Electric Power Systems Research 130 (2016) 230–240

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

Optimal energy management of a distribution network during the

course of a heat wave
Moein Choobineh a , Paulo C. Tabares-Velasco b , Salman Mohagheghi a,∗
a
Electrical Engineering and Computer Science Department, Colorado School of Mines, Golden, CO, United States
b
Mechanical Engineering Department, Colorado School of Mines, Golden, CO, United States

a r t i c l e i n f o a b s t r a c t

Article history: Heat waves are prolonged periods of excessive ambient temperature that may last up to several weeks. In
Received 27 June 2015 addition to posing health threats to the society, these events may easily push the power grid towards its
Received in revised form 6 September 2015 operational limits. The maximum capacity of many energy resources gets negatively affected by excess
Accepted 9 September 2015
temperatures. This can be in addition to the expected loss of life due to operation under harsh conditions.
Available online 29 September 2015
Overhead lines, on the other hand, experience excessive conductor surface temperatures that can dras-
tically reduce their power transmission capacity. To make matters worse, the reduction in generation
Keywords:
and/or transmission capacity will coincide with a rise in electric demand, often attributed to the overuti-
Distributed energy resources
Energy management
lization of air-conditioning systems. This can jeopardize the ability of the power grid to maintain system
Extreme weather event stability. A key to ensuring that the grid continues operating safely and securely is to incorporate the effect
Heat wave of temperature into its operation schedule. In this paper, we propose an optimal generation dispatch strat-
Natural disaster egy for a distribution grid exposed to a heat wave event, while taking into account the dependence of
operational constraints of various components on ambient temperature. We study a power grid equipped
with renewable and non-renewable distributed generation, battery energy storage, and demand respon-
sive loads. We evaluate the effectiveness of our proposed approach on a test system with data acquired
from the heat wave event of July 2006 in Sacramento, CA.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction • The health issues and consequent increase in heat-induced ill-

The last decades have witnessed severe changes in weather pat-
terns. Among extreme weather events, past instances of heat waves
in the United States especially in 2006 and 2012 [1,2], as well as the
2003 heat wave in Europe and some parts of Asia [3,4] and the 2009
heat wave in Australia [5] have gained widespread attention. Heat
waves are meteorological events that are generally considered as
a “prolonged period of excessive heat” [5]. Some researchers have
defined heat waves as temperatures ranging above a certain thresh-
old for two or more consecutive days, typically with heat threshold
between 95 and 99 percentiles [6]. However, in general there is no
common definition for a heat wave, and its interpretation may vary
with location, since people living in different regions usually have
a different perception of heat [5].
Several recent studies indicate considerable adverse effects that
can be caused in the aftermath of a heat wave. In general, these can
be attributed to any extended extreme temperature scenario:
∗ Corresponding author. Tel.: +1 303 2733501.
E-mail address: smohaghe@mines.edu (S. Mohagheghi).
nesses and mortality [1,4–6],
• Significant growth of power consumption especially for cooling
purposes and possible failure to meet the energy needs due to
fuel shortage [5,7],
• Increased greenhouse gas emissions per unit of energy consumed
[5],
• Failure in operation and availability of critical infrastructures
such as water sanitation systems [5] and transportation systems
[8],
• Dependency of steam power plants on cooling water which
becomes physically scarce during hot periods [9],
• Increased possibility of wildfire occurrence and consequent dam-
ages to the electrical and energy infrastructure [10].
Growth in electric demand, the need to meet the demand of
critical infrastructure, and the affected capacity or availability of
generation units are issues that are closely tied in with energy dis-
patch of the grid. Hence, adequate provisions should be planned
ahead in order to manage the energy resources in the most effective
fashion during the course of a heat wave. Extreme temperatures
have significant impacts on different power system components.

http://dx.doi.org/10.1016/j.epsr.2015.09.010
0378-7796/© 2015 Elsevier B.V. All rights reserved.
 M. Choobineh et al. / Electric Power Systems Research 130 (2016) 230–240 231

