Installation of a digital governor on the single Pelton unit at the 91-MW Stanislaus facility allowed Pacific Gas & Electric to increase unit efficiency by 2% overall and 3% when operating in the lower power ranges.
By Matthew Roberts and Mark E. Nunnelley
Multiple needle impulse turbines, among the most efficient hydroelectric turbine designs, often are restricted from reaching their full operating potential because of control limitations in governor systems. Mechanical, analog and early digital governors cannot separate control of the deflector and individual needles, resulting in such operating obstacles as poor needle control and the inability to realize maximum turbine efficiency. Modern digital controllers and hydraulic controls allow for precise, independent positioning of the deflector and each needle, and thus needle sequencing. With these features, all the aforementioned shortcomings can be overcome.
Installation of a digital governor at Pacific Gas & Electric Co.’s 91-MW Stanislaus powerhouse in late 2009 illustrates the benefits gained. Stanislaus is capable of rapid loading and unloading, meaning the unit receives constant and significant load setpoint changes throughout the day. The mechanical governor was previously controlled via pulses that resulted in severe over/under shooting of the load setpoint. Now, a setpoint is sent over a communications link to the digital governor, which can quickly change and hold load. Implementation of needle sequencing has allowed PG&E to operate the turbine at higher efficiencies than previously possible.
Mechanical governors and multiple needle impulse turbines
Mechanical, analog and early digital governors control Pelton turbines using the same equipment as for Francis and Kaplan turbines. The governor directly actuates the deflector servomotor in response to deviations from the speed setpoint. A needle/deflector cam provides the needle distributing valves (one per needle) with a setpoint that is dependent on the deflector servomotor position. The shape of this cam serves two purposes: to position the needles slightly open for synchronization and to keep the deflector positioned just out of the needle stream while at load.
This control method exhibits two critical limitations: needle control is indirect, and there is no ability to individually position the deflector and each needle. The most apparent effects of indirect needle control are:
– Lost motion in the cables, links and levers that connect the needle controllers to the deflector servomotor results in sluggish needle response and a large needle position deadband.
– Poor needle control results in power hysteresis when controlling load. In the most favorable operating mode, a modest deviation in load from the setpoint is tolerated. When tight load control is required, operators and SCADA remote terminal units often attempt to make frequent small changes to the governor’s speed adjuster, which can result in premature wear of governor control linkages and associated equipment.
– Governor droop is typically set to 5% when connected to the main grid. The governor primarily acts on the deflector position, and deflector droop is therefore linear throughout its operating range. However, the needle-to-deflector relationship results in a non-linear droop setting for the needles. A legacy governor can exhibit a droop response ranging from 50% (little to no response to speed changes) at low loads to less than 1% at high loads (excessive needle movement in response to speed changes). When operating at 1% droop, it is not uncommon to see ±10% needle movement and power generation due to normal grid variations.
Two consequences of the fixed deflector/needle position relationship are:
– Inability to realize maximum turbine efficiency. The efficiency of impulse turbines often drops at lower load levels (see Figure 1). However, when operating multiple needle units with fewer needles, some low end efficiency can be regained. On a unit operating evenly throughout its load range, total operating efficiency gains of up to 2% can be achieved.
– Inability to hold an isolated load. The needle/deflector cam will position the deflector in the water stream only while at zero load. Once loaded, the deflector moves out of the stream and governing action is primarily accomplished via needle movement (while the deflector may be able to divert water from the runner on a sudden frequency rise, there is no ability to pick up more load on a frequency drop). Because of the slow response time of needles, it is usually not possible to hold a local load of any significant size with an impulse turbine.
Stanislaus is a one-unit plant on the Stanislaus River near Angels Camp, Calif., that was commissioned in March 1963 with an Allis-Chalmers six-needle Pelton impulse turbine and a General Electric generator. The turbine nameplate rating is 113,000 hp at 1,525 feet of head, and the generator nameplate rating is 91 MW, 91 MVA, 0.9 power factor at 13.8 KV. The original governor controller was a Woodward mechanical cabinet actuator for multiple nozzle impulse turbines, which operated the single deflector servomotor and each of the six needle servomotors.
