Artificial Neural Networks Applied to Power Systems

摘要

Artificial neural networks (ANNs) are a relatively new branch of artificialintelligence. Based on models of the human brain, these networks offer features that are copied from the brain, such as pattern-recognition-based operation, content addressability, spontaneous generalization, fault-tolerant behavior and training to presented examples. These properties can be very useful when applied to the automation of processes in a technical environment. The aim of this thesis is to indicate the usefulness of this new technique when applied to power systems. Three case studies have been carried out that are selected from existing processes in the power system domain. The case studies are adaptive distance protection, load forecasting and alarm processing. These subjects are selected because they are not (yet) covered by existing methods, or because existing methods do not perform satisfactorily. The common elements of the selected cases are the non-linear behavior, the problems concerning the analysis of the process and the potential corruption of signals. The properties of ANNs indicate the usability of this technique in such processes. The general conclusions of this research are, therefore, that artificial neural networks are a useful addition to existing techniques, and that the application of ANNs is simple when the basic principles of them are understood.