نشریه هوش محاسباتی در مهندسی برق
سال پانزدهم شماره 2 (تابستان 1403)

In C-RAN architecture, all computational processing is performed in the central baseband unit (BBU) pool, while radio operations are carried out in the remote radio heads (RRHs). The central BBU pool is connected to the RRHs by fronthaul links. Therefore, by separating the processing unit and the radio units, the clustering structure of RRHs can be designed to adapt to network changes. This paper deals with the problem of radio and computation resource allocation to maximize weighted sum rate. To reduce complexity, we use user-centric clustering and appropriate training resources allocation. Additionally, to lessen channel training overhead, an incomplete model of channel state information is considered, in which only intra-cluster channel state information is estimated. By replacing a sticky lower bound of user data rate in the main problem, the beamforming vectors under the constraints of computational and radio resources are designed in three steps. In the first step, the weighted sum rate maximization problem is solved under maximum radio transmitted power constraints by weighted minimum mean square error method. Then, in the second step, using a proposed greedy algorithm to allocate computational resources to users. In the third step, fronthaul capacity constraints are applied by another greedy algorithm.

By integrating the power and natural gas systems (IPGS), these two systems can synergistically provide electric and gas power through mutual energy sharing. However, disruptions occurring in one system can have detrimental effects on the optimal performance of the other system. Hence, the assessment of IPGS reliability becomes imperative. In this article, a reliability evaluation model is proposed that utilizes machine learning algorithms to tackle uncertainties associated with the failure rate of IPGS components in the presence of gas storage resources. To calculate system reliability indices, we employ Sequential Monte Carlo Simulation (SMCS) and an optimal load-shedding program. Moreover, a random forest (RF) algorithm is adopted to classify elements based on their importance in upholding the overall system's reliability. The proposed model is implemented using MATLAB, GAMS, and Python software, with the IEEE 14-bus system and the 10-node natural gas system as case studies. The simulation results illustrate that by considering the gas storage resources, the reliability of the power system and natural gas can be improved by 2.12 and 8.3%, respectively, and the overall reliability of IPGS is improved by 5.25%. Also, the prioritization of these resources is determined in IPGS.

Identifying the fault location in the transmission lines and fixing the problem in the shortest possible time is of particular importance to reduce blackouts and increase the reliability of the network. One of these fault location methods is the impedance-based method. The precise values of transmission line parameters and the Thevenin impedance of the network are unknown due to various issues such as temperature change and network structure, which can affect the accuracy of this method in locating the fault in both uncompensated and compensated transmission lines. Therefore, in this article, the proposed method is an intelligent impedance-based method using the variable neighborhood search algorithm, regarding the uncertainty in the line parameters and the Thevenin impedance of the network in locating single-phase-to-ground faults. In the proposed method, only one-end-current data is used to estimate the location of the fault, which does not require a telecommunication platform and results in increasing the reliability of the location method. The simulation results show that the method has high accuracy in fault location in transmission lines even though the line parameters and Thevenin impedance of the network are unknown. Moreover, the resistance, the angle and the location of the fault, and the level of compensation do not have a significant effect on the accuracy of the fault location.

Energy storage systems have an important effect on increasing the flexibility of power grids and are necessary to achieve the goals of smart grids. So far, many researches have been conducted in the field of optimal utilization of batteries, but most of them have studied the effects of batteries only on the side of the transmission or distribution network. In the present study, the effect of the installation and optimal operation of energy storage systems on both sides of the transmission and distribution network has been investigated. This paper presents a bi-level optimization model for integrated energy management in transmission and distribution smart grids along with demand management issues. Considering that both proposed models are mixed integer linear programming models, it is difficult to solve them by direct methods. Therefore, a method based on the change of variables and KKT equations is presented to solve these problems, and globally optimal solutions to these problems are guaranteed. Several integrated systems that include transmission and distribution networks are considered to evaluate the proposed model and method, and the results show that the proposed model is efficient and shows the positive effect of modeling energy storage systems on both sides of the network.

