You are on page 1of 6

Hierarchical Energy Management Strategy for a

Community of Multi Smart Homes


Sima Aznavi Poria Fajri Md. Rasheduzzaman
Electrical and Biomedical Engineering Department Electrical and Biomedical Engineering Department Dept. of Polytechnic Studies
University of Nevada, Reno University of Nevada, Reno Southeast Missouri State University
Reno, NV, USA Reno, NV, USA Cape Girardeau, MO, USA
s.aznavi@nevada.unr.edu pfajri@unr.edu mrasheduzzaman@semo.edu

Abstract—In this paper, a hierarchical energy management resources operation and household electricity consumption tasks
strategy for a community of multiple smart homes is proposed. are scheduled based on real-time electricity pricing. A charging
This strategy reduces the stress on the power grid caused by scheme is also adopted to reduce the peak demand from the grid.
uncoordinated energy scheduling of smart homes. Different Authors in [6] focused on smoothing out the load variance in a
stochastic operational conditions for plug-in electric vehicles household by coordinated charging of existing PEVs. They
(PEVs) and household battery storages are considered. First, cost- accomplished to reduce the variance of the load power and
effective day-ahead energy schedules under these operational energy losses by regulating the charging profiles of the PEVs.
conditions are calculated by individual home energy management To improve the efficiency and reduce the average cost of
systems (HEMSs). Then, a centralized decision making unit
electricity purchase, a distributed real-time electricity allocation
determines the day-ahead schedule for each household with the
aim of reducing the stress on the grid. Comprehensive results
is proposed in [7] for grid connected residential microgrids.
obtained from simulations show that the proposed hierarchical Research in [8] shows that the U.S. grid peak demand can
energy management strategy not only maintains a low energy increase by a factor of 2.5 to 6.5 due to PEV charging, assuming
purchase cost for each individual household, but also smooths the that every household in the U.S. owns only one PEV. This peak
overall grid power profile requested by all smart homes combined. load increases energy demand from the transmission grid and
can lead to further aging of the power system equipment and
Keywords—Day-ahead energy schedule, hierarchical energy frequent tripping of the relays during overload conditions. Thus,
management, stochastic fractal search (SFS) optimization, smart independent decisions on energy scheduling of households can
home, plug-in electric vehicles (PEV). cause undesired effects such as contingencies and instabilities in
the network [9]. Therefore, the interactions between the energy
I. INTRODUCTION management decisions of multiple households would be a
Smart homes equipped with renewable energy sources have challenging issue in future power management.
developed significantly within the past few years and play a In recent years, more attention has been directed to the
significant role in facilitating the penetration of renewable development of better optimization algorithms for smart home
energies and energy saving technologies. At the same time, energy management frameworks. Based on these efforts, it is
owing to the advancements in transportation electrification, clear that optimization for individual smart homes may have
plug-in electric vehicles (PEVs) have increasingly grown in multiple solutions that slightly differ in terms of energy cost. As
recent years. As flexible resources, the integration of PEVs with a result, a study on the selection of the best energy scheduling
smart buildings is creating new opportunities for the energy for a community of households is a worthy effort. Given this
management infrastructure. However, stochastic and background, this paper proposes a two-stage energy
uncoordinated PEV charging activities of smart households may management strategy for a system consisting of multiple smart
lead to undesired stress on the distribution system. homes. The goal of the proposed strategy is to reduce the energy
The existing research work has made good contributions to purchase cost for each household, while at the same time,
the energy management in smart buildings. An energy lowering the stress on the power grid. In the first stage, day-
management method in a building is proposed in [1] to reduce ahead energy scheduling is performed and the cost of purchasing
the impact of PEV charging on the grid with different charging energy from the grid is optimized separately through each
strategies. In [2] a hierarchical community management system household’s energy management unit. Such a shortsighted
is proposed that is based on scheduling combined cool and heat strategy may threaten the global benefits of the system. Thus, in
power plants and demand response (DR). DR is presented in [2] the second stage, the day-ahead energy schedules are sent to a
to smooth the electric tie-line power fluctuations by managing central decision making unit. The whole community is served by
loads and distributed generations. Fair cost distribution among the central decision making unit which is also responsible for
multiple smart homes and the microgrid through energy determining the power exchanges with the distribution grid. It
management system is presented in [3]. A two-layered energy aggregates the energy schedules from individual energy
management strategy to smooth the fluctuations of the electric management units. Then, it organizes those schedules to reduce
power exchanges with respect to uncertainties is presented in [4]. the harsh fluctuations of the overall day-ahead energy imported
An optimal scheduling of smart home energy consumption is from the grid. This is an effective approach to alleviate the stress
studied in [5] with the aim of minimizing the forecasted day- on the network by selecting the right combination of smart home
ahead energy usage cost. To achieve this, distributed energy energy schedules and distributing the required energy over the
24-hour period.

