AI-based intelligent battery design for space and nuclear applications

Öz

Space and nuclear missions impose extreme and often contradictory requirements on energy storage: temperatures as low as −270 °C, total ionizing doses exceeding 300 krad, mission durations spanning decades, and mass budgets measured in single kilograms. No single technology satisfies all these constraints simultaneously, necessitating a principled, data-driven approach to technology selection and design. We present the Unified Battery Intelligence Framework (UBIF), a five-phase AI pipeline integrating data from NASA, IAEA, NREL, and the Materials Project. Phase 1 builds a structured database of 10 technologies across 15 features. Phase 2 trains a Gradient Boosting Regressor (R^2=0.895±0.031) and a Random Forest Classifier (F1=1.000) for mission-specific selection. Phase 3 applies Pareto optimization to 2,000 candidates, yielding 17 non-dominated designs. Phase 4 builds digital twins for RTG and Li-ion systems with Isolation Forest anomaly detection (contamination 3%). Phase 5 deploys Bayesian Optimization (BO) with a Gaussian Process surrogate over a thermoelectric material space. The framework produces: a validated mission recommendation engine benchmarked against five mission profiles; a Pareto design frontier spanning 180–510 Wh kg-1 at 2–18 years lifetime; reliable anomaly detection within two data samples of onset; and BO-proposed thermoelectric candidates with predicted ZT > 2.71 — a 4.2× improvement over heritage SiGe (ZT=0.65). These results converge in the conceptual BTES-X hybrid design combining betavoltaic, thermoelectric, and solid-state subsystems, targeting 30-year, 1-Mrad-tolerant operation at 45 Wh kg-1 system level. UBIF demonstrates that an integrated AI pipeline — spanning data engineering, supervised learning, Pareto optimization, digital twin modelling, and material discovery — enables a transition from reactive technology selection to proactive intelligent design of energy storage for extreme environments.

Anahtar Kelimeler

• Battery intelligence framework
• Digital twin
• Multi-objective optimization
• Radioisotope thermoelectric generators
• Space energy storage

Uzay ve nükleer uygulamalar için yapay zeka tabanlı akıllı pil tasarımı

Öz

Uzay ve nükleer görevler, enerji depolama konusunda aşırı ve çoğu zaman çelişkili gereksinimler ortaya koymaktadır: -270 °C kadar düşük sıcaklıklar, 300 krad'ı aşan toplam iyonlaştırıcı dozlar, on yıllarca süren görev süreleri ve tek kilogram cinsinden ölçülen kütle bütçeleri. Hiçbir teknoloji bu kısıtlamaların tümünü aynı anda karşılayamaz; bu da teknoloji seçimi ve tasarımı için ilkeli, veri odaklı bir yaklaşımı gerektirir. NASA, IAEA, NREL ve Malzeme Projesi'nden gelen verileri entegre eden beş aşamalı bir yapay zeka hattı olan Birleşik Batarya Zeka Çerçevesi'ni (UBIF) sunuyoruz. 1. Aşama, 15 özellik üzerinden 10 teknolojiden oluşan yapılandırılmış bir veri tabanı oluşturur. 2. Aşama, göreve özgü seçim için bir Gradyan Artırma Regresörü (R^2=0.895±0.031) ve bir Rastgele Orman Sınıflandırıcısı (F1=1000) eğitir. 3. Aşama, 2000 adaya Pareto optimizasyonu uygulayarak 17 baskın olmayan tasarım üretir. 4. Aşama, İzolasyon Ormanı anomali tespiti (kirlenme %3) ile RTG ve Li-ion sistemleri için dijital ikizler oluşturur. 5. Aşama, termoelektrik malzeme uzayında Gauss Süreci vekil modeli ile Bayes Optimizasyonu (BO) uygular. Çerçeve şunları üretir: beş görev profiline göre kıyaslanabilen doğrulanmış bir görev öneri motoru; 2-18 yıl ömürde 180-510 Wh kg-1'i kapsayan bir Pareto tasarım sınırı; başlangıçtan itibaren iki veri örneği içinde güvenilir anomali tespiti; ve tahmini ZT > 2,71 olan BO tarafından önerilen termoelektrik adaylar — eski SiGe'ye (ZT=0,65) göre 4,2 kat iyileşme. Bu sonuçlar, 45 Wh kg-1 sistem seviyesinde 30 yıl, 1 Mrad toleranslı çalışma hedefleyen betavoltaik, termoelektrik ve katı hal alt sistemlerini birleştiren kavramsal BTES-X hibrit tasarımında birleşir. UBIF, veri mühendisliği, denetimli öğrenme, Pareto optimizasyonu, dijital ikiz modelleme ve malzeme keşfini kapsayan entegre bir yapay zeka hattının, aşırı ortamlar için enerji depolama sistemlerinin reaktif teknoloji seçiminden proaktif akıllı tasarımına geçişi mümkün kıldığını göstermektedir.

