Unsupervised neural network for forecasting alarms in hydroelectric power plant

摘要: Power plant management relies on monitoring many signals that represent the technical parameters of the real plant. The use of neural networks (NN) is a novel approach that can help to produce decisions when integrated in a more general system. In this paper we introduce a NN module using an ART-MAP to discriminate different situations from the plant in order to prevent future malfunctions. A special process to generate of a complete training set has been designed. This process is developed in order to deal with the absence of data in abnormal plant situations. This module belongs to a more general system for predictive maintenance that has been implemented and incorporated in an hydroelectric plant.