Design and Implementation of a Load Scheduling

Embedded System for Off grid Solar Power Systems

Kizito Masaba, Amini Ntakirutimana and Taha Selim Ustun
School of Electrical and Computer Engineering
Carnegie Mellon University
{kmasaba and antakiru}@andrew.cmu.edu

Abstract— Power drives most activities in health centers and renewable energy off grid systems in such regions an alternative
homes. However, its availability and high cost make it an worth considering[2]. The operation and maintenance costs of a
expensive commodity. Off grid power systems like PV-battery renewable energy based off grid system, especially solar, are
setups are more sensitive to fluctuations in the generation and lower due to the availability of the resource. The system also
load. Power generation in such systems is also intermittent due to generates clean energy which is important in environmental
dynamic weather changes and other natural factors which require conservation and control of the carbon content in the atmosphere
inevitable use of batteries. Batteries are the most expensive part of [3].
off-grid solutions and their use has to be optimized. Furthermore,
these systems, like any other power source, have maximum With the steady drop in the prices of Solar PV, installation
capacities beyond which they should not operate. If these of an off grid solar system has become cheaper than ever. The
limitations are not well observed, usage of the available power remote regions in Africa are seizing the opportunity and several
cannot be maximized and the power equipment may be damaged. health centers and public institutions are creating their own off
In this regard, this paper proposes an embedded system solution grid systems in order to offer more services to the
that schedules appliances to run when there is available power community[4]. However, there is still hesitation in the use of
from solar panels. The appliances are connected to the smart Solar PV due to the high upfront cost of batteries that store
scheduler through a dedicated smart plug that powers the excess energy for use at times of less or no generation from PV.
appliance depending on the conditions set by the smart scheduler.
The cost of operation and maintenance of the batteries is also
This solution not only ensures that loads are served at the earliest
quite high in terms of management. In [5], the optimal
possible availability, but also avoids blackouts in the system.
Furthermore, smart utilization of the resources minimizes the
conditions for maximizing the battery life cycle in a photovoltaic
battery need and hence, the investment costs. system are discussed. One of the conditions is maintaining a
Keywords—load scheduling; off grid systems; photovoltaic good Battery Array to Load ratio (A: L) to ensure that the
systems; smart plug; smart scheduler; load control batteries are not over loaded. However, such a solution requires
large number of batteries, hence large investment costs. The
NOMENCLATURE charge and discharge times of the batteries also have to be
Constant Considerable power to serve the loads observed. When these conditions are not observed, PV systems
become costly in terms of maintenance.
Earliest time of day for obtaining
Latest time of day for obtaining The challenges raised call for the need to reduce on the
Power demand for any load number of batteries required to run the Solar PV off grid system
Running duration for each load and also maintain the batteries at optimal conditions in order to
Max. number of loads that can be served by minimize the cost of running an off grid solar PV system.
, Start time for ith scheduled load Demand side management became more feasible with the
, Finish time for the ith scheduled load tremendous opportunities that ICT brings in the energy sector.
Total Power being consumed by each scheduled load A lot of work has been done in this field with major focus on
Array list of all finish times , reducing the peak load on smart grids. Most of the work suggests
Time when smart scheduler receives a request algorithms [6] and models that can be implemented to minimize
Identifier for the incoming load the cost of energy and also cut the peak load on the grid.
However, they all point at managing terminals that are
I. INTRODUCTION connected to the grid. The utilization of solar power in off grid
Electricity (or electrification) plays a key role in boosting the PV systems is barely considered in their models. The work in
country’s economy. In Africa, the electrification rate is slowly [7] discusses various load scheduling models and ranks them
growing and most regions are still far off the grid. This is according to the percentage of the peak load they shave off the
attributed to the geographical size and infrastructural challenges grid. This is important when choosing the scheduling model to
in these regions [1]. Consequently, they rely on kerosene, implement.
batteries, candles and wood for their energy. Diesel generators In [8], several models are created to show how various
are also commonly used in commercial centers and factories. classes of home appliances consume power. The work in [9]
However, these hydro-carbon sources get depleted and are considers the off grid systems in demand side management by
harmful to humans. This makes the implementation of proposing systems that incorporate distributed energy resources

