Optimizing Electricity Load and Cost for Demand Side Management in Smart Grid

This paper proposes a mechanism for OELC (Optimizing Electricity Load and Cost) for smart grid. The load of every smart home is predicted one-hour prior to their actual usage. To fulfill PL (Predicted Load) of each consumer, multiple resources of electricity are considered, including RE (Renewable Energy) resources. Furthermore, cost to get PL from multiple resources is calculated. In proposed model 3-4 smart homes are grouped in the form of clusters. To reduce the amount of electricity bills, system also allows privileges to share electricity between adjacent smart homes within a cluster. To validate the OELC mechanism, extensive numerical simulations are conducted which shows a significant reduction in electricity load and cost for electricity consumers. In future, to enhance the functionality of OELC, security from cyber-attacks can be considered.


INTRODUCTION
almost 10% of overall household income [1]. Studies recommend that demand and the price of electricity will continue to increase in future [6].
The easiest way for consumers to cut their electricity bills is to simply reduce their electricity consumption during peak hours, when electricity rates are high [7].
Unfortunately, this dynamic pricing places a burden on electricity consumers to constantly monitor price signals, and then alters the pattern of their electricity consumption to reduce costs without disturbing their daily life [8][9].
This approach is challenging, since electricity consumers are unaware of power consumption of individual devices Authors E-Mail: (ayesha.afzaal@lcwu.edu.pk, mohsinsage@gmail.com) * Lahore College for Women University, Lahore, Pakistan.
and resist thinking about it [10]. Thus, electricity consumers may not respond appropriately to price changes, and they cannot gain the cost-saving benefits of off-peak hours [11].
Over past few years, significant research was conducted on demand side management, which focuses on reducing costs of electricity to end consumers. In particular Hemant proposed an algorithm based on rebound peak condition.
A threshold is defined for grid station. Servers are continuously monitored grid as its capacity reaches allowable limit, it stops remaining electrical appliances [12]. The power will be provided to remaining appliances in the next time slot of off-peak hours. In this way grid remains stable and the cost of electricity for end consumer is reduced [3]. Spiliotis et. al. [13] proposed an AP (Action  [18][19].
Literature shows that lots of methods have already been proposed to reduce cost of electricity and peak to average ratio but the existing methods require user personal information, authority to shift users load or disturb their comfort level [20][21][22][23]. Moreover, most of the other cost reduction techniques for smart grid focuses on appliance scheduling, which will ultimately affect the comfort level of electricity consumers [24][25]. There is a need of a mechanism, which will reduce cost of electricity bills for consumers while maintaining their conform level along, keeping users personal information and authorities secure [26][27][28].
In this paper OELC is proposed. OELC will encourage electricity consumers to switch between small-sized RE resources and electricity supplier companies, to satisfy their demand. In OELC, every smart home is equipped with RE resources behaves like mini micro-grid. Smart home users can get electricity not only from their own mini micro-grid but also from the neighboring smart homes mini micro-grids.
In this way each smart home user is buyer as well as seller of electricity. The system will attempt to fulfill upcoming demand of electricity for each smart home from their own RE resources. If demand is not completely satisfied, system will consider interested neighboring smart homes for trading of electricity (sell extra units of electricity). At the end, electricity supplier companies will be considered for the remaining demand. In this way, all the demand is not transferred to the traditional power grid. The overall goal of this research is to reduce load and cost of electricity for end consumers. Moreover, implementing security features in smart grid is out of the scope of this paper as already a lot of work is done in this domain. In future, smart meter and electricity consumption profile of each user can be secured using ARMET [29] technology.
The rest of the paper is organized as follows. Section II gives details of OELC mechanism as well as some infrastructure changes required to implement OELC. The functioning of OELC is analyzed under various settings using a case study in section III. Finally, the conclusion and future recommendations are drawn in section IV.

MATERIALS AND METHOD
In this paper a novel mechanism for OELC is proposed which will not only overcome the load on the grid by considering multiple resources of electricity, but also reduce the amount of electricity bills for end consumers by performing electricity trading between adjacent smart homes.

Infrastructure Requirement for Proposed System
A small infrastructure changes are required to implement OELC-I sent the PL t i of each user i for time slot t to OELCII, which will calculate the cost to satisfy PL from multiple resources of electricity. It is considered that there are three main resources of electricity for every smart home user.


In-house mini Micro-Grid  Neighboring smart homes' mini micro-grid  Electric grid

Details of OELC
OELC is explained in six steps given below: Step-1: The upcoming electricity demand of consumers is calculated by using our previously proposed agentbased weighted average prediction algorithm [30]. System consists of two types of agents; the forecaster agent and a group of expert agents having some associated weights.
Weights are assigned to expert agents according to the Where f j,t is the advice of j th expert agent, belongs to the set of expert agents' advices (f 1,t ; ...,f n,t ) and n is total number of expert agents and w (j,t-1) is the weight assigned to j th expert agent in the previous time slot.
Weights of expert agents are revised after every time slot depending on the accuracy of their prediction.
Moreover, when time slot t occurs; actual load L t i of electricity consumer i for time slot t is compared with predicted loads PL t i . The objective is to minimize the difference between these two load profiles. To re-assign the weights of expert agents, actual load L t i is also compared with each expert agent' advise. Expert agents' weights are increased whose advice was closer to actual load. Agents weight reassignment is also explained in Algorithm-1.
Step-2: OELC-I will compute cost to fulfill from in-house mini micro-grid. It is calculated using Equation (2): Where C hmg is the cost to get PL from in-house mini microgrid, P hmg is price for bidding and H hbm is hourly payback amount of investment. P hmg is the sum of multiple factors as shown in Equation (3).