Anahtar Kelimeler

• Çok amaçlı optimizasyon
• Dijital ikiz
• Pil zekası çerçevesi
• Radyoizotop termoelektrik jeneratörler
• Uzay enerji depolama

Kaynakça

1. Stone EC, Cummings AC, McDonald FB, Heikkila BC, Lal N et al. Voyager 1 observes low-energy galactic cosmic rays in a region depleted of heliospheric ions. Science 2013; 341 (6142): 150-153. https://doi.org/10.1126/science.1236408
2. Grotzinger JP, Crisp J, Vasavada AR, Anderson RC, Bell CJ et al. Mars Science Laboratory mission and science investigation. Space Science Reviews 2012; 170 (1-4): 5-56. https://doi.org/10.1007/s11214-012-9892-2
3. Farley KA, Williford KH, Stack KM, Bhartia R, Chen A et al. Mars 2020 Mission overview. Space Science Reviews 2020; 216 (8): 142. https://doi.org/10.1007/s11214-020-00762-y
4. Tarascon JM, Armand M. Issues and challenges facing rechargeable lithium batteries. Nature 2001; 414 (6861): 359-367. https://doi.org/10.1038/35104644
5. Bugga RV, Smart MC, Whitacre J, West W. Lithium ion batteries for space applications. 2007 IEEE Aerospace Conference, Big Sky; 1-7, MT, USA; 2007. https://doi.org/ 10.1109/AERO.2007.352728.
6. Jones JP, Jones SC, Billings KJ, Pasalic J, Bugga RV, Krause FC, Smart MC, Brandon EJ. Radiation effects on lithium CFX batteries for future spacecraft and landers. Journal of Power Sources 2020; 471: 228464. https://doi.org/10.1016/j.jpowsour.2020.228464.
7. Bennet GL. Space nuclear power: opening the final frontier. In: 4th International Energy Conversion Engineering Conference; San Diego, CA, USA; 2006. AIAA Paper 2006-4191.
8. Severson KA, Attia PM, Jin N, Perkins N, Jiang B et al. Data-driven prediction of battery cycle life before capacity degradation. Nature Energy 2019; 4 (5): 383-391. https://doi.org/10.1038/s41560-019-0356-8

