30 May 97
Columbia and Snake Rivers hydroelectric station
Phase 2 - Probabilistic risk analysis techniques
powerhouses (Vo et al. 1995a,b). The purpose of
and software were used to complete this phase. The
the analysis is to evaluate the risks associated with
postulated initiating events (loss-of-load, internal
events that would require a nonroutine shutdown at
flooding upstream of the wicket gates, and internal
hydroelectric stations and involve an inability to
flooding downstream of the wicket gates) were
close the intake gates within the time normally
modeled in event trees. Systems required to
allotted to close. The Corps guidance for rapid
respond to these events were modeled in fault trees.
closure of the intake gate is the 10-minute closure
The fault tree component failure rate information
rule which requires intake gates to be capable of
was taken from the Phase 1 database. The model
closure within 10 minutes in the case of a flooding
accounted for the minimum time that a component
or overspeed event. The ability to meet the
could operate, the minimum time that a component
10-minute closure rule is questionable for
could fail, and time-based recovery actions. In
hydroelectric stations that have their intake gates
addition, the model accounted for the different plant
removed or raised from the original design position.
conditions which exist in the field (i.e., differences
The intake gates at some hydroelectric stations on
in design, operations, etc.). This latter feature
the Columbia and Snake Rivers have been removed
allowed 48 different field conditions to be modeled.
or raised to improve fish guidance.
The results of this phase were "frequency profiles."
These profiles reflect the frequency of remaining in
b. This project provides a general prob-
a potentially damaging event versus time after
abilistic risk assessment (PRA) for hydroelectric
initiation (e.g., frequency of having loss-of-load
stations. The results of the PRA are being
conditions which last 5 minutes, 10 minutes,
synthesized with an economic consequence analysis
15 minutes, etc.). At present the model only
to produce results in terms of economic risk.
handles the three events of concern for the project.
Results of this study can be used for policy and
Other events are possible/plausible which could
decision making. This project was broken down
have application to a rehabilitation evaluation.
into four phases. A separate report was or will be
Only systems and components required to mitigate
issued for each phase. Each phase offers
the events of concern were modeled. With the
information and/or processes that individually could
addition of new events, more system models and
be useful to rehabilitation evaluations.
components could be required. The existing system
models may also require additional detail. As noted
Phase 1 - This phase involved collection and
above, the model can handle 48 major design
analysis of relevant hydroelectric power equipment
features (based on governor type, intake gate
failure data. Reviews of failure data from generic
design, emergency wicket gate closure, etc.). The
sources were conducted and data were collected
selected design features were found to be adequate
from a survey of hydroelectric stations and an
for differentiating between possible plant response
expert panel elicitation. For each component the
to the events of concern. With the addition of new
sources were combined using a Bayesian process.
events, more design features may be required.
The resulting failure rate values are generic for the
components over their expected life. Failure rate
Phase 3 - Given an event occurring for a
functions (i.e., failure rate vs component age) were
not developed. Failure rate functions would need
damage of a certain level (a damage state) will
to be developed to support both major maintenance
occur. There is an economic cost associated with
and rehabilitation programs. In addition, the
each level of damage. Phase 3 collected informa-
failure information for electrical components
tion to (1) delineate the different damage states,
generally came from nuclear related sources.
Further research into the applicability and possible
state given an event lasting a set time, and
development of hydroelectric-specific electrical
(3) estimating the cost associated with each damage
component failure rates could be warranted.
state. This information was collected from a