Hydro-Québec in Canada and Svenska Kraftnät in Sweden have used a new power system frequency control test bench to accurately model the prime mover and speed governor of a synchronous generator.
By Marc Langevin, Sébastien Guindon, Catarina Là¶fqvist and Evert Agneholm
Marc Langevin, Eng., PhD, is specialist in power system studies with OPAL-RT Technologies. Sébastien Guindon is an engineer with Hydro-Québec Production. Catarina Là¶fqvist is an electrical engineer at the Swedish national grid (Svenska Kraftnät) in the department of security and preparedness and is project manager for the BERTA project. Evert Agneholm, PhD, is an electric power engineer at Gothia Power.
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.
Digital models of speed governors and turbines are necessary to carry out stability studies of an alternating current (AC) generator in an islanded power system, as well as general stability studies of a large power system. Such models have improved with evolving knowledge, requirements and computer simulation technologies. However, these models are not individually validated on-site because it is impossible to ensure only one generating unit will react to a power system disturbance. Even a load or generating unit trip in the proximity of a unit under test will lead to a general power system reaction. Identifying modeling parameters has required highly accurate calculations using data that manufacturers are reluctant to provide or costly and arduous on-site tests.
Speed control parameters should be set to contribute to the power system frequency stability and avoid increasing natural electromechanical power oscillations. To do this, engineers used to carry out stability studies using available digital models. On-site settings were often unusable because, in practice, it was impossible to perform tests representing the real accuracy and robustness of the speed control settings.
OPAL-RT Technologies — in collaboration with Hydro-Quebec, Svenska Kraftnat and Gothia Power — is experimenting with new methods for conducting on-site tests on generating units that should contribute to better:
— Understanding of hydro-generator behavior;
— Modeling of hydraulic turbines and speed governors, for improved transient stability studies; and
— Speed governor settings for improved frequency stability.
The company has developed a test bench called RT-LAB BERTA. RT-LAB, for real-time laboratory, is OPAL-RT Technologies’ real-time operating software for high-performance digital simulators. BERTA is a French acronym for Banc d’Essais pour Regulateurs, Turbines et Alternateurs, which means Test Bench for Regulators, Turbines and Alternators.
Marc Langevin with OPAL-RT developed RT-LAB BERTA. He has more than 30 years of experience in power system analysis at Hydro-Quebec and Hydro-Quebec International and as a private consultant for engineering companies.
He represented Hydro-Quebec TransEnergie (the Quebec transmission system owner) on the task force on operation committee of the Northeast Power Coordinating Council (NPCC). The idea of developing a new test bench for speed governors, and eventually excitation systems, came to his mind as he started working for Hydro-Quebec’s subsidiary TransEnergie Technologies, which was mandated to commercialize the power system real-time simulator developed by IREQ, Hydro-Quebec’s research institute.
Langevin was concerned by the frequency control and simulation problems he faced when he was responsible for the speed governor settings at Hydro-Quebec. Despite many on-site tests, using sophisticated equipment to measure power and wicket gate servomotor behavior, results did not lead to adequate and accurate speed governor and turbine model achievement. Those tests were expensive and were not easy to promote considering the results.
Because the frequency behavior also revealed unsatisfactory damping, following a major generation rejection, he was concerned there was no method to validate that the speed governor settings were adequate to ensure any investigated unit contributed to the frequency stability of the entire system, instead of withdrawing stabilizing energy from the grid. Hence, he imagined developing a test bench that would facilitate power plant on-site tests and generate results significant enough to identify and validate dynamic governor and turbine models and, most of all, ensure adequate settings that contribute to the power system frequency stability.
OPAL-RT and Hydro-Quebec Production (which has a total hydro capacity of about 36,000 MW) have been partners since the early development of RT-LAB BERTA.
Besides the usual instrumentation necessary for conducting conventional open-loop tests on speed governors, RT-LAB BERTA is equipped with a real-time simulator that allows it to drive the governor with various generated signals, among which is a frequency signal corresponding to the tested unit operating in isochronous mode, that is, if it alone was supplying a load equal to its generated power.
The integrated real-time simulator makes it possible for the governor and turbine to react as though they were operating in an islanded power system. It is then possible to adjust the speed governor and validate the governor and turbine models in the same test session. Moreover, the simulator includes a PSS (power system stabilizer) emulator that allows users to determine whether the PSS contributes to the whole power system frequency stability.
An interface using National Instrument’s Labview platform facilitates testing.
