Generation Expansion Planning (GEP) is a part of expansion planning problem in power systems. The GEP is defined as the problem of determining when, where, what type and how much capacity of new power plants should be constructed over a long term planning horizon to meet forecasted demand. Least cost GEP is known as a highly constrained and non-linear discrete dynamic optimization problem. The main purpose of the problem is to minimize the total cost of generation system expansion plan within pre-specified reliability criteria. In order to achieve this objective, in this thesis, GEP is modeled as an optimization problem in which the objective function is to minimize the total investment, operation, and outage (energy not served) costs of power system as well as salvage value of investment costs. Generation system reliability is assessed and provided by means of Expected Energy Not Served (EENS) and Loss of Load Probability (LOLP) indices. Probabilistic Production Simulation () is done using the Equivalent Energy Function (EEF) method. Production system simulation at each stage of planning horizon calculates the expected energy produced by each unit. The reliability indices LOLP and EENS are also determined by the used simulation procedure. To solve the GEP problem, a new Modified Shuffled Frog Leaping (MSFL) algorithm is proposed in this thesis. The SFL algorithm is a meta-heuristic optimization method that mimics the memetic evolution of a group of frogs seeking for a location with maximum amount of available food. It consists of a frog leaping rule for local search and a memetic shuffling rule for global information exchange. Some issues in the original SFL algorithm may cause it not to be able to find the global optimum in some optimization problems and may get trapped in local optimums. To overcome this problem, a new frog leaping rule and a new strategy for frog distribution into memeplexes is introduced to improve the local exploration and performance of the original SFL algorithm. To show the effectiveness of the MSFL algorithm, it is applied to a test system with 15 existing power plants and 5 types of new candidates, for a 12-years and a 24-years planning horizon. The original SFL algorithm and the Genetic Algorithm (GA) are also applied to solve the GEP problem. Simulation results show the advantages of the proposed MSFL algorithm over the original SFL and GA. The proposed MSFL algorithm could also achieve an order of magnitude of improvement, especially for the GEP problems in larger scale, than other two algorithms. Therefore, the dimensionality is not the key factor and it can be employed as a planning tool for long-term generation expansion planning in a real-scale systems. Keywords: Generation Expansion Planning, Probabilistic Production Simulation, Shuffled Frog Leaping Algorithm, Combinatorial Optimization, Power System Reliability