Artificial Neural Networks for Energy Management System Applicability and Limitations of the Main Paradigms
The practical applicability and limitations of the main neural paradigms is revised by studying three main questions: when can a particular problem be approached by means of Artificial Neural Networks? Which is the most suitable neural paradigm for a particular problem? How must the available information be presented to the implemented network? As a case of study, three operations involved in an Energy Management System, instances of general problems often solved by neural networks, have been considered: Load Forecasting, State Estimation and Contingency Analysis. The analysis of the attempted neural solutions brings to light the features and limitations of the main neural paradigms, such as Multilayer Perceptrons with Backpropagation, Radial Basis Function networks, Hopfield networks and Self-Organising Maps.
Gonzalo JOYA, Francisco GARCÍA-LAGOS, Miguel A. ATENCIA, Francisco SANDOVAL
Artificial Neuronal Network, Energy Management System, Neural Paradigms Applicability, Input Selection