A constraint-coordinated hierarchical catboost framework for structurally consistent multi-target cost prediction in substation engineering projects

Tianqiong Chen; Hongda Chen; Jiaxiang Wen; Huijuan Huo; Zhikai Yang; Jing Duan; Cheng Xin; Zihao Sun; Jiyuan Zhang; Tianqiong Chen; Hongda Chen; Jiaxiang Wen; Huijuan Huo; Zhikai Yang; Jing Duan; Cheng Xin; Zihao Sun; Jiyuan Zhang

doi:10.3934/jimo.2026175

Journal of Industrial and Management Optimization

2026, Volume 22, Issue 10: 5074-5105. doi: 10.3934/jimo.2026175

Previous Article Next Article

Research article Special Issues

A constraint-coordinated hierarchical catboost framework for structurally consistent multi-target cost prediction in substation engineering projects

1.
State Grid Economic and Technological Research Institute Co., Ltd., Beijing 102209, China
2.
State Grid Sichuan Power Company, Chengdu 610041, Sichuan Province, China
3.
School of Economics and Management, North China Electric Power University, Beijing 102206, China
4.
State Grid Sichuan Economic Research Institute, Chengdu 610000, China

Received: 27 July 2026 Revised: 01 September 2026 Accepted: 10 September 2026 Published: 17 September 2026
68T05, 90C20

Accurate and structurally consistent cost prediction is essential for investment control, budget review, and refined management in substation engineering projects. Although machine-learning models have shown strong potential for engineering cost estimation, conventional single-stage prediction models usually treat multiple cost indicators as independent regression targets and may ignore the accounting identities among cost components. In this study, we proposed a constraint-coordinated hierarchical CatBoost framework for multi-target cost prediction using 2820 real substation project records collected from the State Grid system in 2024. The architecture decomposed the prediction task into a structured representation of heterogeneous engineering attributes, intermediate-variable completion, target-wise cost prediction, and reliability-weighted engineering-constraint coordination, with the final-account reserve fund treated separately as a sparse residual-type accounting target. The experimental results showed that the proposed framework achieved a global R² of 0.9521 and a global wMAPE of 17.06%, outperforming six conventional benchmark models. Although it did not uniformly surpass a strong single-stage CatBoost baseline in all global statistical metrics, it substantially reduced engineering-identity residuals and improved consistency between disaggregated cost components and aggregate investment indicators. These results indicate that the framework balances statistical accuracy with engineering accounting consistency and can support machine-learning-assisted cost estimation and decision support in power infrastructure projects.
- substation engineering,
- engineering cost prediction,
- CatBoost,
- hierarchical modeling,
- engineering constraints,
- applied machine learning
Citation: Tianqiong Chen, Hongda Chen, Jiaxiang Wen, Huijuan Huo, Zhikai Yang, Jing Duan, Cheng Xin, Zihao Sun, Jiyuan Zhang. A constraint-coordinated hierarchical catboost framework for structurally consistent multi-target cost prediction in substation engineering projects[J]. Journal of Industrial and Management Optimization, 2026, 22(10): 5074-5105. doi: 10.3934/jimo.2026175

Related Papers:

Abstract

Accurate and structurally consistent cost prediction is essential for investment control, budget review, and refined management in substation engineering projects. Although machine-learning models have shown strong potential for engineering cost estimation, conventional single-stage prediction models usually treat multiple cost indicators as independent regression targets and may ignore the accounting identities among cost components. In this study, we proposed a constraint-coordinated hierarchical CatBoost framework for multi-target cost prediction using 2820 real substation project records collected from the State Grid system in 2024. The architecture decomposed the prediction task into a structured representation of heterogeneous engineering attributes, intermediate-variable completion, target-wise cost prediction, and reliability-weighted engineering-constraint coordination, with the final-account reserve fund treated separately as a sparse residual-type accounting target. The experimental results showed that the proposed framework achieved a global R² of 0.9521 and a global wMAPE of 17.06%, outperforming six conventional benchmark models. Although it did not uniformly surpass a strong single-stage CatBoost baseline in all global statistical metrics, it substantially reduced engineering-identity residuals and improved consistency between disaggregated cost components and aggregate investment indicators. These results indicate that the framework balances statistical accuracy with engineering accounting consistency and can support machine-learning-assisted cost estimation and decision support in power infrastructure projects.

