<p>In view to the common challenges in thermal-fluid topology optimization, including the high dimensionality of the design space, mesh dependence, and topological feature control, this study develops a thermal-fluid topology optimization approach based on anisotropic Material Field Series Expansion (aMFSE). This method characterizes flow channel topology by constructing anisotropic high-dimensional bounded material field functions with spatial correlation. Utilizing the Karhunen–Loève series expansion, the material field is parameterized into a linear combination of characteristic modes and their corresponding coefficients. The modal truncation strategy is employed to truncate low-order modes, reducing the dimension of the optimization design space. Based on the discrete adjoint method, a gradient-driven thermal-fluid topology optimization framework is established, effectively addressing the thermal-fluid optimization problem that couples the Navier–Stokes equations with the energy equation. The effectiveness of the proposed method is demonstrated through several 2D and 3D optimization cases. Results indicate that the inherent spatial correlation of the material field function produces smooth flow channel boundaries without the need for additional filtering techniques. Under identical volume fraction constraints, the aMFSE method achieves a 1.75% reduction in the objective function compared to the density-based approach. Moreover, by decoupling the design variables from the computational mesh, the aMFSE method reduces the number of design variables by 92.74%, greatly enhancing computational efficiency. Additionally, by adjusting the correlation length components in different directions, topological characteristics can be effectively controlled. The robustness of the aMFSE framework is further validated through flow channel designs under various thermal loading conditions.</p>