You are on page 1of 7
Cage International Conference on Large High Voltage Electric Systems 112, boulevard Hauremann — 76008 Psi 1988 Session — 28th August - 3rd September ° 12-02 Pref. Subj. 2 PREVENTION OF TANK RUPTURE DUE TO INTERNAL FAULT OF OIL-FILLED TRANSFORMER T. KAWAMURA* M. UEDA K. ANDO. T. MAEDA University of Tokyo The Tokyo Electric Chubu Electric The Kansai Electric Power Co, Inc. Power Co, Ine Power Co. Inc Y. ABIRU M. WATANABE K MoRITsU Toshiba, Lid Hitachi, Lee Mitsubishl Electric Corp. (Japan) ‘Sumnat 2, Main Characteristics Affecting Tank Rupture Fundanontal characteristics affecting tank 2-1, Decomposition gas Amount generated by Internal rupture were studied and computation method of pressure rise during fault was established, Tt vas proved that the utilization of dlaphragn type conservator as pressure reducing space is effective when tank is 80 stifr. The dynamic pressure tests on full scale tanks using the explosive powder vere carried out and justified the analysis and improving measures. Keywords Transformer, O11, Tank, Rupture, Failure, Fault L, Introduction In order to prevent transformer failures, an extensive study has been made ané lead to much improvements in design, production and maintenance technics. The diagnosis methods, such as dissolved gas analysis, for detecting abnormalities in their early stage ‘were developed and have bean widely utilized, However, failures may still happen, even Lf the most advanced technologies are adopted. ‘When an snternal fault takes place in an oil filled transformer, a large volume of decomposition gas is generated by are, resulting in the rapid increase of tank pressure. In case of a heavy fault, the pressure relief devices cannot keep up with the gas generation and a tank rupture may cccur. The tank rupture causes an cutflov of a large anount of oil and nay also cause a fire by the contsct of hot combustible gas with oxygen in outside air, which Vill give serious effects to environsent. ‘A very deep concern exists in Japan about these public hazards, as many substations are located very close to the human residence and underground sub~ stations are increasing in urban districts. In this regard, a systenatic study was carried out in Japan by the organization called "Society of Electrical Cooperative esearch", which is supported by universities, research institute, utilities and manufacturers. The results of the stidy are present= ed in this paper. (a)__ Are Voltage The are voltage is not related very much to the are current, but changes by the arc length, elec- trode shape, pressure, etc. Pigure 1 shows the relationship between the are voltage and arc length. For exanple, the flashover distance at a 275K lead terminal is estimated to be 200 to 300 tam, and the are voltage can be estimated to be 5000V. 10,000) 5,000 — 3,000|—1— 2,000} 1,000}—- | soo LF Sy nae chart plots values ee neasured by the japanese transformer nanvfacturers, Are voltage (W) dl L ‘0 3030100700300 500"T,000 Are length (mm) Figure 1, Are voltage (b) Are Current ‘An are current ie a fault current of the system at that point in a line terminal fault inside @ transformer, It can be calculated based on the tranaformer Supedanee viewed fron the short esreuse tap ae far ae shore elreuit fault between tape inaide the winding 1s concerned. For example, if a fault has occurred on a load terminal of 275KV, the arc current is calculated based on the short-circuit capecity at che 275LV terminal of the transformer. (©). Are nergy Amount Figure 2 shows are vaveforns during a fault current and are voltage * Institute oF Industrial Science,University of Tokyo. 22-1 Roppongi Tchome,Minato-ku, Tokyo 106, Japan 12.02 ‘The are energy amount E per unit second vhich occurs at this time ean be calculated using the Following equation: Ave current / pave voltage yf ola Figue 2. Ave corent andar voltage B Haff 27 vol3 te sind feat = 09+ v1 (Wesel) wo where E: Are energy amount per unit tine (hil see/see) + Are voltage (V) Are current effective value (kA) Fault duration time (see) + Frequency (I/sec) (@)_ deconposition Gas Generation Amount per Unit ‘ave Energy ‘A decomposition gas generation anount of 0.5 £/ki sec (at normal pressure and 2000° k) under a normal pressure at the tine of an accident is considered ‘appropriate. (@)_ Fault Duration Tine the fault duration time is decided taking into consideration the protective relay operating tine, cireuit breaker breaking tine, and other factors 2-2, Tank Expansion Characteristics ‘The tank expansion anount is calculated by individually obtaining pressure characteristics including the elastic areas when the pressure is lov and the plastic areas vhen the pressure 1s high based fon the following approaches. 