Search

Due to the complex nature of processing units and the variability in raw oil properties, traditional first-principle methods are not always effective in predicting the output of these units without adapting them. Data-driven models have become popular alternatives because they are more adaptable to the variability in processing units, especially when large amounts of data are available. Feature selection is a crucial aspect of data-driven modeling, especially for processing units. It involves identifying and selecting the most important variables influencing the unit's output. The selected variables are then used as inputs to the model, which helps to reduce noise and improve model accuracy. Accordingly, this article presents a hybrid method that combines feature selection with data-driven modeling to create an accurate model for predicting the output of a Residue Fluid Catalytic Cracking (RFCC) process unit. The Metaheuristic Ant Colony Optimizer (MACO) algorithm was used to extract the best subset of features from the collected data, which were then used with the Artificial Immune Genetic Algorithm with Local Vaccination (AIGA-LV) method to select the optimal features. Finally, the Random Forest (RF) method was used to predict the products of the RFCC process unit. The data were collected from a refinery plant in central Iran for nine months. The proposed model was efficient in feature selection and flexible in handling different scenarios and products. It improved the accuracy to 99.17% at its best and improved the primary model by 3–20%. Therefore, this method could be useful for predicting the output of various industrial processes, especially the RFCC and FCC process units. [ABSTRACT FROM AUTHOR]

In this study, we proposed a new metaheuristic algorithm called Electrical Search Algorithm (ESA). The proposed algorithm is based on the movement of electricity in high-resistive areas such as wood, glass, and gases. ESA has a unique initialization scheme that only one agent initializes at the lower and upper bounds of the search space, which creates structures called poles. After that, ESA uses unique exploration and exploitation strategies to search. The search mechanism is based on electrons moving to opposite poles. ESA differs from other metaheuristics compared to its initialization scheme, pole search mechanism, and update strategy of the best solutions. ESA was tested with the "100-Digit Challenge" benchmark functions in the IEEE-CEC-2019, four well-known benchmark functions, and an np-hard clustering problem. For the clustering problem, we used four well-known datasets: Iris, Wine, Seeds, and Hepatitis C Virus. ESA was compared with seven different metaheuristic algorithms on these well-known benchmark functions, and the results of the clustering problem were compared with the K-Means algorithm. Additionally, Friedman Signed Rank and post hoc Wilcoxon Test were run to show the significance of the results. In all of the well-known benchmark functions, ESA either offered the best results or similar results to other compared algorithms. The score of the ESA on the IEEE-CEC-2019 benchmark functions shows us that even with the minor evaluation numbers, ESA can achieve similar results to the competing algorithms. Results show that ESA has a robust mechanism for not trapping in local points and moves slow but persistent rate. [ABSTRACT FROM AUTHOR]

This paper addresses the problem of optimal location and sizing of distributed generators (DGs) in direct current (DC) grids. To solve it, we propose an optimization approach with an objective function that aims to reduce power losses due to energy transport, while considering all the constraints that represent DC grids in a distributed generation environment. For the mathematical formulation of the problem, we used a mixed-integer nonlinear programming (MINLP) model, which allowed us to evaluate the impact of all possible configurations (i.e., location and size of DGs in the DC network) on the objective function and the constraints. The solution method proposed here is a master–slave strategy that implements a hybrid solution methodology that combines a genetic algorithm (GA) and the vortex search algorithm (VSA). The GA is in charge of solving the location problem in the master stage, and the VSA is responsible for sizing the DGs in the slave stage. To evaluate the effectiveness and robustness of the proposed GA/VSA methodology, we employed two test systems (i.e., 21 and 69 buses) considering a maximum penetration of distributed generation equal to 40% of the power generated by the slack buses. Furthermore, we also implemented nine other hybrid methodologies based on metaheuristic techniques (proposed in the literature for solving the problem addressed here) to make comparisons. All the solution methods used and proposed in this paper are based on sequential programming to avoid the need for specialized software and thus reduce the complexity and cost of the solutions. The effectiveness of the proposed solution was evaluated in two scenarios: (1) peak power demand and (2) variation in power generation and demand associated with photovoltaic generation and user demand in Medellín, Colombia. The results demonstrate that the GA/VSA methodology achieved the best results in terms of solution quality and processing times in all the test scenarios proposed in this study. [ABSTRACT FROM AUTHOR]

