Gas Storage at Calgary

Natural Gas Storage Assets:
Value, Trading Strategy and Financial Investment
Youyi Feng1 James Pang2
1
Department of System Engineering and Engineering Management
Chinese University of Hong Kong, Hong Kong, China
2
Department of Mathematics and Statistics
University of Calgary, Calgary, Alberta, Canada
Seminar of Finance Lab, 2008
Y. Feng, Z. Pang Gas Storage Valuation

Outline
1 Introduction
Fundamentals of Natural Gas Storage
Literature Review
2 The Model
Valuing a Peak-load Storage Facility
Optimal Trading Strategy in a Dynamic Context
Valuing a Base-Load Storage Facility
Valuing a Firm Storage Contract
3 Numerical Analysis
Example 1: Valuing a natural gas storage facility
Example 2: Valuing a firm storage contract

Outline
1 Introduction
Literature Review
2 The Model

Outline
1 Introduction
Literature Review
2 The Model

1 Storage’s role is changing

Traditional role: balancing the variability in demand and smoothing the
production/supply.
After deregulation: independent charged service and arbitrate instruments
2 Natural Gas Storage Players
Local distribution company (LDC) (obligated to serve, focuses on operational
needs )
Energy marketer (Arbitrager and speculator)
3 Operational Characteristics
Base (cushion) and working gas capacities;
Deliverability, injection/withdrawal rates
Cycling
4 Cost Structure
Injection/withdrawal variable cost (≤ 2% per mmBtu)
Injection/withdrawal fuel cost (≤ 1%)

1 Classification by Operational Characteristics

Depleted Reservoir Storages (Low capital investment, 50% base gas, low
deliverability, 80% )
Aquifer Storages (high deliverability, 80% base gas, less popular)
Salt caverns or salt dome storages (low base gas [25%], high deliverability,
high capital investment)
2 Classification by Functions
Base load facilities (Depleted gas storages)
Peak load storage facilities (Salt caverns)
3 Natural gas storages add value in two ways.
Arbitrage mechanism that allow the exploration of the time spread (buy low
in summer and sell high in winter). (Intrinsic value)
Operational flexibility to react to price fluctuations. (Extrinsic value)
Hedging affects optionality
Valuing the storage with spot trading strategy (instead of forward-based
valuation).

Relation to American and Swing Options
Storage contract is similar to American options in timing decision of the

exercises (injection/withdraw rights); but with more decision options and
volumetric restrictions at any time.
Similar to swing options with multiple exercises rights but with physical
constraints and transaction costs.

Leasing Contracts of Natural Gas Storage
1 Term of Service, Injection/Withdrawal Dates

2 Capacities
Maximum Storage Quantity (MSQ) ($MDth)
Maximum Daily Injection Quantity (MDIQ) ($ per Dth/d)
Maximum Daily Withdrawal Quantity (MDWQ) ($ per Dth/d)
3 RATES AND CHARGES (monthly)
Monthly Reservation Charge:
Capacity Reservation Rate ($ per Dth/mo)
Deliverability ($ per Dth/mo)
Usage Charge:
Injection Rate ($ per Dth)
Withdrawal Rate ($ per Dth)
Inventory Charge ($ per Dth/day)
Authorized Overrun Charge
Fuel Reimbursement: In Kind

Related Literature
1 Real options: Dixit and Pindyck [1994], Trigeorgis [1996], Smit and
Trigeorgis [2004], and Amram and Kulatilaka [2005])
2 Existing literature on gas storage valuation
Financial derivative based valuation: rolling intrinsic approach (Gary and
khandelwal [2004]), calendar spread call options (Eydeland and Wolyniec
[2003], virtual storage (Ahn et al. [2002]).
Computational contribution: finite difference approach (Ahn et al. [2002],
Thompson et al. [2003], Weston [2002], Chen and Forsyth [2006]), Monte
Carlo simulation approach (see e.g., Ludkovski and Carmona [2005],
Boogert and de Jong [2006]) and stochastic programming (Nowak and
Römisch [2000]).
3 Inventory models: Song and Zipkin (1993) and Li, Porteus and Zhang
(2001).
4 Our focus: Characterizing the structure of the storage model and the
optimal spot trading strategy, and the impacts of operational
characteristics and market dynamics, to exploit its financial implications
as the fundamentals of investment analysis.

