Figure 1. Classic Strategies used for DSM
This paper presents a survey on Demand Side Management techniques used to reduce the energy waste, postpone the construction of new plants, reduce costs or electricity bill, and reduce total power demand during peak demand periods. One of the major goals of DSM is reducing consumption during peak hours and shifting load to off-peak hours. Several algorithms and techniques for load shifting have been reported in researches. Different approaches have been suggested to solve the demand response problem using linear and dynamic programming techniques. There are different types of optimization techniques as Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Evolutionary Algorithm (EA), Game theoretic techniques and others. Researches are performed on different optimization techniques to reduce peak load demand, reduce operational cost, reduce PAR (Peak to Average Ratio), and reduce the discrepancy between power supply and demand.
DSM is described as the planning and monitoring of utility's activities designed to encourage customers to adapt their electricity consumption routines by considering the timing and the level of electricity demand. Thus, help the customers to use electricity more efficiently.
The load shapes which indicate the daily electricity demands of consumers between peak and off peak times can be shown in Figure 1. There are three DSM techniques, i.e. Peak clipping (reduction in the peak demand), Valley filling (increased demand at off-peak), and load shifting (demand shifting to non-peak period) [58, 70].
Figure 1. Classic Strategies used for DSM
A smart grid is an electrical network that manages electricity demand in a sustainable, reliable, and economic manner. The smart grid was introduced with the aim of overcoming the weaknesses of conventional electrical grids by using smart net meters. Demand Side Management (DSM) is an important feature of SG to improve energy efficiency, reduce the peak average load, and minimize cost. In future smart grids, all buildings would have a part called Central Scheduling Unit (CSU) and all customers in flats would be installed with digital smart meters, that is equipped with digital and programmable control unit. The smart meter has communications with the CSU and vice versa as shown in Figure 2 [4, 19, 20, 25].
Figure 2. An Assumed Demand Side Management System
Optimization techniques are used to minimize the peak load, minimize cost taking into account user's individual preferences for the loads by setting priorities, and preferred time intervals for load scheduling and Reduce PAR (Peak to Average Ratio).
It depends on customer's awareness to reduce their consumption to improve efficiency of equipment and process.
In Demand Side Management (DSM) techniques, the load is shifted from peak hours to off peak hours which results in financial benefit to residential customers. The optimization should reduce the total costs of the system.
The flexible devices should be shifted to off peak period according to Load Priority Technique (LPT), taking into account user's individual preferences for the loads by setting priorities and preferred time intervals for load scheduling.
DSM using load scheduling can be done in two levels, viz. Day-ahead and Real time scheduling as in Figure 3.“The Day Ahead Shifting” technique is proposed to minimize total costs of the system.
Figure 3. DSM using Load Scheduling
Filling valleys can be done using energy storage technologies by using different electricity generation systems as a small Combined Heat and Power (CHP) plant, a battery system and a photovoltaic (PV) system, using renewable energy to store surplus energy in times of higher generation and to provide energy in times of lower generation.
Domestic appliances of an individual consumer (or a group of consumers) are classified by their importance. When a given condition is reached, single loads (i.e., appliances) can be disconnected by a central management system according to Load Priority Technique (LPT) and utility should encourage consumers by using different electricity tariffs, for instance increase electricity price during peak hours and decrease electricity price during off-peak hours, and residential consumers can be encouraged to adapt their behavior.
The components of DSM are illustrated in Figure 4.
Figure 4. Components of DSM [53]
This survey shows different types of optimization techniques used to minimize the peak load and cost. Different researches were performed on each type and are compared with other techniques to show better solution.
GA was first proposed by Holland [22]. GA based scheduling algorithm is used to solve cost optimization problem, reduce PAR (Peak to Average Ratio), and increase the efficiency of smart grid by increasing the utilization of spinning reserves. This strategy is beneficial for SG. Large number of appliances in smart grid increases the complexity of the problem. GA have potential to solving these types of complex problems [4, 23, 45, 52].
