Regional Routing Model Review:
C) Model Formulation and Scenario Analysis
Frank SouthworthOak Ridge National LaboratoryOak Ridge, TN 37831
NETS Program ReviewDecember 12, 2005Washington DC
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
Calibration, Forecasting & Scenario Analysis
County/Port Based Mode/Route/Market Choice Model(s)*
Calibration
Forecast / Scenario Based Commodity Flows, Costs and Benefits
Forecasting,Scenario Analysis
* = Simultaneous or nested mode and destination choice linked to capacity constrained route assignment
Base Case Computed Flows, Costs (and Benefits)
County/Port Based Commodity Production /Consumption Forecasts
NETS Tier 1 Regional Economic Activity Forecasts/ Scenarios
Changes in Demands
Fuel, labor, I&M costs by vehicle /vessel types (C,M,V) (Data)
Changes to Network Conditions (Capacities) and Mode/Route Costs
Network Changes
Mode Specific Rate Estimation Models
Transit times, distances
dollar/ton shipment rates
Changes in Supply
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
Estimate Commodity Production (O) & Consumption (D) by Region
Connect Os and Ds to Estimate O-to-D Commodity Flows
Assign O-to-D Flows to Modes & Routes
Re-Estimate O-to-D Costs per Ton
Estimate O-to-D Costs per Ton
Re-Estimate O-to-D Flows
Iterate toConvergence
The Basic Flow and Cost Estimation Process
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
Vak
)d( S XVZMinimize ak
akXV 0, ),(
+ m 1/ βm ( Xi jm ln Xi j
m) + m 1/ λm ( Xi jkm ln Xi j
km)
Prototype Regional Routing Model Formulation:
subject to:
Vak = ∑i ∑j ∑r δ i,j
akr X i j kr for all links, a, and modes k, in the
network
∑r X i j
rk = X i j k for i=1,2,...I, and j=1,2,...J
∑k X i j
k = X i j for i=1,2,...I, and j=1,2,...J
Vak ≥ 0
X i j kr ≥ 0
V = a link volumeX = an O-D flow volumeS = link transportation time orgeneralized cost
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
Freight Destination Choice (Flows Modeling):
Xi jm = Oi
m Dj m F[ βm, ci j
m ] Aim Bj
m
Aim = 1/{j Bj
m Djm
F [ βm, c i j m ]} for all i
Bj m = 1/{i Ai
m Oim F [ βm, i j
m ]} for all j c
i j ( Xi jm di j
m } / i,j Xi jm = d * m
where
and
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
Freight Mode Choice:
Xijkm = Xij
m *[exp(-θm ckm)/ ∑kεK(m) exp(-θm ckm)]
cijkm = α0 + α1 rij
km + α2 Sijkm + α3 vij
km
where (for example):
and,
c ijm = -(1/ λm ) ln {k exp (-λm c i j
km )}
(links mode and destination choice)
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
Components of Freight Costs that Need Modeling:
Number of different “legs” to a journey
Shipper/receiver perceived costs per leg:
freight rate transit time service reliability
Congestion effects
i.e. congestion transit time and reliability direct plus indirect (i.e. rate) effects on costs
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
14
14.5
15
15.5
16
16.5
17
17.5
18
18.5
19
0 500 1000 1500 2000
S=
Tra
ve
l T
ime
(M
inu
tes
)
Traffic Volume (Vehicles/hour)
S = S0 * (1+ 0.15 (V/C)**4)
x =Kilotons/Year (in thousands)
CC/2
S=
Min
ute
s o
f D
ela
y/T
ow
k
S = k /((C/x) –
200
400
600
800
1000
1200
1400
1600
1800
2000
0 5 10 15 20 25 30 35 40 45 50
CC/2
k
–1)
200
400
600
800
1000
1200
1400
1600
1800
2000
0 5 10 15 20 25 30 35 40 45 50
Lock Congestion Function Highway Congestion Function
Sa = Sao * [ 1 + θ1* Va + θ2 *( Va / Capa)γ ]
“Generic” link congestion function for use in Toy Model:
Network Link Transit Time (Congestion) Functions
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
Components of Freight Movement Costs
Freight Rates In-Transit Times Service Quality (Reliability)
Variable Operating Costs (Annualized)
Line-haul costsLoading/unloading costsEquipment utilization costs Commodity carrying costsAdministrative costs Shipment distance
Cargo type (commodity, weight, volume)Labor rateFuel price Carrier/operator type Equipment typeCompany type (private/for-hire; size)Contract type (duration)Facility type, Location,Operating licenses, fees and taxes
Capital Operating Costs
Generalized Transportation Costs (Shipper Perspective)
Level 1: (Statistical)
Level 2: (Engineering)
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
10
1112
2313
14
16
15
1718
19
20
21
22
24
25
30
26
27
28
29
31
32
33
1
2
45
7
8
9
3 6
42
Port A
Port B
Port C
Lock a
Lock b Lock c
TruckRailInland WaterDeep Water
Traffic Centroids(10 in all)
41
34
40
35
39
3637
38
40 Network Nodes (44)
Key:
Links (110 in all):
44
43
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
1
2
3
4 5
6
7 8
9
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
$$
$
$$
$
$
$
$
$
$
$
$
$
$
$
$
$
$
$
$
$ $
$
$ $
$
$
$
$
$
$
$
$
$
$
$ $
$
$
$
$
$$
#
#
#
#
#
#
##
#
#44
42
41
40
3837
36
33
32
3130
29
27
26
25
24
2221
20
19
17
16
1514
13
12
10
9
87
6
54
3
2
1
R R M T o y N etwo rkPort A
Port B
Port C
F o re ig nP o rt
L o ck1
Lock 2L o ck 3
Rail
Rail
Rail
R iver 1
River 2
C oastline0 10 20 30
M iles
FIELD_9
200000 100000 50000Model Run # 1
Origin Mode Split = 36.7% water 63.3 % rail
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
$$
$
$$
$
$
$
$
$
$
$
$
$
$
$
$
$
$
$
$
$ $
$
$ $
$
$
$
$
$
$
$
$
$
$
$ $
$
$
$
$
$$
#
#
#
#
#
#
##
#
#44
42
41
40
3837
36
33
32
3130
29
27
26
25
24
2221
20
19
17
16
1514
13
12
10
9
87
6
54
3
2
1
R R M T o y N etwo rkPort A
Port B
Port C
Fore ignP ort
Lock1
Lock 2Lock 3
Rail
Rail
Rail
R iver 1
R iver 2
C oastline0 10 20 30
M iles
FLOW S
150000 75000 37500Model Run # 2
Origin Mode Split = 74.9% water 25.1% rail
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
0
20000
40000
60000
80000
100000
120000
140000
160000
180000
200000
An
nu
al T
on
s R
ecei
ved
1 2 3 4 5 6 7 8 9 10
Destination Region
Destination Flow Comparisons
data input
model#1
model#2