IEEE Power & Energy Magazine - November/December 2016 - 39

30 november 2004, when the hi warrick sensor initiated
shutdowns with the druck sensors indicating a water level of
1,595 ft, a 60-s time delay was added to the warrick sensor
unit trip logic in the Plc on 1 december 2004, and the sensors were physically raised. drawings dated 10 december
2004 indicate that the sensors were placed at 1,596.7 and
1,596.9 ft, although no explanation for this placement was
provided. later, on 15 February 2005, the Plc logic for
the hard shutdown trips based on the warrick sensors was
changed from parallel to series, meaning that both sensors
now needed to detect water for a pumping shutdown to be
triggered. (loss of communication from either sensor would
permit the other sensor to initiate a hard shutdown.)
several unexplained alarms from the warrick sensors
were recorded over the next several months, although no
shutdown commands were issued, due to the 60-s time delay
and the series logic. three of the alarms occurred while
thunderstorms were in the area, leading investigators to
believe that the sensor triggers were a result of induced voltages. another alarm that occurred on 2 november 2005 and
lasted for 9 s was never explained.
On 25 september 2005, as remnants of hurricane rita
were passing through the area, waves generated by winds
came over the top of the reservoir wall; workers described
the overflow as "niagara Falls at the northwest corner of
the reservoir." the druck sensor indicated a water elevation of 1,596 ft. Generator #2 was started to bring down
the water level. erosion ruts of 6-12 in were found on the
outside of the parapet wall; the damage was repaired by
filling with gravel.
two days later, the plant superintendent noted that the
water level was just 4 in below the top of the parapet wall.
the druck sensor indicated a level of 1,596 ft, but the actual
level was at least 1,596.7 ft. One of the three druck sensors
was found to be "a foot higher" and was removed from the
averaging; the calculated water level increased by 0.2 ft with
this change. an additional adjustment of 0.4 ft was inserted
into the Plc logic to reconcile the difference between the
measurements from the druck sensors and the observed
water level on the parapet wall.
the positions of the warrick sensors were "verified"
three days later as 4 and 7 in below the top of the parapet
wall. Because they were located on a higher section of the
wall, however, the actual elevations of the sensors were
1,597.4 and 1,597.7 ft, higher than the lowest point of the
containment wall.
On 4 October 2005, the guage piping that held the sensors was found to have failed. the pipes were visibly bowed
by 5 ft to the side, meaning that the druck sensors were
at a higher elevation than when originally installed. the
maximum allowed elevation was lowered to 1,594 ft on 7
October, with corresponding changes in the Plc logic. an
inspection by a diver on 11 October indicated that the pipes
could be straightened, but a new system of anchoring the
pipes to the reservoir's sides was needed. Plans were made
november/december 2016

and materials were obtained for this work, but a time for
the work was not scheduled, owing in part to the diver's
lack of availability before the end of the year.
as pumping started in the late hours of 13 december
2005, a sudden 1.9-ft drop in the water level was recorded,
even though both units were pumping at the time. the
water level generally rose at the rate of 1.5 in per minute
while both units were pumping. this abrupt drop in the
recorded water level suggests a further deflection of the
sensor piping. the druck sensors reported a water level of
1,591.9 ft when Unit #2 stopped pumping at 4:43 a.m. and
1,593.4 ft when Unit #1 stopped pumping at 5:16 a.m.. the
actual level was 4 ft higher than that indicated by the sensors, as evidenced by the flood. although the revised plant
operating procedure assumed an error of 2 ft in the elevation readings and the pumping control operated to a druck
elevation reading of 1,594 ft or lower, an additional error of
2 ft in the elevation readings from the druck pressure sensors had evidently developed, leading to the overtopping of
the reservoir.
reconstruction of the upper reservoir began in late 2007
and took two and a half years. a complete rebuild of the dam
was deemed necessary. the new dam consists primarily of
roller compacted concrete (rcc) and was the largest rcc
dam in the United states under construction at the time. the
top of the new dam is at elevation 1,601 ft with a parapet
wall extended up another 3.5 ft, permitting a maximum
water level of 1,597 ft in the upper reservoir. the reservoir
also includes multiple water level sensors, continuous video
monitoring, and a spillway.
references to hundreds or even thousands of megawatts
are commonplace in discussions about the electric power
system. this incident demonstrated graphically what hundreds of megawatts can do when they are released in an
uncontrolled manner.

For Further Reading
Ferc taum sauk investigation team. (2006, apr. 28). report of findings on the overtopping and embankment breach
of the upper dam: taum sauk pumped storage project,
Ferc no. 2277 [Online]. available: http://www.ferc.gov/
industries/hydropower/safety/projects/taum-sauk/staff-rpt.asp
rizzo associates. (2010). taum sauk upper reservoir rebuild, project description [Online]. available: http://www.
rizzoassoc.com/cms/images/projectdescriptions/taum%20
sauk%20reservoir.pdf
r. Boyle. (2010, apr. 18). ameren's huge taum sauk project nears completion, St. Louis Business J. [Online]. available: http://www.bizjournals.com/stlouis/stories/2010/04/19/
focus4.html

Biography
Eric H. Allen is with new electricity transmission software solutions, sudbury, Massachusetts.
p&e

ieee power & energy magazine

http://www.ferc.gov/ http://www http://www.rizzoassoc.com/cms/images/projectdescriptions/taum http://www.bizjournals.com/stlouis/stories/2010/04/19/

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - November/December 2016