System owner requirements
The North American Electric Reliability Corporation (NERC) is developing a reliability standard on frequency control. The purpose of Standard MOD-027 is “To verify that the turbine/governor and load control or active power/frequency control model and the model parameters, used in dynamic simulations that assess Bulk Electric System (BES) reliability, accurately represent generator unit real power response to system frequency variations.”1
Requirement #2 of this standard is: “Each Generator Owner shall provide, for each applicable unit, a verified turbine/governor and load control or active power/frequency control model, including documentation and data … to its Transmission Planner in accordance with the periodicity specified in MOD-027 Attachment 1.”
Many producers object to the cost and difficulties of providing such verified models. Besides, for commissioning purposes, they consider that conducting tests on a generating unit is unnecessary, unproductive and expensive. On the other side, important system owners and experienced producers are aware of the risk of power system instability and potential collapse due to inaccurate dynamic stability simulation and/or inappropriate frequency control. Many transmission system owners have implemented multi-level load shedding schemes because they can’t rely on appropriate speed governor response to frequency deviation.
In its standard #60308 on testing of hydraulic turbine control systems,2 the International Electro-technical Commission (IEC) defines tests that should be performed for hydraulic turbine controls, including validating the capacity of operating in islanded network. Although they should not be considered mandatory, IEC gives some hints on how to perform isolated network field tests. IEC suggests: “Instead of carrying out real isolated grid field tests, to develop an intermediate method, based on an on-line isolated grid simulator.”2
Real isolated grid field tests are expensive and practically impossible: intentionally causing frequency disturbance in a given area is certainly not safe or appreciated by utilities’ customers.
As a member of NPCC, Hydro-Quebec Production needed to provide to Hydro-Quebec TransEnergie the required models and parameters for the hydraulic turbine and speed governors. These models must be in accordance with IEEE standards “Dynamic Models for Steam and Hydro Turbines in Power System Studies” and “Hydraulic Turbine and Turbine Control models for System Dynamic Studies.”
The provided models shall be usable by Hydro-Quebec TransEnergie engineers for accurate frequency stability simulations. On-site test and simulation results must be compared to prove that the provided models and associated parameters refer to actual installed equipment, not exclusively to theory and/or planning specifications.
On-site tests were essentially open-loop tests, sometimes including mechanical torque measurements using strain gauges. Setting up such instrumentation was time-consuming and thus expensive, as the tested unit must be out of service, engendering considerable revenue losses: most of Hydro-Quebec’s hydro generators are large units (more than 100 MW).
Hydro-Quebec Production was looking for a cheaper way of measuring mechanical torque and a better way of performing on-site tests that would result in useful information for modeling and frequency stability insurance. When Langevin proposed the BERTA solution, Hydro-Quebec Production engineers did not hesitate long before recommending it to their management.
Dynamic model identification and validation
Despite disposing of data resulting from open-loop tests characterizing the turbine power in steady state and the gate servomotor response to a command input, and computed parameters such as the water starting time, engineers don’t always succeed in reproducing the behavior of the prime mover system through off-line simulation. Closed-loop tests, in particular, often reveal non-linear behaviors that were not anticipated through off-line simulations based on non-validated models. Engineers then need to investigate why their model doesn’t fit reality. Nevertheless, these kinds of on-site tests allow taking note that simulations done with governor and prime mover models do not always provide results conforming to reality.
In a series of closed-loop tests performed at Hydro-Quebec’s 780-MW Eastmain 1A hydro plant, the results with the current gain values showed good stability and no reason to worry about the plant’s contribution to grid stability. However, in an attempt to reduce the frequency deviation resulting from a sudden load disturbance, the speed controller proportional gain was increased from 1.5 to 2. The actual results were far from expected.
Figure 1 on page 30 shows the frequency deviation resulting from a sudden 5% load shedding in an islanded grid, the tested unit alone supplying a load corresponding to 80% of its rated MVA. Figure 2 on page 30 shows the mechanical power. Electric power, which corresponds to the load, starts at 81% and suddenly drops to 76%. Of course, this represents a dramatic case. But it clearly means that we can no more rely on classical turbine and speed governor models for correctly simulating the dynamics of a hydroelectric plant in an important load/generation rejection case.
Improving frequency stability
Frequency stability problems are becoming a major concern. Even in well-developed countries, an increase of frequency stability problems is observed, as was recently reported in a study requested by the U.S. Department of Energy: “Over the past decade, the North American Electric Reliability Corporation (NERC) has observed an increase in frequency stability problems. For example, frequency response in the Eastern Interconnection has deteriorated significantly over this period, so that progressively smaller power disturbances are able to induce significant frequency deviations…”3
Several causes have been suggested, and these are among those the study identifies:
— Decrease of total moment of inertia due to the addition of smaller rotating units, slow rotating generators (wind farms) and non-rotating power sources.
