Extreme icing disasters increasingly undermine the reliability of integrated power and heat networks by causing line outages

supply shortages and sharp thermal load fluctuations. To address these challenges

this paper proposes a comprehensive optimisation framework that exploits the spatiotemporal flexibility of data centres for coordinated electric–thermal demand response under uncertain icing disasters. An improved spatiotemporal inverse distance weighting method combined with a Gaussian copula is first developed to reconstruct the joint spatial temporal dependence of icing variables

whereas a cellular automaton is introduced to capture icing-driven fault propagation and generate representative stochastic scenarios. Based on these scenarios

an integrated electric–thermal demand response model is formulated

jointly leveraging data centre load migration and waste heat recovery

with an objective function incorporating operating cost

response benefits

icing-related damage and social penalties. A case study in Chun'an County

Zhejiang Province

validates the proposed framework. Compared with fixed icing assumptions and decoupled power–heat scheduling

the method reduces residential electricity load shedding to 36.34%

increases data centre task migration utilisation to 67.21% and improves waste heat recovery efficiency to 73.45%. The results demonstrate that data centre flexibility can significantly enhance the resilience and reliability of multi-energy systems under extreme icing disasters.