REVISTA DYNA ENERGÍA

Key words:

Sistema de Agua de Alimentar Caldera, Modelo Híbrido, Red Neuronal Artificial, Regresión Estadística, Optimización.Boiler Feed Water Pumps System, Hybrid Model, Artificial Neural Network, Regression Statistics, Optimization.

Article type:

ARTICULO DE INVESTIGACION / RESEARCH ARTICLE

Section:

RESEARCH ARTICLES

A method is proposed to increase energy efficiency in the boiler feed water pumps system of power plant based on Rankine cycle. Where is used a hybrid model based on parallel implementation of a statistical regression (REG) and an artificial neural network (MLP), for model the dependence of the current consumption and the position of the speed variator, in function of the discharge pressure and flow delivered by each pump. The hybrid model is used for its good accuracy in prediction of the variables modeled compared with REG and MLP methods. For the fit of the models is used real measurements of operation, obtained of the Supervisory Control and Data Acquisition (SCADA) system, where is applied a method data processing, eliminating noise that can affect the fit of the models. Obtained the models, an optimization method is used to determine the optimal distribution of flows delivered by each pump, minimizing the current consumption of the ensemble of pumps that is in service. To validate the proposed method, a study case is used and the results obtained are analyzed.