One of the major problems faced by the world lies on how to fight back climate change without disrupting the economy. In a scenario where fossil fuels are still dominant over renewable sources of energy, if development growth is to be maintained, doing a solid and economically viable transaction between sources of energy is primordial. Carbon captured coupled with CO2-EOR & Storage rises as a bridge, in which it would still be feasible to utilize existing fossil fuel infrastructure while addressing the climate change mitigation goals. The focus of this study is evaluate, at a multiscale level, the tradeoffs related to CCS coupled with CO2-EOR & Storage within the United Arab Emirates. The interplay between the reservoir level parameters (CO2 breakthrough time, injection rate, etc.) and the country level aspects (CO2 capture potential, desired oil production, etc.) merits further investigation which has not been previously studied.
Limestone reservoirs contain almost half of the world hydrocarbon proven reserves with 90% of them detected to be mixed-wet to oil-wet. By linking some reservoir features represented in reservoir management, fluid types, rock types, reservoir vertical and areal heterogeneities and drive mechanisms, it was assumed that more than 50% of oil trapped and bypassed, even after secondary recovery, in carbonate rocks are yet to be recovered. Therefore, different techniques of enhanced oil recovery (EOR) have been introduced to carbonate reservoirs to reduce residual oil saturation (Sor) as much as possible and increase recovery. The suitable EOR technique to be used is selected mainly based on fluid/rock types and initial reservoir conditions. Chemically modified water flooding is one of the most recent and propitious EOR techniques. Recently, many studies suggested that modifying the injected brine chemistry could decrease the residual oil saturation. Various mechanisms behind chemically modified water flooding were suggested but not yet proved. Literature has pointed out that wettability alteration could be the predominant mechanism behind chemically modified water flooding. Other suggested mechanisms include emulsification and entrainment, emulsification and entrapment, IFT reduction between oil/brine interfaces, rock dissolution and electrical double layer expansion. These mechanisms were shown effective by studying them experimentally. However, the success of the aforementioned mechanisms has not yet been proved in field scale. Understanding these mechanisms is a must. The objective of this research is to thoroughly investigate the negative/positive impact of the synergy between two different types of chemically modified water flooding combined with EDTA on oil recovery in limestones, considering what have been achieved in the literature. The chemicals being used are Sodium Hydroxide and Trisodium Phosphate. Several parameters are varied under controlled laboratory environment to reveal their possible effects. These parameters include oil composition, composition of injected brines, pH and temperature. A detailed oil analysis, including SARA analysis, gas chromatography, acid and base numbers of a crude oil, is performed. XRD and CT-scanning are also conducted to entirely characterize the core plugs used; in order to get a full picture of the rock/brine interactions. Core flooding tests, Amott spontaneous imbibition, zeta potential of crude/brine and rock/brine coupled with effluent analysis are performed to reveal more deep conclusions regarding the optimum conditions at which chemically modified water flooding could yield as best EOR potential as possible.
Knowledge of fluid saturation at any one cross-section is very often required when studying fluid behavior in porous media. It becomes more challenging in case of microscale geometry with heterogeneous morphology, as in complex geological samples. Therefore, to solve deficiency, we used the latest advancement in uCT and MRI imaging in enabling higher resolution of pore network saturation measurement. Our research effort in addressing the challenge has led to a novel solution.
For oil recovery applications, the interfacial interactions for two-phase flow, specifically water/oil interaction with rock, have been extensively investigated with the advances in the mesco- /microscale experimental and numerical research approaches. However, as the key to understanding the water-alternating-gas (WAG) injection, foam injection and other immiscible fluids injections for enhancing oil recovery, the more complicated the interactions among ternary fluids and solid surface is rarely reported. In this work, we target on the ternary fluid system, water, gas and oil, to show the role of interfacial forces in the phase flow behavior and residual oil mobilization under immiscible WAG injection. The lattice Boltzmann model is developed for threephase flow simulation based on the classical Shan-Chen type multicomponent model. The effect of the gas wettability on the oil mobilization and extraction efficiency is also discussed. This work also offers a methodology in the studying other ternary fluids relevant application, such as oilinfused surface in energy-water nexus and droplet generation in microfluidics.
