It is well known that the rate of penetration is a key function for drilling engineers since it is directly related to the final well cost, thus reducing the non-productive time is a target of interest for all oil companies by optimizing the drilling processes or drilling parameters. These drilling parameters include mechanical (RPM, WOB, flow rate, SPP, torque and hook load) and travel transit time. The big challenge prediction is the complex interconnection between the drilling parameters so artificial intelligence techniques have been conducted in this study to predict ROP using operational drilling parameters and formation characteristics. In the current study, three AI techniques have been used which are neural network, fuzzy inference system and genetic algorithm. An offset field data was collected from mud logging and wire line log from East Baghdad oil field south region to build the AI models, including datasets of two wells: well 1 for AI modeling and well 2 for validation of the obtained results. The types of interesting formations are sandstone and shale (Nahr Umr and Zubair formations). Nahr Umr and Zubair formations are medium –harder. The prediction results obtained from this study showed that the ANN technique can predict the ROP with high efficiency as well as FIS technique could achieve reliable results in predicting ROP, but GA technique has shown a lower efficiency in predicting ROP. The correlation coefficient and RMSE were two criteria utilized to evaluate and estimate the performance ability of AI techniques in predicting ROP and comparing the obtained results. In the Nahr Umr and Zubair formations, the obtained correlation coefficient values for training processes of ANN, FIS and GA were 0.94, 0.93, and 0.76 respectively. Data sets from another well (well 2) in the same field of interest were utilized to validate of the developed models. Datasets of well 2 were conducted against sandstone and shale formations (Nahr Umr and Zubair formations). The results revealed a good matching between the actual rate of penetration values and the predicted ROP values using two artificial intelligence techniques (neural network, and fuzzy inference technique). In contrast, the genetic algorithm model showed overestimation/ underestimation of the rate of penetration against sandstone and shale formations. This means that the optimum prediction of rate of penetration can be obtained from neural network model rather than using genetic algorithm and genetic algorithm techniques. The developed model can be successfully used to predict the rate of penetration and optimize the drilling parameters, achieving reduce the cost and time of future wells that will be drilled in the East Baghdad Iraqi oil field.
This work is licensed under a