It is essential to implement a high-performance drive control system in a six-phase induction motor (6PIM) to benefit from the advantages of reduced torque ripple, reduced harmonic current, reduced current amplitude per phase, and improved reliability. In this regard, two control schemes based on the model predictive torque control (MPTC) method are proposed in this paper. In addition, a fuzzy logic controller (FLC) is used in the speed control loop. Furthermore, in the first control scheme, during each switching period, a combination of an active switching vector and a null switching vector is applied to the inverter; so that the duty cycle of the active switching vector is optimally calculated. In the second control scheme, a virtual voltage (VV) vector composed of a large switching vector and a relatively large switching vector with certain duty cycles is applied to the 6PIM so that related switching vectors’ effects in the x-y subspace which models the motor losses lead to zero. The proposed methods are evaluated by simulation using the MATLAB/Simulink software. The results prove the effectiveness of the presented methods. The advantages of the first method consist of fully exploiting the DC link voltage and reducing the torque ripple. In addition, the advantages of the second method include reducing the current ripple and improving efficiency.

Euler-Lagrange systems include a wide class of dynamic systems such as electrical, mechanical and robotic systems. In this paper, a disturbance observer-based control methodology is developed for a multi-agent network including several agents with Euler-Lagrange motion dynamics. The control objective is to reach the same position by all agents in both leader-less or leader-follower situations. The proposed control scheme estimates the disturbance term due to the dynamic uncertainties, and the estimated value is utilized in the control law as a feedforward term. In contrast to the existing adaptive control methodologies for handling uncertainties, the proposed scheme guarantees satisfactory performance in the presence of uncertainty without the need for a dynamic structure of the system. Stability analysis of the proposed scheme is done using the Lyapunov theory. The effectiveness of the proposed method is validated by simulation studies for regulation and tracking objectives in leader-follower cases and consensus objectives in leaderless scenarios.

One of the leading causes of death in today’s world is arterial events, which can occur in the form of myocardial infarction, stroke, fainting, etc., leading to heart failure and premature death. Hence, it is essential to inform people before the occurrence of unpleasant events to prevent and warn of abnormal conditions. In this paper, we used a deep learning approach to predict arterial events during the weeks/ months before the event using 5 minutes of ECG signals recording by extracting the temporal and frequency characteristics. For this purpose, the long short-term memory (LSTM) neural network was used due to the possibility of learning long-term dependencies to quickly identify and prevent these types of events. A deep belief network (DBN) was also used to represent and select more effective and efficient features of the recorded dataset. This approach is briefly called LSTM-DBN. Four publicly-available datasets in the field of health care were used to evaluate the proposed approach. These data were collected from ECG signal recording wearable sensors along with demographic characteristics. The prediction results of the LSTM-DBN network with 88.42% mean accuracy compared to other deep learning approaches and traditional classifications indicate the superiority of the proposed approach.

Reconfiguration along with the optimal allocation of DG resources and DSTATCOM units in power distribution networks is one of the effective tasks to improve power quality indicators in distribution networks, that as a non-linear combination scenario with high complexity, has been less discussed in research. This problem is a challenge with a discrete-continuous nature in the field of electrical power engineering, the solution of which requires powerful techniques in the field of optimization. To solve the proposed problem, a new hybrid algorithm named (GA-HOA) is introduced in this article. Fuzzy logic has also been used to find the most optimal solution among the obtained Pareto solutions. Considering a multi-purpose objective function including reducing losses, improving voltage stability, reliability, as well as operating costs, solving the proposed scenario along with four other scenarios using a new combined algorithm and a number of the proposed optimization algorithms has been evaluated on a distribution network with 70 buses. Also, different scenarios have been implemented on two distribution networks with 33 and 84 buses to reduce active power losses and compared with the outputs obtained from other methods in various articles. The obtained results show the success of the new hybrid algorithm in solving the proposed complex scenario along with providing a new topology for switches, locations, and optimal capacities for DG and DSTATCOM units with lower losses, more stable voltage, more reliability, and suitable operating cost. Also, the obtained results indicate the relative superiority of the proposed hybrid algorithm in finding the optimal solution and proper convergence time compared to other compared methods.