Experimental evaluation and ANN modeling of a recuperative micro gas turbine burning mixtures of natural gas and biogas
Experimental evaluation and ANN modeling of a recuperative micro gas turbine burning mixtures of natural gas and biogas
复制标题
燃烧天然气和沼气混合物的回热式微型燃气轮机的实验评估和人工神经网络建模
DOI：
10.1016/j.apenergy.2013.11.074
复制
发表时间：
2014
期刊：
影响因子：
11.2
通讯作者：
P. Mørkved
中科院分区：
文献类型：
--
作者：
H. Nikpey;M. Assadi;P. Breuhaus;P. Mørkved
以前发表的研究已经解决了在使用沼气(即低热值燃料)时对发动机进行修改的问题。这项研究的重点是绘制出微型热电联产(CHP)装置中沼气和天然气混合燃料中可能的沼气份额，而不需要对发动机进行任何修改。这有助于减少现有热电联产装置的二氧化碳排放，并有可能避免昂贵的沼气升级到天然气质量以及发动机改装。此外，这种方法允许在沼气成分最终变化和/或沼气短缺的情况下使用天然气作为后备解决方案，从而提供更好的可用性。在本研究中，对商用100千瓦微型燃气轮机(MGT)在满负荷和部分负荷下的性能和排放进行了实验评估，当供给不同的天然气和沼气混合物时。MGT配备了额外的仪器，并使用气体混合站通过用二氧化碳稀释天然气来从零沼气供应所需的燃料混合物到尽可能高的水平。以甲烷和二氧化碳(摩尔分数)分别为0.6和0.4的典型沼气组分为参照物，计算了供气混合气中相应的沼气含量，介绍了用于实验活动的试验台装置，并报告了结果，展示了燃烧沼气和天然气混合物对MGT性能和排放的影响。结果表明，与天然气燃烧情况相比，电效率几乎没有变化，运行参数没有明显变化。研究还表明，在满负荷运行时，燃烧天然气和沼气的混合物有助于显著减少工厂的二氧化碳排放量约19%。考虑到实验过程中获得的大量数据，建立了基于人工神经网络(ANN)的数据驱动模型来模拟MGT的性能。使用平均相对误差(MRE)来评估所建立的神经网络模型对训练过程中未使用的实验数据的预测精度。结果表明，所建立的神经网络模型能够较高精度地预测MGT的性能参数，大部分样本的误差在1%以内。在Microsoft Visual Basic中为ANN模型创建了图形用户界面(GUI)。图形用户界面是一种用户友好的工具，用于工厂的建模和状态监控。
Previously published studies have addressed modifications to the engines when operating with biogas, i.e. a low heating value fuel. This study focuses on mapping out the possible biogas share in a fuel mixture of biogas and natural gas in micro combined heat and power (CHP) installations without any engine modifications. This contributes to a reduction in CO2emissions from existing CHP installations and makes it possible to avoid a costly upgrade of biogas to the natural gas quality as well as engine modifications. Moreover, this approach allows the use of natural gas as a “fallback” solution in the case of eventual variations of the biogas composition and or shortage of biogas, providing improved availability.In this study, the performance and emissions of a commercial 100 kW micro gas turbine (MGT) at full and part loads are experimentally evaluated when fed by varying mixtures of natural gas and biogas. The MGT is equipped with additional instrumentation, and a gas mixing station is used to supply the demanded fuel mixtures from zero biogas to the maximum possible level by diluting natural gas with CO2. A typical biogas composition with 0.6 CH4and 0.4 CO2(in mole fraction) was used as reference, and corresponding biogas content in the supplied mixtures was computed.This paper presents the test rig setup used for the experimental activities and reports the results, demonstrating the impact of burning a mixture of biogas and natural gas on the performance and emissions of the MGT. The results indicate that the electrical efficiency is almost unchanged and no significant changes were observed in operating parameters, comparing with the natural gas fired case. It was also shown that burning a mixture of natural gas and biogas contributes to a significant reduction in CO2emissions from the plant by about 19% at full load operation. Given the extensive data obtained during the experimental tests, a data-driven model based on an artificial neural network (ANN) was developed to simulate the performance of the MGT. The mean relative error (MRE) was used to evaluate the prediction accuracy of the developed ANN model with respect to experimental data which were not used during the training. It was demonstrated that the ANN model can predict the performance parameters of the MGT with high accuracy and the error of most samples is less than 1%. A graphical user interface (GUI) was created for the ANN model in Microsoft Visual Basic. The GUI is presented as a user-friendly tool for modeling and condition monitoring of the plant.