Nomenclature
Ptsub,max maximum allowable power passing through the
A. Indices transformer at the distribution substation at time
b index used for branches (lines) period t (kW)
i index used for distributed generation (DG) units; W
Pi,t output power of wind turbine i during time period
may be used for a solar panel, a wind turbine, or t (kW)
a traditional dispatchable DG unit SHGC solar heat gain coefficient
j index used for demand responsive (DR) loads SOCkB,max maximum limit of battery k state-of-charge (SOC);
k index used for battery energy storage systems battery is not allowed to be charged to higher than
m index used for the buses (nodes) in the network. this level
Node m may indicate node i with DG, node k with SOCkB,min minimum limit of battery k SOC; battery is not
battery, node j with DR load, or node q with non-DR allowed to be discharged to lower than this level
load. Hence, indices i, j, k and q may all be viewed as Tta ambient temperature during time period t (◦ C);
sub-indices for this index assumed to be the same for all components in the
q index used for non-controllable loads (i.e. non- power grid
demand responsive) TR PV reference temperature (◦ C)
t index used for time tTD time period during which battery works at temper-
ature T (months); assumed to be the same for all
B. Parameters battery systems
Ai area swept by the rotor of wind turbine i (m2 ) vi,t wind speed at wind turbine i during time period t
Bm mth row of the network susceptance matrix (m/s)
CP Albert Betz constant ˇjDR , jDR cost function coefficients for DR load j ($/kW and
ckB price of using battery k over its lifetime; this
$/kW2 )
value can be determined based on the total cost
␦k self-discharge rate for battery k
of purchasing and maintaining the battery and the B,max
Pk,T maximum power correction factor for battery k due
maximum amount of ampere hours (A h) it can
provide over its lifetime ($) to ambient temperature Tta (%)
DG,max
ctsub price of power purchased from the grid at time Pi,T maximum power correction factor for DG i due to
period t ($/kW) ambient temperature Tta (%)
fuel Line,max
ci price of fuel for generation unit i ($/m3 , $/l, $/kg) Pb,T maximum capacity correction factor for line b due
fuel
Hi net thermal value of the fuel (kW h/m3 , kW h/l, to ambient temperature Ta t (%)
kW h/kg) ˚i,t incident solar irradiance at PV panel i during time
kPV PV temperature coefficient of power (◦ C−1 ) period t (W/m2 )
KT temperature correction factor for the lifetime of a ˚STC solar irradiance at STC (W/m2 )
battery at temperature Tta B,c
k
charging efficiency of battery k (%)
LFk nominal battery lifetime for battery k (months) B,d
k
discharging efficiency of battery k (%)
NB number of batteries DG generation efficiency for DG unit i (%)
i
NDG number of DG units
W wind turbine efficiency (%)
NDR number of DR loads
a air density (kg/m3 )
NL number of loads
NT number of operation time steps
C. Variables
PkB,Total total capacity that battery k can deliver over its life- B
Ck,t operation cost of battery k during time period t ($)
time (kW h) DG
PkB,max maximum allowable power level for battery k (kW) Ci,.t operation cost of DG unit i during time period t ($)
DR
Cj,t cost of DR load j at time t ($); DR is modeled as
PkB,min minimum allowable power level for battery k (kW)
negative demand
PiDG,max maximum allowable power generation level for DG
CLFk,T corrected lifetime for battery k due to ambient tem-
unit i (kW)
perature Tta (months)
PiDG,min minimum allowable power generation level for DG
Pm,t injected active power to the network at node m dur-
unit i (kW); it is assumed not to be cost effective for
ing time period t (kW)
the DG to generate power below this level B,c
Pk,t amount of charging power provided to battery k
PjDR,max maximum allowable power reduction level for DR
during time period t (kW)
load j (kW) B,d
Pk,t amount of power discharge provided by battery k
PjDR,min minimum allowable power reduction level for DR
during time period t (kW)
load j (kW); it is assumed not to be cost effective for DG
Pi,t active power provided by DG unit i during time
the DR load to provide demand reduction below this
period t (kW); DG may be a wind resource (denoted
level
L as W) or a solar resource (denoted as PV)
Pm,t active demand for the load connected to bus m dur- DR
Pj,t active power of DR load j during time period t (kW)
ing time period t (kW)
Line
PbLine,max maximum allowable power flow through line b Pb,t active power flowing thorough line b at time t (kW)
(kW); also known as line rating or line ampacity Ptsub power injected to the grid from the distribution sub-
PV
Pi,t power output of PV panel i during time period t (kW) station at time period t (kW)
B
PiPV,STC power provided by PV panel i under standard test SOCk,t state of charge of battery k during time period
condition (STC) (kW) t (%)
232 M. Choobineh et al. / Electric Power Systems Research 130 (2016) 230–240
uB,c
k,t
binary variable that indicates battery k is charging
during time period t (1: being charged, 0: not being
charged)
uB,d
k,t
binary variable that indicates battery k is dis-
charging during time period t (1: being discharged,
0: not being discharged)
vDR
j,t
binary variable that specifies DR load j is being cur-
tailed during time period t (1: being curtailed, 0: no
curtailment)
ım,t mth column of the voltage angle matrix at time
period t
First and foremost, system demand is highly dependent on temper-
ature variations [11,12]. For instance, Chen [12] showed that the
power consumption of Taipei city can increase by up to 22% when
the temperature rises by 5 ◦ C. This increase is mainly attributed
to the increased loading of air conditioning systems. On the other
hand, generation capacity of most energy resources will vary with
temperature. For synchronous generators, maximum power limit
will be adversely affected by significant temperature changes. The
efficiency of a generator during extreme heat is lower than its effi-
ciency during normal operating conditions, and its natural cooling
capability significantly decreases. In addition, high temperatures
are typically associated with lower air density, which can cause
combustion problems in smaller scale diesel engines. This inverse
relation between temperature and generator power output which
occurs due to lack of enough oxygen levels can be modeled by a
derating factor [13]. Energy storage systems, on the other hand,
respond to extreme heat differently. The capacity of most batter-
ies (e.g. sealed lead-acid batteries) is not only dependent on the
discharge rate, but also on the ambient temperature. As temper-
ature rises, the electrochemical activity in the battery increases;
and electrical capacity that can be extracted from it increases as a
result [14,15]. The dependence of battery capacity on the temper-
ature can also be modeled by a temperature coefficient [16]. High
temperatures also have considerable impacts on the battery life-
time, and many researchers have tried to explicitly or empirically
model the effect of temperature on degradation of batteries [17,18].
In [19], this phenomenon has been modeled for lead acid batter-
ies. In [20], it has been shown how depth-of-discharge and high
temperatures can both affect the degradation of Li-ion batteries.
Photovoltaic (PV) panels are also sensitive to ambient tempera-
ture due to the properties of the semiconductor material in the
PV cells. Increasing the temperature reduces the band gap of the
PV cell and consequently its open circuit voltage. Short circuit cur-
rent, on the other hand, is mostly a function of solar irradiance
and varies with temperature only slightly. Therefore, the dc power
generated by a PV panel can be affected by extreme temperatures
[21]. Last but not the least, excess heat can dramatically increase
the surface temperature of overhead conductors, which in turn
results in conductor sag. Temporary sag of the conductor reduces
its distance to ground and increases the chances of flashover. At
the same time, most conductors have a threshold beyond which
excess elongation leads to permanent annealing of the steel core
[22]. Fig. 1 illustrates some example diagrams for the DG maxi-
mum power, battery capacity, PV generation output, and overhead
line ampacity as a function of variations in ambient temperature.
It is noteworthy that except for the PV generator (whose power
versus temperature variations is a well-known mathematical equa-
tion), the effect of temperature on the rest of the components are
usually provided in the form of discrete observations. Whenever
applicable, we have extrapolated this data in order to be able to
model the temperature–capacity relation over a range of ambient
temperatures.
The reduction in the available capacity of most energy resources
together with the increase in demand may prevent some power
utilities from meeting the demand during a heat wave period.
Under such conditions, the utility may have to revert to the often
expensive spot market of electricity to acquire the needed energy.
This could manifest itself as large price spikes in the electricity mar-
ket. For instance, during the heat wave in August 2003, prices in the
day-ahead Dutch APX market reached over D1000 for 1 MW of elec-
tricity delivered during peak hours on August 11 and 13, compared
to the average peak price of approximately D49 over the 20-day
period before August 11 [23]. In addition, even if energy supplies
are not affected by the event, the power transfer capability of lines
connecting generation resources to loads may be reduced. These
further emphasize the need for an efficient and optimal energy dis-
patch in the power grid during the course of extreme temperatures.
The objective of this paper is to dispatch distributed energy
resources (DER), i.e. distributed generation (DG) units, battery
energy storage systems (BESS) and demand responsive (DR) loads
in the distribution network during the course of a heat wave event.
DR loads provide a platform for voluntary demand reduction and
can help the utility maintain the balance between generation and
demand while avoiding the involuntary load shedding. The out-
come of the proposed methodology would be a series of dispatch
set-points that consists of the level at which the generation and
energy storage systems need to inject power and the level to which
DR loads need to be curtailed. All this will be determined given
the demand forecast of the system as well as the available power
generation and transfer capacity across the grid. Since high temper-
ature rises affect the DER units and overhead lines in the system, a
temperature-dependent operational framework is proposed here.
2. Proposed model
In this section, the mathematical models for DG, BESS and DR
loads are described. Whenever applicable, the impact of tempera-
ture on the component capacity has also been modeled. In the rest
of this section, each time step of analysis is assumed to be one hour,
which means the power levels (in kW) would be equivalent in value
to the energy levels (in kW h). In the absence of this assumption, all
power values would have to be multiplied by the duration of the
time step in order to derive the energy level.
2.1. Dispatchable DG units
The operational cost of a DG unit can be expressed as a function
of the fuel price as well as its thermal value [24,25]. Eq. (1) repre-
sents the general relation that can be adapted for a diesel generator,
a microturbine, or any other fuel-based DG unit.
DG
DG
cifuel Pi,t
Ci,t = × (1)
Hifuel DG
i
Output power of the generator in the above equation will be
limited by its maximum allowable power PiDG,max . As mentioned
before, PiDG,max varies with temperature; therefore, the corrected
maximum power based on the ambient temperature should be
used instead. To do this, we have used the data available in [13] and
have fit a linear curve to it, as indicated in (2). It should be noted
that this particular DG unit has been chosen for demonstration
purposes, and different units may demonstrate slightly different
behavior with temperature variations.
DG,max
Pi,T = −0.47 · Tta + 111.6 (2)
 M. Choobineh et al. / Electric Power Systems Research 130 (2016) 230–240 233