1) Generally, the elastic limit pressure is roughly 2 kg/en?. However, the pressure at which the cal~ culated Stress becomes 30 kg/am? is individually caleulated. 2) The expansion coefficient in the elastic area 1s caleulated based on the deformation amount during a vacuum test of the tanks 3) The expansion coefficient plastic deformation after the Pressure is surpassed. The increase 1s estinated to be twice the expansion coefficient of the elastic area. ‘Appreache: Expansion characteristics obtained in a characteristic test at hydrostatic pressure. Expansion characteristics obtained in stress analysis spreading over elastic and plastic ‘The pressure value obtained in a full-scale test can be covered by the analysis value if analysis is made assuming the increase to be twice. In accordance with these approaches, twice va get a8 a Value on the safe side. sncreases due to elastic limit tank 2-3, Tank Strengeh It has been considered that the tank strength is decided by the strength of the reinforcing beans if the joining flange parts are reinforced and that the critical value 1s estimated to be approxinately 3kg/en? based on evaluation of generated stresses dn 2 clastic areas. Hydrostatic tank and explosive combustion tests conducted ac full-scale teste show that the flange part reinforcing mecoures generally used at present in Japan such ae C shaped clamps (Refer to Figure 3) and Joint reinforeenent are nore effective than originally anticipated and that a tank strengh of roughly Skg/em? can be obtained by auch methods. Upper tank ( shaped claxp| Welded part Reinforcing bess Welded part Lower tank @ [} Figure 3. © shaped clamp 2-4. Severity of an Internal Faule The coverity of an internal fault depends mainly on the are energy and the tank expansion coefficient. The are energy becomes larger and the tank expansion coefficient becomes enaller, the severity becomes bigger. Generally, an internal fault at = high voltage Line lead of s split type 3phase transformer is very severe. 3. Numerical Analysis for Pressure Rise in Tra former Tanke 3-1, Pressure Rise Analyete without Oi1 Motion If the influence of the ofl notion is smell, then the internal pressure builds up uniformly in the whole tank during the arcing faults. The traneforser Eongeieuted with only one tank can be treated by this sathenstical model. In the case of a com pletely oil filled transformer, the generated and Compressed arcing gas volume ie equal to the expand- ing Voluse of the tank, Therefore @) ‘A + expansion coeffictent of tank (ca2/kg/em*) Instantaneous pressure rise (kg/cm?) C+ arcing gas generating rate at normal (cn/feassee) are poser (kw.sec/sec) time’ (sec) If 1 1s assuned that the expansion coefficient of tank A is constant against the pressure the above equation becomes 7) o On the other hand, since the expansion coeffi- cient A is varied non-linearly as shown in the previous section, the instantaneous pressure rise in Eank P(t) cen be also calculated exactly by nuner- ical methods from equation (2). If the peak value of the instantaneous pressure P(t) is to be denoted by Ppeak, Ppeak is mean value of the pressure rise in the tank. Practically, the pressure builds up with the oscillation and ususiform Getribueion in the tank because of the kinetic = effects, Therefore, maximum pressure rise Pnax developed in the tank ie expressed se Pnaz = DLE + Peake “ where DLP which is called dysamie load factor ts 2 function of the dixension and stiffness of the tank, the location of the fault, the arcing gas generation rate. For the high energy faults on the large transformer, it was cleared chat the dynamic load factor is approxinately 1.3 from the theoretical and ‘experinontal results. According to this theory, Figure 4 and 5 show an allowable fault current ageinet the are duration without tank rupture for a ofl filled large trans~ former. “Those results shov if the high-speed efreuit breaker will interrupt the fault within 60 te 80 nsec, violent Failure with tank rupture can be prevented. The pressure relief device, further, cannot reduce the internal pressure developed” by such high-energy fault. These reeules vill be intvoduced by the theory shown in the next 3-2 section, 6 Tank Eapanston sq oert. = 49 E ad = 20; * 10; Tenk Strength: Skg/en? | eelveheages sooo ab. ; Yo its sd 8 Bo hve Duration fasec) Migure 4, Allowable Poult Current for Tank with Vartous Erpanston coeff event 6 50} are Voltage Za E sol rare 19 th: Skg/em? Tank Expansion Goeff.