978-1-5090-3358-4/16/$31.00 ©2016 IEEE

Authorized licensed use limited to: SRM Institute of Science and Technology. Downloaded on August 28,2024 at 04:50:26 UTC from IEEE Xplore. Restrictions apply.
in the peak load control and management of the grid. However, where P is the power generated by the PV module at time t, A
this does not focus on the maximization of power usage of the is the area of the PV module and R is the solar radiation at t. k
energy generated by the local energy resource but rather is a constant that caters for the energy losses due to temperature.
determine when to switch to the grid or remain on the local Ideally, the power generation of a PV module throughout the day
energy source, and also calculate the power to feed the grid. In varies as shown by the dotted line in Fig.1.
[10], an embedded system is implemented to perform load
scheduling of home appliances basing on the pricing
information of energy from the grid. This also focuses on
minimizing energy costs and cutting the peak load. The system
collects power usage plans from the user and price of power
from utility to create a scheduling model for the home
appliances. The research in [11] presents an implementation of
a smart plug that automatically collects appliance’s power
consumption information and also switches the appliances.
Control of power wastage through power consumed by Figure 1: Overloading and underutilization of PV systems
appliances while in standby mode is implemented in that
research using embedded systems. This also aims at cutting The loads running from t to t and t to t are overloading
demand and the cost of power. the system whereas power is being wasted between t and
t since the demand is less than the available power. To solve
Generally, most previous work aims at managing the grid, these two challenges, an appropriate load scheduling algorithm
but not much has been done to manage and optimize the usage has to be implemented to ensure that there is always a balance
of power generated by off grid systems. The work in [12] between the power generated and the available demand. This
presents a smart metering system using Bluetooth as the algorithm can be created based on four different possibilities.
communication medium. This is also pertinent in this research, The first possibility is the loads having the same power demand
especially when it comes to the interconnection of the scheduler and run for the same duration. The second case is where the
to the loads. The smart scheduling approach of this paper has loads have the same power demand but run for different
been proposed in [13] and implemented in [14] using computer durations. The loads may also have different power demands but
systems. However, given the energy requirements and required run for the same duration, or they may have different power
expertise to run the system, that implementation may not be demands and run for different durations. The fourth possibility
feasible in such communities. Therefore an embedded system is the most practical since it considers the diversity in power
based approach has been assumed to maximize the usage of demand and operation of the loads.
power generated by an off-grid PV system and improving its
reliability. This aims at minimizing the need for back up A. Same Power Demand, Same Running Duration Algorithm
batteries since most loads run at a time when the PV array This algorithm implements the first scenario discussed in the
generates enough power. At times of less or no power, the previous section. This is the least probable and simplest
system maintains the optimal conditions for the batteries to scenario. However, since the major focus of this paper is
avoid being drained by the loads. Moreover, it avoids embedded system development, this scenario is selected for
overloading the system and consequent brownouts/blackouts. implementation but should other algorithms be developed, they
This system has a major benefit of maximizing the use of power can easily be implemented in the embedded system presented
generated by the PV system since it ensures that the loads run here.
during the time of maximum generation. This solves the
challenge of power wastage, hence reducing the break-even To model this algorithm, a constant considerable amount of
timeframe. power for scheduling,P and the minimum and maximum
time of day, t and t , respectively, when this power can
This paper is divided into four sections. Section II discusses be generated are assumed. The scheduled loads are stored in an
the proposed load scheduling algorithm for a PV off grid system. array and each incoming load is added to the schedule array with
Section III details the design and implementation of the respect to the indices of the currently served loads in the array.
proposed embedded system. The conclusion and future work are For any incoming load, its start and finish time can be modeled
discussed in Section IV. as stated in (2) and (3).
II. PROPOSED LOAD SCHEDULING ALGORITHM ×
, = (% ) + (2)
The challenges being addressed in this paper are under
utilization of the generated power and overloading the PV
system. To put this into perspective, the daily PV power , = (% ) +( + 1) × (3)
generation is studied. The total power generated by a PV system
depends on the area and yield of the PV module, and the average The start time of the i load, t , , is determined by
solar radiation on the PV panel at any given time. This is adding the finish time of the load that is in position i%SL of
represented by (1).
the array to the product of and load running duration,
P = A×k×R (1)
d . The load finish time,t , , on the other hand is the sum
of the finish time of the load at the same position and the running

