Adriano Peres de Morais

Alarm processing and fault diagnosis in power systems using Artificial Neural Networks and Genetic Algorithms

This work approaches relative aspects to the alarm processing problem and fault diagnosis in system level, having as purpose filter the alarms generated during a outage and identify the equipment under fault. A methodology was developed using Artificial Neural Networks (ANN) and Genetic Algorithms (GA) in order to resolve the problem. This procedure had as initiative explore the GA capacity to deal with combinatory problems, as well as the ANN processing speed and generalization capacity. Such strategy favors a fast and robust solution.

Performance Evaluation of the Adaptive Loss of Field Protection in Synchronous Generators by means of the Positive Offset Method

This paper aims to evaluate the performance of the adaptive loss-of-field protection in synchronous generator based on the rate of change of the apparent reactance. It was accomplished by considering the well known Positive Offset method. Comparisons were made by means of computer simulations. The qualitative analysis is carried out to point out the advantages and disadvantages of each method according to different generators parameters. In this way, it was identified circumstances in which one method is better than other, in order to aid the protection engineers in choosing the more appropriate.

New setting of loss of excitation protection in P-Q plan in order to maximize the operation area of the capacity curve of the synchronous machine

This paper presents a method for maximizing the operation area of the capacity curve of a synchronous machine. The technique is set in P-Q plan in order to adjust more features of minimum excitation and practical stability limits of the capacity curve. Thus, the voltage limiter can be intended to the left of the capacity curve and so the machine may have an operating gain in this region. The new setting is done primarily on the generator capacity curve, plan P-Q, and subsequently analyzed in the R-X plan.

Fault location in distribution systems: A Method Considering the Parameter Estimation Using a RNA Online

This paper proposes a new methodology for single-phase short-circuit location in power distribution systems. The feeder behavior during single-phase faults is modeled in terms of the distance of the fault, the network parameters and the currents and voltages measured only at the substation. Thus, the proposed method does not require the installation of any additional measurement equipment in the network. The capacitances of the network were also considered in the model used, making it closer to the real system. The main innovation of the proposed method is the use of a neural network, which does not require prior training, to estimate the unknown parameters of nonlinear equations that model the feeder behavior during faults. The performance of the proposed method was evaluated in the test system IEEE 34 bars through the variation of fault resistance and load feeder. Furthermore, it was also evaluated the influence of the estimated short-circuit current in the determination of the fault location. The good results obtained, combined with simplicity and low cost of implementation, make the proposed method promising for application in a real feeder.

Transients detection and classification in distribution networks for high impedance faults identification

Protection systems used in distribution networks are not able to detect short circuits with high contact resistance due to reduced currents generated, endangering the population and degrading the quality of the energy supplied. In this sense, this paper presents a new technique for detecting high impedance faults (HIF). The major advantage of this method is the not needing of additionally equipment installation in the distribution network, which reduces the cost of its implementation. Moreover, the proposed methodology is immune to any electrical transients generated in the system, such as capacitor bank switching, power transformers and inputs and output loads.

Integrated use of artificial neural networks and genetic algorithms for problems of alarm processing and fault diagnosis in power systems

This work approaches relative aspects to the alarm processing problem and fault diagnosis in system level, having as purpose filter the alarms generated during a outage and identify the equipment under fault. A methodology was developed using Artificial Neural Networks (ANN) and Genetic Algorithms (GA) in order to resolve the problem. This procedure had as initiative explore the GA capacity to deal with combinatory problems, as well as the ANN processing speed and generalization capacity. Such strategy favors a fast and robust solution.

New Methodology for Identification of Sympathetic Inrush for a Power Transformer using Wavelet Transform

This paper presents a new methodology to identify the transient of sympathetic inrush between power transformers that aims to prevent the protection system from being affected by this transient. The developed method is based on the variation of the detail coefficients of different levels of DWT decomposition. During the sympathetic inrush condition, the detail coefficients present a different variation from that which occurs during an internal fault condition. The proposed methodology is able to identify the sympathetic inrush conditions without the need to set a threshold, making the methodology intelligent and avoiding unwanted trip of the protection system. The proposed methodology was thoroughly tested and proved to be promising in the identification and discrimination between sympathetic inrush and internal faults.

Curve fitting analysis of time-current characteristic of expulsion fuse links

This work aims to modeling the Time-Current Characteristics (TCC) of expulsion fuse links through mathematics equations. To achieve this, the Least Square Method has been used and the Residual Sum of Squares (RSS) as a comparative parameter. From this parameter, it is possible to identify models that have the better adjustment in global scope considering the different nominal current ratings, speed ratios and manufacturers. The coefficients from the equations with the minor RSS will be available, so researchers of distribution system protection can use it in future works.

A comparative analysis of loss of excitation protection methods for synchronous generators

Synchronous generators require to be protected against loss of excitation (LOE) because this event threatens the generator and power system stability. Over the years, several methods have been proposed for LOE detection. Thereby, this paper aims to show the performance of the loss of excitation protection methods in synchronous generators. Comprehensive simulation studies of total LOE and partial LOE are accomplished for various generator conditions and system disturbances. The performed studies demonstrate depending of initial generator loading, a partial loss of excitation may damage the machine. Still, in order to test the methods security to Stable Power Swings have been done. The results show that majority of the methods discussed presented difficulties in the detection of partial loss of excitation.

Hybrid Method for Protective Devices Placement, Sizing and Coordination in Electric Distribution Systems

This work presents a hybrid method to solve the protective devices allocation problem in electric distribution systems. Current methods only consider optimization results, without taking into account standards established by electric energy companies. These standards reflect distribution systems particularities observed by these companies. Firstly, recloser placementis determined through an optimization process, aiming to reduce reliability indices. Fuse links are placed according to heuristic rules established by electric companies. To validate the proposed methodology, a comparison with three other references from the literature is carried out, showing good results and proving the methods effectiveness.