+ 6008/kg/cm” Gee ee ey a a CY ‘Are Duration (asec) Allowable Fault Curvent for Yarious Are Voltage Figure 5. 3-2. Pressure Rise Analyeis with O11 Motion In the case where the transformer 1s divided in some tanks and connected with lead ducte such as the split type 3+phase transformer shown in Figure 8, the “Kinetic energy of the oil has a considerable influence on the pressure rise. The mathematical model is shovn in Figure 6, 12.02 Tne Tanke Dutt Dect Preaure “QU Dep. Presre OLDE. 7 ur Pie a Expansion Coll. A, Expansion Col sy Figure 6, Machematical Model for Pressure Rise In this case the following three equations are introduced from the volume balance according to the state of the tank. () For the tank including faule point av-830= SGebh Ses Uil0-8- 009 ° (2) for ehe tank wishout fault Ap PHO= Ser Uri 8 -UO © (3) For the tank with gas space AFA) 4 “Ust-8 “Ud pg o cross section area of duct (cn?) displacement of of1 (cn) volume of gas space (cm?) For the oil at the duct the momentum equation te given as follows 0-0 E09 #8 a B49 140! © specsde vetghe ofoat Chglee™) length of duct (cm) scioctg ar oat cee floes) ‘These equation can be also solved by nunerteal methods and the instantaneous pressure P(t) can be caleulated. In addition, the maximum pressure rise in the tank including fault point mist be given by equation (4). Figure 7 shows an allowable fault current against the are duration without tank rupture for split type 3-phase transformer. This result indicates that there ia a possibility of violent failure with tank rupture on this transformer. 60 Tonk Strength + Ske/en? Tank Expansion Coeff. : 3002/kg/en? sol Lead Duct Length : 2.5m 7 Are Voltage : 5000 g = wo Eo E30 Lead Duct é Cross Section 4 1.0m? 10] 0 3008 350200350 Are Duration (asec) Figure 7, ALlovable Fault current of Split Type 3-Phase Transformer 3-3. Effect of Pressure Reducing Space hon the expansion coefficient of the tank is 20 small that the pressure rice might exceed the permissible value such as the split type J-phaso transformer, it 1o recomended to utilize = dia phragm conservator as a pressure reducing space. The 12.02 practical structure applied to the S-phase transformer ts shown Figure 8. Ty Diaphrags Pressure Reducing Space; 7 Conservator Gas Volume + Min. 20002 “F connecting puct Connecting Duct — Connecting Duct "tous Section split type 210m? Ack Cross Section ength + 500 am cd ; Pressure ase Connecting Duct (aieees ing Due Redveing Practical Structure of Pressure Reducing Space Figure 8. ‘The effect of the pressure reducing space is calculated by the previous equations, and the calculation result is) shova in Figure 9 And this ‘Study shows that the connecting duct should be shore With enough cross section area and the conservator Should be designed vith sufficient air space. Reducing Rate of Tonk Pressure (@) Conservator Pressure (¢) Tank Pressure Gas Space Volune (2) Transformers Split Type 3-Phase Transformer Tank Expansion Coeff. : 3008 /kg/en? Phase Connecting Duct; Cross Section : 0.64 m? Length + 2.5m Connecting Ducts (Grose Section’: 1.0m Length + 1.0m Figure 9, Effect of Pressure Reducing Space 4. Results of Full se Test 4-1, Outline of the Models and Internal Pressure Rise Siaulation 4-1-1, Scale of the Models The transformers as stated in 2-4 were chosen fas the experimental study topic, and two tank models each coupatible with a 275kV" split type S-phase transformer, core type and shell type, with a capacity equivalent £0 that of a real unit (300NVA, I-phase), were used to make investigations on what behavior of pressure they would take in the event of a fault. The transformer full scale tank models (see Figure 10) were fabricated with enphasis placed on the folloving points: =i = the transformer tank mist have a strength and an expansion coefficient equivalent to those of the factusl product transformer, (it must have such provision ss a C-shaped clamp.) = the transformer tank must be designed to allow the addition of an adjacent tank that simulates another phase of tank, This is necessary to check the Phase-to-phase tank connection structure (1.e., Inter-phase connecting duct) ef 2 split type I-phase traneforner. = The duct which connects che tank with che conser vator must be so designed as to be replaced for checking its effect. = For simulation of the transforner content, a steel box is Fixed inside the tank. 4-1-2, Simulation of Ares by Powder Simulation of a pressure rise within the tank that would