ECE 802, secion 605 The effec of he fire deecors on gas urbine reliabiliy By: Mohammed Ben-Idris upervisor: Professor Mira December, 12, 2009
Ben-Idris 1 Absrac. Reliable fire, smoke, and gas leakage deecors are an essenial par of fire proecion in gas urbines. Also, mainaining he reliabiliy of gas urbines is imporan in avoiding cascading failures especially when gas urbines are conneced o he grid in he case of emergency or a peak loads. In his projec, he effec of malfuncioning in he fire proecion sysem has been inroduced. Also, a reliabiliy logic diagram for he deecor has been proposed. Two-ou-of-hree deecor configuraion is used in his projec as well as using one deecor and comparing he resuls. To improve faul deecion reliabiliy, wo-ou-of-hree deecor configuraion is recommended. This projec is divided ino hree main pars: firs, reliabiliy analysis of he hree sae deecors (normal operaion, open circui, and shor circui.) econd, reliabiliy analysis of gas urbine conneced o generaor which feeds load and sands alone. Third par is he reliabiliy analysis of he overall sysem which consiss from gas urbine, generaor, load, and fire proecion sysem. The problem Reliabiliy modeling of wo or more sysems working ogeher and have compleely differen asks or funcions is no sraigh forward. Reliabiliy of gas urbine is a measure of providing he required power o he grid when i is needed and he reliabiliy of he proecion sysem is a measure of is abiliy o deec fauls and no sending false alarms in he case of here is no faul. These wo sysems canno be conneced ogeher in one reliabiliy block diagram unless connecing failure dependen branches beween he gas urbine and he proecion sysem which makes he analysis much harder. In his projec, Mone Carlo imulaion is used o solve he problem of dependencies. The sysem is divided ino wo sysems and simulaed for all possible condiions. ysem configuraion This sysem consiss from wo sysems: (1) energy conversion sysem which composed of gas urbine wih is faciliies, fuel saions (gas and ligh fuel), generaor, and load; (2) Fire proecion sysem.
Ben-Idris 2 For energy conversion sysem, o make he analysis simple and eliminaing he grid effecs, he sysem is assumed o be sand alone sysem. The sysem configuraion is shown in figure 1. Fuel aion Turbine Generaor Load Figure 1 The proecion sysem of he urbine consiss from fire, smoke, and gas leakage deecors. They are divided ino hree groups according o heir funcion: - Gas leakage, fire, and smoke faul a he urbine leads o ripping he urbine. - Fire and smoke faul a he ligh saion leads o change o Gas operaion mode. - Gas leakage, fire, and smoke faul a he gas saion leads o change o ligh fuel mode. The fire proecion sysem configuraion is shown in figure 2. Ligh Fuel aion Fire deecor Gas Turbine Combusion chamber moke Deecor moke Deecor Change o Gas Change o ligh fuel Derae Trip Fire Deecor Gas Leakage Deecor Fire deecor moke Deecor High emperaure sensor Gas Leakage Deecor Gas aion Figure 2 Deecors In many combinaorial reliabiliy models he behavior of a sysem and is componens is assumed o be binary (funcioning and failed) [1]. However, here are hree modes of operaion for deecors: (1) normal operaion in which he deecor deecs he faul and sends a signal o rip he sysem if here is a faul and no making any acion when here is no faul, (2) failure o operae or missing he faul in which he deecor fails o see he faul, and (3) unwaned ripping
