Control adaptativo para arranque de generadores de vapor mediante redes neuronales dinámicas

in this document it will be explain the modeling and adaptive control by means of auto-recurrent neural networks for steam generators start-up with the purpose of obtain a minimum in the start-up time, the variance and the failures for manual settings of the process variables. the primary start-up process variable to control is the temperature, but in addition will be control the oxygen for guarantee the maximum performance of the burner combustion. both variables were controlled setting the relation between the natural gas and combustion air. the training data are empiric of the process and it was sampled every 2 minutes. the process and controller simulation was realized in matlab. the controller actions obtained in the simulations had concordance with a normal start up of the steam generator.