Generalizable Temperature Nowcasting with Physics-Constrained RNNs for Predictive Maintenance of Wind Turbine Components

TL;DR

This paper presents a physics-constrained RNN approach for wind turbine temperature nowcasting that improves generalization in predictive maintenance, especially with limited data, by incorporating partial system knowledge into neural networks.

Contribution

It introduces a simple, efficient physics-constrained deep learning method for wind turbine temperature prediction using partial system information, enhancing generalization.

Findings

Improved generalization to unseen environments.

Better performance in low data scenarios.

Effective integration of physics knowledge into neural networks.

Abstract

Machine learning plays an important role in the operation of current wind energy production systems. One central application is predictive maintenance to increase efficiency and lower electricity costs by reducing downtimes. Integrating physics-based knowledge in neural networks to enforce their physical plausibilty is a promising method to improve current approaches, but incomplete system information often impedes their application in real world scenarios. We describe a simple and efficient way for physics-constrained deep learning-based predictive maintenance for wind turbine gearbox bearings with partial system knowledge. The approach is based on temperature nowcasting constrained by physics, where unknown system coefficients are treated as learnable neural network parameters. Results show improved generalization performance to unseen environments compared to a baseline neural network, which is especially important in low data scenarios often encountered in real-world applications.