Ben-Idris 3 or false alarm in which he deecor sends a rip signal in he case of no faul. From he reliabiliy poin of view, he second and hird modes are failure modes. The reliabiliy of wo-sae componens can be increased by connecing he componens in parallel as many as required o achieve he desired reliabiliy index. In he muli-sae devices, however, ha is no necessary rue. As we will see in he nex analysis, he probabiliy of false alarms may increase wih increasing he number of componens conneced in parallel. Deecors reliabiliy modeling There are hree saes for fire, smoke, and gas deecors which are rue alarm (TA), false alarm (FA), and missed alarm (MA) [1], wih probabiliies p TA, q FA, q MA respecively. These probabiliies sum up o one p TA + q FA + q MA = 1 (1) We can consider he sae ha he deecor sends a rue alarm as a normal operaing sae wih probabiliy P n, he sae ha he deecor sends a false alarm as shor circui sae wih probabiliy q s, and finally he sae ha he deecor misses he faul as open circui sae wih probabiliy q o. Therefore, equaion (1) becomes p n + q s + q o = 1 (2) Figure 3 shows he proposed deecor s logic configuraion Faul P n Trip q O No signal q Figure 3 In order o increase he reliabiliy of deecing he faul, one way is o connec wo deecors in parallel. The sysem probabiliies in his case will be p n + q s + q 2 o = p 2 n + q 2 s + q 2 o + 2p n q s + 2p n q o + 2q s q o (3) The probabiliy of deecing he faul is p 2 n + 2p n q o The probabiliy of sending false alarm is q 2 s + 2p n q s + 2q s q o 2 The probabiliy of missing he faul is q o
Ben-Idris 4 From hese probabiliies, i is obvious ha he probabiliy of missing he faul decreases when wo deecors are conneced in parallel. However, he probabiliy of sending a false alarm and he probabiliy of working normally depend on he dominan one. The probabiliy of geing a false alarm can be reduced by making a condiion ha he signal should no rip he sysem unless he wo deecors send a rip signal which is equivalen o series sysem or 2-ou-of-2 sysem. In he case of 2-ou-of-2, he sysem probabiliies are p n + q s + q 2 o = p 2 n + q 2 s + q 2 o + 2p n q s + 2p n q o + 2q s q o (3) The probabiliy of deecing he faul is p 2 n + 2p n q s 2 The probabiliy of sending false alarm is q s The probabiliy of missing he faul is q 2 o + 2p n q o + 2q s q o From hese probabiliies we can see ha even hough he probabiliy of geing a false alarm decreased he probabiliy of missing he faul increased. By connecing hree deecors in parallel, any deecor sends a signal leads o rip he sysem. The all sae probabiliies are as follows p n + q s + q 3 o = p 3 n + q 3 s + q 3 o + 3p n q 2 s + 3p n q 2 o + 3p 2 n q s + 3p 2 n q o + 6p n q s q o + 3q 2 2 s q o + 3q s q o The probabiliy of deecing he faul is p 3 n + 3p n q 2 o + 3p 2 n q o The probabiliy of sending false alarm is q 3 s + 3p n q 2 s + 3p 2 n q s + 6p n q s q o + 3q 2 2 s q o + 3q s q o 3 The probabiliy of missing he faul is q o Despie he fac ha he probabiliy of missing he faul decreases by cubic roo, he probabiliy of geing a false alarm increases as a funcion of oher probabiliies. Anoher way o solve his problem is by making a condiion ha he rip will occur unless 2-ou-of-3 of he deecors send a rip signal. The 2-ou-of-3 probabiliies become: The probabiliy of deecing he faul is p 3 n + 3p 2 n q s + 3p 2 n q o + 6p n q s q o The probabiliy of sending false alarm is The probabiliy of missing he faul is q s 3 + 3p n q s 2 + 3q s 2 q o
Ben-Idris 5 q 3 o + 3p n q 2 2 o + 3q s q o From hese probabiliies we can see ha he probabiliy of deecing he faul has grealy improved and he probabiliies of geing a false alarm and missing he faul has significanly decreased. Despie he fac ha reducing he probabiliy of geing a false alarm is imporan, increasing he probabiliy of deecing he faul is more imporan in order o proec he sysem from damage as well as proecing he operaors. To compromise beween he cos and reliabiliy, 2-ou-3 fulfils he requiremen. Figure 4 shows he relaionship beween he numbers of deecors wih each configuraion (a) (b) (c) Figure 4. (a) True Alarm, (b) False Alarm, (c) Missed Alarm The energy conversion sysem reliabiliy modeling I is hard o model every par of he gas urbine and ha needs analysis of a huge number of saes. In his projec, however, he gas urbine iself is considered as one block wih a fixed failure and repair raes. In he oher hand, he auxiliaries are caegorized according o heir funcions.
