Orthogonal learning particle swarm optimization for power electronic circuit optimization with free search range

Abstract

Power electronic circuit (PEC) always consists of a number of components such as resistors, capacitors, and inductors which have to be optimized in order to obtain good circuit performance. In current studies, the search ranges of these components are always pre-defined carefully by expert designers, making it difficult for practical applications. In this paper, the search space is freely set to the commonly used ranges and an efficient orthogonal learning particle swarm optimization (OLPSO) is applied to optimally design the PEC with such search space. OLPSO uses an orthogonal learning (OL) strategy for PSO to discover useful information that lies in the personal historical best experience and the neighborhood's best experience via orthogonal experimental design. Therefore, OLPSO can construct a more promising and efficient exemplar to guide particle to fly better towards the global optimal region. OLPSO is implemented to optimize the design of a buck regulator in PEC. The optimized results are compared with those obtained by using a genetic algorithm (GA) approach and those obtained by using PSO with traditional learning strategy. Results show that the OLPSO algorithm is more promising in the design and optimization of the PEC with large search space. Moreover, the simulations results demonstrate the advantages of OLPSO by showing that the circuit optimized by OLPSO exhibits better startup and large-signal disturbance performance when compared with the one optimized by GA.

Resource-Aware Distributed Differential Evolution for Training Expensive Neural-Network-Based Controller in Power Electronic Circuit

“…According to Wilcoxon's rank-sum test at a 0.05 significance level, RADDE performs significantly better than other algorithms. In addition, given an acceptable fitness value of 150 following the suggestion of Zhan and Zhang [39], RADDE succeeded in 25 out of the 30 runs, showing its strong reliability. Although RADDE has similar best and worst values to OLPSO, RADDE obtained 16 runs (which is more than half of the 30 runs) with fitness values larger than 192.632 and five runs below 150, resulting in a median value close to the best and a large Std Dev.…”

“…4. It is a buck regulator with two parts, i.e., a classical buck converter as the PCS and a proportional-plus-integral controller as the FNC [37]- [39]. In the PCS part, we assume that R L , r C , and r E are known in advance, and the components L and C are 200 and 1000 μF, respectively [37].…”

“…Particularly, the input voltage range v in is set as [20 V, 60 V], the output load R L is set in the range of [2 , 12 ], the nominal output voltage is 5 V±1%, the switching frequency is 20 kHz, and the maximum ST is 20 ms. The search ranges for resistors R 1 , R 2 , R 3 , and R 4 are set to be [100 , 100 k ], and the search ranges for capacitors C 2 , C 3 , and C 4 are set to be [0.1 μF, 100 μF] following [39].…”

“…GA [37], particle swarm optimization (PSO) [40], and orthogonal learning PSO (OLPSO) [39] are taken for comparisons since they have been successfully applied to PEC optimization and are representative. In the GA, the population size is set as 30, the crossover rate is set as 0.85, and the mutation probability is set as 0.25.…”

A Study on Various Particle Swarm Optimization Techniques used in Current Scenario

“…The Mac-span declared by the algorithm is specified in the Mac-span of the proposed algorithm and compared. Well known solution [28]. Two PSO algorithms are available in the literature Their PSOs were mixed with simulated one"s analgesics.…”