Research Of Interval Optimization Algorithm And Its Application In Dynamic Optimization

Interval optimization algorithm is an optimization method which based on interval analysis and different from traditional ones. It uses the interval variable as an unit to take part in the computation and have also formed a series of operation rules gradually. In the field of controlling, the interval method which has the encompassment character can provide a new way to process some difficulties, no matter in the research of theory and in the application of engineering. PSO method using random search technique which base on the cooperation of mean-agent, can greatly improve the global search capability. However, the randomness can always make the algorithm trap into local optimum.In this paper, the certainty superiority of the interval optimization method and the powerful global search capability of the PSO were combined, the new algorithm generated and then has been modified. Considering many actual control problems are in dynamic environments, much effort have been made to do the research of the dynamic optimal problems based on the interval method. A lot of work has been made to study these aforementioned questions, several aspects of the research achievement followed:Firstly, the basic theory of the interval optimization method has been relearned deeply, the already existed problems have been mentioned through some examples. A new interval puzzle has been presented named as the vector operation of the interval vector.Secondly, combine the respective superiority of the interval method and PSO algorithm, a new modified interval optimizaiton algorithm (MIOA) based on the dynamic interval shrink was proposed.10 testing functions were used to simulate, the results predicted that the method based on the dynamic interval shrink mechanism was better than the traditional interval method and the one based on non-dynamic interval method, about 10-20 times faster than the above two.Then, more simulations were carried out to test how PSO control parameters, interval particle numbers and the initial interval width influence the effect of the modified interval method. The results predict that the modified interval method is not sensitive to the PSO control parameters, simulations also point out some other methods which can help to get a better result.At last, in allusion to dynamic problems, "gridding test" was devised to perceive whether the environment changed and "the best interval expansion" was designed to respond the changed environment. The interval dynamic optimal algorithm (IDOA) was proposed base on the MIOA, many tests were carried out in the DF1 dynamic environment, and the results indicated that the IDOA performance well in both periodic changed environment and non-periodic changed environment. IDOA can track the changed environment accurately and Acc, Ada index keep in a low level.In the end of the paper, new research directions which based on the above-mentioned were proposed as the further job.