The Model
1 Introduction
Literature Review
2 The Model

Model Specifications
Consider a marketer who is rational and risk-neutral and rents a storage

facility to trade in energy market, aiming to maximize total expected profit
over the finite horizon of her rental. The financial market is complete and
the marketer is a price-taker.
Time horizon T , risk free interest rate r .
Inventory level of working gas at time t is Xt with the realization x.
Assume that the maximum injection rate α(x) is strictly decreasing and
concave in x and the maximum withdrawal rate β(x) is strictly increasing
and concave in x.
Let q ∈ [−β(x)1{x >0} , α(x)1{x <M} ] be the amount of gas being injected
to (q > 0) or release from (q < 0) storage.
Fuel charges (proportion) for injection/withdrawal L(q) = ρi q + + ρw q − ,
ρi , ρw ∈ [0, 1).
Injection and withdrawal charge C(q) = ci q + + cw q −
Inventory holding cost h(x, s) is increasing and convex in x for any spot
price s.
Terminal value ν(x, s) is non-decreasing and concave in x for any s.

Dynamics of Spot Price
Risk-adjusted natural gas spot price follows a Markov process {St }

which can be described by the following stochastic differential equation:
K
X
dSt = µ(St , t)dt + σ(St , t)dzt + χk (t, St , ξk )dNk (t)
k =1
(For the sake of simplicity, we consider only the single-factor model.)

Some canonical examples:
GBM Process: µ(St , t) = µSt , σ(St , t) = σSt
Ornstein-Uhlenbeck Process: µ(St , t) = κ(η − St ), σ(St , t) = σ or
σ(St , t) = σSt .
Geometric Mean-Reverting Process:
µ(St , t) = κ(η − log(St ))St , σ(St , t) = σSt

Dynamics of Storage System
Let π(t, Xt , St , qt ) the instant payoff rate at time t given the inventory level Xt ,
spot price St and the injection/withdrawal rate qt . Then, the dynamics of
payoff rate and inventory level can be represented as follows.

 Injection (qt > 0): π(t, Xt , St ; qt ) = −St qt (1 + ρi ) − ci qt − h(Xt , St ) dXt = qt dt
Store (qt = 0): π(t, Xt , St ; qt ) = −h(Xt , St ) dXt = 0
 Withdrawal (q < 0): π(t, X , S ; q ) = −S q (1 − ρ ) + c q − h(X , S ) dXt = qt dt
t t t t t t w w t t t

The Stochastic Control Problem

RT
Define J u (t, x, s) = E e−r (τ −t) π(τ, Xτu , Sτ ; qτu )dτ + e−rT ν(XTu , ST )S(t) = s

t
for any admissible strategy u ∈ U .

The optimal stochastic control problem is
J(t, x, s) = sup J u (t, x, s)
u∈U
subject to the dynamic of

K
X
dSt = µ(St , t)dt + σ(St , t)dz + χ(t, St , ξk )dNk , S0 = s,
k =1
dXtu = (qtu + L(qtu , Xtu ))dt, qtu ∈ [−β(Xtu )1{X u >0} , α(Xtu )1{X u <M} ],
t t
and terminal and boundary conditions

J u (T , x, s) = ν(x, s), (1)
lim J u (t, x, s) → 0, and u
lim Jss (t, x, s) → 0, (2)
s→∞ ss s→0

The HJB Equation
0 = L J(t, x, s) + max {−s[q + L(q)] − C(q, x) + qJx (t, x, s)} − h(x, s)

q∈[−β(x ),α(x )]
where
K
1 2 X
L J(t, x , s) = Jt (t, x , s)+ σ (s, t)Jss (t, x , s)+µ(x , t)Js (t, x , s)+ λk [EV (t, x , s+χk (t, x , ξk ))−V (t, x , s)]
2 k =1
The optimal control:


