Fle:nrel overfll004 fnal report1 rev02.doc January 31, 2008 Revsed: March 20, 2008 TASK ORDER KLDO-5-55636-01 PROPANE TANK OVERFILL ANALYSIS: QUANTIFYING THE SCOPE OF THE ISSUE AND OUTLINING BEST PRACTICES TO PREVENT OVERFILLS TECHNICAL REPORT ON FIELD TESTING TECHNICAL MONITOR: MS. MARGO MELENDEZ By: Mr. Alex Spataru, The ADEPT Group, Inc. I. Background The use of propane vehcles can enhance our energy securty and mprove ar qualty. Today, propane vehcles are most often used for school and shuttle bus applcatons, mass transt as well as lght -duty truck applcatons throughout the Unted States. To maxmze the emssons and energy securty benefts, as well as to ensure safe operaton, the on-board propane tanks must be refueled properly. The Natonal Renewable Energy Laboratory (NREL), Department of Energy (DOE), and the Natonal Propane Gas Assocaton (NPGA) have recently become aware of reported cases where fuel tanks on propane vehcles have been overflled, potentally resultng n emssons from pressure release valves (fuel consumpton and local and global ar polluton ssues); or, n rare cases, cause a potental safety hazard. The frequency of overfllng on-board tanks s not well understood. Lquefed petroleum gas (LPG, LP Gas, or propane) has a hgh thermal coeffcent of expanson. Ths means that when there s a hgh rse n temperature the lqud n the tanks wll sgnfcantly expand. Most tanks are desgned to vent LP Gas n such case as to avod sgnfcant pressure buld-up n the tank. Ths release however, leads to wasted fuel and negatve envronmental mpacts. To avod undesrable releases, the common practce s to fll on-board tanks to a maxmum of eghty percent of total volumetrc capacty. Work conducted snce md-2005 by the ADEPT Group, Inc. (ADEPT) and earler by South West Research Insttute (SwRI) [PERC Docket 11200: Motor-Fuel OPD Falure Mode Evaluaton, 2004) and CEODEUX (domnant maker of Overfll Preventon Devces or OPD s) [per conversaton wth D. Lawson, CEODEUX, 2008] ndcates that vehcular propane tanks are occasonally flled above the eghty percent mark. These overflled tanks, when ventng, pose envronmental and potental safety rsks, as well as a safety rsk f the release valve fals to open when t s supposed to do so. ADEPT and Adept Scence & Technologes, LLC (ASCENT), completed a technology demonstraton project for the Propane Educaton & Research Councl 1 of 11
(PERC) n 2005 whch documented that at VIA Metropoltan Transt (VIA) n San Antono, TX, ~16% of ther on-board tanks were regularly overflled [PERC Docket 11653: Acoustc Stop-Fll Instrument for LP Gas Tanks, 2006]. In 2006 and 2007, smlar nvestgatve and dagnostc work [ADEPT] on transt buses operated by the Los Angeles Department of Transportaton DASH buses and by Port Arthur Transt demonstrated that ths problem was not lmted to VIA. These fndngs confrmed the need to assess the extent of overfllng n propane vehcles across the country. II. Project Objectve Ths project s objectve s to conduct a statstcally sgnfcant survey to answer two questons: (1) Are LP Gas on-board tanks overflled throughout the US, and (2) If the answer to the frst queston s Yes, then answer the queston of how sgnfcant ths condton mght be. Ths objectve dd not nclude dagnoss of the problem, just an evaluaton of fleets to determne the extent to whch the stuaton exsts. Based on the fndngs, a best practces gudelne for refuelng wll be prepared. Subsequent work to dagnose the root cause/s of on-board tank overfllng may be ndcated. III. Statstcal Approach A. Propane Vehcle Market Ths study s to estmate the frequency of the tank overflled condton n LP Gas powered vehcles n US based fleets. On-road LP Gas vehcles operatng as fleets n the US nclude: transt buses, school