An Energy Efficient Residential Load Management System for Multi-Class Appliances in Smart Homes M. B. Rasheed1, M. Awais1, N. Javaid1,*, Z. Iqbal2 , A. Khurshid2, F. A. Chaudhry3, F. Ilahi2 1 COMSATS
Institute of Information Technology, Islamabad 44000, Pakistan Agriculture University, Rawalpindi 46000, Pakistan 3 COMSATS Institute of Information Technology, Wah Cantt 47040, Pakistan *
[email protected], www.njavaid.com 2 Arid
Abstract—Demand Side Management (DSM) mechanism is used for the implementation of different strategies to encourage residential users to reduce electricity bill as well as energy demand. There is also a close relationship between the consumer and utility for equally benefiting to both in terms of grid stability and bill reduction. Extensive research is undertaken now a days in order to make practical implementation on the possible use of different DSM strategies to regulate the energy demand and carbon emission reduction in the World. The major objective of this work is to study the DSM-based approaches which could be helpful in achieving significant electricity demand reduction at the electricity distribution network which is directly connected to the commercial and residential sector especially. In this work, we use an optimization algorithm to obtain the optimal solution for residential electricity load management in a typical household setting. There are two major tasks of this algorithms; firstly, electricity bill minimization of residential user in time of use pricing models, secondly, peaks reduction of demand curve (peak shaving) which will eventually minimize the investment cost of utility including, peak power plants, and transmission lines. Three types of smart appliances are considered; without delay, delay of one hour, delay of five hours. To validate the effectiveness of the proposed algorithm, mathematical models of appliances based on their length of operation time is developed. Index Terms—Demand Response; Optimization; Integer Programming; Smart Grid; I. I NTRODUCTION Due to continuous and rapid increase in electricity demand especially in residential sector, the electricity transmission and distribution system face some difficulties in conventional grid infrastructure. With the development of intelligence and two way communication technologies in Smart Grid (SG), empowers the conventional grid in decision making and automation. This results the benefits to the users as well as utility in the form of bill and energy demand reduction which eventually stabilize the grid [1]. Demand Response (DR) is one of the key element to make effective implementation of smart grid infrastructure. Changes in electricity consumption pattern from normal usages by the residential users in response to varying electricity price over time (i.e, hourly, day-ahead) or incentive based pricing models to reduce the power consumption is known as DR [2] [3] [4]. Currently, DR programs and techniques are actively studied and implemented under different pricing models such as; Time of Use (TOU), and Real Time
Pricing (RTP) in order to reduce electricity consumption as well as bill. In [2] [5], authors implement DR program to facilitate the residential users to schedule their daily load in order to save electricity bill. A mixed integer non-linear optimization technique under TOU pricing model along with incentive based mechanism is used which encouraged the users to schedule their load accordingly. Incentive based mechanism efficiently utilized by the users to save up-to 25% bill as compared to that of TOU model. Moreover, for grid stability, DR programs along with reward/incentive schemes are also very effective. Green energy and distributed generation such as; photovoltaic and wind energy also play an important role in fulfillment of residential energy management using DR programs. In high peak hours, executive users who can pay an extra bill without considering DR programs, distributed generation can be used as a supplement to make it possible to effectively implement DR programs. In [6], authors implement DR program by incorporating distributed generation as well as green energy to make grid stable and reliable for long life. For elastic power consumption with some flexibility in scheduling, appliance power consumption patterns and interior point method is used for optimization and scheduling. It is clear form results that, elastic power consumption patterns provide extra benefits to the users in terms of bill reduction. One of the key reasons for global warming and climate change is due to electricity generation by means of coal and diesel. To overcome this issue, renewable energy generation along with some conventional generation is an attracting field of research now a days and a lot of efforts have been done in this field by different researchers. However, due to lack of controlled infrastructure and uneven nature of climate especially in north USA, make this technology uncertain for long time usage. Moreover, this is not a feasible and long term solution when combined with DR and Demand Side Management (DSM) programs. Another major reasons for the lack of implementation is its capital cost which might not be affordable especially developing countries. In developed countries, this technology is beneficial and feasible due to government inclement and support. So, one of the alternative solutions for renewable energy along with conventional generation is tidal power generation [7]. Tidal power is differ from
solar wind energy and it is very attractive in a small scale generation and management. There are two major benefits while using tidal energy; 1) less storage infrastructure is required 2) environment friendly. Residential sector is one of the major energy consumption contributor and there needs a powerful energy efficient infrastructure to efficiently utilize the available energy. After successful implementation of DR programs, HEM attract the researchers to investigate and explores these programs. With HEM, authors [8], [9] expected to increase the stability and reliability of the smart grid and deployment of DSM programs [10], [11], [12] will enhance the energy efficient electricity utilization and bill reduction. In this work, we present the HEM algorithms for multi-class appliances using DR and TOU pricing mechanism to reduce the energy consumption and bill reduction. We considered a typical house in which three types of appliances are categorized and used in the simulation. Another good approach of our proposed algorithms is that we do not considered distributed and renewable energy generation and storage. Grid energy is efficiently utilized among all types of appliances.
