(203d) Vision-Based Solar Irradiation Prediction for the Control of Solar-Thermal Reactors | AIChE

(203d) Vision-Based Solar Irradiation Prediction for the Control of Solar-Thermal Reactors

Authors 

Saade Saade, E. - Presenter, University of Colorado
Clough, D. E., University of Colorado
Weimer, A. W., University of Colorado at Boulder


One of the biggest challenges that solar thermochemical processes face is dealing with the intermittent nature of solar irradiation. Changes in solar irradiation directly affect the performance of solar-thermal reactors causing unnecessary shutdowns and start-ups, complications in the purification processes downstream of the reactor, and damage to the reactor materials due to thermal shock. Being able to maintain a continuous high performance operation, even in the presence of passing clouds, is one of the main concerns of the feasibility of solar-thermal processes. 

In order to address these issues, a control system that manipulates the flow rates into the reactor is required. The proposed control system is a model predictive controller, which makes use of a model of the system in order to determine the manipulation required to minimize the error between the process and a reference trajectory.  In addition to a model of the system and current measurements of the controlled variables, the model predictive control system requires knowledge of the disturbances. In this case, the available solar irradiation is treated as a measured disturbance and it is estimated using images of the sky. In order to do this, a digital charge coupled device (CCD) camera (YES Total Sky Imager TSI-880) captures images of the sky and analyzes them to determine the presence of clouds and their characteristics (density of white, speed at which they are moving, time before they cover the sun, etc.). This information is combined with other atmospheric measurements, such as wind speed and relative humidity, and used to estimate the solar irradiation at future steps. The predictions of the disturbances are used, in combination with a dynamic model of the process, to determine the required manipulation at every time step. In this paper, the algorithm used to determine the presence of clouds is described, as well as the method used to estimate the solar irradiation at future time steps.