This document summarizes a risk analysis of options for a wastewater treatment plant for a gold mine. A Monte Carlo simulation model analyzed the costs of three options: a used skid plant, a new skid plant, and a new fixed plant. The analysis found that a new skid plant had the lowest expected cost and was least impacted by parameter variations, making it the recommended option. A sensitivity analysis identified key cost drivers to incorporate into the Monte Carlo modeling. Risk measures including value-at-risk and mean upper semi-deviation supported choosing the new skid plant as the lowest risk option.
CH&Cie white paper value-at-risk in tuburlent times_VaR
Waste Water Treatment Risk Analysis
1. PROJECT REPORT
Risk Preferences Analysis of a Waste Water
Treatment Installation
Ashwanidua
AaronFriedlander,RobertMayers,Ashwani Dua,RohanSanas,Andrew McCormack,
LavinaChoudhary
Prof. Ricardo A. Collado
BIA-670
Risk Management: Methods and Practice
Fall 2015
2. 1
Table of Contents
1. Project Description and Background.........................................................................................................2
1.1. Wastewater Plant Model..................................................................................................................2
1.2. Sensitivity Analysis...........................................................................................................................4
1.3. Assumptions....................................................................................................................................4
2. Model Description and Analysis................................................................................................................7
3. Explanation of Performed Analysis ...........................................................................................................8
4. Results....................................................................................................................................................8
4.1. What If Analysis...............................................................................................................................8
4.2. Monte CarloAnalysis,VaR and Semi Deviations.................................................................................9
4.3. Risk Measures:...............................................................................................................................10
4.3.1. Value at Risk:.............................................................................................................................10
4.3.2. Mean Upper Semi-Deviation:......................................................................................................12
5. Conclusions and Recommendations .......................................................................................................15
6. Participants and Responsibilities ............................................................................................................17
3. 2
1. ProjectDescription and Background
The gold mine construction is almost nearing its completion when the team discovered
that the water dischargecontaminant level is much higher than the original forecasts.
This will lead to enormous environmentaland regulatory implications. Also, fromthe
governmentrules and regulations perspective high levels of contamination may lead to
financial loss in the formof fine or result in complete shutdown. Regulation requires
improved facilities to recover heavy metals fromthe mine water discharge. So the team
came up with the solution to bring in a Skid Wastewater Treatment module. This
module requires Investment, Operating Costs, and can potentially delay the Mine start-
up. Below is the model which team has come up with to decide which treatment plant
should be used.
1.1.Wastewater Plant Model
Analysis shows 3 main alternatives based on input factors.
A. Buy a used skid wastewater plant
B. Buy a new skid wastewater plant
C. Buy a new fixed wastewater plant
In order to choosebest alternative, wehave set tradeoffs between few factors.
4. 3
Tradeoffs:
D. Month Delays, Costs, Payoffs
E. Initial investment, Installation, Operating cost, Salvage value
Figure 1 - Tradeoffs
In order to perform a Monte Carlo Simulation given inputs, constraints, and cost
function, we have set a goal to usedifferent risk preferences to guide our decision
making process. We areinterested in understanding how our decisions depend on our
given parameters.
5. 4
1.2.Sensitivity Analysis
We will performa sensitivity analysis by considering a rangeof values for our
parameters. In order to do this, we vary each parameter independently of the others
and observethe outcomes. An exploration of our model reveals that our key parameters
are:
• Time to complete other distribution
• Time to complete distribution
• Acq + InstallCost
We focus on testing these parameters with an incremental percentage change. This is a
formof stress testing our model and evaluating the parameters sensitivity. This allows
us to evaluate our optimal decision (the one that minimizes cost) subjectto our
parameters risk preference.
1.3.Assumptions
Below are few assumptions wehavemade in order to performthe simulation.
AboveModel is further extended using Monte Carlo Distribution.
1) Model assumes Time to Complete alpha as 2.
2) Multiplier for Time to Complete Beta Distribution for the calculation is 27.5 with
1 constant.
3) Sigma for Normal Distribution is 0.632
6. 5
4) Alpha for Time to Complete Beta Distribution is 2 with 3.067 Constant.
Below are the tornado graphs for each plant type, which show how sensitivethe plant
costs are to each parameter:
Figure 2- Whatif Analysis - Fixed Plant
8. 7
2. Model Description and Analysis
Monte Carlo simulation was used to model the outcome of the wastewater treatment
plant options. The simulations ran in Excel using the Palisades Risk module. The Monte
Carlo simulations were run for each scenario to producean estimated cost.
The cost for each scenario is calculated using this equation:
Cost= AqInstallCost+(Operating Years x Operating Cost)+ (Months Delay x Opportunity
Loss)-SalvageCost
Where:
Months Delay = MAX( Months to InstallPlant -Months to Complete Other Activities , 0)
Each scenario was measured in the following ways: Expectation, Value at Risk, and
Mean upper semi deviation. Expectation (E[C]) is the expected cost. This is a risk-neutral
measure; the plant with the lowestexpected cost is the best choice. Value at Risk (E[C]
+αVaR[C]) uses the expected value and a risk measure(VaR). The α in the VaRformula is
the quantile, which is 95% sincewe are interested in costs aboveaverage. This method
is risk aversesinceit also accounts for the chance of a higher cost. Mean Upper Semi
Deviation (E[C] +α E[[C – E[C]]+]) calculates the expected cost plus the averageof costs
greater than expected (costs lower than expected are not added). This method is also
risk averse, as it only accounts for the average costover the averagecost.
9. 8
3. Explanation of Performed Analysis
To ensurethe correctdecision is made, management has to incorporateuncertainties in
the projectthat can pose potential risks to the projects and associated outcomes. As risk
can be related to any aspectof the project, simulation helps with executing multiple
scenarios by selecting randomvariables.