of time (1 < t ≤ T ) and known in advance to EMC. Unit energy price in each hour ia denoted by pu . Then, total energy consumption cost of all appliances in time T is denoted by;
II. S YSTEM M ODEL
At ∈ [Tf − Ts ] ∀t = [Ts ≤ t ≤ Tf ].
In the proposed mode, we consider three types of N appliances which consume energy in a 24 hours time period. Each device is controlled by Energy Management Controller (EMC) which takes energy price signals directly from utility via smart meter. We divide the total scheduling time period (e,g., day) into 24 qual time slots T having one hour duration. EMC calculates the starting Ts and finishing Tf time intervals as well as energy consumption of each appliance in a given time interval without exceeding power capacity C. The energy consumption during all time intervals is defined as; Ei = [ei t1 , ei t2 , ei t3 , ....nTn ],
(1)
where, ei t1 is the amount of energy consumption of appliance i in time interval t1 and so on. Sum of all time intervals is denoted by T − n. By considering scheduling horizon, each smart appliance n has Ts , Tf , and Length of Operation Time (LOT). Each appliance n can have scheduling time horizon between [Ts , Tf ]. However, scheduling range of each appliance also depends on its LOT. We can write also; t > 0, and t < Tn ,
E(n) =
(
0 0
(2)
if t ≥ Tf if t ≥ Ts
∀t = [t1 , t2 , t3 , ...Tn ],
T X N X
Et pt .
(3)
Time of Use (TOU) pricing scheme is used in our model where, energy consumption prices vary in each sub-interval
(4)
t=1 n=1
III. T YPES OF A PPLIANCES We consider a home where N number of smart appliances are used having different LOT requirements. In order to better energy management and scheduling, we divide these appliance into four classes based on flexibility of operation time and constraints.
A. Non-Shiftable Appliances Class-1 Non-shiftable appliances A, TV, fridge, and heater that have strict starting and ending time limits in a given LOT. Due to non-flexible nature, we cannot shift these appliances in any particular time slot. One such appliances run, they must complete their operation time without taking into account the price and energy limit. (5)
Objective function for class-A appliances is given as; Obj = min
T X A X
EtA pt .
(6)
t=1 n=1
s.t:
t = [Ts ≤ t ≤ Tf ],
(6a)
EtA ≤ Enmax
(6b)
C ≤
∀, A = [A1 , A2 , ..An ], EtA
≤
Enmin .
(6c)
Where, electricity unit price for appliances A is same in a given scheduling interval. Constraint 6b shows that the total energy consumption of appliances A cannot exceed from the available given capacity C limit.
B. Shiftable Appliances Class-2 Shiftable group of appliances B such as dish-washer, washing machine, electric cooker, and iron are those that can be shifted within specified time limit if the available energy limit in a particulate time slot is less than the required energy. In this category, we assigned one hour time flexibility to each appliance. In other words, appliances can be scheduled within maximum and minimum Ts and Tf time limits respectively. Bt ∈ [Tf − Ts + 1] ∀ t = [Ts ≤ t ≤ Tf + 1].