In 2016, Qihang Li et al., made a comprehensive view of unit commitment problem which involves cost factors of intermittent wind energy as well as the dispatching load from EVs into total scheduling cost of the system. Through complete modeling of the unit commitment constrains and the price mechanism of demand response, a genetic algorithm was proposed to solve this problem [37].
In [29] Particle Swarm Optimization (PSO) was first introduced by Eberhart and Kennedy as an optimization technique. The state space is represented as a swarm of particles. Each particle is assigned fitness value. Thus, the swarm best particle is the particle with highest fitness, the rest of the swarm is updated iteratively in a way that results in moving the swarm toward the best solution [5].
In 2009, A. Pedrasa et al., wrote a paper using Binary Particle Swarm Optimization (BPSO) to schedule a significant number of varied interruptible loads over 16 h [5].
In 2011, J. Agneessens et al., wrote a paper using Binary Particle Swarm Optimization (BPSO) to find a curtailment schedule in order to smoothen the voltage along a distribution feeder [1].
In 2015, T. Logenthiran et al., used particle swarm optimization to minimize cost taking into account user's priorities and to find an optimal load schedule for different devices over a day [38] and this technique is also used by Sanjaya Kumar Nayak et al. to minimize the peak load and minimize cost [49].
In 2016, Ishan Gupta et al., solved DSM problem using Particle Swarm Optimization (PSO) in three area loads of smart grid, i.e. residential, commercial, and industrial in order to minimize cost and the peak demand [21].
In 2016, Hsin-Hui Kuo et al., used Particle Swarm Optimization (PSO) to compute the optimal activity schedule based on user preferences in nearly real-time [23].
In 2016, Bilal Hussain used Particle Swarm Optimization (PSO) to compute optimized schedules for home appliance and local distributed energy resources [24].
In [40], Maniezzo introduced a new discrete combinatorial optimization technique that mimics the ants' colonies behavior. ACO is a meta-heuristic optimization approach. That used for solving problems that can be reduced to finding good paths through graphs. The state space is searched by ants. These ants visit the graph nodes. Ants communicate by depositing pheromones along their path. The amount of pheromones deposited is inversely proportional to the length of the path. The likelihood of the path is directly proportional to the amount of deposited pheromones on that path. Thus, ants are most likely to select the ones with a higher pheromone value. This behavior causes the convergence to optimal solution [40].
In 2015, André Silva et al., used ACO to solve the problem of scheduling tasks to minimize peak-times and cost [57]. In 2016, Sahar Rahim et al. used ACO for electricity bill reduction and minimization of PAR while considering user satisfaction [53].
Game theory is used to handle DSM problems and capture the competition between users. The two types of Game Theoretic Techniques are shown in Figure 5.
Figure 5. Game Theoretic Techniques
In 1999, M. Fahrioglu et al. used Game Theory to analyze the interaction between the different users and the utility to encourage customers to reveal their true value of power [17].
In 2010, A.H. Mohsenian-Rad et al. proposed an incentive-based optimal methodology for energy consumption scheduling. The proposed algorithm optimizes the energy cost, and balances the total residential load for multiple users sharing a common energy source [48].
In 2011, Shengrong Bu et al. proposed a gametheoretical decision making scheme for electricity retailers in the smart grid, where real-time pricing DSM is used [9].
In 2011, Chen Chen et al. proposed a smart a Real-Time Pricing RTP based power scheduling scheme for residential power usage using a Stackelberg game model to reduce peak load and the variance between demand and supply [11].
In 2012, Jiang Chen et al. used a Stackelberg game approach to deal with Demand Response (DR) scheduling, the Stackelberg game approach based on real-time pricing (RTP) [12].
In 2012, Walid Saad et al. made a comprehensive overview of the applications of game theory in the smart grid [55].
