Graham E. Broadbent, Tony Williams
Pipelines Authority of South Australia
Glenside, South Australia
An optimization system for operation of the Moomba Adelaide gas-pipeline network has been developed by the operator, Pipelines Authority of South Australia (PASA), through use of a steady-state model to predict pipeline compressor configurations and setpoints.
The system has allowed PASA greater accuracy in operations of the pipeline network even when subject to highly transient flows. A concluding article provides a detailed description of the model used.
PASA NETWORK
The Pipelines Authority of South Australia is the owner and operator of the natural gas-pipeline system supplying Adelaide and areas in South Australia (Fig, 1).
Gas sales are monitored by computer via meter stations equipped with orifice or rotary meters. The computer system essentially generates gas billing in energy terms on an hourly basis.
PASA operates two maintenance depots, one at suburban Dry Creek and the other at Peterborough, 150 miles north of Adelaide.
A permanently manned control center adjacent to the Glenside head office enables pipeline-control staff to monitor and control the pipeline systems continuously to ensure adequate sales gas and safe, efficient operation of facilities.
Compressor units are remotely controlled and monitored from the control center through telemetry and radio systems maintained by PASA.
PASA's pipeline system consists of the 450 miles of 22 in. pipeline from Moomba to Adelaide and a number of lateral pipelines totaling 170 miles.
The main pipeline has seven compressor stations, each consisting of two compressor units rated at approximately 4,000 hp/unit and at a regular spacing of 60 miles.
Gas is supplied to two major consumers at the end of the pipeline.
The Electricity Trust of South Australia (ETSA) consumes about 60% of the transported gas for power generation. Most of the remainder is distributed by the South Australian Gas Co. (Sagasco) for domestic and industrial use.
Because of the high domestic nature of the demand, large and somewhat unpredictable variations can occur on an hourly, daily, and seasonal basis. There is little or no contractual limitation on the customers for these demands although a penalty applies on exceeding a maximum daily quantity.
Because of the essential state-utility nature of PASA and the major customers, no attempt is made to restrict sales if operationally possible.
There are no on-line storage facilities or LNG plants, a condition that results in the transients and demand variations being taken up, to a small degree, by linepack but essentially by compressor horsepower availability.
Fig. 2 illustrates typical time-varying load on daily and weekly bases. Seasonal variations may give rise-to sales of 170-320 MMcfd.
Sales to the Electricity Trust of South Australia in midsummer can be very high because of the significant air conditioning component of the Trust's power demand, and similarly peak sales in midwinter are due to gas and electric domestic heating.
OPTIMIZATION NEED, AIMS
The PASA pipeline network is remotely monitored and controlled by the control center at Glenside in, Adelaide.
The areas of monitoring include compressor-station performance, all gas accounting, pipeline operation, and microwave/radio link supervision. This is achieved with several scada (supervisory, control, and data acquisition) systems which feed into a central host scada computer-system network.
Analog data, such as pressures, temperatures, speeds, flows, and vibrations, are updated on request and each minute and stored in the system's data base. Response time is less than 1 sec.
Alarms and statuses are scanned for change about eight times a minute and likewise stored.
These data are archived each minute and available for a month for detailed analysis. Hourly data are permanently archived for other historic analysis.
The overall responsibility for gas accounting lies with the control center and includes supply and customer billing, PASA use (fuel gas, blow downs, relief valve operations, etc.), and an item termed "gas unaccounted."
The initial need for pipeline optimization was not apparent other than a concern that perhaps operation of the facilities was less efficient than it might have been. The real problem appeared to be in the areas of accountability, repeatability, and operational consistency.
The need for tighter overall control was triggered to some degree by budget requirements and accountability.
In the area of compressor fuel use alone, very little repeatability could be noted in any performance measurement attempted, even in a simple example of fuel gas used per quantity of transported gas.
There were many contributing variables, including the unpredictable method of pipeline operation, the calculation of fuel gas itself, and the subsequent revelation that the maintenance scheduling of the compressors generally dictated the pipeline operation.
Analysis of any aspects of pipeline operation, compressor operation, and gas accounting would obviously require a much higher level of consistency for the decision-making process to be achieved.
The program set out to achieve the following aims:
- To discover if any improvements could be made in the operation and control of the pipeline
- To use existing facilities (of which there were many) without significant operational or capital expenditure
- To achieve significant and tangible levels of accountability and performance measure on operational decisions for later use in optimization strategy
- To develop tools and devices as necessary
- To develop an operating strategy in which the techniques used could be consistent from an operator's point of view, with observable and measurable aims, but one in which the operational parameters could be adjusted with little observable impact on the operating philosophy
- To keep it simple.