occur in the event of a fault in a transformer nay be implenented by applying a current into the tank causing are discharge to develop. However, such an electrical test vas found impossi- ble because of the Limitation to the shorting generator capacity and of the difficulty to obtain fn adequate test place, Thus our simulation was perforned using a method of causing powder to burn ina tank filled with water. Conservator Connecting duct (Pressure reducing chenber) Adjacent tank oy Ge ® 18 LD 1° o. ‘rank mode{ Simulated trans- Povder combustion 1 former content ve model (a) Core Type conservator (ressure redveing chanber) Connecting duet is a @ +o rat wide | sioasaed ecans- Sat Powder combustion vessel (b) shell Type Figure 10, Construction of Transformer Tank Model The adequate capability of this simulating the rise in the internal pr tank caused by arco in ofl has already b (eee REF.) nothod of Simulation of arc energy generated in the event of an dnternal fault vas performed as stated below, Using ae the basis the results of tests using small-scale tank model (including the ability to generating a maximum target pressure in a target time) (Q). Test for gathering the effect of pre: data needed for analysis of jure reducing space. Combustion vessels 26 shown in Figure [1 were used, and che anount of power per vessel was Set at 500g. Thens = For the target pressure about 1.5 kg/em? (equivaient co about 1,000,000 kVA in terms of short circuit capacity), one vessel vas ignited aca time. = For the target pressure about 2 kg/cm? (equivalent to about 2,000,000 kVA in terms of short circuit capacity,’ to vessels were ignited at the same tine. (2) Test for verifying the threshold tank strength The anount of powder equivalent to a short cireuit capacity of 15,000,000 kVA was set at 7,300 g. Three combustion vessels (each having an amount of powder of 2,500 g) vere ignited at the same time. The point of arcs being generated in the event of a transformer internal fault was assumed to be the 275 kV lead vhich would encounter the highest rise in internal pressure. The combustion vessels vere placed in an intermediate position for the core type nodel, and at the top for the shell type model. The case placed on top of the core type model vas also tested with a view to analyzing the effect on the fault point. Tenition Lead pressure Detection Port Gas Discharge Ports (72 M12 Holes) Screen Tube Screen Tube | Conbustion Vessel Figure 11, Construction of Powder Combustion Vessel 4-2, Test Results 4-2-1, Comparison with Analytical Results ‘The behavior of internal pressure found from the results of the above testing has the following characteristics: = The pressure rise begins in the tank having the fasaumed fault point, then developing to the pressure reducing chamber within the conservator and to the adjacent tank. 12.02 = The pressure other than that in the tank having the assumed fault point takes an oscillation waveform with = long period which may be caused Saialy by the motion of ‘the fluid, = Por places distant from the fault point within the tank, the pressure values measured at various points are nearly equal. Thus it is considered that uniform pressure develope in such places. = The pressure near the assumed fault point takes a Gifferent waveform when observed innediately after occurrence of the simlated fault, In addition, various pressures vith short periods overlap with each other at some points. This probably is attributable to the fact that, near the assused fault point, the fluid moves three-dimensionally, resulting in localized overlap of complex motions. Figure 12 shows comparisons between the pressure waveforms (shown in dotted lines) derived from the analytical method stated in the previous section and those (shown in solid lines) obtained from the full-scale test. The folloving points are Found: = The pressure ar a poine distant fron che assuned fault point can be given by an analytical equation considering the fluid's motion of inertia. Measured dat “= Analyeteal values TS tial values [TS Pressure sear Sroned ful poine Precoure tn cen! of tank whieh Ioeat pre Hpbtained with novenene, of f ‘vse taken Eel 5 AE Scoot pe reducing QL eee a * i Time (ms) (@) core Type 1m cop ani (having romed Foul point) 1.3 dynamic leading factor ani (eonmected to top tani) a ‘Mid taken Pressure reducing pace (comerted fo cop cank) we rime (ms) (b) sheti Type Figure 12, Comparisons between Analytical Valu and Measured Data (Are Energy: 142,000 kil, with 3,000-liter Pressure reducing Space)

You might also like