In 2014, Hazem M. Soliman and Alberto Leon-Garcia used game theory principles in order to analyze the interaction between utility and the varies users assuming smart grid communication infrastructure and the presence of storage elements. The utility decides its prices taking into consideration the reaction of the customers, and where the users schedule their energy in order to minimize their cost based on the prices given by the utility [59].
In 2013, Ryohei Arai et al. used a differential game and an optimal control problem to formulate the three control schemes, i.e. decentralized open-loop control, decentralized feedback control and centralized control [2].
In 2016, Chathurika P. Mediwath and David B. Smith used non-cooperative Stackelberg game between community energy storage (CES) device and users with rooftop photovoltaic panels to minimize their personal daily energy costs [44].
In 2016, Feng Ye et al. used centralized scheme to minimize power generation cost and also proposed game theoretical approaches to motivate customer by extra savings if participating and to minimize global Peak to Average Ratio (PAR) of power usage in the grid [64].
In 2016, Panagiotis D. Diamantoulakis et al. used a Stackelberg game to help the operator maximizes its Profit, and to optimize utilization of the renewable power to reduction of the power price for the consumers [16].
In 2010, Jichen Zhang et al. used a multi-period stochastic mixed integer programming model for power generation scheduling in a day-ahead electricity market. They used a heuristic solution methodology to finding good quality feasible solutions within reasonable computation time [67].
In 2011, T.T. Kim and H.V. Poor used stochastic dynamic programming to reduce the conflict between power supply and demand and to minimize the cost [30].
In 2012, Zhi Chen et al. compared stochastic optimization and robust optimization approaches to real time demand response management for residential appliances. The stochastic optimization is used for minimizing electricity bill for the entire day, the robust optimization take into consideration the price uncertainty intervals for minimizing the worst-case electricity payment [14].
In 2015, Atefeh Zomorodi Moghadam et al. used a stochastic framework to facilitate DR aggregator participation into the electricity market and aims to maximize its profit for day-ahead through a stochastic price-based self-scheduling problem [47].
In 2016, Enxin Yao et al. used an autonomous energy consumption scheduling algorithm to schedule the operation of deferrable loads to reduce the peak load and reduce the reverse power flow with high-power generation from the PV units, used stochastic programming to formulate an energy consumption scheduling problem to reduce cost [63].
In 2016, S. Sofana Reka and V. Ramesh used a two stage stochastic process with the target of maximizing the profit and minimizing the cost [54].
Differential Evolution was originally developed by Storn and Price in 1997 [60].
In 2014, André R.S. Vidal et al. used an Evolutionary approach to solve the problem of DSM on smart grid using concept of day-ahead load shifting to obtain the lowest possible cost of energy [62].
In 2012, Daniele Miorandi and Francesco De Pellegrini used an Evolutionary Algorithm for analyzing and predicting the adoption of DSM solutions by users under two different pricing schemes [46].
In 2012, Thillainathan Logenthiran et al. used load shifting technique for demand side management of future smart grids and used a heuristic-based Evolutionary Algorithm (EA) to solve this minimization problem [39].
In 2014, Ingo Mauser et al. used an Evolutionary Algorithm to optimize the energy consumption with respect to variable tariffs, load limit signals, and the user's behavior [43].
In 2016, Pranjal Pragya Verma et al. used Differential Evolution to solve a lossless optimal power flow DC-OPF for IEEE-24 bus (RTS) system for duration of 24 hours without considering ramp constraints or system losses with an objective to minimize the total system generation cost [61].
In 2013, Italo Atzeni et al. formulated the day-ahead grid optimization problem to minimize user's cumulative monetary expense for buying/producing his energy needs, using a game theoretical approach, and study the existence of the Nash equilibria [3].
