https://doi.org/10.65770/DSSF5027
ABSTRACT
Increasing adoption of artificial intelligence (AI) in healthcare raises energy-demand questions that are especially acute in resource-limited regions. Because operational, application-level energy-metering data for medical AI systems in Sub-Saharan African hospitals are not yet publicly available, this study develops a climate-calibrated simulation framework that couples real facility profiles (bed capacity for six Nigerian teaching hospitals, from institutional and published sources) with real long-term climate normals (temperature and humidity for each hospital city) and a transparently parameterized medical-AI workload overlay (diurnal, weekly, and holiday patterns informed by prior literature). Statistical models (ARIMA, SARIMA, Holt-Winters exponential smoothing) and machine-learning models (a purpose-built LSTM and a hybrid LSTM-ARIMA) were evaluated on the resulting 58-month (January 2020-October 2024) simulated series using a consistent walk-forward, one-step-ahead protocol. Seasonal statistical models achieved the lowest forecast error (Holt-Winters MAPE=0.36%, SARIMA MAPE=0.37%), outperforming the hybrid LSTM-ARIMA (MAPE=0.51%) and a standalone LSTM (MAPE=0.61%), while the hybrid was most accurate at very short (1-day) horizons. Diurnal peak-hour (0800-1400) demand exceeded off-peak demand by 35%, and dry/harmattan-season demand exceeded the annual mean by 1.4-5.3% depending on institution and local climate. Rather than asserting a single best algorithm, this study’s central contribution is the framework itself: a reusable, real-data-calibrated method for generating defensible medical-AI energy-demand scenarios where metered data do not yet exist, together with comparative evidence that the relative advantage of statistical versus machine-learning forecasting depends on how strongly deterministic the underlying seasonal structure is. The framework and comparative findings offer a starting point for energy planning and policy-making for AI-enabled healthcare across Sub-Saharan Africa (SSA), pending validation against real operational meter data.
References
- [1] Schwartz, R., Dodge, J., Smith, N. A., & Etzioni, O. (2020). Green AI. Communications of the ACM, 63(12), 54-63. https://doi.org/10.1145/3381831
- [2] Wu, C. J., Raghavendra, R., Gupta, U., Acun, B., Ardalani, N., Maeng, K., Chang, G., Behram, F. A., Huang, J., Bai, C., Gschwind, M., Gupta, A., Ott, M., Melnikov, A., Candido, S., Brooks, D., Chauhan, G., Lee, B., Lee, H. H. S., … & Hazelwood, K. (2022). Sustainable AI: Environmental implications, challenges and opportunities. Proceedings of Machine Learning and Systems, 4, 795-813. https://doi.org/10.5555/3520312.3520365
- [3] Adepoju, O. E., Adejumo, A. O., & Oyeleye, B. A. (2023). Readiness of Nigerian teaching hospitals for artificial intelligence deployment: Infrastructure assessment and policy implications. African Journal of Health Policy and Systems, 12(2), 145-162. https://doi.org/10.4314/ajhps.v12i2.8
- [4] Nwosu, C. E., & Okereke, C. (2022). Energy infrastructure challenges in Nigerian teaching hospitals: Implications for digital health transformation. West African Journal of Medicine, 39(8), 897-904. https://doi.org/10.55891/wajm.v39i8.1567
- [5] García-Martín, E., Rodrigues, C. F., Riley, G., & Grahn, H. (2021). Estimation of energy consumption in machine learning. Journal of Parallel and Distributed Computing, 134, 75-88. https://doi.org/10.1016/j.jpdc.2019.07.007
- [6] Li, C., Tao, Y., Ao, W., Yang, S., & Bai, Y. (2023). Improving forecasting accuracy of daily enterprise electricity consumption using a random forest based on ensemble empirical mode decomposition. Energy, 165, 1220-1227. https://doi.org/10.1016/j.energy.2018.10.113
- [7] Dodge, J., Prewitt, T., Tachet des Combes, R., Odmark, E., Schwartz, R., Strubell, E., Luccioni, A. S., Smith, N. A., DeCario, N., & Buchanan, W. (2022). Measuring the carbon intensity of AI in cloud instances. Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency, 1877-1894. https://doi.org/10.1145/3531146.3533234
- [8] Henderson, P., Hu, J., Romoff, J., Brunskill, E., Jurafsky, D., & Pineau, J. (2020). Towards the systematic reporting of the energy and carbon footprints of machine learning. Journal of Machine Learning Research, 21(248), 1-43. https://doi.org/10.5555/3455716.3455964
- [9] Strubell, E., Ganesh, A., & McCallum, A. (2020). Energy and policy considerations for modern deep learning research. Proceedings of the AAAI Conference on Artificial Intelligence, 34(9), 13693-13696. https://doi.org/10.1609/aaai.v34i09.7123
- Patterson, D., Gonzalez, J., Le, Q., Liang, C., Munguia, L. M., Rothchild, D., So, D., Texier, M., & Dean, J. (2022). Carbon emissions and large neural network training. arXiv preprint arXiv:2104.10350. https://doi.org/10.48550/arXiv.2104.10350
- Topol, E. J. (2023). High-performance medicine: The convergence of human and artificial intelligence. Nature Medicine, 25(1), 44-56. https://doi.org/10.1038/s41591-018-0300-7
- Chen, F., Taviani, V., Malkiel, I., Cheng, J. Y., Tamir, J. I., Shaikh, J., Chang, S. T., Hardy, C. J., Pauly, J. M., & Vasanawala, S. S. (2023). Variable-density single-shot fast spin-echo MRI with deep learning reconstruction. Magnetic Resonance in Medicine, 89(4), 1450-1466. https://doi.org/10.1002/mrm.29521
- Campanella, G., Hanna, M. G., Geneslaw, L., Miraflor, A., Silva, V. W. K., Busam, K. J., Brogi, E., Reuter, V. E., Klimstra, D. S., & Fuchs, T. J. (2020). Clinical-grade computational pathology using weakly supervised deep learning on whole slide images. Nature Medicine, 25(8), 1301-1309. https://doi.org/10.1038/s41591-019-0508-1
- Sendak, M. P., Gao, M., Brajer, N., & Balu, S. (2020). Presenting machine learning model information to clinical end users with model facts labels. NPJ Digital Medicine, 3(1), 41. https://doi.org/10.1038/s41746-020-0253-3
- Box, G. E. P., Jenkins, G. M., Reinsel, G. C., & Ljung, G. M. (2015). Time series analysis: Forecasting and control (5th ed.). John Wiley & Sons. https://doi.org/10.1002/9781118619193
Download all article in PDF