Due to the growing number of Internet users and smart handheld devices, the demand for databased services increases rapidly over the past several years. Caching-centric networks and caching techniques, such edge caching, have gained numerous attention as they provide efficient and effective methods to maintain a high quality of service. In fact, only a few contents are popular and win the majority of viewers, thus caching them reduces the latency and download time. In this view, integrating popularity prediction into caching, effectively increases network utilization as well as user satisfaction rate. In this paper, we propose the method based on long short-term memory (LSTM) for anticipating video contents demand. Unlike neural networks, LSTM is capable of finding correlations among training instance and memorizing long and short dependencies. The model hyperparameters are selected using Tree of Parzen Estimators to improve the overall performance.
This paper presents an efficient blind channel estimation technique for time-varying orthogonal frequency division multiplexing (OFDM) systems. New frame structure is proposed, where different modulation schemes are employed to estimate the time-varying channel coefficients. Amplitude shift keying (ASK) and phase shift keying (PSK) modulation schemes are utilized to modulate particular pair of subcarriers over consecutive OFDM symbols. Exact closed-form expression for the symbol error rate (SER) of the ASK symbols is derived and corroborated with Monte Carlo simulations to evaluate the performance of the proposed technique and compare it with the pilot based OFDM system. Analytical and simulation results show that the proposed estimator can provide estimation with accuracy and computational complexity that are comparable to pilot based estimators.
Quintillion bytes of data is created every day from various organizations in real time. The ability to analyze real-time data would benefit these organization to make better decisions and improve their efficiency. However, this objective is less efficient with the traditional machine-learning practices. In this paper, we propose an efficient algorithm for selection of features from a feature stream by online feature grouping. This technique will be useful in big data analytics due to its efficiency and scalability. The main contribution of this research is to solve the issue of the extremely high dimensionality of big data by delivering the streaming feature grouping and selection algorithm. We have implemented this algorithm and evaluated using benchmark dataset against state-of-the-art streaming feature selection algorithms and feature grouping technique. The results showed superior performance in terms of prediction accuracy.
Cloud network monitoring data is dynamic and distributed. Signals to monitor the cloud can appear, disappear or change their importance and clarity over time. Machine learning (ML) models tuned to a given data set can therefore quickly become inadequate. A model might be highly accurate at one point in time but may lose its accuracy at a later time due to changes in input data and their features. Distributed learning with dynamic model selection is therefore often required. Under such selection, poorly performing models are retired or put on standby while new or standby models are brought in. In this paper, we propose such methodology for automatic ML model selection and tuning that automates the model build and selection and is competitive with existing methods. In particular, we create a Cloud DevOps architecture for autotuning and selection based on container orchestration and messaging between containers, and take advantage of a new autoscaling method to dynamically create and evaluate instantiations of ML algorithms. The proposed methodology and tool are demonstrated on cloud network security datasets.
As the industry prepares for the 1000x mobile data challenge and the resulting spectrum crunch that LTE networks are anticipating in the licensed spectrum, the extension of LTE to the unlicensed spectrum (LTE-U) has been proposed as a promising solution. However, this extension is challenged by the problem of coexistence with incumbent unlicensed systems, especially WiFi. Several LTE channel access mechanisms for enabling fair and friendly coexistence with WiFi have been proposed and evaluated. This paper further explores the design of LTE-U - WiFi coexistence mechanisms from the perspective of video streaming. Specifically, two channel access mechanisms are proposed that take into account the difference in the availability of the LTE licensed and unlicensed channels along with the structure of the encoded video file. The performance of the proposed access mechanisms are evaluated via simulations and shown to outperform the classical Listen Before Talk (LBT) channel access mechanism.
Biological movements provide inspiration to solve several real-world problems, resulting in the rise of several bio-inspired optimization techniques. However, to develop an efficient bio-inspired technique, it is necessary to first understand the underlying physical statistics and models. While most biological movements could be generalized as "random" at first glance, deeper empirical studies suggest that the movements in organisms follow a much more complex randomness than a simple Brownian particle movement. This paper first introduces the variations of randomness observed in biological walks and statistical models used to imitate them, and then proposes on a possible tangent of an ant colony inspired biased random walk that could help implement efficient searches in unknown environments.