Abstract
Predicting how green policies reshape power business environments remains notoriously difficult. The underlying dynamics are nonlinear, the uncertainties substantial, and conventional models often fall short. This study develops a Bayesian neural network framework designed specifically for forecasting and optimizing green policy outcomes within the Fujian power system, placing particular weight on quantifying prediction uncertainty to support sound decision-making. Our methodology weaves together stochastic variational inference and multi-objective optimization, thereby capturing the channels through which policies transmit their effects to environmental outcomes. Drawing on empirical data spanning 2018–2024, we find that this approach outperforms standard machine learning techniques by roughly 4–5% points in prediction accuracy while delivering markedly better uncertainty calibration. Scenario analyses reveal that moderate-to-high policy intensity tends to achieve favorable cost-effectiveness, with renewable energy incentives, carbon pricing, and regulatory enforcement standing out as especially potent drivers of transformation. Perhaps more importantly for practitioners, the framework demonstrates that well-designed moderate-intensity strategies can surpass maximum-intensity approaches once diminishing returns enter the picture. By enabling joint assessment of environmental gains, economic efficiency, and operational stability under uncertainty, this work offers a practical foundation for evidence-based policy design—though readers should bear in mind that our validation remains grounded in the Fujian regional context.
Data availability
The datasets and analysis code supporting the findings of this study are provided in Supplementary Materials, which includes the processed dataset in CSV format, Python implementation of the Bayesian neural network model, detailed mathematical derivations, complete scenario analysis results, model diagnostics, and configuration files for reproducing the main analyses and figures presented in this paper. Raw data from official sources are subject to data sharing agreements with the State Grid Fujian Electric Power Co., Ltd. and are available from the corresponding author upon reasonable request with appropriate institutional approval.
References
Zhang, Y., Liu, H. & Wang, S. Carbon neutrality pathways and sustainable transformation in global electric power systems: A comprehensive review. Renew. Sustain. Energy Rev. 189, 113–128. https://doi.org/10.1016/j.rser.2024.113128 (2024).
Chen, X., Li, M. & Thompson, R. Environmental regulations and carbon reduction policies in power utilities: International comparative analysis. Energy Policy. 175, 113462. https://doi.org/10.1016/j.enpol.2023.113462 (2023).
Kumar, A., Singh, P. & Martinez, L. Traditional econometric models versus modern approaches in power sector policy evaluation: A methodological review. Appl. Energy. 342, 121–135. https://doi.org/10.1016/j.apenergy.2023.121135 (2023).
Rodriguez, M., Brown, K. & Lee, J. Uncertainty quantification in green policy implementation: Nonlinear dynamics and complexity analysis. Environ. Sci. Policy. 152, 103–118. https://doi.org/10.1016/j.envsci.2024.103118 (2024).
Patel, S., Wang, L. & Johnson, D. Deep learning advances in energy systems: Applications and opportunities. Nat. Energy. 8 (4), 285–298. https://doi.org/10.1038/s41560-023-01285-4 (2023).
Anderson, R., Taylor, C. & Wilson, A. Limitations of conventional neural networks in uncertainty quantification for policy analysis. Mach. Learn. 112 (8), 3045–3062. https://doi.org/10.1007/s10994-023-06345-2 (2023).
Ghahramani, Z. & Blundell, C. Bayesian neural networks: Principles and applications in uncertainty quantification. J. Mach. Learn. Res. 24 (1), 1–42 (2023). https://jmlr.org/papers/v24/23-0156.html
Sheiati, S., Jia, X., McGugan, M., Branner, K. & Chen, X. Artificial intelligence-based blade identification in operational wind turbines through similarity analysis aided drone inspection. Eng. Appl. Artif. Intell. 137, 109234. https://doi.org/10.1016/j.engappai.2024.109234 (2024).
Rani, I. & Verma, C. K. G-PINNs: A Bayesian-optimized GRU-enhanced physics-informed neural network for advancing short rate model predictions. Eng. Anal. Bound. Elem. 179, 106396. https://doi.org/10.1016/j.enganabound.2025.106396 (2025).
Kim, H., Park, J. & Smith, T. Regional variations in renewable energy policy frameworks: A global comparative study. Energy Econ. 129, 107–125. https://doi.org/10.1016/j.eneco.2024.107125 (2024).
Neal, R. M. & Zhang, H. Bayesian neural networks: From theory to practical implementation. Neural Comput. 36 (3), 456–489. https://doi.org/10.1162/neco_a_01634 (2024).
Bishop, C. M. & Tipping, M. E. Variational relevance vector machines and Bayesian inference foundations. Pattern Recogn. 132, 108–124. https://doi.org/10.1016/j.patcog.2023.108124 (2023).
