Some of the challenges associated with the multi-objective optimization on a modern power system were addressed in this work. Optimization of reactive resources was performed in order to simultaneously optimize several criteria: transmission losses, distribution losses, voltage stability, etc. The optimization was performed simultaneously on the entire power system; transmission and distribution subsystems included. The inherent physical complexity of modeling together transmission and distribution systems was considered first. After considering all pros and cons for such a task, a model of the entire power system was successfully developed. The inherent mathematical complexity of high-dimensional optimization space was handled by introducing the decoupling principle. System is first decoupled in several independent models and optimizations were performed on each part of the system. An algorithm is developed that properly combines the independent solutions to reach the overall system optima. The principle of algorithm synthesis is used to reduce the size of the solution space. Deterministic algorithms are used to locate the local optima which are subsequently refined by probabilistic algorithm. The algorithm is applied on a "real-life" test system and it is shown that the obtained solutions outperform the solution obtained with the conventional algorithms.