978-1-5090-6684-1/18/$31.00 ©2018 IEEE 176


II. SMART HOME ENERGY MANAGEMENT SYSTEM energy cost. By taking advantage of this, the owners and the grid
Over the past few years, classical residential building would benefit from this flexibility through selecting the
structures have evolved into smart homes equipped with household energy profiles that not only keep the owners’ energy
advanced energy features. Smart homes are small energy cost down, but also their interaction can flatten the grid power
systems which may be equipped with distributed generation, profile at PCC. Thus, a two-stage hierarchical energy
energy storage and loads. Coupled with the vehicle-to-grid management strategy for a community of smart homes is
technologies, smart homes will play an important role in energy presented in this paper with the goal of reducing the energy
conservation of future smart cities. Hence, the energy purchase cost for the end users, while minimizing the stress on
management of smart buildings has become a global priority in the power grid. The two stages of the proposed hierarchical
recent years. energy management system are explained below and its overall
scheme is illustrated in Fig. 1.
Home energy management system (HEMS) of a smart home
attempts to manage the energy flow by scheduling the energy of A. Stage 1
existing renewable sources and grid. The key functions of This stage consists of individual HEMS for each household
HEMS are to monitor, control, and optimize this energy flow which provide cost-effective day-ahead energy scheduling while
within the household components, while meeting the goals such satisfying household load demand and operational constraints.
as comfort and economic incentives of the owner. Several The operational constraints are discussed in Section A.2. Several
approaches exist for the HEMS to achieve smart home energy schedules may exist for each household that all satisfy
operational goals. Optimization techniques are typically used to user benefits and slightly differ in energy cost. Thus, in this
determine the energy schedule over a given time horizon with a stage, different day-ahead energy schedules that yield the lowest
specific objective. The main advantage of an optimization cost and are in the same price range are selected from the
technique is providing the optimal or near-optimal solutions optimization results. This increases the flexibility of the central
with respect to any objective. decision making unit in stage 2, to choose the final energy
schedule for the community based on the results of stage 1. Thus,
In neighborhoods consisting of several smart homes, a in stage 1, several candidate schedules with low operating costs
neighborhood energy management system enables the smart are selected from the optimization results to be sent to the second
homes equipped with individual HEMSs to communicate and stage.
coordinate their actions through communication paths [9]. In this
paper, an energy management system for a community of smart A.1 Objective Function
homes located in a neighborhood area is presented. This In the first stage, the objective of the HEMS equipped with
community of smart homes resembles a hierarchically controlled optimization is defined such that the total buying cost from the
grid connected microgrid from the energy management grid is minimized. For this study, a one-day cycle is considered.
viewpoint. In a hierarchically controlled microgrid, the The 24-hour period of a day is divided into smaller time intervals
centralized control system has the main responsibility for (Δt = 30 minutes) which are the smallest units of time used by
energy scheduling of sources, participating in the energy market, the energy management strategy. 15 minutes and 1 hour
providing ancillary services, and so on [10]. Likewise, the sampling times have also been examined during the simulation
hierarchical control system for a community of smart homes development phase. It was observed that shorter time span,
manages the electricity usage of all smart homes. According to considerably increased the simulation time and calculation
this infrastructure, the central decision making unit has direct burden, while a 1 hour sampling time slightly decreased the
access to the whole information of individual HEMSs and precision of calculations. The total cost is calculated to be the
assigns the energy scheduling for each household based on the sum of the hourly costs of all instants over a 24-hour period and
operational strategy of the community. Moreover, the central is given as
decision making system is responsible for determining the
energy exchanges with the grid. The decisions carried out by the ∑ ∗ (1)
central decision making unit are then sent to each HEMS and the
strategy is applied. where is the absorbed energy from the grid in kWh and
¢
III. PROPOSED CENTRALIZED HIERARCHICAL ENERGY is the grid price in ( ) at hour h.
MANAGEMENT
A.2 Constraints
In a community of multi smart homes, the harsh fluctuations
of imported power from the network can impose an additional In the optimization process, the operating constraints of the
burden on the distribution grid. Therefore, smoothing the PEV, household battery storage, and system total power balance
fluctuations of power at the point of common coupling (PCC) should be considered.
will benefit both the grid and the community. HEMS equipped A.2.1 Power Balance
with optimization algorithms can effectively reduce the cost of
purchasing the energy for each household. However, this does The amount of generated energy should equal the consumed
not necessarily guarantee a smooth profile at the PCC. On the energy at every instant, h. This is given by
other hand, the results of each optimization may yield multiple
daily energy schedules that have very little cost difference. Thus, (2)
a cost-effective energy management strategy may have several
solutions in terms of the day-ahead profile that slightly differ in