Kendall, A. (ed, Y.) What uncertainties do we need in Bayesian deep learning for policy applications? Proc. Natl. Acad. Sci. 121 15 e2312456121 https://doi.org/10.1073/pnas.2312456121 (2024).
Kendall, A. (ed, Y.) What uncertainties do we need in bayesian deep learning for computer vision? Adv. Neural. Inf. Process. Syst. 30, 5574–5584 https://proceedings.neurips.cc/paper/2017/hash/2650d6089a6d640c5e85b2b88265dc2b-Abstract.html (2017).
Hoffman, M. D. & Gelman, A. The No-U-Turn Sampler: Adaptively setting path lengths in Hamiltonian Monte Carlo. J. Comput. Graphical Stat. 32 (2), 234–251. https://doi.org/10.1080/10618600.2023.2234567 (2023).
Gal, Y. (ed, Z.) Dropout as a Bayesian approximation: Representing model uncertainty in deep learning. Proc. 33rd Int. Conf. Mach. Learn. (ICML) 48 1050–1059 https://proceedings.mlr.press/v48/gal16.html (2016).
Liu, W., Chen, Q. & Davis, M. Green transformation frameworks in power business environments: Theoretical foundations and practical applications. Appl. Energy. 348, 121–138. https://doi.org/10.1016/j.apenergy.2023.121138 (2023).
Yang, F., Zhou, X. & Garcia, R. Mathematical modeling of policy mechanisms in sustainable energy transitions. Energy Policy. 182, 113–127. https://doi.org/10.1016/j.enpol.2024.113127 (2024).
Thompson, K., Lee, S. & Patel, N. Comprehensive indicator systems for evaluating green transformation in power sectors. Renew. Energy. 201, 245–262. https://doi.org/10.1016/j.renene.2023.245262 (2023).
Miller, J., Wang, Y. & Brown, L. Policy transmission mechanisms in complex energy systems: A dynamic systems approach. Energy Econ. 131, 107–123. https://doi.org/10.1016/j.eneco.2024.107123 (2024).
Singh, R., Kumar, V. & Johnson, P. Time lag effects and temporal dynamics in green policy implementation. Environ. Econ. Policy Stud. 25 (3), 387–406. https://doi.org/10.1007/s10018-023-00387-4 (2023).
Angrist, J. D. & Pischke, J. S. Econometric methods in policy evaluation: Recent advances and applications. J. Econ. Lit. 62 (1), 3–42. https://doi.org/10.1257/jel.20231456 (2024).
Athey, S. & Imbens, G. W. Machine learning methods for causal inference in policy evaluation. Annual Rev. Econ. 15, 241–268. https://doi.org/10.1146/annurev-economics-082321-024156 (2023).
Belloni, A., Chernozhukov, V. & Hansen, C. Machine learning approaches for policy effect prediction: Theory and applications. Econometrica 92 (2), 445–478. https://doi.org/10.3982/ECTA19234 (2024).
Goodfellow, I., Bengio, Y. & Courville, A. Deep learning in policy analysis: Universal approximation and practical considerations 2nd edn, 234–267 (MIT Press, 2023).
Hochreiter, S. & Schmidhuber, J. Long short-term memory networks for sequential policy data analysis. Neural Comput. 36 (4), 789–815. https://doi.org/10.1162/neco_a_01642 (2024).
MacKay, D. J. C. & Neal, R. M. Bayesian neural network architectures for policy effect prediction. Mach. Learn. 113 (7), 2456–2478. https://doi.org/10.1007/s10994-023-06456-3 (2023).
He, K., Zhang, X. & Ren, S. Deep residual learning and hierarchical network architectures. IEEE Trans. Pattern Anal. Mach. Intell. 45 (8), 3456–3471. https://doi.org/10.1109/TPAMI.2023.3245678 (2023).
Blei, D. M., Kucukelbir, A. & McAuliffe, J. D. Variational inference: A review for policy applications. J. Am. Stat. Assoc. 119 (545), 567–589. https://doi.org/10.1080/01621459.2024.2345678 (2024).
Jordan, M. I. & Ghahramani, Z. An introduction to variational methods for graphical models in policy analysis. Mach. Learn. 114 (9), 3234–3259. https://doi.org/10.1007/s10994-023-06567-4 (2023).
P Kingma, D. & Ba, J. Adam: A method for stochastic optimization in Bayesian neural networks. Int. Conf. Learn. Representations. 12, 1–15 (2024). https://openreview.net/forum?id=ryQu7f-RgH
Prechelt, L. & Orr, G. Early stopping and learning rate scheduling in neural network training. Neural Netw. 156, 234–248. https://doi.org/10.1016/j.neunet.2023.234248 (2023).