We denote the scheduling time interval Tsch of appliance n and each appliance has its maximum Enmax and Enmin energy consumption limit respectively in all time intervals. Enmin ≤ Ent ≤ Enmax Ts ≤ t ≤ Tf .
Etn =
(7)
Where, Nt denotes the scheduling interval of a given appliance in time horizon T . Then, B th appliances have 24 shiftable pattern which can be randomly selected based on available energy limit of in a particular time slot. It is also clear from above equation that appliance in this category are one hour flexible because they can be schedule LOT + 1 hour with respect to given operation time. In order to select any particular pattern for optimization, [13] uses Binary Linear Programming
(BLP) technique. Similarly, objective function for class-B appliances is given as; Obj = min
T X B X
EtB pt .
1.9
×10 -5
1.8
(8)
t=1 n=1 1.7
t = [Ts ≤ t ≤ Tf + 1],
(8a)
EtB ≤ C Enmax ≤
(8b)
∀, B = [B1 , B2 , ..Bn ], EtA ≤ Enmin .
Price ($)
s.t:
(8c)
1.5
Here, given electricity unit price for appliances B is same in a given sub-interval. Constraint 9b denotes that the total energy consumption of appliances in class-A cannot exceed from the available given capacity C limit.
1.4
1.3 0
C. Shiftable Appliances Class-3
Ct ∈ [Tf − Ts + 3] ∀, t = [Ts ≤ t ≤ Tf + 3].
(9)
Objective function for class-C appliances is given as; Obj = min
EtC pt .
(10)
t=1 n=1
s.t:
5
10
15
20
25
Time of day (hour)
Appliances in this category C are more flexible as they can put into wait for more time as compared to class-1 and class2 respectively. Device such as coffee-maker, toaster, washing machine, etc, are kept in this category because EMC can schedule these appliance when you are out of home and there is no need to turn-on immediately. In this category, user can place any device without disturbing the user comfort. Moreover, it is more desirable that appliances finish their job when there is off-peak hours in order to save electricity bill.
T X C X
1.6
t = [Ts ≤ t ≤ Tf + 3],
(10a)
EtC ≤ C Enmax ≤
(10b)
∀, C = [C1 , C2 , ..Cn ], EtA ≤ Enmin .
(10c)
It is clear form constraint 12b that the total energy consumption of appliancesC cannot exceed from the available given capacity C limit. Moreover, constraint 12a shows that appliances have five hours flexibility in given scheduling horizon. IV. L OAD O PTIMIZATION AND S CHEDULING In this section, we discuss the TOU pricing model in which energy signal is only received from utility. The main objective is to design and simulate an optimization model in order to reduce electricity bill of residential users as well as peaks reduction of utility to stabilize the grid. To increase the user comfort, we proposed three types of appliances including non-shiftable, shiftable within one hour, and shiftable with more than one hour time flexibility. Users can select any model according to his comfort and energy requirements. However, during low pricing time intervals, users cannot use more electricity than the given threshold. So, the overall
Fig. 1. TOU price signal.
objective function is to minimize the electricity cost of all types of appliances by taking into account the LOT, scheduling time flexibility as well as Enmax and Enmin energy limits respectively. T X N X Obj = min EtN pt . (11) t=1 n=1
s.t:
t = [Ts ≤ t ≤ Tf ],
(11a)
EtN ≤ C
∀, N = [n1 , n2 , ..Nn ],
(11b)
Enmax ≤ EtN ≤ Enmin , 0 < t ≤ 24.
(11c)
It is clear that in order to solve the objective function (4) to minimize the overall energy consumption, we have to solve the sub-problems given in section (II). It is also noted that the all sub-problems are convex in nature and we can solve it by using any standard optimization technique. In addition, it can also be seen that to reduce the electrically bill, energy consumption should also be minimized by considering and satisfying all the constraints. V. PARAMETER S ETUP To evaluate and validate the proposed energy optimization and scheduling algorithm, performance of the unscheduled appliances is compared with that of scheduled case. We considered total N number of household appliances having initial energy En which is randomly distributed among all appliances. In unscheduled case, once an appliance is on, it will always remains on until it completes its LOT without taking into account the maximum and minimum energy threshold limits and time flexibility. All simulation parameters used in this work are randomly generated. 1) Starting time of appliances (Ts ): Ts of all appliances in each class is randomly selected between sub-interval time [1, 2, 3, ......T ].