Multiple simulations being run increases the confidencein modelling. Incorporation of
Risk methods like Var and Mean Semi Deviation further assists with the decision making
by incorporating maximum loss at the given confidence level. Mean Semi-standard
deviation is a popular measure of downsideuncertainty or risk .
4. Results
As stated above, team performed both What if analysis and Monte Carlo simulation.
1000 iterations of the selected scenario (recommended by What if analysis) were
simulated to understand the probabilities that different outcomes would occur.
4.1.What If Analysis
Based on the What if Analysis (Ranked analysis), the parameters that impact the model
most wereidentified as following for Skid and Fixedoption.
1. Time to complete other distribution
2. Time to complete distribution
10. 9
3. Acq + InstallCost
For Usedoption Parameters wereidentified as :
1. Time to complete other distribution
2. Time to complete distribution
3. Multiplier
4. Acq + InstallCost
Based on the maximum impact across options, the following parameters wereselected
for Monte Carlo Analysis:
1. Time tocomplete other distribution
2. Time tocomplete distribution
3. Acq+ Install Cost
4.2.Monte Carlo Analysis
Risk Neutral Evaluation – Expectations - Using DecisionTools @Risk, weadjusted our
parameter rangeover a 30% increment to test cost. Cost range varied from -15% to
15%. Outcomes vary with each Monte Carlo Simulation. Expectation of Costis defined
as the risk-neutralmeasureof our model obtained by a Monte Carlo Simulation,
decision with the minimum cost. We obtain our results from the @Risk simulation
report. 95% confidence interval for our cost function was selected.
Results based on 1000 iterations with Monte Carlo Simulation:
What if Analysis Results
11. 10
Figure 5 - Expectations
As highlighted: Skid Plant was identified as the preferred option based on the
simulation.
4.3.Risk Measures:
After identifying the parameters and simulation run, various risk methods wereapplied
to understand the impact of risk preferences for decision.
4.3.1. Value at Risk:
For VAR, Projectwith minimum value is recommended as the evaluated function is a
cost function.
95% confidenceinterval was selected for our costrange. Model was created in excel
with Palisades Monte Carlo 95% level observations. Costincreased for the parameters
with increasein alpha.
Figure6 to 9: VAR charts
13. 12
4.3.2. Mean Upper Semi-Deviation:
Further Mean Upper Semi- Deviation was applied to supportthe project evaluation,
Team selected -10% to 10 % change range. Based on the Mean Semi Deviation, for Acq+
InstallMean Upper Standard, at 1650, preferred projectis Fixed. Based on the other
parameters skid option is most costeffective.
Figures 10 to 16: Mean Semi Deviation.
15%
-10%
-5%
+5%
+10
%
+15
%
$-
$1,000.00
$2,000.00
$3,000.00
$4,000.00
0
1
Variation
Cost
Alpha
Time To Complete Minimum Value
15%
-10%
-5%
+5%
+10%
+15%
16. 15
5. Conclusions and Recommendations
The main goal of the project was to create a model that would provideinsight into
which wastewater treatment module solution had the minimum cost. The model
analyzed the costfunction for different risk preferences, (Risk Neutral Cost Expectations,
Value at Risk, and Mean Upper Semideviation) which provided comparisons and
1200
1300
1400
1500
1600
1700
1800
1900
0 1
Acq+Install Mean Upper Standard
Deviation 10%
Fixed -10% Used -10% Skid -10%
1600
1650
1700
1750
1800
1850
1900
1950
0 1
Acq+Install Mean Upper Standard
Deviation 10%
Fixed +10% Used +10% Skid +10%
17. 16
sensitivity analysis. The sensitivity analysis was instrumentalin helping to understand
how certain parameters (Time to complete other distribution, time to complete
distributions, acq. + install cost, opportunity loss, operating years) affected the outcome
of the model.
In conclusion, the model showed that the costs associated with buying a new skid
plant are less than the costs associated with buying a used skid plant or buying a fixed
skid plant. This means the recommendation is to install a new skid for the wastewater
treatment. The sensitivity analysis for the risk neutral costexpectations showed that
buying a new skid plant is the optimal choice (lowestcost) for each of the three
parameters used – acq. + install, time to complete other, and time to complete. The
sensitivity analysis for the mean upper semideviation was in accordancewith the risk
neutral cost expectations. Itshowed that the costvalues werelower when buying a
new skid plant compared to buying a used skid plant or a fixed skid plant. This was
consistentfor all of the input parameters.
Other tests that could have been run to help in the decision could have included more
input parameters for sensitivity analysis and more increments during the sensitivity
analysis. Also, the main factor in determining our optimal solution was cost. Itwould
have been interesting to see whatthe optimal solution would have been if there were
other main factors involved, such as time (e.g. we need minimum cost but the skid plant
18. 17
needs to be running within 3 months of acquiring it), long-termvalue (e.g. weneed
minimum cost but the skid plant needs to operate for at least 10 years), or production
(e.g. we need minimum costbut the skid plant needs to be able to treat at least XX
gallons of water per day). There are multiple factors and parameters that could be
added to the simulation model as futuretests.
6. Participants and Responsibilities
Model/ Data / Charts Aaron Friedlander, Robert Mayers,
AshwaniDua, Rohan Sanas, Andrew
McCormack, Lavina Choudhary
Poster Aaron Friedlander, Robert Mayers,
AshwaniDua, Rohan Sanas, Andrew
McCormack, Lavina Choudhary
Report Aaron Friedlander, Robert Mayers,
AshwaniDua, Rohan Sanas