Erratum to: Real-time transient stability status prediction using cost-sensitive extreme learning machine

Zhen Chen, Xianyong Xiao, Changsong Li, Yin Zhang, Qingquan HuNeural Computing and Applications 2016 · DOI

This document serves as a formal correction to the research paper "Real-time transient stability status prediction using cost-sensitive extreme learning machine," originally published in the journal Neural Computing and Applications. The original study introduced a specialized machine learning approach—Cost-Sensitive Extreme Learning Machines (CS-ELM)—designed to predict the transient stability of power grids in real-time.

The Problem: Accuracy in Technical Documentation

In technical research, particularly in fields like power systems engineering where mathematical precision is paramount, accuracy in documentation is as critical as the research itself. This erratum addresses clerical errors regarding equation numbering and figure citations in the 2015 publication. These corrections are vital for researchers attempting to replicate the mathematical derivation of the stability prediction model or interpret the performance comparisons between cost-sensitive and cost-blind methods.

Context: Real-time Transient Stability Status Prediction (RTSSP)

The original research aimed to solve the challenge of RTSSP. In power systems, "transient stability" refers to the ability of the grid to remain in synchronism after being subjected to a major disturbance, such as a short circuit or line failure. Because these events occur within milliseconds, prediction must be both extremely fast and highly accurate to allow for effective remedial actions.

Methodology and Mathematical Framework

The authors utilized the Extreme Learning Machine (ELM), a type of feed-forward neural network known for its extremely fast training speeds, making it suitable for real-time applications. The erratum clarifies the mathematical framework:

  • Linear Equations: The training process for the ELM is equivalent to solving a specific linear equation, now correctly identified as Eq. 5.

  • Optimization: The Lagrangian function for the optimization problem is associated with Eq. 10, rather than Eq. 9 as previously stated.

  • Binary Classification: Since RTSSP is a binary problem (Stable vs. Unstable), the output layer of the neural network contains only one node ($m=1$).

The Importance of Cost-Sensitivity

A core argument of the original study is the superiority of cost-sensitive methods over "cost-blind" ones. In critical infrastructure like power grids, different types of errors have vastly different consequences. Failing to predict an unstable state (a false negative) is significantly more "costly" than a false alarm (a false positive).

By using a cost-sensitive approach, the model assigns different penalties to these errors. The erratum corrects a figure reference from Fig. 6 to Fig. 7 in the results section, ensuring that the visual evidence correctly supports the claim that cost-sensitive methods achieve lower overall costs and lower False Detection Rates (FDR).

Conclusion

While the fundamental conclusions of the original paper remain unchanged, these corrections ensure the structural integrity of the research. By providing the correct mathematical pointers and figure references, the authors facilitate a clearer understanding of how CS-ELM can be effectively applied to maintain the stability of modern power systems.

Summary updated

BibTeX
@article{chen2016erratum,
  author = {Zhen Chen and Xianyong Xiao and Changsong Li and Yin Zhang and Qingquan Hu},
  title = {{Erratum to: Real-time transient stability status prediction using cost-sensitive extreme learning machine}},
  journal = {Neural Computing and Applications},
  year = {2016},
  doi = {10.1007/s00521-015-1926-8},
}
Figure