You are on page 1of 6

Optimization-based Home Energy Management in the Presence of

Solar Energy and Storage


Sattar Mohammadi*, Mahmoud Momtazpour**, and Esmaeil Sanaei***
Electrical Engineering Department, Sharif University of Technology
*sattar@ee.sharif.ir
**momtazpour@ee.sharif.ir
***sanaei@sharif.edu

Abstract: New technologies such as smart homes and [5], the authors investigated the performance analysis of a
appliances and renewable energy production have been
attracting much attention in recent years. In practice, the IEEE 802.15.4 based sensor network for smart grids
management of these technologies to get the minimum communications.
household payment has become a challenge due to the time- Some researchers studied the importance of the
varying electricity price. In this paper, we propose a real time
energy management system to manage the appliances and behaviour of appliances in home energy management. A
storages. The solar energy is charged in the storages and used recent study classifies appliances into four classes and
together with the grid electricity to supply the appliances. The provides mathematical models to describe their behaviour
proposed management system uses an ILP-based Home Energy
Management (ILPHEM) optimization engine. ILPHEM [6]. In [7], the effect of user behaviour on the energy
schedules appliances requests and storage usage based on consumption of home appliances is studied. In [8], the
updated real time input data for minimum household payment. authors propose a simulator that can predict the power
Experimental results show that the proposed ILPHEM model
can reduce the household payments by 20% compared to the consumption profiles of individual appliances.
best conventional approach. Some researches concentrate on demand management
to reduce the peak load and household payment. In [9],
Keywords: Home Energy Management, Integer Linear the authors evaluated the performance of an in-home
Programming, Responsive appliances, Home Area energy management (iHEM) application and compared it
Network. with an optimization-based residential energy
management (OREM) scheme whose objective is to
1. Introduction minimize the household payment. In [10], a residential
Nowadays, most of the electrical energy is generated load commitment (LC) is presented to achieve the
from fossil fuels. Generating electricity from fossil fuels minimum household payment. The LC problem was
comes with some challenges, such as increasing fuel costs formulated to determine on/off status of responsive
and undesired environmental impacts. Moreover, a appliances, charging/discharging cycles of Plug-in
remarkable part of the electricity is consumed in homes. Hybrid Electric Vehicles (PHEVs).
On the other hand, the need for electrical energy in home In [11], an Intelligent Energy Distribution
is growing continuously [1]. Therefore, reducing energy Management (iEDM) scheme is proposed to monitor
consumption in homes became an important motivation changing environmental variables and manage solar
toward smart grids. power flexibly.
In the past decade, many researches have been done in Most of the researchers have considered a single
different aspects of smart grids, such as demand aspect of the home energy management. In this paper we
managements and the integration of renewable energies consider more than one aspect of the problem. We
to home. For example, financial and technical aspects of propose an optimization-based home energy management
using solar energy in different parts of the world have system that manages appliances and storage to achieve
been evaluated in [2,3]. Some researchers have minimum household payment. This system uses an
investigated the implementation of the communication Integer Linear Programming (ILP)-based model called
infrastructures and analyzed their performance. In [4], ILPHEM as an optimization engine. The proposed
comparisons of communication technologies and network ILPHEM model has the following features:
architectures for machine to machine (M2M) 1) It considers the availability of solar energy and
communications in home area networks were studied. In local storage in the smart home.

978-1-4673-5634-3/13/$31.00 ©2013 IEEE


Fig. 1: The overall diagram of the proposed home energy management system. EMU sends Data Request packets to Smart meter, Solar Panels and
Storage to request electricity price, current solar energy production rate and amount of stored energy respectively. After retrieving the requested data
along with the updated list of appliance requests, the EMU computes the optimal schedule of grid and storage energy supplies to the requests.