Ben-Idris 6 The gas urbine Gas urbines simply consis from hree main pars: compressor, combusion chamber, and he urbine as shown in figure 5. I has wo modes of operaion, naural gas mode and ligh fuel mode. The reliabiliy of he urbine and is operaing hours are funcion of he fuel ype. For simpliciy, in his projec, all valves, gaes, cables, wirings, and swiches are assumed o be perfec. For each mode of operaion, he urbine is represened wih a differen block. Also, he ransiion from one mode o anoher is considered o be insananeous and perfec. Figure 6 shows he urbine reliabiliy block diagram, where, TG means gas operaion mode and TF means ligh fuel operaion mode. Combusion Chamber G Compressor Turbine Figure 5 TG (a) TF (b) The auxiliaries Figure 6 There are many auxiliaries ha are necessary for gas urbine. These auxiliaries are o be reliable o mainain he gas urbine reliabiliy. Table 1 shows he main auxiliaries ha considered in his projec and heir funcions. Table 1 Componen Funcion Number of he componens Main fuel pump Injecing ligh fuel o he urbine 1 Waer cooling pump Cooling he bearings and ceiling air 1-ou-of-2 Oil pump Circulaing oil o he bearings and he moveable pars 1-ou-of-2 Auxiliary ransformer Providing necessary power from he generaor oupu 1 UP Providing necessary power in case of no power 1 ince all hese auxiliaries are imporan for he urbine o operae, for reliabiliy block diagram, hey are conneced in series wih each oher wih he urbine. However, main fuel pump
Ben-Idris 7 is required only when he urbine operaes in a ligh fuel mode. The UP are in a sandby mode unless he auxiliary ransformer fails or here is no power oupu from he generaor. Also, in his projec, he swiching beween he auxiliary ransformer and UP is assumed o be perfec. Figure 7 shows he auxiliaries reliabiliy block diagram. AT WC OP UP WC OP MF Figure 7. AT: Auxiliary Transformer, WC: Waer Cooling Pump, OP: Oil Pump, and MF: Main Ligh Fuel Pump. The Generaor The generaor is considered o be one block. Also, he generaor is assumed o feed local load and is no conneced o he grid. Proecion sysem reliabiliy of he generaor and he generaor s auxiliaries are assumed o be included in he same block. The generaor reliabiliy block diagram is shown below. G The load The load is assumed o have wo saes, zero MW and raed load. L The gas saion and he combusion chamber Gas saion and combusion chamber reliabiliy block diagram is shown in figure 10. The gas supply is considered o be unlimied. The only signals ha considered in his projec from he gas saion and he combusion chamber are fire, smoke, and gas leakage fauls. ince he gas saion and he combusion chamber deecors are he same, hey are shown in figure 8 wihou any differences where GL denoes gas leakage deecor, F denoes fire deecor, and M denoes smoke deecor. Ligh fuel saion The ligh fuel saion reliabiliy block diagram is shown in figure 10. The ligh fuel supply is considered o be unlimied. The only signals ha considered in his projec from he
Ben-Idris 8 ligh fuel saion are fire and smoke fauls. Figure 9 shows he ligh fuel saion deecor reliabiliy block. GL 2-ou-of-3 F 2-ou-of-3 M 2-ou-of-3 GL F M GL F M Figure 8 F 2-ou-of-3 M 2-ou-of-3 F M F M Figure 9 Overall energy conversion sysem reliabiliy represenaion I is obvious ha his sysem consiss from wo disinguishable and failure dependen subsysems which are he energy conversion sysem and he proecion sysem. The energy conversion sysem reliabiliy block diagram is shown in figure 10. Perfec swich Auxiliaries MF F TF AT WC OP G L G TG UP WC OP Figure 10. F: ligh fuel saion, G: gas saion Mehodology The energy conversion sysem and he proecion sysem are dependen sysems. In his projec, he effec of he proecion sysem on he energy conversion sysem is considered. There is no
Ben-Idris 9 such way ha can connec he reliabiliy block diagram of hese wo sysems ogeher unless using some branches ha represen hese dependencies. Mone Carlo imulaion is used o evaluae he overall sysem reliabiliy indices because i is easy o generae random variable for each componen and hen use he probabiliies o solve he problem of he differences and he dependency. Following is he proocol of he sysem operaion which is followed in using Mone Carlo imulaion: 1. The urbine normally works in gas mode. 2. The energy conversion sysem works properly if here is no faul. 3. The conrol sysem rips he urbine if: a. There is a failure in he generaor. b. The wo waer cooling pumps fail. c. The wo oil circulaing pumps fail. d. The auxiliary ransformer and he UP boh fail. 4. If a leas wo of he hree fire deecors, wo of he hree smoke deecors, or wo of he hree gas leakage deecors a he gas saion deec a faul, he conrol sysem changes he urbine o ligh fuel mode. 5. If he urbine is working in a ligh fuel mode (afer changing from gas mode) and if he faul a gas saion is cleared, he conrol sysem changes he urbine back o gas mode. 6. If a leas wo of he hree fire deecors or wo of he hree smoke deecors a he ligh fuel saion deec a faul and if he faul a he gas saion sill un-cleared, he sysem rips he urbine. 