Time series forecasting based on optimized LLM for fault prediction in distribution power grid insulators

TL;DR

This work tackles the risk of power-grid outages from leakage-current rise in contaminated insulators by introducing a hybrid forecasting framework that combines an input-stage noise filter with a time-series large language model (timeLLM) whose hyperparameters are optimized via a tree-structured Parzen estimator in Optuna. Among multiple filters tested, Empirical Wavelet Transform (EWT) provided the best denoising, enabling accurate forecasting by the optimized timeLLM, which achieved a short-term RMSE of $2.24\times10^{-4}$ and a medium-term RMSE of $1.21\times10^{-3}$, outperforming state-of-the-art DL models. The approach demonstrates robust performance across multi-horizon forecasts and exhibits favorable stability over 50 random initializations, suggesting practical utility for predictive maintenance in distribution grids. The results highlight a scalable framework that integrates signal processing, transformer-based temporal reasoning, and principled hyperparameter search, with potential extensions to real-time deployment and cross-domain validation in smart-grid contexts.

Abstract

Surface contamination on electrical grid insulators leads to an increase in leakage current until an electrical discharge occurs, which can result in a power system shutdown. To mitigate the possibility of disruptive faults resulting in a power outage, monitoring contamination and leakage current can help predict the progression of faults. Given this need, this paper proposes a hybrid deep learning (DL) model for predicting the increase in leakage current in high-voltage insulators. The hybrid structure considers a multi-criteria optimization using tree-structured Parzen estimation, an input stage filter for signal noise attenuation combined with a large language model (LLM) applied for time series forecasting. The proposed optimized LLM outperforms state-of-the-art DL models with a root-mean-square error equal to 2.24$\times10^{-4}$ for a short-term horizon and 1.21$\times10^{-3}$ for a medium-term horizon.

Figures (8)

• Figure 1: Proposed optimized LLM model.
• Figure 2: Leakage current in an insulator with artificial contamination.
• Figure 3: Comparison of application of filters to the original signal.
• Figure 4: Results of multi-objective optimization.
• Figure 5: Trial rank of hyperparâmeters.