Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Predicting Estimated Times of Restoration for Electrical Outages Using Longitudinal Tabular Transformers

Bogireddy Sai Prasanna Teja, Valliappan Muthukaruppan, Carls Benjamin

TL;DR

This work tackles the challenge of delivering accurate and timely Estimated Times of Restoration (ETR) during severe outages by exploiting longitudinal outage data. It introduces a Longitudinal Tabular Transformer (LTT) that processes sequences of incremental updates for each outage event, updating ETR predictions as new information arrives. The approach is augmented with a customer-focused asymmetric loss, new evaluation metrics (UPR, OPR-8, WAE, CSI), and interpretability tools (attention visualization and SHAP) to align predictions with real-world customer satisfaction. Empirical results across 34,000 storm events from three utilities show that LTT outperforms traditional ML/DL baselines in accuracy and reliability, while offering transparent, explainable predictions that can improve communication and trust with customers.

Abstract

As climate variability increases, the ability of utility providers to deliver precise Estimated Times of Restoration (ETR) during natural disasters has become increasingly critical. Accurate and timely ETRs are essential for enabling customer preparedness during extended power outages, where informed decision-making can be crucial, particularly in severe weather conditions. Nonetheless, prevailing utility practices predominantly depend on manual assessments or traditional statistical methods, which often fail to achieve the level of precision required for reliable and actionable predictions. To address these limitations, we propose a Longitudinal Tabular Transformer (LTT) model that leverages historical outage event data along with sequential updates of these events to improve the accuracy of ETR predictions. The model's performance was evaluated over 34,000 storm-related outage events from three major utility companies, collectively serving over 3 million customers over a 2-year period. Results demonstrate that the LTT model improves the Customer Satisfaction Impact (CSI) metric by an average of 19.08% (p > 0.001) compared to existing methods. Additionally, we introduce customer-informed regression metrics that align model evaluation with real-world satisfaction, ensuring the outcomes resonate with customer expectations. Furthermore, we employ interpretability techniques to analyze the temporal significance of incorporating sequential updates in modeling outage events and to identify the contributions of predictive features to a given ETR. This comprehensive approach not only improves predictive accuracy but also enhances transparency, fostering greater trust in the model's capabilities.

Predicting Estimated Times of Restoration for Electrical Outages Using Longitudinal Tabular Transformers

TL;DR

This work tackles the challenge of delivering accurate and timely Estimated Times of Restoration (ETR) during severe outages by exploiting longitudinal outage data. It introduces a Longitudinal Tabular Transformer (LTT) that processes sequences of incremental updates for each outage event, updating ETR predictions as new information arrives. The approach is augmented with a customer-focused asymmetric loss, new evaluation metrics (UPR, OPR-8, WAE, CSI), and interpretability tools (attention visualization and SHAP) to align predictions with real-world customer satisfaction. Empirical results across 34,000 storm events from three utilities show that LTT outperforms traditional ML/DL baselines in accuracy and reliability, while offering transparent, explainable predictions that can improve communication and trust with customers.

Abstract

As climate variability increases, the ability of utility providers to deliver precise Estimated Times of Restoration (ETR) during natural disasters has become increasingly critical. Accurate and timely ETRs are essential for enabling customer preparedness during extended power outages, where informed decision-making can be crucial, particularly in severe weather conditions. Nonetheless, prevailing utility practices predominantly depend on manual assessments or traditional statistical methods, which often fail to achieve the level of precision required for reliable and actionable predictions. To address these limitations, we propose a Longitudinal Tabular Transformer (LTT) model that leverages historical outage event data along with sequential updates of these events to improve the accuracy of ETR predictions. The model's performance was evaluated over 34,000 storm-related outage events from three major utility companies, collectively serving over 3 million customers over a 2-year period. Results demonstrate that the LTT model improves the Customer Satisfaction Impact (CSI) metric by an average of 19.08% (p > 0.001) compared to existing methods. Additionally, we introduce customer-informed regression metrics that align model evaluation with real-world satisfaction, ensuring the outcomes resonate with customer expectations. Furthermore, we employ interpretability techniques to analyze the temporal significance of incorporating sequential updates in modeling outage events and to identify the contributions of predictive features to a given ETR. This comprehensive approach not only improves predictive accuracy but also enhances transparency, fostering greater trust in the model's capabilities.
Paper Structure (31 sections, 18 equations, 5 figures, 1 table)

This paper contains 31 sections, 18 equations, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Experimental Setup
  3. Data
  4. Baseline Comparisons
  5. Asymmetric Loss Function for Customer Satisfaction
  6. Formulation of the Asymmetric Loss Function
  7. Evaluation Metrics
  8. Under-prediction Rate (UPR)
  9. Over-prediction Beyond 8 Hours Rate (OPR-8)
  10. Weighted Asymmetric Error (WAE)
  11. Customer Satisfaction Impact (CSI)
  12. Longitudinal Tabular Transformer
  13. Problem Setup
  14. Feature Representation
  15. Categorical Features
  16. ...and 16 more sections

Figures (5)

  • Figure 1: Storm-Realted Outage Counts between Jan 2022 to Jan 2024 for OPCO-1, OPCO-2, and OPCO-3.
  • Figure 2: Storm-Outage Related Customer Satisfaction Survey Analysis
  • Figure 3: Architecture of Longitudinal Tabular Transformer
  • Figure 4: Attention Weights Heatmap for OPCO-1 LTT Model
  • Figure 5: Top 5 features identified through SHAP analysis across the first five event revisions for OPCO-3, highlighting the evolving impact of key features on the predicted Estimated Time of Restoration (ETR).