Preventive control formulation in electrical distribution energy systems based on fuzzy logic and neural network

Abstract

This work aims to develop an intelligent system to the preventive control (proactive system) in distribution electrical systems. It is a procedure to identify critical situations in advance, of faults that can damage the components and lead the system to interrupt the supplying to the users. This proactive system is based on fuzzy logic and neural networks, in special the ART (Adaptive resonance Theory) family architecture, i.e., the ARTMAP-fuzzy neural network. This structure is chosen due to be stable and plastic, assuring to obtain a solution and introducing the continuous training, respectively. Furthermore, the solutions are obtained very fast, being very interesting to applications in real time. The continuous training is an alternative that allows extracting the knowledge without needing to restart the training when new pattern are available to file in the weight matrices, which is different from the usual neural networks. Therefore, it is possible to use a reduced set of pattern on the training phase and as the analysis is being effectuated the extraction of the knowledge is continued, i.e. a system that is improved as the time goes on. The system starts introducing as principal contingencies the voltage disturbances. New kind of contingencies can be introduced during the developing of the work. Thus, it deals with a system that can permanently incorporate new information to become an efficient inference system. (AU)