Research Article  Open Access
Pawan Singh, Baseem Khan, "Smart Microgrid Energy Management Using a Novel Artificial Shark Optimization", Complexity, vol. 2017, Article ID 2158926, 22 pages, 2017. https://doi.org/10.1155/2017/2158926
Smart Microgrid Energy Management Using a Novel Artificial Shark Optimization
Abstract
At present, renewable energy sources (RESs) integration using microgrid (MG) technology is of great importance for demand side management. Optimization of MG provides enhanced generation from RES at minimum operation cost. The microgrid optimization problem involves a large number of variables and constraints; therefore, it is complex in nature and various existing algorithms are unable to handle them efficiently. This paper proposed an artificial shark optimization (ASO) method to remove the limitation of existing algorithms for solving the economical operation problem of MG. The ASO algorithm is motivated by the sound sensing capability of sharks, which they use for hunting. Further, the intermittent nature of renewable energy sources is managed by utilizing battery energy storage (BES). BES has several benefits. However, all these benefits are limited to a certain fixed area due to the stationary nature of the BES system. The latest technologies, such as electric vehicle technologies (EVTs), provide all benefits of BES along with mobility to support the variable system demands. Therefore, in this work, EVTs incorporated grid connected smart microgrid (SMG) system is introduced. Additionally, a comparative study is provided, which shows that the ASO performs relatively better than the existing techniques.
1. Introduction
As the world is transforming from the conventional grid system to the smart grid system, renewable energy sources’ integration has become a vital issue in the current situation. The intermittency and climate dependency of RESs make their integration more complex and difficult. A microgrid (MG) offers an efficient way to incorporate distributed RESs in large electric power systems for supplying the continually growing demand. The smart microgrid (SMG) consist of RESs (especially wind turbine (WT) and solar photovoltaic (SPV)), BESs, electric vehicle technologies (EVTs), and electrical demands. BESs and EVTs have a bidirectional battery charging system as well as automatic sensors for detecting over or undergeneration. SMG coalesced with RESs, BESs, and EVTs is a preferable alternative to manage the increased power demand and is a supplement to the centralized smart power grids [1]. Recently, there are rising issues and concerns regarding the instability and discontinuity of RESs in the MG. Therefore, the MG central operator (MGCO) recommends the incorporation of BES in the MG for accumulating surplus power and feedback to the MG during the peak load. Further, the latest EVTs (battery electric vehicle (BEV) and plugin hybrid vehicle (PHEV)) along with BES play a vital role to store excess power during high availability. The advantages of EVTs are their mobility and ability to supply the stored power in the energy deficient areas during peak hours. Therefore, the computation of the suitable capacity of BES, BEV, and PHEV is highly essential for an economized operation of SMG. Currently, the attention is shifted towards more efficient fuel cell technologies (FCTs), such as automotive fuel cell electric vehicle (FCEV) and stationary FC power generator (FCPG) [2, 3], due to their numerous advantages (less CO2 emissions, extremely less noise and vibrations). FCEV and FCPG have a lower maintenance cost, as they consist of fewer rotating parts. Additionally, FCTs do not require recharging, unlike BESs. FCTs are expandable and can be grouped for the required power rating [4]. FCTs decrease manufacturing, shipping, and security concerns related to MG [5].
The economical operation of SMG is one of the most significant optimization problems for MGCO, wherein the economical power output of BES, FCPG, microturbine (MT), EVTs, and RESs is computed by fulfilling all equality and inequality constraints. Various studies are performed for the economical operation of MG. Mitra [6] proposed a method to estimate the capacity of BES for removing the intermittency problem of DGs. SA is applied by Ekren and Ekren Banu [7] to optimize the capacity of DGs and minimize the total cost. Mohammadi et al. [8] applied a variation of GA to find an optimal power output as well as the cost of MG under the deregulated electricity market. BahmaniFirouzi and AzizipanahAbarghooee [9] developed an IBA for the economical operation of MG.
Various studies focus on the economical operation problem of MG without taking into account the effect of optimum sizing of BES. Chakraborty et al. [10] applied linear programming for the economical operation of MG and improved the BES’s charge states. Sortomme and ElSharkawi utilized PSO for the economical operation of MG [11]. Niknam et al. [12] utilized the honeybee mating optimization for the economical operation of MG, which includes PV, WT, and FC. Sharma et al. [13] presented a comparative study of metaheuristic techniques for microgrid optimization.
Nowadays, EVTs are the integral part of MG technology. Various optimization studies are performed on EVTs incorporated MG systems. Laureri et al. [14] focused on EVTs and their active participation in grid optimization by adopting power modulation under V2G mode. Bai et al. [15] designed a twoway charging system for bidirectional flow of current to convert it as a mobile energy storage system. Chen et al. [16] proposed a microgrid layout of EVT charging station, which combines with renewable energy sources and battery storage. Gunter et al. [17] presented a framework and an optimization technique to design utility coupled MG with BES, distributed generation, and PHEV chargers.
Conventional optimization techniques have a number of limitations together with continuity and derivability of the objective function. Existing methods which are based on metaheuristic search techniques [19, 20] can be regarded as appropriate alternatives for handling optimization problems since these metaheuristic methods provide the best global solution, handle large constraints, and are derivativefree. These methods have some shortcomings like getting stuck in local optima and take a long time to find global optima; consequently, selection of the appropriate evolutionary algorithm has great importance. In previous algorithms, the main limitation is to handle problems with a large number of variables and various constraints.
This paper proposed an artificial shark optimization (ASO) method to remove the limitation of existing algorithms for solving the economical operation problem of MG. The ASO algorithm is motivated by the sound sensing capability of sharks which they use for hunting. ASO is capable of providing extremely reasonable results of various standard functions in comparison with other renowned metaheuristic techniques. The global and local searching capabilities of ASO algorithm are far better than the former optimization method. These capabilities of ASO have encouraged the authors to utilize this algorithm to find the economical operation of SMG (EOSMG). The result obtained from ASO is compared with the results of GA, SA, and PSO to show its feasibility. The key contributions of this work are summarized as follows: () developing a novel ASO technique for solving EOSMG, () introducing V2G, G2V, and V2H technology based on SMG system, and () comparative analysis of developed ASO and existing methods on EOSMG and standard benchmark functions.
Section 2 describes the problem formulation of EOSMG. In Section 3, a novel optimization method called artificial shark optimization is proposed. Section 4 provides the results and discussion followed by the conclusion.
2. Problem Formulation
The required objective function and constraints for the formulation of an economical operation problem of SMG (EOSMG) are as follows.
Minimize total costs:where The total cost of the SMG is the summation of the following costs:
(A) Operation cost of utility, BES, BEV, PHEV, and FCEV
(B) Fuel, operation, and maintenance cost of DGs
(C) Startup costs of FC, FCEV, and MT
(D) The cumulative total cost per day of batteries used in BES, BEV, and PHEV
The overall cost of batteries depends on the sum of fixed cost (FX) and maintenance cost (MC), where FX and MC are onetime purchasing cost and a variable annual maintenance cost, respectively. Therefore, the total cost, proportional to the size of batteries, is , where is the maximum capacity of a battery. The study is conducted for a day and the cost of operation is computed on an hourly basis (24 values); therefore, is modeled in €ct/day. Let and be the rate of interest for funding the installed battery and its lifetime, respectively; then, the in €ct/day has been calculated as follows [18]:The proposed operation cost minimization problem handles the constraints as presented in Table 1.

