MMF-EDRM: A Multi-Modal Fusion and Dual-Risk Modeling Framework for Renewable Energy Project Financing and ECM/DCM Pricing

Zijie Zhong; Guyue Gui; Ye Mou

doi:10.63313/EBM.9186

Authors

Zijie Zhong Nanjing University, Nanjing, China Author
Guyue Gui University of Chicago, Chicago, IL, USA Author
Ye Mou Statistics Program, GSAS, Columbia University, New York, NY, USA Author

DOI:

https://doi.org/10.63313/EBM.9186

Keywords:

Multi-Modal Deep Learning, Renewable Energy Investment, Risk-Return Modeling, ECM/DCM Pricing, Multi-Factor Model, Attention Mechanism

Abstract

Renewable energy infrastructure investment faces substantial uncertainty arising from heterogeneous data sources, including environmental variability, policy interventions, and financial market fluctuations. Traditional valuation methods, such as discounted cash flow (DCF) and conventional multi-factor models, are limited in their ability to integrate unstructured information and to jointly capture risk–return dynamics in financing decisions. To address these challenges, this paper proposes MMF-EDRM (Multi-Modal Fusion and Dual-Risk Energy financing model), a unified framework for renewable energy project financing valuation and ECM/DCM pricing. The model integrates multi-modal deep learning with multi-factor statistical modeling to establish a risk–return–pricing linkage for infrastructure assets. Specifically, satellite imagery, meteorological time series, policy documents, and structured financial indicators are encoded through modality-specific networks and fused via a cross-modal attention mechanism. A heteroscedastic dual-head architecture is then employed to jointly estimate expected returns and conditional risk uncertainty. These outputs are further incorporated into an extended multi-factor pricing formulation to derive equity valuation (ECM) and credit spread estimation (DCM). The framework is optimized using a multi-task learning objective that aligns prediction and pricing consistency. Empirical results on the constructed multi-modal renewable energy dataset show that MMF-EDRM reduces RMSE for return prediction by 17.6% compared with CNN-LSTM baselines, while achieving 14.0% improvement in equity pricing error and 12.7% reduction in credit spread prediction error. In addition, the model attains a lower NLL of 0.742, indicating superior risk calibration under volatile policy conditions. These results confirm the effectiveness of the proposed framework.

References

[1] Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks[J]. Advances in neural information processing systems, 2012, 25.

[2] Shan S, Li C, Wang Y, et al. A deep learning model for multi-modal spatio-temporal irradiance forecast[J]. Expert Systems with Applications, 2024, 244: 122925.

[3] Anand S, Sonkar S K. MULTI-SOURCE DATA FUSION FOR WIND FARM POWER FORE-CASTING THROUGH AI[J]. Journal of the Balkan Tribological Association, 2024, 30(3).

[4] Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[J]. Advances in neural information processing systems, 2017, 30.

[5] Fama E F, French K R. Common risk factors in the returns on stocks and bonds[J]. Journal of financial economics, 1993, 33(1): 3-56.

[6] Carhart M M. On persistence in mutual fund performance[J]. The Journal of finance, 1997, 52(1): 57-82.

[7] Li W, Yang R, Hu Z. Forecasting Surface Facilities Investment Based on Factor Analysis and Multiple Regression Analysis[C]//International Field Exploration and Development Conference. Singapore: Springer Nature Singapore, 2023: 334-344.

[8] He K, Wang H, Du J, et al. Forecasting electricity market risk using empirical mode decomposition (EMD)—based multiscale methodology[J]. Energies, 2016, 9(11): 931.

[9] Gu S, Kelly B, Xiu D. Empirical asset pricing via machine learning[J]. The Review of Financial Studies, 2020, 33(5): 2223-2273.

[10] Heaton J B, Polson N G, Witte J H. Deep learning in finance[J]. arXiv preprint arXiv:1602.06561, 2016.

[11] Dixon M F, Halperin I, Bilokon P. Machine learning in finance[M]. New York, NY, USA: Springer International Publishing, 2020.

[12] Kendall A, Gal Y. What uncertainties do we need in bayesian deep learning for computer vision?[J]. Advances in neural information processing systems, 2017, 30.

[13] Xu S, Jiang L, Gu B. Design and Validation of a Smart Neuromorphic System Architecture for Algorithmic Trading[C]//Proceedings of the 2nd International Symposium on Integrated Circuit Design and Integrated Systems. 2025: 127-136.

[14] Fan P, Li H, Hu M. Profit-Oriented Production and Pricing Optimization for Manufacturing Enterprises Using Proximal Policy Optimization[J]. Economics and Management Innovation, 2026, 3(2): 8-17.

[15] Li X, Li Z, Lin X. Automated Implementation of Machine Learning-Based Causal Inference in Product Operations Decision Making[C]//Proceedings of the 2nd International Symposium on Integrated Circuit Design and Integrated Systems. 2025: 204-209.

[16] Li Z, Li X, Lin X. Design and Implementation of a Platform for Business Intelligence Knowledge Mining and Graph Construction Based on Deep Learning[C]//Proceedings of the 2nd International Symposium on Integrated Circuit Design and Integrated Systems. 2025: 137-141.

[17] Jiang G, Zhao S, Zhong Y, et al. Investment advisory robotics 2.0: Leveraging deep neural networks for personalized financial guidance[C]//2025 International Joint Conference on Neural Networks (IJCNN). IEEE, 2025: 1-7.

[18] Luo R, Hu J, Sun Q. Group Anomaly Detection and Risk Control of Commodity Sales Volume Data Based on LSTM-VAE Framework[J]. Journal of Computer, Signal, and System Research, 2025, 2(7): 48-57.