Table of Contents

1. Wind production in the deterministic case.......................................................................................2

1.1 Methodology for Deterministic Wind Production Simulation...................................................2
1.2 Wind Speed Distribution Modeling...........................................................................................2
1.4. Conversion of Wind Speed to Power Output Using the Turbine Power Curve........................3
1.7 Example graphs for a given location and turbine model............................................................4
2. Power prices in the deterministic case.............................................................................................5
2.1 Capture prices.............................................................................................................................5
2.2 Structure of market forecast.......................................................................................................5
2.3 Example graphs for a onshore wind capture prices in a given country......................................6
3. Wind production and prices in the simulation case..........................................................................7
3.2 Stochastic Modeling of Production and Price Distributions......................................................7
3.2.1 Production Standard Deviation Estimation.........................................................................7
3.2.2 Price Standard Deviation Estimation..................................................................................8
3.3 Preservation of Correlation in Stochastic Simulations...............................................................8
3.4. Generation of Simulated Production and Prices........................................................................8
4. Interest rates.....................................................................................................................................9
5. Construction delay..........................................................................................................................10
1. Wind production in the deterministic case
1.1 Methodology for Deterministic Wind Production Profile
This methodology outlines the deterministic profile of wind energy production, serving as the

foundational step for subsequent stochastic analyses in a project finance context. The approach

models wind speeds using a Weibull distribution, incorporates seasonal variations, applies a turbine

power curve to translate wind speeds into electrical power, and aggregates hourly production to

determine monthly energy generation. The final capacity factor is then normalized and adjusted for

system losses.

1.2 Wind Speed Distribution Modeling

Due to its proven ability to capture the inherent variability and skewness observed in wind speed

data, a Weibull distribution is widely used in wind energy studies. Wind speeds are assumed to

follow a Weibull distribution characterized by a scale parameter λ and a shape parameter k . The

probability density function (PDF) for wind speeds is given by:

() [ ( )]
k−1 k
k v v
f ( v ; k , λ )= exp −
λ λ λ

where v represents the wind speed. For wind speed, the shape parameter k typically ranges between

2.0 and 2.4, while the scale parameter λ can be estimated from the mean wind speed v using the

relationship:

v
λ=
Γ 1+ ( 1k )
Where Γ ( ⋅ ) represents the gamma function.

1.3 Incorporation of Seasonal Variability

To reflect seasonal fluctuations in wind conditions, a multiplicative seasonal adjustment factor sm

and a multiplicative hourly adjustment sh are applied for each hour h of each month m. This factor
scales the theoretical average wind speed, producing the adjusted expected wind speed for each

hour h in month m:

v m ,h=v × s m × sh

This approach ensures that the model captures variations in wind speeds that typically occur due to

seasonal meteorological patterns.

1.4. Conversion of Wind Speed to Power Output Using the Turbine Power Curve
The translation of wind speed into electrical power output is governed by the turbine's power

curve, which defines power generation across different wind speed ranges. This power curve

consists of three critical speed thresholds:

 Cut-in speed ( v cut-in): The minimum wind speed required for the turbine to start generating

power.

 Rated speed ( v rated): The wind speed at which the turbine reaches its maximum (rated)

power output.

 Cut-off speed ( v cut-off): The maximum wind speed beyond which the turbine ceases

operation for safety reasons.

The power output P ( v ) at a given wind speed v is modeled using the following piecewise function,

which is approximated to increase linearly between the cut-in and the rated wind speeds:

For each hour h in month m , the corresponding power output is computed as:

Pm , h=P ( v m , h )

1.5 Aggregation of Hourly Power Output to Monthly Energy Production

To determine the total energy production over a given month, the hourly power outputs are

aggregated. For a month m with H m hours, the total energy production Em is given by:
 Hm
Em = ∑ Pm , h
h=1

The maximum theoretical energy production for the month, assuming continuous operation at rated

power, is expressed as:

E max ,m=P rated × H m

Using this, the capacity factor (C F m) for month m is calculated as:

Em
C F m=
Emax , m

This metric provides a measure of the turbine’s performance relative to its maximum possible

output over the month.

