LNG INCIDENT IDENTIFICATION A COMPILATION AND ANALYSIS BY THE INTERNATIONAL LNG IMPORTER S GROUP

LNG INCIDENT IDENTIFICATION A COMPILATION AND ANALYSIS BY THE INTERNATIONAL LNG IMPORTER S GROUP IDENTIFICATION DES INCIDENTS GNL UNE COLLECTE ET UNE ANALYSE REALISEE PAR LE GROUPE INTERNATIONAL DES IMPORTATEURS DE GNL A. Acton BG Group U.K GIIGNL Study Group Chairman F. Katulak Tractebel LNG North America J. L. Deveautour Gaz de France K. Sakamoto Osaka Gas Company, Ltd. M. Kan Tokyo Gas Co., Ltd. ABSTRACT Safety in LNG Operations has been of utmost importance to the industry since its inception. The International LNG Importer s Group (GIIGNL) first began a comprehensive study of incidents relating to the global LNG industry in 1992 and has recently updated its work during 21 to include all incidents reported up to that time. The work has been undertaken by a Study Group comprising representatives from GIIGNL Member companies in the USA, Europe and Japan who obtained data from all the GIIGNL Members. The goals of the study were as follows: To identify actual incidents of LNG or vapor release for possible inclusion in the hazard analysis of new, modified, or existing facilities. To advise on the severity of the identified LNG incidents to assist in evaluation of their importance and potential consequences. To provide information on the circumstances under which the identified LNG incidents occur (and their frequency where possible) to assist in the evaluation of their relevance to the particular LNG facility under review. Incidents have been categorized by type, function being performed, and severity. A frequency and trend analysis has been performed on the data presented in order to identify trends and draw conclusions. The general trend is for a decrease in the relative number of events where significant amounts of hydrocarbons are released. This indicates continual improvement in mitigation measures and procedures at facilities. PS7-3.1

The most important fact drawn from this comprehensive study for the global LNG industry is that there have been no reports of offsite damage resulting from an incident at an LNG facility. This is important both in demonstrating the industry s commitment to safety and is demonstrable evidence of the adequacy of design and procedures employed by the industry, as well as the codes and regulations which govern LNG facilities. RESUME La sécurité dans l exploitation des installations de GNL a toujours été de la plus haute importance dans cette industrie depuis sa création. Le Groupe International des importateurs de GNL (GIIGNL) a initié en 1992 une étude détaillée des incidents qui se sont produits dans l industrie du GNL. Ce travail a été mis à jour en 21 pour prendre en compte tous les incidents enregistrés jusqu à cette date. Cette étude a été réalisée par un groupe de travail comprenant des représentants des compagnies membres du GIIGNL des USA, d Europe et du Japon qui eux-mêmes ont pu récolter des données en provenance de tous les membres du GIIGNL. Les objectifs de cette étude sont les suivants : Identifier les incidents avec fuites de GNL ou rejets de gaz pour les inclure dans les études de dangers des installations nouvelles, modifiées ou existantes. Conseiller sur la sévérité des incidents GNL identifiés pour permettre une évaluation de leur ampleur et de la gravité de leurs conséquences. Fournir des informations sur les circonstances dans lesquelles ces incidents de GNL se sont produit (et leur fréquence si possible) afin d aider à l évaluation de leur occurrence pour les installations de GNL à étudier. Les incidents ont été classés par type, fonction et par gravité. Une analyse de leur occurrence et des tendances a été réalisée dans le but d identifier les évolutions et de pouvoir tirer de premières conclusions. La tendance générale est une diminution relative des incidents conduisant à une fuite significative de GNL. Cette constatation souligne les améliorations constantes dans les mesures de prévention et dans les procédures d exploitation. Du point de vue de l ensemble de l industrie du GNL, cette étude détaillée établit qu il n y a pas eu de conséquence à l extérieur du site suite à un incident survenu sur les installations GNL. Ce constat est à souligner à double titre. D une part cela démontre l engagement de l industrie dans la sécurité des installations et d autre part cela souligne l adéquation des procédures utilisées ainsi que des codes et des réglementations qui régissent les installations de GNL. 1. INTRODUCTION The International LNG Importer s Group (GIIGNL) has developed a database of incidents of LNG release or vapour release at the LNG facilities their Members own and operate. The key objective of the work is to provide real data for use in future hazard assessments. The database covers the period from the beginning of the commercial LNG Industry in 1965 up until late 2. PS7-3.2