177
Household 1 ... Household (2)‐(k‐1) ... Household k

Stage1: Day‐ahead Cost


Objectives Day‐ahead data forecasts Day‐ahead data forecasts Objectives
Load profile, PV output Load profile, PV output
Constraints Grid price Grid price Constraints

Reduction
Day‐ahead dispatch Day‐ahead dispatch
Minimum daily cost Minimum daily cost

Day‐ahead Day‐ahead Day‐ahead Day‐ahead


... ...
schedule 1 schedule n schedule 1 schedule n

Stage2: Grid Power


Smoothing
STD (1) ... STD (nk)

Minimum STD

Day‐ahead schedule ... Day‐ahead schedule


of household 1 of household k

Fig. 1. Flowchart of the proposed hierarchical energy management system.

where is the output power of the PV at hour h. and . The constraints for the battery storage and
and are the power of the household battery and PEV at PEV are given in (7) and (8).
hour h, respectively. is the imported power from the ∗
grid and is the household load at hour h. All values are 1 (5)
in kW.

The battery and PEV power limits can be described by (3) 1 (6)
and (4), respectively.
_ _ (7)
_ _ (3)
_ _ (8)
_ _ (4)
where and are household battery storage and
where _ and _ are the minimum allowed PEV battery capacity in kWh.
charging/discharging power of the household battery and PEV, B. Stage 2
respectively. _ and _ are the maximum allowed
In this stage, selected day-ahead energy schedules obtained
charging/discharging power of the household battery and PEV,
for each household from stage 1 are analyzed with the aim of
respectively.
minimizing the stress on the grid. The central decision making
A.2.2 State of Charge unit organizes the household schedules to reduce the standard
Equations (5) and (6) represent the state of charge (SOC) deviation (SD) of the cumulative day-ahead imported energy
calculations for the battery and PEV. Due to lifetime from the grid. It aggregates all the day-ahead schedules received
considerations and to avoid over charging or deep discharging, by the HEMSs. Different combinations are evaluated such that
the SOC for both PEV and battery is limited to a range set by each combination consists of only one schedule from every
household. In the decision making stage, the SD of the grid

178
power profile is considered as the criterion for selecting the best The initial SOC of PEVs were randomly selected between 35%-
profile according to (9). 50% and the initial SOC of each household battery storage was
randomly chosen within the range of 30%-45%. The household
:
battery capacity was considered to be 14 kWh. Other system
,
(9) parameters used for this simulation are listed in Table I. The
, , ,∀ ∈ 1,2, …
final desired SOC of the PEV was also set to 70 2% for
simulations to guarantee enough energy for the next trip. For
where, is the best profile with the least SD. , this simulation, the utility grid price and load data, shown in
is the 24-hour power profile at PCC as a result of the ith Figs. 3 and 4, were extracted from PJM and ComEd's and for
combination which includes only one day-ahead schedule from March 31st, 2017, respectively [12], [13].
each household. n is the number of candidate schedules from
each household and k is the number of households. TABLE I. SYSTEM PARAMETERS USED FOR SIMULATION
The combination which minimizes the total SD guarantees a Parameter Value Parameter Value
smooth profile at the PCC. Once this combination is found, its SOCmax of PEV 90%
SOCmax of household
80%
composing candidate profiles are assigned to each battery storage
corresponding HEMS as the final day-ahead energy schedule for SOCmin of household
SOCmin of PEV 30% 20%
battery storage
that household.
Final desired
70 2% PEV battery capacity 30 kWh
SOC of PEV
IV. MULTI SMART HOME TEST SYSTEM UNDER STUDY
In this paper, a community that consists of ten smart homes
(k=10) is considered. Each household consists of home
appliances, a PEV, PV panels, and a household battery storage
system. The community is connected to the grid and can import
unlimited power from the grid. It is assumed that power flow
from the grid is unidirectional and the smart homes do not sell
back energy to the grid. It is also assumed that the PEV and the
battery both have bidirectional power flow capability and can
charge or discharge, within their limits, at any given time. Fig. 2
shows the structure of the system comprised of multiple smart
homes. The energy schedule of each household assigned by the
central decision making unit is implemented by the
corresponding HEMS which is responsible for the control of
energy exchange between the PV, home appliances, PEV,
battery storage, and the utility grid.
Fig. 3. Photovoltaic generation and load demand.

