Page 277 - Sustainable On-Site CHP Systems Design, Construction, and Operations

250 C o ns truction

Significant improvement in reliability can only be obtained when historical price
data are drawn from several completed CHP projects, even when sacrificing some
comparability. Since the available database is finite, it often becomes a trade-off between
sample size and homogeneity. Unfortunately, the precise nature of this trade-off is
unknown, but as the uncertainties are high, a limited number of samples (for instance,
fewer than five) are probably inadequate.
A word of caution is worth noting here. Often, cost data mask the impact of regional
differences in construction prices and differences in size, quality, complexity, and con-
structibility of CHP plant projects. Professional skill and judgment is therefore needed
in the careful selection of projects similar to the proposed project.

Dealing with Contractor Cost Uncertainties
In evaluating cost issues prior to undertaking actual construction either through
competitive bidding, negotiating not-to-exceed or guaranteed maximum cost bids
from a preselected group of qualified contractors or design-build firms, it is hardly
ever the case where all activities experience either their best or worst line item cost
during a given CHP plant project. The general result is a distribution of reasonably
foreseeable CHP plant project construction costs falling in the range of actual costs
most likely to be encountered. One approach used by many successful general or
specialty contractors is to employ probability theory when considering three critical
cost points: minimum cost, maximum cost, and most likely cost together with every-
thing that goes on in between.
The process of selecting a probability distribution sometimes presents difficulties
for the in-house construction risk manager. To choose the correct probability distribu-
tion, the following cautions are generally advised:
1. Be sure to list everything known about the variable sought including all
applicable the conditions affecting the variable.
2. Seek to better understand the basic types of probability distributions and where
they can best apply.
3. Be careful to choose the probability distribution that best characterizes the cost
variable being sought.

Use of Probability Distributions
Distribution types which have been commonly used to evaluate probabilities of various
data distributions under a variety of circumstances include Triangular, Uniform,
Poisson, Normal, Exponential, Geometric, Hypergeometric, Lognormal, Beta, and Weibull.
Distributions can either be continuous or discreet.
The most important factor in price forecasting is uncertainty. Cost estimation is
more of an art than science since it involves both intuition and experienced judgment.
There exists no objective test of the probability that a particular cost determination will
result since it is the sum of many factors. Therefore an objective evaluation of its accu-
racy is possible only by the use of statistical techniques.
Probability theory allows future uncertainty to be expressed by a number, so
that the uncertainty of different events may be directly compared. Information about the