Authorized licensed use limited to: SRM Institute of Science and Technology. Downloaded on August 28,2024 at 04:50:26 UTC from IEEE Xplore. Restrictions apply.
duration d multiplied by + 1. Whether the load is availability in the socket. Needless to say, the socket on the
smart plug is where the load is plugged. The high level
waiting or running, the user can cancel the operation at his/her architecture of smart scheduler and smart plug are shown in
discretion. In this case, the load is removed from the array and Figure 3.
the wait time for the waiting loads have to be adjusted
accordingly.
III. DESIGN AND IMPLEMENTATION OF THE SMART
SCHEDULING SYSTEM
This section discusses the whole process of the proposed
smart scheduling system design and implementation. The last
subsection summarizes the experiments done with the
implemented system and results obtained.
A. System Requirements
The system has a smart scheduler that schedules all
appliances based on their power rating and priority using a
scheduling algorithm mentioned. The connection between the Figure 3: Smart Plug (Left) and Smart Scheduler (Right) High Level
Architecture
smart scheduler and appliances is through a network of smart
plugs onto which each appliance is plugged. Each appliance C. Operation of the Smart Plug
connects to a specific smart plug that has been specially All appliances have smart plugs to which they are connected.
programmed for it. The smart plug contains power consumption Once plugged, the smart plug sends the ID of the connected
and run duration information for that appliance which it appliance to the smart scheduler so that the appliance can be
transmits to the smart scheduler while sending a schedule added to the schedule. This request to the smart scheduler can
request. The smart plug controls the connection of the appliance only be made when the button on the smart plug is pushed ON
to the smart scheduler and the power source. It communicates by the operator. The smart scheduler then responds with the time
with the smart scheduler that determines the time when to switch the appliance has to wait until it gets powered. On plugging the
the appliance on or off. Communication between these two appliance, the tricolor LED lights red. When the button is
devices can be of any technology while Bluetooth is used for this pushed the smart plug sends the schedule request to the smart
particular implementation. The conceptual diagram is shown in scheduler which responds with the time this particular load has
Figure 2. to wait before it is served. At this point, the LED lights yellow.
The wait time is displayed and counted down from the LCD until
the start time for the appliance to run is reached. It is then that
the smart plug immediately triggers the relay to power the
appliance. At this moment, the tricolor LED lights green. The
appliance then runs for its time slot and switched OFF thereafter.
The user may as well decide to stop the appliance before its time
slot expires. This is done by simply pushing the button the
second time. When the button is pushed, a cancel schedule
request is sent to the smart scheduler which removes the
appliance from the schedule list and the appliance is turned off.
The flowcharts in Figure 4 summarize the operation of the smart
Figure 2: Conceptual Diagram for the proposed Scheduling Embedded plug.
System
D. Operation of the Smart scheduler
B. System Design
The only work of the smart scheduler is to run a load
As already mentioned, the system is a network of smart plugs scheduling algorithm that ensures that at any instant, the
and a central smart scheduler, with Bluetooth as the medium of available power and demand are balanced. In this case, the
communication. A PIC16F887 microcontroller is being used as algorithm discussed in section II is implemented. Once the smart
the control unit in both the smart plug and smart scheduler. A scheduler is powered, it gets P ,t ,t for the
HC-06 Bluetooth module manages the communication and a 16 day,P and d , and starts waiting for schedule requests
x2 LCD is used to display the feedback information. The smart from the smart plugs. On receiving the request, P is
plug has got extra features that enable it to control the load and calculated and the available power is determined by getting the
also give feedback to the user. The extra features are the relay, difference between P and P . If the available power is
tricolor LED and the socket. The micro controller of the smart enough to serve the load, it is powered instantly. Otherwise, it is
plug triggers the relay to switch the load depending on the scheduled to wait for a finite time. In cases where if a load is
commands received from the smart scheduler. The tricolor LED scheduled shall run beyond t , the schedule request is
shows the state of the plugged load. The Green color of the LED dropped. The same applies to schedule requests whose power
implies that the Load is being served by the smart scheduler, demands are higher than P . The smart scheduler
yellow represent the waiting state, whereas the red shows power maintains an array of finish times , t , for both running and