— Modification of load types. The power system is less naturally damped because of the nature of new loads that don’t behave in the good direction as the frequency deviates. “As the load in North America changes … it includes more electronics and variable-speed drives that do not demonstrate the same beneficial frequency-power relationship as inductive motors.”3
— Bad generation control practices. Many producers prefer to operate at best efficiency versus “what is optimal for the overall power system.”
Because of their inherent characteristics, hydroelectric plants with upstream reservoirs are generally considered best-suited for primary frequency control. They can easily tolerate deviations of their operating points. So what criteria should guide the power engineer when setting (or recommending settings for) the speed/power control of a hydraulic unit?
The first criteria should be to ensure its operating stability in an islanded power system, regardless of its actual operating mode. Due to the dynamics of pressure in a penstock, the so-called water hammer effect that transiently drives the mechanical torque in the opposite direction of water flow, stable isochronous operation of hydro generators is often more challenging than expected.
The second criteria should be to avoid contributing to natural electric power oscillations. This typically occurs when the control gains, mainly the derivative gain, are too high or when the speed and/or electric power signals driving the speed governor control are not properly filtered. The small oscillations of these two signals in the frequency range between 0.4 Hz and 3 Hz, characterizing synchronous operation in a multi-generator system, can reverberate on the displacement of the turbine gates and generate destabilizing mechanical power oscillations.
Meeting those two criteria might look easy as power engineers often rely on off-line simulation and usual open-loop tests for recommending speed controller settings. But, as revealed by closed-loop testing, an apparently stable system may actually be dramatically unstable. Closed-loop testing, simulating the tested unit behavior in islanded operating mode, is rendered easy thanks to hardware in the loop (HIL) technology.
Using HIL provides more data, helps validate stability
When using a pre-programmed signal to drive a controller, we realize an open-loop test. Whatever the characteristics of the controller and controlled system, the response is forced by the pre-programmed input signal. In the case of an electric power system, this response will also be determined by the behavior of the other synchronized units. Extracting the individual characteristic behavior of the tested unit from this type of test is not easy.
On the other hand, using a real-time simulator inside a test bench makes it possible to generate and inject into the speed governor a signal that is computed according to the current power variation of the generator. If this input signal results from simulated speed computed according to the turbine power variation and rotating inertia, we are performing a closed-loop test, corresponding to the operation in islanded network. The system response could therefore be considered as the natural response to an external disturbance, actually a load variation, as if the generating unit was supplying this load alone, i.e. as if operating in islanded (isochronous) mode.
The main added value of BERTA is to provide this feature that ensures the speed governor is correctly set for isochronous mode frequency stability and for contributing to the whole power system stability. If a closed-loop test conducted with the test bench reveals instability or under-damped oscillations, this means the tested unit deteriorates the whole system stability. As too many generating units exhibit similar behaviors, the whole system stability is endangered. When 50% of its total rotating inertia originates from units that would be unstable in isochronous operating mode, a system is on the edge of a major outage.
Some results from the use of BERTA
Since 2009, Hydro-Quebec Production has used RT-LAB BERTA to validate governor settings that were recommended following off-line simulations, to make sure that they contribute to Quebec’s grid stability.
BERTA recently demonstrated its efficiency in a series of on-site tests at Eastmain 1A. Tests on a 260-MW unit were conducted to identify and validate the speed governor and hydraulic turbine dynamic models to be used in the utility’s power system transient stability program. Its integrated islanded operation real-time simulator, directly connected to the generating unit speed governor, was used to identify and validate the speed governor and turbine models and find adequate settings for islanded operation, thus ensuring the unit will contribute to the whole power system stability. As was the case with former on-line tests in two other hydroelectric plants, the islanded operation simulator made it possible to observe poorly damped gate servomotor oscillations that cannot be revealed by conventional open-loop tests.
Sebastien Guindon, the lead engineer in speed governor testing at Hydro-Quebec Production, says: “Hydro-Quebec Production used BERTA to test hydraulic turbines and their governors in four hydropower plants. The last, but not least, was at Eastmain 1A power plant. Through real-time simulation of islanded system operation, using the real controllers, turbine and generating unit, it was shown how the use of classical theoretical models can be hazardous. It is recommended to perform tests for model identification and also for simulation behavior especially in isolated network to ensure that the controller settings contribute to network stability.”
