L.E. Borges da Silva

A fuzzy knowledge-based system for bus load forecasting

The authors describe an alternative approach to short-term load forecasting. The approach merges traditional mathematical techniques and fuzzy concepts in a knowledge base. The rules of this knowledge base are devised using the historical data of the bus and are represented by fuzzy conditional statements. An example using real data from an actual power system is presented.<<ETX>>

Knowledge-base reduction based on rough set techniques

Knowledge acquisition is one of the most difficult tasks during the construction of an expert system. Usually, the experts have difficulty in explaining to the knowledge engineers how they solve a given problem. This fact may result in superfluous information about some specific points, while in other points, only an incomplete set of information is available to prepare the knowledge base. This paper presents a contribution to help knowledge engineers to manipulate and reduce knowledge bases for power system operation problems using a systematic approach. The approach is based on rough set theory. An illustrative example is presented in this paper.

Neural net adaptive schemes for DC motor drives

The use of neural networks in a speed control loop applied to a DC motor is described. The proposed technique makes use of the learning capability of neural networks to implement an autoadaptive control structure. This capability allows the network to learn the dynamic behavior of the SCR-driven DC motor. This identification network is then used to train another network as the process controller, so that the process output follows the reference signal. An adaptation scheme for working conditions is also presented. Its performance is verified through testing of physical parameter variations and noise presence, showing the applicability of the system.<<ETX>>

Optimizing the Series Active Filters Under Unbalanced Conditions Acting in the Neutral Current

This paper discusses the limitations of the series active power filter operating in unbalanced three-phase four-wire systems and proposes the insertion of series active impedance to neutral wire. The control of this active impedance uses a modified synchronous reference frame controller (MSRFC) in order to mitigate the flow of zero sequence harmonic components. The performance of the combined system, constituted of two series active filters, is simulated with and without the insertion of a series active filter to the neutral wire and the results arc presented. The simulation confirms the viability of proposition and the discussions show the importance of using this proposed structure when series active filters, based on three-phase synchronous reference frame controller (SRFC), is used as control strategy in unbalanced four-wire systems.

A simple and ultra-fast DSP-based space vector PWM algorithm and its implementation on a two-level inverter covering undermodulation and overmodulation

The paper proposes a simplified algorithm of space vector modulation (SVM) for a two-level voltage-fed inverter. The algorithm is faster and more flexible than any other algorithm so far proposed in the literature. Another relevant advantage of this approach is that a single algorithm covers the undermodulation and overmodulation regions including the square-wave operation with almost no additional computation cost The simplification of the algorithm is achieved by exploring (1) the coupling between the angle and amplitude of the command voltage vector in the final result of the algorithm, (2) extrapolation of the undermodulation strategy into the overmodulation region, (3) reduction of the number of equations to calculate the switching times, and (4) simplification of the sector identification strategy. As a result, the algorithm can be implemented by using only basic mathematical operators (addition, subtraction and multiplication), Boolean logical operators, and a small one-dimensional look-up table. This simplified algorithm permits SVM implementation at higher switching frequency without loosing any good features of the conventional SVM algorithm. Matlab/Simulink simulation results and laboratory experimental results using DSP type TMS320F240 for a Volts/Hz controlled induction motor drive with 20 kHz switching frequency indicate excellent performance of the system.

Improving performance of slip-recovery drive: An approach using fuzzy techniques

Conventional slip-recovery drives are plagued with a low overall power factor, due to the naturally commutated power convertor in the rotor circuit. One solution consists of adding two GTOs to the converter bridge, in order to transform it into a fully controllable inverter, which allows for independent control over voltage and displacement factor. When appropriately controlled, this converter is capable of producing reactive power in just the right amount to fully compensate the reactive power consumed by the power induction machine. Unity power displacement factor is then attainable over most of the operating range. A controller based on a fuzzy logic expert system is proposed.<<ETX>>

Real-time determination of the best interval of calculation for moving averages used for DC value extraction in active power filter control methods

In many cases, active power filters based on the i/sub d/-i/sub q/ and p-q control methods can have their dynamic response improved and their computational effort reduced by using moving averages instead of filters for the DC value extraction. An optimal dynamic performance, however, can not be achieved without the knowledge of the shortest calculation interval for the moving averages. The best interval depends on the harmonic content of the load currents and/or supply voltages that will be compensated. This paper presents a method for the instantaneous determination of the shortest calculation interval of the moving averages. The paper describes the strategy and presents the simulations.

Simulation of a neural net controller for motor drives

This paper describes the use of neural networks in a speed control loop applied to a DC motor. The proposed technique makes use of the learning capability of neural networks to implement an auto-adaptive control structure. Such capability allows the network to learn the dynamic behavior of the SCR-driven DC motor. This identification network is then used to train another network as the process controller, so that the process output follows the reference signal. An adaptation scheme for working conditions is also presented. Its performance is verified through testing of physical parameter variations and noise presence, showing the applicability of the system.

Using partial discharge as sample signal source to identify contamination moisture pattern in power transformer insulating oil

The properly behavior of power transformers is highly dependent of the insulation health. So, the insulation condition assessment can provide information to a system of diagnosis and prognosis, allowing for planning interventions at the appropriate time to avoid unexpected interruptions. This paper presents a technique that can detect moisture contamination in the insulating oil using acoustic signals of partial discharge. In this work, a piezoelectric sensor is mounted in a reservoir with insulating oils with different levels of moisture. A device to simulate partial discharges is immersed in those oils. The detected signals are processed using Discrete Wavelet Transform to extract time-frequency characteristics which are processing using the Principal Component Analysis in order to identify patterns of water content in insulating oil. The results are presented and the methodology has shown a very good technique to be implemented as on line monitoring system.

Detection of partial discharge in power transformers using Rogowski coil and multiresolution analysis

Partial discharge detection in power transformers is discussed using a new approach that exploit the broad band of the Rogowski coils and the potential of two signal processing tools: discrete wavelet transform and empirical mode decomposition. Detecting and analyzing incipient activities of partial discharge can provide useful information to diagnostics and prognostics about transformer insulation. So, partial discharge signals embedded in the electric current at ground conductor are measured using the Rogowski coil. These signals are submitted to noise suppression and the partial discharges waveforms are extracted through different ways: using discrete wavelet transform and using empirical mode decomposition. These two methods are compared and their advantages discussed.