Jane Labadin

Feature Selection based on Mutual Information

The application of machine learning models such as support vector machine (SVM) and artificial neural networks (ANN) in predicting reservoir properties has been effective in the recent years when compared with the traditional empirical methods. Despite that the machine learning models suffer a lot in the faces of uncertain data which is common characteristics of well log dataset. The reason for uncertainty in well log dataset includes a missing scale, data interpretation and measurement error problems. Feature Selection aimed at selecting feature subset that is relevant to the predicting property. In this paper a feature selection based on mutual information criterion is proposed, the strong point of this method relies on the choice of threshold based on statistically sound criterion for the typical greedy feedforward method of feature selection. Experimental results indicate that the proposed method is capable of improving the performance of the machine learning models in terms of prediction accuracy and reduction in training time.

A least-square-driven functional networks type-2 fuzzy logic hybrid model for efficient petroleum reservoir properties prediction

Various computational intelligence techniques have been used in the prediction of petroleum reservoir properties. However, each of them has its limitations depending on different conditions such as data size and dimensionality. Hybrid computational intelligence has been introduced as a new paradigm to complement the weaknesses of one technique with the strengths of another or others. This paper presents a computational intelligence hybrid model to overcome some of the limitations of the standalone type-2 fuzzy logic system (T2FLS) model by using a least-square-fitting-based model selection algorithm to reduce the dimensionality of the input data while selecting the best variables. This novel feature selection procedure resulted in the improvement of the performance of T2FLS whose complexity is usually increased and performance degraded with increased dimensionality of input data. The iterative least-square-fitting algorithm part of functional networks (FN) and T2FLS techniques were combined in a hybrid manner to predict the porosity and permeability of North American and Middle Eastern oil and gas reservoirs. Training and testing the T2FLS block of the hybrid model with the best and dimensionally reduced input variables caused the hybrid model to perform better with higher correlation coefficients, lower root mean square errors, and less execution times than the standalone T2FLS model. This work has demonstrated the promising capability of hybrid modelling and has given more insight into the possibility of more robust hybrid models with better functionality and capability indices.

Improving the prediction of petroleum reservoir characterization with a stacked generalization ensemble model of support vector machines

Despite successful applications of ensembles, the petroleum industry has not benefited enough.SVM is promising but its performance depends mostly on the regularization parameter.We propose an SVM ensemble with diverse opinions on the regularization parameter.The proposed model outperformed Random Forest but competitive with SVM Bagging.There is great potential for ensemble models in petroleum reservoir characterization. The ensemble learning paradigm has proved to be relevant to solving most challenging industrial problems. Despite its successful application especially in the Bioinformatics, the petroleum industry has not benefited enough from the promises of this machine learning technology. The petroleum industry, with its persistent quest for high-performance predictive models, is in great need of this new learning methodology. A marginal improvement in the prediction indices of petroleum reservoir properties could have huge positive impact on the success of exploration, drilling and the overall reservoir management portfolio. Support vector machines (SVM) is one of the promising machine learning tools that have performed excellently well in most prediction problems. However, its performance is a function of the prudent choice of its tuning parameters most especially the regularization parameter, C. Reports have shown that this parameter has significant impact on the performance of SVM. Understandably, no specific value has been recommended for it. This paper proposes a stacked generalization ensemble model of SVM that incorporates different expert opinions on the optimal values of this parameter in the prediction of porosity and permeability of petroleum reservoirs using datasets from diverse geological formations. The performance of the proposed SVM ensemble was compared to that of conventional SVM technique, another SVM implemented with the bagging method, and Random Forest technique. The results showed that the proposed ensemble model, in most cases, outperformed the others with the highest correlation coefficient, and the lowest mean and absolute errors. The study indicated that there is a great potential for ensemble learning in petroleum reservoir characterization to improve the accuracy of reservoir properties predictions for more successful explorations and increased production of petroleum resources. The results also confirmed that ensemble models perform better than the conventional SVM implementation.

An Improved Mathematical Model for Computing Power Output of Solar Photovoltaic Modules

It is difficult to determine the input parameters values for equivalent circuit models of photovoltaic modules through analytical methods. Thus, the previous researchers preferred to use numerical methods. Since, the numerical methods are time consuming and need long term time series data which is not available in most developing countries, an improved mathematical model was formulated by combination of analytical and numerical methods to overcome the limitations of existing methods. The values of required model input parameters were computed analytically. The expression for output current of photovoltaic module was determined explicitly by Lambert W function and voltage was determined numerically by Newton-Raphson method. Moreover, the algebraic equations were derived for the shape factor which involves the ideality factor and the series resistance of a single diode photovoltaic module power output model. The formulated model results were validated with rated power output of a photovoltaic module provided by manufacturers using local meteorological data, which gave ±2% error. It was found that the proposed model is more practical in terms of precise estimations of photovoltaic module power output for any required location and number of variables used.