Stanislaus, controlled via the California independent system operator, is used for load following. For that reason, the generation setpoint for the unit is in constant flux. With the mechanical governor, the ISO sent a megawatt setpoint to PG&E’s master station. The master station sent raise or lower pulses to the Stanislaus powerhouse RTU. The RTU output relays energized to raise or lower the speed adjuster on the governor. The inherent sluggishness of the mechanical actuator system provided slow response to this megawatt setpoint and in most cases would not reach desired load before the setpoint changed again. This resulted in the governor almost constantly being in a state of change. Output variations at Stanislaus occur up to 100 times per day and often span the full operating range of the unit.
Attempt at needle sequencing
The frequent operation of Stanislaus at low load levels makes it an ideal candidate for needle sequencing. Needle sequencing allows impulse turbines to operate with a selection of their needles to maximize efficiency. Six-needle turbines are typically operated with two, four or six needles in use, depending on load. Turbine efficiency would then follow the highest of the three curves seen in Figure 1. Because the efficiency gains are greatest at low loads, the gains at Stanislaus are estimated to be upwards of 2%.
Around 1990, PG&E attempted to add a piggyback digital controller that would allow Stanislaus to operate with 2/4/6 needle sequencing. A Woodward 517 digital controller was installed that operated by manually forcing two or four needle controllers closed when required. To maintain constant load on the unit with fewer needles, the speed adjustment was moved to compensate for the addition or subtraction of the extra needles. For the California ISO to control load, pulses were sent to the controller, which would then control the governor’s speed adjuster.
Unfortunately, the nature of the purely mechanical and hydraulic control made operating with needle sequencing unacceptable. Transitions were harsh and resulted in significant output disruptions. The inherent slack still present in the mechanical linkages, levers and restoring cables caused hysteresis in the speed adjustment versus megawatt relationship. As this difference increased, the governor would begin to hunt between what the programmed curves in the controller indicated the speed adjustment should be and what was required to achieve the megawatt setpoint. The setpoint was conveyed to the digital controller via raise and lower pulses from the RTU. The controller would ramp the load up or down and only transition between needle pairs when the setpoint crossed the defined threshold. When the load change requests were large, this resulted in excessively long load on and load off times. These shortcomings led to the decision to abandon the piggyback control system and restore the governor to its original configuration.
Decision to replace the mechanical governor
The mechanical governor configuration continued to operate satisfactorily, and PG&E was able to maintain the equipment in good operating condition. However, the mechanical equipment has some drawbacks, with many complaints from PG&E operations and maintenance personnel:
– Annual disassembly, cleaning, and replacement of worn mechanical parts, as well as frequent readjustment of the governor, resulted in high costs;
– The governor quickly slipped out of calibration after being overhauled;
– While both paralleling and trying to achieve megawatt setpoint, the governor exhibited constant hunting. This hunting resulted in excessive time to parallel, which was often critical during a black start; and
– RTU control relay and speed adjustment motors often lasted only 12 to 18 months due to the excessive adjustments. These failures resulted in a total loss of control by the ISO.
These problems, in conjunction with the inability to operate the needles sequentially, led PG&E to decide to convert to a programmable logic controller-based digital governor in 2009. The digital governor would replace all elements of the mechanical governor. Needle sequencing would be done directly, not through cams, cables and linkages. Control modernization would also improve unit serviceability and reliability. The improved governor would enhance the ability of the Stanislaus powerhouse to participate in grid restoration after local or large area grid power outages.
In 2008, PG&E released a procurement specification for a governor digital upgrade for bidding by vendors. In early 2009, the project was awarded to American Governor Co. By November 2009, the new equipment was installed, commissioned, recertified by the California ISO and released for service.