1.6 Adjustment for System Losses

To derive the net capacity factor, adjustments must be made to account for various system losses,

including transmission losses, turbine availability, and operational inefficiencies. If L represents the

total loss factor (expressed as a fraction), the net capacity factor is computed as:

NC F net ,m =C F m × ( 1−L )

This final value represents the deterministic net capacity factor for month m , incorporating both the

available wind resource and system performance constraints.

1.7 Example graphs for a given location and turbine model

2. Power prices in the deterministic case
2.1 Capture prices
In the deterministic simulation framework, power prices are projected based on forecasts provided

by a market advisor. These forecasts specifically represent capture prices for onshore wind in the

selected country.

Capture prices refer to the actual prices that a given technology, such as onshore wind, effectively

receives for its generated electricity. Unlike wholesale market prices, which represent the average

electricity price across all generation technologies, capture prices adjust for factors that influence

the realized revenue of intermittent renewable energy sources.

A primary driver of this adjustment is cannibalization, which occurs when an increasing

share of wind energy in the electricity mix depresses the market price during periods of high wind

generation. Since wind turbines often generate power simultaneously due to similar meteorological

conditions, high wind output can lead to periods of oversupply in the electricity market. This

oversupply results in lower market-clearing prices precisely when wind farms are producing the

most electricity. As a result, the average price that wind generators "capture" is lower than the

broader wholesale price.

2.2 Structure of market forecast

The market advisor provides three forecast scenarios for capture prices to reflect potential market

conditions and uncertainty:

  Central Case: This represents the expected or base-case price trajectory, assuming

reasonable market conditions based on current policies, supply-demand dynamics, and

expected technological deployment rates.

 High Case (90th percentile): This scenario reflects a favorable price outlook, where capture

prices are in the upper decile of possible outcomes. It assumes conditions such as stronger

electricity demand, slower renewable capacity deployment, or higher fossil fuel prices,

leading to reduced cannibalization and higher realized prices for wind power.

 Low Case (10th percentile): This scenario represents a downside price trajectory, where

capture prices are in the lower decile of expected outcomes. This may be driven by factors

such as accelerated renewable energy deployment, prolonged periods of low electricity

demand, or market saturation effects, leading to greater cannibalization and lower prices for

wind generation.

In summary, the use of capture price forecasts, rather than wholesale prices, ensures that the

simulation reflects the economic reality faced by wind generators, particularly in markets where

renewable penetration is increasing.

2.3 Example graphs for a onshore wind capture prices forecast in a given country

Monthly prices - base case

160.0
140.0
120.0
100.0
80.0
60.0
40.0
20.0
-
25 26 27 29 30 31 33 34 35 37 38 39 41 42 43 45 46 47 49 50 51 53 54 55 57 58 59
/20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20 /20
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
4/ 8/ 12/ 4/ 8/ 12/ 4/ 8/ 12/ 4/ 8/ 12/ 4/ 8/ 12/ 4/ 8/ 12/ 4/ 8/ 12/ 4/ 8/ 12/ 4/ 8/ 12/
 Price average per month
80.00
70.00
60.00
50.00
40.00
30.00
20.00
10.00
-
1 2 3 4 5 6 7 8 9 10 11 12

3. Wind production and prices in the simulation case

This section describes the simulation framework used to generate monthly production and price

series by incorporating stochastic variability while preserving the observed correlation between

these variables. The methodology builds upon the deterministic base cases for production and prices

and introduces randomness in a statistically coherent manner.