Ensemble Learning Model for Petroleum Reservoir Characterization: A Case of Feed-Forward Back-Propagation Neural Networks

Conventional machine learning methods are incapable of handling several hypotheses. This is the main strength of the ensemble learning paradigm. The petroleum industry is in great need of this new learning methodology due to the persistent quest for better prediction accuracies of reservoir properties for improved exploration and production activities. This paper proposes an ensemble model of Artificial Neural Networks (ANN) that incorporates various expert opinions on the optimal number of hidden neurons in the prediction of petroleum reservoir properties. The performance of the ensemble model was evaluated using standard decision rules and compared with those of ANN-Ensemble with the conventional Bootstrap Aggregation method and Random Forest. The results showed that the proposed method outperformed the others with the highest correlation coefficient and the least errors. The study also confirmed that ensemble models perform better than the average performance of individual base learners. This study demonstrated the great potential for the application of ensemble learning paradigm in petroleum reservoir characterization.

Reaction-diffusion generic model for mosquito-borne diseases

Diseases which are transmitted by vector mosquitoes are major health problems in many countries. Although many mathematical models for diseases had been formulated, they are customized. As these diseases are spread by a common vector, similarities in the disease transmission are notable hence it will be beneficial to construct a general model which encompasses the epidemiology aspects and transmission of mosquito-borne diseases. In this paper, a SI (Susceptible-Infectious) generic model for mosquito borne diseases is formulated. The model is made up of partial differential reaction-diffusion equations which incorporate both the human and mosquito populations. Numerical simulation of this model is presented.

Ensemble model of Artificial Neural Networks with randomized number of hidden neurons

Conventional artificial intelligence techniques and their hybrid models are incapable of handling several hypotheses at a time. The limitation in the performance of certain techniques has made the ensemble learning paradigm a desirable alternative for better predictions. The petroleum industry stands to gain immensely from this learning methodology due to the persistent quest for better prediction accuracies of reservoir properties for improved hydrocarbon exploration, production, and management activities. Artificial Neural Networks (ANN) has been applied in petroleum engineering but widely reported to be lacking in global optima caused mainly by the great challenge involved in the determination of optimal number of hidden neurons. This paper presents a novel ensemble model of ANN that uses a randomized algorithm to generate the number of hidden neurons in the prediction of petroleum reservoir properties. Ten base learners of the ANN model were created with each using a randomly generated number of hidden neurons. Each learner contributed in solving the problem and a single ensemble solution was evolved. The performance of the ensemble model was evaluated using standard evaluation criteria. The results showed that the performance of the proposed ensemble model is better than the average performance of the individual base learners. This study is a successful proof of concept of randomization of the number of hidden neurons and demonstrated the great potential for the application of this learning paradigm in petroleum reservoir characterization.

A Hybrid of Functional Networks and Support Vector Machine models for the prediction of petroleum reservoir properties

This paper presents an innovative hybrid of Functional Networks and Support Vector Machines (FN-SVM) as an improvement over an existing Functional Networks and Type-2 Fuzzy Logic (FN-T2FL) hybrid model. The former is more promising as it combines two existing techniques that are very close in performance and well known for their computational stability and fast processing. This proposed FN-SVM hybrid model benefits from the excellent performance of the least-square-based model-selection algorithm of Functional Networks and the non-linear high-dimensional feature transformation capability that is based on structural risk minimization and Tikhonov regularization properties of SVM. Training and testing the SVM component of the hybrid model with the best and dimensionally-reduced variables from the input data resulted in better performance with higher correlation coefficients, lower root mean square errors and further less execution time than the standard SVM model. A comparison of FN-SVM with the existing FN-T2FL, using the same data and operating environment, showed that the FN-SVM is more accurate and consumes less time.

Computational Modeling and Simulation of EHD Ion-Drag Pumping Using Finite Difference Method

The theoretical modeling of EHD pumping is a complex process governed by the electrostatic and hydrodynamic partial differential equations. The exact solution of these equations is quite difficult therefore, the numerical methods are used to investigate and simulate the EHD pumping. In most of the cases the numerical solution is obtained by the available simulation packages based on the finite element methods that limit the analysis with built in functions. In this paper for the first time EHD ion-drag pumping at the micro scale is simulated by using finite difference method. A user defined code is written in MATLAB and the interactive simulation patterns for electric potential, electric field, velocity field and pressure field are obtained and also compared with finite element method. It was found that the finite difference simulation is significantly in agreement that of by finite element method. The former provides more control and ease to analyze and predict the performance of ion-drag micro pumps.

3D Modeling and Simulation of Electrohydrodynamic Ion-Drag Micropump with Different Configurations of Collector Electrode

This paper presents 3D simulation work for ion-drag micro pump. The commercial simulation software Comsol Multiphysics 4.2 was used to model and simulate ion-drag micro pump for three different configurations of collector electrode. The purpose of the simulation was to investigate the ways to improve the performance of the micro pump and analyze the effect of the electric field, fluid velocity and pressure gradient on the different design of the micro pump. The initial simulation patterns reveal the behavior of the electric field, fluid velocity and pressure gradient in the domain of each of the design of the micro pump. From the simulation results it is observed that at the top of the flow channel the fluid flow and pressure field is low because the driving electric body force is stronger near the electrodes surface and weaker at the top regions of the micro channel.