This work investigates the performance of intelligent models on the forecasting of permeability in heterogeneous reservoirs. Production logs are used to compute loss functions for regression in the algorithms optimization process. A flow profile interpretation method is used to remove wellbore skin effects from the measured flow rate. Additionally, a segmentation technique is applied to high-resolution ultrasonic image logs which provide not only the image of mega and giga pore systems but also identify the permeable facies along the reservoir. The image segmentation jointly with other borehole logs provides the necessary input data for the models training process. The estimations presented herein demonstrate the algorithms ability to learn non-linear relationships between geological input variables and a reservoir dynamic data even if the actual physical relationship is complex and not known a priori. Though the preprocessing stages of the procedure involve some data interpretation expertise, the algorithms can easily be coded in any programming language, requiring no assumptions on physics in advance. The proposed procedure provides more accurate permeability curves than those obtained from conventional methods, which may fail to predict the permeability measured on drill stem tests (DSTs) conducted in dual-porosity reservoirs. The novelty of this work is to incorporate dynamic production logging (PL) data into the permeability estimation workflow using machine learning algorithms.