Fig. 2. Multi smart home test system under study.

In this paper, stochastic fractal search (SFS) optimization


Fig. 4. Grid hourly price.
algorithm is applied to optimize the day-ahead energy cost. SFS
was first introduced in [11] and is used to solve both constrained
V. SIMULATION RESULTS
and unconstrained, either single objective or multiobjective
optimization problems based on Fractal concept [10]. In this study, two scenarios were considered. In the first
Simulations were carried out for various scenarios with scenario it was assumed that all the households were only
different PEV arrival times and initial PEV and household equipped with individual HEMSs of stage 1. Fifty optimizations
battery SOCs. In this case study, ten random scenarios were were performed and the solution with the lowest cost was
selected as the operation conditions of the ten households. The selected as the final day-ahead profile. Equation (1) was
PEVs arrival time were randomly chosen between hours 16:00- considered as the objective function and day-ahead cost
minimization for each household was performed considering the
20:00 and it was assumed that they leave at 8:00 the next day.

179
operational constraints described in (2)-(8). In the second
scenario, the hierarchical energy management system, equipped
with stage 2 was added to the system with the objective of
minimizing SD of the grid power at PCC. Fig. 5 and Fig. 6 show
the minimum cost of the day-ahead schedules for households 1-
5 and 6-10 of scenario 1, respectively.

Fig. 7. Three candidate schedules with the lowest price for household 3.

In stage 2, the central decision making unit calculates the SD


of the grid profile for all combinations. Since each combination
includes ten daily schedules, and only one candidate from every
Fig. 5. The cheapest day-ahead schedules for households 1-5. household takes part in the combination, therefore, 310 scenarios
are analyzed in the second stage. Finally, the combination with
the least SD is selected as the final combination and based on
this result, the daily scheduling for each household is assigned
by the central decision making unit. Table III shows the
candidate schedules for each household sent to stage 2, with the
schedules that contributed to SD minimization highlighted as the
final outcome.

TABLE III. CANDIDATE PROFILES AND OPTIMUM OUTCOME SCHEDULES FOR


SCENARIO 2
Cost (¢) Cost (¢)
Profile
1 2 3 1 2 3
Number
Household 1 172 174 175 Household 6 187 188 190
Fig. 6. The cheapest day-ahead schedules for households 6-10. Household 2 176 177 178 Household 7 182 183 186
Household 3 180 182 183 Household 8 177 178 180
Household 4 182 183 184 Household 9 176 177 179
In scenario 2, the first stage was considered to be the same Household 5 182 183 184 Household10 172 173 174
as scenario 1, however, three candidate schedules (n=3) with the
lowest operating costs were selected amongst the fifty solutions Fig. 8 shows the community’s daily power profile at the PCC
of each household. These candidates were sent to the second for both scenarios. According to these results, the SD for
stage. Table II shows the cost of energy for the three candidates scenario 1 is 6.25 kW, while this value is 1.64 kW for scenario
of each household, obtained from stage 1 which include 2, which is much lower in comparison. The results also show
minimum and near minimum prices. According to Table II, it several peaks in grid power for scenario 1, while scenario 2 has
can be seen that the difference between the total daily cost of a smoother profile. Further analysis of the results reveals that
purchasing energy from the grid is negligible for the three after hour 8:00 when the PEVs leave, the grid profile for both
selected candidates. Therefore, the user’s benefit is not scenarios are almost the same. During this time, the imported
sacrificed by choosing one schedule over another. As an power from the grid starts to decrease. On the other hand, with
example, Fig.7 shows three candidate schedules for household the absence of the majority of PEVs during hours 12:00-18:00
3. By comparing the results of this figure and the results obtained and higher PV generation during this period, the required grid
from Table II, it can be seen that although the cost difference is power is at its lowest value. However, during hours 20:00-8:00
less than 2%, the optimization yields totally different day-ahead and in the presence of all PEVs, since the generated PV power
profiles. is at its lowest value, the required power is mostly imported from
TABLE II. OPTIMUM RESULTS FOR THREE CANDIDATE PROFILES the grid. On the other hand, during this interval, the residential
Cost (¢) Cost (¢) load is also low and therefore the PEVs have the greatest effect
Profile
1 2 3 1 2 3 on the daily profile. From this viewpoint, it can be concluded
Number that the proposed energy management system is more effective
Household 1 172 174 175 Household 6 187 188 190
in reducing the peaks and smoothing the gird profile during the
Household 2 176 177 178 Household 7 182 183 186
Household 3 180 182 183 Household 8 177 178 180 time when PEVs are present.
Household 4 182 183 184 Household 9 176 177 179
Household 5 182 183 184 Household10 172 173 174

