A Brief Review of Microgrid Surveys, by Focusing on Energy Management System

: Microgrids are new concepts in power systems that can upgrade current power systems due to their technical, economic, and environmental advantages. In addition, the increasing penetration of renewable energies and their use in microgrids have increased the complexity of these new grids in terms of planning and operation. Along with numerous research and practical projects built in different countries with multiple applications, countless types of research have also been performed relying on different aspects of MGs. In this paper, based on a review of studies and review articles related to MGs, an attempt has been made to evaluate and report the optimal energy management of MGs, based on what is addressed in the literature. In addition, the most critical surveys on various topics of MGs are introduced as a guide for researchers to draw a road map for future works.


Introduction
A microgrid (MG), as the basic structure of the smart grid (SG) concept, can be defined as a local electrical grid, mainly in the low-voltage distribution system, containing renewable and non-renewable energy sources, controllable (dispatch-able) loads, energy storage systems (electrical or thermal), electric vehicles, combined heat and power (CHP) units, control and communication systems, and different strategies such as demand response programs, which can operate in grid-connected or islanded modes [1][2][3].
The microgrid has many advantages for both the consumer and the power generation companies.From the consumer's point of view, it can simultaneously provide electricity and heat, increase reliability and resiliency, reduce greenhouse gas emissions, and improve quality [4].In addition, from the point of view of the power companies, using microgrids can manage the demand and, therefore, can postpone the need for new facilities for power system expansion.
Furthermore, with the electricity growth and the need for higher power quality, the electricity industry has moved toward using new technologies.On the other hand, privatization, competition, and restructuring in the electricity market have forced planners and operators to small-scale generation units, increasing energy efficiency and minimizing total investment and operation costs.For this purpose, one of the efficient solutions is using renewable-energy-based distributed generation sources in new formworks, such as microgrids [5].This also helps in reducing fossil fuels and greenhouse emissions.Furthermore, the power losses will be crucially decreased, creating more flexibility and providing various services to consumers [6].
From the beginning of the introduction of these emergent grids, many types of research and studies have been performed to explain their basics, advantages, and disadvantages, and to examine the challenges ahead in different tasks.Various fields of planning, operation, control, and protection of the MGs are prominent topics that are focused on by scientists, planners, and operators.

Energy Management and Review of Surveys
The energy management system (EMS) is a complex system that prepares the necessary actions to minimize/maximize predefined objective functions in an MG, considering the relevant constraints and limitations.
In addition, the leading technologies of non-renewable DGs are categorized as follows: • Reciprocating engines [19];
As an important note, all of the renewable DGs are from AC type, except the fuel cell, solar, or photovoltaic systems [12,19].
In another classification, Muqeet et al. [15] presented the MG components based on what is described in Table 2. Battula et al. [12] also divided the supply and demand management strategies into two categories of demand control (price-based and direct load) and energy generation control (with and without RES).
Table 3 compares various costs of energy storage systems [21,22] and their environmental impacts [23].Table 4 also describes some technical characteristics of different ESS technologies [23].It should be noted that these data were recompiled using reported data from various references.

The Advantages of Microgrids
The MG brings several advantages to the power industry.From a grid point of view, the main advantage of an MG is that it is treated as a controlled entity within the power grid, which can be operated as a single aggregated load [24].Based on the literature, the following subjects can be addressed in this regard [24][25][26][27]:

•
Operation or investment issues: reducing both electrical and physical distance between generating units and consumers may contribute to: -Improving the reactive power support of the whole system, thus enhancing the voltage profile; -Operating in both modes of connected to the main grid or islanded, which increases the supply reliability of consumers; -Separating and isolating itself from the utility, during a grid disturbance, which helps in continuous operation of MG; -Reduction in distribution and transmission feeder overloading; -Reducing the power losses in distribution and transmission sub-system; -Reducing/postponing the investment expansion for large-scale generation and transmission systems; -Cost saving: utilization of waste heat in CHP units, which increases the energy efficiency and finally leads to a decrease in overall costs; -Integrating RES low-voltage distribution grids; -Increasing power system reliability by reducing customer outage and service restoration time.• Environmental issues: -Physical proximity between DGs and consumers may increase consumer information on more rational use of energy; -Reduce the GHG emissions.

