WATER INJECTED LM 1600 INSTALLATION AND OPERATING EXPERIENCE

determine and These tests showed that the expected emission reductions had been achieved and allowed the optimum water injection flow rate to be accurately established for a range of operating conditions.


C.^1ulu_ M" 1
In 1992 Alberta Natural Gas Company Ltd ("ANG") installed a General Electric LM 1600 gas turbine at its Moyie compressor station in southeastern British Columbia, Canada. The unit was packaged by Dresser Rand with a General Electric supplied power turbine. To comply with provincial emissions permitting requirements, and in response to growing environmental concerns, the gas turbine was installed with water injection for exhaust stack NO, control. Water was obtained from an underground well and, after treatment to bring the water to a condition specified by General Electric, was injected into the combustion chamber of the gas turbine. After commissioning, extensive on-site testing was conducted to determine the emissions from the unit using three different techniques, at a variety of load and water injection rates. These tests showed that the expected emission reductions had been achieved and allowed the optimum water injection flow rate to be accurately established for a range of operating conditions. BACKGROUND ANG owns and operates a major natural gas transmission pipeline passing through the southeastern comer of British Columbia. This pipeline is 170 km (106 miles) long and is part of the Alberta to California gas transmission system used for the export of Canadian produced natural gas. It forms the interconnection between the NOVA system in Alberta and the Pacific Gas Transmission system in Idaho. When initially installed and commissioned in 1961, the system utilized reciprocating compressors totalling 10.4 MW (14,000 HP) at a single compressor station. In 1966 the first Avon gas turbine driven Coberra package was installed, followed by five more similar units in following years, increasing the number of stations to three. In 1981, the reciprocating compressors were replaced by a Solar Mars driven compressor set. In 1991, one of the ageing Avon units at the Elko compressor station was replaced with an LM 1600 driven compressor set, and in 1992 a second similar replacement was undertaken at the Moyie station.
Following the recent expansion of the system, which was completed in 1993, the ANG system now comprises of two parallel loops of 914 mm (36 inch) and 1067 mm (42 inch) pipe, ten gas turbine driven compressor units, totalling over 123 MW (164,000 HP), and an export meter station (see figure 1). The majority of the pipeline loops are owned by Foothills (South BC) Ltd. The system is designed to transport a daily contract volume of 69.4 million cubic metres per day (2,450 million scf/d).

REGULATORY REQUIREMENTS
Federal and Provincial initiatives have produced guidelines and legislation that restrict the amount of contaminants that can be emitted into the atmosphere, and which dictate emissions monitoring requirements. Federal guidelines have been produced under the jurisdiction of the Canadian Council of Ministers of the Environment ("CCME"), a forum which combines the 13 member governments working as partners in developing nationally consistent environmental standards. These guidelines, which have been developed over a number of years through a multi-stakeholder consultation process, were issued in December, 1992, as part of the Canadian NO I/VOC Management Plan, and apply to new combustion turbines which receive final provincial orregional regulatory environmental approval to construct on or after 30 November 1994. For the two-year period of implementation of the guideline, the emission allowances for 3 -20 MW sized non-peaking units are relaxed.
In the Province of British Columbia, the Ministry of Environment, Lands and Parks ("BCMELP") has developed a best available control technology (`%ACT ') policy, and based on this produced an emissions criteria for gas turbines. These regulations were issued in December, 1992, and are administered by the local waste management branch. The emission criteria are effective from the date of issue, and compliance is necessary to obtain an emissions permit for the installation.
For gas turbines burning natural gas as a fuel, the BCMELP emission criteria for gas turbines are summarized in   were still in an early stage of development and did not require the g as turbine to be installed with emissions reduction. However, by the time it became necessary to seek approval for the second unit replacement for installation at the Moyie compressor station, developmentof the new regulations was well advanced, andBCMELP expressed concern regarding the 140 ppm emission level of the LM 1600. In addition, ANG was already starting to work with BCMELP on the preparation of an Environmental Impact Assessment (EIA) for the upcoming pipeline expansion project, which would entail the installation of an additional 3 new gas turbines, and public consultation also was underway on a proposed 130 MW cogeneration plant at the Crowsnest compressor station. Consequently, because BCMELP would not allow uncontrolled emissions, and were not prepared to wait until 1995/6 forthe dry lowNO solution, the secondreplacement unit was ordered for installation with water injection to reduce the NO, emissions to 42 ppm. CO emissions were predicted to increase from 25 ppm to 57 ppm as a result of the water injection.
Installation of the LM 1600 with water injection involved a number of interesting design challenges, including water sourcing, water treatment, site layout, disposal of waste from the water treatment plant, and extending the exhaust stack to permit mass flow measurement.
Due to the critical nature of the station in maintaining the required contract pressure at the Canadian -United States border, the LM 1600 was installed in a new building so that the Avon Coberra unit would not have to be retired until the new unit was fully commissioned. The layout of the station is depicted in figure 2. The new installation comprised of three new buildings: • compressor building, housing the gas turbine and compressor package, complete with air inlet filter and silencer, exhaust silencer, oil consoles and water injection skid; • unit auxiliary building, housing unit control panel, RLU, and unit motor control centres; • water treatment building. In addition to these buildings, a new water well was drilled to provide water for the water injection to supplement the well already on site that provided water for general use. Precast concrete trenching was installed for the wiring to connect the three buildings listed above with the existing utility building and station control room.
The LM 1600 requires up to 42 litres per minute (11 USgpm) for water injection. The water treatment plant was designed specifically for the water expected from the wells located on the site, based on several samples that were taken, and to meet the water quality specification provided by General Electric (table 2).

