@article{LIU2026, 
author = {Songkai LIU and Siyu CHENG and Pan SU and Yanzhang LI and Hao QIN and Changhe CHEN},
title = {Deep learning-based transient stability constrained optimal power flow with wind power uncertainty},
year = {2026},
journal = {Electric Power Engineering Technology},
volume = {45},
number = {5},
pages = {15-26},
keywords = {wind power uncertainty, power system, transient stability constrained optimal power flow (TSCOPF), complementary ensemble empirical mode decomposition with adaptive noise (CEEMDAN), improved multi-layer perceptron (IMLP), improved weighted mean of vectors (IINFO) algorithm},
url = {https://www.sciopen.com/article/10.12158/j.2096-3203.2026.05.002},
doi = {10.12158/j.2096-3203.2026.05.002},
abstract = {A deep learning-based transient stability constrained optimal power flow (TSCOPF) model with wind power uncertainty is proposed to address stability challenges faced by traditional TSCOPF models in power systems with increasing renewable energy integration. A cascaded prediction method combining complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), empirical wavelet transform (EWT), and long short-term memory (LSTM) networks is firstly developed to enhance wind power prediction accuracy and robustness. An improved multi-layer perceptron (IMLP) model is then employed to establish the mapping relationship between system operating states and transient stability index (TSI) for rapid and accurate transient stability assessment. An improved weIghted mean of vectors (IINFO) algorithm based on Lévy flight is subsequently adopted to solve the TSCOPF optimization problem. Simulation experiments are finally conducted on modified IEEE 39-bus and IEEE 68-bus test systems. Results demonstrate that the proposed cascaded method achieves superior prediction performance compared to traditional methods in wind power forecasting. The proposed TSCOPF model is verified to maintain stable system operation under wind power integration conditions. The improved IINFO algorithm exhibits significantly faster convergence speed and lower optimization costs than other optimization algorithms.}
}