3.1 Deterministic Base Cases and Correlation Estimation

Let X base represent the deterministic (base-case) monthly production and Y base the deterministic

monthly price. The relationship between these two variables is quantified using the empirical

correlation coefficient, denoted as ρ , which is computed as:

Cov ( X base , Y base )

ρ=
σ X σY

where Cov ( X base , Y base ) represents the covariance between monthly base-case production and

prices, while σ X and σ Y denote the standard deviations of monthly base-case production and prices,

respectively. This correlation coefficient serves as a key parameter to ensure that the stochastic

variations in production and price remain consistent with observed historical patterns.
3.2 Stochastic Modeling of Production and Price Distributions
For each forecasted month, production and price are modeled as random variables drawn from

normal distributions. The mean of each distribution corresponds to the deterministic base-case

value, while the standard deviation is estimated separately for production and prices.

3.2.1 Production Standard Deviation Estimation

The standard deviation of production, σ prod, is estimated based on the variability of hourly

production within a given month. If the set of hourly production values in a month is denoted as

\{ P 1 , P2 ,… , P H \}, where H represents the total number of hours in the month, then σ prodis

computed as:

σ prod=std ( P1 , P2 , … , P H )

This approach ensures that the simulated production variability is consistent with observed intra-

monthly fluctuations.

3.2.2 Price Standard Deviation Estimation

The standard deviation of prices, σ price, is estimated using the high and low case price forecasts

provided for each month. These forecasts are assumed to represent a specified percentile range,

typically the 90th percentile (high case) and 10th percentile (low case). Given these percentile

bounds, σ price is approximated as:

High−Low
σ price=
2z

where z is the z-score corresponding to the percentile difference. For the 10th and 90th

percentiles, the value of z is approximately 1.28.

Alternatively, if historical monthly price data are available, σ price can be directly computed

as the standard deviation of those observed values.

3.3 Preservation of Correlation in Stochastic Simulations

To ensure that the simulated production and price values maintain the observed correlation ρ , a

method based on correlated standard normal variables is employed.

First, two independent standard normally distributed random variables are generated:

 Z prod, representing production variability.

 Z ind, an independent standard normal variable.

The standard normal variable corresponding to price, Z price, is then constructed using the following

transformation:

Z price= ρ Z prod + √ 1−ρ2 Z ind

This ensures that Z price maintains the desired correlation ρ with Z prod while preserving the stochastic

nature of the simulation.

3.4. Generation of Simulated Production and Prices

The final simulated monthly production and price values are derived by scaling the standard normal

variables with their respective standard deviations and adding the deterministic base-case values:

Prod sim =X base +σ prod × Z prod , Price sim=Y base +σ price × Z price

These formulas ensure that the simulated values:

1. Are centered around the deterministic base-case values while introducing stochastic

variability.

2. Exhibit variability consistent with observed fluctuations in production and prices.

3. Maintain the correlation between production and prices.

4. Interest rates
A short rate, such as EURIBOR or SONIA, is assumed as the base rate for the project financing,

with an interest rate margin added to determine the periodic interest rates which determine the

interest payment owed to the lenders for the period. By modeling the stochastic behavior of the

short rate, interest rate uncertainty can be introduced into the model. This approach captures the

inherent volatility and mean-reverting nature of short rate, thereby allowing the simulation to reflect

realistic fluctuations in financing costs over time. The dynamics of the short rate are modeled using

the Vasicek framework. The Vasicek model is a mean-reverting stochastic process that captures the
tendency of short-term interest rates, such as short rate, to revert toward a long-run average. The

continuous-time dynamics of the short rate r t are described by the stochastic differential equation

d r t =a ( b−r t ) dt + σ d W t

where a is the speed of mean reversion, b is the long-run mean level of the interest rate, σ is the

volatility, and d W t is the increment of a standard Brownian motion. In this formulation, the term

a ( b−r t ) represents the deterministic drift that pulls the rate toward the mean b , while σ d W t

introduces random fluctuations around this path.

