Session: 06-01 Thermochemical Redox Cycles: Demonstration and Operation

Paper Number: 115024

115024 - Demonstration of a Model Predictive Control for a Cluster of Solar Chemical Batch Reactors 

Chemical batch reactors driven by concentrated solar energy require regular adaptation of the mass and heat flows provided to them. For a quasi-continuous production of chemicals or fuels and to make best use of the solar resource, a receiver comprised of an array of batch reactors is envisaged on a solar tower. Other than for a typical concentrated solar power (CSP) plant, the heliostat field needs to provide a non-uniform and time-varying solar flux density distribution to the solar receiver in this case. For an efficient operation, it needs to be optimized how the available power of the heliostat field is distributed to the reactors. For each reactor, it must be decided when to proceed to the next process step, which results in a challenging discrete optimization problem. The fact that the spillage of one reactor can hit the other reactors and that the overall solar resource is not constant further complicates the control task.

Examples for solar batch reactors are the ones with a fixed, directly irradiated monolithic redox material in thermochemical redox cycles for hydrogen production, for example in the Sun2Liquid and HydroSol projects. In these reactors, the redox material is first reduced under solar irradiation at temperatures above 1400°C and then oxidized with less or no irradiation at lower temperatures of about 900°C by water vapor, carbon dioxide or a mixture of these gases. Hydrogen, carbon monoxide or synthesis gas is produced, respectively.

Here, we present different control approaches for these solar batch reactors and focus on the most sophisticated one, which is currently being demonstrated at the solar tower Jülich. In this approach, a model-predictive controller (MPC) optimizes the plant operation within the next 15 minutes. Making use of a simplified reactor model, it simulates the future states of all reactors. It is fed by predictions for the available solar power at the receiver, obtained from the cloud camera system WobaS producing probabilistic nowcasts coupled with the ray-tracing tool STRAL for the heliostat field. The MPC also receives live sensor data from the reactors. In the demonstration, we use one real thermochemical reactor in the 250 kW scale and four reactor mockups due to cost reasons. On the mockups, the solar flux density is measured and fed into a detailed reactor model, from which virtual sensor data in the same probe locations as in the real reactor is generated. The MPC sets the desired fluxes for the reactors and sends them to the heliostat field control system HeliOS, which sets the aim points for the heliostats.

Successful hardware-in-the-loop simulations of the entire system are presented. They show plausible operation of the MPC for the five reactors at the solar tower Jülich. Interesting insights not only for the operation of thermochemical batch reactors, but also for operating solar chemical reactors on a solar tower in MW scale in general are shared. An outlook to the experimental demonstration at the solar tower Jülich is given and first test results are shown.

Presenting Author: Johannes Grobbel German Aerospace Center (DLR)

Presenting Author Biography: Johannes Grobbel studied Mechanical&Energy Engineering at RWTH Aachen, Germany and UC Davis, CA, USA. He joined DLR in 2014 for his Master thesis on convective losses of cavity receivers. In the following, he worked on a vacuum solar particle receiver for hydrogen production. Motivated by this project, he investigated the use of the Discrete Element Method (DEM) to simulate solar particle receivers in his dissertation, which he defended in 2019. In this context, he developed DEM heat transfer models and determined DEM contact model parameters. Currently he is working in several solarchemical projects, mainly regarding the solar fuel production via solar thermochemical redox cycles.