(672d) A Decision Support Tool and Closed-Loop Startup Framework for a Hydrogen Plant

Hydrogen is one of the most extensively used chemical components, with wide range of applications in petroleum and chemical industries. An economic way to commercially produce hydrogen is to utilize steam methane reforming [1]. The process forms an intricate network of material and energy due presence of vast number of material and heat flows. Startup of a hydrogen plant involves implementation of a predefined recipe that consists of a series of events taking place over a finite duration of time. Specifically, it involves continuous actions such as ramping the natural gas fuel to reach a sufficient reformer exit temperature, regulation of nitrogen flow in the reactor tubes, circulation of natural gas feed and steam etc. Further, discrete decisions such as starting up the PSA, recirculating the tail gas from the PSA and ramping up to full capacity in several stages are also part of the startup recipe. Based on the decisions that are taken at various stages in the startup, the startup time can vary significantly. Startup is typically categorized as cold startup and warm startup, based on the initial state of the plant from which the plant is required to be driven to nominal operation. The cold startup will be central to the discussion in this work, where the plant is started from completely shutdown conditions.

Existing startup recipes have been built heuristically from past successful startups, with the possibility of optimizing the startup procedure remaining unexplored. Further, due to the uncertainty associated with the startup duration, the availability of the product cannot be predicted accurately. Motivated by these considerations, in this work we first developed a decision support tool (DST) that can assist the operator during startup to predict the completion of startup. One key recognition that we make is that due to the nature of startup/shutdown phase, they can be treated as a batch like operation with variable lengths, where the objective is to drive the process from one state to the other. Thus, batch subspace identification approach [2] is used for data-driven modeling of the hydrogen plant [3]. Development of a data-driven LTI model, that can handle variable batch duration, paves the way for reliable prediction of the startup evolution and its utilization in a tractable optimization formulation. The DST in real time predicts the timeline of the startup by utilizing the data-driven model and state-trajectories from historical batches. This paper will present the proposed algorithm and results from plant startup.

Secondly, we addressed the problem of synthesis of an optimal startup procedure for the hydrogen production plant. A key requirement for operating procedure synthesis (OPS) is a good model for the startup process. There exist several modeling approaches that are well suited for small scale processes (see [5] for an excellent review) but difficult to implement to large chemical units. Out of these approaches, dynamic simulation based strategies are most prominent. In these approaches, dynamic simulations are used to select the best startup scenarios. In the present work, startup is optimized such that it improves not only the process economics but also its completion time using dynamic optimization. The optimization problem for the present application is an optimal control problem where the objective is to reach the desired terminal state in minimum time, while adhering to various constraints (such as the bounds on reformer exit temperature [6]) and making discrete decisions along the way. To this end, first a high-fidelity test bed model of the entire plant is developed in UniSim, capable of simulating the startup and shutdown phase, with appropriate adaptation of the plant SOP. Note however that such detailed first principles models pose computational challenges when directly embedded in optimization problems, and thus existing SOP synthesis approaches utilize heuristics to determine optimal profile [7-10]. Thus, in this work, several simulated startups are performed to generate the training data. Then, an LTI data-driven model of the process using batch subspace identification [2-4] is identified. The identified data-driven model is subsequently utilized within an optimization framework to synthesize the startup SOP. Through closed-loop startup, we are able to achieve startup in significantly less time. Results from the closed-loop startup and the corresponding optimization problem will be presented.

[1] Kroschwitz, J., Seidel, A., 2006. Kirk-Othmer Encyclopedia of Chemical Technology, No. V. 20 in Encyclopedia of Chemical Technology. Wiley.

[2] Corbett, B., Mhaskar, P. Subspace identification for data-driven modelling and quality control of batch processes, AIChE Journal 62 (2016) 1581-1601.

[3] Garg, A., Corbett, B., Mhaskar, P., Hu, G., Flores-Cerrillo, J. (2017), High Fidelity Model Development and Subspace Identification of a Hydrogen Plant Startup Dynamics, Computers and Chemical Engineering, 106 183-190.

[4] Garg, A., Mhaskar, P. (2017), Subspace Identification Based Modeling and Control of Batch Particulate Processes, Industrial and Engineering Chemistry Research, 56 (26) 7491-7502.

[5] Batres, R., Soutter, J., Asprey, S.P., Chung, P., 2002. Operating procedure synthesis: science or art. Knowl. Eng. Rev. 17 (3).

[6] Du, M., Mhaskar, P., Zhu, Y., Flores-Cerrillo, J., 2014. Safe-parking of a hydrogen production unit. Ind. Eng. Chem. Res. 53 (19), 8147–8154.

[7] Scenna, N., Ruiz, C., Benz, S., 1998. Dynamic simulation of start-up procedures of reactive distillation columns. Comput. Chem. Eng. 22, S719–S722.

[8] Reepmeyer, F., Repke, J.-U., Wozny, G., 2004. Time optimal start-up strategies for reactive distillation columns. Chem. Eng. Sci. 59 (20), 4339–4347.

[9] Yang, X., Xu, Q., Li, K., Sagar, C.D., 2010. Dynamic simulation and optimization for the start-up operation of an ethylene oxide plant. Ind. Eng. Chem. Res. 49 (9), 4360–4371.

[10] Zhao, Y., Zhang, J., Qiu, T., Zhao, J., Xu, Q., 2014. Flare minimization during start-ups of an integrated cryogenic separation system via dynamic simulation. Ind. Eng. Chem. Res. 53 (4), 1553–1562.