References

[1]	K. Sun, D. Zhang, J. Wang, W. Mao, Data-driven and physically informed power grid dispatch decision-making method, Sustain. Energy Grids Netw., 42 (2025), 101644. https://doi.org/10.1016/j.segan.2025.101644 doi: 10.1016/j.segan.2025.101644
[2]	O. Bertozzi, G. A. Ooi, M. A. Aftab, N. Mohammed, M. S. Venkatraman, G. P. Adam, et al., Data-driven planning of mixed-generation power systems: Towards 100% RES-based grids, Int. J. Electr. Power Energy Syst., 172 (2025), 111207. https://doi.org/10.1016/j.ijepes.2025.111207 doi: 10.1016/j.ijepes.2025.111207
[3]	Z. Ge, C. Yan, J. Guo, C. Zhang, Evaluating the embodied carbon emissions of power infrastructure construction towards carbon neutrality: A study of Fujian Province, China, Energy Strategy Rev., 64 (2026), 102096. https://doi.org/10.1016/j.esr.2026.102096 doi: 10.1016/j.esr.2026.102096
[4]	S. Y. Teng, C. C. V. Nooten, J. M. V. Doorn, A. Ottenbros, M. A. J. Huijbregts, J. J. Jansen, Near real-time predictions of renewable electricity production at substation level via domain adaptation zero-shot learning in sequence, Renew. Sustain. Energy Rev., 186 (2023), 113662. https://doi.org/10.1016/j.rser.2023.113662 doi: 10.1016/j.rser.2023.113662
[5]	A. Gao, J. Zheng, F. Mei, Y. Liu, Toward intelligent demand-side energy management via substation-level flexible load disaggregation, Appl. Energy, 367 (2024), 123361. https://doi.org/10.1016/j.apenergy.2024.123361 doi: 10.1016/j.apenergy.2024.123361
[6]	M. Alipour, J. Aghaei, M. Norouzi, T. Niknam, S. Hashemi, M. Lehtonen, A novel electrical net-load forecasting model based on deep neural networks and wavelet transform integration, Energy, 205 (2020), 118106. https://doi.org/10.1016/j.energy.2020.118106 doi: 10.1016/j.energy.2020.118106
[7]	B. A. Atalay, K. Zor, An innovative approach for forecasting hydroelectricity generation by benchmarking tree-based machine learning models, Appl. Sci., 15 (2025), 10514. https://doi.org/10.3390/app151910514 doi: 10.3390/app151910514
[8]	İ. Levent, G. Şahin, G. Işık, W. G. J. H. M. V. Sark, Comparative analysis of advanced machine learning regression models with advanced artificial intelligence techniques to predict rooftop PV solar power plant efficiency using indoor solar panel parameters, Appl. Sci., 15 (2025), 3320. https://doi.org/10.3390/app15063320 doi: 10.3390/app15063320
[9]	Y. Jiang, J. Hu, S. Mei, Y. Zhou, Power engineering cost prediction model based on BP neural network, Procedia Comput. Sci., 261 (2025), 1057–1064. https://doi.org/10.1016/j.procs.2025.04.685 doi: 10.1016/j.procs.2025.04.685
[10]	H. Lu, M. Fang, Y. Zhang, X. Li, BayesShrink threshold estimation-based multi-metric space clustering algorithm for power grid project cost data, Results Eng., 29 (2026), 109134. https://doi.org/10.1016/j.rineng.2026.109134 doi: 10.1016/j.rineng.2026.109134
[11]	C. Wang, J. Qiao, Construction project cost prediction method based on improved BiLSTM, Appl. Sci., 14 (2024), 978. https://doi.org/10.3390/app14030978 doi: 10.3390/app14030978
