An Efficient Dynamic Micro Multi-objective Optimization Method And Its Applications On PID Control

In engineering design, real-time control, artificial intelligence and otherengineering fields, there are a wide range of optimization problems that the objectivesand constraints will change over time. And this kind of problem is called dynamicoptimization problem. If the optimization problem has multiple conflictingoptimization goals, then it is known as a dynamic multi-objective optimizationproblem. The complex nature of the dynamic multi-objective optimization problemitself makes this kind of problem very difficult to solve. It requires the algorithm tomake adjustments according to changes in the environment continuously, and searchthe changed Pareto optimal solution set as quickly as possible. This will inevitablyrequire dynamic multi-objective optimization algorithm has high solution efficiency.Especially in the face of increasingly complex engineering problems, many existingalgorithms will also become powerless.To solve this problem, a high efficient dynamic micro multi-objective geneticalgorithm based on the Micro Multi-objective genetic algorithm is suggested. Thealgorithm uses small-scale evolutionary population to improve the solution efficiency,and tracks changes of the environment through an introduction of an environmentdetection mechanism in the evolutionary process to ensure that the algorithm searchnew Pareto optimal solution set quickly. When the population of the currentgeneration accomplishes the genetic manipulation, the detection mechanism will bestarted for environment testing. Calculate the value of all the individual objectivefunctions and the constraint functions, and compare the results with the correspondingfunction value calculated in the previous generation. If there is any difference, weconsider the problem has changed over time. At this time, select the appropriate wayto regenerate a population, and combined with the saved information solutions ofprevious generation and external populations, update non-dominated solutions set intoa new evolutionary process.Then, four different types of dynamic multi-objective optimization test problemsare used to test the performance of the efficient dynamic micro multi-objectivegenetic algorithm. And by comparison with DNSGA-II, an existing dynamicmulti-objective optimization algorithm with good performance, verify the effectivedynamic micro multi-objective genetic algorithm is an algorithm with high solution efficiency and accuracy.Finally, the dynamic micro multi-objective genetic algorithm is applied to amulti-objective optimization problem of a dynamic speed control system of vehiclediesel engine to optimize PID control parameters. The two main performanceindicators of the step response overshoot and rise time is used as the objectives tooptimizer The results showed that when the environment changes, the algorithm canquickly find the Pareto optimal solution set in the new environment, and bycomparison with two existing PID control scheme, proved efficient dynamic micromulti-objective genetic algorithm in solving efficiency obvious advantages.