Fig. 1. Typical behavior of a synchronous generator, a battery system, a PV panel and an overhead conductor as the temperature rises. Components from various manufacturers
may show slight differences in numeric values but the patterns of behavior generally remain the same. The discrete data for the overhead line, the generator and the battery
have been taken from [10,13,19], respectively. Curve fitting has been carried out using the curve fitting toolbox in MATLAB.

2.2. Non-dispatchable DG units has been derived here by curve fitting using the data provided in
[19] for a vented lead-acid battery with 1.215 specific gravity cells.
Two types of non-dispatchable DG units have been considered Again, it should be noted that this particular battery is chosen for
in this paper: wind power and solar power. The output power of proof-of-concept only, and the choice of battery does not alter the
wind energy conversion system can be determined as a function nature of the proposed methodology.
of the wind speed and the swept area of the turbine rotor, among
B,max
 2
other things [26]: Pk,T = −0.016 · Tta + 1.97 · Tta + 60.75 (6)
W
Pi,t = 0.5W C P a Ai v3i,t (3) When it comes to modeling the operational cost of a BESS, sev-
eral approaches have been proposed in the literature. Because the
The output power of a PV generator i at time t can be modeled
lifetime of a battery depends on the total ampere hour (A h) it pro-
based on the generation level under standard test condition (STC)
vides, its cost function is modeled here based on its working A h
[27]. The effect of ambient temperature is included in the relation:
[29,30]. This way, we assume a cost associated with the battery
˚i,t    if it is being discharged. This cost is normalized based on the total
PV
Pi,t = PiPV,STC 1 + kPV · Tta − T R (4)
˚STC energy that the battery can deliver over its entire lifetime. However,
this total energy must be corrected based on ambient temperature
2.3. Battery energy storage system (BESS) [19]:
B,d
The capacity of battery k at time t can be determined based Pk,t
on its capacity at the previous time step, and the consideration of
B
Ck,t = ckB
B,Total
  (7)
Pk CLFk,T /LFk
whether or not it has been charged or discharged during the current
time period. Therefore, the battery state of charge (SOC) is modeled where CLF is defined as follows:
according to the following equation [25,29]:
    LFk
  B,c
Pk,t B,c
k B,d
Pk,t CLFk,T =    (8)
B
SOCk,t = 1 − ık B
· SOCk,t−1 + uB,c
k,t
· − uB,d
k,t
· (5) Tk,t
tTD /KT
PkB,max B,d
k
PkB,max

The battery capacity under various ambient temperatures can be KT is the temperature correction factor for battery lifetime which
expressed based on its nominal capacity (i.e., 100% capacity avail- has been described in IEEE 450 [19]. The extrapolated function indi-
able at 25 ◦ C). The ratio of the available capacity of the battery in (6) cates that this factor follows an exponential behavior that can be
234 M. Choobineh et al. / Electric Power Systems Research 130 (2016) 230–240

focus is on the real-time cost of operation, and we do not consider

the longer term costs such as maintenance and capacity expan-
sion. The problem is formulated while considering the impact of
ambient temperature on the available capacity of various compo-
nents. It is assumed that operational costs incur due to the usage of
any DG, DR or BESS, with the exception of non-dispatchable units
that operate in the maximum power point tracking (MPPT) mode.
Renewable units are therefore not incorporated into the cost func-
tion; however, they are considered in the generation-load balance
equation.
The objective function is therefore defined as follows:
⎧ ⎧ ⎫⎫
⎨ NT ⎨ NDG NB NDR ⎬⎬
min ctsub Ptsub + DG
Ci,t + uB,d B
· Ck,t + vDR C DR
⎩ ⎩ k,t j,t j,t
⎭⎭
t=1 i=1 k=1 j=1

