COMPUTATIONAL INTELLIGENCE IN ELECTRICAL ENGINEERING (INTELLIGENT SYSTEMS IN ELECTRICAL ENGINEERING)
Year:2013 | Volume:4 | Issue:1|Papers Count:6

Cellular positioning methods in urban environments suffer from a significant error due to multipath and fading phenomena. In this paper, regarding the implementation location-based services, one pattern recognition positioning method based on the signal strength is presented, which provides proper results in urban areas. Although many works have been done in this field, the main challenge is to create and to update the database and heavy calculation to estimate the position .In this paper, some practical and intelligent solutions for overcoming the mentioned problems, enhancing the accuracy and reducing the computational load of this positioning method are presented. Specially, some intelligent filtering methods are presented for reducing the search region in the database and also increasing the accuracy of position estimation. Real measurements obtained in Isfahan, illustrate the high accuracy of the proposed techniques. In addition, an advertising service based on the proposed method is presented. User’s interest and local time is considered as well as him/her position for enhancing the efficiency of the mentioned service.

EEG is one of the most important and common sources for study of brain function and neurological disorders. Automated systems are under study for many years to detect EEG changes. Because of the importance of making correct decision, we are looking for better classification methods for EEG signals. In this paper a smart compound system is used for classifying EEG signals to different groups. Since in each classification the system accuracy of making decision is very important, in this study we look for some methods to improve the accuracy of EEG signals classification. In this paper the use of Mixture of Experts for improving the EEG signals classification of normal subjects and patients with epilepsy is shown and the classification accuracy is evaluated. Decision making was performed in two stages: 1) feature extractions with different methods of eigenvector and 2) Classification using the classifier trained by extracted features. This smart system inputs are formed from composites features that are selected appropriate with network structure. In this study tree methods based on eigenvectors (Minimum Norm, MUSIC, Pisarenko) are chosen for the estimation of Power Spectral Density (PSD). After the implementation of ME and train it on composite features, we propose that this technique can reach high classification accuracy. Hence, EEG signals classification of epilepsy patients in different situations and control subjects is available. In this study, Mixture of Experts structure was used for EEG signals classification. Proper performance of Neural Network depends on the size of train and test data. Combination of multiple Neural Networks even without using the probable structure in obtaining weights in classification problem can produce high accuracy in less time, which is important and valuable in the classification point of view.

Electricity exchange is known as the trajectory of power industry restructuring which facilitates achieving a direct fair market and accomplishing the privatization. However, little knowledge about market participants and their decisions may increase the investment risk and affect the economic boom cycle. As the Iranian electricity stock has started working in 2011 and the subscription has been made, it will start its services in near future; so, it would be very helpful to inform the investors about what may happen, and how to direct their portfolio into the satisfactory corridor. One way to arrive capital investment security is to predict insolvency of a business unit. Predicting the possibility of a company’s insolvency not only can prevent losing the principle and capital interest of investing, but also facilitates the most important decision makings. This paper proposes a new model for insolvency prediction of the Iranian electrical firms within future electricity exchange via an artificial bee colony algorithm. To do so, 118 firms among the electrical and energy industrial firms listed in Tehran Stock Exchange (TSE) are assumed; they are used as training data to find a suitable linear classifier. The introduced algorithm is conducted on 40 test firms and obtained results are discussed in several scenarios.

In designing a classifier and estimating the optimum decision hyperplanes, the main goal is often high “recognition score” in the training phase. While other objectives such as “reliability” of decisions and “the number of optimum decision functions” are also very important factors during designing classifiers, which should never be forgotten. In this paper, at first, the central force optimization (CFO) algorithm developed for multi-objective optimization problems (namely, MOCFO), and then a novel multi-objective classifier is constructed based on using the proposed MOCFO and called MOCFO-classifier. In fact, MOCFO-classifier optimizes the above objectives simultaneously. Due to selecting the number of optimal hyperplane in the proposed method, the important issues; “overfitting” and “overtraining” have also been removed. The experimental results on difficult test data show that the proposed multi-objective classifier provides a set of various and optimum options of the hyperplane that separate different classes for making user ideal conditions in regards to selecting mentioned aspects.

The acoustic noise pollution is one of the serious disasters in the current industrialized life. Though traditional solutions based on noise absorption have many different applications, but these methods have low performance for low frequency noises. Active Noise Control (ANC) has been introduced to resolve this problem. In this paper, a new active method is introduced for suppressing acoustic noises based on the reinforcement learning. To achieve this, an algorithm to control periodic noises is suggested. Then, the method is developed further to deal with multi-tonal signals with a large number of harmonics. At the next step, the broad-band signals are considered. The problem is broken into some sub-problems in frequency domain and each is solved via a reinforcement learning approach. In all of the proposed techniques no model for the environment is needed. Combining the reinforcement learning and the traditional methods in ANC for broad-band signals is a new line research considered here. This combination could increase the speed of the response, but some information of the dynamics of the environment is needed. This will cause the system to become compatible with gradual changes of the environment. Simulation results show the effectiveness of the proposed approach.