4
×10 4
4
×10 4 Unscheduled Scheduled
3.5
3
3
Power Consumption (Watt)
Power Consumption (Watt)
Unscheduled
3.5
2.5 2 1.5 1 0.5
2.5 2 1.5 1 0.5
0
0 0
5
10
15
20
25
0
5
Time of day (hour)
(a) Unscheduled power consumption
4 Unscheduled Scheduled
20
25
×10 4 Scheduled Unscheduled Price
Power Consumption (Watt)
3.5
0.5
Price ($)
15
(b) Unscheduled and scheduled power consumption
0.7
0.6
10
Time of day (hour)
0.4
0.3
0.2
0.1
3 2.5 2 1.5 1 0.5
0
0 0
5
10
15
20
25
Time of day (hour)
(c) Unscheduled and scheduled price consumption
0
5
10
15
20
25
Time of day (hour)
(d) Unscheduled and scheduled power consumption with price
Fig. 2. Power and energy consumption between unscheduled and scheduled case
2) Finishing time of appliances (Tf ): Tf is also randomly distributed among sub-interval time of all classes [Tf − Ts + 1], [Tf − Ts + 1], [Tf − Ts + 3] respectively. 3) Length of operation time (LOT ): LOT time of all classes of appliance is selected with some scheduling time flexibility. 4) Power consumption (EtT ): initially, (EtT ) is uniformly distributed among all appliances. 5) Available capacity (C): available capacity limit obtained from the utility in the form of T OU signal.
VI. S IMULATION R ESULTS In this section, we discuss simulation results of the proposed algorithms for HEM. For this purpose, we considered total N = 30 smart appliances in three different cases and the initial power of all appliances is P = 1500W atts. Ts and Tf time is randomly chosen between T = [1, 2, 3, ....N = 24] hours while LOT of each class of appliances if different. For class-1, class-2, class-3, LOT are Ts − Tf , Ts − Tf + 1, Ts − Tf + 3 respectively with some time scheduling flexibility in class-2 and 3. T OU pricing signal is used which is directly obtained from N Y ISO website in which price vary in each hour according to demand requirements from user side fig. (1).
Firstly, we conduct simulations according to T OU pricing for unscheduled case in which we use N appliances of three classes . For this case, we set Ts , Tf , LOT , and Et parameters as discussed in section (parameter discussion) are used. Total unscheduled power consumption for N appliances with random initial values is shown in fig. 2(a). It is clear from figure 1 that, from 10am to 9pm the electricity price is high due to more power usage probability during there hours. Similarly in fig. 2 with initial random values, power consumption rate is also high which seems to be logical without any scheduling policy implementation. To cope for such types of problems, engineers and scientists have been working for many years to invent and implement distributed energy infrastructure. Until now, extensive work has been implemented in smart grid field and results show their effectiveness in the form of better energy utilization, carbon emission reduction for green environment, etc. In our work, to reduce the energy consumption which ultimately reduces the electricity bill, we implement scheduling algorithm considering three type of alliances. For this purpose, we adjust the LOT of each class of appliances so that, users can prioritize their appliances according to the price available energy limit. Fig. 2(b, c) shows the difference between the power consumption and electricity bill in scheduled and unscheduled case respectively. It is clear that our proposed model and scheduling algorithms optimally adjust the working of home appliances. It is also understood that performance of each appliance is also depends on the performance in each sub-interval time. Fig. 2 also show that the algorithms adjust the working of all appliances form on peak hours to off peak hours when electricity bill is low. As in unscheduled case where each appliance is randomly initialized and completes its working without considering prices. Lastly, comparison of power consumption between unscheduled and our proposed case as well as price variations over time is shown in fig. 2(d). It is clear from the figure when we compared unscheduled case with the proposed algorithm which consumes more power when price at time (4am−7am) is low and less power when price at time (10am − 18pm) is relatively high. It is worth mentioning here that the overall power consumption in scheduling case is quite optimal and relatively low. VII. C ONCLUSION
