Master's Thesis
Machine Learning-Based Cost Analysis and Prediction for Power Transmission Line Projects
— 2026
Key information
Authors:
Supervisors:
Published in
June 30, 2026
Abstract
Analyzing the key cost drivers of power transmission line projects—covering project attributes, project scale, physical work quantities, and external engineering conditions—is crucial for preliminary option appraisal and cost control. Traditional cost estimation methods often rely heavily on empirical indicators and similar-project comparisons, and may not adequately capture nonlinear cost patterns arising from interactions among these factors. To address this limitation, this study develops a machine-learning-based framework for cost driver analysis and construction cost prediction, aiming to improve predictive performance and enhance the interpretability of estimation results. First, historical project-level data are cleaned and organised to establish a comprehensive input variable system. Five models, namely Multiple Linear Regression, Random Forest, XGBoost, LightGBM, and CatBoost, are developed and evaluated using a consistent train-validation-test split to compare their predictive accuracy and generalisation performance. Finally, SHAP analysis is applied to identify the relative contribution of input variables to project construction costs. The results show that the ensemble learning models generally outperform the linear baseline model, with CatBoost achieving comparatively strong performance on the independent test set. The SHAP results identify conductor quantity, foundation concrete volume, line length, foundation steel quantity, voltage level, and number of transmission towers as the principal cost drivers. These findings provide data-driven decision support for owner-side preliminary cost estimation, feasibility-study refinement, option appraisal, and cost driver review in power transmission line engineering, based on available or reasonably estimable preliminary engineering information.
Publication details
Authors in the community:
Yuyang Chen
ist1115703
Supervisors of this institution:
Fields of Science and Technology (FOS)
civil-engineering - Civil engineering
Publication language (ISO code)
eng - English
Rights type:
Embargoed access
Date available:
April 26, 2027
Institution name
Instituto Superior Técnico