ESTABLISHING PARAMETERS
In the early stages of the development process, a list of variables was established which affected the overall decision making and accountability processes. These were analyzed with regard to prediction, accuracy, necessity, impact, control parameters, response time, and accountability and performance measures.
The initial list of parameters included hourly demands, transient weekly demands, unpredictable or poorly forecast loads, and other responses to similar situations.
They also included compressor and pipeline maintenance scheduling, little or no operational performance measure, variation in individual compressor performance, irregular and somewhat unpredictable supply pressure, and operational effects on compressor maintenance requirements.
The approach taken was to look at each area in turn and discover in the first place what decisions were being made, what information was being used, and if the decisions were consistent.
This step was followed in many of the cases with an operating instruction and considerable monitoring of the new instruction's performance and adherence.
In the background always lingered the thought that these "expert" instructions could later form the basis of an automated system or at least be part of some flow chart analysis to direct operational decision making.
PARAMETER ANALYSIS
The first aspect of the operation to be addressed was the seemingly inconsistent decisions that were being made. This situation resulted not only from the normal problems associated with a continuous shift system but also from the hierarchal reporting structure of the department.
Areas of responsibility were redefined to increase accountability in the development process.
For operational decision making, a simple operating philosophy was stated which had a sound basis in normal pipeline operating practice. The philosophy consisted of the following eight components:
- Maintain the input pressure as high as possible.
This was power that the suppliers could put into the line and would cost PASA nothing. There were essentially no contract provisions defining the supply pressure.
- The delivery pressure to our customers was to be the least possible without jeopardizing our ability to supply their somewhat unpredictable requirements.
To that end a series of transient studies was performed to establish an operating pressure for the last "required" compressor station on the pipeline which, provided that pressure could be maintained, would cover the transients, and allow some reserve for unpredictable loads.
A series of graphs of delivery pressure vs. time over a few days provided a valid tool for at least some predictability.
Results of the studies also gave a series of response times in case we needed to revert to the "panic" mode of operation. At the same time, we needed to assure customers that system deliverability was being maintained.
- Because customer requirements needed to be more predictable, a greater understanding of their systems was important. Again, contractually there are no requirements for the customers to supply daily sales estimates or, even if they did, to adhere in any way to them.
There was also no need for them to provide notification of some imminent significant change to current demands, although if an increase was needed one customer at least tended to ask. It was apparent that greater customer liaison and understanding was needed.
- Supply demands made on the producers were to become accurate and updated regularly. This aspect was hoped to be one of the benefits the suppliers would see.
No contractual limits were placed on PASA except for a maximum daily quantity (MDQ). Clarification of that definition of MDQ needed to be generated; i.e., was it a maximum hourly rate or an expected daily quantity?
- Compressor-station selection and unit selection were to be more predictable and accountable. Some simple guidelines had to be established to allow prompt and meaningful selection of compressor units.
- Maintenance of compressors was to be scheduled through the control center and a weekly maintenance plan ideally forecast 23 weeks in advance.
Wherever possible this maintenance schedule would be adhered to because considerable effort was put into planning which stations were likely to be needed. Urgent maintenance or unit failure would be responded to and ordered by priority according to needs, safety taking top priority.
- Performance measures had to be established and their validity regularly assessed. Reporting on those measures and responses to problems had to be prompt. For example:
- Minimize fuel gas vs. transported gas on a weekly basis.
- Minimize number of operationally required starts and stops of compressors.
- Minimize and account for the number of operational decisions which led directly to inadvertent compressor shutdowns.
- Minimize the impact and occurrence of high transients on the producers.
- Ensure that gas deliverability was maintained.
- Monitor and assess the impact of the operational changes on various sections within the organization and on customers and suppliers. This area was to include budgeting, maintenance, contract administration, project justification, and so forth.
The following sections focus on a number of these areas and discuss the observations made and techniques used to rationalize and stabilize the operating strategy.
STATION, UNIT SELECTION
As indicated, station and unit selection was essentially driven by maintenance requirements.
Little or no level of priority or absolute need was generated. This approach often meant that several other units would be run in place of one which was taken off for maintenance.
An attempt had been made to schedule maintenance on weekends but on a strict alternate-stations strategy.
This procedure was reviewed and one put in place to plan and schedule maintenance a week in advance. Considerably more accountability was placed on the schedule and on personnel stating what was really needed, as opposed to what was wanted.
The major problem on unit and station selection still had to be addressed. From the philosophy generated on supply and delivery constraints, that is maximizing supply pressure (nominally 950 psi in the first assessment) and operating with a fixed discharge pressure at the last operating compressor station, a statement on the intermediate sections became necessary.