The study covers 38 (mainly) receiving terminals and peak-shaving plants belonging to GIIGNL members. The scope includes all LNG incidents with the potential to cause damage to equipment or injury to personnel. We have not sought to provide a unique list of LNG incidents to be included in all hazard assessments because of the site-specific nature of some LNG plant and equipment and also because operators or regulators may wish to include scenarios for incidents that have not occurred due to the potential severity of their consequences. Other incidents that are well-known within the LNG industry but are outside the scope of this study were reviewed but were not included in the quantitative analysis because they were already reported and also because the available information was insufficiently detailed. The work was led by the Permanent Technical Study Group of GIIGNL which, at the time, comprised representatives from BG, CMS Energy, Tractebel LNG North America, ENAGAS, Fluxys, Gaz de France, Gastransport services, Osaka Gas, GNL Italia, and Tokyo Gas. BG provided the Secretariat for the Study Group. Regional co-ordinators from within the Study Group communicated with companies within their regions and also provided a single source of contact for the collection and distribution of incident information. Tokyo Gas and Osaka Gas acted as Far-Eastern Regional Co-ordinators to collect and process data from about half each of the 11 companies in their region. Tractebel LNG North America and Gaz de France acted as American and European Regional Co-ordinators respectively. Advantica Technologies of the UK prepared the comparative analysis for the Study Group. 2. THE GIIGNL SURVEY Information on incidents was first collected from previous GIIGNL studies and also on recent incidents known to the Study Group Members that had not featured in those studies. This was used to initiate the work but it was immediately recognised that more details were required for any useful analysis of results to be performed. Reports of events considered to be routine operational incidents were disregarded. Information was also reviewed for a number of incidents in the public domain that were technically outside the scope of this study because the plants concerned did not belong to GIIGNL members. Detailed information on incidents was then obtained by a questionnaire that was developed by the Study Group and then circulated by the regional co-ordinators to all of the GIIGNL companies in their regions. Incidents were divided into three categories based on definitions developed for the study, as follows: Category 1 - Releases of Hazardous Material Any release of LNG, LPG, NGL, Liquid Nitrogen or related hydrocarbon gases leading to, or with the potential to lead to, injury to personnel or damage to equipment or buildings either on or off site. Category 2 - Near Misses Any incident involving a hazardous material system where there was no actual release of hazardous material but which had the potential to lead to a release of hazardous material as described in Category 1. Category 3 - Other Incidents of Concern Any incident, not involving a hazardous material system in the LNG plant, but, for example, with the potential to escalate to Category 1 release. PS7-3.3

Companies were asked to report the Immediate and Primary Causes of the incident in order to distinguish between the precipitating event and its basic cause. Companies were also asked to provide a summary of the key points of the incident. It was recognised that a definition of the Immediate and Primary Causes of incidents can be a subjective view so this was supported by detailed reviews of each incident by the Study Group and a final consistency check between similar types of incident by a subgroup of European Study Group Members. One of the most important points in the incident reports was an estimate of the total quantity released. This had proved difficult to specify in the past and so three quantity bands were defined: A: Less than 1 kg B: Greater than 1 kg but less than 1 kg C: Greater than 1 kg Extra information such as date and time of the incident, sequence of events, extent of consequences, details of any casualties, effect on plant production, size of vapour cloud or liquid pool and actions taken to prevent it from happening again, were all obtained where historical records were accurate enough for it to be provided. The study was undertaken in two phases, initially covering the period from the start of LNG importation in 1965 to Q1, 1994 and subsequently from Q2, 1994 to 2. 3. INCIDENT DATA AND ANALYSIS The incident database currently contains 144 Previous incidents for the period 1965 to 1994 and 12 New incidents from 1994 to 2 making an All incident total of 246. 3.1 Analysis by Function The following groupings were used for the definition of equipment functions: Unloading: Storage: Send-out: External: Other: LNG Ship, Jetty & Unloading Facilities. LNG Tanks, In-Tank Pumps & BOG Facilities. Pumps, Vaporisers & Send-out Facilities (including LPG equipment) Equipment outside the control of the terminal, other than the LNG ship. Utility equipment & any other equipment not included above. The breakdown of incidents by function is shown in Figure 1 below. There is no appreciable change in this data from the Previous to New periods and so only All data is shown. PS7-3.4

