Reliability model for frequency converter in electrified railway Yasser A. Mahmood a , A.H.S. Garmabaki a,⇑ , Alireza. Ahmadi a , Ajit Kumar Verma b a Division of Operation and Maintenance Engineering, Luleå University of Technology, Luleå, Sweden b Western Norway University of Applied Sciences, Haugesund, Norway article info Article history: Received 8 January 2017 Received in revised form 20 June 2017 Accepted 2 August 2017 Keywords: Outage-data reporting system Unit state reliability model IEEE STD 762 Frequency converter abstract Reliability analysis of frequency converters based on failures and outages reports constitute an important basis for asset performance and management. Two- and four-state reliability models that recognize the operating characteristics of base load units and peaking units are presented and compared in this study. In this study, a four-state model is modified to a three-state model by combining the ‘needed’ and ‘not- needed’ forced-out states. Moreover, the transitions in the three-state model for power frequency con- verter have been designed according to real operational data. An outage-reporting database modelled considering IEEE STD 762 is presented and compared with the existing failure-reporting database of the case considered here. Furthermore, a method to extract information missing in the failure- reporting database by electrical readings is proposed to meet the requirements of the outage-reporting database. The study found that the results of indexes based on the IEEE four-state model are not reason- able for the frequency converter given their differences with the gas-turbine results under operational conditions. The forced outage rates and availability factors of twelve actual traction frequency converters of Swedish railways network are presented to validate the modified model. Ó 2017 Published by Elsevier Ltd. 1. Introduction The behaviour of an electric power system is stochastic in nat- ure, and therefore, it is logical to assess such systems using proba- bilistic techniques [9]. Probabilistic power-system reliability evaluation is becoming increasingly important in new electric util- ity environments. For ensuring the reliability of power systems, several probabilistic technique measures have been proposed to describe the performance of a generating unit [15]. These perfor- mance measures are valuable for identifying weak areas that need reinforcement, monitoring responses to system design changes and establishing indices that serve as guides for acceptable values in future reliability assessments [9,25]. The [15] standardizes terminology and indices to measure reli- ability, availability and productivity for electric power units. Forced outage rate (FOR) is the one of most common probabilistic indices used for generating units. FOR is a measure of the probabil- ity that a generating unit will not be available due to forced outages and is considered as a conventional unavailability index according to the two-state model. The two-state model is suitable for base load units, but it is unsuitable for intermittent operating unit representation because it delivers unreasonably high unavail- ability index for peaking units [11]. Peaking units normally operate for relatively short periods contrary to base load units that work all the time. Hence to recognize this behaviour, an IEEE task group extended the basic two-state model to a four-state representation for peaking units [16]. The challenge now is to determine the extent to which the four- state reliability model is compatible with operation conditions and transition among states for different peaking units. [10] observed that this model is not reasonable for gas turbines of the Canadian Electricity Association (CEA) in light of the observation of a large number of transitions from the reserve shutdown state to the forced-out but not-needed state. Therefore, they modified the IEEE four-state model by changing the transition states. [5] proposed a model to evaluate the reliability of unified power flow controllers (UPFC). They used three-state and multi-state Markov models to calculate the probability, frequency and duration of states of a UPFC in a composite system reliability evaluation. [19] proposed a three-state reliability model based on the IEEE four-state model for peaking units to consider not only the probability of failure dur- ing operation but also the probability of failure due to synchroniza- tion in a timely manner. [21] proposed a new version of the four- state model according to the load conditions and load uncertainty for evaluating operational reliability. The models proposed in literature are based on the differences with the IEEE four-state model in operating conditions. For instance, starting failure can be consider as such differences. The http://dx.doi.org/10.1016/j.ijepes.2017.08.002 0142-0615/Ó 2017 Published by Elsevier Ltd. ⇑ Corresponding author. E-mail address: amir.garmabaki@ltu.se (A.H.S. Garmabaki). Electrical Power and Energy Systems 94 (2018) 385–392 Contents lists available at ScienceDirect Electrical Power and Energy Systems journal homepage: www.elsevier.com/locate/ijepes