buses, shuttles, small pck-up trucks, and sedans. To make the study as comprehensve as possble, the selecton of the fleets for ths project was made to nclude a varety of vehcular end-uses. Ths tested fleets ncluded: school buses bottled water delvery trucks utlty pck up trucks lmousnes shuttle buses Fleets were of varyng sze and functon. No mass transt fleets were ncluded n ths study as three such fleets were prevously tested under another program [PERC Docket 11653, 2006]. The orgnal project scope called for fve (5) fleets to be tested. Seven (7) fleets were actually tested wthn the same budget. Wthn the LP gas fleets vehcle populaton, there are natural classes of fleets. These fleets were prevously classfed as Small, Medum and Large. Wthn Servces at the nterface of energy, economcs and envronment 2 of 11
a class, fleets share common characterstcs n terms of number of vehcles n a fleet, type of vehcles, OEM, type of servce and type of routes. The EIA publshed the below table for alternatve fueled vehcles. Our own research n md-2007 ndcates that ths table most lkely underreports by ~20% the number of actual fleet vehcles runnng on LP Gas n the US. Based on our own md-2007 survey of fleets operatng throughout the US, we beleve that there are ~2,400 on-road fleet vehcles runnng on LP Gas n the US today. Medum Duty Heavy Duty Fuel Type Vans Pckups Trucks Total Trucks Buses Total Compressed Natural Gas (CNG) 629 189 55 873 50 461 511 Dedcated 298 58 12 368 35 451 486 Nondedcated 331 131 43 505 15 10 25 Electrc (EVC) /a/ 0 0 5 5 0 3 3 Ethanol, 85 Percent (E85) /b/ 132 65 0 197 0 0 0 Hydrogen (HYD) /c/ 1 0 0 1 0 0 0 Lquefed Natural Gas (LNG) 0 10 1 11 0 0 0 Dedcated 0 0 0 0 0 0 0 Nondedcated 0 10 1 11 0 0 0 Lquefed Petroleum Gas (LPG) 683 181 522 1,386 169 465 634 Dedcated 23 92 8 123 9 448 457 Nondedcated 660 89 514 1,263 160 17 177 TOTAL 1,445 445 583 2,473 219 929 1,148 Dedcated 322 150 25 497 44 902 946 Nondedcated 1,123 295 558 1,976 175 27 202 /a/ Electrc vehcles are battery powered and are consdered dedcated. /b/ Ethanol vehcles are flexble-fueled and are consdered nondedcated; the remanng porton of 85-percent ethanol s gasolne. /c/ Hydrogen fuel cells are consdered dedcated hydrogen because hydrogen s the nput fuel. Notes: Dedcated vehcles are desgned to operate exclusvely on one alternatve fuel. Nondedcated vehcles are confgured to operate on more than one fuel. Medum duty ncludes vehcles wth a GVWR of 8,501 to 26,000 lbs. Heavy Duty ncludes vehcles wth a GVWR greater than 26,000 lbs. Source: Energy Informaton Admnstraton, Form EIA-886, "Annual Survey of Alternatve Fuel Vehcle Supplers and Users" The statstcal technque known as cluster samplng s ndcated when the envronment s not completely known and the basc unts beng analyzed are clustered nto groups sharng common characterstcs (.e. vehcle type, functon Servces at the nterface of energy, economcs and envronment 3 of 11
served, area served, etc.) [See the classc reference Cochran, Wllam, Samplng Technques (2 nd edton), John Wley & Sons Inc, New-York, 1963 ]. B. Cluster Samplng The cluster samplng technque conssts of selectng a small number of clusters (fleets n ths case) and, wthn a cluster, analyzng the unts of nterest ether exhaustvely (sngle-stage cluster samplng) or by drawng a random sample of unts wthn each cluster (two-stage cluster samplng). Here, the unts of nterest are the on-board tanks and the two-stage random samplng was used. Notatons: N = total number of clusters (fleets) M = number of unts (tanks) wthn cluster n = number of clusters sampled m = number of unts sampled n cluster p = estmate of the proporton of unts satsfyng the crteron (tank overflled) The unbased estmate of the proporton