(12)

The first term in (12) indicates the cost of power purchased from
the electric utility. The second and third terms denote the opera-
Fig. 2. Temperature correction factor for battery lifetime. tional costs of dispatching the DG units and the BESS, while the
fourth term indicates the costs associated with utilizing DR loads.
The objective function in (12) would be solved subject to the fol-
expressed as in (9) (see Fig. 2). This is expected, since KT is based
lowing constraints:
on the Arrhenius equation.
  ∀m, ∀t : Pm,t = Bm ım,t (13)
KT = 762.26 · exp −0.082 · Tta (9)
∀i, ∀t : Pi,t = Pi,t
DG L
− Pi,t (14)
2.4. Conductor ampacity
∀k, ∀t : Pk,t = uB,d P B,d − uB,c
k,t k,t
P B,c − Pk,t
k,t k,t
L
(15)
For safe operation, the surface temperature of an overhead
conductor must always be below a specific manufacturer- ∀j, ∀t : Pj,t = vDR P DR − Pj,t
j,t j,t
L
(16)
recommended threshold. Exceeding this threshold could lead to
thermal stresses on the conductor, and potential loss of tensile ∀q, ∀t : Pq,t = −Pq,t
L
(17)
strength and annealing. Since the current flow through a con-
∀t : Ptsub ≤ Ptsub,max (18)
ductor increases its surface temperature (due to Joule’s effect),
the available capacity of the conductor is determined by the con- ∀i, ∀t : PiDG,min ≤ DG
Pi,t ≤ PiDG,max DG,max
· Pi,T (19)
ductor’s maximum allowable surface temperature [10]. This calls
for calculating the dynamic thermal rating (DTR) of a conductor ∀j, ∀t : PjDR,min vDR
j,t
DR
≤ Pj,t ≤ PjDR,max vDR
j,t
(20)
under temperature or loading variations. IEEE 738-1993 proposed
a first order differential equation for calculating the DTR of a bare- ∀k, ∀t : SOCB,min
k
≤ SOCBk,t ≤ SOCB,max
k
(21)
conductor overhead line [32]. This equation balances the heat
produced by the Joule’s effect in conductor’s resistance as well ∀k, ∀t : PkB,min uB,d
k,t
B,d
≤ Pk,t ≤ PkB,max uB,d
k,t
B,max
· Pk,T (22)
as the heat gained from solar irradiance with the heat dissipated
through convection and radiation. Based on this model, the follow- ∀k, ∀t : uB,c
k,t
+ uB,d
k,t
≤1 (23)
ing extrapolated function can be approximated for an aluminum
conductor steel reinforced (ACSR) overhead line with the specifi- ∀b, ∀t : |Pb,t
Line
| ≤ PbLine,max · Pi,T
Line,max
(24)
cations provided in [10]:
Constraint (13) indicates the requirement that generation and
Line,max
Pb,T = −0.768 · Tta + 119.45 (10) demand have to be balanced at all times. This is critical for ensuring
frequency stability in the system. Constraints (14)–(17) represent
2.5. Demand responsive (DR) loads active power balance equations at each node, i.e. the total power
injected into each node equals the total power consumed at that
When the balance between available generation and total node. Substation node is not included here because it is assumed
demand is jeopardized, DR loads may be requested by the utility that no load is directly connected to the distribution substation.
to curtail their consumption level by a certain amount. In return, Demand response has been modeled as virtual generation on the
the customers who comply with this request will receive financial right hand side of (16). Eq. (18) shows the upper limit for the power
incentives [33]. This resource can therefore be modeled as a virtual imported through the substation transformer. Eqs. (19) and (20)
generation resource with a quadratic cost function [34] as shown represent the lower and upper limits for capacity available through
in the following equation: DG and DR, respectively. It can be seen that the maximum power