By Function External 6% Others 12% Storage 21% Sendout 34% Unloading 27% Figure 1. Incident Analysis by Function The major functions, unloading, storage and sendout account for the majority of the incidents. However, (ship) unloading is a discontinuous function at conventional import terminals so the frequency of incidents as a function of operational time for this function is the most significant of the three, see Section 3.4. 3.2 Analysis by Category The three Incident categories are: Category 1: Releases of Hazardous Material Category 2: Near Misses Category 3: Other Incidents of Concern The breakdown of incidents by category is shown in Figure 2 below. There is no appreciable change in this data from the Previous to New periods and so only All data is shown. PS7-3.5

Category 3 18% By Category Category 2 14% Category 1 68% Figure 2. Incident Analysis by Category 3.3 Analysis by Function and Category The breakdown of incidents by function and category is shown in Figure 3 below. There is no appreciable change in this data from the Previous to New periods and so only All data is shown. By Function for Each Category 1 8 6 4 2 3 2 1 Storage Unloading Sendout External Others Figure 3. Incident Analysis by Function and Category (Numbers of Incidents Reported) The major functions, storage, unloading and sendout account for the majority of the category 1 incidents (hazardous releases). For the incidents assigned to the Other function category, the most significant numbers were associated with electrical/control equipment failures. PS7-3.6

3.4 Frequency Versus Duration for Each Function For the Previous period from 1965 to 1994, estimates were made based on data from GIIGNL and SIGTTO (Society of International Tanker & Terminal Operators) of the cumulative operating period of the import terminals and peak-shaving sites included in the scope of the study as summarised below. 533 operating site-years (44 for 1965-1974, 179 for 1975-1984 and 31 for 1985-Q1, 1994). 3 6 tank-years used for storage analysis. 2 25 ship voyages used for unloading analysis Total throughput of 868 x 1 9 m 3 (n) of natural gas sent out by import terminals used for send-out analysis. For the New period from Q2, 1994 to 2, operational data was collated via a survey of terminals as summarised below. 28 site-years 1 11 tank-years 12 ship voyages Total throughput of 67 x 1 9 m 3 (n) of natural gas. For the purpose of analysis, the above operating data has been converted into millions of operating hours for each of the functions with the exception of sendout. The frequency of incidents within each function was calculated and the results are shown in Table 1 and Figure 4 below. Storage (Per million hours) Table 1. Incident Frequencies per Function Unloading (Per million hours) Sendout (Per 1 9 m 3 (n) of gas) External (Per million hours) Others (Per million hours) Data Previous 1.34 176.95 54.15 1.71 2.14 New 1.64 166.67 6.96 3.31 1.47 All 1.42 173.13 56.95 2.16 4.47 PS7-3.7

Frequency per Function 'Previous' Frequency per Function 'New' 2 176.95 2 166.67 Incident Frequency (per million hrs) 15 1 5 1.34 1.71 2.14 Storage Unloading External Others Incident Frequency (per milion hrs) 15 1 5 1.64 3.31 1.47 Storage Unloading External Others Frequency per Function Incident Frequency (per milion hrs) 2 15 1 5 173.13 1.42 2.16 4.47 Storage Unloading External Others Figure 4. Incident Frequencies per Function The analysis shows that events during Unloading functions are the most frequent although there is a small decrease in the rate for the New data compared with the Previous data. These events are approximately 1 times more likely per operating hour than for any other function. However, it is important to realise that the unloading operation is relatively infrequent at import terminals, typically 16-2 hours per week. Small increases in incident rates for Storage, Sendout, External and Other functions are apparent between the two data collection periods. Whilst factors such as the increased throughput and ageing equipment could have an influence, the introduction of formal reporting procedures will likely have increased the number of incidents recorded. 3.5 Analysis by Incident Date & Time Incident Date. Using the dates provided in the questionnaire report for each incident, an incident frequency has been estimated for 4 periods of operation since the commercial LNG industry began 35 years before completion of the present survey: The historical incident frequencies per site-year are shown in Table 2 below. PS7-3.8

