Estimation of reservoir rock properties from conventional well log data by using a hybrid particle swarm optimization and neural network approach

The geomechanical and petrophysical parameters of the reservoir such as shear wave velocity, porosity and permeability are regarded as the most important elements in estimating reserves, reservoir simulation, and overall field exploitation strategies. Recently, several methods of artificial intelligence techniques have been used to predict this parameter by using well log data. However, predicting the characteristics of heterogeneous reservoirs always has been facing many problems and an appropriate response is rarely achieved. In this paper, a new methodology is presented for reservoir parameters (shear wave velocity, porosity and permeability) estimation by combining artificial neural network and Particle Swarm optimization (PSO) in Asmari formation of mansuri oilfield. Performance of proposed hybrid scheme was evaluated by comparing the results with the Neural Network and Nero-Fuzzy methods as well as hybrid genetic algorithm–neural network strategy (GA–NN). Comparison of the results shows that PSO-ANN outperforms all the other methods and it can be considered as a powerful tool for reservoir parameters estimation, especially in cases where a precise estimation criterion is crucial.