EM Definition
The energy management system (EMS) is an information and control system that provides the necessary functionality, ensuring that both the generation and distribution systems supply energy at minimized operational costs [28].It is an important procedure to achieve a stable and economic operation of MGs through some optimization techniques that manage and coordinate the dispatchable distributed generators (DGs), energy storage systems, demand responses, and other applied strategies to optimize the objective function (s) of an MG [8,[29][30][31].It is also responsible for maximizing the utilization of the RES, considering their relevant uncertainties [8,32].
An important point to note is that despite the many definitions of this concept, three sub-systems of power, control, and telecommunication play a fundamental role.The subsystems' full coordination and integration are the basis for the development from MGs to smart grids (SGs) and, as a result, smart cities (SCs) and smart societies (SSs) [33].
The operation modes of MGs are addressed as grid-connected and islanded ones [13].
Furthermore, different EMS schemes or structures are mentioned as: The computational stress, privacy, scalability, resilience, and communication infrastructure are important issues for EMS that should be discussed [8].In addition, the functionality of EMS in different views of information modules, scheduling and control, and resilience operation should be addressed [8].
In addition, it is required to compare the different types of EMS control strategies in terms of information access, communication information, function in real-time, the feature of plug and play, expenditure, the structure of the grid, size (number of nodes), tolerance during the fault, infrastructure final nodes, operation flexibility, bandwidth and latencies, quality of service (QoS), connectivity, safety measures, and individuality [12].
The control configurations for ESSs are aggregated, distributed, and hybrid [11].
Muqeet et al. [15], by focusing on the market price fluctuations, limited photovoltaic generations, and controlling different loads in managing uncertainties in a campus MG, detailed some campus MG projects in Al-Akhawayn University, Morocco; Aligarh Muslim University, India; American University of Beirut (AUB), Lebanon; Chalmers University of Technology, Sweden; Clemson University, South Carolina; De Vega Zana, Spain; Eindhoven University of Technology, The Netherlands; Federal University of Pará, Brazil; Federal University of Rio de Janeiro, Brazil; Griffith University, Australia; Illinois Institute of Technology, USA; Nanyang Technological University (NTU), Singapore; North China Electric-Power University, Beijing, China; Science and Technology, Algeria; Seoul University, South Korea; Tezpur University, India; University of Central Missouri, USA; University of Connecticut, Mansfield, Connecticut, USA; University of Cyprus (UCY); University of Genova, Savona Campus, Italy; University of Malta; University of Novi Sad, Serbia; University of Southern California, USA; University of Wisconsin-Madison, USA; Valahia University of Targoviste, Romania; Yuan Ze University, Taiwan.All of the mentioned projects were compared in terms of their loads (campus or building) and their components (PV, BESS, Wind, Biomass, DG, MT, EV, SC, FC, and CHP).
Table 5 shows some details for sample campus MGs with their EMS [15].Some examples of the MG implementations or active experiments were also addressed in [38] for the European Union (EU), Japan, Korea, North America, and Australia.

Objective Functions and Constraints
Muqeet et al. [16] presented the MG EMS objective functions including costs of energy, net present, emission, reliability, operation, investment, controlling the MG frequency, start-up, shut-down, network security, reserves, and demand response, some of which, in each reported reference, may be considered.
Zou et al. [10] described the optimization objectives of EM in IMMGs including operating cost, customer satisfaction, usage of renewable resources, transmission loss, system flexibility, stability, and environmental benefits, which will be studied in off-line and real-time timescales.
Pourbehzadi et al. [13] divided the objectives into environmental, economic, and technical issues.Furthermore, the primary constraints are load balance, real power, and battery energy.The cost model includes the grid, wind power, PV, battery, FC, and MT.
Khan et al. [17] categorized the OF in EM of SMGs into the environmental cost (carbon emissions and penalties for emissions), capital and operational costs (fuel, fuel cell, capital, maintenance, and electrolysis), energy storage cost (charge/discharge, ultra-capacitor, hydrogen storage, hourly capital and storage, efficiency of charge/discharge, and battery), and miscellaneous (frustration costs, dissatisfaction costs, penalty costs for solar and wind, tracking error penalty, and load shedding costs), and all of the relevant mathematical formulations were discussed based on different references.In addition, different constraints were categorized into supply, demand, storage, operation, prices, wind, solar, fuel cell, and carbon emissions.