Table 2. Historical Incident Frequencies Period Incidents Site-years Frequency (Incidents/Siteyear) 1965-1974 15 44.34 1975-1984 52 179.29 1985-1994 94 327.29 1995-2 85 191.45 Total 1965-2 246 741.33 As noted in the previous section, the increase in incident frequency in the most-recent 5-year period may be due to factors such as the increased throughput and ageing of equipment but is likely to be influenced by improved reporting rates following the introduction of formal reporting procedures. Incident Time. Analysis was based on an operational day period from 6: a.m. to 1: p.m. There is no appreciable change in this data from the Previous to New periods and so only All data is shown in Figure 5 below. Incident by Time of Day Night 19% Day 81% 3.6 Analysis by Incident Cause Figure 5. Incidents During the Day and Night Two levels of incident causality were defined in the study, Major Immediate Cause and Main Primary Cause in order to distinguish between the precipitating and the underlying causes of incidents. Major Immediate Cause. Four Major Immediate Causes were defined as follows: Operation (concerning an action, including maintenance) Material (concerning equipment, materials or installation) External (concerning anything out of the influence of the terminal) Unknown (concerning those where the Immediate Cause was never found) PS7-3.9

The results are shown in Figure 6 below: Operation 27% Major Immediate Cause 'Previous' Unknow n 2% Operation 23% Major Immediate Cause 'New' Unknow n 5% External 14% Material 57% External 2% Material 52% Operation 25% Major Immediate Cause Unknow n 3% Material 56% External 16% Figure 6. Incident Analysis by Major Immediate Cause It can be seen that, in general, there are more incidents with a Major Immediate Cause that can be attributed to problems with Material (including equipment and its installation) than those that are caused by Operation (including maintenance), which is the second most-important cause. However, there is a reduction in the proportion of incidents due to Material in the New data compared with the Previous data. The breakdown of categories of incidents for each Major Immediate Cause is shown in Figure 7 below: PS7-3.1

Major Immediate Cause v Category of Incident 'Previous' Major Immediate Cause v Category of Incident 'New' 7 6 5 4 3 2 1 63 26 12 7 5 7 8 8 5 3 Material External Operation Unknow n 1 2 3 7 6 5 4 3 2 1 46 17 7 1 5 5 5 1 1 2 3 Material External Operation Unknow n 1 2 3 Major Immediate Cause v Category of Incident 12 1 8 6 4 2 19 43 13 14 18 1 12 9 1 5 3 Material External Operation Unknow n 1 2 3 Figure 7. Incident Analysis by Major Immediate Cause and Category (Numbers of Incidents Reported) The results show that incidents with Material as a Major Immediate Cause (i.e. defective equipment, materials or installation) tend to involve releases of hazardous substances, and external causes are less likely to bring about an incident involving a release. Main Primary Cause. Four Primary Causes were defined as follows: - Design/Construction, - Operation/Maintenance, - External Cause, - Unknown, For these causality groupings, the Design/Construction cause was restricted to incidents caused by a problem with the initial design and installation of the equipment. Any incidents caused by equipment failure during operation, operator error, poor procedures and poor maintenance were grouped together under Operation/Maintenance. In other words, these incidents were caused as a result of the operation of the terminal and could not be attributed to poor design or installation in the first place. The results are shown in Figure 8 below: PS7-3.11

Unknow n 3% Main Primary Cause 'Previous' External 8% D/C 27% Unknow n 7% Main Primary Cause 'New' External 16% D/C 12% O/M 62% O/M 65% Main Primary Cause External 11% Unknow n 5% D/C 21% O/M 63% Figure 8. Incident Analysis by Main Primary Cause It can be seen that O/M (operation & maintenance) is the biggest Main Primary Cause of incidents and that the proportion of incidents with design and construction as the Main Primary Cause has reduced between the Previous and New periods. Major Immediate Cause Versus Main Primary Cause Material. Figure 9 below shows that for the Previous data, 51 % of incidents cited as having Material as a Major Immediate Cause had a Primary Cause that was a design fault or an equipment failure. For the New data this has increased to 69 %. This suggests an improvement in maintenance, better procedures and less operator error. PS7-3.12

Primary Cause for Major Cause as Material 'Previous' Primary Cause for Major Cause as Material 'New' Op Error 9% Poor Maint 27% Poor Procs 11% Ex/Unknow n 2% Op Error 6% Poor Maint 15% Poor Procs 6% Ex/Unknow n 4% Design/Equip 51% Design/Equip 69% Primary Cause for Major Cause as Material Poor Maint 15% Poor Procs 9% Ex/Unknow n 4% Op Error 7% Design/Equip 59% Figure 9. Primary Causes for Material as Major Immediate Cause Operation. Figure 1, below, shows that for the Previous data, 74 % of incidents cited as having Operation as a Major Immediate Cause had Primary Causes that were due to human actions (Operator Error and Poor Procedures). For New data, this is 1 %. Therefore, it is clear that these kinds of incidents can be reduced by better training of operators and an improvement in procedures. PS7-3.13

