Differential evolution algorithm to improve research

1. The characteristics and existing problems of differential evolution algorithm
Differential evolutionary algorithm has the following four characteristics, is presented in the following:
1) differential evolution algorithm has strong versatility, and low dependence on the characteristic of the 1 with information,
2) the differential evolution algorithm in the search process has the ability of the optimal individual memory,
3) differential evolutionary algorithm can take advantage of the local information of the individual and group global coordination of information search, use the differential strategy on variation as a group can effectively utilize the distribution characteristics of meters to improve the ability to search for,
4) differential evolution algorithm with other algorithms (e.g. mosquitoes swarm algorithm, artificial the swarm algorithm, etc.) for fusion and structure, produce better, more efficient hybrid optimization algorithms,

Differential evolution algorithm is also exist the following problems, in the future study of the algorithm should be deep progress and improve,
1) the search performance of differential evolution algorithm parameters are dependent on sex and bureau search ability is weak,
2) the differential evolution algorithm to improve search efficiency, short time is difficult to ensure that can get global optimal solution
3) differential evolution algorithm with other similar group of intelligent optimization algorithm, the theoretical study is less, the lack of practicability of algorithm design guidelines,


2. The introduction of the adaptive crossover probability
In differential evolution algorithm, crossover probability is a fixed value, the rate of pay and stables value affects the diversity of the algorithm, the size of the search ability and speed, the larger crossover probability can enhance the genetic diversity and global search ability, small cross rate both can accelerate the speed, strengthen the local search ability, with the artificial bee colony algorithm control the parameters of the random way the heart there is a lot of similarities, reference book 6. 3. The practice of 2 in the differential evolution algorithm is introduced in the adaptive crossover probability and its value with the increase of the selected generation number k increasing gradually, make the algorithm global searching ability in early or late in the local search ability and convergence speed are improved, and maintain the population diversity, through a lot of experimental tests such as Mao Running finally determining the CR value is between 0.4 to 0.9, the differential evolution algorithm optimization effect is best, this book take adaptive crossover probability formula is:
CR=20 k + 0.4
Type of k for the current loop algebra, k for the largest circulation algebra,