Optimizing Electricity Load and Cost for Demand Side Management in Smart Grid
Where, C k is the production cost of electricity and k is the number of distributed RE resources in each smart home and x is the available units of electricity in RE. M j is price of energy if stored units of RE are received from gird and j is rate at peak and off-peak hour. S c is per unit storage cost of electricity.
Step-3: Smart homes are gathered to make clusters. Every smart home has an adjacent neighbor. The cost to get electricity from adjacent neighbors is stored in the form of the matrix given below.
C nmg is the cost to get PL from neighboring micro-grid. P is pricing factor, q nmg represents the quantity of electricity transferred from neighboring micro-grid, l loss is the energy losses during transformation of electricity from seller micro-grid (s m ) to buyer micro-grid (b m ). q g is quantity of electricity that neighboring micro-grid get from grid and p g is per unit cost to get electricity from grid. q d is the quantity of electricity discharged and p d is discharging price. l represents the load profile and s is storage profile.
Similarly, s a+ is charging profile and s a-is discharging profile. Step-4: A centralized control OELC -II will control all the OELC-I in each smart home. If PL is not fully satisfied, OELC-I sent request to OELC-II to fulfill remaining electricity demand from the electric grid. OELC-II will calculate the cost to get the remaining PL from the grid using Equation (6).
Where C g is the cost to fulfill remaining PL (r t ) from the electric grid and u t is per unit cost according to time slot t.
Step-5: OELC-II will compare all three costs calculated above. It will select the cheapest one to fulfill the PL but

ALGORITHM-1: WEIGHTED AVERAGE PREDICTION ALGORITHM
1: Input Parameter: Set of expert weights W, w (j,t) weight of j th expert agent, f j,t advice of j th expert agent, actual Load L t i of user i for time t, Threshold T is the allowable difference between actual and predicted load.

11: end while
if PL is not fully satisfied from one resource then move towards the second cheapest resource similarly multiple resources of electricity will be selected starting from cheapest to highest until the demand is not completely satisfied.
Step-6: At the end the electricity bill of each user is calculated. This electricity bill is based on the sum of their units consumed from each resource. It is calculated using Algorithm-2.

RESULTS AND DISCUSSION
This research study was conducted to reduce cost of electricity for end users. The simulation results and performance of OELC mechanism was assessed in this section. The initial data used for simulation was taken from a case study [31]. profile of four different smart homes users according to time slot is portrayed in Fig. 3. As explained in Fig. 3, it is assumed that each user electricity demand is slightly different from each other and the demand of electricity is ALGORITHM-2: UTILITY BILL CALCULATI ON FOR SINGLE USER i 1: Parameters: C hmg is per unit cost to get electricity form home micro-grid, UC hmg is the units available in-house micro-grid, C nmg is per unit cost to get electricity from neighboring micro-grid, UC nmg is the units available neighboring smart homes, C g is the per unit cost to get electricity from grid, TC 1 , TC 2, TC 3 is the cost to get electricity from resource 1, 2 and 3 respectively. Time slot t, user i. and also at beginning hours of the night (7PM-12AM).
According to the proposed OELC mechanism, first the PL of each smart home is fulfilled from in-house mini microgrid. Fig. 4 shows the availability of stored electricity units in 4 different smart homes' mini micro-grids used during simulation. Furthermore, the cost to get PL from each smart home, mini micro-grid C hmg is calculated based on Equation (2).
Moreover, according to OELC, each smart home can buy electricity from their adjacent neighbors within a cluster.
Cost to get electricity from adjacent smart home neighbors C nmg is stored in matrix form. It is calculated based on Equation (4) and updated at the beginning of every hour. In our case study the matrix shown above is used for simulation. Where rows indicate the buyers Furthermore, the cost to get (C g ) PL from the grid (electricity supply company) is also calculated using Equation (6). By implementing Algorithm-2, system will calculate the total cost of each smart home user against their PL as shown in Fig. 5. To define a threshold, two different systems are compared. A system without OELC, in which demands of each electrical appliance of every smart home user are assumed to be fulfilled by electricity supply company only (grid station). The second system consists of dynamic pricing load shifting, in which flexible loads shift to non-peak hours. The simulation results of electricity cost of each smart home user using OELC system, the system without OELC and dynamic pricing load shifting system are described in Fig. 5. It is clear from the results that there is a significant difference in cost of electricity in three different systems. Cost of electricity for each smart home user is reduced using OELC as well as PAR is also reduced as shown in Table 1 Fig. 6.

CONCLUSION
In this work, a mechanism OELC is proposed for electricity load and cost optimization. In OELC the demand for electricity is managed during peak hours by considering multiple electricity providers, including inhouse mini micro-grid as well as neighboring smart home's mini micro-grids. To decrease the cost of electricity for end users, bids are collected from multiple resources of electricity against upcoming PL of each consumer. Preference will be given to cheapest resource then move towards higher ones to fulfill the demand.
Through simulation analysis, OELC is compared with two other systems; dynamic pricing load shifting and system without OELC. Simulation results confirm that along with the reduction in PAR, electricity consumers achieve a significant difference in their electricity bills using proposed mechanism. The proposed work can be protracted in multiple ways. The benefits of multiple electricity provider companies can be considered and the security issues against malicious users can be studied. In future, more work can be done to secure smart meter and electricity consumption profile of each consumer. Weight assigned to j th expert agents in previous time slot (t -1).

NOMENCLATURE
x Available units of electricity in RE.