Particle swarm optimization is a new evolutionary algorithm developed in recentyears. It has been widely used in function optimization, neural network training, fuzzysystem control and other applications. In this paper, the particle swarm optimization wasnot high and the algorithm was easy to get into local optimum in the later stage of PSO, forthe analysis and processing, the parameters selection and speed update formula, as far aspossible to compensate for the shortage of the algorithm. The main contents of this paperare as follows:1. The initial particle algorithm speed setting is not easy to make the fast particleoptimization, algorithm to search for long time defect, put forward the strategy of inertiaweight an exponential function, This method can make the particles in the algorithm rapidconvergence to regional advantages, the later can be accurately searching in the superiorregion, shorten the running time of the algorithm.2. The algorithm is easy to fall into local optimum in the late, the particle velocity isalmost stagnant, a method of velocity perturbation is presented, the method of updatingformula in the end, add a small speed increasing, low speed makes the algorithm pre addthe relative velocity is almost negligible, does not affect the iterative algorithm, low speedadded later to still have a smaller particle velocity, the speed of the search space can beprecise search, avoid falling into local optimal algorithm.3. Design the dominance relationship, use of the dominance relationship to constructthe non dominated set, this method has greatly shortened the algorithm the optimal solutionselected time, improve the efficiency of the algorithm; in the construction of the externalset, application of grid method was used to locate the storage for the particle position,made the particles rapidly converge to the Pareto optimal front requirements; lead into thecrowding distance technology, retain many income between the solutions of the optimumsolution, remove the inferior solution, completed the algorithm of the improve the accuracy;Selected the global optimal extreme value by random external set, ensured the algorithm distribution.4. To test the algorithm using ZDT test function and FON function of two kinds ofclassic, application of diversity, convergence and error ratio compared with the traditionalmulti target data, the image data comparison, algorithm has been improved to a certainextent. The improved algorithm is applied to the data in the optimization of coal tomethanol, temperature, pressure, hydrocarbon ratio as input, CO and CO2conversion rateas output data optimization, through the comparison with the traditional multi-objectiveparticle swarm optimization data, as can be seen, the improved algorithm in this paperimproves the traditional multi-objective particle swarm optimization to a certain extent.In this paper, the multi-objective particle swarm optimization algorithm is improved,and made the improved algorithm more convergence and higher accuracy; the improvedalgorithm is applied to the optimization of methanol data, the simulation experiments showthat the improved PSO algorithm could optimize the methanol data better, and better guideproduction practice. |