In the evolving landscape of modern warfare, an escalating importance is being attributed to vehicle-mounted phased array radars due to their imperative role in information acquisition and mastery. A significant limitation to their performance lies in the challenges of heat dissipation and resultant thermal deformation of the antenna array surface. Current methodologies for predicting thermal deformation and techniques for heat dissipation exhibit inherent shortcomings, rendering them inadequate for varying radar systems and diverse operational environments. To address this, heat transfer characteristics of microchannels on the antenna array surface of vehicle-mounted phased array radar were investigated. A novel design paradigm for the structural optimisation of the antenna array surface was subsequently proposed. In addition, a cutting-edge thermal deformation prediction method utilising a Particle Swarm Optimization (PSO)-Genetic Algorithm (GA)-Back Propagation Neural Network (BPNN) synergy was presented, aiming to overcome the deficiencies observed in existing prediction models. The insights garnered from this research offer potential pathways for the refinement of radar antenna array structures and are poised to pave the way for future advancements in this domain.