Research On Particle Swarm Optimization And Application For Hydraulic Control System

In actual engineering control process,we need to set the controlling parameters by the actual system's ones,and further to adjust these parameters to realise the stability,accuracy and fast speed.At this time,the whole control process has the discrete,non differentiable performance,and the objective function is also showing the characteristics of the game,which set up obstacles to find the optimal parameter configuration.Therefore,it is necessary to find a suitable method to solve the optimization problem of this kind of control system.On the basis of some current particle swarm optimization(PSO),the authors put forward some improved particle swarm optimization algorithms.Aiming at the traditional PSO's problems of premature and slow convergence speed in the later stage,and later most of the improved PSOs' problem of restricting the function,we put forward a various interval chaotic search algorithm which judges particle swarm's global optimal location by the updating state to achieve the switch of different search methods.Some search methods such as cloud search(CS),gradient search(AGS)and piecewise chaotic search(PCS)are combined to obtain a better performance of the algorithm,called mul-ti methods argument particle swarm optimization(MMAPSO).Among them,AGS and CS are used for local search behaviours,and PCS is used for global search behaviour.The three meth-ods can improve the convergence speed of the algorithm,and also have some ability to jump out of the local optima.For different function forms,MMAPSO can adopt the corresponding search strategy to adapt to the convergence.Inspired by the learning behaviour of human beings,the authors propose a classification particle swarm optimization algorithm,which divides the learning community into three cate-gories,each of which uses different learning strategies and directions.A new algorithm is proposed to increase the convergence speed of the traditional compre-hensive learning particle swarm optimization(CLPSO),called lagrange interpolation learning particle swarm optimization algorithm(LILPSO).First of all,according to the best position of each iteration(gbest),the concept of a Lagrange interpolation is introduced to accelerate the convergence of local search.Second,this method is used to replace the traditional CLPSO learning methods,to get a better learning object.A local optimum topology structure is proposed,and it is introduced into the CLPSO al-gorithm,which is called CLPSO-LOT.First,to sort them according to the the rank of particle's position,secondly to find the optima according to their function fitness,and then to generate a extension topology structure.Finally,to selected elements from the topological structure ran-domly to constitute a particle of learning objects.In contrast with CLPSO,this algorithm has a smaller population diversity in the early stage,and a larger search space,so that it can increase the efficiency of search.After comparative analysis on several algorithm,and application on the hydraulic straight-ening machine PID control parameter optimization,and rolling mill AGC synovial control sys-tem parameter optimization,we obtain the better optimization results,among them,LILPSO are best suited for this kind of problem.This illustrates that it can be the ideal algorithm to combin the Lagrange interpolation search method with a mutation to learn each other.