The scope of the project was to remove the control elements of the Woodward mechanical governor and replace these with modern components. Items removed included the flyballs; control column; needle/deflector cam; links, levers and cables connecting the deflector to the needle control valves; and pilot valves. The deflector and needle distributing valves were retained because they were in good working order and modern components would not offer performance improvements.
The digital governor controller uses two Allen-Bradley Controllogix PLCs in a custom redundant configuration. Allen-Bradley’s standard approach to redundancy allows for power supply and central processing unit duplication but does not provide backup for failure of an input/output module. The approach taken by American Governor allows for full redundancy of the PLC, such that no single component could cause failure of the entire governor system. This is made possible by use of an Ethernet/IP communication link between the processors, which transfers critical process information. A fault in the active PLC will result in a bump-less transfer to the backup controller.
To communicate with PG&E’s master station, each PLC is equipped with a ProSoft Technology DNP 3.0 Master/Slave Communications Interface Module. The link is used to transmit a load setpoint to the governor and provide unit indication and governor status information (such as modes and alarms) to the remote dispatching center.
In addition to redundant PLCs, redundant human-machine interfaces, power supplies and position feedback transducers on the deflector servomotors were used. No redundancy of the needle control equipment was included because the needles themselves are redundant (the unit can continue operating at a curtailed load level in the event of a needle failure). The only single point of failure is in the deflector hydraulic controls. This was an accepted risk and could be mitigated through the addition of a redundant hydraulic control manifold stack.
The seven hydraulic control manifolds, also referred to as electro-hydraulic interface manifolds, are designed to have redundant failsafe operation. Each manifold includes a Bosch proportional valve (which acts as a pilot valve to the existing distributing valves) and a standard directional spool valve (which acts as a shutdown solenoid). To position a servomotor, the shutdown valve must be energized, which allows the proportional valve to control. The proportional valve is configured with a “fourth position,” which acts to position the servomotor in the safe position (closed) in the event of a power loss.
While there is much to be gained through installation of a PLC-based governor, it is the precise control achieved by the separate proportional valves that provides the most noticeable benefits on multiple needle units. The ability to accurately position each servomotor (often to within 0.001 inch) remedies all issues previously associated with poor needle control. Additionally, with the capability of controlling the deflector and each needle independently, new features and control schemes can be achieved.
Precise load control
To implement load following with the mechanical governor, PG&E used a proportional integral derivative control algorithm in its SCADA system. This control loop would compare actual megawatts to a setpoint received from the California ISO and send raise and lower pulses to the governor’s speed adjustment motor. This control method can be classified as a cascade control system, because the outer loop (SCADA PID) controls power output while the inner loops consists of the unmodified speed governor. While cascaded control loops are common, this system tends to be overly sluggish to respond and prone to overshoots and hunting.
This can be seen in Figure 2, which is a SCADA log of generation megawatts (green) versus ISO setpoint (red) over a one-hour period. The average delay to achieve setpoint is about 2 minutes, as can be seen at 17:15 and 17:55. Overshooting of the setpoint is witnessed at 17:27 and appears to be most significant after frequent setpoint changes. This is most likely due to incomplete reset of the dashpot at the time of setpoint change. Additionally, noticeable hunting of several megawatts can be witnessed between 17:40 and 17:50, during a period of no setpoint change.
The cause of this poor performance lies in three areas. First, mechanical governors cannot both load quickly and provide stable servomotor control. This is due to the nature of the dashpot, which is either stable in the normal state or quick to respond when in the bypass state. Secondly, the significant physical backlash present in the mechanical actuator system, as well as the aggressive needle/deflector relationship, results in sluggish and inconsistent needle control. Finally, the outer loop controller is subject to all inherent delays of the needle servomotor timing and water starting time. With long lags invariably comes either slow response or instability.