[12]	A. Mahmoodzadeh, H. R. Nejati, M. Mohammadi, H. Hashim Ibrahim, M. Khishe, S. Rashidi, et al., Developing six hybrid machine learning models based on Gaussian process regression and meta-heuristic optimization algorithms for prediction of duration and cost of road tunnels construction, Tunn. Undergr. Space Technol., 130 (2022), 104759. https://doi.org/10.1016/j.tust.2022.104759 doi: 10.1016/j.tust.2022.104759
[13]	L. Chen, C. Xu, W. H. Lim, A. Sharma, S. S. Tiang, K. S. Chong, et al., Transparent and reliable construction cost prediction using advanced machine learning and explainable AI, Eng. Sci. Technol. Int. J., 70 (2025), 102159. https://doi.org/10.1016/j.jestch.2025.102159 doi: 10.1016/j.jestch.2025.102159
[14]	F. Antoniou, K. Konstantinidis, Client-oriented highway construction cost estimation models using machine learning, Appl. Sci., 15 (2025), 10237. https://doi.org/10.3390/app151810237 doi: 10.3390/app151810237
[15]	U. Park, Y. Kang, H. Lee, S. Yun, A stacking heterogeneous ensemble learning method for the prediction of building construction project costs, Appl. Sci., 12 (2022), 9729. https://doi.org/10.3390/app12199729 doi: 10.3390/app12199729
[16]	Y. S. Ko, B. G. Lee, A study on machine learning-based cost estimation models for AI training data construction, Appl. Sci., 16 (2026), 2891. https://doi.org/10.3390/app16062891 doi: 10.3390/app16062891
[17]	H. Borchani, G. Varando, C. Bielza, P. Larrañaga, A survey on multi-output regression, WIREs Data Min. Knowl. Discov., 5 (2015), 216–233. https://doi.org/10.1002/widm.1157 doi: 10.1002/widm.1157
[18]	S. L. Wickramasuriya, G. Athanasopoulos, R. J. Hyndman, Optimal forecast reconciliation for hierarchical and grouped time series through trace minimization, J. Am. Stat. Assoc., 114 (2019), 804–819. https://doi.org/10.1080/01621459.2018.1448825 doi: 10.1080/01621459.2018.1448825
[19]	M. Waqar, Y. W. Kim, Y. C. Byun, A hybrid deep learning framework for multivariate energy forecasting and peak load prediction in electric vehicle charging infrastructure, Appl. Energy, 402 (2026), 126964. https://doi.org/10.1016/j.apenergy.2025.126964 doi: 10.1016/j.apenergy.2025.126964
[20]	W. Sfar, L. Amhaimar, A. Khalidi, B. Talbi, A hybrid long-term photovoltaic power prediction model integrating a BiLSTM network with residual correction via CatBoost, Results Eng., 29 (2026), 108898. https://doi.org/10.1016/j.rineng.2025.108898 doi: 10.1016/j.rineng.2025.108898
[21]	Z. Zhou, Y. Dai, M. Leng, A photovoltaic power forecasting framework based on attention mechanism and parallel prediction architecture, Appl. Energy, 391 (2025), 125869. https://doi.org/10.1016/j.apenergy.2025.125869 doi: 10.1016/j.apenergy.2025.125869
[22]	H. Liu, H. Wu, H. Jin, Y. He, Adaptive forecasting of photovoltaic power based on dual-type models' ensemble and online error correction, Appl. Energy, 408 (2026), 127397. https://doi.org/10.1016/j.apenergy.2026.127397 doi: 10.1016/j.apenergy.2026.127397
