Second-day tests
The primary objective of the second series of tests was to evaluate the impact of MSP on the unit's ability to ramp to load. A secondary objective was to collect data to support the possible design of a solution using model predictive control by a POMS. Because test time was at a premium, pursuing both goals simultaneously seemed warranted. For this reason, operators manually controlled the primary steam temperature control loops of burner tilts, the final superheat temperature, and the number of burners in service.
Figure 7 illustrates the results of the load ramp tests conducted during the second night. Start conditions are shown in white and correspond to the white timeline on the far left. Pressure was ramped from 2,000 psig to 1,500 psig at 30 psi/min while varying the number of burners in service. Steam temperatures were effectively maintained within a narrow range. There was a subsequent drop in steam temperatures due to an operator's initial misunderstanding of the inverse relationship of final superheat controller output and the position of spray valves. Later, using manual control, he was able to maintain steam temperature relatively close to the setpoint for the remainder of the load ramp tests.
7. Second-day test results. The increase in first-stage temperature of approximately 34 degrees F from the full-pressure start condition represents a reduction in temperature variation from 170F to 136F from low load to full load. This yielded a 60% increase in the allowable rate of change of temperature for the 10,000-cycle cyclical life expenditure curve. Source: LCRA
The load was first ramped from 60 MW to 90 MW while pressure was increased from 1,500 psig to 1,650 psig. Then load was ramped to 130 MW at 1,850 psig. The concern was that the combined effect of building load and pressure would cause high steam temperatures. Peak steam temperatures were readily controlled, but a dip did occur following the ramp, indicating a need for further control action. The subsequent load reductions and load increases all resulted in smaller steam temperature variation, due to improved operator control action.
Test results showed an increase in first-stage temperature of about 34 degrees F from the full-pressure start condition. This represents a reduction in temperature variation from 170 degrees to 136 degrees from low load to full load. According to the HP rotor temperature SALI curve, the reduced temperature range provides a 60% increase in the allowable rate of change of temperature for the 10,000-cycle curve (from 280 to 440 F/hr) and an allowable increase in average ramp rate from 7.4 to 14.6 MW/min between 60 MW and 330 MW.
There is a very steep gradient of first-stage temperature at high load. If the maximum load for fast ramping is limited to 315 MW, the range of first-stage temperature change is reduced by 20 degrees F. Applying MSP and this load limit for fast ramping reduces the HP rotor temperature range to 116F for an allowable rate of temperature change of more than 800 degrees F/hr for the 10,000-cycle curve, or an average ramp rate of 29.3 MW/min. In fixed-pressure operation, the 150F variation for a 315-MW maximum load permits a temperature change of 340F/hr, corresponding to a ramp rate of 9.6 MW/min.
During the final load ramp test, load was added at 18 MW/min from 60 MW to 130 MW and then back again. The throttle-pressure setpoint was ramped at 45 psi/min up from 1,500 psig to 1,650 psig, and then to 1,850 psig during the increasing load periods. Pressure was ramped from 1,850 psig to 1,500 psig at 30 psi/minute during the load down-ramping. The pressure ramp rates, which were specified by LCRA engineering, are consistent with the start-up steam drum limitations of 200 degrees/hr for increasing temperature and 150 degrees/hr for decreasing temperature. The total range of superheat temperature variation was 977F to 1,008F while load varied from 59 MW to 131 MW and pressure varied from 1,488 psig to 1,858 psig.
There was no explicit testing conducted at high load. However, the unit was ramped from low load to full load during the afternoon of October 28. As the unit was ramped at 18 MW/min, the steam temperature increased to 1,025F with the spray valves full open. The load ramp was stopped until the steam temperature dropped to an acceptable value, and then continued. Figure 8 shows the next set of trend lines, 20 minutes later. A load of 315 MW was reached at 12:51:32, or 20 minutes after the start of the ramp, as indicated by the white timeline.
8. Fast load ramp. A load of 315 MW was reached in 20 minutes as part of a load ramp test, as indicated by the white timeline. Source: LCRA
The first-stage metal temperature reached 918.2F at 12:51:32, representing a 150-degree temperature rise in 20 minutes, or 450F/hr. Based on the SALI curve, this corresponds to a CLE of 0.02%, or a 5,000-cycle curve. Although the unit was ramped at 18 MW/min at times, the average ramp rate was 11.25 MW/min. The slow temperature response of the first stage produced a lower-than-normal temperature for the 90-MW condition at the start of the fast load ramp. This contributed to the large first-stage metal temperature change.
These data indicate that the 18 MW/min ramp rate may exceed the allowable CLE 10,000-cycle curve if load is ramped from low load to nearly full load. If steam temperatures had not limited the load ramp, it is likely that a faster rate of temperature change would have produced even greater cyclic life expenditure.
With MSP in place, the temperature variation would have been approximately 120 degrees instead of 150 degrees, due to a higher first-stage metal temperature at low load. The 10,000-cycle curve allows an 800F/hr rate change for operation, which corresponds to an allowable 25 MW/min ramp rate.
Achieving these fast ramp rates with MSP depends on the ability to control steam temperatures. In this test, steam temperatures reached 1,025F with tilts low and superheat sprays at maximum. Because gas header pressure was high, it is unlikely that burners could have been turned off. Another control variable would have to have been used to lower transient temperature.
Writing final specs
Space limitations preclude presenting the details of other tests, of which there were many. However, here are the final system specifications that were implemented in the Sim Gideon control system that maximizes Unit 3's ramp rates:
- Implement MSP controls with a low setting of about 1,500 psig.
- Increase the low-load superheat temperature setpoint from 995F to 1,000F.
- Lower the maximum load in the fast ramp range from 330 MW to 315 MW and configure the 315 MW to 330 MW load range for slow ramping.
- Use the POMS to coordinate multivariable control of the throttle-pressure setpoint, the number of burners in service and their tilts, superheat sprays, and IFGR damper position.
- Upgrade the POMS computer for faster control and compatibility with Connoisseur software.
- Use an on-line algorithm to calculate and accumulate CLE.
Currently, Sim Gideon staff are implementing the "Turbine One-Button Start" project to further automate plant and unit control. The project envisions taking the steam turbine from a cold start to closure of the breaker at 3,600 rpm. Modules to be installed as part of the retrofit will also start the vacuum pumps and reset the turbine. After the breaker is closed, the modules will control load holds and load ramp rate, and assume control of the turbine's control valves from the stop valves. The system can operate in advisory mode as well as in control mode.
Significantly, the design of the control system will also require capture of "institutional knowledge" that must be quantified in order to automate the plant. When the project is completed, the consistency of start-up and unit operations should improve, start-up times should be reduced, and LCRA's new dispatch center should be able to operate the units directly.
The author would like to acknowledge the LCRA Sim Gideon operations, engineering, and management teams for making the testing and implementation of the MSP approach at Sim Gideon Unit 3 so successful. Special thanks go to John Lax for his in-depth analysis of the HP rotor temperature limitations and sliding-pressure boiler constraints, to Keith Trolinger for test coordination and guidance on ramp rate issues, to Robert Chapa for control coordination and Wonderware data collection, to Steve Slater for instrumentation and control guidance, and to Matthew Henderson for the expert boiler control during these tests. A final shout-out goes to Don Labbe of Invensys/Foxboro, who also did yeoman's work.
—David Runkle is production manager of LCRA's Lost Pines Power Park. He can be reached at drunkle@lcra.org.
Email
Comments
Print
Reuse
