The inexorable inclusion of high temperature fuel cells in the energy grid of the world’s major economies accentuates the needs for more robust systems capable of tolerating non-stationary conditions with minor degradation. It seems undeniable that the latter needs to be addressed mainly by basic and applied research on the fuel cell stack components (electrodes, interconnects, etc…), yet there is still space for performance improvement by optimising system integration. This paper encompasses the experimental procedure used to fully characterise a 6-cell intermediate temperature solid oxide fuel cell (IT-SOFC) short stack when operating in non-stationary conditions and in dissimilar load demands, fuel flows, fuel utilisations and temperatures, simulating to some extent a real-life appliance capable of following electrical load and/or heat demand. The spanning of such input factors allows generating a space-matrix of variables – voltages, temperatures and output gas compositions – that describe the system’s dynamic and stationary performance considering null degradation. This experimental approach can produce appropriate values so as to generate simple autoregressive models suitable for real-time one-step-ahead predictions of the output variables. In this work, the most effective mathematical correlation between the numerous inputs and outputs resulted to be a non-linear autoregressive exogenous (ARX) model built by means of a treepartition algorithm (i.e. binary search trees). The normalised root mean square error (NRMSE) of the calculated fit of the train data was above 94% in all of the cases denoting that the model simulates exceptionally well the system’s outputs in the studied ranges. Additionally, the model was validated with test data, more particularly a dynamic current-voltage curve, and the results show how the simulated points follow remarkably well the experimental results. It must be noted that the electrical current imposed to the SOFC stack in the test data ranged from 5A to 20A, which is the range in which the model performs best; nevertheless outside these bounds the model is capable of simulating exceptionally well the voltage output, especially at high current densities. This dynamic model can be of great use for preliminary design & control of power plants incorporating nominal size stacks from the same manufacturer/integrator by increasing the understanding of how the plant will perform when in transient conditions. Furthermore, the present model is capable of assessing online the degradation rate of a fuel cell stack by comparing the experimental results with the expected ones from the model.
On the Dynamics of Solid Oxide Fuel Cell Stacks: Preliminary Model-Driven Monitoring / BOIGUES MUNOZ, Carlos; Mcphail, Stephen J.; Comodi, Gabriele. - In: ECS TRANSACTIONS. - ISSN 1938-6737. - 75:(2017), pp. 1-7. (Intervento presentato al convegno PRiME 2016 tenutosi a Honolulu, USA nel Oct. 2 - 7, 2016) [10.1149/07537.0001ecst].
On the Dynamics of Solid Oxide Fuel Cell Stacks: Preliminary Model-Driven Monitoring
BOIGUES MUNOZ, CARLOS;COMODI, Gabriele
2017-01-01
Abstract
The inexorable inclusion of high temperature fuel cells in the energy grid of the world’s major economies accentuates the needs for more robust systems capable of tolerating non-stationary conditions with minor degradation. It seems undeniable that the latter needs to be addressed mainly by basic and applied research on the fuel cell stack components (electrodes, interconnects, etc…), yet there is still space for performance improvement by optimising system integration. This paper encompasses the experimental procedure used to fully characterise a 6-cell intermediate temperature solid oxide fuel cell (IT-SOFC) short stack when operating in non-stationary conditions and in dissimilar load demands, fuel flows, fuel utilisations and temperatures, simulating to some extent a real-life appliance capable of following electrical load and/or heat demand. The spanning of such input factors allows generating a space-matrix of variables – voltages, temperatures and output gas compositions – that describe the system’s dynamic and stationary performance considering null degradation. This experimental approach can produce appropriate values so as to generate simple autoregressive models suitable for real-time one-step-ahead predictions of the output variables. In this work, the most effective mathematical correlation between the numerous inputs and outputs resulted to be a non-linear autoregressive exogenous (ARX) model built by means of a treepartition algorithm (i.e. binary search trees). The normalised root mean square error (NRMSE) of the calculated fit of the train data was above 94% in all of the cases denoting that the model simulates exceptionally well the system’s outputs in the studied ranges. Additionally, the model was validated with test data, more particularly a dynamic current-voltage curve, and the results show how the simulated points follow remarkably well the experimental results. It must be noted that the electrical current imposed to the SOFC stack in the test data ranged from 5A to 20A, which is the range in which the model performs best; nevertheless outside these bounds the model is capable of simulating exceptionally well the voltage output, especially at high current densities. This dynamic model can be of great use for preliminary design & control of power plants incorporating nominal size stacks from the same manufacturer/integrator by increasing the understanding of how the plant will perform when in transient conditions. Furthermore, the present model is capable of assessing online the degradation rate of a fuel cell stack by comparing the experimental results with the expected ones from the model.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