DR
 2 available from DG may be corrected due to temperature variations.
Cj,t = ˇjDR Pj,t
DR
+ jDR Pj,t
DR
(11) Eqs. (21)–(23) demonstrate the operational constraints of the bat-
tery. Eq. (23) indicates that during each time step, a BESS can be
2.6. Problem formulation either in charging mode or discharging mode, but not both. Finally,
Eq. (24) indicates that the power flowing through each line should
The objective here is to minimize the total operational cost of the be limited by its thermal rating.
network, while ensuring the balance between demand and avail- It should be pointed out that this model considers a DC power
able generation is maintained. It should be emphasized that our flow model where active power is taken into account and reactive
 M. Choobineh et al. / Electric Power Systems Research 130 (2016) 230–240
Table 1
Difference between new and existing homes.
New Existing
Wall insulation (m2 K/W) R-2.29 + R0.88 Uninsulated
Attic insulation (m2 K/W) R-5.28 R-2.29
Windows U/SHGC (m2 K/W) 2.1/53 4.77/0.63
Infiltration (ACH@50 Pa) 4 20
Foundation Uninsulated slab-on-grade
A/C (SEER) 13 8
power equations are ignored. This is often acceptable due to the
long-term nature of the analysis (i.e. daily to weekly). However,
for a more detailed analysis, reactive power equations can also be
incorporated into the model (AC power flow). First and foremost,
this introduces node balance equations and lower/upper limit con-
straints defined for reactive power. In addition, Eq. (13) must be
expanded to also include the relationship between reactive power
and node voltage magnitudes. Decoupling between active and reac-
tive power can be assumed in order to maintain the linearity of the
equations. However, the main change to the model as a result of AC
power flow approach would be the introduction of line flow limits
in terms of the apparent power. This would result in second order
constraints.
3. Modeling demand
One of our goals is to model demand variations with temper-
ature. To do this, we have analyzed energy consumption models
for different types of buildings. Naturally, increase in temperature
would result in a rise in energy consumed by the air conditioning
(A/C) systems, which would in turn depend on: (1) the size of the
house, (2) how well insulated and air-tight it is, and (3) the house’s
A/C equipment performance, as the coefficient of performance
(COP) decreases with an increase in outdoor air temperature. For a
realistic analysis, we have simulated the consumption of a variety
of houses under different ambient temperatures. We have consid-
ered both current practices for new homes as well as designs for
pre-1980 homes. All houses are assumed to have central A/C and
heating systems. Table 1 summarizes the difference between new
and existing houses. All new ones are assumed to be either similar
or more energy-efficient than the B10 Benchmark from the Build-
ing America House Simulation Protocols developed by NREL [35].
The B10 Benchmark is consistent with: (1) the 2009 International
Energy Conservation Code (ICC 2009), (2) federal appliance stan-
dards of Jan. 1, 2010, and (3) lighting and miscellaneous electric
loads (MELs) characteristics common in homes built in 2010. Exist-
ing houses followed the House Simulation Protocols for Pre-retrofit
cases of existing (old) buildings. It assumes houses were built
between 1950 and 1979, with A/C units dating between 1981 and
1991. All houses are assigned a cooling setpoint of 76 ◦ F as stated
in the benchmark. Wall insulation of R-2.29 + R0.88 m2 K/W (or R-
13 + R5 in h ft2 ◦ F/Btu) stands for R-2.29 fiberglass batts inside the
wall cavity in a stud frame house with additional R-5 of extruded
Table 2
Difference between detached, attached and apartment homes.
Detached homes
New
Finished area (m2 ) 232
A/C Size (Tons) 5 6
Plug loads (kW h/y) 3270
Ratio of AC energy and total house energy use at peak time 0.73
Ratio of aggregated AC energy and total house energy use 0.57
over the analyzed week
235
Fig. 3. Hourly profiles for occupants in the house (Occupants), lights (Lights),
cooking (Range) and plug-in loads (Mels) over a 24-h period.
polystyrene as continuous exterior insulation. This is better than
B10 for a house located in Sacramento. Table 2 shows the dif-
ference between detached, attached (townhouse) and apartment
homes. It also shows that most of the electric energy use during
heat waves is due to A/C equipment, with existing houses being
the most susceptible to heat waves.
The building models have been developed in Building Energy
Optimization (BEopt) 2.3, which is a program used for designing
residential buildings and identifying cost-optimal efficiency pack-
ages of whole-house energy savings. BEopt 2.3 uses DOE building
energy simulation program EnergyPlus version 8.1 that considers
thermal interactions between the outdoor air, sky and ground, as
well as gains from inside the house. Fig. 3 shows the hourly profiles
for different internal gains per Building America House Simulation
Protocols. Internal gains peak around 4 pm to 10 pm when people
arrive home and start cooking, in addition to other activities that
increase the internal gains in homes.
For demonstration purposes, the building energy simulations
used actual weather data from Weather Analytics for the week of
July 19–27, 2006 associated with the heat wave in Sacramento, CA.
Fig. 4 shows the cooling electric consumption for each home as well
as the outdoor air temperature. All houses follow a similar trend to
outdoor air temperature and have higher energy use as this tem-
perature increases. The house with higher cooling energy house is
the existing detached home, despite the fact that the new detached
home is larger and has higher internal gains. This is because the
existing house has higher exterior gains due to a less insulated
envelope and in addition, has a less efficient A/C system. The same
trend is observed with the townhouse and the apartment homes.
This is not surprising, since residential buildings are more suscepti-
ble to outdoor environment than commercial buildings, thus more
susceptible to extreme weather events such as heat waves. It is
important to note that the apartment home only simulates 1 apart-
ment, thus a simple calculation for a multi-family building would
require multiplying this number by the number of apartments in
the building.
Attached home (both sides) Apartment (center)
Existing New Existing New Existing
119 149 124 83 62
3 4 1 1.5
2730 2980 2890 1459 1406
0.88 0.64 0.87 0.51 0.77
0.77 0.48 0.74 0.36 0.66
236 M. Choobineh et al. / Electric Power Systems Research 130 (2016) 230–240

Fig. 4. Cooling electric consumption for new and existing detached home, townhouse and apartments alongside the outdoor air temperature for week of July 19–27, 2006.

