Explainable AI-based Multi-Branch Hybrid Model for Short-term Energy Demand Forecasting

Akpan; E. J.; Sako; D. J. S.; Nwiabu; N. D.

doi:10.56201/wjimt.v10.no10.2026.pg127.139

📄 Download Paper

Explainable AI-based Multi-Branch Hybrid Model for Short-term Energy Demand Forecasting

Akpan, E. J., Sako, D. J. S., Nwiabu, N. D.

Abstract

Energy demand forecasting occupies a fairly central place in modern power system operation, and yet two problems keep resurfacing in the literature devoted to it. Weather variables, more often than not, get folded into forecasting models only partially, and the deep learning architectures that now dominate the field remain, for the most part, opaque to the people who must actually rely on their output. This paper presents an explainable multi-branch hybrid Convolutional Neural Network and Long Short-Term Memory (CNN-LSTM) model for energy demand forecasting in a smart grid system. Historical demand, weather and calendar information, and a set of engineered lag and rolling statistical features were each routed through a separate convolutional-recurrent branch before the three resulting representations were concatenated and passed through fully connected layers to produce a twenty-four-hour-ahead forecast. Interpretability was addressed through SHAP (Shapley Additive Explanations), applied by way of a GradientExplainer, so that predictions could be attributed to individual branches and features across the forecast horizon rather than at a single fixed point alone. The proposed model was benchmarked against three baselines trained under matched conditions, a standalone LSTM, CNN, and a sequential CNN- LSTM, using hourly data drawn from Panama's national electricity grid. On aggregate test-set metrics, the multi-branch model (RMSE 86.69, MAE 60.687, MAPE 5.169%, R2 0.780) performed almost identically to, though not decisively better than, the considerably simpler LSTM baseline (RMSE 86.33, MAE 60.465, MAPE 5.173%, R2 0.782). SHAP attribution showed that historical demand accounted for roughly 63.6% of overall importance, weather and calendar variables for 24.3%, and engineered lag features for the remaining 12.2%, with this ranking shifting toward cyclical hour-of-day encodings as the horizon lengthened.

Keywords

Grid systemenergy demandexplainable AICNNLSTMelectricityforecasting

References

Abedin, J., Ikram, A. I., Islam, M. R., Islam, M. J., Ahemad, F., & Elias, M. (2025). Deep learning model for electrical load forecasting considering meteorological effect. Paper presented at the 2023 International Conference on Next-Generation Computing, IoT and Machine Learning (NCIM 2025), Gazipur-1707, Bangladesh. Abimbola, O. (2025). Energy demand forecasting with deep learning for smart cities. International Journal of Research Publication and Reviews, 6(4), 5950-5959. Aguiar-Pérez, J. M., & Pérez-Juárez, M. Á. (2023). An insight of deep learning based demand forecasting in smart grids. Sensors, 23(3), 1467. Ariyo, B. O., Adesina, L. M., Ogunbiyi, O., Musa, A., Ojuolape, B. J., & Balogun, M. O. (2025). AI-driven demand forecasting for enhanced energy management in renewable microgrids: A hybrid LSTM-CNN approach. Journal of Engineering Research and Reviews, 2(1), 1- 11. Box, G. E. P., Jenkins, G. M., & Reinsel, G. C. (2015). Time series analysis: Forecasting and control (5th ed.). Wiley. Deb, C., Zhang, F., Yang, J., Lee, S. E., & Shah, K. W. (2017). A review on time series forecasting techniques for building energy consumption. Renewable and Sustainable Energy Reviews, 74, 902-924. Doshi Velez, F., & Kim, B. (2017). Towards a rigorous science of interpretable machine learning. arXiv preprint arXiv:1702.08608. Eren, Y., & Küçükdemiral, İ. (2024). A comprehensive review on deep learning approaches for short-term load forecasting. Renewable and Sustainable Energy Reviews, 189, 114031. Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8), 1735-1780. Hong, T., & Fan, S. (2016). Probabilistic electric load forecasting: A tutorial review. International Journal of Forecasting, 32(3), 914-938. Karima, N. N., Ahmad, S., Islam, A., Huda, A. S. N., Awalin, L. J., Halim, S. A., Mokhlis, H., Ali, M. S., & Mubarok, S. (2026). Enhancing short-term load forecasting using hyperparameter-optimized deep learning approaches. Energies, 19(3), 705. Kim, T.-Y., & Cho, S.-B. (2019). Predicting residential energy consumption using CNN-LSTM neural networks. Energy, 182, 72-81. Lago, J., Marcjasz, G., De Schutter, B., & Weron, R. (2021). Forecasting day-ahead electricity prices: A review of state-of-the-art algorithms, best practices and an open-access benchmark. Applied Energy, 293, 116983. Nabavi, S. A., Mohammadi, S., Motlagh, N. H., Tarkoma, S., & Geyer, P. (2024). Deep learning modeling in electricity load forecasting: Improved accuracy by combining DWT and LSTM. Energy Reports, 12, 2873-2900. Qureshi, M., Arbab, M. A., & Rehman, S. ur. (2024). Deep learning-based forecasting of electricity consumption. Scientific Reports, 14, 6489. Sarker, M. A. A., Shanmugam, B., Azam, S., & Thennadil, S. (2024). Enhancing smart grid load forecasting: An attention-based deep learning model integrated with federated learning and XAI for security and interpretability. Intelligent Systems with Applications, 23, Article 200422. Shaikh, A. K., Nazir, A., Khan, I., & Shah, A. S. (2022). Short term energy consumption forecasting using neural basis expansion analysis for interpretable time series. Scientific Reports, 12, 22562. Shi, H., Miao, K., & Ren, X. (2023). Short-term load forecasting based on CNN-BiLSTM with Bayesian optimization and attention mechanism. Concurrency and Computation: Practice and Experience, 35(17), e6676. Taylor, J. W., & McSharry, P. E. (2007). Short-term load forecasting methods: An evaluation based on European data. IEEE Transactions on Power Systems, 22(4), 2213-2219. Ucar, M. T., & Kaygusuz, A. (2025). Short-term energy consumption forecasting analysis using different optimization and activation functions with deep learning models. Applied Sciences, 15(12), 6839. Ullah, K., Ahsan, M., Hasanat, S. M., Haris, M., Yousaf, H., Raza, S. F., Tandon, R., Abid, S., & Ullah, Z. (2024). Short-term load forecasting: A comprehensive review and simulation study with CNN-LSTM hybrids approach. IEEE Access, 12, 111858-111881. Wang, Y., Chen, Q., Kang, C., & Xia, Q. (2019). Clustering of electricity consumption behavior dynamics toward big data applications. IEEE Transactions on Smart Grid, 7(5), 2437-2447. Yang, X., Zhao, S., Li, K., Chen, W., Zhang, S., & Chen, J. (2025). An optimized method for short-term load forecasting based on feature fusion and ConvLSTM-3D neural network. Frontiers in Energy Research, 12, Article 1501963. Zhang, G., Eddy Patuwo, B., & Hu, M. Y. (2018). Forecasting with artificial neural networks: The state of the art. International Journal of Forecasting, 14(1), 35-62. Zhang, Y., Liu, Z., & Wang, J. (2021). Deep learning for smart grid energy forecasting: A comparative study of LSTM and Transformer models. Renewable and Sustainable Energy Reviews, 144, 111024.

Explainable AI-based Multi-Branch Hybrid Model for Short-term Energy Demand Forecasting

Abstract

Keywords

References

More Articles from WORLD JOURNAL OF INNOVATION AND MODERN TECHNOLOGY