نشریه مهندسی برق
سال پنجاه و دوم شماره 2 (پیاپی 100، تابستان 1401)

In this paper, the robust finite-time tracking for a class of nonlinear systems comprising interconnected double integrator subsystems is discussed. This particular class of nonlinear systems is able to describe and model a group of practical plants such as industrial robot manipulators, autonomous underwater vehicles (AUVs), autonomous marine vessels, unmanned aerial vehicles (UAVs), and inverted pendulums. By developing the nonsingular terminal sliding mode control (NTSMC) method and defining innovative nonlinear sliding manifolds, control inputs are designed in order to convert the aforementioned system to the canonical nonlinear form and, in consequence, two significant goals including the finite-time tracking objective and the global finite-time stabilization of the closed-loop system (subjected to unbounded disturbances and uncertainties) are provided and guaranteed. Furthermore, a remarkable relation is derived to estimate the convergence finite time regarding the mentioned tracking problem. This relation reveals that the convergence finite time extremely depends on the values of arbitrary constants of the designed control inputs. Finally, the proposed robust control scheme is numerically simulated onto two-link robot manipulator and simulation results illustrate that the designed control inputs properly fulfill the finite-time tracking objective.

Exhaustive testing of software systems with a large number of input parameters and combinations between them often causes the problem of combinatorial explosion. Combinatorial t-way testing is a technique that generates an array of test cases to maximize combinations covering of between input parameters. Generating a minimum covering array is an optimization problem that many strategies based on metaheuristic algorithms such as teaching and learning based optimization, particle swarm optimization, and genetic and cuckoo search algorithms have been used for solving it. Although these strategies have produced smaller covering arrays, complete minimization has not yet been performed. In this paper, we propose a new strategy based on the success-history and linear population size reduction based adaptive differential evolution algorithm (so-called LSHADE), which is one of winners of IEEE CEC competitions, to generate minimum covering array. The results of Friedman mean rank show that the LSHADE strategy has the first rank in terms of generating the covering array with the lowest size and the lowest average number of algorithmic calls, compared to mathematics based strategies such as TConfig, greedy strategies such as IPOG, Jenny and PICT and meta-heuristics such as GS, TLBO, HC-BAT, PSTG, WOA , BAPSO and GSTG. While, in terms of the average number of fitness function evaluations and the average runtime, this strategy has the first rank after the GS strategy. Moreover, the convergence diagrams confirm the high convergence speed of this strategy compared to the other meta-heuristic strategies.

In this paper, an optimal scheduling method for an Active Distribution Network (ADN) in the presence of Soft Open Points (SOPs) is presented. In the proposed method, all types of energy transactions, as well as the reactive power exchange between the ADN and the Upstream Network (UN) and the microgrids, are modeled. Scheduling is formulated in a 24-hour time horizon in which all costs and revenues of the ADN in energy transactions, such as the cost of purchasing power from the UN and the revenue from selling power to it along with the power loss cost, and the DG's fuel cost are modeled. In this method, the exact equations of AC load flow, operational constraints of SOPs and DGs, and energy transaction constraints such as power factor limitation, are modeled in a mixed-integer second-order conic programming model. Therefore, it is expected that the obtained results are the optimal global response that meets the scheduling constraints. To evaluate this claim, the operational scheduling of the modified 33-bus IEEE distribution system connected to two microgrids in different scenarios in the presence of high penetration of renewable sources has been investigated. The simulation results show the efficiency of the proposed framework.

The lack of clarity regarding the participation of transmission system operators and their collaboration with distribution system operators in local flexibility markets is one of the challenges of these markets. So in this study for managing transmission system unpredicted congestions due to sudden increase in system load with demand side resources, a bilevel flexibility market is presented. The first level of this market is related to providing TSOs with required services and only DSOs can participate in this level. The second level consists of local markets in each distribution system that are managed by DSOs and prosumers submit their flexibility bids in these markets. In this paper, DSO and TSO coordination is based on DSO-managed models. This means that DSO aggregates all flexibility bids in distribution networks and modifies prosumers' cost functions considering distribution network loss minimization. The DSO then submits these bids to the first level flexibility market. Modifying prosumers' cost functions is an optimization problem that is solved using water cycle optimization algorithms.

Automatic keyphrase generation plays an important role in many text analysis and natural language processing tasks. Many existing methods are bound to select keyphrases from the terms and phrases that are present in the target text. This handicap could be overcome using sequence-to-sequence methods. However, many such methods need huge datasets for training which pose a challenge for low-resource languages such as Persian. Transfer learning where a pre-trained model is adapted to a new task specified with a smaller dataset is very useful in such circumstances. In this paper, we present a sequence-to-sequence method utilizing a transformer model for Persian keyphrase generation. Accordingly, a corpus of 70K Persian scientific abstracts and their corresponding keyphrases have been gathered. A pretrianed MT5 mdel is fine-tuned on this corpus for the task of Persian keyword generation. The resulted model is compared to several other keyphrase generation methods. The results indicate that the proposed method can outperform existing methods at least by 2.71 percent.

Feature selection is one of the most important techniques in machine learning and pattern recognition, which eliminates redudant features and selects a suitable subset of features. This avoids overfitting when building the model and improves the model performance. In many applications, obtaining labeled data is costly and time consuming, while unlabeled data are readily available. Therefore, semi-supervised feature selection methods can be used to consider both labeled and unlabeled data in the feature selection process. In this paper, a semi-supervised sparse feature selection method is proposed based on hessian regularization and Fisher discriminant analysis which selects the appropriate features using the labeled data and the local structure of both labeled and unlabeled data. In the proposed method, an objective function based on semi-supervised scatter matrix and l2,1-norm is presented for feature selection which considers the correlation among features. To solve the proposed objective function, an iterative algorithm is used and its convergence is experimentally and theoretically proved. The results of the experiments on five data sets indicate that the proposed method improves the selection of relevant features compared to other methods used in this paper.

Identification of the exact location of an exon in a DNA sequence is an important research area of bioinformatics. The main issues of the previous signal processing techniques are accuracy and robustness for the exact locating of exons. To address the mentioned issues, in this study, a method has been proposed based on deep learning. The proposed method includes a new preprocessing, a new mapping method, and a multi-scale modified and hybrid deep neural network. The proposed preprocessing method enriches the network to accept and encode genes at any length in a new mapping method. The proposed multi-scale deep neural network uses a combination of an embedding layer, a modified CNN, and an LSTM network. In this study, HMR195, BG570, and F56F11.4 datasets have been used to compare this work with previous studies. The accuracies of the proposed method have been 0.982, 0.966, and 0.965 on HMR195, BG570, and F56F11.4 databases, respectively. The results reveal the superiority and effectiveness of the proposed hybrid multi-scale CNN-LSTM network.