At the heart of the new digital governor is a standard PID controller. The control algorithm used is speed regulation (also referred to as speed droop with megawatt feedback) and is consistent with loops seen in Standard 1207 from IEEE.1 Megawatt feed forward is included and is the critical feature that allows for smooth and rapid loading without any overshoot or hunting. In contrast to a dashpot bypass, which is essentially just a change in controller gains, feed forward allows quick, manual movements to be made without compromising unit stability.
After the digital governor was commissioned, the delay to reach the megawatt setpoint was greatly reduced and is now consistent with the programmed ramp rates: 28 MW/minute to pick up load and 16.5 MW/minute to shed load. (While it appears that there remains an offset between the setpoint and actual megawatts, the setpoint is changing faster than the unit can physically react. The megawatts are in fact tracking setpoint changes as they occur.) It is also apparent that there is little overshoot and no hunting present during periods of stable demand. This can be attributed to the direct communication of California ISO’s load demand between the PG&E master station and the governor, as well as the use of the feed forward circuit.
Successful needle sequencing depends on a seamless transition between needle pairs, which is transparent to operations and maintenance personnel. This means power output should not be disturbed during a transition, and undue wear should not be put on the needles or needle distributing valves.
The sequencing algorithm developed by American Governor meets both these requirements. The key to achieving this is through fully independent control of each needle. The governor PID control output is a single needle setpoint. This setpoint is then handled by the needle sequencing algorithm that controls which needles are used and at what rate needle pairs open or close to transition on or off. By transitioning needles at the proper rate, total flow through the needles matches what is required by the PID, regardless of the needles’ state of transition.
During a California ISO recertification load on test, the power setpoint was ramped up from no load to full load in just under three minutes. An internal ramp ensures that load changes happen at the defined and certified ramp rates. The actual power output follows the ramp and cleanly trims into the setpoint. Each of the three needle pairs can be seen to ramp open until a transition is required, at which point the next pair opens and the needle pairs match. During a transition, the open pair(s) of needles either closes or holds in place, as required, to maintain a smooth ramp in power output.
By implementing this control scheme, the overall operating efficiency has been increased by up to 2%, with gains greater than 3% in the lower power ranges (see Figure 3).
Isolated loads (water wasting mode)
An additional benefit offered by the digital governor is the ability to hold isolated loads in a water wasting mode. Under normal conditions, the deflector is out of the water stream and all governing action is achieved with the needles. This is adequate for power control on a stable grid but would not work for speed control of an isolated load. The needles are simply not fast enough to respond to sudden load changes.
With the digital governor, the needles can be biased open further than required to hold line frequency at 60 Hz. The deflectors would then enter the streams and modulate water flow onto the turbine to control frequency. This is where the term water wasting comes from: More water than is required to hold frequency is passed through the needles in order for the deflectors to both remove and add power to the turbine.
At Stanislaus, this is used to hold 12 MW of local load in the event all three high-voltage lines are lost. With this control scheme, it would be possible to hold any load up to the generator rating.
There are many other benefits to converting to a digital governor.2 Maintenance costs have dropped significantly due to removal of the numerous mechanical control elements and the governor no longer requiring annual recalibration. With the removal of mechanical pilot valves, oil leaks have been virtually eliminated. And if future expansion of the governor’s capabilities or operating modes is required, the customizable nature of the PLC makes this a low-cost addition.
1. IEEE Guide for the Application of Turbine Governing Systems for Hydroelectric Generating Units, Standard 1207, Institute of Electrical and Electronics Engineers, Ann Arbor, Mich., 2011.
2. Clarke-Johnson, Roger, and Scott Ginesin, Overhaul or Upgrade: Governor Decision Factors, www.americangovernor.com.
This article has been evaluated and edited in accordance with reviews conducted by two or more professionals who have relevant expertise. These peer reviewers judge manuscripts for technical accuracy, usefulness, and overall importance within the hydroelectric industry.
Matt Roberts is project manager and engineering manager with American Governor Co. Mark Nunnelley is maintenance supervisor with Pacific Gas & Electric Co.
More HRW Archives Issue Articles