 α(x)
 if (t, x, s) ∈ IR
q ∗ (t, x, s) = −β(x) if (t, x, s) ∈ WR

0 if (t, x, s) ∈ NT.
where
IR = {(t, x, s) ∈ D : Jx (t, x, s) > s(1 + ρi ) + ci },
WR = {(t, x, s) ∈ D : Jx (t, x, s) < s(1 − ρw ) − cw },
NT = {(t, x, s) ∈ D : s(1 − ρw ) − cw ≤ Jx (t, x, s) ≤ s(1 + ρi ) + ci }.

The Structure Properties
Theorem
1 J(t, x, s) is strictly concave in x for any fixed (t, s).

2 For any fixed (t, s), there is an inject-up-to gas load level I(t, s) and a
withdraw-down-to level W (t, s), which are given by

I(t, s) = inf x ∈ [0, M) : Jx (t, x, s) ≤ (1 + ρi )s + ci ,

W (t, s) = inf x ∈ (0, M] : Jx (t, x, s) ≤ (1 − ρw )s − cw .
Further, I(t, s) ≤ W (t, s). It is optimal to inject gas at rate α(x) if and
only if x < I(t, s), and to withdraw gas at rate β(x) if and only if
x > W (t, s), and to do nothing otherwise. That is,

 a(t, x, s) if x ∈ [0, I(t, s))
q ∗ (t, x, s) = 0 if x ∈ [I(t, s), W (t, s)] .
−w(t, x, s) if x ∈ (W (t, s), M]


Structure of Optimal Trading Strategy
x
M
Withdrawal
β (x)
W(t,s)
No Transaction
I(t,s)
Injection
α (x)
0
s

The Model
1 Introduction
Literature Review
2 The Model

Some Facts
Is it always true to buy low and sell high ?

Monotonicity on Price Dynamics
Assumption
(A) The spot price process {S(t) : t ≥ 0} is stochastically increasing
in the sense that for any s1 ≥ s2 and any s ≥ 0 and t > t ′
P[S(t) ≥ s|S(t ′ ) = s1 ] ≥ P[S(t) ≥ s|S(t ′ ) = s2 ].
Theorem
Suppose (A) holds.

1. If µ(t, s) = κ(η − s)(mean-reverting drift), q(t, x, s) is decreasing in s.
2. If µ(t, s) = µs (exponential growth drift) and µ > r holds, q(t, x, s) is
increasing in s.

Pressure of Contract Duration
Theorem
When implementing the optimal operating strategy, the present value of
dynamic marginal storage profit e −rt Jx (t, Xt , St ) is a super-martingale, i.e., for
any time interval δ ∈ (0, T − t),
E[e−r (t+δ) Jx (t + δ, Xt+δ , St+δ )|St = s, Xt = x] ≤ e−rt Jx (t, x, s). (3)
Theorem
Suppose h(x, s) = 0 and ν(x, s) = 0.

J(t, x, s) is increasing in x for any (t, s), J(t, x, s) is decreasing in t for
any (x, s). Moreover, Jx (t, x, s) is decreasing in t.
For any x and s, q(t, x, s) is decreasing in t).

The Model
Valuing a Base-Load Storage Asset
1 Introduction
Literature Review
2 The Model

Valuing a Base-Load Storage Asset
Low injection injection and deliverability rates ( seasonal storage such as

depleted reservoirs )
Single-cycle per year which turnovers twice annually.
The injection season is from April to October and the withdrawal season
is from November to March in the following year.
The HJB equation: for t ∈ [τ, T ] and x < M,
0 = L˜J(t, x, s) − h(x, s) + max {−s(1 + ρi )q − C(q, x) + qJx (t, x, s)},
q∈[0,α(x )]
and for t ∈ [0, τ ] and x > 0,

0 = L˜J(t, x, s) − h(x, s) + max {−s(1 − ρw )q − C(q, x) + qJx (t, x, s)}.
q∈[−β(x ),q]

Valuing a firm storage contract
1 Introduction
Literature Review
2 The Model

Sensitivity of Operational Constraints
Theorem
Defining a firm storage contract with C = (α, β, M, ci , cw , p).