of unts of nterest s gven by the formula: Equaton III.1 N pˆ = M p nm and the formula for the varance s: Equaton III.2 2 = N n N + M m Vˆ 1 2 1 1 1 2 1 N M p M j p j M p 1 p 2 M N n n 1 n j n M m 1 To obtan a good estmate of p t s desrable that all clusters be of approxmately equal sze. For that reason, the classes Large, Medum and Small are analyzed separately. The results of the above descrbed statstcal analyss methodology are shown below n Secton VII. Statstcal Analyss of the Collected Data. ( ) Servces at the nterface of energy, economcs and envronment 4 of 11
IV. Feld Tests Methodology The testng protocol was developed from pror experence wth three propane powered mass transt fleets as well as wth help from members of a specal task force (TEF-1714) of the Technology, Standards and Safety (TS&S) Commttee of the Natonal Propane Gas Assocaton (NPGA). TS&S created ths task force after learnng of the overflled tanks fndngs on the frst three tested fleets. At each ste (except Metro Cars n Taylor, MI whch ssued prnted recepts of the volume pumped rather than a meter dsplay) the below 12 steps were followed: 1. The vehcle was verfed to be on level ground when t pulled up to the fuel dspensng unt (pump). The drver, or person who reflled the vehcle, hereafter operator, was told that the refuelng event would be observed. 2. The operator was asked to report the on-the-dash level gauge dal readng, whch was verfed by the ADEPT. The level ndcaton on the on-tank gauge dal was also recorded, f one was present and accessble. 3. The Maxmus Overfll Dagnostc Instrument (ODI) and/or the Maxmus Contnuous Level Gauge (CLG) were used to get accurate lqud level readngs from the on-vehcle propane tanks on all tanks suspected to be overflled. An osclloscope was perodcally used to valdate the Maxmus nstruments. The sequence of vehcles beng fueled was not predetermned; vehcles were tested n the random order that they pulled to the pump. These readngs were compared wth the level ndcatons on the dash board and/or on the on-tank dal/s. The relablty of on-dash or on-tank level ndcator dals s sgnfcant to the extent that, f relable, they could provde a secondary means to check the lqud level n the event of OPD falure. However, 47% of these level ndcators dd not functon relably [based on checks wth the Maxmus nstruments, and/or on reports from the drver / fuelng operator]. At the request of LP Gas ndustry representatves, the ntal volume of fuel n the tank was ncluded as part of the data sheet. As the project progressed, the frequency of ths partcular step was lowered and then elmnated as t became apparent that t was suffcent, n a sngle tank system, to retroactvely calculate, f needed, what the lqud level was at the start of the refuelng process nsde the tank. To collect data for ths calculaton requred recordng the number of gallons pumped n (from the calbrated pump meter) at the tme when lqud came out of the Fxed Lqud Level Gauge (FLLG) and/or the number of gallons pumped n (from the prnt tcket f a pump gallons n dynamc dsplay was not avalable), the max capacty of the tank beng flled, and the fnal amount of gallons nsde the tank (measured wth the ODI and/or the CLG nstrument/s). 4. The refuelng process was observed wth the FLLG open. The FLLG s a valve attached to a dp tube whch extends nto the LP Gas tank. The lower Servces at the nterface of energy, economcs and envronment 5 of 11
