SYSTEM OPTIMIZATION-CONCLUSION FLEXIBILITY, SAVINGS CHIEF RETURNS OF NEW PIPELINE SYSTEM

Graham E. Broadbent, Tony Williams
Pipelines Authority of South Australia
Glenside, South Australia

Pipelines Authority of South Australia (PASA) has developed an optimization system for operation of the Moomba Adelaide gas-pipeline network.

The system employs a steady-state model to predict pipeline compressor configurations and setpoints.

The system allowed PASA greater accuracy in operations of the pipeline network even when subject to highly transient flows. The initial article in this series on the system covered strategies employed in the system's development.

This concluding article provides a detailed description of the system.

EFFECTS OF TRANSIENTS

Why can steady operation of the pipeline and compressors be more efficient?

Fig. 1 shows a relative variation of frictional loss associated with a sinusoidal transient flow.

The conservative transient of 10% of average flow is typical of those experienced along the pipeline if the line is operated on discharge-pressure control and to some degree maximum-power control.

The trends are plotted as relative to the steady-state parameters of constant flow, constant pipeline pressure, and the associated constant frictional loss. This of course is the ideal situation with no temperature effects or gas quality changes.

The flow swings sinusoidally about the steady-state value, and the average pressure takes essentially an equal but opposite reaction. Observation of the relative variation in frictional loss over the transient period shows a higher peak loss at the high flow part of the transient than the peak "gain" during the low-flow period.

A simple integration over the full period indicates an overall increase in frictional loss of about 1.4%, in this example, over the loss associated with the same average flow rate at steady conditions.

This occurs essentially because of the squared relationship of frictional loss to flow rate. This extra frictional loss must be made up by extra power at the next compressor station, and so on for each section.

The effect of transient flow on the operating point of the process compressor is described in the following subsections.

Most of the influences of the transient flow and transient inlet pressure are cumulative and tend to make situations even less efficient than they may be if considered individually. The effects are also different, depending on the mode of control of the compressor and pipeline system.

Maximum pipeline pressure. For situations in which the compressor is controlling at maximum pipeline pressure, the operating point on the compressor wheel map is influenced by changes in the head as a result of changes in suction pressure (due to change in pressure drop in the upstream section).
The operating point in the X direction, the inlet flow, is influenced in a cumulative way because of an increase in suction pressure and decrease in pipeline flow, and vice versa. This gives rise to a movement in the operating point (Fig. 2a) which is somewhat elliptical and can take the unit into the less efficient areas of operation as shown by the efficiency lines.
Intuitively, operation across these boundaries must be less efficient than if the compressor were to operate at a fixed point in the maximum-efficiency area. An assumption made here is that the power plant (in our case a gas-turbine engine) is efficiently matched to the compressor. This is a reasonable assumption.
In a limiting case the operating point may be driven to the left so much as to impinge on the recycle or surge control which also wastes considerable power.
Maximum power/maximum pressure. The next stage from that just described may be a situation in which there is insufficient engine power to maintain maximum pipeline pressure (or the discharge pressure setpoint),
In this case the compressor operating point moves as in Fig. 2b where for some period of the transient the compressor will control on maximum speed.
This tends to reduce the variation in operating point but the discussion of maximum pipeline pressure still applies. And an additional influence in this case is that for some period of time when the unit is controlling on maximum speed rather than the required pressure, the downstream pipeline efficiency is reduced to a small degree.
Maximum power. The next stage when the pipeline operation forces the unit always to operate at maximum speed is shown in Fig. 2c. In this case the operating point "floats" on the pipeline conditions and oscillates back and forth on the maximum-speed line again, possibly across into the lower efficiency regions.

Conditions of pipeline flow and compressor configuration can move all these scenarios significantly to the left or right of the maximum-efficiency region and can result in consistently high compressor recycle, or significant reduction in head in the other extreme, giving rise to poor pipeline operating efficiency in the downstream section.

