Predictive Control Model for Radiant Heating System Based on Neural Network

A predictive control model of radiant floor heating system is developed based on BP (back propagation) neural network, Which consists of input layer(14), hidden layer(16) and output layer(1). The model is trained with the experiment data and the on-line correction predictive control is conducted. The maximum relative error between the indoor temperature given by the on-line correction predictive control and that measured in the experiment is only some 6%. The model will be used to improve the control accuracy of radiant floor heating system and the level of indoor thermal comfort by controlling temperature.