Lightweight aggregate concrete (LWAC) enhances sustainability in the construction industry. Its lower density reduces transportation-related emissions, while its improved thermal insulation contributes to energy efficiency in buildings by reducing heating and cooling demands and improving indoor thermal comfort. In addition, LWAC exhibits superior fire resistance, durability, and high-temperature performance compared to conventional concrete. However, due to its heterogeneous nature, LWAC exhibits complex thermal behaviour that cannot be accurately captured using homogenised material assumptions. At the mesoscale, the size, distribution, and thermal properties of lightweight aggregates embedded in the mortar matrix play a crucial role in heat transfer mechanisms. Therefore, advanced mesoscale modelling approaches are required to properly understand and optimise its thermal performance. This study presents a 3D mesoscale numerical model to evaluate the effective thermal conductivity of LWAC incorporating expanded clay aggregates. The material is modelled as a heterogeneous composite consisting of a cement-based mortar matrix and randomly distributed spherical lightweight aggregates. Aggregate granulometry is obtained from sieving tests and described using the Rosin–Rammler function. This granulometry and the laboratory mix proportions are used to develop a MATLAB script that generates Representative Volume Elements (RVEs) using a take-and-place algorithm. The generated RVEs are imported into a Finite Element framework to simulate steady-state heat transfer. The thermal conductivity of the mortar matrix is obtained experimentally using a modified transient plane source technique, while the thermal properties of the lightweight aggregates are taken from manufacturer data. These parameters are used as model inputs. The numerically predicted effective thermal conductivity of LWAC is validated against experimental measurements performed on LWAC specimens using the same technique. The proposed hybrid experimental–numerical approach enables an accurate prediction of LWAC thermal behaviour while reducing experimental effort, supporting the development of energy-efficient construction materials through mesoscale analysis.