A Data-Driven Long Time-Series Electrical Line Trip Fault Prediction Method Using an Improved Stacked-Informer...

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Data Science

Background:

• Effective monitoring of electrical equipment and power grids is crucial for safe power transmission and distribution.
• Accurate fault prediction in long time-series data is essential for ensuring power system reliability.
• Existing methods like Recurrent Neural Networks (RNN) and Long Short-Short Term Memory (LSTM) networks have limitations in handling real-world long sequences.

Purpose of the Study:

• To propose a data-driven method for electrical line trip fault sequence prediction.
• To address the limitations of current methods in predicting long time-series sequences.
• To improve the accuracy and efficiency of fault prediction in power systems.

Main Methods:

• Development of an improved stacked-Informer network architecture.
• Integration of gradient centralized (GC) technology with an optimizer, replacing the Adam optimizer.
• Utilizing long time-series data from a wind and solar hybrid substation for experimental validation.

Main Results:

• The stacked-Informer network effectively extracts underlying features from long time-series data.
• The proposed method demonstrates superior generalization ability and faster training efficiency compared to previous approaches.
• Experimental results confirm improved fault sequence prediction accuracy in real-world scenarios.

Conclusions:

• The improved stacked-Informer network offers a robust solution for electrical line trip fault sequence prediction.
• The method enhances the safety and operational efficiency of power transmission and distribution systems.
• This data-driven approach provides a significant advancement in predictive maintenance for power grids.