For practical implementation in a Monte Carlo simulation with monthly time steps, the

1
continuous-time model is discretized. Denoting the time step by Δ t (with Δ t= for monthly
12

intervals), the discrete-time version of the Vasicek model is given by

r t +1=r t +a ( b−r t ) Δ t+ σ √ Δ t ϵ t

where ϵ t is a standard normal random variable, ϵ t ∼ N ( 0 , 1 ). The first term, r t , carries the

current rate forward; the second term, a ( b−r t ) Δ t , adjusts the rate toward the long-run mean b over

the time step; and the third term, σ √ Δ t ϵ t , adds stochastic variability scaled by the volatility σ and

the square root of the time increment.

The base case for the simulation is established using the prevailing short rate, which serves

as the initial value r 0 . The parameters a , b , and σ are calibrated from historical short rate data. For

example, b is set equal to the historical long-run average of short rate, σ is estimated from the

historical standard deviation of the rate, and a is chosen to reflect the observed speed at which short

rate reverts to its mean.

At each monthly time step, the interest rate is updated according to the discretized equation

above. By iterating this process over the forecast horizon, a stochastic path for short rate is

generated that captures both its mean-reverting behavior and its inherent volatility. This simulated

short rate path is then used as an input in the overall Monte Carlo simulation of the project’s cash

flows.
 In summary, the modeling of short rate via the Vasicek model ensures that the simulated

interest rates are anchored to a deterministic base rate while incorporating realistic dynamics of

mean reversion and random fluctuations. The discrete-time formulation

r t +1=r t +a ( b−r t )
1
12 √
+σ
1
ϵ
12 t

provides a tractable and theoretically sound method for simulating short rate over monthly intervals,

making it suitable for long-term financial analysis in a Monte Carlo framework.

5. Construction delay
The following methodology describes the modeling of months of delay as a stochastic input in the

financial model for the Monte Carlo simulation. It is assumed that the delay, measured in months,

follows a Poisson distribution. The Poisson distribution is well suited for modeling the number of

discrete events occurring over a fixed time period, when these events occur independently and at a

constant average rate.

Let D denote the random variable representing the number of months of delay. The Poisson

probability mass function (PMF) is given by

k −λ
λ e
P ( D=k )= , k =0 , 1 ,2 , … ,
k!

where λ is the average number of months of delay, estimated from historical data or expert

judgment.

In this framework, the parameter λ represents the mean delay (in months) expected in the

project schedule. For each simulation run, a value for D is drawn from the Poisson distribution,

thereby providing a discrete delay duration. In practical terms, if the average delay is assumed to be

2 months (i.e., λ=2), then the probability of observing 0, 1, 2, 3, etc., months of delay is calculated

using the PMF above.

In order to obtain a simulated delay value, denoted Dsim , from the Poisson distribution, the inverse

transform method is applied. This method starts with a random draw U from a uniform distribution
on the interval [ 0 , 1 ] and uses the cumulative distribution function (CDF) of the Poisson distribution

to determine the corresponding discrete outcome.

The Poisson distribution with parameter λ has the probability mass function (PMF)
k −λ
λ e
P ( D=k )=
k!

and the corresponding CDF is given by

k
λ j e−λ
F ( k )=∑
j=0 j!

The simulation process proceeds as follows:

1. Draw a random number U from the uniform distribution on [0 ,1] , i.e., U ∼ Uniform ( 0 ,1 )

2. Set k =0 and compute the initial cumulative probability

0 −λ
λ e −λ
F ( 0 )= =e
0!

3. While F ( k ) <U , increment k by 1 and update the cumulative probability

k −λ
λ e
F ( k )=F ( k−1 ) +
k!

4. Once F ( k ) ≥ U , set the simulated delay Dsim =k .

This algorithm ensures that the discrete value k is selected with the correct probability

according to the Poisson distribution. By drawing from a uniform distribution and using the

cumulative probability, the inverse transform method maps the continuous uniform variable U into

a discrete outcome that represents the number of months of delay.