2) It determines on/off status of responsive future trend of power consumption of the appliances is
appliances as well as the usage intervals of predicted since is essential for storage scheduling. In the
storage and grid electricity for each appliance. proposed model, 24 hours of a day is divided into 240
3) It reschedules appliances and storage in real time decision time intervals. In the other hand, the EMU
if the input data change to ensure minimum attempts to reschedule the current requests every six
household payment. minutes. Nevertheless, the decision time interval can be
4) The ILPHEM can handle all pricing policies. reduced for seamless decision making at the cost of more
The rest of the paper is organized as follows. In computational complexity.
Section 2, we introduce the proposed home energy
3.1 Input data and variables
management system. Section 3 formulates the proposed
ILPHEM model. In Section 4, the experimental results TABLE I and TABLE II describe sets and indices as
are provided and then followed by the conclusion in well as decision variables used in our optimization
Section 5. problem. Once the values of the decision variables are
2. Home energy management system determined by the proposed ILPHEM model, EMU issues
A Home Energy Management System (HEMS) the scheduling commands to the appliances.
consists of several parts. The Energy Management Unit
TABLE I: Sets and indices
(EMU) is the central controller that schedules the
appliances. As can been seen in Fig 1, The EMU receives Sets / Indices Description
following data through Home Area Network (HAN). i Appliance index
• Responsive appliances data: Start time, Stop I Set of appliances
time and Serving time of appliances services r Appliance request index
• Current appliances status R Set of appliance requests
n Time slot index
• Current Battery Status
N Set of timeslots
• Current power production rate of solar panels
• Current electricity price that is provided by the
TABLE II: Decision variables
smart meter
EMU predicts some other required data. It uses history Decision variables Description
of production and consumption of energy in home to Binary variable where 1 means
x(i,r,n) allocating electricity to ith appliance
predict future power consumption of non-responsive
in the rth request and the nth interval
appliances and future power production of solar panels. Binary variable where 1/0 means
3. ILP Home Energy Management model allocating electricity from
b(i,r,n)
(ILPHEM) storage/grid to ith appliance in the
rth request and the nth interval
Appliances in this model are divided to responsive and
non-responsive groups. In the responsive group, the
As stated in TABLE II, variable x decides whether the
allowable start and finish times of the appliance service is
appliance is turned OFF or ON, while variable b
considered to be known. In the non-responsive group, the
determines whether the appliance is supplied from
storage or grid in the case it is turned ON. For example, if , , , , 0 4
x(2,3,85) = 1 and b(2,3,85) = 1, EMU sends ‘ON with : (5)
storage electricity’ command to the 2nd appliance for its , 1, ,
3rd request in the 85th timeslot. TABLE III describes data (6)
inputs and constant parameters. Some of these data inputs , , , , 1 , , 1
are predicted and the rest are received via HAN.
Based on Equation (1) and Equation (2) , the variables are
TABLE III: Input Data, predicted data and constants binary. Equation (3) and (4) are for responsive
appliances. These equations constrain appliances to their
Input data Description
allowable service time. In these equations, s resembles
Realistic energy consumption of ith non-
