Applying neural network-based tuner to optimize parameters of PI-controller for heating furnace functioning in different modes

We propose a neural network-based tuner for online optimization of parameters of an automatic PI-controller for heating furnace control. The tuner consists of two neural networks responsible for adjusting coefficients KP and KI for furnace heating and cooling processes respectively. We develop a structure of a neural tuner and show by model experiments that such a tuner can be applied to control heating furnaces with the different value of the time constant. A muffle electric heating furnace functioning in different loading modes has been chosen as a plant. Having made our experiments, we conclude that such an optimizer helps to achieve about 23% decrease of time length and 19% decrease of energy consumption for each schedule in comparison with a conventional PI-controller.