4. Simulation results that this approach has been used for proof-of-concept and demon-
stration purposes only and does not affect the generality of the
In order to evaluate the proposed temperature-dependent proposed solution.
dispatch model, two case studies are conducted which solve In our simulations, the actual solar irradiance and wind speed
the problem of active power dispatch (minimize (12) subject data have been used for the time period and geographical loca-
to constraints (13)–(24)) with and without the effect of ambi- tion under study [37,38]. The following parameters have been used
ent temperature. To determine the dispatch without the effect for each PV panel: PPV,STC = 200 W, ˚STC = 1000 W/m2 , kPV = 0.004
of temperature, ambient temperature has been set at 25 ◦ C. (◦ C−1 ), and TR = 25 ◦ C. It is assumed that all PV panels operate in
This value provides the nominal ratings for all components. The the MPPT mode. Wind turbine parameters have also been cho-
problem has been solved using LINDOGLOBAL solver in GAMS sen according to [39] which are as follows: W = 40%, Cp = 0.593,
software. a = 1.225 kg/m3 , A = 353 m2 , and cut-in and cut-out wind speed:
2.7 m/s and 25 m/s. Figs. 6 and 7 illustrate the total PV and wind
power projections with and without temperature consideration
4.1. Case study over a 3-day time period between July 22–24, 2006. Total PV out-
put differs by about 10% when actual outdoor temperatures are
The case study presented here is related to the period of exces- evaluated, while wind power shows no dependency on outdoor air
sive temperatures in July 19–27, 2006 in the city of Sacramento, CA. temperature.
Whenever possible, actual data corresponding to this time period
have been used. In the absence of actual data on the distribution
network of the city of Sacramento, the IEEE 33-bus test distri-
bution system [36] has been used for proof-of-concept purposes
(see Fig. 5). The locations of the nodes are heuristically determined
based on their respective average demand and the population den-
sity map of the city. Then, housing types (buildings, townhouses,
and single family residences) are heuristically assigned based on
proximity to the downtown area or the suburbs, as well as the pop-
ulation density. Finally, data on the median household income is
consulted to heuristically assign ratios of old houses versus new
ones, i.e. it has been assumed that the ratio of new to old houses
increases as the median household income increases. For example,
in [36], bus 24 has the joint second highest demand. Hence, it is
assigned to the location shown in Fig. 5, which is one of the areas
with the highest population density in the city of Sacramento. Since
the area is located between the downtown area and the suburbs
(closer to the former), it is heuristically assumed that it consists of
mostly townhouses (50%), compared to apartments and single fam-
ily houses (20% and 30%, respectively). Also, the bus is associated
with an area in the city with a relatively high median household
income of $80,000; therefore, the ratio of new constructions to the Fig. 5. Schematic diagram of the system under study. Background image is extracted
old ones is considered to be around 80%. It should be emphasized from Google maps.
 M. Choobineh et al. / Electric Power Systems Research 130 (2016) 230–240
Fig. 6. Total PV power generation with and without temperature effect consider-
ation.
The network under study has been expanded to include DER
units, i.e. DGs, BESSs and DRs. The related cost coefficients and
other parameters for these units are provided in the Appendix (see
Tables A1–A3). It should be noted that all numeric values for the
DER unit costs are used here for demonstration purposes and do
not alter the nature of the problem. Adopting higher or lower values
would change the output of the problem, but not the methodology.
4.2. Results
The three hottest days (i.e. 22–24 July 2006) during the heat
wave event have been considered for the study (a total of 72 h).
All the simulation results are shown for two scenarios: with and
without temperature effect consideration. Regardless of the sce-
nario chosen, loads are assumed to vary based on temperature (as
derived in Section 3).
Fig. 8 shows the amount of power purchased from the grid for
the two scenarios. It can be seen that when the temperature effect
is not considered, the power imported from the grid is kept at its
Fig. 7. Total wind power generation.
237
Fig. 8. Amount of power purchased from the grid with and without temperature
effect consideration.
maximum level for most of the time; however, considering tem-
perature effect changes the dispatch decision especially during the
hours of 11–18, 35–42 and 59–65 (corresponding to about 11 am to
7 pm) when the temperature exceeds 40 ◦ C. This is mainly caused by
the effect of temperature on the main overhead line which is sup-
plied from the substation. This fact further emphasizes that even
when generation capacity is available, it may be of limited use if
sufficient capacity to transmit it to the loads is not available.
Also, it can be seen that during hour 24 (midnight) the power
purchased from the grid has reduced for both cases. This phe-
nomenon slightly repeats in hours 48 and 72 as well. This is due to
the reduction in demand in those hours, which means less power
is purchased from the grid. Of course, the demand remains low
from midnight until approximately 6 am the next day. However, the
variable electricity rate we have assumed for the simulations (see
Table A4) is higher during midnight compared to the early hours
of the day, which is why this pattern is not observed during those
hours. Another interesting observation for hour 24 is that grid par-
ticipation in the dispatch is higher when the temperature is ignored
compared to when it is taken into account. This is because high
temperatures in this hour reduce the maximum capacity of DGs by
almost 20.58 kW, which means the participation of the grid in the
dispatch problem must increase by this same amount (20.58 kW).
Fig. 9 illustrates the total DG power injected into the grid with
and without temperature effect consideration. As expected, total
DG power reduces by up to 7% whenever temperature is higher than
40 ◦ C. This can be attributed in part to the reduction in DG maxi-
Fig. 9. Total dispatchable DG power generation with and without temperature effect
consideration.
238 M. Choobineh et al. / Electric Power Systems Research 130 (2016) 230–240
Fig. 10. Total BESS charging and discharging power with temperature effect con-
sideration.
mum capacity. However, there are some time periods in the exact
solution (with temperature effect) that the DG power injection
increases to levels above those associated with approximate solu-
tion (without temperature effect). Essentially, this happens because
during these hours, BESS units start charging in order to be able to
assist with grid operation during hot hours.
Figs. 10 and 11 depict BESS dispatch results for both charg-
ing and discharging modes, as well as their SOCs. As can be seen
the SOCs for all batteries remain within bounds during the analy-
sis timeframe. It should also be noted that when the temperature
effect is not considered, BESSs are not dispatched in the opera-
tion program. This can be because the problem assumes sufficient
generation and transmission capacity is available, and also because
batteries are in general considered to be more costly than the tra-
ditional grid or the DG units.
Fig. 12 shows the total power curtailed through DR loads. Similar
to the previous results, DR loads would be dispatched during hot
hours, with the highest participation being around hour 38 when
the temperature is at its maximum (i.e. 50.1 ◦ C). Similar to the case
of BESS units, no DR has been dispatched when temperature effect
is not considered.
Fig. 11. BESS SOC with temperature effect consideration.
Fig. 12. Total DR load curtailment with temperature effect consideration.
Finally, Fig. 13 illustrates the flow of power through four lines
which are fed from the four DG units. For lines 18–17 and 33–32
(see Fig. 5 for reference on the locations of these lines), the tem-
perature effect leads to reduction in the power flow through them.
This is because the DGs connected to these two lines always inject
power (in both scenarios) and there is no large demand connected
to these buses to consume the power locally. However, the situa-
tion is different for lines 25–24 and 22–21. For the latter, there are
some time periods when the line flow considering the temperature
effect is even higher than the case with no temperature consider-
ation. As can be seen, these time periods always coincide with the
times that power flow for line 33–32 decreases. Moreover, solar
power also reduces during these time periods (because of temper-
ature rise). Therefore, DG unit of bus 22 tries to compensate for
this power reduction. The same argument can be stated about line
25–24 which tries to compensate for the power drop in DG unit
at bus 18, as well as some portion of the solar power reduction.
The negative power flow of these lines is associated with the times
when DG unit connected to bus 22 does not generate power and
the consequent need to supply the heavy loading of bus 25 from
other resources.
The cost comparison for the two scenarios has been shown in
Table 3. It can be seen that considering the effect of temperature
increases the operational cost by 2.6%. This increase amount is
mostly due to battery and DR participation. It is noteworthy that
this difference in cost is paid to guarantee the load-generation bal-
ance is maintained at all times. Otherwise, if temperature effects
are ignored, lack of balance between load and generation may take
place, leading to emergency actions such as load shedding.
4.3. Discussion
Analysis of the simulation results indicates that when consider-
ing the effects of temperature on the energy dispatch, the dispatch
cost can be expected to increase. This is natural since the effect of
temperature is modeled in terms of more limitations on the capac-
ity of power grid components, which would naturally move the
Table 3
Cost comparison between the two scenarios.
Cost ($) Without temperature With temperature
effect consideration effect consideration
Grid power cost 16,842 16,635
DG cost 11,220 11,525
DR cost 0 537
BESS cost 0 109
Total cost 28,062 28,807
 M. Choobineh et al. / Electric Power Systems Research 130 (2016) 230–240 239

