Innovative advanced control of large cryogenic systems to save energy and reduce carbone emission – CRYOGREEN

An approach based on existing dynamic modeled will be used. Physical models will be simplified in order to get control oriented models. A generic tool to recover control-oriented models will be offered. Based on the control oriented modeling, a division into sub-systems of the refrigerator will be carried out.

We plan to use a distributed control architecture optimized communication. It is based on the identification of the relationship between the constrained optimal control and the sequence of past measurements. The optimal control scenarios will be generated through the control oriented simulator where time varying MPC will be used.

Validation phases of a midsize cryogenic refrigerator available at CEA-SBT are planned. The final tests will be performed to validate the complete development. They will be made on a refrigerator simulator or on a large cryogenic refrigerator (depending on availability), both installed at CERN. This test will help validate the scalability of our solution.

1- A Matlab / Simulink / Simscape library named simCryogenics to model any refrigerator has been developed with the objective to obtain linear models to be used for control. This tool is based on a library of cryogenic components to be assembled to obtain the model of the refrigerator. After extraction of the operating point, simCryogenics extracts the linear system model. Initially this tool can be used to quickly set PID controls (which was not originally envisaged). It will be used for the time varying MPC controls derived based on obtained linear models. We used a generic product on all installations.

2- A tool for implementing constrained predictive control with time variant parameters in a programmable logic controller was developed. Our tools meet real time constraints. It will implement the various control methods on low computing power targets and low memory capacity.

3- A method of distributed control with optimized communication based on the identification of the relationship between the optimal control and the sequence of past measurements has been validated on a case study. The method achieved the control objectives (meaning acceptable accuracy) with only three sub-models against more than one hundred used in existing tool (MPT toolbox).

The final outlook of the project are to have tools for energy optimization of large refrigerators within the constraints related to operations: control command based on programmable logic controllers, management of degraded operating modes, control subsystem one by one, etc. The methods used will have to adapt to the different types of refrigerators (more pressure levels, different number of Brayton cycle ...)

A paper was published in the ICIT'2015 conference. This paper proposes a distributed control solution applied to the cryogenic refrigerator model of SBT. The main advantage of this distributed control architecture is to allow to have a modular implementation of the control algorithms.
A second paper was published Journal of process control. This paper proposes a solution to estimate the heat load entering the refrigerator. The advantage of this method is to have an estimation of the thermal load without a priori knowledge. This estimate will be used for optimization of advanced control cryogenic refrigerator.

Large cryogenic systems (e.g. cryoplant of the Large Hadron Collider), extract large heat loads at low temperature. This process requires much power at room temperature and cryogenic users and manufacturers have been for a long time aware of the importance of the energy efficiency of these devices. Indeed, these large cryoplants are optimized for a certain design point, and it has been possible in the past years to reach an efficiency of 20% of the Carnot efficiency in the LHC cryoplant. However, such large cryoplants are tricky to control, and subjected to some instabilities as soon as the heat loads change significantly above a certain time scale. Moreover, when heat loads change, the optimum efficiency, reached at the design point, is no longer guaranteed, as the optimum efficiency is the result of a complex compromise between the operations of different components. It would be of major interest for cryogenic users to have at their disposal a tool ensuring that the electrical consumption of the cryoplant will always be minimum. This is the objective of this project. In this project, we propose to develop a totally new control system for large refrigerators, which amazingly still use so far very simple PIDs, in spite of their complexity. Such a control system could also be used in any complex cryogenic system, where heat loads are not constant in time. Our approach is the following: first we will base our control system on a dynamic modeling of a large cryoplant. In the recent years, steady improvements were made in the dynamic modeling of such complex devices, and we will build upon recent results obtained at CERN and at CEA Grenoble. Based on such modeling, we will divide the refrigerator into different subsystems. Then the description of these subsystems will be based on “control oriented” simplified models, which will still describe accurately the subsystems, but will require much less memory and calculation than the general physics driven model of the refrigerator. This will enable an easier implementation of these simplified models in a PLC. Each subsystem will be controlled by its local controller, with various interactions with its neighbors. In this project, we plan to use a “Parametrized Distributed in Time Model Predictive Control” scheme, which seems the most promising local multivariable controller, and the most suited to our constrained environment. This new approach is a generalization of the MPC scheme, which is likely to enable an optimized control within a real time system such as a cryogenic refrigerator. The different subsystems will need to exchange information between each other, and some decisions will be needed, in order to solve possible conflicts. This is a typical case, where a cooperative control architecture would bring a major improvement. In this domain, GIPSA-lab has a prominent experience, and will therefore bring its skill in the definition of such an innovative architecture. Within this project, we therefore plan to develop a totally innovative control scheme of a cryogenic refrigerator. All the different steps will need to be carefully experimentally validated, and this will be done in the medium-sized cryogenic refrigerator available at CEA-SBT. This refrigerator is totally dedicated to R&D, and will be made available for verifying the models, testing local controllers, and evaluating the benefits brought by the innovative architecture developed within this project. A medium sized company will be in charge of the software translation to PLC, enabling an efficient and rapid development of the controller. Final tests, either on a simulator of a CERN refrigerator, or, if it is available, on a large cryogenic refrigerator itself based at CERN, will be performed to finally validate the whole development. This project needs an exceptional variety of prominent expertise(cryogenics,thermodynamic cycles, control systems and PLCs), which are gathered in the CRYOGREEN project submitted here.