A heuristic method of variable selection based on principal component analysis and factor analysis for monitoring in a 300 kW MCFC power plant

https://doi.org/10.1016/j.ijhydene.2012.04.135Get rights and content

Abstract

In a commercialized 300 kW molten carbonate fuel cell (MCFC) power plant, a univariate alarm system that has only upper and lower limits is usually employed to identify abnormal conditions in the system. Even though univariate alarms have already been adopted for system monitoring, this simple monitoring system is limited for using in an extended monitoring system for fault diagnosis. Therefore, based on principal component analysis (PCA), a recursive variable grouping method for a multivariate monitoring system in a commercialized MCFC power plant is presented in this paper. In terms of development, since a principal component analysis model that contains all system variables cannot isolate a system fault, heuristic recursive variable selection method using factor analysis is presented here. To verify the performance of the fault detection, real plant operations data are used. Furthermore, comparison between type 1 and type 2 errors for four different variable groups demonstrates that the developed heuristic method works well when system faults occur. These monitoring techniques can reduce the number of false alarms occurring on site at MCFC power plant.

Highlights

► Multivariate monitoring system for a MCFC power plant is demonstrated. ► Comparative study of various variable groups within several system faults. ► Conventional principal component analysis (PCA) is not working to fault detection. ► Monitoring performance is increased by heuristic variable selection method.

Introduction

A molten carbonate fuel cell (MCFC) power plant is an eco-friendly electricity generation system. Among several kinds of fuel cells for power plants using polymer electrolyte membrane fuel cells, such as solid oxide fuel cells and so on, MCFCs stand out as the most available technology for commercial release with respect to the cost and their capacity for electric energy generation. The advantages of an MCFC power plant are the absence of industrial noise and discernible cleanness, meaning that installation in downtown urban areas is possible. The main source of noise is mostly from the air blower. There are no air pollutants such as NOx and SOx; therefore, the MCFC system is the most promising technology that is close to commercialization [1], [2], [3], [4].

In order to optimize design parameters, MCFC stack modeling was carried out [5], [6]. These studies focused on the state estimation of the stack for understanding phenomena, i.e. effects of input parameters, temperature profiles on the stack. Research on analysis of MCFC systems also was demonstrated. Control strategies for a pilot scale power system were developed. Some studies of the pilot scale power system built mathematical models through experimentation, but the other studies developed similar mathematical models without employing experimentation results [7], [8], [9], [10], [11]. The performance of a combined MCFC system was investigated for a comprehensive thermodynamic analysis [12].

The power generation system in an MCFC plant was developed by POSCO Power, Inc. The specification of the commercialized MCFC power plant is summarized in Table 1. The plant's electric generating capacity is 330 kW maximum and decreases 5 kW every 6 months due to the stack degradation. The MCFC power plant primarily consists of three major units: a fuel processor, a stack, and electric power conditioning shown in Fig. 1. The fuel processor is composed of a fuel pre-convertor, which reforms natural gas to hydrogen with a steam reforming reaction, and a heat recovery system to increase the overall system efficiency. In the fuel processor, liquefied natural gas (LNG) with the primary treatment is mixed with steam through the humidifier, which removes the impurities in the desulfurizer and the particle filter. The LNG is then fed to the pre-convertor. The stack that generates the direct current is made up of hundreds of piled-up large cells, which are separated by bipolar plates. Details on the general stack and system are reported in the literature [13], [14] (Table 2).

The MCFC power plant is composed of the stack and the balance of plant (BOP), which is commonly divided into the electrical BOP and the mechanical BOP. Each component has dozens of sensors that typically estimate the temperature, flow rate, and pressure. Due to an overabundance of sensors, false alarms occur quite frequently and simultaneously. Both batch and continuous systems have many single variables. Simple monitoring system only detects abrupt faults. Complicated faults involving interlock control system and drift of multiple variables are not detected [15], [16]. For this reason, an advanced monitoring system for a plant size system should be considered. In this paper, we present a multivariate statistical monitoring system, considering the performance of fault detection with various variable groups, based on the principal component analysis technique.

Section snippets

Theory

The most well-known traditional monitoring technique for a chemical process is the statistical process control (SPC) method. The main purpose of this method is to monitor the performance of the process fundamentally observing whether the state of the process is in control. The state of control is defined as having certain variables remain close to their desired values and the only source of variation being a common cause, i.e. process disturbance and set points change. This traditional method

Monitoring system

The PCA technique described in Section 2 provides the basis of design a multivariate monitoring system. In order to conduct off-line test for real plant data, data preprocessing and validation are necessary. In the simulation studies, process variables are not frequently changed [22], [23]; otherwise, real operations data may have noise and abrupt change that is due to sensing errors [24]. The general procedure for building up a monitoring system is shown in Fig. 2. In the following, the

Results and discussion

First, the off-line monitoring results are presented for the fault detection performance, from the plant operations data, along with the trip history. Subsequently, the comparison result with other variable groups were reported in order to validate the improvement in the monitoring performance in terms of type I and type II error rates [26]:TypeIerrorrate=FalsepositivecasesAllnormalcasesTypeIIerrorrate=FalsenegativecasesAlltripcaseswhere false positive refers to the cases in which a normal

Conclusions

In this paper, off-line analysis for the performance estimation of a multivariate monitoring system in a MCFC power plant is carried out. In terms of development, the basic PCA model cannot work for all the variables since all 163 variables cannot represent the system status well. In order to implement an advanced monitoring technique for the commercialized process, a heuristic recursive variable selection method for PCA modeling is developed. The comparison results of the 4 variable groups are

Acknowledgment

This research was supported by the second phase of the Brain Korea 21 Program in 2012, Institute of Chemical Processes in Seoul National University, Strategic Technology Development and Energy Efficiency & Resources Development of the Korea Institute of Energy Technology Evaluation and Planning (KETEP) grant funded by the Ministry of Knowledge Economy (MKE) and grant from the LNG Plant R&D Center funded by the Ministry of Land, Transportation and Maritime Affairs (MLTM) of the Korean government.

References (27)

Cited by (28)

  • The progress of fuel cell for malaysian residential consumption: Energy status and prospects to introduction as a renewable power generation system

    2021, Renewable and Sustainable Energy Reviews
    Citation Excerpt :

    The cost maintaining of the balance of power (BoP) operations is very high to preserve carbonate electrode chemistry. Hence, The corrosion state "of corrosive carbonate electrolyte" consists of hazard characteristics [126,127]. For the stationary application, phosphoric acid fuel cells (PAFCs) have been commercialized in the mature stage.

  • A heuristic method of road traffic compensation based on delay function gradient

    2019, Measurement: Journal of the International Measurement Confederation
    Citation Excerpt :

    Traffic jams will cause huge economic and social damage, with congestion charges in the U.S. urban area at $101 billion in 2016. According to a recent study, the cost of congestion in the Australian capital city was estimated at about $1.75 billion in 2015, compared with an annual congestion cost of 1.28 billion, a significant increase [1–6]. Therefore, it is very important to study the problem of traffic congestion.

  • Variable selection methods in multivariate statistical process control: A systematic literature review

    2018, Computers and Industrial Engineering
    Citation Excerpt :

    An automotive manufacturing dataset was the case studied by González and Sánchez (2010) and a real cigarette production was analyzed by Zhao and Gao (2017). Giannetti et al. (2014) and Jeong et al. (2012) developed their methods to deal particularly with complex, and very specific, manufacturing settings, such as foundry environment and molten carbonate fuel cell power plant, respectively. Finally, Wang and Tsung (2009) tested their method on simulated data.

View all citing articles on Scopus
View full text