end of ths dp tube s typcally located at the 80% full level (although there s an accepted tank manufacturer tolerance of 75% to 82% full for the poston of the lower end of the dp tube). Durng fllng, once lqud reaches the bottom of the dp tube, lqud LP Gas s released through the open valve (as a whte cloud) whch ndcates that the 80% full level has been reached. 5. The # of gallons pumped nto the tank when the OPD trggered was recorded. 6. We also noted n the comments secton f the OPD stopped the flow before or soon after lqud LP Gas spewed through the FLLG. We frequently noted the # of gallons pumped n when lqud frst came out of the FLLG (n case ths nformaton mght later be useful). 7. If the OPD dd not stop the n-flow of LP Gas nto the vehcle after lqud LP Gas was released for some tme through the FLLG, we asked the operator to stop the pump and recorded the amount of fuel pumped n up to that pont. 8. Wth the FLLG stll open, the operator was asked to try to pump a few more gallons n the tank (the actual # of gallons added depended on the maxmum tank capacty and was lmted to less than 10% of ths capacty to avod severe overfllng, f possble) to attempt to force the OPD to trgger. The gallons pumped n ths addtonal refuelng step were recorded. 9. The lqud level ndcators were read and recorded, f accessble and operatonal. 10. Vehcles suspected to have an overflled tank were pulled away from the pump and parked on level ground. Fuel level ndcatng nstruments were then properly postoned. Fve to ten mnutes were allowed to pass before the readngs were taken. Ths was to ensure that any lngerng sloshng nsde the tank dd not affect the readngs, and to allow for further level equalzaton between mult-cylnder tanks. The tank temperature was recorded. Wth the ODI and the CLG, and occasonally the osclloscope, the lqud level nsde the cylnder was accurately measured. Readngs from these two dfferent acoustc nstruments (whch use dfferent measurng technques) were compared to double-check the LP Gas volume n the tank. The fuel levels were recorded. 11. Vehcles wth confrmed (from step 10) overflled tanks, f not already flled to ~100%, were brought back to the pump to try to trgger the OPD wth addtonal fuelng. If the OPD trggered on ths thrd refuelng attempt, the vehcle was not recorded as havng an overflled tank, however, the fleet manager was notfed. If the OPD dd not trgger on ths thrd pumpng acton, the vehcle was recorded as havng an overflled tank. Servces at the nterface of energy, economcs and envronment 6 of 11
(12) Comments (e.g. the presence of an unrelable mechancal level ndcator dal) were occasonally noted on the data sheet. All the vehcles tested had only one tank systems on-board, so no solaton nbetween cylnders were needed. A one tank system means that, although t may have multple cylnders, t was equpped wth one OPD, one pressure relef valve, one outage gauge, and ts cylnders were connected wth welded metal ppes that allow unmpeded and rapd equalzaton of lqud and vapor between cylnders. The above descrbed procedure added lttle tme to the normal refll process at each fleet/s ste on vehcles where the possblty of an overfll was not ndcated. On vehcles where overflls were ndcated, the tank examnaton wth ultrasonc nstrument/s (ODI and/or the CLG) took anywhere from 24 to 33 mnutes. V. Test Stes The fleets ncluded n ths study are shown below. These fleets were selected because of ther type of operaton, type of vehcles, and wllngness to partcpate. Fleet Name Fleet Ste Fleet Sze # of vehcles n fleet Mnmum # of vehcles to be tested Tanks/ vehcle Dallas County Schools - Pat Lancaster, TX Large 168 17 1 or 2 Raney Servce Center Dallas County Schools - Dallas, TX Large 154 15 1 or 2 Don Shelds Servce Center Metro Cars - Van Fleet Taylor, MI Medum 47 5 1 Metro Cars - Town Car Taylor, MI Small 20 2 1 Fleet Sparkletts - Gardena Gardena, CA Medum 30 3 1 or 2 Dstrbuton Center Sparkletts - Van Nuys Dstrbuton Center Van Nuys, CA Medum 24 2 1 or 2 UCLA CTS - Unted Rentals Los Angeles, CA Small 1 1 1 The number of vehcles sampled at each ste was based on the total number of vehcles n the fleet. To be statstcally meanngful, sample szes should be proportonally smlar to overall fleet sze. Gven the total avalable tme at each ste, the number of LP Gas powered vehcles n each fleet to be tested and the antcpated test duraton, a sample sze of 10% of the fleet populaton was used. Servces at the nterface of energy, economcs and envronment 7 of 11