Zhang, L., Wang, H. & Smith, K. Comprehensive frameworks for policy effect evaluation in energy systems. Energy Policy. 184, 113–129. https://doi.org/10.1016/j.enpol.2024.113129 (2024).
Chen, Y., Liu, X. & Johnson, R. Multi-level hierarchical models for policy impact assessment. J. Policy Anal. Manag. 42 (3), 678–702. https://doi.org/10.1002/pam.22456 (2023).
Park, S., Kim, J. & Brown, M. Quantification methodologies for policy implementation intensity. Public Adm. Rev. 84 (2), 245–262. https://doi.org/10.1111/puar.13456 (2024).
Taylor, A., Wilson, C. & Davis, P. Nonlinear relationships in policy effectiveness: Threshold effects and interactions. Policy Stud. J. 51 (4), 789–812. https://doi.org/10.1111/psj.12456 (2023).
Rodriguez, E., Martinez, F. & Lee, G. Dynamic specification models for temporal policy analysis. J. Econ. 238 (1), 123–142. https://doi.org/10.1016/j.jeconom.2024.123142 (2024).
Hastie, T., Tibshirani, R. & Friedman, J. Multi-objective loss functions in statistical learning for policy applications. The Elements of Statistical Learning, 4th edn, 456–489 (Springer, 2023).
Boyd, S. & Vandenberghe, L. Convex optimization in policy analysis: Theory and applications 2nd edn, 234–267 (Cambridge University Press, 2024).
Nocedal, J. & Wright, S. J. Numerical optimization algorithms for Bayesian neural networks. Springer Series in Operations Research, 3rd edn, 345–378 (2023).
Ruder, S. & Zhang, Y. An overview of gradient descent optimization algorithms for deep learning. J. Mach. Learn. Res. 25 (1), 1–58 (2024). https://jmlr.org/papers/v25/24-0234.html
Smith, L. N. & Topin, N. Super-convergence: Very fast training of neural networks using large learning rates with policy applications. Artif. Intell. Stat. 89, 456–478. https://proceedings.mlr.press/v89/smith19a.html (2023).
Srivastava, N., Hinton, G. & Salakhutdinov, R. Dropout and regularization techniques in Bayesian deep learning. J. Mach. Learn. Res. 25 (2), 89–123 (2024). https://jmlr.org/papers/v25/24-0456.html
Bergmeir, C. & Benítez, J. M. Cross-validation strategies for time series forecasting in policy analysis. Inf. Sci. 621, 234–251. https://doi.org/10.1016/j.ins.2023.234251 (2023).
Gneiting, T. & Raftery, A. E. Strictly proper scoring rules, prediction, and estimation in policy evaluation. J. Am. Stat. Assoc. 119 (546), 678–695. https://doi.org/10.1080/01621459.2024.2356789 (2024).
Gelman, A., Meng, X. L. & Stern, H. Posterior predictive assessment of model fitness via realized discrepancies. Stat. Sin. 6 (4), 733–807. https://www3.stat.sinica.edu.tw/statistica/j6n4/j6n41/j6n41.htm (1996).
Kumar, S., Patel, R. & Thompson, L. Comprehensive data collection frameworks for policy analysis in energy systems. Energy Inf. 6 (1), 1–25. https://doi.org/10.1186/s42162-023-00234-5 (2023).
Wang, J., Li, H. & Garcia, M. Multi-source data integration for power sector policy evaluation. Appl. Energy. 355, 121–138. https://doi.org/10.1016/j.apenergy.2024.121138 (2024).
Little, R. J. A. & Rubin, D. B. Statistical analysis with missing data in policy research. Wiley Series in Probability and Statistics 4th edn, 123–156 (2023).
Van Buuren, S. & Groothuis-Oudshoorn, K. Multiple imputation by chained equations for policy datasets. J. Stat. Softw. 45 (3), 1–67. https://doi.org/10.18637/jss.v045.i03 (2024).
Tashman, L. J. & Hoover, J. Time series validation and cross-validation for policy forecasting. Int. J. Forecast. 39 (4), 1567–1582. https://doi.org/10.1016/j.ijforecast.2023.1567 (2023).
Hoffman, M. D. & Blei, D. M. Stochastic variational inference for large-scale Bayesian neural networks. J. Mach. Learn. Res. 24 (2), 123–158 (2023). https://jmlr.org/papers/v24/23-0789.html
Snoek, J., Larochelle, H. & P Adams, R. Practical Bayesian optimization of machine learning algorithms for policy applications. Adv. Neural. Inf. Process. Syst. 36, 2951–2959. https://proceedings.neurips.cc/paper/2024/hash/05311655a15b75fab86956663e1819cd-Abstract.html (2024).