Authorized licensed use limited to: SRM Institute of Science and Technology. Downloaded on August 28,2024 at 04:50:26 UTC from IEEE Xplore. Restrictions apply.
waiting loads. The array is refreshed every second to ensure that
t , that expires is removed from the array. The flow charts
in Figure 5, 6 and 7 summarize the algorithm.

Figure 5: Flow chart showing the operation of the smart scheduler

Figure 4: Flowchart showing the operation of the Smart Plug

E. System Implementation
The implementation was done on a Lab X1 Programmer kit
using MikroC. Table 1 lists the components that were used to
implement both the smart plug and smart scheduler. Figure 8
shows the lab setup of the smart plug. When a button is pushed,
a request is sent to the smart scheduler through the Bluetooth
module. The smart scheduler responds with the time the smart
plug should wait before it switches ON the load. This wait time
is counted down from the LCD. Once the time is up, the load is
turned ON and the usage time counter is started. The load runs
for a specific duration, as instructed by the smart scheduler and
is then switched OFF once the appointed usage time expires.
TABLE I: LIST OF SYSTEM COMPONENTS

Component Type
Micro Controller PIC16F887
Bluetooth Modules HC-05 and HC-06
LCD 16x2
4 Relay Module SRD-05VDC-SL-C
PIC Lab Kit X1
LEDs Red, Yellow and Blue
Figure 6: Flowchart for scheduling a load

Authorized licensed use limited to: SRM Institute of Science and Technology. Downloaded on August 28,2024 at 04:50:26 UTC from IEEE Xplore. Restrictions apply.
 F. Experiment and Results
The load scheduling embedded system prototype was
successfully developed and tested in the laboratory. In the
experiment, LEDs were connected to the Relays of different
smart plugs and schedule requests sent to the smart scheduler.
The smart scheduler sent back different wait times to each smart
plug. The smart plugs counted down the wait time and turned
ON their respective LED when the wait time had elapsed. All
the LEDs kept lighting for the same duration and were
automatically turned OFF by their respective smart plug when
the running duration had elapsed.

IV. CONCLUSION AND FUTURE WORK

This paper discusses the design and implementation of a load
scheduling embedded system for off grid solar systems. This
implementation is necessary in the management of a PV system
as it maximizes the use of power being generated. It also
minimizes the size of the storage system required to save the
excess power, since the loads are scheduled at the time of
generation. Hence the installation cost is greatly reduced.
Moreover, the solution minimizes wastage of power generated
since it ensures consumption of most amounts during
generation.
As future work, the feature for collecting data about the
frequency of loads on the schedule list and generated power shall
be added. This shall enable the system to make more intelligent
decisions like prioritizing loads and reporting performance of
the PV system. As illustrated in Figure 9, load 1 is first
scheduled and runs for its time slot. Then it requests for schedule
again at the same time with load 6. So the smart scheduler gives
load 6 higher priority over load 1 because the latter has run once
more than its counterpart.