VI. Feld Test Data Summary The below table summarzes the feld data collected from seven fleets: Fleet Locaton Test Date Dallas County Schools - Pat Raney Servce Center LP Gas Vehcles n Fleet Tanks Tested Tanks per Vehcle % of Fleet Tested Tanks Overflled % Overflled Overflls Verfed wth CLG Lancaster, TX 10/28/2007 168 31 1 18% 4 12.9% 2 Dallas County Schools - Don Shelds Servce Center Dallas, TX 10/29/2007 154 48 1 31% 5 10.4% 5 Metro Cars - Van Fleet Taylor, MI 11/14/2007 20 7 1 35% 1 14.3% 1 Metro Cars - Town Car Fleet Taylor, MI 11/14/2007 47 3 1 6% 0 0.0% NA Sparkletts - Gardena Dstrbuton Center Sparkletts - Van Nuys Dstrbuton Center UCLA CTS - Unted Rentals Gardena, CA 11/28/2007 36 9 1 25% 4 44.4% 4 Van Nuys, CA 11/29/2007 36 6 1 17% 3 50.0% 3 Van Nuys, CA 11/29/2007 1 1 1 100% 1 100.0% 1 Total 462 105 18 17.1% 16 In total, seven (7) fleets and 105 vehcles were tested. VII. Statstcal Analyss of the Collected Data As dscussed above, a cluster samplng technque was used. The data contaned two outlers: (1) Metro Cars Medum fleet of Town Cars) wth a szable number of vehcles but only three vehcles/tanks could be properly tested, and (2) UCLA CTS pck-up truck fleet had only one vehcle and thus belongs n the Small category. Consequently, the Small category was elmnated as a cluster because a cluster, by defnton, must have at least two elements. There were two phases of fleet nspecton work: [Phase I: Samplng of 3 fleets (PERC Docket 11653, 2006 and related ADEPT work conducted durng the Servces at the nterface of energy, economcs and envronment 8 of 11
testng of the Maxmus ODI); and Phase II: Samplng of 7 fleets (the DOE study)]. The dstrbuton of the data collected over a two (2) year perod allowed for fve (5) separate studes. Phase I data was ncluded n the studes ndcated to provde greater sgnfcance for the statstcal analyss: Study 1: Phase I excluded / Metro Cars (Med) Town Car fleet (TMM) ncluded (6 fleets sampled) Study 2: Phase I excluded / TMM excluded (5 fleets sampled) Study 3: Both Phases are ncluded / TMM ncluded (9 fleets sampled) Study 4: Both Phases are ncluded / TMM excluded (8 fleets sampled) Study 5: Both Phases,/TMM and TMC excluded (7 fleets sampled). (TMC = Taylor, MI, Metro Cars - Small fleet of shuttle vans) Each of these analyses provded a unque probablty of the tank overflled condton (the lkelhood that a tank would be overflled). These probabltes were calculated by Equaton III.1 (above). Large Medum Small Study 1 0.083 0.147 N/A Study 2 0.042 0.196 N/A Study 3 0.159 0.147 0.111 Study 4 0.159 0.196 0.111 Study 5 0.159 0.176 0.111 Studes 1 and 2 do not contan a Small as the Small fleets from Phase I studes were not ncluded. Also of note s that for the Small category, n Studes 3 and 4, a smple random samplng estmate was used takng PAT to be representatve of the entre Small populaton. Study 5 s shown n the event an estmate of the tank overflled condton s of nterest wthout the fleets at Taylor, MI. Ths s provded because the fndngs were later dsputed by the Metro Cars fleet manager and the LP Gas provder to Metro Cars. Usng the populaton szes for Large, Medum and Small as relatve, estmates of the probablty for the overall populaton of LPG vehcles was calculated, usng Equaton III.1 (above). p est. Study 1 0.11 Study 2 0.11 Study 3 0.15 Study 4 0.17 Study 5 0.16 The most relable study s Study 4, as t contans the largest number of observatons and excludes outlers that ntroduce undue varablty to the estmates and also cause large varances. Servces at the nterface of energy, economcs and envronment 9 of 11
