Development of Artificial Neural Networks (ANNs) to Synthesize Petrophysical Well Logs

Porosity is one of the fundamental petrophysical properties which should be evaluated for hydrocarbon bearing reservoirs. Petrophysical well logs are the most essential instruments for the evaluation of hydrocarbon reservoirs. There are three main petrophysical logging tools for porosity determination namely: neutron, density and sonic well logs. Porosity can be determined using each of these tools; however, a precise analysis requires a complete set of these tools. Log sets are commonly either incomplete or unreliable for many reasons (i.e. incomplete logging, measurement errors and loss of data owing to unsuitable data storage). To overcome this issue, the current study presents an intelligent technique using Artificial Neural Networks (ANN) to synthesize petrophysical well logs including: neutron, density and sonic logs. To accomplish this, the petrophysical well logs data collected from six wells was utilized for constructing optimum ANN model and a seventh well data from the field was employed to evaluate the reliability of the model. The proposed methodology is presented with an application to field information of a carbonate oil reservoir, located in Persian Gulf, Iran. The corresponding correlation was obtained through the comparison of synthesized log values to real log amounts. The results demonstrate that ANNs are successful in synthesizing petrophysical well logs with a high degree of accuracy.