Given the diversity of styles and sizes in apparel, marker planning which aims to arrange and move all the pattern parts of garments onto a long thin paper before the cutting process is a very important process for the apparel industry. In order to decrease the wastage of fabric after the cutting process, the marker layout essentially needs to be as compact as possible. In this paper, hybrid heuristics are proposed to conduct and achieve an optimized marker layout and length. First, a moving heuristic is presented as a new packing method to arrange and move the patterns without overlapping; here, an initial marker is presented to calculate the length. This heuristic considers multiple rotated angles and flipping positions of the patterns in order to obtain more diverse arrangements. With different arrangements, there is a higher chance of achieving an optimized marker layout and length. To further improve the solution received from the moving heuristic, soft computing algorithms are taken into account, including the genetic algorithm (GA), simulated annealing (SA), and hybrid genetic algorithm-simulated annealing (HGASA) to achieve a shorter length in the marker layout. The best marker length can be obtained from HGASA, which can save almost 28% of the length. Special cases with specific combinations in rotated angles are considered so that the industry can make informed choices on the algorithms suitable to address the marker planning problem. [ABSTRACT FROM AUTHOR]

Feature transformation methods can be used effectively to improve the performance of the machine learning techniques, such as classification and clustering, via enhancing the discrimination among the samples. Additionally, binary representation of features may be necessary in several circumstances. For example, distributed systems may have limited bandwidth, storage, and energy that necessitate rough quantization of the measurements. This paper proposes a random projection-based feature transformation method that maps the data points from the original space into a new binary space. In the proposed transformation method, the random projection process is formulated as an optimization problem. A modified chaotic version of the sine–cosine algorithm is used to find the optimal random projector. The performance of the proposed transformation method is validated using 15 standard UCI benchmark datasets against the linear discernment analysis. Moreover, it is compared to other metaheuristic-based feature reduction methods including the gray wolf optimizer, the genetic algorithm, and the antlion optimizer. The obtained results assure the superiority of the proposed transformation method in terms of classification accuracy and storage requirements. [ABSTRACT FROM AUTHOR]

Complex road networks, inaccurate GPS receiver output, low sampling rate and many other associated issues pose real challenges for map matching process. Genetic algorithms have recently been trialed for rendering GPS fix on digital maps. This manuscript introduces an improvised genetic algorithm named as post-processing genetic-inspired map matching (GiMM) algorithm. The proposed GiMM intends to mitigate inherent challenges associated with originally proposed genetic algorithm for map matching. The fitness function used by GiMM makes use of Bucket Dijkstra's and fast dynamic time wrapping (FDTW) algorithms to render GPS information on digital maps. Bucket Dijkstra's suggests the shortest path available in between two points, and FDTW is responsible for comparing two data series. Unlike traditional genetic algorithm for map matching, GiMM was evaluated on sparse as well as dense GPS data. The performance of the GiMM algorithm was evaluated in real time using OpenStreetMap data and GPS dataset mapped onto a road network of 82 km. GiMM uses population size, generation count, accuracy and execution time as input parameters. A maximum accuracy of 99.4% with root-mean-square error 0.06 was observed, whereas a minimum of 60% accuracy was observed at 0.47 root-mean-square error. Number of iterations and population size were concluded to be the most influential parameters for the performance of genetic algorithms for map matching. [ABSTRACT FROM AUTHOR]

