BUCK, BOOST AND BUCK-BOOST CONVERTER DESIGNS WITH VARIOUS METAHEURISTIC METHODS

Abstract: One of the basic circuit structures in the field of power electronics is DC-DC converters. As these design steps require many mathematical operations, these problems are hard to solve by hand. In addition, choosing the proper component values is always curial when adopting the computer-based designs to the real-world. In this study, the software is developed for the designs of buck, boost and buckboost DC-DC converters via metaheuristic algorithms that calculate the parameters of the circuits. The components of the specified DC-DC converters are selected via the software with a user-friendly interface, under the desired criteria from the industrial series (E12, E24 and E96), by using eight different metaheuristic algorithms (artificial bee colony, differential evolution, genetic algorithm, particle swarm optimization, cuckoo search, harmony search, lightning search and gray wolf optimizer). The designs and analyses of DC-DC converters that are chosen according to the type and features (determining/selecting the components in accordance with the specified industrial series) can perform easily, fast and effectively through the software developed for this purpose.


INTRODUCTION
DC-DC converter circuits, which carry a certain level of DC voltage to an another level (boosting or stepping down of voltage), are very important in the field of power electronics. The main design steps can be summarized: achieving the desired voltage level at the output, calculation of the components with different topologies of the circuits for minimum voltage fluctuation, determination of the critical components of the circuit for a continuous current and selection of the optimal component values in accordance with the industrial series.
Circuit components calculated by solving the design steps consisted of the systems of equations with multiple variables may not be existed in the industrial series. This requires all calculations to be made again by selecting the appropriate industrial series. It is challenging and time-consuming to perform all the design steps by hand as the complexities of the problem rise. Metaheuristic algorithms are able to solve these kinds of the problems quickly (León- Aldaco et al., 2015). In addition, as conventional methods require derivative information and are highly sensitive to initial points in some cases, metaheuristic algorithms are one step ahead of them in solving complex problems. In these optimization processes, metaheuristic algorithms can be utilized effectively and efficiently (Simon, 2013;Price et al., 2005;Dasgupta, 1997;Yang, 2014;Yang, 2010;Vatansever et al., 2015). The basic properties and advantages of metaheuristics can be found in the related studies (Vasant, 2012;Du and Swamy, 2016).
In this work, artificial bee colony (ABC) (Karaboga, 2005), differential evolution (DE) (Storn and Price, 1995), genetic (GA) (Goldberg, 1989), particle swarm optimization (PSO) (Kennedy and Eberhart, 1995) This paper is organized as follows: DC-DC converters and the used metaheuristic algorithms are briefly described in Section 2-3. The designed simulator and its applications (results) are given in Section 4 and finally, the study conclusions are detailed in Section 5.

DC-DC CONVERTERS
The DC-DC converters changing the DC voltage levels at the input and output can be divided into three groups:  Buck (step-down): DC/DC switching voltage regulator that reduces the input voltage.  Boost (step-up): DC/DC switching voltage regulator that increases the input voltage.  Buck-boost: DC/DC switching voltage regulator that reduces or increases the input voltage. The main topologies and some equations of the converters are summarized in Table 1 (Rashid, 2013;Gürdal, 2008).

METAHEURISTIC ALGORITHMS
Metaheuristic algorithms are inspired from various methodologies, such as biologic and social behaviors in the nature, as in the artificial bee colony, genetic, differential evolution, particle swarm optimization algorithms, or physic-based like the lightning search algorithm. These algorithms are frequently used in the field of optimization related to social and engineering sciences. The pseudo codes of the algorithms used in this study are also summarized in Table 2

Critical inductance
Voltage fluctuation    In the first application, the buck converter that reduces to (frequency of the circuit is , desired voltage fluctuation in the output is ) is designed by using E12 series. The results of the first application are given in Table 3. As can be shown in this table, considering component values found by the algorithms, ABC, GA and GWO reach the similar results and GWO appears to be slightly faster in terms of run-time. Besides, the resistor values of the algorithms are the same, except DE and LSA; however, their resistor values are still compatible with E12 series. Comparing the values of obtained by the algorithms, ABC, GA, HS and GWO algorithms achieve the similar results that are far from the results obtained by DE and LSA algorithms. When regarding the outcomes of , the highest value belongs to DE with 1.667 A. while the lowest one is obtained by PSO algorithm, having 0.942 A. In terms of run time, GWO is able to design the circuit faster than the other competitors, getting a run time of 0.091 sec., and DE is behind GWO with a small difference which is only 0.007 sec. It is worth mentioning that all the algorithms have a capable of designing the circuit in a reasonable time. In the second application, the boost converter that raises the voltage from to (frequency of the circuit is , desired voltage fluctuation in the output is ) is designed in accordance with E24 series and the comparative results can be shown in Table 4. The resistor values of the boost converters are found to be 36 Ω by ABC, DE, HS and LSA algorithms, 27 Ω by GA, 51 Ω by PSO algorithm and 47 Ω by GWO algorithm, respectively. In terms of and values, it can be seen that ABC, DE and HS algorithms achieve very similar results in the range of 14 to 17 A. As different to the first application, DE is the fastest algorithm which achieves the results in 0.058 seconds whereas GWO is the second fastest with a run time of 0.095 sec, and ABC comes after GWO, which is almost two times slower than DE. In the third application, the buck converter is designed with E96 series by using the same input values of the first application. The results given in Table 5 show that DE is the best algorithm with respect to run time as similar to second applications whereas PSO is the slowest one with a run time of 0.527 sec. which is close to that of GA. When comparing Table 5 with Table 3 in terms of the resistor values found by the algorithms, it can be seen that the probability of the finding different resistor rises as the number of E-series (from E12 to E96) used increases. This is because E96 series have more possible combinations of component values than E12 series. It can be observed from the both tables that the run time of most of the algorithms slightly increases. Generally speaking, the results show that the resistor values found by the algorithms are totally compatible with the industrial series and the run time of all the algorithms is remarkably low for all the applications.

CONCLUSION
Determining proper component values is a crucial part on the design of the circuits and components chosen from the industrial series make the designs easier, more reliable and flexible to realize. Besides, obtaining the component values with parallel or series connections can be time-consuming while designing the circuits. In this work, the resistor values of DC-DC converters, which are compatible with the industrial series (E12, E24, E96), have been found and some current values have been calculated via the eight different algorithms called ABC, DE, GA, PSO, CS, HS, LSA and GWO. The software developed for this purpose can design the related circuits comparatively with the selected metaheuristic algorithms and is able to find proper component values automatically, easily, fast and effectively. As a future work, it is planned to develop a web platform that will automatically identify all components of DC-DC converters with up-to-date metaheuristic algorithms.