Figure 7: Flowchart for cancelling a scheduled load

Figure 8: Smart Scheduler (Left) and Smart Plug (Right) in their default states Figure 9: Prioritizationof Loads based on Service History

Authorized licensed use limited to: SRM Institute of Science and Technology. Downloaded on August 28,2024 at 04:50:26 UTC from IEEE Xplore. Restrictions apply.
 ACKNOWLEDGMENT [8] D. Li, X. Liu, S. Lin, L. Cui, and J. Ren, “Design and load
control strategy of smart domestic electric system,” Proc.
This work is supported by Carnegie Mellon University under 2014 Int. Conf. Intell. Green Build. Smart Grid, IGBSG 2014,
Smart Embedded Solutions Grant, 2016. no. 51177098, 2014.
[9] R. Sureshkumar and K. R. Remya, “An Efficient Energy
REFERENCES Management and Household Appliances Scheduling for
[1] N. Cook, R. J. Campbell, P. Brown, and M. Ratner, Residential Customers using Embedded System,” Int. J.
“Powering Africa : Challenges of and U . S . Aid for Emerg. Technol. Adv. Eng., vol. 4, no. 1, pp. 623–628, 2014.
Electrification in Africa,” Congr. Res. Serv., 2015. [10] M. Chen and C. Lin, “Design and Implementation of a Smart
[2] K. S. Lackner, “Comparative Impacts of Fossil Fuels and Home Energy Saving System By Multi-Microprocessors,”
Alternative Energy Sources,” Issues Environ. Sci. Technol., Consum. Electron. - Taiwan (ICCE-TW), IEEE Int. Conf. 6-8
pp. 1–41, 2010. June 2015 ,Taipei, Taiwan, pp. 410–411.
[3] J. Goldemberg, “The Case for Renewable Energies,” Int. [11] M. Jahn, M. Jentsch, C. R. Prause, F. Pramudianto, A. Al-
Conf. Renew. Energies, February, 2004, Bonn, p. 16. Akkad, and R. Reiners, “The energy aware smart home,” 5th
[4] M. B. Philippa Marshall, Phil Wallace, “Module 14: Demand- Int. Conf. Futur. Inf. Technol. (FutureTech), IEEE., pp. 1–8,
side management sustainable energy regulation and 2010.
policymaking for africa,” Sustain. Energy Regul. [12] B. S. Koay, S. S. Cheah, Y. H. Sng, P. H. J. Chong, P. Shum,
Policymaking Africa, Energy Sect. Africa, pp. 8–9. Y. C. Tong, X. Y. Wang, Y. X. Zuo, and H. W. Kuek, “Design
[5] J. P. Dunlop and B. Farhi, “Recommendations for and implementation of Bluetooth energy meter,” Information,
Maximizing Battery Life in Photovoltaic Systems : A Review Commun. Signal Process. 2003 Fourth Pacific Rim Conf.
of Lessons Learned Authors,” Sol. Energy Power to Choose, Multimedia. Proc. 2003 Jt. Conf. Fourth Int. Conf. 15-18
no. 321, pp. 1–7, 2001. Dec. 2003, vol. 3, pp. 1474–1477 vol.3.
[6] S. R. Ayyubi, “Grid Planning : Agent Based Approach for [13] D. Niyigena, C. Habineza, and T. S. Ustun, “Smart Energy
Early Notification of Air Conditioning Loads to Smart Grid,” Management for PV Energized Health Centers in
IEEE 10th Conf. Ind. Electron. Appl. (ICIEA),2015, Underserved Communities,” 4th Int. Conf. Micro Gener.
Auckland, New Zeal., pp. 1772–1777. Relat. Technol. (Microgen 4), 28-30 Oct. 2015, Tokyo, Japan.
[7] M. N. Ullah, N. Javaid, I. Khan, A. Mahmood, and M. U. [14] D. Niyigena, C. Habineza, and T. S. Ustun, “Computer-based
Farooq, “Residential Energy Consumption Controlling Smart Energy Management System for Rural Health
Techniques to Enable Autonomous Demand Side Centers,” IEEE Int. Renew. Sustain. Energy Conf. (IRSEC),
Management in Future Smart Grid Communications,” 2013 2015, Marrakesh, Morocco.
Eighth Int. Conf. Broadband Wirel. Comput. Commun. Appl.,
pp. 545–550, 2013.

Authorized licensed use limited to: SRM Institute of Science and Technology. Downloaded on August 28,2024 at 04:50:26 UTC from IEEE Xplore. Restrictions apply.