Neural simulation of a solar thermal system in low temperature

The operation of a low-temperature solar thermal system using artificial neural network (ANN) models of its components (flat-plate solar collector, internal heat exchanger, and stratified tank), and its workings with dynamic and static modes, has been simulated. The ANN models of these components, used as blocks, have been previously formulated using the experimental data of solar irradiance, ambient temperature, flow and temperature of the working fluid and water supplied to the tank, and stratification temperatures in eight levels of the tank, measured under the continental Mediterranean climate conditions of central Iberian Peninsula. The simulation, executed in intervals of 1 minute, was run on 2 days for each month of 1 year. The f-chart method was used to validate the neural simulation under the same conditions (without stratification) for 10 years, resulting in an average deviation of the performance of 1.85%. The results for 1 day at stratification temperatures show a root mean square error value of 0.77°C in dynamic operation mode and of 0.13°C in static operation mode.