M. Szewczyk

Neural networks on transformer fault detection evaluating the relevance of the input space parameters

Following a number of studies that have employed different forms of neural network models to perform dissolved gas-in-oil analysis (DGA) of transformer bushings, this manuscript focuses on evaluating the relevance of the parameters that form part of the model input space. Using a multilayer neural network initially populated with all the 10 input parameters (10V-Model), a matrix containing causal information about the possible relevance of each input parameter is obtained. The information from this matrix is proven to be valid through the construction and testing of another two, separate, multilayer networks. One networks input space is populated with the 5 most relevant parameters (MRV-Model), while the other is populated with the 5 least relevant parameters (LRV-Model). The obtained classification accuracy values are as follows: 100% for the 10V-Model, 98.5% for the MRV-Model, and 53.0% for the LRV-Model.

Distance Protections in the Power System Lines with Connected Wind Farms

In recent years there has been an intensive effort to increase the participation of renewable sources of electricity in the fuel and energy balance of many countries. In particular, this relates to the power of wind farms (WF) attached to the power system at both the distribution network (the level of MV and 110 kV) and the HV transmission network (220 kV and 400 kV)1. The number and the level of power (from a dozen to about 100 MW) of wind farms attached to the power system are growing steadily, increasing the participation and the role of such sources in the overall energy balance. Incorporating renewable energy sources into the power system entails a number of new challenges for the power system protections in that it will have an impact on distance protections which use the impedance criteria as the basis for decision-making. The prevalence of distance protections in the distribution networks of 110 kV and transmission networks necessitates an analysis of their functioning in the new conditions. This study will be considering selected factors which influence the proper functioning of distance protections in the distribution networks with the wind farms connected to the power system.

ANN-Based Fault-Type and Location Identification System for Autonomous Preventive-Restoration Control of Complex Electrical Power System Plants

In modern electrical power systems (EPS) an increasing importance of generating nodes, such as: pump storage plants or combined power plants including gas-turbine as well as steam-turbine generators, is observed. Such objects posess composed but highly redundant structure offering a significant opertion flexibility e.g. the adjacent start-up system use in a fault case in the dedicated start-up system, or two generating sets co-operation with one step-up transformer. The essential features of these objects are: very good power regulation abilities, short start-up time and internal structure modification possibility. The above features determine the growing importance of such objects in the EPS operation control process, aiming at the optimal operation mode as well as fast and effective restoration after a system fault occurrence, which may be ensured by the adaptive protection and control system (APCS) - sufficiently reliable both in normal operating conditions and in a fault case. One of the basic functional modules included in APCS is the autonomous preventive restoration control module (APRCM), developing the object control rules and preventing it from a fault or limiting possible fault effects via fast restoration of generation and regulation abilities.

Adaptive decision-taking of protection systems in networks with dispersed power generating sources

Incorporating renewable energy sources into the power system is connected with a number of new challenges, i.e. for the power system protections. Thus, several problems have been identified in addition to the factors which occur in such systems without the connected i.e. wind farms. Taking the presence of wind farms into account, the selectivity and the speed of decisions of distance protection will probably be very limited. One of the possible solutions will be creating protection systems which use data from the selected area of the power system network to increase the correctness of decision-taking by adapting settings of protection functions to the changing conditions. This paper will discuss the proposal of such a solution as well its simulation results.

A neural network approach to adaptive protective systems problem in the complex power generating units

Practical successes have been achieved with neural network models in a variety of domains, including energy‐related industry. The large, complex design space of electrical power systems (EPS) is only minimally explored in current practice. The satisfactory results that nevertheless have been obtained testify that neural networks are a robust modeling technology; at the same time, however, the lack of a systematic design approach implies that the best neural network models generally remain undiscovered for most applications. This paper presents an approach to an adaptive protective systems problem in the complex power generating units. First, we demonstrate the complex interdependencies between various parameters of EPS protection systems. We then present an approach, based on protection and adaptation criteria, for designing a neural network based adaptive protection system. © 2000 John Wiley & Sons, Inc.

Neural Network-Based Adaptive Protection System

Practical successes have been achieved with neural network models in a variety of domains, including energy-related industry. The large, complex design space of electrical power systems (EPS) is only minimally explored in current practice. The satisfactory results that nevertheless have been obtained testify that neural networks are a robust modeling technology; at the same time, however, the lack of a systematic design approach implies that the best neural network models generally remain undiscovered for most applications. This chapter presents an approach to an adaptive protective systems problem in complex power generating units. First, we demonstrate the complex interdependencies between various parameters of EPS protection systems. Then, we present an approach, based on protection and adaptation criteria, for designing a neural network based adaptive protection system.

Programming Work Simulator of the Ann-Based Integrated Protection System for the Reversible Hydro Generating Sets

This paper presents a programming work simulator worked out for verification of the new artificial neural networks - based approach to the decision making systems for the power protection automation purposes.

Intelligent, ANN-Based Identification System of Complex Generating Sets Operating Mode

The effective operation of the electrical power system (EPS) protection devices is mainly determined by a correctness of generated decisions, classifying actual state of the protected object to one of the two event classes (i.e. normal or fault), which may be realized only if the following conditions are considered: the acquisition of extended information about the protected object is enabled, correct identification of the object operating mode, data processing and decision making speed, the ability of the protective system adaptation to the actual object state and its “insensitivity” to any kind of interference. On the other hand the use of digital technology allows new approach to the protected object state identification and protection functions adaptation. It may be achieved by implementing additional criteria (both measuring and logical) and by use of new computational methods, based upon artificial intelligence such as artificial neural networks (ANNs).

ANN-based Power Unit Protective System

Numerous changes introduced to the input and measuring elements of modem digital protective arrangements considerably improved their reliability and access to the extended information about processes in the protected object, they did not cause however any essential improvement of the protection decision-making. To extend the possibilities of the decisionmaking units of the protective systems the ANN structures are proposed for the proper fault detection, on the base of the measuring signals delivered from the protected object and its environment. In the generator-transformer units the criterial quantities (voltage and current) and their components measurement is performed for the protected object classification to one of the event classes: 9 protected unit normal condition, 9 external fault, 9 internal fault.