Model of room storage heater and system identification using neural networks

This paper presents two approaches used to develop a model of Room Storage Heater. The first one consists of a dynamic model of the RSH developed by the authors using the results obtained from tests performed in a calorimetric chamber. The model was verified against the results obtained during five different charge-discharge test periods. The second approach is a new concept based on Neural Networks applications. In this approach, we suppose that we do not have a description of the RSH itself. The input data in a neural network training are as follows: the immediate paste bricks temperature, the room temperature, the electric power input and the on/off activation function of the fan. The energy released and the current brick temperature were the neural network outputs. The results of two training and test procedures are presented. In the first procedure we use the results of the tests performed in the calorimetric chamber which are sufficient to develop the dynamic model but they appear not adequate for the neural networks application. Consequently, the second NNs training and test were conducted with the modified training data set which was obtained by the simulations performed using the RSH dynamic model. Two comparison are presented : comparison of the NNs and simulation results and comparison of the NNs and calorimetric chamber test results. The NNs model accuracy seems to be very good. It is comparable with the dynamic modelization methods