As implied by the previous discussions, the ideal operating point for the compressor is a constant one in the most efficient region. In real terms it's at the high-speed point in the high-efficiency area as a result of the engine characteristics.

It has been demonstrated here that steady operation of the pipeline can lead to higher efficiency and that relatively steady operation of the compressors in an efficient operating range can be individually more efficient.

But what is the overall effect in transportation efficiency of the process of using the last compressor station to knock out the transients? This mode of operation may in a global sense be quite inefficient.

Again with reference to the wheel map, assume that the steady-state operating point is that of Point A in Fig. 2d.

The effect of using this unit to soak up the transient with horsepower at steady inlet flow and fixed suction pressure is the path described by Point B.

This vertical path can still be selected to be reasonably within the more efficient region of the map provided that the unit can operate over the necessary horsepower range.

STEADY-STATE METHODOLOGY

The process of optimization of the gas-pipeline operations depended on a number of factors.

Various classical methodologies were available for assessing the operation of the gas pipeline: transient simulation, linear programming, steady-state simulation, a combinational approach, and relaxation techniques.

The methodology chosen was combinational optimization with steady-state theory (COST).

Optimization by the COST method involves simply costing all suitable combinations of pipeline compressor configurations at various discharge-pressure setpoints assuming a steady-state flow throughout the pipeline.

A pipeline-optimization program using the COST method was produced, originally written in Fortran 4 to run on a CMAD mainframe but was converted to Microsoft Fortran to allow the use on desktop computer. It is currently running on an NEC APC4 Powermate 1 with 640K, EGA, and math co-processor.

The validity of choosing any given compressor unit in a configuration is assessed at various stages throughout a computer run by comparison against a predetermined set of criteria.

Nested loops ensure that all possible compressor-unit combinations at varying setpoints are assessed. Certain assumptions are made during execution as well as prior to execution to omit unlikely configurational choices and improve the speed of the program.

It is clear that as the number of stations increases so do the combinations and therefore the resultant run times. However, unit unavailability, start/stop costs, preferred unit and experience, results in typical run times of 10 sec to 30 min on our existing equipment.

PROGRAM, DEVELOPMENT

The optimization program has 14 major modules.

Main: Chooses unit combination and provides setpoint stepping for each unit.
Optsuc: Calculates suction pressure based on the equation:
Ps = ( Pd2 - a.545.4.Qstd2) - 14.696
Tav + 491.67
where:
Ps = Suction pressure, psig
Pd = Absolute discharge pressure, psi
Qstd = Corrected flow, MMcfd
Tav = Average gas temperature, F.
a = Constant found by analysis of real data for a pipeline section approaching steady state
Interp: Linearly interpolates points between compressor speed lines
Optzft: Calculates gas compressibility based on the AGA NX-19 calculation
Optclc: Defines compressor unit operating envelope
Optcmp: Calculates discharge pressure, head, suction flow; determines unit recycle; and calculates transmitted power and fuel-gas usage.
Transmitted horsepower is given by the following equation:
const-H.Qsuct-SG.Psuct-FPV2
P =
Tsuct
where:
P = Transmitted power, hp
H = Station head, ft-lb/lb-mass
Qsuct = Suction flow, MMcfd
SG = Gas specific gravity
Psuct = Suction pressure, psig
FPV = Gas non-ideality constant
Tsuct = Suction temperature; or
const = Conversion constants and air density
Optcve: Interprets results from compressor interpolation, i.e., outside or bordering on known envelope
Optdtn: Data-entry routine defining unit availability, set points, and other operating criteria
Optfpn: Interpolates within and outside known fuel-gas horsepower relationship
Optmde: Interprets operational mode, i.e., no compressor Unit A, Unit B, or operation of units in series at a station
Optpak: Calculates linepack for a section between compressor stations
Optpip: Input general pipeline sectional properties, e.g., diameters, lengths, etc.
Optprn: Outputs configurations in order of fuel minimization
Optstr: Stores the 20 most fuel-efficient configurations.

The reason for such a simplistic approach was originally as a means of assessing the feasibility of producing a pipeline-optimization program with simplifications of present pipeline and compressor models.