In this study, a novel usage of four metaheuristic approaches Genetic algorithm (GA), Simulated annealing (SA), migrating bird optimization algorithm (MBO) and clonal selection algorithm (CSA) are applied to multidimensional two-way number partitioning problem (MDTWNPP). MDTWNPP is a classical combinatorial NP-hard optimization problem where a set of vectors have more than one coordinate is partitioned into two subsets. The main objective function of MDTWNPP is to minimize the maximum absolute difference between the sums per coordinate of elements. In order to solve this problem, GA is applied with greedy crossover and mutation operators. SA is improved with dual local search mechanism. MBO is specialized as multiple flock migrating birds optimization algorithms. CSA is applied with problem specific hyper mutation process. Furthermore, all instances are solved using an integer linear programming model which was previously presented in the literature. In the experiments, four metaheuristic approaches and integer linear programming model are used to solve 126 datasets with different sizes and coordinates. As a brief result, the GA and SA approaches designed for this problem outperformed all other heuristics and the integer programming model. Both the performance of GA and SA approaches are in a competitive manner where GA and SA yielded the best solution for 56 and 65 out of 125 datasets, respectively. [ABSTRACT FROM AUTHOR]

This paper discusses a bi-objective reliability optimization of a switch-mode active and cold standby k-out-of-n series-parallel system in which a switch is installed to add a redundant component when one of the components fails. The system failure rate is not only dependent on the number of its working components, but reduces when more monitory resources are allocated. As the bi-objective optimization problem is shown to belong to the class of NP-hard problems, a multi-objective meta-heuristic algorithm, namely multi-objective evolutionary algorithm based on decomposition (MOEA/D) with a novel solution structure is developed to solve large-scale problems. Since there is no benchmark available in the literature, the well-known multi-objective evolutionary algorithm called the non-dominated sorting genetic algorithm (NSGA-II) is utilized to validate the solutions obtained. The parameters of both algorithms are tuned by the Taguchi method where the AHP-TOPSIS method is applied to compare the performances of the parameter-tuned algorithms in terms of several multi-objective performance measures. The results are in support of MOEA/D. [ABSTRACT FROM AUTHOR]

This paper presents an approach to solve the unit commitment problem using a newly developed Multi-agent Evolutionary Programming incorporating Priority List optimisation technique (MAEP-PL). The objective of this study is to search for generation scheduling such that the total operating cost can be minimised when subjected to a variety of constraints, while at the same time reducing its computational time. The proposed technique assimilates the concepts of Priority Listing (PL), Multi-agent System (MAS) and Evolutionary Programming (EP) as its basis. In the proposed technique, deterministic PL technique is applied to produce a population of initial solutions. The search process is refined using heuristic EP-based algorithm with multi-agent approach to produce the final solution. The developed technique is tested on ten generating units test system for a 24-h scheduling period, and the results are compared with the standard Evolutionary Programming (EP), Evolutionary Programming with Priority Listing (EP-PL) and Multi-agent Evolutionary Programming (MAEP) optimisation techniques. From the obtained results and the comparative studies, it was found that the proposed MAEP-PL optimisation technique is able to solve the unit commitment problem where the total daily generation cost is effectively minimised and the computation time is reduced as compared to other techniques. [ABSTRACT FROM AUTHOR]

This paper deals with the problem of hybrid flow shop scheduling. In this investigation, we considered group scheduling within the area of sequence-dependent family setup times and two objectives of minimizing makespan and total tardiness are taken into consideration simultaneously. Due to the computational complexity in solving these set of problems with multiple objectives, metaheuristics has a high priority, because these algorithms are capable of solving combinatorial problems in a reasonable time. This study focuses on three multi-objective algorithms, multi-objective genetic algorithm, sub-population genetic algorithm-II and non-dominated sorting genetic algorithm-II, to solve the mentioned problem. In order to investigate the effectiveness and efficiency of applying the noted metaheuristics for such an NP-hard problem, we evaluate non-dominated solution sets obtained via each algorithm through some evaluation metrics. [ABSTRACT FROM AUTHOR]