Robust control of a solid oxide fuel cell for combined heat and power applications

Abstract

The reduction of energy consumption and CO2 emissions is one of the main environmental issues in the present-day society. Distributed combined heat and power systems (CHP-systems) are one of the most promising technologies for the achievement of this goal. The reason is that the power is produced near to the site of consumption, minimizing the heat loss during transport. Additionally, the waste heat generated in the process (which in the case of centralized electricity production becomes generally lost) can be recovered for heating purposes or warm water supply. As a consequence, the total efficiency of the energy conversion process is increased, resulting in a reduction of the CO2 emissions and energy costs. Solid oxide fuel cells (SOFC) are the CHP technology with the highest electrical efficiency, achieving values of up to 60 %. Additionally, they operate directly with natural gas, enabling their installation in most industrial, commercial and residential buildings without the necessity of additional infrastructure. The natural gas consumed in Germany is a mixture of gases extracted from different sources, resulting in a gas composition which varies over time and site of consumption. The consequence is a fluctuation of the heating value of up to 30 %, which can not be easily measured nor predicted. The mentioned variation is a risk for SOFC systems, since their degradation and efficiency are very sensitive to changes in the gas composition. However, the current control strategies for SOFC systems are still calibrated to one type of natural gas and do not regard the mentioned fluctuations. For instance, the system studied in this work has been calibrated with methane, leading to optimal operating conditions for this type of gas. However, it has been estimated that if it was operated with other gases, the evaluation criteria regarding the degradation (fuel utilization and oxygen to carbon ratio) would have achieved deviations up to 25 % from the set point. The aim of this work is to solve this problem by developing alternative control strategies which take changes in the natural gas composition into account. The mean amount of carbon atoms per molecule of fuel is identified as the key property for this issue, since it can be related analytically to the degradation criteria of fuel utilization and oxygen to carbon ratio. Thus, if this property of the fuel is identified, the control parameters can be determined and the system can work under optimal conditions, enabling the desired lifetime and efficiency of the system. For this purpose, different detection strategies of the natural gas carbon content have been developed, by relating the non-measurable mean amount of carbon atoms per molecule to other measurable gas properties. The most relevant strategies are based on the measurement of the inlet natural gas calorific value (calorific value strategy), on the oxygen mass fraction of the exhaust gas (oxygen strategy) or on the humidity and carbon dioxide mass fraction of the exhaust gas (humidity strategy). The mentioned strategies have been evaluated by using a complete system model developed in Modelica language, which includes a stack, a fuel processor and an air processor. The models have been parameterized and validated with experimental data. The evaluation results show that the usage of the humidity strategy does not improve the results compared to the reference strategy, which is the control strategy of the system without any adaptation to the gas composition. However, by using the oxygen strategy the deviation of the evaluation criteria from their set point is reduced to a maximum of 5 % and by applying the calorific value strategy it decreases to a maximum of 5.5 %. The detection of the mean amount of carbon atoms per molecule is experimentally tested for the humidity strategy and for the oxygen strategy. In the first case the carbon content in the gas can be identified with a mean error of 2.5 % and a maximal error of 15 %, while in the second case its detection is not possible. Thereby, the behavior observed at the simulation results is confirmed by the experimental tests. The next step is to test the successful strategies in real systems connected to the natural gas network, in order to evaluate their effect on efficiency and long-term robustness of the system.