At the Eastmain 1A powerhouse, the PSS emulator integrated in the BERTA simulator made it possible to demonstrate how valuable it is to use a multi-band PSS in reducing the frequency deviation following a major power/load unbalance disturbance. Figure 3 shows the effect of a multi-band PSS on the frequency deviation amplitude and stability. Considering the governor settings that were used during these two tests, it is obvious the multi-band PSS has a great stabilizing effect on the tested unit. It would have been impossible to observe this effect in open-loop and closed-loop tests without PSS emulation.
The use of BERTA for conducting on-site tests in power plants has been described by Hydro-Quebec TransEnergie as an appropriate tool for identifying and validating dynamic models of speed governors and turbines. More specifically, the transmission system owner agrees that the on-site real-time simulation of the islanded operating mode is a good method for analyzing the tested unit’s contribution to the system frequency stability.
Svenska Kraftnat is the Swedish transmission system owner, responsible for securing electricity supply during stressed situations. In rare cases, this means establishing islanded operation of regional parts of the power system. At the company’s request, the islanded mode simulator was improved to include a virtual islanded grid comprising up to five generating units and five dynamic loads that can be paralleled with the tested unit. The behavior of the tested unit in reaction to load shedding and/or generating unit trip can therefore be simulated.
As a consultant for Svenska Kraftnat, Evert Agneholm from Gothia Power conducted many on-site tests using BERTA. He says: “Island operation where a unit alone or together with a few units shall take care of the frequency and voltage control in the island is normally the most challenging mode of operation for a production unit. To secure a stable and robust control of the units involved, it is extremely important to test and verify that the units and their controllers can run in islanded operation. Testing real islanded operation is normally difficult to perform as the grid owners are afraid of using real load during tests. By using a HIL simulator such as BERTA, realistic tests can be performed without risking disturbances for the customers.”
He continues: “During tests performed with BERTA, stability problems, related to bad tuning of the turbine governor, have been discovered. During the tests, the parameter settings have therefore been changed to provide a stable, robust operation.”
Several tests were performed on individual units to study their contribution to the frequency control of the synchronized Scandinavian power system. Other tests were also performed where two and three big hydro units were operated in parallel by BERTA and generated power variations that could be clearly seen on the grid frequency, i.e., tests on the overall frequency control of the synchronized Scandinavian power system.
In a series of tests conducted on a Francis unit, a superimposed frequency deviation signal of 0.1 Hz amplitude and sinusoidal time period varying from 6 to 600 seconds was injected in the tested unit speed governor. Figure 4 on page 37 shows the results using a sinusoidal input signal with a 60-second time period.
Figure 5 on page 40 provides an example of a Bode plot resulting from a series of tests. As expected, the gain in terms of MW/Hz is continuously decreasing as the frequency of oscillation increases. There is also a significant change in the phase. These tests were performed at three loadings and it can be seen that the gains vary according to the loadings. The reason for this difference is related to the fact that the speed governor uses the wicket gate opening as feedback signal and not the power. As the relationship between gate opening and power is not linear, the MW/Hz gain varies with the loading. In synchronous mode, with gate feedback, the electric power signal is kept outside the control loop. This is a good practice from the stability point of view but an inconvenience for maintaining the power setpoint. In Sweden and Norway, only a few hydro units use active power as feedback signal.
Several islanded operation tests were performed in Sweden, with good results.
Many on-site tests in power plants, using HIL simulation integrated into OPAL-RT’s RT-LAB BERTA, have demonstrated the benefits of investing time and money to identify adequate models for power system stability studies and check the robustness of speed governor settings. In recent years, Hydro-Quebec Production and Svenska Kraftnat have shared their experience with OPAL-RT Technologies in order to improve BERTA.
Both believe that conducting speed governor on-site tests in power plants is worth the investment. Using HIL technology allows conducting closed-loop tests that provide results quite more significant than conventional open-loop tests.
Besides continuously improving the current application for speed governor tests, OPAL-RT is presently developing an equivalency for excitation systems.
Cost of the BERTA test bench, including two to five days of training, varies from $80,000 to $120,000, depending on options. OPAL-RT also provides the services to conduct tests in the power plant and to deliver a technical report, including recommendations on dynamic models and speed governor settings.
1“Verification of Models and Data for Turbine/Governor and Load Control or Active Power/Frequency Control Functions,” Standard MOD-027-1, North American Electric Reliability Corporation, 2014.
2“Hydraulic Turbines — Testing of Control Systems,” Standard 60308, second edition, International Electrotechnical Commission, 2005.
3“Frequency Instability Problems in North American Interconnections,” National Energy Laboratory, 2011.
More HR Archives Issue Articles