Based on the best available information regarding the most efficient compressor-unit operation, a selection was made on the basis that units would be run at maximum speed.
The units were selected with a basic graphic technique of tracing a flow line (based on sales predictions) which gave a pressure drop in the pipeline and essentially a suction pressure at the next station. A simple calculation based on expected differential across the station established if this unit could or should be run.
This process was continued for each section of pipe until the last station (as indicated by the transient studies) was able to maintain the required discharge pressure. The pressure drops in the sections at different flow rates were established from historic data analysis, with transient analysis programs used to establish roughness and other parameters.
This technique and an associated on-line pressure profile display was the start of the many steps to more predictable and efficient pipeline operation. A basic philosophy was in place and a tool for the more predictable pipeline operation was able to be used and assessed.
It was important to establish an early mechanism of accountability. Therefore, fuel gas vs. transported gas trending was established on a weekly basis and a comparative tally of operationally (as opposed to maintenance) required starts and stops of compressor units.
The former measure of fuel economy was directed at senior management to support the considerable effort being directed at the exercise. The anticipated reduction in compressor starts and stops was aimed at the compressor maintenance personnel since they could see that as a benefit in the improved reliability of the units.
It wasn't long before the sheet of paper with the pressure gradients and flow lines was replaced by a personal computer program which did the steady-state calculations at request and became a significant part of the predictability of the operation.
The operation of this program and the subsequent optimization program will be described in a following article.
COMPRESSOR OPERATING CRITERIA
Fig. 3 shows the compressor configuration in the PASA pipeline network.
The analysis process in the area of compressor operations led to querying the reasons that, during compressor configuration changes, the system was so vulnerable to shutting down a station and subsequently creating the "domino effect" (several stations shutting down further upstream of one which had been stopped). This occurred particularly on weekends when pipeline flow rates were lower.
The cause in general of the domino effect was high unit discharge gas temperature due essentially to high recycle of the unit slowing down to unloading speed and subsequently oscillating on and off-line.
The nature of this problem became obvious when the pressure profile was viewed historically for the few hours before the initial compressor shutdown. It was found that in many cases the need for the initial configuration change was not determined early enough to avoid the inevitable.
This domino effect often had a severe effect on the supplier even to the point of his having to flare processed gas in large quantities if advance warning was not given. This resulted in the supplier's reluctance to operate into our pipeline at the higher pressure because it reduced the time it allowed him to respond to the situation.
Procedures were updated to ensure as much prior warning of the situation as possible and to quantify the likely effect on his flow rates.
At the same time, methods were devised to predict vulnerable situations and control them.
The PC program was important in analyzing worst case scenarios with regard to flow changes and the likelihood of a domino effect and the most vulnerable spot.
The problems of the domino effect or, at best, the units' loading and unloading had four causes:
- A compressor being shut down while running at full power, thus creating a huge flow transient upstream.
- An inappropriate compressor configuration for the expected pipeline flows.
- Compressor units running with high unit recycle and being extremely vulnerable to reduced pipeline flow (recycle gas not cooled).
- The unit being allowed to operate at a low speed, close to the unloading speed.
Each of these aspects was addressed.
Before the compressor was shut down, the unit was slowly ramped back close to its unloading speed. At first this was done by the operator issuing a series of setpoint changes, but these were often at irregular intervals.
A program subsequently was written for the scada system to ramp a unit back automatically at a slow rate to avoid excessive recycle on that unit and to reduce the size of the transient in the line. Significant sophistication was later added to this program.
The selection of compressor configurations was already being assisted by the program, but units with better recycle characteristics could be selected for critical or vulnerable situations.
FUEL VS. HP
We needed to validate one of the first stated performance criteria, that the compressors run most efficiently at maximum power.
Thorough investigation of this aspect revealed it to be misleading at best. It depended on the definition of compressor maximum speed: Was the interpretation to be process compressor? engine compressor package? compressor speed? engine speed? And did the statement allow for the compressor-recycle characteristics?
It was decided that our performance measure of a compressor should be dollars of fuel and effective station transmitted horsepower.
Somewhat surprisingly with the many variables of pipe losses, ambient temperature variations, compressor efficiency, engine efficiency, and so forth, the whole station characteristic was essentially linear when fuel was plotted against transmitted horsepower over the full speed range of the compressor.
The problem of operating the compressors at low speed, or at least having pipeline conditions which required it, was somewhat reduced by better overall predictability and control. The scada compressor-setpoint ramp program was modified to give some minimum level of speed control.
A speed was selected that was slightly above the unloading speed. It was used in situations when vulnerable operation existed.