7. If a leas wo of he hree fire deecors, wo of he hree smoke deecors, or wo of he hree gas leakage deecors a he combusion chamber deec a faul, he sysem rips he urbine. Componen s daa ome daa are brough from IEEE sandards which are given in some sources, some from oher sources, and some are assumed. Table 2 shows he reliabiliy daa which are aken from sources for some componens. In his projec, moors are assumed idenical. Table 3 shows he assumed deecors probabiliy of failures. The deecors are assumed idenical, un-repairable, and hey are available a he power
Ben-Idris 10 plan sorage. Table 4 shows he assumed probabiliy daa for he gas and ligh fuel saions and he assumed probabiliies of occurring fire, smoke, and gas leakage. Table 2 Componen λ / year μ / year Availabiliy ource Aux. Transformer 0.1 28 0.996 [2] Generaor 4.5625 1095 0.995 [2] Moor 0.1087 51.5 0.9979 [2] UP 0.16667 730 0.99977 [2] Turbine / gas mode 2.9796 146 0.980 [3] Turbine / ligh fuel mode 16.718 146 0.897 [4] Table 3 Normal mode probabiliy False alarm probabiliy Missed alarm probabiliy Case 1 Case 2 Case 3 Case 1 Case 2 Case 3 Case 1 Case 2 Case 3 Gas deecor 0.95 0.90 0.85 0.02 0.04 0.06 0.03 0.06 0.09 Fire deecor 0.95 0.90 0.85 0.02 0.04 0.06 0.03 0.06 0.09 moke deecor 0.95 0.90 0.85 0.02 0.04 0.06 0.03 0.06 0.09 Table 4 (a) Componen Availabiliy remarks Gas saion 0.95 Ligh fuel saion 0.98 Table 4 (a) Probabiliy Fire 0.001 moke 0.001 Gas leakage 0.001 remarks The load is assumed o have wo saes: 0 MW and raed power. The rae of ransiion from 0 MW o raed power is assumed o be 80 per year and he rae of ransiion from raed power o 0 MW is assumed o be 40 per year. Resuls The sysem is esed for wo cases: case sudy (1) using one deecor for each ype; case sudy (2) using wo-ou-of-hree deecors. Case sudy (1): In his case, he sysem is esed when he number of deecors is one for each ype. Deecor s probabiliies were changed as shown in able 3 and he resuls are shown in able 5.
Ben-Idris 11 Table 5. ysem availabiliy wih changing deecors characerisics Case 1 Case 2 Case 3 ysem availabiliy 0.89773 0.8425 0.8134 Probabiliy of geing false alarm 0.000098 0.00027 0.00062 Probabiliy of no deecing he faul 0.000000023 0.00000094 0.0000063 Case sudy (2): In his case, he sysem is esed when he number of deecors is wo-ou-of-hree for each ype. Deecor s probabiliies were changed as shown in able 3 and he resuls are shown in able 6. Table 6. ysem availabiliy wih changing deecors characerisics Case 1 Case 2 Case 3 ysem availabiliy 0.95423 0.94056 0.92987 Probabiliy of geing false alarm 0.0000000000 0.0000000008 0.000000039 Probabiliy of no deecing he faul 0.0000000000 0.0000000000 0.0000000000 Conclusion In his projec, he effec of fire proecion sysem accuracy on gas urbine is evaluaed. In his regard, wo differen proecion sysem configuraions wih differen probabiliies were considered. The resuls showed ha he use wo-ou-of-hree configuraion is sufficien o proec he sysem and does no have noiceable negaive impac on he sysem reliabiliy even when he probabiliy of sending false alarm and he probabiliy of no deecing he faul were large. Moreover, by using Mone Carlo imulaion, he difficulies of failure dependency are solved. References [1] Yang, F., Deyun, X., and irish,., Opimal ensor Locaion Design for Reliable Faul Deecion in Presence of False Alarms. ensors 2009, doi:10.3390/s91108579, pp. 8579-8592. [2] Bollen, M., Lieraure earch for Reliabiliy Daa of Componens in Elecric Disribuion Neworks. Eindhoven Universiy of Technology Research Repors, 1993. [3] Ei, M., Ogaji,., and Prober,., Reliabiliy of Afam Elecric Power Generaing aion, Nigeria. Applied Energy, Volume 77, Issue 3, March 2004, pp. 309-315. [4] Esrada., Jorge, Redondo., Noemi, Ruiz., Juan, and Barquin., Julian, Including Combined- Cycle Power Plans in Generaion ysem Reliabiliy udies. 8 h Inernaional Conference on Probabilisic Mehods Applied o Power ysem, Iowa ae Universiy, 2004.