[23]	L. Zhang, D. Jánošík, Enhanced short-term load forecasting with hybrid machine learning models: CatBoost and XGBoost approaches, Expert Syst. Appl., 241 (2024), 122686. https://doi.org/10.1016/j.eswa.2023.122686 doi: 10.1016/j.eswa.2023.122686
[24]	M. Shafiuzzaman, M. Safayet Islam, T. M. Rubaith Bashar, M. Munem, M. Nahiduzzaman, M. Ahsan, et al., Enhanced very short-term load forecasting with multi-lag feature engineering and Prophet-XGBoost-CatBoost architecture, Energy, 335 (2025), 137981. https://doi.org/10.1016/j.energy.2025.137981 doi: 10.1016/j.energy.2025.137981
[25]	Z. Colak, Hybrid machine learning approach in forecasting renewable energy production: Extra Trees, CatBoost and LightGBM-based stacking model, Renew. Energy, 260 (2026), 125167. https://doi.org/10.1016/j.renene.2025.125167 doi: 10.1016/j.renene.2025.125167
[26]	W. Xiang, P. Xu, J. Fang, Q. Zhao, Z. Gu, Q. Zhang, Multi-dimensional data-based medium- and long-term power-load forecasting using double-layer CatBoost, Energy Rep., 8 (2022), 8511–8522. https://doi.org/10.1016/j.egyr.2022.06.063 doi: 10.1016/j.egyr.2022.06.063
[27]	S. Ruan, Y. Chen, G. Lu, Z. Li, S. Chen, C. Wang, et al., Multifactor interpretability method for offshore wind power output prediction based on TPE-CatBoost-SHAP, Comput. Electr. Eng., 123 (2025), 110081. https://doi.org/10.1016/j.compeleceng.2025.110081 doi: 10.1016/j.compeleceng.2025.110081
[28]	X. Guo, P. Zeng, X. Xiong, G. Wang, Y. Cui, Short-term wind power forecasting methods based on machine learning: A review and case study, Energy Rep., 14 (2025), 3753–3782. https://doi.org/10.1016/j.egyr.2025.10.040 doi: 10.1016/j.egyr.2025.10.040
[29]	A. Wang, Q. Yu, J. Wang, X. Yu, Z. Wang, Z. Hu, Electric load forecasting based on deep ensemble learning, Appl. Sci., 13 (2023), 9706. https://doi.org/10.3390/app13179706 doi: 10.3390/app13179706
[30]	Y. Zhang, Z. Wang, M. Xi, Y. Zhao, J. Liu, Data-driven prediction of residual flexural capacity in corroded RC beams using PSO and GA-optimized CatBoost ensemble models, Eng. Res. Express, 7 (2025), 035129. https://doi.org/10.1088/2631-8695/adfcb3 doi: 10.1088/2631-8695/adfcb3
[31]	Y. Zhang, Z. Wang, J. Liu, Y. Li, Z. Huang, X. Yu, A metaheuristic-driven categorical boosting framework with interpretability for high-precision prediction of mechanical properties in corroded reinforced concrete beams, Eng. Appl. Artif. Intell., 163 (2026), 112804. https://doi.org/10.1016/j.engappai.2025.112804 doi: 10.1016/j.engappai.2025.112804
[32]	Q. Xu, C. Zhang, S. Geng, S. Wang, J. Li, J. Liu, et al., An explainable ensemble learning framework for flexible pavement roughness prediction under multi-climate stressors, Case Stud. Constr. Mater., 23 (2025), e05402. https://doi.org/10.1016/j.cscm.2025.e05402 doi: 10.1016/j.cscm.2025.e05402
[33]	S. Xiong, C. Nie, Y. Lu, J. Zheng, Research on price prediction of construction materials based on VMD-SSA-LSTM, Appl. Sci., 15 (2025), 2005. https://doi.org/10.3390/app15042005 doi: 10.3390/app15042005

Reader Comments

Your name:*

Email:*
© 2026 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)