1. At a time t, J(t, x, s; C) is concave in (x, α, β, M) and supermodular in
(x, −α), (x, β) and (x, M), respetively .
2. Optimal thresholds I(t, s; C) and W (t, s; C) are increasing in M and β,
respectively, and decreasing in α. Thus, q(t, x, s) = a(t, x, s) − w(t, x, s)
is increasing in M and β and decreasing in α.
3. J(t, x, s; C) is submodular in (α, ci ) and (β, cw ), respectively.

Optimal Capacity Decision
Let p be the marginal capacity reservation fee, then the investment

return: Z T
Π(C) = J(0, x0 , s0 ; C) − e −rt pMdt
0
Theorem
1. Π(C) is increasing and concave in α and β, and concave in M. In
addition, it is submodular in (α, ci ), (β, cw ), and (M, p), respectively.
2. For fixed (α, β, ci , cw ), if there is no limit on the contractual capacity, then
the optimal capacity of an FSS contract, M ∗ (p), is determined by the
first-order condition:
∂Π(M ∗ (p), α, β, p, ci , cw )
= 0.
∂M
In addition, M ∗ (p) is decreasing in p.

Numerical Analysis
1 Introduction
Literature Review
2 The Model

Example
A natural gas storage facility
Capacity (M=1Bcf), injection cost ($0.0218/mmBtu), withdrawal cost ($
0.0195/mmBtu). There is 3.59% injection fuel loss and no withdrawal
fuel cost.
Maximum injection/withdrawal rates are stepwise functions:
x/M α(x)(Btu/month) x/M α(x)(Btu/month)
0 − 10% 0.167M 0 − 10% 0.250M
50 − 100% 0.140M 10 − 16% 0.333M
16 − 30% 0.375M
30 − 35% 0.475M
35 − 100% 0.500M
Table: Injection/Withdrawal Rates (Ratchets) a
a
This example is provided by Dr. Kevin G. Kindall, Commercial Division, ConocoPhillips

Example
Price Calibration
We refer to Thompson et al. (2003) for a calibrated gas price process
with jumps:
dSt = 0.25(2.5 − St )dt + 0.2St dWt + (ξ − St )dQ.
where Q is a Poisson process with intensity 2 and ξ ∈ N(6, 4).

The risk-free interest rate is 10% per year and T = 1 (year).
The terminal value function ν(x, s) = 0.

Value of Storage
7
x 10
1.8
1.6
1.4
Value of Storage ($)
1.2
0.8
0.6
0.4
0.2
1
20
15
0.5
10
Working Gas in Storage (Bcf) 5 Gas Price ($/MMBtu)

0 0
Figure: The Value of Storage

Net Withdrawal Rate Net Withdrawal Rate
6
6
5
5
Net Withdrawal Rate (Bcf/Year)

Net Withdrawal Rate (Bcf/Year)
4
4
3
3
2
2
1
1
0
0
−1
−1
−2
−2
−3
−3
1
20
1 20 15
15 0.5
10
0.5 10 Working Gas in Storage (Bcf) 5 Gas Price ($/MMBtu)
Working in Gas in Storage (Bcf) 5
Gas Price ($/MMBtu) 0 0
0
1
Figure: Net withdrawal rate at t= 0 Figure: Net withdrawal rate at t= 4
Net Withdrawal Rate Net Wihdrawal Rate
6 6
5 5
Netwithdrawal Rate (Bcf/Year)
Net Withrawal Rate (Bcf/Year)