180
[2] X. Xu, X. Jin, H. Jia, X. Yu, and K. Li, “Hierarchical management for
integrated community energy systems,” Applied Energy, vol. 160, pp.
231–243, 2015.
[3] D. Zhang, S. Liu, and L. G. Papageorgiou, “Fair cost distribution among
smart homes with microgrid,” Energy Conversion and Management, vol.
80, pp. 498–508, 2014.
[4] X. Jin, J. Wu, Y. Mu, M. Wang, X. Xu, and H. Jia, “Hierarchical microgrid
energy management in an office building,” Applied Energy, vol. 208, pp.
480-494, 2017.
[5] D. Zhang, N. Shah, and L. G. Papageorgiou, “Efficient energy
consumption and operation management in a smart building with
microgrid,” Energy Conversion and Management, vol. 74, pp. 209–222,
2013.
[6] L. Jian, H. Xue, G. Xu, X. Zhu, D. Zhao, and Z. Y. Shao, “Regulated
charging of plug-in hybrid electric vehicles for minimizing load variance
Fig. 8. Day-ahead imported power from the grid for scenario 1 and 2. in household smart microgrid,” IEEE Transactions on Industrial
Electronics, vol. 60, no. 8, pp. 3218–3226, 2013.
[7] M. H. K. Tushar, C. Assi, and M. Maier, “Distributed Real-Time
VI. CONCLUSIONS Electricity Allocation Mechanism for Large Residential Microgrid,” IEEE
In this paper, a hierarchical energy management strategy for Transactions on Smart Grid, vol. 6, no. 3, pp. 1353–1363, 2015.
a community of multi smart homes was proposed. This strategy [8] A. Ghavami, K. Kar, and A. Gupta, “Decentralized Charging of Plug-in
Electric Vehicles with Distribution Feeder Overload Control,” IEEE
is equipped with a centralized decision making unit which is Transactions on Automatic Control, vol. 61, no. 11, pp. 3527–3532, 2016.
responsible for reducing the stress on the grid at PCC. By [9] B. Celik, R. Roche, S. Suryanarayanan, D. Bouquain, and A. Miraoui,
organizing the day-ahead schedules obtained from each “Electric energy management in residential areas through coordination of
household’s HEMS, it minimizes the SD of the overall imported multiple smart homes,” Renewable and Sustainable Energy Reviews, vol.
energy from the grid. Simulation results showed that the 80, pp. 260–275, 2017.
proposed centralized hierarchical strategy is effective in [10] S. Aznavi, P. Fajri, M. Benidris, and B. Falahati, "Hierarchical droop
flattening the grid power profile at PCC, especially during the controlled frequency optimization and energy management of a grid-
connected microgrid," in 2017 IEEE Conference on Technologies for
presence of PEVs. Sustainability (SusTech), Phoenix, AZ, USA, 2017, pp. 1-7.
REFERENCES [11] H. Salimi, “Stochastic Fractal Search: A powerful metaheuristic
algorithm,” Knowledge-Based Systems, vol. 75, pp. 1–18, 2015.
[1] J. Van Roy, N. Leemput, F. Geth, J. Buscher, R. Salenbien, and J. Driesen, [12] Available online: https://hourlypricing.comed.com/live-prices/predicted-
“Electric vehicle charging in an office building microgrid with distributed prices/
energy resources,” IEEE Transactions on Sustainable Energy, vol. 5, no.
4, pp. 1389–1396, 2014. [13] Available online: http://www.pjm.com/markets-and-operations/ops-
analysis/historical-load-data.aspx

181

Powered by TCPDF (www.tcpdf.org)

You might also like