COMPUTATIONAL INTELLIGENCE IN ELECTRICAL ENGINEERING (INTELLIGENT SYSTEMS IN ELECTRICAL ENGINEERING)
Year:2022 | Volume:13 | Issue:2|Papers Count:8

Permanent Magnets (PM), which are one of the most vulnerable and important components of permanent magnet machines, are at risk of demagnetization for reasons such as overheating and corrosion. One of the most critical tasks of engineers in the industry is to prevent the machine's operation in the demagnetization condition and replace the machine's defective PMs as much as possible. Demagnetization causes linkage flux reduction, electromagnetic force changes, and machine heating. This paper presents a new approach for demagnetized PM diagnosis in a permanent magnet linear synchronous machine using wavelet packet transform. Due to the high sensitivity to the flux changes, the electromotive force signal was selected as a detection signal. By examining this method in different demagnetization cases, it is determined that this method has a proper performance. In this paper, the finite element software (Maxwell) and MATLAB were used to simulate the machine and analyze the data, respectively.

In this paper, the performance of the neural network in diagnosing an induction motor situations (safe, bearing outer race, broken rotor bar and stator short circuit faults) has been evaluated using fusion ofthe pre-processed-current-and-voltagesignals information. Moreover, the robustness of the proposed approach is evaluated against the unbalanced power source and dry running of the electro-pump. Results indicate that the proposed neural data fusion has better reliability in this fault detection application, also provides more robust behavior.

One way to improve the distribution network resilience is to use mobile generators (MEGs) in the network. Such generators with the ability to move in the network and supply power to different parts increase the possibility of supplying important loads. However, due to the limited fuel of these generators, their movement will be limited, which will reduce their efficiency. Therefore, in this paper, considering this type of generators and also their limited fuel, an algorithm for their optimal displacement during the disaster is presented. Also, by considering fueling vehicles and planning to move them to inject fuel into these generators, an attempt has been made to reduce the effect of fuel restrictions on these generators. Finally, by simulating the proposed algorithm on the IEEE 33-bus network, taking into account the daily variable loads, the effectiveness of the proposed algorithm on improving the resilience of the distribution network is shown.

The Automatic Voltage Regulator (AVR) system is one of the important systems in power grids. Proportional – Integral (PI) controller is usually used in industries to control the AVR system. However, to increase the flexibility or improving the performance of the PI controller, fractional order PID (FOPID) or combined PID and fuzzy controllers can be used. A new method for the control of AVR has been proposed in this paper. For this purpose, a Fuzzy-PID controller with optimized membership functions has been employed. Input membership functions along with gains of Fuzzy-PID controller have been optimized with imperialist competitive algorithm (ICA) to control the AVR system. Optimizations have been done with various fitness functions. To compare, simulations have been carried out by optimized fractional order PID (FOPID) controller, too. Results of simulations have illustrated that the proposed controller has better performance compared to the FOPID controller according to settling time, overshoot, fitness function and the maximum of control signal values in all cases in such a way that the minimum improvement of overshoot, settling time, the maximum value of control signal, and fitness function value have improved 59%, 33%, 65%, and 3%, respectively. Moreover, to verify the robustness of the proposed controller to system uncertainties, simulations have been carried out by changing the parameters of the system. Results of simulations have shown the perfect performance of the proposed controller.

This paper deals with the optimal set of directional overcurrent relays using intelligent methods when the problem has multi-objective functions with different constraints. Due to the linear and nonlinear constraints in the coordination of the directional overcurrent relays, the method used is based on Pareto Envelope-based Selection Algorithm to obtain optimal values of plug setting, time multiplier setting, and relay characteristic curve by considering all governing constraints setting parameters of the network primary and backup relays. In this study, there is a simultaneous time reduction between the performance of the main and backup relays. Besides, in this method, while solving a multi-objective function without the need for weight factors, a penalty function is used to check the correct coordination of the primary and backup relays. The proposed method is implemented in two different standard networks, i. e. 3-bus and modified IEEE 14-bus networks. The results show the efficiency of the method used and the optimality of the values compared to other algorithms.

In high voltage cables, due to the mutual induction between the core and the sheath as well as the high capacitance of the cable, the fault location in alternative current (AC) cable is more complicated than the head transmission line. By using distance protection scheme for AC transmission line, the seen impedance by the relay has a nonlinear behavior with respect to fault location. In this paper, with the help of extreme learning machine (ELM), the fault locating algorithm is implemented by using the measured values of voltage and current of core and sheath on both sides of the cable. The proposed algorithm can detect the non-linear and complicated relations between measured quantities and fault location. In the system under study, at first, the core to sheath faults are simulated in the PSCAD/EMTDC software considering different fault resistances and different fault distances. Then, in order to train the intelligent core of the proposed method, input vectors are extracted for different conditions and a desirable output is considered corresponding to the fault distance. Examination of the results obtained from the use of various intelligent tools shows the superiority of the ELM over the ANN and SVM in terms of accuracy of and learning speed.

This paper presents an intelligent method for online voltage stability margin (VSM) assessment using optimized adaptive ANFIS. Harris Hawks Optimization Algorithm (HHOA) is used to train the ANFIS and conventional wavelet transform (WT) is also applied as a feature extraction technique on the network voltage profile. The network voltage profile is used as the main data to estimate VSM because it contains the necessary information about the network structure, load levels, production pattern, and control system performance in the network. Using wavelet transform technique with high resolution, the necessary features for entering the ANFIS block are extracted, but due to the variety and multiplicity of these features, especially for large networks, the Principal Component Analysis (PCA) method is used to select the appropriate features and remove additional data. The characteristic of this hybrid algorithm is that it can be used both in dynamic and static conditions of the network. Finally, the proposed VSM estimation algorithm is applied to the 39-bus and 118-bus IEEE test systems, and its results are evaluated. The comparison of the results with other VSM methods shows that the proposed algorithm is effective for large power grids.

Residential load forecasting plays an important role in management and planning of modern smart grids. Accurate residential load forecasting is needed in planning to keep demand and supply balanced. This paper presents a mid-term residential load forecasting method based on feature selection to solve the linear regression problem. In this way, for feature selection to perform the regression, the neighborhood component analysis method is used. For this purpose, an optimization problem is designed, and the problem is solved using the LBFGS algorithm. The AMPds2 dataset is used to implement the proposed method, and the results were compared with the results of the other six forecasting methods. Comparisons were made through mean squared error, root mean squared error, and mean absolute percentage error. The simulation results confirm the effectiveness of the proposed method for accurate residential load forecasting.