The confdence ntervals for Large, Medum and Small fleets are shown below: p-large: 11% ± 7% p-med: 19% ± 13% p-small: 11% ± 11% A 0.70 confdence level apples to the above results. Ths means that for each of the fleet sze categores above, there s a 70% chance that the probablty of a tank beng overflled s wthn the ndcated confdence nterval. VIII. Conclusons & Fndngs 1. A lttle more than one out of every sx fleet vehcles tested was overflled. 2. Tanks are sgnfcantly overflled n Small, Medum and Large fleets, regardless of ther end-use or of OEM. 3. Drver and/or re-fuelng staff can beneft from addtonal tranng on safe refuelng practces. Very few, f any, were aware of the rsks assocated wth overfllng. 4. Mechancal lqud level gauges on the tested vehcles had a very hgh falure rate (nearly 50%). The fact that the drver and/or the re-fueler cannot rely on the level ndcator n the vehcle and/or on the tank/s aggravates the overflled stuaton. 5. Tanks were overflled due to apparent falure of the Overfll Preventon Devce (OPD) as well as n one case where the OPD was ntentonally removed because t stopped fllng too soon (on prevous occasons the fuel ran out and the delvery truck could not fnsh ts route). IX. Recommended Next Steps ADEPT wll: 1. Prepare a Best Practces outlne for fleet vehcles re-fuelng. 2. Submt ths outlne and the fnal report to NPGA and to leaders of the automotve sector of the US LP Gas ndustry, as well as to the supervsors of the tested fleets. 3. Submt ths outlne and the fnal report to the Texas Ralroad Commsson (TRC). TRC oversees the safety of all LP Gas vehcles n Texas. Texas has about half of all the LP Gas powered vehcles n the US. 4. Suggest to the LP Gas ndustry to: (a) ntensfy tranng for safe refuelng practces, and (b) conduct perodc verfcaton of the proper functonng of OPD s. 5. Submt a Tentatve Interm Amendment (TIA) to NFPA (Natonal Fre Protecton Assocaton) recommendng that NFPA 58 (the Bble for the LP gas ndustry) requre that OPD s be perodcally checked for proper performance. Servces at the nterface of energy, economcs and envronment 10 of 11
6. Make CSA and UL aware of ths study s fndngs. CSA and UL both test and certfy OPD s. 7. Make the on-board tank manufacturers aware of ths study s fndngs. Addtonal Follow-on Work OPD s are mechancal devces. There are several factors that could lead to the falure of such devces ncludng the length of servce (one hypothess from ths work s that more recently nstalled OPDs fal less than older OPDs), presence of contamnants n the fuel tank, devce desgn, nstallaton, tank orentaton, etc. It s recommended that these falure modes be further nvestgated. X. Acknowledgements ADEPT wshes to thank the staff at each of the test stes for ther cooperaton and the tme they dedcated to ths project. We are partcularly grateful to Mr. Bobby Belsher and Mr. Terry Woodfn (Dallas County Schools), Mr. Jeff Pardonnet and Mr. Dan Payne (Metro Cars), Mr. Norm Berry (Sparkletts), and Mr. Marshall Donkn (UCLA Communcatons Technology Servces). We would also lke to thank the staff of the Natonal Propane Gas Assocaton and all members of the NPGA TS&S Task Force TEF-1714 for ther contrbutons to ths project. Lastly we would lke to thank Ms. Margo Melendez and the staff at NREL for all ther assstance and contrbutons throughout ths project. Servces at the nterface of energy, economcs and envronment 11 of 11