RE(i,n) s(i,k) and f means f(i,k) which is itself equal to
responsive appliance in nth timeslot
s(i,k)+dmax(i,k). Equation (5) determines electrical energy
C(n) Electricity price in the nth timeslot
consumption of non-responsive appliances. For
Average energy consumption of ith responsive appliances, the servicing timeslots must be
E(i,r)
appliance in its rth request continuous. Therefore, Equation (6) constrains responsive
Realistic solar energy that is produced appliances to be served continuously. On the other hand,
RPV(n)
in nth timeslot if x(i,r,n-1) is 1 and x(i,r,n) is 0, x(i,r,m) is forced to be 0
Solar energy profile, which is a mixture for all m>n. This means that if an appliance gets OFF
of realistic solar energy produced in the after it is served, it can’t get ON once again for the same
past and predicted solar energy for the request. The optimization problem has following
PV(n)
future timeslots. That is: constraints for the storage.
for n > t: PV(n) = PPV(n)
for n ≤ t: PV(n) = RPV(n) ∑ ∑ , , (7)
Non-responsive appliance energy ∑ ∑ ∑ ∑ , (8)
consumption profile, which is a mixture , ,
of realistic energy consumption in the _ _ (9)
EE(i,n) past and predicted energy consumption
for the future timeslots. That is: ∑ ∑ , , 6 60 _ _ (10)
for n > t: EE(i,n) = PE(i,n) , , , , (11)
for n ≤ t: EE(i,n) = RE(i,n)
Earliest allowable start timeslot of ith Equation (7) constrains allocating electricity from
s(i,r) storage in the mth timeslot to the stored energy in the
appliance in its rth request
Allowable delay of ith appliance in its rth storage. Bat[m] which resembles ... is calculated in
dmax(i,r) Equation (8). Equation (9) and (10) determine the
request
Latest allowable finish timeslot of ith
maximum stored energy capacity and maximum
f(i,r) appliance in its rth request. That is:
discharge power respectively. In Equation (10), 6*60
f(i,r)=s(i,r)+dmax(i,r)
resembles the length of one timeslot in seconds. Based on
Equation (11), if there is no request from the appliances,
Number of timeslots required to service
d(i,r) the storage is not allowed to supply any electrical
ith appliance in its rth request
electricity.
Predicted data Description
Predicted energy consumption of ith 3.3 Objective function
PE(i,n)
non-responsive appliance in nth timeslot In our proposed home energy management system, the
Predicted solar energy that is produced goal is to minimize the household payment. Since the
PPV(n)
in nth timeslot problem is solved in every timeslot during a day, the
Constant data Description objective function should be defined for each timeslot
Max_cap_stor Maximum power capacity of storage individually. Therefore, being in the tth timeslot of a day,
Max_out_pow Maximum output power of storage the objective function can be defined as:
3.2 Optimization Problem Constraints Minimize ∑ ∑ ∑N C n E i, r x i, r, n b i, r, n
Constraints for appliances regarding the proposed
4. Experimental Results
optimization problem are as follows:
0 , , 1
We consider a home with seven appliances. Four of the
(1)
appliances are non-responsive with the specification
0 , , 1 (2) summarized in TABLE IV. For simplicity, we assume
only one request for each appliance. The solar energy
∑ , , , (3)
production pattern during a day is obtained from [12].
TABLE IV: Non-responsive appliances data In Fig. 2, the household payment values are gathered
and illustrated for the mentioned case approaches.
Appliances Hours/day Usage(kW)
Experimental results show that the proposed ILPHEM
Air condition 6 0.8
model decreases the household payment by 20% in
Refrigerator 9 0.6
comparison with the third approach.
PC 8 0.3
TV 8 0.3
3
TABLE V describes the remaining responsive

