Data-Driven Optimization of Intermittent Gas Production in Mature Fields Assisted by Deep Learning and a Population-Based Global Optimizer

In gas wells, decreased/unstable production can occur due to difficult-to-predict dynamic effects resulted from late-life phenomena, such as liquid loading and flooding. To minimize the negative impact of these effects, maximize production and extend the wells’ lifetime, wells are often operated in an intermittent production regime. The goal of this work is to find the optimum production and shut-in cycles to maximize intermittent gas production as a decision support to operators. A framework suitable for single and multiple wells was developed by coupling a Deep Learning forward model trained on historical data with a population-based global optimizer, Particle Swarm Optimization (PSO). The forward model predicts the production rates and wellhead pressure during production and shut-in conditions, respectively. The PSO algorithm optimizes the operational criteria given operational and environmental objectives, such as maximizing production, minimizing start-up/shut-in actions, penalizing emissions under several constraints such as planned maintenances and meeting a contract production value. The accuracy of the Deep Learning models was tested on synthetic and field data. On synthetic data, mature wells were tested under different reservoir conditions such as initial water saturation, permeability and flow regimes. The relative errors in the predicted total cumulative production ranged between 0.5 and 4.6% for synthetic data and 0.9% for field data. The mean errors for pressure prediction were of 2-3 bar. The optimization framework was benchmarked for production optimization and contract value matching for a single-well (on field data) and a cluster of wells (synthetic data). Single-well production optimization of a North Sea well achieved a 3% production increase, including planned maintenances. Production optimization for six wells resulted in a 21% production increase for a horizon of 30 days, while contract value matching yielded 29/30 values within 3% of the target. The most optimum, repeatable and computationally efficient results were obtained using critical pressure/gas flowrates as operational criteria. This could enable real-time gas production optimization and operational decision-making in a wide range of well conditions and operational requirements.

Application of Capacitance-Resistive Model With Global Optimizer

The need to do a history matching in a deltaic environment with a total of 550 compartmentalized channel reservoirs of Field X brings such heavy challenges in terms of time consumed and uncertainties present. The capacitance-resistive model (CRM) rooted in signal processing between the injection and production rate was chosen to determine connectivity between injectors and producers (f_ij) and flood efficiencies for portions of the field (f_F). These constants become key insights for validating the dynamic synthesis of the reservoirs. CRM relies solely upon production and injection data. Two different control-volumes for CRM, CRMT and CRMP, were solved using a global non-deterministic solver which elevate differential evolution algorithm. The parameters’ solved was then validated with the observed liquid rate of the wells. Several techniques such as using system-wide R2 as the objective function, removal of inactive wells and distance-based weighting were used to improve the validation of the proxy model. The methods were applied to validate analyzed reservoirs with divided regions based on earlier analysis. A first CRM run was presented in this paper to test the algorithm prepared. Then another CRM run were demonstrated in this paper to show how they confirm the compartmentalization within a reservoir when compared to the reservoir’s pressure-over-time plot and earlier manual production-injection data analysis. This paper exemplified the strength of CRM itself which is to describe large-scale system in a way that circumvents geologic modeling and saturation matching with short to moderate computation time, as well as improvements applied to help the optimization process.

Application of Genetic Algorithm for Binary Optimization of Microstrip Antennas: A Review

<abstract> <p>Researchers have proposed applying optimization techniques to improve performance of microstrip antennas (MSAs) in terms of bandwidth, radiation characteristics, polarization, directivity and size. The drawbacks of the conventional MSAs can be overcome by optimizing the antenna parameters while keeping a compact configuration. Applying a global optimizer is a better technique than using a local optimizer or a trial and error method for performance enhancement. This paper discusses genetic algorithm (GA) optimization of microstrip antennas presented by the antenna research community. The GA optimization procedure, antenna parameters optimized by using GA and the optimization objectives are presented by reviewing the literature. Further, evolution of GA in the field of MSAs and its significance are explored. Application of GA optimization to design broadband, multiband, high-directivity and miniature antennas is demonstrated with the support of several case studies giving an insight for further developments in the field.</p> </abstract>

Global optimization based on active preference learning with radial basis functions

