DTR prediction model for transmission lines based on dual decomposition and jointly improved DBO-optimized BiLSTM
Year: 2026
Type: article
Abstract: Accurate prediction of the dynamic thermal rating of overhead transmission lines is important for safely increasing line capacity. To address the strong volatility of line-rating data and the sensitivity of neural-network performance to parameter settings, this paper proposes a forecasting model based on dual signal decomposition and an improved dung beetle optimizer optimized bidirectional long short-term memory network. First, complete ensemble empirical mode decomposition with adaptive noise is used for primary decomposition of the line-rating series. Sample entropy is then calculated to identify high-complexity components, and variational mode decomposition is applied to these components for secondary decomposition. The reconstructed components are predicted separately and then superposed. Finally, Latin hypercube initialization and a dynamic weight factor are introduced into the dung beetle optimizer to optimize the key network parameters. Experiments using measured data from a 220 kv transmission line in Guangzhou in January, April, July, and October show that the proposed model achieves stable one-step 15 min forecasting performance across seasons. Repeated trials, ablation experiments, comparison experiments, and decomposition-order validation further demonstrate the effectiveness and interpretability of the proposed modules. Recursive multi-step results are reported as a supplementary short-horizon sensitivity analysis rather than as the main accuracy evidence.
Source: Discover Applied Sciences
Authors: Jiewen Shen, Shuai Wang, Zhendong Zhu, Zhijian Liu
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7236778697