It was recognized at this time that the compressor station setpoint control system lacked the requirements of smooth control. The discharge-pressure setpoint controller had significant hysteresis and was insensitive to small adjustments.
There was no direct speed control and this seemed necessary for controlling tight situations. It appeared that minimum speed control would be an advantage to avoid the loading/unloading problem of the units.
The station PLC [programmable logic controller], only used when both units were required to be run at the station (infrequent by now because of lower flows and improved operating technique), was reprogrammed with minor hardware modification to:
- Include the station discharge pressure control. This gave far more sensitivity and more stable control and improved ramp scheduling.
- Allow for individual unit speed control. This was seen as a benefit by the compressor maintenance section to allow units to be controlled remotely with respect to reducing vibrations or performing remote performance checks.
- Include a minimum-speed control locally so that a unit being controlled on discharge-pressure setpoint was not allowed to be driven into the load/unload cycle except in the normal start-up and shutdown. The remote speed setpoints can override the minimum-speed control, and pipeline overpressurization is protected by the addition of a minimum speed/maximum pressure shutdown.
The minimum-speed addition has proved most significant in the control of the pipeline.
By this time, significant fuel savings were being made and the number of operationally required compressor stops and starts had been reduced as well as the number of situations of compressor shutdowns resulting from high recycle or the unit unloading problem.
There were still problems, however, because of the transient nature of the sales and the sales unpredictability.
SALES PREDICTIONS
The customers were approached regarding their sales trends and predictions. An improvement was made in the communication of known changes due to plant failure or maintenance scheduling.
But the basic cause of the transients and sales unpredictability was the domestic nature of the customers. The base industrial load was not the most significant contributor. Typical sales trends on hourly and weekly bases are shown in Fig. 2.
Considerable analytical effort was put into association of ambient temperature, day/night variations, etc., against estimated temperatures and estimated sales.
Although some correlations were found, the whole process became lost in a mountain of equations and statistical analysis. It surpassed the realm of comprehension. The nature of the South Australian domestic customer was difficult to model.
An alternative approach was taken, somewhat similar to the compressor model, and that was to look at the flow data and try to find some characteristic of the data to use as an estimate.
The initial idea came from one of the operators who indicated a good rule of thumb he had used was that 42% of the sales for the day had been delivered at 4 p.m. when he started his shift.
His concept was put to the test on several years of data. With some tuning and recognition that Mondays, Fridays, and weekends were different and that seasonal variations had to be considered, a model which could predict daily sales for the current day became available. The model was of course extended to predict sales at any time of the day.
Its accuracy is not always absolute (usually 4%) but the early awareness of significant variation to expected customer estimates is powerful knowledge.
Another traditional operating method is to maximize linepack. This might imply an optimum operation under some circumstances.
But we have found that in our situation of a highly compressed line and large daily transients, the overall linepack can be a very misleading performance measure.
It is very important, however, during the transition between compressor configurations, to match the exit linepack from one configuration to the entry linepack of the new one.
This particularly applies in the region of the configuration change.
If not done properly, it can inflict significant transients on the adjacent stations and on the supplier if within that area.
TRANSIENTS IN SALES PROFILE
Another aspect to be addressed is the variable nature of the pipeline operation resulting from the transient-flow demand.
First elements of control theory taught us that steady operation is more efficient than transient or cyclic. Because steady-state analysis is also much easier to perform, consideration was given as to how we might make transients into steady state.
There are no reservoirs or peak-load LNG plants to draw on, and the pipeline linepack is not significantly high enough to be effective. The only available excess parameter was compressor horsepower.
It was compressor horsepower that was creating huge transients in the first place when units were stopped and started at full power. Perhaps we had tamed the beast sufficiently to put the control of horsepower to a better advantage.
At this time also the gas suppliers were beginning to see significant advantages in the predictability of the pipeline operation and were beginning to do their own "optimization" of field and plant. They began asking for steady and predictable gas demands, and thus began the steady-state optimization and operation of the pipeline. The idea was simple: operate the pipeline at steady flows to the last operating compressor station and soak up the flow transient with horsepower variation at that station and to some degree the linepack downstream of that station.
The first attempt at this was quite disconcerting (although predictable as the transient studies had shown). The last station was controlled essentially on a suction pressure (via the discharge-pressure controller) which stopped the flow transient from going upstream.
As the delivery flows during the day were increasing. however, and the delivery pressures decreasing, the discharge pressure of the compressor station was also decreasing.
Visually on the pipeline-pressure profile this was a very disconcerting phenomenon and scepticism abounded.
The mathematics held true, however, and another step was made in the stable, predictable, and efficient operation of the pipeline.
Copyright 1990 Oil & Gas Journal. All Rights Reserved.
