Real time controls

In 2000 BPA initiated a program to develop a control system using the high speed, high quality measurement that phasors provided.  One of the limiting conditions in the Western grid for north to south power transfer is the possibility of a large generation loss in the southwest.  This would create a large N-S swing that could cause the regions to go out of step.  The contingency limits the allowable power flow scheduling under certain conditions.  A control that could detect this type of  incident and take action quickly enough to prevent out of step could allow higher power flow at certain times.  The basic detection plan was based on the fact that during a large swing, the intertie voltages will go way down and the nearby generation will increase VAR output to support intertie voltage depress.  This proposed control, developed under the direction of Carson Taylor at BPA, included two detection schemes.  One scheme used a weighted average of bus voltage measurements from PMUs along the intertie.  When the average exceeded a threshold it issued a trip using an inverse time/magnitude accumulator.  The other used the same weighted voltage average but combined that with another weighted average of VAR outputs from generation that supported the intertie.  A fuzzy logic engine combined these two indicators to output control signals.  Input from the controller was from the PMUs through the PDC at the control center.  Output was through the existing RAS circuits.  Control actions could include capacitor insertion, reactor trips, and generation trips.

The system was fully prototyped and ran in monitor mode for 4 years.  After the initial development stage, it only issued a trip one time, and this was the only incident that was significant enough that it should have issued a trip.  The worst case contingency for generation loss in the SW used in most studies is the loss of two generators at Palo Verde.  This is plausible based on WECC criteria, and is larger than other outages.  In this unusual incident in 2004, not just two but all 3 Palo Verde units were lost in addition to several other generators in the area.  The swing was large and very comparable with the model of such a swing [Figure 10], but the system stayed intact largely because the N-S power flow was well below the limit.  It provided a good test of the control and showed that both detection algorithms worked very well.  Incidents of this type and magnitude are exceedingly rare-fortunately-so implementing and demonstrating reliable and secure controls with a high degree of confidence is difficult.