Limit
Test Method Total matter, ppm, max 5 ASTM D1888 Dissolved matter, ppm, max 3 ASTM D1888 pH (1) Note: (1) pH and/or conductivity shall be measured when water is free of carbon dioxide.  Figure 3. Water is fed to the plant from the submerged well pump and is then filtered through three sets of filters -multi-media, activated carbon, and cartridge filters. Surge tanks regulate the flow of water from the well. After filtration the water is conditioned to  remove calcium (softened), and then sterilized in an ultra-violet unit. Next the water is heated to 25°C (77°F) and is fed to the reverse osmosis ("RO") unit. The RO unit operates at 2965 kPa (430 psig) and produces a continuous waste stream of 27.6 litres perminute (7.3 USgpm). The water is then stored in a 9084 litre (2,400/US gallon) atmospheric storage tank. Pumps, controlled from the injection skid located in the compressor building, feed the water through a series of portable exchange deionizers comprising of cation, anion and mixed bed filters, and then on to the compressor building for injection. The portable exchange deionizers are removed periodically for off-site regeneration. Water quality is continuously monitored for hardness, silica, and resistivity.
Dealing with the waste from the water treatment package presented the greatest technical challenge, and several options were examined, in consultation with BCMELP, prior to selecting a shallow well for the disposal of the waste water. These included surface discharge, rock pit and disposal into a stream in the vicinity of the plant. The total waste water stream amounts to approximately 38 litres per minute (10 USgpm), and is disposed of through a shallow disposal well. This waste, which comes from the slip-stream from the reverse osmosis unit and from the backflush of the inlet filters and water softener, is comprised of concentrated solids removed from the feed water. The treated water is pumped to the water injection skid that was supplied with the gas turbine package. This skid is located in the compressor building beneath the gas turbine exhaust stack. It comprises of a small storage tank, injection pump, filter and flow meter. The water then goes to the gas turbine where it is injected into the combustor via a second set of nozzles. For water injection the LM 1600 uses a fuel manifold and nozzles designed for duel fuel service (gas and liquid fuel). The liquid fuel nozzle is used for the water.
The unit control panel is programmed to provide a water schedule giving a fixed fuel to water ratio utilizing a feed-back signal from the water flow meter. The system is designed to allow adjustment of the water-to-fuel ratio from 0 to 1.1, and upsets such as loss of water to the turbine will cause an alarm, but do not shut down the unit. The water system can be re-started with the turbine on line.
In order to conduct annual emissions tests BCMELP required that a stack extension be installed to enable exhaust duct traverses to be conducted. This extension, depicted in figure 4, was to have an effective upstream and downstream unobstructed length of 2.5 diameters and 0.5 diameters respectively relative to the traverse plane. Five traverse nozzles were required, based on effective duct diameter, to enable a traverse grid to be established. These requirements were met by installing a permanent 3.05 m (10 ft) extension, together with a removable temporary extension. A platform and ladder was also installed.

CONSTRUCTION AND COMMISSIONING
Construction of the installation commenced in February, 1992, with the gas turbine arriving in April. Installation continued throughout the summer, and commissioning of the unit was completed in August. The water treatment system and injection unit, however, were not commissioned at that time due to the lack of a waste disposal permit, and permission was obtained from BCMELP to operate the unit dry. The disposal permit was received in November and commissioning of the water treatment plant was completed in readiness for the extensive emissions testing planned for December. This commissioning was somewhat hampered by initial control difficulties with the water injection skid, and the control philosophy of the gas turbine had to be modified to prevent unnecessary unit shutdowns that were occurring as aresult water supply interruptions. These involved keeping the unit running at full load even when the water is abruptly shut off. We found it to be unnecessary to trip the machine or step-to-idle in the event of loss of water.
Modifications were also required to the water treatment package in order to obtain a suitable time period between off-site regeneration of the deionization beds. This was achieved by installing a caustic storage and injection system to convert dissolved CO 2 in the water to bicarbonate alkalinity, which is then removed in the RO unit. Apart from the the controls and water quality problems noted above, the only other significant problem encountered involved inability to adequately monitor and control the water-to-fuel ratio. This was resolved by installing a water flow meter on the injection skid to allow measurement of the the water flow, and feed-back control of the water-to-fuel ratio. Once the intial problems had been resolved,  Figure 6: CO Emissions unit availability increased to an acceptable level. Water injection has had no significant impact to date on engine performance, reliability or component life. Reliability of the water treatment system, however, still remains less than ideal, and the unit operates on occasion without water injection. All such occurances must be reported to BCMELP. When the unit is running on T48 (exhaust gas temperature) limit, the use of water injection increases the power output by up to 5 percent.