Fig. 13. Power flows through four lines connected to DG units, with and without temperature effect consideration. Dashed and solid lines indicate with and without
temperature considerations, respectively.

solution away from the more relaxed (no temperature) problem.
However, during the course of a natural disaster such as a heat
wave, operational cost is usually of secondary importance. Instead,
care must be taken to ensure the balance between load and gener-
ation is maintained.
If an operator chooses to ignore the temperature effects, the grid
may become exposed to severe risks. First and foremost, ignoring
the “real-time” capacity of generators and running them closer to
or above this limit could potentially lead to severe loss of life, or
in more serious cases, failure of the component. More importantly,
ignoring the available capacity of the overhead lines can cause one
or more to become overloaded. This is expected since the system
may experience heavy loading conditions due to excessive usage of
A/C units. Overloading a line, especially over prolonged periods of
time, leads to increased conductor sag which could in turn result
in flashover and loss of service.
Finally, our analysis showed the invaluable role that DR loads
can play under such conditions. Without the need for major capital
investment and with capability to deploy fast, these resources can
help the utilities compensate for their inadequacy of supply.
5. Concluding remarks
An operation dispatch strategy was proposed in this paper for
energy management of a power distribution system exposed to
a heat wave event. Whenever applicable, the components were
modeled by considering the effect of excess temperatures on their
operational constraints. It was shown that ignoring the effect of
temperatures can lead to significant overestimation of available
resources, both generation and transmission. This can jeopardize
the security and reliability of the system.
Acknowledgements
Financial support by Colorado School of Mines (CSM) is greatly
appreciated for the completion of this work. CSM was not involved
in the technical aspects of the work.
Appendix A.
Tables A1–A4

Table A1
Locations and characteristics of DG units.

DG type Bus number PiDG,min (kW) PiDG,max (kW) Hi (kW h/fuel unit) Fuel cost ($/fuel unit) DG
i

Dispatchable 18 0 500 13.88 0.7 0.6

22 0 600 10.53 0.8 0.55
25 0 600 13.88 0.7 0.6
33 0 500 13.88 0.7 0.6

Non-dispatchable Solar 3, 10, 30 0 N/A N/A N/A N/A

Wind 16, 21 0 N/A N/A N/A N/A
240 M. Choobineh et al. / Electric Power Systems Research 130 (2016) 230–240

Table A2
Locations and characteristics of batteries.

Bus number PkB,min (kW) PkB,max (kW) Min. SOC Max. SOC B,c
k
, B,d
k
␦k ckB ($)

6, 14 0 300 30% 100% 0.8 0.0025 40

Table A3 [16] S. Barsali, M. Ceraolo, Dynamical models of lead-acid batteries: implementation