AND
F UTURE W ORK
In this work, we studied a residential electricity load management problem for different class of appliances in T OU pricing environment. Different class of appliances are characterized based on their LOT and energy consumption patterns. Firstly, we proposed and developed mathematical models for all classes and design an optimization algorithm in order to reduce the overall electricity bill as well as peaks reduction during low price hours. Moreover, to facilitate the user in terms of comfort, time scheduling flexibility is introduces in each class so that users can adopt any models based on their requirements. To sum-up, our proposed algorithm optimally scheduled the smart appliances based on energy consumption patterns, T OU pricing signals, and LOT which effectively
reduced the overall power consumption and electricity bill. We believe that, our proposed algorithm with multi-class appliances would be very helpful when practically implemented. Moreover, in future work, we aim to enhance this model for more classes with variable scheduling flexibility, Real Time Pricing (RT P ) models, user’s presence at home, etc. R EFERENCES [1] T. J. Lui, W. Stirling, and H. O. Marcy, “Get smart,” Power and Energy Magazine, IEEE, vol. 8, no. 3, pp. 66–78, 2010. [2] K. Hamilton and N. Gulhar, “Taking demand response to the next level,” Power and Energy Magazine, IEEE, vol. 8, no. 3, pp. 60–65, 2010. [3] Z. Zhao, W. C. Lee, Y. Shin, and K.-B. Song, “An optimal power scheduling method for demand response in home energy management system,” Smart Grid, IEEE Transactions on, vol. 4, no. 3, pp. 1391– 1400, 2013. [4] Y. Huaguang, L. Bin, C. Songsong, M. Zhong, L. Dezhi, L. Jiang, and H. Guixiong, “Future evolution of automated demand response system in smart grid for low-carbon economy,” Journal of Modern Power Systems and Clean Energy, vol. 3, no. 1, pp. 72–81, 2015. [5] C. Vivekananthan, Y. Mishra, G. Ledwich, and F. Li, “Demand response for residential appliances via customer reward scheme.,” IEEE Trans. Smart Grid, vol. 5, no. 2, pp. 809–820, 2014. [6] C.-R. Chen and M.-J. Lan, “Optimal demand response of smart home with pv generators,” International Journal of Photoenergy, vol. 2014, 2014. [7] D. Anna, “Community smart grid utilizing dynamic demand response and tidal power for grid stabilization,” Smart Grid and Renewable Energy, vol. 2013, 2013. [8] M. Castillo-Cagigal, A. Guti´errez, F. Monasterio-Huelin, E. Caama˜noMart´ın, D. Masa, and J. Jim´enez-Leube, “A semi-distributed electric demand-side management system with pv generation for selfconsumption enhancement,” Energy Conversion and Management, vol. 52, no. 7, pp. 2659–2666, 2011. [9] M. Chaabene, M. B. Ammar, and A. Elhajjaji, “Fuzzy approach for optimal energy-management of a domestic photovoltaic panel,” Applied energy, vol. 84, no. 10, pp. 992–1001, 2007. [10] P. Palensky and D. Dietrich, “Demand side management: Demand response, intelligent energy systems, and smart loads,” Industrial Informatics, IEEE Transactions on, vol. 7, no. 3, pp. 381–388, 2011. [11] D. Setlhaolo, X. Xia, and J. Zhang, “Optimal scheduling of household appliances for demand response,” Electric Power Systems Research, vol. 116, pp. 24–28, 2014. [12] A. Fakhrazari, H. Vakilzadian, and F. F. Choobineh, “Optimal energy scheduling for a smart entity,” 2014. [13] Z. Zhu, J. Tang, S. Lambotharan, W. H. Chin, and Z. Fan, “An integer linear programming based optimization for home demand-side management in smart grid,” in Innovative Smart Grid Technologies (ISGT), 2012 IEEE PES, pp. 1–5, IEEE, 2012.