4 4
3 3
2 2
1 1
0 0
−1 −1
−2 −2
−3 −3
1 1
20 20
15 15
0.5 0.5
10 10
5 Working Gas in Storage (Bcf) 5
Gas Price ($/MMBtu)
Working Gas in Storage (Bcf) 0 0 Gas Price ($/MMBtu) 0 0
1 3
Figure: Net withdrawal rate at t= 2
Figure: Net withdrawal rate at t= 4

Numerical Analysis
1 Introduction
Literature Review
2 The Model

Example
Price calibration and contract specification
We refer to de Jong and Walet (2003) for a calibrated gas price process :
d log(St ) = 17.1(log(3) − log(St ))dt + 1.33dWt .
Total capacity (8 Bcf), duration (T = 1 year), initial inventory level

(x0 = 4Bcf ).
The terminal payoff ν(x, s) = −2 · s · max{4 − x, 0}.
Maximum injection rate α(x) = 0.06 · 365(Bcf /year ) and the maximum
withdrawal rate β(x) = 0.25 · 365(Bcf /year ).
injection/withdrawal costs ci = $0.02/MMBtu, cw = $0.01/MMBtu,
injection/withdarwal fuel charges ρi = ρw = 1%.
The risk-free interest rate is 0.06% per year.

Example
Sensitivity to price characteristics
7 7
x 10 x 10
1.8
2 1.6
Value of Storage (US$)

Value of Storage (US$)
1.4
1.5
1.2
1 1
s=1
s=2
s=3 s=1
0.8
s=2
s=4
0.5 s=3
s=5
s=4
s=6 0.6
s=5
s=7 s=6
s=7
0 0.4
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 0 2 4 6 8 10 12 14 16 18 20
Volatility (σ) Mean−reverting Rate (κ)
Figure: Sensitivity to volatility Figure: Sensitivity to mean-reverting rate

Example
Sensitivity to operational constraints
Value of Storage (V (0, 4, s)) (106 $)
α β M Initial Spot Price ($/MMBtu)
1 2 3 4 5 6 7
0.01 0.25 8 2.5423 2.3357 2.2564 2.4056 3.1484 4.8093 7.2195
0.05 0.25 8 9.4928 8.5354 8.1254 8.5653 9.9344 12.0395 14.7544
0.1 0.25 8 13.1912 11.5808 10.8936 11.5381 13.0373 15.1968 17.9282
0.15 0.25 8 15.1745 13.1512 12.274 13.0264 14.5814 16.7584 19.4908
0.2 0.25 8 16.4279 14.1359 13.1374 13.9542 15.5374 17.724 20.4561
1 2 3 4 5 6 7
0.06 0.06 8 6.3395 5.3021 4.988 5.4565 6.0913 6.9094 7.8978
0.06 0.1 8 7.9626 6.8845 6.4894 7.0029 7.8918 9.0997 10.5956
0.06 0.15 8 9.1611 8.0631 7.6244 8.1445 9.2624 10.85 12.8533
0.06 0.2 8 9.9335 8.8256 8.3655 8.8745 10.152 12.035 14.4373
0.06 0.25 8 10.4716 9.3591 8.8823 9.3736 10.7798 12.9017 15.6221
1 2 3 4 5 6 7
0.06 0.25 5 7.1643 6.8641 6.9375 7.9434 9.5578 11.7898 14.5758
0.06 0.25 6 8.699 7.9732 7.78 8.572 10.1022 12.2908 15.0512
0.06 0.25 7 9.7442 8.7767 8.4163 9.0376 10.4985 12.6499 15.388
0.06 0.25 8 10.4716 9.3591 8.8823 9.3736 10.7798 12.9017 15.6221
0.06 0.25 10 11.3535 10.0849 9.4718 9.7825 11.1075 13.1852 15.8795

Summary
Valuation of natural gas storage needs to capture the extrinsic value as

well as intrinsic value.
Understanding the impacts of operational flexibility and market dynamics
on the options value and trading strategy is important.
How to identify momentum opportunity is key to success for a marketer.
We are still thinking about ......