Locations and characteristics of DR loads.
Bus number PjDR,min PjDR,max (kW) ˇj j
26–29, 31, 32 0 40% of the total 0.2 0.001
demand at the
bus it is
connected to
Table A4
Price of electricity purchased from the grid (repeats for each day).
Hour 1–6 7–11 12–16 17–21 22–24
Price ($/kW h) 0.06 0.08 0.10 0.12 0.10
References
[1] B.D. Ostro, L.A. Roth, R.S. Green, R. Basu, Estimating the mortality effect of the
July 2006 California heat wave, Environ. Res. 109 (2009) 614–619.
[2] O. Jay, M.N. Cramer, N.M. Ravanelli, S.G. Hodder, Should electric fans be used
during a heat wave? Appl. Ergon. 46 (2015) 137–143.
[3] C. Kousky, Informing climate adaptation: a review of the economic cost of
natural disasters, Appl. Ergon. 46 (2014) 576–592.
[4] W. Huang, H. Kan, S. Kovats, The impact of the 2003 heat wave on mortality in
Shanghai, China, Sci. Total Environ. 408 (2010) 2418–2420.
[5] J. Zuo, S. Pullen, J. Palmer, H. Bennetts, N. Chileshe, T. Ma, Impacts of heat waves
and corresponding measures: a review, J. Cleaner Prod. 92 (2015) 1–12.
[6] A.G. Barnett, S. Hajat, A. Gasparrini, J. Rocklov, Cold and heat waves in the United
States, J. Environ. Res. 112 (2012) 218–224.
[7] A. Sakka, M. Santamouris, I. Livada, F. Nicol, M. Wilson, On the thermal per-
formance of low income housing during heat waves, J. Energy Build. 49 (2012)
69–77.
[8] B. Thom, J. Cane, R. Cox, C. Farrell, P. Hayes, R. Kay, A. Kearns, D. Low Choy,
J. McAneney, J. McDonald, M. Nolan, B. Norman, J. Nott, T.T. Smith, National
Climate Change Adaptation Research Plan for Settlements and Infrastructure,
National Climate Change Adaptation Research Facility, Gold Coast, QLD, 2010.
[9] A. Pechan, K. Eisenack, The impact of heat waves on electricity spot markets, J.
Energ. Econ. 43 (2014) 63–71.
[10] M. Choobineh, B. Ansari, S. Mohagheghi, Vulnerability assessment of the power
grid against progressing wildfires, Fire Saf. J. 73 (2015) 20–28.
[11] D.H.O. McQueen, P.R. Hyland, S.J. Watson, Monte Carlo simulation of resi-
dential electricity demand for forecasting maximum demand on distribution
networks, IEEE Trans. Power Syst. 19 (Aug. (3)) (2004) 1685–1689.
[12] C.S. Chen, M.S. Kang, J.C. Hwang, C.W. Huang, Temperature effect to distribution
system load profiles and feeder losses, IEEE Trans. Power Syst. 16 (4) (Nov. 2001)
916–921.
[13] 1370GQMA Data Sheet, Cummins, [Online]. Available at: http://www.
cumminspower.com/www/Commercial/SparkIgnited/d-3266.pdf.
[14] K. Kutluay, Y. Cadirci, Y.S. Ozkazanc, I. Cadirci, A new online state-of-charge
estimation and monitoring system for sealed lead-acid batteries in telecom-
munication power supplies, IEEE Trans. Ind. Electron. 52 (5) (2005 Oct)
1315–1327.
[15] M. Chen, G.A. Rincon-Mora, Accurate electrical battery model capable of pre-
dicting runtime and I–V performance, IEEE Trans. Energy Convers. 21 (2) (2006
Jun) 504–511.
issues, IEEE Trans. Energy Convers 17 (1) (2002 Mar) 16–23.
[17] C.J. Rydh, B.A. Sanden, Energy analysis of batteries in photovoltaic systems. Part
I: Performance and energy requirements, J. Energy Convers. Manage. 46 (2005)
1957–1979.
[18] M. Uno, K. Tanaka, Accelerated charge–discharge cycling test and cycle life
prediction model for supercapacitors in alternative battery applications, IEEE
Trans. Ind. Electron. 59 (12) (2012) 4704–4712.
[19] IEEE, IEEE recommended practice for maintenance, testing and replacement of
vented lead-acid batteries for stationary applications, in: IEEE P450-R10, Apr.
2001, 2001.
[20] I.J. Fernandez, C.F. Calvillo, A. Sanchez-Miralles, J. Boal, Capacity fade and ageing
models for electric batteries and optimal charging strategy for electric vehicles,
J. Energy 60 (2013) 35–43.
[21] E. Skoplaki, J.A. Palyvos, On the temperature dependence of photovoltaic mod-
ule electrical performance: a review of efficiency/power correlations, J. Sol.
Energy Eng. 83 (2009) 614–624.
[22] T. Seppa, Accurate ampacity determination: temperature-sag model for oper-
ational real-time ratings, IEEE Trans. Power Delivery 10 (3) (1995) 1460–1470.
[23] R. Huisman, The influence of temperature on spike probability in day-ahead
power prices, J. Energy Econ. 30 (2008) 2697–2704.
[24] J. Chen, X. Yang, L. Zhu, M. Zhang, Study on microgrid multi-objective economic
operation optimization, Int. J. Adv. Comput. Technol. 5 (2013) 1218–1229.
[25] H. Wu, X. Liu, M. Ding, Dynamic economic dispatch of a microgrid: mathe-
matical models and solution algorithm, Electr. Power Energy Syst. 63 (2014)
336–346.
[26] G.M. Masters, Renewable and Efficient Electric Power Systems, John Wiley &
Sons, Inc, Hoboken, NJ, 2004.
[27] J.G. Vlachogiannis, Probabilistic constrained load flow considering integration
of wind power generation and electric vehicles, IEEE Trans. Power Syst. 24 (4)
(2009 Nov) 1801–1817.
[29] B. Zhao, X. Zhang, J. Chen, C. Wang, L. Guo, Operation optimization of stan-
dalone microgrids considering lifetime characteristics of battery energy storage
system, IEEE Trans. Sustainable Energy 4 (2013 Oct) 934–943.
[30] Y. Riffonneau, S. Bacha, F. Barruel, S. Ploix, Optimal power flow management
for grid connected PV systems with batteries, IEEE Trans. Sustainable Energy 3
(2011 Jul) 309–320.
[32] IEEE, IEEE standard for calculating the current-temperature of bare overhead
conductors, in: IEEE 738, Nov 2006, 2006.
[33] S. Mohagheghi, J. Stoupis, Z. Wang, Z. Li, H. Kazemzadeh, Demand response
architecture-integration into the distribution management system, in: Proc.
IEEE International Conference on Smart Grid Communications, Oct. 2010,
Gaithersburg, MD, USA, 2010.
[34] R. Palma-Behnke, J. Cerda, L.S. Vargas, A. Jofré, A Distribution company energy
acquisition market model with integration of distributed generation and load
curtailment options, IEEE Trans. Power Syst. 20 (Nov. (4)) (2005) 1718–1727.
[35] E. Wilson, C. Engebrecht Metzger, S. Horowitz, R. Hendron, Building America
house simulation protocols, in: NREL/TP-5500-60988, Mar. 2014, pp. 91.
[36] N. Rajaei, M.H. Ahmed, M.M.A. Salama, R.K. Varma, Fault current management
using inverter-based distributed generators in smart grids, IEEE Trans. Smart
Grid 5 (Sept. (5)) (2014) 2183–2193.
[37] National Renewable Energy Laboratory (NREL) Data Set, [Online]. Avail-
able at: http://rredc.nrel.gov/solar/old data/nsrdb/1991-2010/hourly/list by
state.html.
[38] Weather Underground History and Data Archive, [Online]. Available at:
http://www.wunderground.com/history/airport/KSAC/2006/7/22/Daily
History.html.
[39] Polaris America LLC 100 kW Wind Turbine Specification, [Online]. Available
at: http://www.environmental-expert.com/products/model-100kw-wind-
turbines-388404.