This paper proposes a method for solving optimization problems in which the decision-maker cannot evaluate the objective function, but rather can only express a preference such as “this is better than that” between two candidate decision vectors. The algorithm described in this paper aims at reaching the global optimizer by iteratively proposing the decision maker a new comparison to make, based on actively learning a surrogate of the latent (unknown and perhaps unquantifiable) objective function from past sampled decision vectors and pairwise preferences. A radial-basis function surrogate is fit via linear or quadratic programming, satisfying if possible the preferences expressed by the decision maker on existing samples. The surrogate is used to propose a new sample of the decision vector for comparison with the current best candidate based on two possible criteria: minimize a combination of the surrogate and an inverse weighting distance function to balance between exploitation of the surrogate and exploration of the decision space, or maximize a function related to the probability that the new candidate will be preferred. Compared to active preference learning based on Bayesian optimization, we show that our approach is competitive in that, within the same number of comparisons, it usually approaches the global optimum more closely and is computationally lighter. Applications of the proposed algorithm to solve a set of benchmark global optimization problems, for multi-objective optimization, and for optimal tuning of a cost-sensitive neural network classifier for object recognition from images are described in the paper. MATLAB and a Python implementations of the algorithms described in the paper are available at http://cse.lab.imtlucca.it/~bemporad/glis.

Adaptation, Comparison, and Improvement of Metaheuristic Algorithms to the Part-of-Speech Tagging Problem

Part-of-Speech Tagging (POST) is a complex task in the preprocessing of Natural Language Processing applications. Tagging has been tackled from statistical information and rule-based approaches, making use of a range of methods. Most recently, metaheuristic algorithms have gained attention while being used in a wide variety of knowledge areas, with good results. As a result, they were deployed in this research in a POST problem to assign the best sequence of tags (roles) for the words of a sentence based on information statistics. This process was carried out in two cycles, each of them comprised four phases, allowing the adaptation to the tagging problem in metaheuristic algorithms such as Particle Swarm Optimization, Jaya, Random-Restart Hill Climbing, and a memetic algorithm based on Global-Best Harmony Search as a global optimizer, and on Hill Climbing as a local optimizer. In the consolidation of each algorithm, preliminary experiments were carried out (using cross-validation) to adjust the parameters of each algorithm and, thus, evaluate them on the datasets of the complete tagged corpus: IULA (Spanish), Brown (English) and Nasa Yuwe (Nasa). The results obtained by the proposed taggers were compared, and the Friedman and Wilcoxon statistical tests were applied, confirming that the proposed memetic, GBHS Tagger, obtained better results in precision. The proposed taggers make an important contribution to POST for traditional languages (English and Spanish), non-traditional languages (Nasa Yuwe), and their application areas.

Global Evolution Commended by Localized Search for Unconstrained Single Objective Optimization

Differential Evolution (DE) is one of the prevailing search techniques in the present era to solve global optimization problems. However, it shows weakness in performing a localized search, since it is based on mutation strategies that take large steps while searching a local area. Thus, DE is not a good option for solving local optimization problems. On the other hand, there are traditional local search (LS) methods, such as Steepest Decent and Davidon–Fletcher–Powell (DFP) that are good at local searching, but poor in searching global regions. Hence, motivated by the short comings of existing search techniques, we propose a hybrid algorithm of a DE version, reflected adaptive differential evolution with two external archives (RJADE/TA) with DFP to benefit from both search techniques and to alleviate their search disadvantages. In the novel hybrid design, the initial population is explored by global optimizer, RJADE/TA, and then a few comparatively best solutions are shifted to the archive and refined there by DFP. Thus, both kinds of searches, global and local, are incorporated alternatively. Furthermore, a population minimization approach is also proposed. At each call of DFP, the population is decreased. The algorithm starts with a maximum population and ends up with a minimum. The proposed technique was tested on a test suite of 28 complex functions selected from literature to evaluate its merit. The results achieved demonstrate that DE complemented with LS can further enhance the performance of RJADE/TA.

Ab initio derived force fields for Zeolitic Imidazolate Frameworks: MOF-FF for ZIFs

<div> <div> <div> <p>In this paper we parameterized in a consistent way a new force field for a range of different zeolitic imidazolate framework systems (ZIF-8, ZIF-8(H), ZIF-8(Br) and ZIF- 8(Cl)), extending the MOF-FF parameterization methodology in two aspects. First, we implemented the possibility to use periodic reference data in order to prevent the difficulty of generating representative finite clusters. Second, a more efficient global optimizer based on the covariance matrix adaptation evolutionary strategy (CMA-ES) was employed during the parameterization process. We confirmed that CMA-ES, as a state-of-the-art black box optimizer for problems on continuous variables, is more suitable for force field optimization than the previous genetic algorithm. The obtained force field was then fully validated with respect to static and dynamic properties. Much effort was spent to ensure that the FF is able to describe the crucial linker swing effect in a large number of ZIF-8 derivatives. For this reason we compared our force field to ab initio molecular dynamic simulations and found an accuracy comparable to those obtained by different exchange–correlation functionals. </p></div></div></div><div><div><div> </div> </div> </div>