ABC algorithm is a stochastic population-based algorithm that was introduced by Karaboga through mimicking the behavior of bee swarm in 2005. The honey bees in swarm are classified into three groups: employed bee, onlooker bee, and scout bee. Employed bees are responsible for searching foods around food sources and sharing food information with onlooker bees. After getting the food information, the onlooker bees choose food sources and calculate the nectar amount to calculate solution quality. In ABC, a food source represents a possible solution to the optimization problem, and more nectar means better solution [6, 26-28].
In [69], Y. Zhang et al. proposed an improved ABC algorithm named WGABC for numerical optimization, which introduced a linear decreasing inertial weight and a new food generation scheme at scout bee stage based on the framework presented in [71].
In 2015, Yanyu Zhang et al. used an improved ABC algorithm to schedule the operations of home appliances according to the day ahead electricity price to minimize electricity cost [68].
In 2015, Vijaya Margaret and Kuma Rao used ABC algorithm to schedule the residential loads for various hours of the day to encourage consumers to participate in demand response through incentives [41].
In 2015, Mousa Marzband et al. used Artificial Bee Colony optimization algorithm for economic dispatch. An Artificial Neural Network combined with a Markov Chain (ANN-MC) approach is used to predict non dispatchable power generation and load demand considering uncertainties [42].
In 1996,C.N. Kurucz et al. used Linear Programming (LP) to schedule commercial, industrial, and residential load to optimize the amount of system peak load reduction [34].
In 1998, Kah-Hoe Ng and Gerald B. Sheble used a Linear Programming (LP) to schedule group of customer load type to maximize the profit and a profit-based load management is used to examine generic direct load control scheduling [50].
In 2012, Ziming Zhu et al. used an Integer Linear Programming to schedule both the optimal power and the optimal operation time for power-shift able appliances and time-shiftable appliances to minimize the peak hourly load [72].
In [32] Yu and Kouvelis used a Robust Discrete Optimization to find a solution that minimizes the worst case performance under a set of scenarios for the data [7, 8].
In 2017, Cuo Zhang et al. used a Two-Stage Robust Optimization (TSRO) to optimize the proposed microgrid coordination strategy, which maximizes the total profit with operational limits [66].
In 2017, Javier Zazo et al. used Robust Worst-Case Analysis to coordinate the energy load so that users minimize their monetary expenditure [65].
In 2010, Antonio J. Conejo et al. used a robust optimization to reduce the electricity bill of a consumer that integrates the proposed procedure in its EMS (Energy Management System) [15].
In 2013, Nandkishor Kinhekar et al. used Non- Conventional Optimization technique to reduce utility operational cost and system peak load [31].
In 2010, L. Chen et al. used distributed demand response algorithms to shape power demand to match power supply [13].
In 2010, Pedram Samadi et al. distributed algorithm to minimize the cost and keeping the total power consumption below the generating capacity [56].
In 2010, S. Caron and G. Kesidis used a distributed algorithm to reduce the total cost and Peak-to-Average Ratio (PAR) of the system [10].
In 2011, Na Li et al. used a distributed algorithm for the utility company and the customers consider households that operate different appliances including PHEVs and batteries to compute this optimal prices and demand schedules [36].
In 2011, Nikolaos Gatsis and Georgios B. Giannakis used distributed sub gradient method to minimizing the total cost electricity plus the total user dissatisfaction (social welfare) [18].
In 2017, Chaojie Li et al. used a distributed algorithm for sparse load shifting in demand-side management with a focus on the scheduling problem of residential smart appliances to minimize billing cost and the Peak-to Average Ratio (PAR) [35]. Table 1 presents a comparison between different optimization methods used for DSM.
Table 1. Comparison between Different Optimization Techniques used in DSM
This survey shows different optimization techniques that solve the demand response problem using linear and dynamic programming techniques. Optimization techniques are used to reduce peak load demand, reduce operational cost and reduce PAR (Peak to Average Ratio). Researches were made for comparison between different optimization techniques and also different strategies used for DSM to get better response in order to minimize the discrepancy between power supply and demand.