HOMER analysis
The optimal configuration of MG is selected, including solar PV, wind turbine system, and ESS Abdi et al. [1] addressed the objective functions of the OPF including active power generation cost, reactive power generation cost, power supplied to the grid from an external utility, active power losses, carbon emission, load curtailment, tap position and capacitor bank switching, social welfare, reserve cost, and load adjustment.Furthermore, the constraints were divided into constraints of active power (generation or power supply to the load), reactive power, voltage, current (thermal rate or maximum capability), voltage angle, tap position, capacitor bank switching, curtailment, Reserve, Flowing AC power to DC grid, and vice versa.

Optimization Techniques
Khan et al. [17] categorized the optimization methods in EM problems into dynamic programming, mixed integer programming (MIP) (mixed integer non-linear programming, mixed integer quadratic programming, and mixed integer linear programming), stochastic programming, non-linear programming, and integer programming.
Muqeet et al. [15] addressed the optimization techniques for EM: linear optimization, HOMER analysis, genetic algorithm, energy scheduling optimization problem (ESOP), fast Fourier transform (FFT), charging/discharging algorithm, generalized reduced gradient (GRG) algorithm, forecasting method, P2P trading mechanism, NSGA-II (Non-dominated Sorting Genetic Algorithm-II), interval optimization, OPF (optimal power flow) technique auction algorithm CPLEX solver, LabVIEW analysis, and Newton-Raphson technique swarm intelligence approach.Furthermore, the optimization methods are compared based on their advantages and disadvantages, applications, and objectives.Among the optimization algorithms, we can mention the deterministic methods, including MILP, dynamic programming (DP), and MINLP; metaheuristic methods, including PSO, GA, and artificial fish swarm; artificial intelligence methods, including artificial neural network and Fuzzy logic; other methods such as Manta Ray optimization and Harris Hawks optimization.
Aguilar et al. [18] reviewed the relevant recent research to improve EM for smart buildings using artificial intelligence (AI) methods.For this purpose, they first introduced the concept of "Autonomous Cycles of Data Analysis Tasks" (ACODAT), which defines the need for an autonomous management system for some specialized tasks, including monitoring, analysis, and decision-making to reach defined objectives, such as the energy efficiency.They were previously applied in different domains, such as smart classrooms, smart cities, and industry.
In addition, the methodology for the development of a data mining application (MI-DANO) for implementing the ACODAT architectures is presented, which includes three phases: phase 1, which specifies the ACODAT for the problem to be solved; phase 2, which prepares the data for the data analytics tasks (i.e., the extraction and transformation operations of the data); phase 3, which implements all the data analytics tasks of the autonomic cycle.Then, the authors classified the AI techniques in monitoring tasks to K-means, Fuzzy rules or rule-based approaches, ANN, regression algorithms, support vector regression (SVR), density-based spatial clustering of applications with noise (DBSCAN), random forest (RF), evolutionary algorithms, K-nearest neighbor (KNN), and transfer kernel learning.In addition, the main AI techniques used in analysis tasks are classified as: LSTM, auto-encoder, regression neural networks, suggesting convolutional networks (CNN), extreme learning machine (ELM), regression forest, SVR, RF, radial basis functions network, autoregressive integrated moving average (ARIMA), Seasonal autoregressive integrated moving average (SARIMA), and eXtreme gradient boosting (XGBoost).The used AI techniques in control tasks are deep learning (DL), multi-agents, Fuzzy, Fuzzy ANN, and multi-objective.Different techniques in optimization tasks are mentioned as: multi-objective approaches, evolutionary approaches, Fuzzy models, ELM, multi-agent systems, Z-number, reinforcement learning, and bio-inspired approaches.The most important AI techniques in scheduling tasks are bio-inspired approaches, deep reinforcement learning (DRL), and multi-agent systems.
Muqeet et al. [16] focused on the idea of the advanced energy management system (AEMS) to smooth energy flow in a campus MG.For this purpose, the authors introduced some relevant projects and addressed the optimization techniques including high-reliability distribution system (HRDS), control and management system operation, MILP, MICP, PSO, TLBO, multi-agent system (MAS)-based, two-stage stochastic programming, MINLP, and NSGA.
Abdi et al. [1] addressed different mathematical-based methods suggested for the OPF problems: distributed and parallel OPF (DPOPF), multiphase OPF (MOPF), OPF approach based on linearization and approximation, OPF approach based on considering storage devices, unbalanced three-phase OPF (TOPF), alternating direction method of multipliers (ADMM), OPF based on the simultaneous formulation of the post-contingency flows, and uncertainty-based OPF models.These approaches can be handled in terms of system type (three phases versus single phase), system balance (balanced versus unbalanced), operational state (islanded versus grid-connected mode), network topology (distribution versus transmission), programming model (dynamic versus static), control strategy (centralized versus decentralized control), multi-agent versus central agent, solution algorithm (mathematical approach versus heuristic algorithm), and realistic description (deterministic versus uncertainty).