9. Attia PM, Grover A, Jin N, Severson KA, Markov TM et al. Closed-loop optimisation of fast-charging protocols for batteries with machine learning. Nature 2020; 578 (7795): 397-402. https://doi.org/10.1038/s41586-020-1994-5
10. Furmanchuk A, Saal JE, Doak JW, Olson GB, Choudhary A, Agrawal A. Prediction of seebeck coefficient for compounds without restriction to fixed stoichiometry: A machine learning approach. J Comput Chem. 2018; 39 (4): 191-202. https://doi.org/10.1002/jcc.25067.
11. NASA. Lunar Gateway Power Architecture Study. NASA/TM-2018-219980. Washington, DC, USA: National Aeronautics and Space Administration, 2018.
12. NREL. Battery Performance Data and Life Cycle Analysis. NREL/TP-5400-75077. Golden, CO, USA: National Renewable Energy Laboratory, 2019.
13. IAEA. Nuclear Data Services: Decay Data. Vienna, Austria: International Atomic Energy Agency, 2022. Available at: www-nds.iaea.org
14. Jain A, Ong SP, Hautier G, Chen W, Richards WD et al. The Materials Project: a materials genome approach to accelerating materials innovation. APL Materials 2013; 1 (1): 011002. https://doi.org/10.1063/1.4812323
15. Friedman JH. Greedy function approximation: a gradient boosting machine. Annals of Statistics 2001; 29 (5): 1189-1232. https://doi.org/10.1214/aos/1013203451
16. Breiman L. Random forests. Machine Learning 2001; 45 (1): 5-32. https://doi.org/10.1023/A:1010933404324
17. Deb K, Pratap A, Agarwal S, Meyarivan T. A fast and elitist multiobjective genetic algorithm: NSGA-II. IEEE Transactions on Evolutionary Computation 2002; 6 (2): 182-197. https://doi.org/10.1109/4235.996017
18. Christensen J, Albertus P, Sanchez-Carrera RS, Lohmann T, Kozinsky B et al. A critical review of Li/air batteries. Journal of the Electrochemical Society 2012; 159 (2): R1-R30. https://doi.org/10.1149/2.086202jes
19. McKay MD, Beckman RJ, Conover WJ. A comparison of three methods for selecting values of input variables in the analysis of output from a computer code. Technometrics 1979; 21 (2): 239-245. https://doi.org/10.2307/1268522
20. Hammel TE, Bennett R, Otting W, Fanale S. Multi-mission radioisotope thermoelectric generator (MMRTG) program overview. In: Proceedings of the 10th Symposium on Space Nuclear Power and Propulsion; Albuquerque, NM, USA; 1993. pp. 1-5.
21. Smith K et al. Lithium-ion battery life model with electrode characterisation. Journal of Power Sources 2010; 195 (8): 2851-2864. https://doi.org/10.1149/1945-7111/ac2ebd
22. Wang J et al. Cycle-life model for graphite-LiFePO4 cells. Journal of Power Sources 2014; 269: 858-867. https://doi.org/10.1016/j.jpowsour.2010.11.134
23. Liu FT, Ting KM, Zhou ZH. Isolation forest. In: Proceedings of the 8th IEEE International Conference on Data Mining; Pisa, Italy; 2008. pp. 413-422. https://doi.org/10.1109/ICDM.2008.17
24. Williams CKI, Rasmussen CE. Gaussian Processes for Machine Learning. Cambridge, MA, USA: MIT Press, 2006.
25. Shahriari B, Swersky K, Wang Z, Adams RP, de Freitas N. Taking the human out of the loop: a review of Bayesian optimisation. Proceedings of the IEEE 2016; 104 (1): 148-175. https://doi.org/10.1109/JPROC.2015.2494218
26. Snyder GJ, Toberer ES. Complex thermoelectric materials. Nature Materials 2008; 7 (2): 105-114. https://doi.org/10.1038/nmat2090
27. He J, Tritt TM. Advances in thermoelectric materials research: looking back and moving forward. Science 2017; 357 (6358): eaak9997. https://doi.org/10.1126/science.aak9997
28. Zhao LD, Lo SH, Zhang Y, Sun H, Tan G et al. Ultralow thermal conductivity and high thermoelectric figure of merit in SnSe crystals. Nature 2014; 508 (7496): 373-377.
29. Slack GA. New materials and performance limits for thermoelectric cooling. In: Rowe DM (editor). CRC Handbook of Thermoelectrics. Boca Raton, FL, USA: CRC Press, 1995. pp. 407-440.
30. Grieves M, Vickers J. Digital twin: mitigating unpredictable, undesirable emergent behaviour in complex systems. In: Kahlen FJ, Flumerfelt S, Alves A (editors). Transdisciplinary Perspectives on Complex Systems. Cham, Switzerland: Springer, 2017. pp. 85-113.
31. Ding JF, Zhang YT, Xu R, Zhang R et al. Review on lithium metal anodes towards high energy density batteries. Green Energy & Environment 2023; 8 (6): 1509-1530. https://doi.org/10.1016/j.gee.2022.08.002