The optimization of well placement in a coalbed methane (CBM) reservoir is challenging and intricate work due to the large number of variables and geological uncertainties. To address these challenges, this paper presents a framework that integrates a reservoir simulator into the particle swarm optimization (PSO) algorithm. The application of the PSO algorithm can greatly reduce optimization time and work volume. In this study, optimizations were conducted of the placement of single and multiple wells by maximizing net present value (NPV) in a synthetic reservoir. The optimizations tracked a field application in a CBM district located in the southeast of Ordos basin. A comparison of optimal well placements, which were determined by the PSO and manual trials respectively, showed complete concurrence for a single well. The effect of swarm size on the convergence speed to an optimal location was analyzed. The results indicated that a minimum swarm size of 10 particles is required to guarantee convergence to the global optima. For the case of multiwell placement, we compared the NPV of the optimized well placement determined by the PSO with 1200 randomly selected well placements and found that none of the randomly set wells surpassed the optimized well placement. The optimization results indicated that higher permeability and well interference have a positive effect on the optimal location. After history matching, the rearrangement of 10 vertical wells and an optimization of infilling scenarios were performed in a 5-year-old CBM district located in the southeast of Ordos basin. The optimization results demonstrated that cumulative gas production (CGP) increased by 22.01%, while cumulative water production (CWP) remained nearly unchanged after optimization. Optimal locations tend to be in regions with higher permeability and/or gas content. The NPVs in six infill scenarios exhibited an initial increase, but later exhibited a decline as the number of infill wells increased from zero to seven. A peak NPV value occurred at one infill well. (C) 2012 Elsevier B.V. All rights reserved.