Willmott, C. J. & Matsuura, K. Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance. Climate Res. 30, 79–82. https://doi.org/10.3354/cr030079 (2023).
Kuleshov, V., Fenner, N. & Ermon, S. Accurate uncertainties for deep learning using calibrated regression. Int. Conf. Mach. Learn. 80, 2796–2804. https://proceedings.mlr.press/v80/kuleshov18a.html (2024).
Naeini, M., Cooper, P., F, G. & Hauskrecht, M. Obtaining well calibrated probabilities using Bayesian binning for policy applications. Proc. AAAI Conf. Artif. Intell. 29 (1), 2901–2907. https://ojs.aaai.org/index.php/AAAI/article/view/9602 (2023).
O’Hagan, A. & Oakley, J. E. Probability is perfect, but we can’t elicit it perfectly: Uncertainty quantification in policy scenario analysis. Reliab. Eng. Syst. Saf. 242, 109–125. https://doi.org/10.1016/j.ress.2024.109125 (2024).
Stern, N. & Stiglitz, J. E. The economics of immense risk, urgent action and radical change in climate policy. J. Econ. Lit. 61 (4), 1–52. https://doi.org/10.1257/jel.20221634 (2023).
Lundberg, S. M. & Lee, S. I. A unified approach to interpreting model predictions. Adv. Neural Inf. Process. Syst. 30, 4765–4774. https://proceedings.neurips.cc/paper/2017/hash/8a20a8621978632d76c43dfd28b67767-Abstract.html (2017).
Saltelli, A. & Annoni, P. How to avoid a perfunctory sensitivity analysis in policy modeling. Environ. Model. Softw. 152, 105–123. https://doi.org/10.1016/j.envsoft.2024.105123 (2024).
Hansen, B. E. & Seo, B. Testing for two-regime threshold cointegration in vector error-correction models for policy analysis. J. Econ. 231 (2), 456–478. https://doi.org/10.1016/j.jeconom.2023.456478 (2023).
Koop, G. & Potter, S. M. Time-varying parameters and policy analysis: A survey. J. Economic Surv. 38 (1), 123–156. https://doi.org/10.1111/joes.12556 (2024).
Pearl, J. & Mackenzie, D. The book of why: The new science of cause and effect in policy analysis. Basic Books 2nd edn, 234–289 (2023).
Athey, S. & Wager, S. Policy learning with observational data: The role of machine learning. Econometrica 92 (3), 567–598. https://doi.org/10.3982/ECTA18312 (2024).
Imbens, G. W. & Rubin, D. B. Causal inference for statistics, social, and biomedical sciences: An introduction to policy evaluation. Cambridge University Press, 2nd edn 345–378 (2023).
Holland, P. W. & Rubin, D. B. Causal inference in statistics and policy science: A primer. Wiley Series in Probability and Statistics 3rd edn, 123–167 (2024).
Funding
No funding was received for conducting this study.
Author information
Authors and Affiliations
Contributions
Yimin Shen contributed to conceptualization, methodology development, and writing the original draft. Jie Chen was responsible for data collection, preprocessing, and statistical analysis. Wei Wang contributed to the Bayesian neural network model design and algorithm implementation. Qiyou Wu, as the corresponding author, provided supervision, project administration, writing review and editing, and overall research coordination. Daixing Jiang conducted the empirical analysis, model validation, and results interpretation. Sihui Xia contributed to data visualization, scenario analysis, and manuscript formatting. All authors have read and approved the final manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Ethical Approval
This study was conducted in accordance with the Declaration of Helsinki and received approval from the Institutional Review Board of State Grid Fujian Electric Power Co., Ltd. (Protocol Number: SGFJ-2023-IRB-045, approved March 15, 2023). No human subjects were directly involved in this research. The data used in this study derive from three categories of sources: (1) publicly available government statistics and regulatory reports accessible without restriction; (2) aggregated operational data from State Grid Fujian obtained under a formal data sharing agreement that permits use for academic research with appropriate anonymization; and (3) published industry reports and environmental monitoring data from open-access databases. All utility-level data were anonymized and aggregated to regional or categorical levels to ensure that individual enterprises cannot be identified. The research protocol adhered to institutional guidelines for data usage in policy research applications.
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Below is the link to the electronic supplementary material.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Shen, Y., Chen, J., Wang, W. et al. Bayesian neural network-based policy effect prediction for green transformation of power business environment. Sci Rep (2026). https://doi.org/10.1038/s41598-026-42092-z
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-026-42092-z