Primary Cause for Major Cause as Operation 'Previous' Primary Cause for Major Cause as Operation 'New' Ex/Unknow n 13% Design/Equip 13% Poor Procs 22% Ex/Unknow n % Design/Equip % Poor Procs 31% Poor Maint % Op Error 43% Poor Maint % Op Error 78% Primary Cause for Major Cause as Operation 'New' Ex/Unknow n 8% Design/Equip 8% Poor Procs 27% Poor Maint % Op Error 57% Figure 1. Primary Causes for Operation as Major Immediate Cause 3.7 Analysis by Release Quantity Analysis of the Category 1 incidents (the only category that includes actual releases of hazardous material) produced the results shown in Figure 11 below: PS7-3.14

>1kg 36% Quantity of Release 'Previous' <1kg 49% 1-1kg 19% Quantity of Release 'New' >1kg 8% 1-1kg 15% <1kg 73% Quantity of Release >1kg 25% 1-1kg 17% <1kg 58% Figure 11. Incident Analysis by Release Quantity Further analysis by Send-out, Storage and Unloading functions produced the additional results shown in Figure 12 below: Function for each Release Quantity 'Previous' Function for each Release Quantity 'New' 5 5 4 4 3 2 Unloading Storage Sendout 3 2 Unloading Storage Sendout 1 1 <1kg 1-1kg >1kg <1kg 1-1kg >1kg Function for each Release Quantity 1 8 6 4 2 Unloading Storage Sendout <1kg 1-1kg >1kg Figure 12. Release Quantity for Unloading, Storage and Sendout Functions (Numbers of Incidents Reported) PS7-3.15

There is a general trend between the Previous and New data away from incidents involving the larger releases of hazardous materials. The improvement for unloading facilities is particularly striking. 3.8 Incident Analysis by Gravity Incidents have been classified according to their gravity by cross-referencing the incidents of release of hazardous material (Category 1 Incidents) against the range of consequences reported, i.e. Explosion (E), Fire (F) or Rapid Phase Transition (RPT).There is no appreciable change in this data from the Previous to New periods and so only All data is shown in Figure 13 below. Gravity of Event 11% No E,F,RPT E,F,RPT 89% Figure 13. Incident Analysis by Gravity 4. CONCLUSIONS Since the commercial LNG import industry began in 1965, there have been no known instances of damage outside the LNG facility concerned nor of catastrophic LNG storage tank failure leading to loss of containment integrity. This study has confirmed these facts. A total of 246 incidents of releases of hazardous material, near misses and other incidents of concern have been reported and analysed in this comprehensive study of GIIGNL Member s facilities covering the 35 year period from 1965 to 2. The frequency of the reported incidents over the whole period is very low,.33 incidents per site-year. Incidents involving the release of hazardous material (hydrocarbons) are 68 % of all those reported. The gravity of the events reported is variable and only 11 % resulted in an explosion, fire or rapid phase transition. The study has been undertaken over two periods from 1965-1994 and from 1994-2 and it has therefore been possible to evaluate safety performance improvements. There is a trend towards a decrease in the relative number of events where significant PS7-3.16

quantities of hydrocarbon have been released. This demonstrates better mitigation measures and operational procedures at LNG import facilities. The frequency of events reported has increased in some areas. Whilst this is likely to be influenced by the introduction of formal reporting procedures, factors such as increased throughput and ageing of equipment could also be responsible. The analysis shows that incidents during LNG unloading functions are the most frequent. These incidents are approximately 1 times more likely per operating hour than for any other function but do not contribute excessively to the total number because LNG unloading is a relatively infrequent operation at LNG import facilities. The severity of incidents during LNG unloading has deceased dramatically recently, which shows that LNG Importers have been successful in reducing this risk. The relative number of incidents where the cause was considered to be outside of the influence of the terminal has increased. Nine out of the 2 incidents are related to shipping, 5 to adverse weather (storms, lightning), 2 to external interference (digging up pipelines), an electrical supply interruption, a vehicle impact, an adverse pressure build up in a tank and a World War I bomb discovered during excavation works near an LNG site. Many fewer incidents (19 %) occur during the night than during daytime hours. The LNG Importer s Group (GIIGNL) believes that it has developed a comprehensive database of incidents that will be useful for the future design and development of safe LNG facilities. GIIGNL is committed both to improving further the reporting of incidents and to maintaining its database up-to-date for the general good of the LNG Industry. PS7-3.17