a. moradzadeh

In determining the optimal points of production drilling, it is important to identify areas of suitable reservoir quality. For this purpose, the use of geochemical data, which is usually small in number, is common. This data discontinuity creates information gaps. If one uses more continuous data so that its modeling accuracy is suitable, the drilling could be then performed with more success. In this study, seismic and well logs data were used to classify the quality of gas facies by two non-parametric statistical (Parzen) and supervised deep learning techniques (long-term short-term memory network (LSTM)). The LSTM network was then also optimized by two heuristic optimization methods (Imperialistic competition algorithm and Whale algorithm). The obtained results indicate that both methods produce good results in classification so that the modeling accuracy of gas facies quality using supervised deep learning technique (87%) is more than that of the non-parametric Parzen (83%) method. Moreover, the application of optimization algorithms has increased the classification accuracy. The best accuracy is related to the LSTM network optimized with the imperialistic competition algorithm (90%). Geochemical reports and well cores data show the high validity of these models.

The Frequency Response Analysis (FRA) technique has advantages in identifying faults related to power transformers, but it suffers from the interpretation of frequency responses. This paper presents an approach based on statistical indices and Artificial Neural Network (ANN) methods to interpret frequency responses. The proposed procedure divides frequency responses into four frequency regions based on frequency resonances and anti-resonances. Then, Lin’s Concordance Coefficient (LCC) index is used as one of the most appropriate numerical indices to extract features of the four frequency regions. Finally, the Multilayer Perceptron (MLP) neural network is trained by the extracted features to identify and differentiate the types of winding faults. Besides, other intelligent algorithms such as Support Vector Machine (SVM), Extreme Learning Machine (ELM), Probabilistic Neural Network (PNN), and Radial Basis Function (RBF) neural network have been employed to compare the classification results. The proposed techniques have been practically implemented. The Axial Displacement (AD) and Disk Space Variation (DSV) faults are applied as two common mechanical faults in different locations and intensities on the 20kV windings of a 1.6MVA distribution power transformer and their corresponding frequency responses are calculated. Frequency responses calculated from the AD and DSV faults constitute the MLP input data set. The network is trained with part of the input data, and the rest of the data is allocated to validate and test the network. The results show that the suggested method has more proper performance than others using the phase component of the frequency responses in interpreting frequency responses and separation and identifying various mechanical fault types of transformer windings.

One of the most essential factors involved in unconventional gas reserves for drilling and production is a suitable quality facies determination. The direct core and geochemical analyses are the most common methods used for studying this quality. Due to the lack of this data and the high cost, the researchers have recently resorted to the indirect methods that use the common data of the reservoir (including petro-physical logs and seismic data). One of the major problems in using these methods is that the complexities of these reproducible repositories cannot be accurately modeled. In this work, the quality of facies in shale gas is zoned using the deep learning technique. The applied method is long short-term memory (LSTM) neural network. In this scheme, the features required for zoning are automatically extracted and used to model the reservoir complexities properly. The results of this work show that zoning is done with an appropriate accuracy (86%) using the LSTM neural network, while it is 78% for a conventional intelligent MLP network. This specifies the superior accuracy of the deep learning method.