior of oil temperature data due to complex environmental impacts, a novel hybrid prediction model is proposed integrating triangulation topology aggregation optimizer (TTAO), variational mode decomposition (VMD), fuzzy entropy (FE), phase space reconstruction (PSR), gated recurrent unit (GRU), and Informer. The raw temperature series is first decomposed into multi-frequency subsequences using TTAO-optimized VMD (TVMD). These subsequences are then reconstructed using FE-based hierarchical clustering and classified using the maximum Lyapunov exponent. For chaotic sequences, PSR is applied to uncover their underlying dynamics. For non-chaotic sequences, relevant lag terms are selected to enhance predictive accuracy. Finally, the processed reconstructed sequences are input to the GRU-Informer model to predict the next time step, and the final oil temperature prediction results are obtained after the superposition of all component prediction results. Empirical analysis demonstrates that the hybrid model proposed in this paper exhibits superior predictive performance. Compared with the traditional neural network, the <i>R</i> square (<InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="11227_2025_7438_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="21" /> </InlineMediaObject> <EquationSource Format="TEX">${R}^2$</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mrow> <mi>R</mi> </mrow> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation>) of the proposed model increases by 25.51%, the root mean square error decreases by 82.99% and the mean absolute error decreases by 84.3%. The superiority of the proposed method is further validated through the Diebold–Mariano test, which assesses its performance from a statistical perspective. Lastly, the discussion section demonstrates the necessity of proper parameter selection, comparing the model’s complexity and operational efficiency, as well as the approach for industrial applications of the model. This research offers a novel high-precision method and a new strategy for predicting gearbox oil temperature. This research offers a promising and robust approach for wind turbine gearbox oil temperature forecasting.</p>

The hybrid prediction of wind turbine gearbox oil temperature based on chaotic theory

期刊论文
发表日期：

2025年5月29日

查看全文

Cheng Huang,
Shaojuan Ma,
Changlin Xu,
Xinyi Xu

摘要

Accurate oil temperature prediction is essential for real-time monitoring and maintenance of wind turbine gearboxes. Given the randomness and chaotic behavior of oil temperature data due to complex environmental impacts, a novel hybrid prediction model is proposed integrating triangulation topology aggregation optimizer (TTAO), variational mode decomposition (VMD), fuzzy entropy (FE), phase space reconstruction (PSR), gated recurrent unit (GRU), and Informer. The raw temperature series is first decomposed into multi-frequency subsequences using TTAO-optimized VMD (TVMD). These subsequences are then reconstructed using FE-based hierarchical clustering and classified using the maximum Lyapunov exponent. For chaotic sequences, PSR is applied to uncover their underlying dynamics. For non-chaotic sequences, relevant lag terms are selected to enhance predictive accuracy. Finally, the processed reconstructed sequences are input to the GRU-Informer model to predict the next time step, and the final oil temperature prediction results are obtained after the superposition of all component prediction results. Empirical analysis demonstrates that the hybrid model proposed in this paper exhibits superior predictive performance. Compared with the traditional neural network, the R square ( ${R}^2$ $R^{2}$ ) of the proposed model increases by 25.51%, the root mean square error decreases by 82.99% and the mean absolute error decreases by 84.3%. The superiority of the proposed method is further validated through the Diebold–Mariano test, which assesses its performance from a statistical perspective. Lastly, the discussion section demonstrates the necessity of proper parameter selection, comparing the model’s complexity and operational efficiency, as well as the approach for industrial applications of the model. This research offers a novel high-precision method and a new strategy for predicting gearbox oil temperature. This research offers a promising and robust approach for wind turbine gearbox oil temperature forecasting.