Development time for such a program was seen to be short and results easily assessed by comparison with real time data.

The alternative of optimization by linear programming was considered but would have required some lead time for the theoretical aspects of linearization of nonlinear equations, handling of local maxima and minima, and mathematical stability to be learned and addressed.

Equations for pipe drop and compressor simulation during program development were simple so that run times could be kept to a minimum. Constants were grouped and calculated from real time data.

Separation and dependencies of the various factors within the equations occurred only after inconsistencies appeared or a refinement was dictated by operational circumstances.

During the development of the optimization model, traditional viewpoints had to be overcome and fresh approaches taken. Some of the aspects which were investigated included:

The concept of compressor station-head flow curves rather than unit-head flow data
The effect of ground temperature on pipeline pressure drops
The relationships between compressor horsepower and station fuel usage
Compressor-station recycle line definition
Compressor-station maximum-speed line definition
Efficiency and characteristics of reduced-speed running of gas turbines
Head-speed interpolation within known boundaries
Simulation of series compressor operation
Compressor-unit operational limitations
Gas-sales predictive methods
Transient analysis of the pipeline at various demand profiles
Optimum operating pressures for the delivery end of the pipeline
Calculation of natural gas fp and ft factors from specific gravity for fpr calculations.

The ease with which this analysis was performed can be attributed to the large amount of telemetered data available on line and archived, the installation of programs on-line to assess concepts or theories being investigated, and the dedicated logging of information by gas-control personnel during events of significance.

SHATTERED TRADITIONS

It was evident from the initial optimization runs that the traditional viewpoint of running the gas turbines at maximum speed was the first assumption to question.

The model invariably chose units running at less than maximum speed and at some flows indicated it was more efficient to run at reduced speed with one more unit than running all units full speed.

The concept of station operating curves rather than unit also was a deviation from normal modeling practice. This incorporates station inefficiencies or losses into the compressor model.

It could possibly-and correctly-be argued that "series operation accuracy" is affected, but the effect is minimal when compared to actual operating data.

Fuel-gas usage has been linked to various unit parameters such as speed, efficiency, and head.

Real time data indicate a well-defined linear relationship between fuel-gas usage and transmitted horsepower with possible plateauing effects for gas turbines with bleed valves rather than variable inlet guide vanes.

Each compressor station is assigned a characteristic array of head, actual flow, and speed defining an operational envelope for interpolation, departing from traditional head flow functions.

All data are taken from actual operating points and updated on a regular basis to account for changes in engine or compressor performances and station inefficiencies, leaking main line valves, for example.

COST EXAMPLE, SOLUTION

The COST method can be illustrated by the following example.

Consider a 124-mile natural-gas pipeline with a compressor station containing two gas-turbine compressor units at the 62-mile point which can be operated individually or in series.

Actual operating curves using discrete elements, as analyzed by the real time data-acquisition system, are shown in Fig. 3 for Units A and B, respectively. The relationship between transmitted horsepower and gas-turbine fuel usage rate is indicated by Fig. 4.

Transient-flow studies have indicated that, under the assumption of an average hourly flow, a steady-state pressure of 580 psig should be aimed for.

Given that minimum unit suction pressure is 435 psig, pipeline inlet pressure 870 psig, average gas temperature 59 F., average gas flow 190 MMcfd, and with unit series operation being unavailable, what are the 20 most fuel-efficient ways of operating the pipeline?

Optimization run restrictions can be summarized as:

Steady-state flow: 190 MMcfd

Inlet pressure: 870 psig

Outlet pressure: 580 psig

Minimum pressure: 435 psig

The solution should provide the most economical unit and its discharge-pressure setpoint.

The permutations begin by calculating the suction pressure at the compressor station based on the pressure drop equation previously mentioned. No units operating are assumed initially, and then the outlet pressure is calculated and compared with the minimum outlet-pressure restriction.

A failure to meet this restriction results in the abandonment of this configuration.

