Plaza

Taking the reliability of power supply as a constraint and adopting the cost as an objective function, many researchers [37–45] have investigated and proposed various approaches for de- termining the optimal sizes of an IMG. Their recommended optimization techniques are mainly (1) graphical construction method, (2) probabilistic approaches, (3) enumerative/iterative tech- nique, and (4) stochastic & heuristic techniques.

• Keeping one decision variable (e.g., size of WPS) fixed and varying the other (e.g., size of PVS), graphical construction method has been utilized to determine the sizes of a battery bank and a PV array [37] and to calculate the sizes of WPS and PVS [46]. This method has utilized only two decision variables in the optimization processes and the cost function of the IMG has been taken a linear function that combines the decision variables. This method is not effective for the problems that involve more than two decision variables.

1.4. Control Strategies and Optimal Sizing of IMGs 17

limit of a battery bank in a wind-PV-battery system and to asses the performances of a PV-wind system, respectively. Probabilistic methods are generally the simplest sizing methods. The disadvantage of the probabilistic approach is that the approach cannot represent the dynamic changing performances of the IMG. Thus, the obtained results by these methods are not generally be the most suitable solution.

• By minimizing production cost, linear programming approaches have been utilized to optimize the component sizes of IMGs [39], [48], [49] and to determine optimal design of a autonomous and grid-connected hybrid wind-PV power system [50]. Reference [40] has used a computer programming technique to calculate the maximum number of storage days and minimum areas of PV array for a PV-diesel-battery system. However, the linear programming approaches are not effective for a large and complicated IMG. The approaches may end up the optimization process in a suboptimal solution and have required a large computational effort.

• By employing enumeration and/or iterative schemes and minimizing costs upon main- taining the power supply reliability to a desired value, many researchers [41, 44, 51–58] have proposed the single objective optimal sizing approaches for numerous configura- tions of an IMG. Among the aforementioned studies, the most [41], [53], [54], [56], [58] have optimized the sizes of very small IMGs (e.g., a single house/motel) where the daily average load (DAL) is less than 75 kWh, some others [51], [52], [55] have determined the optimal sizes of small IMGs that have the DAL between 100 kWh and 700 kWh, and the rest [44], [55] have calculated the optimal sizes of medium IMGs that have the DAL between 800 kWh and 2000 kWh. Reference [41] has proposed a simple numerical algo- rithm to determine the optimum generation capacities and storage for a hybrid wind-PV- battery system. Based on loop iteration, reference [59] has presented a general method- ology for technical-economic analysis of an autonomous renewable power system. Few of the aforementioned studies have utilized PV-wind-diesel-battery configuration while the rest have employed PV-wind-battery/fuel cell in the configurations. All of the afore-

18 Chapter 1. Introduction

mentioned studies have neither utilized more than one PMSs for simulating the IMGs nor optimized the PMSs as well and they have accounted in small number (e.g., two to four) of decision variables. None of the above studies have adopted converter and/or battery charger size(s) as decision variable(s). In the aforementioned studies, the effect of reac- tive power on component sizes during the optimal sizing studies has not been taken into account, though the reactive power issues (e.g., the voltage deviations) in an IMG are investigated in many dynamic studies and control designs [12], [19]. Many researchers have proposed methods to minimize the intermittencies of solar irradiance [60] by anal- ysis and wind speeds by utilization of BESS [61]. The aforementioned optimal sizing studies have not investigated the impact of RERs’ intermittency on LCC.

• Among the stochastic and heuristic techniques, GA [3], [42], [62], [63], particle swarm optimization (PSO) [64], [65] simulated annealing (SA) [66], and tabu search [67] have been used for determining optimal sizes of an IMG. Assuming the total cost to be an objective function, GAs are utilized in [3] to determine the optimum number of PV modules, wind turbine generators, and battery banks and to develop a hybrid optimiza- tion GA (HOGA) program [26] for simultaneously determining the optimal sizes and control strategies of a PV-diesel system, while reference [62] has presented a methodol- ogy for the optimal sizing of a PV-wind-battery system. Reference [68] has employed GA to jointly optimize the sizes and operations of a hybrid-PV system while refer- ence [69] has utilized GA for investigating optimal sizes and economical analysis of a wind-microturbine-PV-battery hybrid system. By ensuring a low LPSP and minimiz- ing the annualized costs, reference [42] has suggested a GA to optimize the sizes of a PV-wind-battery system. Taking the total costs as an objective function and based on stochastic gradient search, reference [66] has used SA algorithm to optimize the sizes of PV-wind-battery system. Though the GAs are generally robust in finding global opti- mal solutions in multi-modal and multi-optimization process, the attainment of a global minimum with a small number of population is sometime uncertain. Reference [70] has proposed tabu search technique to optimize the sizes of a PV-wind-diesel-battery system

1.4. Control Strategies and Optimal Sizing of IMGs 19

with a large number of decision variables where the peak load has been 120 kW. Stan- dard PSO has the shortcomings that the calculation time is too long and it is easy to fall into local optimal solution [64]. The most studies have utilized chronological simulation scheme for determining the optimal sizes. The aforementioned studies have either used typical meteorological year (TMY) or typical day or typical month time series weather data to simulate the IMG. The typical day or typical month time series cannot provide satisfactory solution due to avoiding the seasonal variations while the TMY time series needs large central processing unit (CPU) time. Therefore, the TMY-based methods are computational intensive as well as time consuming.