Household payment for one day ($)


appliances.
2.5
TABLE V: Responsive appliances data (r = 1)
2
Appliances kW E S d dmax
Clothes dryer 4.5 0.45 100 8 140 1.5
Washing Machine 1.2 0.12 130 5 110
Dishwasher 0.4 .040 1 5 100 1
We considered a three-level tariffs which are obtained
from [13] as illustrated in TABLE VI. The ILP model is 0.5
solved by using GAMS. The simulations are done on a
Core i5 computer with 4GB DDR3 memory. Each 0
simulation that was performed for each timeslot took Case 1 Case 2 Case 3 Case 4 ILPHEM
about 2 seconds.
TABLE VII: Three level tariff Fig. 2: Comparison of household payment of one day in
different case studies.
TOU period Time (hour) TOU rate (¢/kWh)
Off-peak [24pm-7 am] 9.866 In Fig. 3, total energy consumption of appliances is
Mid-peak [7 am-10am] 11.453 depicted for the third approach. Total energy
On-peak [10am-20pm] 14.958 consumption of appliances in the case of our proposed
Mid-peak [20pm-24pm] 11.453 ILPHEM model is also depicted in Fig. 4. As it can be
Off-peak [23pm-24pm] 9.866 seen from these figures, our proposed method shifts the
4.1 Household Payment servicing time of responsive appliances from ON-peak
period to OFF-peak to reduce the household payment.
To compare the obtained results with the conventional
For example, Cloth dryer servicing time is shifted from
approaches, we consider the following cases. In the first
10am to 23pm.
approach, there is a home with no controllable appliances
In Fig. 5, storage status is illustrated for case 3. In this
and no storage for storing the solar energy. It means that
case, the storage discharging begins from 65th timeslot. In
the responsive appliance cannot postpone the requests. In
Fig. 6, the storage status of ILPHEM model is depicted.
the second approach, the appliances are controllable, but
In this case, the storage discharging begins from 100th
there is no storage. In the third conventional approach,
timeslot. It can be seen that ILPHEM model decides to
the appliances are not controllable, but there is a storage
discharge storage in peak timeslots despite of case 3
that stores solar energy. However the storage cannot be
where the storage is discharged by appliances whenever
scheduled to supply the energy in the best possible
there are enough stored electrical energy.
interval and the appliances use the stored energy just after
the storage gets charged with enough energy. The last
conventional approach is the case where the appliances 0.6
are not controllable, however there is a storage that can
Total appliances energy usage (kWh)

0.5
be optimally scheduled to supply energy in the best
possible interval based on the predicted solar energy and 0.4
predicted workload.
TABLE III Different conventional approaches what are used for 0.3
comparison with our proposed ILPHEM model.
0.2
Controllable There are Controllable
appliances storage storage 0.1
Case 1 NO NO NO
Case 2 YES NO NO 0
Case 3 NO YES NO
1
15
29
43
57
71
85
99
113
127
141
155
169
183
197
211
225
239

Case 4 NO YES YES Periods


ILPHEM YES YES YES
Fig. 3: Total energy consumption pattern of appliances for case 3
0.25 possible interval in a seamless way. In this section, to
provide better insights into the effectiveness of proposed
method, we compared the seamless real time scheduling
Total appliances energy usage (kWh)

0.2
with a planned offline scheduling in the case of new
request. For the sake of simplicity, we just add a new
0.15 responsive appliance request with the following
specifications:
0.1 • snew(i=2,r=2)=200
• dnew(i=2,r=2)=5
0.05 • dmax_new (i=2,r=2)=40
• dmax_new(i=2,r=1) = dmax_old(i=2,r=1) – 40
0
ILPHEM reschedule the requests as it receives the new
1
21
48
73
85
104
114
127
137
147
157
167
177
197
207
217
227
237
Periods
request based on the updated input data. In TABLE VIII,
the result of the reschedule attempt made by ILPHEM is
Fig. 4: Total energy consumption pattern of appliances for ILPHEM
compared with an offline planned scheduling to show the
effectiveness of the real time reschedule attempts.
0.5 TABLE VIII shows that if the washing machine requests
0.45 electricity at 130th timeslot and changes allowable delay
0.4 of its servicing time for the first request, ILPHEM will be
Storage discharging (kWh)

0.35 able to reschedule the appliances for optimal household


0.3 payment. But the offline scheduler cannot reschedule
0.25 appliances and the washing machine servicing time is
0.2 done without delay.
0.15
0.1 TABLE VIII Responsive scheduling before and after 130th timeslot
0.05
0
Servicing time Servicing
Appliance by offline time
60
68
76
84
92
100
108
116
124
132
140
148
156
164
172
180
188
196

scheduler by ILPHEM
Periods Clothes dryer 231-238 231-238
Washing
Fig. 5: Storage discharging pattern for Case 3 131-135 155-159
Machine(r=1)
Washing
0.25 201-205 231-235
Machine(r=2)
Dishwasher 1-5 1-5
0.2 Household
Storage discharging (kWh)