Water/Fuel Ratio
The disposal permitreceived from the BCMELP WasteManagement branch contained requirements for monitoring of the disposal stream on a quarterly basis, and a well and adjacent creek (upstream and downstream of the station) monthly. The well and creek were also to be tested for a six month period prior to commissioning of the new water treatment system. This was to enable BCMELP to confirm that the waste stream composition was as predicted, and to monitor what impact, if any, the disposal well would have on the creek. To date, no adverse environmental have been observed.

EMISSIONS TESTING
In order to establish the emission levels of the unit, to comply with permit requirements to conduct annual testing, and to demonstrate to BCMELP alternative methods of engine exhaust mass flow measurement, extensive emissions testing was conducted in December 1992. This testing was conducted over a number of days by Cubix Corporation from Texas, and engine measurements and calculations were performed by General Electric from Cincinnati. A mobile laboratory was installed on site to conduct on-line gas analysis. BCMELP were also on site and witnessed the testing.
Three types of tests were conducted at a range of water injection rates ranging from zero to over 100% as follows: • mass flow was calculated using measurements that were taken during traverses with a picot tube. This was inserted into the ports in the exhaust duct measuring section. Exhaust concentrations were also measured, thus enabling exhaust mass flow to be calculated. Due to the difficulty and the time involved in obtaining these measurements, this method was used for only five of the test points; fuel flow measurements, combined with exhaust gas concentrations of both NO and CO, were used to perform stoichiometric calculations to determine exhaust gas emissions; calculations were performed by General Electric, based on engine parameters, to determine mass flows and emission rates.
A test matrix was established to record exhaust emissions and mass flow rates over a wide range of operating conditions as shown in

Engine Power
Idle X 20% X 30% X X X 50% X X X 70% *X X X X X X X 90% X X X X X X X X 100%+ *X X *X X X X *X + 100% load = 11,511 kW (15,430 hp) * Pitot traverses conducted at these conditions  Test methods were approved by BCMELP prior to testing, and sampling and analysis procedures used conformed to EPA methods. Figure 5 shows NO z emissions versus water to fuel ratio, and figure  6 shows CO emission against water to fuel ratio, both for various loads, based on stoichiometric calculations. Both NO. and CO are shown in ppmv corrected to 15% oxygen content. It can be clearly seen that NO. levels decrease, and that CO levels increase, with increasing water to fuel ratios. The optimum water to fuel ratio to keep both NO. and CO levels below the BCMELP permitted values is 0.6 at all load levels.
Differences were observed between mass flow calculated using exhaust gas stoichiometry, and from calculations based on stack pitot tube velocity traverses. These differences were eventually reconciled when "effective" stack area was used instead of an "actual" stack area, using an area coefficient of 0.95.
Calculations were also conducted by General Electric using the gas turbine design "cycle deck". Gas turbine operating parameters were measured and then fed into the cycle deck computer program. The program calculated exhaust mass flow rates and exhaust gas composition that correlated extremely well with measured readings. Figure 7 compares the results of the three test methods. Based on these results, for the purposes of emission inventories, process evaluation, and air modelling, BCMELP accepted the mathematical model (cycle deck) for flow calculations, and EPA method 19 (of 40 CFR 60) for concentration measurements (stoichiometric calculations). For compliance testing, if required, flow measurement would have to be conducted using either 40 CFR 60 or British Columbia methods. Future testing will be conducted in an attempt to establish the General Electric computer model as a method acceptable to BCMELP for concentration measurement. This would provide a simple continuous emissions monitoring technique.

FUTURE CONTROLS
Three additional LM 1600 driven compressor units were installed by ANG in 1993, also with water injection. Two of these units were installed at the Crowsnest compressor station, and one at Moyie. Two additional units are currently proposed for a 1995 installation. It is anticipated that these will be installed without water injection, but be purchased with controls to make the units suitable for dry low NO, conversion as soon as kits become commercially available.

CONCLUSION
Water injection has proven to be an effective method of reducing NOxemissions on gas turbines, with the added benefit of power enhancement under certain ambient conditions. However, significant problems were encountered, including obtaining an adequate water source, producing water of suitable quality on a consistent basis, and disposing of the waste water from the water treatment plant. Installation of the water treatment and injection facilities entailed additonal capital, commissioning and operating costs, and resulted in an increase in CO emmissions.
Testing established the correct fuel to water ratio for optimum operation of the unit, established alternative methods for exhaust mass flow measurement, and provided the first steps for continuous emissions monitoring based solely on measured engine parameters.