3. Artificial Shark Optimization Algorithm
Metaheuristic optimization algorithms are best suited to solve the complex engineering problems as they depend on simple and easily implantable concepts, bypass local optima, and do not need gradient data.
3.1. Motivation
Nature finds solutions in a better manner than humans. Modeling biological, physical, or behavioral phenomena of natural objects provides better capabilities to solve optimization problems. Such optimization methods are categorized as nature inspired metaheuristic algorithms. The great hunting capabilities of sharks have become a motivation to propose this optimization algorithm. Sharks are great swimmers with fantastic sudden and smooth turning capabilities. Sharks swim in the ocean in groups and search for food by identifying the proper location of prey. Sharks have strong sensing potential; they can hear their prey in the water even from approximately 3000 feet away [21]. This makes sharks a deadly hunter that survived for millions of years even before the dinosaurs.
The sound of the prey helps sharks to locate the appropriate position of their prey. The shark receives a fraction of these sounds due to different noises, available between the shark and the prey. The movement of the shark depends on the sounds of nearby () or faraway () prey. The motion of a shark is random until the probability of finding prey is less than or equal to the threshold value (). If the probability goes more than , the shark starts moving in a shrinking spiral motion to grab the prey.
3.2. Mathematical Model
In the initial phase, the shark searches for potential prey in a random motion. If the probability of getting the prey is high, then it circulates in a shrinking spiral motion to grab the prey.
3.2.1. Probability
The shark is a hungry animal; it searches for prey until potential prey is not located. The probability of finding prey is an increasing real number and it ranges from 0 to 1. The threshold value of probability to find prey is . If the shark estimates the probability to be more than the threshold value, it starts moving in a shrinking spiral motion to grab the prey.
3.2.2. Random Motion
The minimum and maximum audible sounds from nearby (faraway) prey are and ( and ), respectively. A function returns a random value in the range . Figure 1 shows the random motion of a shark.
The sound of nearby prey for iteration iter is formulated as The sound of faraway prey for iteration iter is formulated asThe speed that defines the amount of change applied to the shark is modeled asThe position of a shark is evaluated as
3.2.3. Shrinking Spiral Motion
Let the position of the prey (the best solution achieved so far) be and the previous location (position) of the shark be ; then, the distance between them is modeled asThen, the position of a shark is evaluated aswhere , , and is a function that returns a random value in the range . Figure 2 presents the spiral motion of the shark.
3.3. Pseudocode of the Artificial Shark Optimization Algorithm
See Algorithm 1.

3.4. Performance Assessment
To observe the efficiency of the proposed ASO technique, fifteen renowned unimodal, multimodal, and composite test functions are utilized. The parametric values of () and () are assumed to be 2 and 0 for this study, respectively. The formulation of these test functions is described in Tables 2(a), 2(b), and 2(c) [22].
(a) Unimodal benchmark functions  
 
(b) Multimodal benchmark functions  
 
Dim in indicates the count of variables.  
(c) Composite benchmark functions  
