A Hybrid Multi-Objective Optimization Model for Vibration Tendency Prediction of Hydropower Generators

Area of Science:

• Power Systems Engineering
• Mechanical Vibrations
• Artificial Intelligence

Background:

• Hydropower generator units (HGUs) are critical for power grid stability, and their vibration signals offer insights into operational status.
• Predictive maintenance of HGUs relies on accurate vibration tendency prediction, but existing methods often prioritize either stability or accuracy.
• A simultaneous focus on both prediction stability and accuracy is crucial for effective HGU health monitoring.

Purpose of the Study:

• To propose an intelligent vibration tendency prediction method for HGUs that achieves both strong stability and high accuracy.
• To integrate signal preprocessing, feature selection, and prediction within a multi-objective optimization framework.
• To enhance the predictive maintenance capabilities for hydropower generator units.

Main Methods:

• Empirical wavelet transform (EWT) for raw sensor signal decomposition into modes.
• Sample entropy-based reconstruction for refactoring signal modes.
• Gram-Schmidt orthogonal (GSO) process for important feature selection.
• Kernel extreme learning machine (KELM) for refactored mode prediction.
• Multi-objective salp swarm algorithm for synchronous optimization of GSO and KELM parameters.

Main Results:

• The proposed method successfully integrates multiple techniques within a multi-objective optimization framework.
• Feature selection using GSO and prediction via KELM, optimized by the salp swarm algorithm, demonstrated effectiveness.
• A case study on mixed-flow HGU data confirmed superior performance in predicting vibration tendency stability and accuracy compared to traditional methods.

Conclusions:

• The developed intelligent method offers a robust solution for predicting HGU vibration tendencies by optimizing for both stability and accuracy.
• This approach advances predictive maintenance strategies for hydropower infrastructure.
• The integrated multi-objective optimization framework provides a promising direction for complex machinery health monitoring.