Forecasting Algorithms and Energy Management Strategies
In Ref. [9], focusing on accurately forecasting power generation and load-to-energy management in MGs, forecasting algorithms for the power supply side and load demand were addressed.The forecasting techniques in two categories of hybrid and single models were discussed.In hybrid models, the combination of artificial neural network (ANN) with wavelet, Fuzzy logic, support vector machines (SVMs), and genetic algorithm (GA) were detailed.The single models were classified into artificial intelligence (AI) (including SVM, Fuzzy logic, and ANN) and parametric techniques (including statistic methods and the Kalman filter).
Battula et al. [12] discussed the forecasting techniques for EM of MGs in terms of different parameters (including load, price, and weather).They categorized them into different types based on the required foresting period, including very short-term (from multi seconds to 30 min, used for the dynamic control of RES to the load requirements); short-term (from 30 min to 6 h, used for energy scheduling of different sources and storage devices); medium-term (from 6 h to one day, used for market pricing); long-term (from one day to one week, used for maintenance and load dispatch).In another classification, the forecasting models were categorized based on the used model, linear (including time series; dynamic programming (state space and ARMA)) and non-linear (SVM; Markov chain; stochastic; Fuzzy neural; ANN (supervised, unsupervised, and reinforced)).

Uncertainty Modeling of EM
Pourbehzadi et al. [13] presented a summary of the proposed approaches in terms of MG Type (AC, DC, AC/DC, grid-connected, and islanded) and solution methodology (deterministic, probabilistic (probability density function/uncertainty model), single/multiobjective, correlation, and EV).They mentioned some methods that were suggested for uncertain optimization such as: analytical hierarchical process, artificial bee colony/improved differential evolution algorithm, clonal selection algorithm, GA, improved teaching-learning optimization algorithm, modified firefly optimization algorithm, modified teaching-learning algorithm/fuzzy-based clustering, NSGA II, PSO/self-adaptive probabilistic mutation, probabilistic load flow, robust optimization, self-adaptive modifies honey bee optimization/fuzzybased clustering, stochastic programming approach, stochastic programming/cuckoo optimization algorithm, and two-stage stochastic integer programming/robust optimization.Furthermore, different probability density functions/uncertainty models used for modeling the uncertain parameters are: Latin hypercube sampling, real-world values, MCS, Weibull distribution/MCS, 2m-PEM, 2m + 1 PEM, Gaussian mixture model/chance constrained, chance constrained, scenario-based, unscented transformation.In addition, the modeled uncertain parameters were described as power generation of RES, load demand, solar radiation, wind speed, fuel price, load growth, component outage, market price, islanding, a daily-driven distance of PHEVs, electricity price, wind speed, forecast error of active and reactive loads, power loss cost factor, customer interruption cost, failure rate, repair rate, cost function coefficients, the active power output of conventional units, active power flow of transmission lines, and correlated loads.
Abdi et al. [1] stated that deterministic techniques cannot consider the impacts of different uncertainties, arising from the high penetration of RESs, load demand forecasting errors, etc.For this purpose, and in a technical categorization, they suggested using three main methods of MCS, analytical, and approximate methods.The first one is accurate as handling non-linear and complex problems, but it is computationally expensive.Analytical methods, which are based on some mathematical simplifications, can overcome the deficiency of MCS.The first-order second-moment method, Taylor series expansion method, Cumulant method, common uncertain source method, discretization method, and PEM are some methods to model both shortages of the two mentioned methods.
Pourbehzadi et al. [13] focused on IBM ILOG CPLEXs, Matlab/Simulink, and CPLEX software for EM in an MG:

Blockchain Technology
Dinesha and Balachandra [39] mentioned that due to the increased penetration of DERs and to overcome climate change, a shift from centralized large-scale generation units to distributed small-sized networks is mandatory.This transition needs to use the distributed ledger technology (DLT) management of energy, information, and money data, by applying blockchain technology due to its numerous advantages of having privacy protection, and facilitating accurate, fast, and real-time settlement of financial transactions.The authors investigated the possibility of developing blockchain-enabled smart microgrids (BSMGs), including the process flows and transaction protocols.For this purpose, after addressing the significance of blockchain technology in SMGs, its different applications in the energy industry were detailed.Then, a review of some projects, including Ethereum, HyperLedger, Tendermint, and open-source projects was performed.Furthermore, the suggested structural (including constituents of a BSMG and layers in BSMGs) and operational (consists of process flow, pricing mechanisms, interoperability between heterogeneous blockchain platforms, and proposed setup of BSMGs) frameworks for BSMGs were discussed.Finally, the authors presented some qualitative and quantitative criteria for the evaluation and benchmarking of the proposed model.

Challenging Issues
Based on the literature, the most important challenging issues in EM of MGs are categorized as follows:

•
Optimization of the large-scale problems of MMGs [8].

•
Forecast both power and load demand in MGs [9].

•
Modeling the spatial correlation between different renewable power generation sources [10].

•
Appropriate modeling of uncertain parameters such as renewable resources, load demand, and electricity price [10,14].

•
Modeling the active power trading and sharing in IMMGs [10].

•
Applied research on xEVs for worldwide deployments, including the EVs speed, the required standards, the connectivity of the EVs with variable-source-based MGs, and the departure and arrival pattern of xEVs [14].

•
Need for addressing some well-known problems in traditional power systems in MGs, such as the profit-based unit commitment problem (PBUCP) [40] • Further investigation of the cost-effective EM techniques for SMG networks [17].

•
MG controls at the global MG level: islanding detection, re-synchronization with the upstream network, power quality management, protection from internal faults, and planning new MGs [11].
• MG controls at MG components level: need the critical ESS components for periodic maintenance, minimizing generation fluctuations, communication between nodes and control agents, and limitations in plug-and-play capabilities of different components [11].

•
Using machine learning to handle the large amount of front-end data produced by MGs and ESSs [11].

•
Further investigations on the effect of the conventional grid on greenhouse gas emissions [12].

•
Monitoring: the need for real-time semantic features for EM, optimal sensor location strategies, and an occupant detection system [18].

•
Analysis: the need for semi-supervised approaches for classification of the loads, consumers, multivariate forecasting time series approaches, and fault detection methods [18].

•
Management and decision making: the need for dynamic, adaptive, and distributed control schemes, and smart real-time energy consumption scheduling [18].

•
Maximizing the utilization of green sources [16].

•
Reduce the use of utility power [16].

•
Different optimization issues in the operation of multi-carrier energy systems, such as integrated optimal power and gas flow (IOPGF) problems [41,42].

•
Load curtailment costs related to different operating conditions [1].

•
Modeling the impact of the market practice on reactive power costs [1].

•
Applying novel and comprehensive methods mainly based on heuristic algorithms [1].

•
New objective functions linked to the storage devices allocation problems [1].

•
Issues and challenges for implementing an interoperable smart microgrid (ISM), including: alerts and alarms, bandwidth, channel analysis, data fluctuations, data privacy/security, data rate, distance coverage, energy efficiency, latency, link budget, link failures, network topology, node placement, power consumption, receiver sensitivity, scalability, spectrum usage, standards applicability, system cost, system migration, technology access, throughput, and typical framework [34].

The Surveys on Other Fields of MGs
Based on the literature, there are many survey papers regarding the MGs.Hereafter, some of these references are addressed.Interested readers are referred to this long list for further investigations.