How to price the storage interruptible contract?
Storage facility as a risk management instrument
A contracting game between a risk-averse storage owner and a risk-neutral
marketer
How storage capacity affects the energy (gas/electricity) prices?
Applying to oil markets, soft commodity markets and even electricity markets
(through pumped-storage hydropower stations).

Gas Storage at Calgary

Uploaded by

Copyright:

Available Formats

Gas Storage at Calgary

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Gas Storage at Calgary

Uploaded by

Copyright:

Available Formats

Natural Gas Storage Assets:

Value, Trading Strategy and Financial Investment

Youyi Feng1 James Pang2

Seminar of Finance Lab, 2008

Y. Feng, Z. Pang Gas Storage Valuation

Y. Feng, Z. Pang Gas Storage Valuation

Y. Feng, Z. Pang Gas Storage Valuation

Y. Feng, Z. Pang Gas Storage Valuation

1 Storage’s role is changing

Y. Feng, Z. Pang Gas Storage Valuation

1 Classification by Operational Characteristics

Y. Feng, Z. Pang Gas Storage Valuation

Storage contract is similar to American options in timing decision of the

Y. Feng, Z. Pang Gas Storage Valuation

1 Term of Service, Injection/Withdrawal Dates

Y. Feng, Z. Pang Gas Storage Valuation

Y. Feng, Z. Pang Gas Storage Valuation

Y. Feng, Z. Pang Gas Storage Valuation

Consider a marketer who is rational and risk-neutral and rents a storage

Y. Feng, Z. Pang Gas Storage Valuation

Risk-adjusted natural gas spot price follows a Markov process {St }

(For the sake of simplicity, we consider only the single-factor model.)

Y. Feng, Z. Pang Gas Storage Valuation

Y. Feng, Z. Pang Gas Storage Valuation

for any admissible strategy u ∈ U .

subject to the dynamic of

and terminal and boundary conditions

Y. Feng, Z. Pang Gas Storage Valuation

0 = L J(t, x, s) + max {−s[q + L(q)] − C(q, x) + qJx (t, x, s)} − h(x, s)

The optimal control:

Y. Feng, Z. Pang Gas Storage Valuation

1 J(t, x, s) is strictly concave in x for any fixed (t, s).

Y. Feng, Z. Pang Gas Storage Valuation

Y. Feng, Z. Pang Gas Storage Valuation

Y. Feng, Z. Pang Gas Storage Valuation

Is it always true to buy low and sell high ?

Y. Feng, Z. Pang Gas Storage Valuation

P[S(t) ≥ s|S(t ′ ) = s1 ] ≥ P[S(t) ≥ s|S(t ′ ) = s2 ].

Suppose (A) holds.

Y. Feng, Z. Pang Gas Storage Valuation

Suppose h(x, s) = 0 and ν(x, s) = 0.

Y. Feng, Z. Pang Gas Storage Valuation

Y. Feng, Z. Pang Gas Storage Valuation

Low injection injection and deliverability rates ( seasonal storage such as

and for t ∈ [0, τ ] and x > 0,

Y. Feng, Z. Pang Gas Storage Valuation

Y. Feng, Z. Pang Gas Storage Valuation

Defining a firm storage contract with C = (α, β, M, ci , cw , p).

Y. Feng, Z. Pang Gas Storage Valuation

Let p be the marginal capacity reservation fee, then the investment

Y. Feng, Z. Pang Gas Storage Valuation

Y. Feng, Z. Pang Gas Storage Valuation

Y. Feng, Z. Pang Gas Storage Valuation

dSt = 0.25(2.5 − St )dt + 0.2St dWt + (ξ − St )dQ.

where Q is a Poisson process with intensity 2 and ξ ∈ N(6, 4).

Y. Feng, Z. Pang Gas Storage Valuation

Working Gas in Storage (Bcf) 5 Gas Price ($/MMBtu)

Figure: The Value of Storage

Y. Feng, Z. Pang Gas Storage Valuation

Net Withdrawal Rate (Bcf/Year)