This paper presents an approach to topology, shape, and size optimization of transmission towers. The key feature of the presented approach is the independent variation of the number and heights of tower panels. For that purpose, a modified version of a Genetic Algorithm with a variable length genome is utilized. A novel method of two-stage mutation is developed for handling a variable length genome. Structural optimization of a 110 kV transmission tower is performed as a demonstration with the key result of 10% cost reduction.