Abstract Steam reforming is the most widespread process for the generation of hydrogen-rich synthesis gas from light carbohydrates. One of the operation and design challenges is that the feed materials are various from natural gas, shale gas, to liquid gas, or naphtha. The optimal operating conditions identified under one type of feed may not be functional well under another one. Even within the same type feed, the compositions may vary large enough in continuous operation that requires online operating conditions tuning. Currently, two types of control strategies are applied to address the issue: one is feedforward control that the feed composition is measured in advance so that the operating conditions could be set accordingly. However, the performance would rely on the accuracy of the process model, and may not be robust on feed composition disturbances. The other is simple PID feedback control that the conditions are adjusted by the product composition feedback. However, the complexity of the multi-input multi-output (MIMO) steam reforming may bring challenges to the PID controller design. In this paper, a Bayesian-based model-set management method is introduced for constructing a statistically superior model set for online model-based control application. First, a number of steam reforming models are developed on various feed types or compositions. Then the optimal operating conditions are identified on the weighted model predictions. During online application, the measured product compositions are feedback to adjust the weights by a Bayesian-based statistical method. A repeated use of the method keeps the weights updated constantly based on the newly available system data, which makes the model-set-based prediction more precise, robust, and adaptive to various feeds automatically. The efficiency of the method is demonstrated by studying a steam reforming system with different feeds.
[1]
R. Hughes,et al.
The kinetics of methane steam reforming over a Ni/α-Al2O catalyst
,
2001
.
[2]
G. Froment,et al.
Methane steam reforming: II. Diffusional limitations and reactor simulation
,
1989
.
[3]
Jia Li,et al.
Bayesian-based on-line applicability evaluation of neural network models in modeling automotive paint spray operations
,
2006,
Comput. Chem. Eng..
[4]
G. Froment,et al.
Methane steam reforming, methanation and water‐gas shift: I. Intrinsic kinetics
,
1989
.
[5]
Valerie Dupont,et al.
Kinetics study and modelling of steam methane reforming process over a NiO/Al2O3 catalyst in an adiabatic packed bed reactor
,
2017
.