1.28 ($) 1.26 ($)


payment
0.15

0.1 5. Conclusion

0.05
In this paper, we proposed an ILP-based home energy
management model to manage appliances and storage in
the presence of solar energy. We formulated the
0
constraints of appliances and the storage. ILPHEM
97
105
111
120
128
135
141
147
153
159
165
171
177
189
199
206

decides for appliances and the stored energy in storage


Periods based on the updated data. We showed that the proposed
ILPHEM model reduces household payment by 20%
Fig. 6: Storage discharging pattern in the case of ILPHEM model
compared to the best conventional approaches.
4.2 Responsive Scheduling
In a real scenario, the input data is changed in real
References
time. For example, actual and predicted produced solar
energy may differ in several timeslots. Also, an appliance [1] J. J. Conti, P. D. Holtberg, J. A. Beamon, A. M. Schaal, G. E.
Sweetnam, and A. S. Kydes, Annual energy outlook with projections
may have several requests with different specifications to 2035, report of U.S. Energy Information Administration (EIA), Apr.
during a day. As mentioned before in Section 2, our 2010 [Online]. Available: http://www.eia.doe.gov
proposed home energy management system is able to use [2] A. Al-Salaymeh, Z, Al-Hamamre, F. Sharaf, M.R. Abdelkader,”
Technical and economical assessment of the utilization of photovoltaic
the predicted request data and predicted solar energy data systems in residential buildings: the case of Jordan”, Energy
to optimally schedule the new requests in the best Conversion and Management, 51 (2010), pp. 1719–1726.
[3] T.J. Hammons, “Integrating renewable energy sources into
European grids,” in proc. 2006 universities power engineering conf.,
vol. 1, pp. 142-151.
[4] D. Niyato, L. Xiao, and P. Wang, “Machine-to-machine
Communications for Home Energy Management System in Smart
Grid,” IEEE Commun. Mag., vol. 49, no. 4, pp. 53-59, April 2011.
[5] S. L. Ullo, A. Vaccaro, and G. Velotto, “Performance Analysis of
IEEE 802.15.4 based Sensor Networks for Smart Grids
Communications”,Journal of Electrical Engineering: Theory and
Application, Vol. 1, No.3, Apr. 2010, pp. 129-134.
[6] Lee, Jang-Won, and Du-Han Lee. "Residential electricity load
scheduling for multi-class appliances with Time-of-Use pricing."
GLOBECOM Workshops (GC Wkshps), 2011 IEEE. IEEE, 2011.
[7] S. Ghaemi, G. Brauner, “User behavior and patterns of electricity
use for energy saving,” Internationale Energiewirtschaftstagung an der
TU Wien, IEWT, Feb. 2009.
[8] S. Park, H. Kim, H. Moon, J. Heo, and S. Yoon, “Concurrent
Simulation Platform for Energy-aware Smart Metering Systems,” IEEE
Trans. Consumer Electron., vol. 56, no. 3, pp. 1918-1926, Aug. 2010.
[9] M. Erol-Kantarci and H. T. Mouftah, “Wireless Sensor Networks
for Cost- Efficient Residential Energy Management in the Smart Grid,”
IEEE Trans. Smart Grid, vol.2, no.2, June 2011, pp. 314–25.
[10] Rastegar, Mohammad, Mahmud Fotuhi-Firuzabad, and Farrokh
Aminifar. "Load commitment in a smart home." Applied Energy (2012).
[11] Hong, Insung, Byeongkwan Kang, and Sehyun Park. "Design and
implementation of intelligent energy distribution management with
photovoltaic system." Consumer Electronics, IEEE Transactions on 58,
no. 2 (2012): 340-346.
[12] Parker, Danny S. "Very low energy homes in the United States:
Perspectives on performance from measured data." Energy and
buildings 41.5 (2009): 512-520.
[13] Baltimore gas and electric three-level summer’s tariffs. [Online].
Available:
http://www.bge.com/portal/site/bge/menuitem.dc72e1697738765822b7
5475da6176a0.

You might also like