Unit A is then assumed as running (If the suction pressure criteria is met; otherwise, no valid configurations would result.) with a discharge pressure 10 psi higher than the calculated suction pressure. The unit parameters are then interpolated by transforming standardized flow and station differential to station head, suction flow, and transmitted horsepower.

The differential is increased in 7-psi steps and the minimum outlet pressure restriction tested.

At each new differential, tests are performed to determine if the calculated data are outside compressor interpolation limits or above maximum speed or below minimum speed limitations. Calculated flows less than station operating-curve data indicate that the unit is recycling and as such is supplying horsepower commensurate with the minimum station flow at the same head.

The differential is increased on the unit until the maximum unit head is achieved (i.e., heads greater than maximum unit speed cause the unit to assume the maximum speed head) or the station discharge pressure setpoint is achieved. Interpolated speeds below the minimum-unit speed result in the unit being assumed unable to run.

On completion of all interactions for Unit A, the whole process is then considered for Unit B. A "best configuration" array is updated based on the resultant fuel usage of any valid configuration passing all of the pipeline and unit operating criteria.

Fig. 5 shows the configuration-setup screen and Fig. 6 shows the alternative output screens. Fig. 6a shows a selected station's head/flow curve and fuel gas vs. horsepower, indicating the operating point on each. Fig. 6b is the alternative pipeline profile data output screen.

FINAL ASSESSMENT

PASA's experience in developing the optimization system has led to the following conclusions:

The development process of an optimization strategy is in itself a significant contributor to more economic operation.
Having real compressor characteristics for the individual units and using these as input to the optimization routine is of significant benefit.
Maintaining steady flow conditions on as much of the pipeline as possible is significant in fuel savings and unit operating stability.
Accurate sales forecasting for current day and week is important but, for some ranges of flows, does not significantly affect fuel optimization. Early warning of significant change is important.
Linepack in itself is not necessarily a good operating guide but is important in terms of timing for configuration setup or changes.
The optimization system must include a list of performance measures.

Pipeline optimization includes not only the optimal operation of the pipeline facilities but also the ability to predict and setup situations which will be optimal solutions for even the most adverse operating conditions. This applies to fuel use as well as in such areas as maintenance scheduling, pipeline repair, and spares inventory.

The use of a steady-state model to predict pipeline compressor configurations and setpoints has become an important tool in assessing, costing, and optimizing operations on PASA's gas-transmission pipeline, even subject to highly transient flows.

The success of using an iterative method rather than linear-programming methods depended somewhat on the simple nature of the pipeline system. This should not detract from the overall effectiveness of the method, however, because its simplicity provided credibility and therefore an impetus for change.

Development of the model could proceed while it still provided significant results. Savings were recognized soon after its implementation because of the unification of decision making it provided.

The flexibility resulting from having the best 20 configurations allowed for the uncertainty in gas-sales' estimates by providing efficient higher horsepower configurations.

BIBLIOGRAPHY

Bertelrud, Arild, "Simplified Pressure Prop Calculations For Gas Pipelines Using SI Units," Pipeline & Gas Journal workbook series, Energy Publications, 1983.

Campbell, John M., Gas Conditioning and Processing, Vols. 1 and 2.

Chilton, Cecil A., and Perry, Robert H., Chemical Engineers Handbook, 5th Edition.

Flanigan, Orin, "Transient Flow: Can it Conserve Compressor Fuel," AGA Distribution/Transmission Conference, May 18, 1982.

GT-22/CDP-416 Training Manual, Ingersoll-Rand.

Percell and Ryan, "Steady State Optimization of Gas Pipeline Network Operation," PSIG, 1987.

Rachford, H.H., "Good Operating Strategy Saves Compressor Fuel in Natural Gas Transmission," SPE 4695, 1973.

Staroselsky, Naum, and Lawrence Ladin, "COMPRESSOR CONTROL-1: Parallel centrifugal gas compressors can be controlled more effectively," OGJ, Nov. 3, 1986, pp. 78-82.