基于组合混沌策略自适应量子微粒群的Volterra核辨识算法

Volterra series identification method based on adaptive quantumbehaved particle swarm optimization combined with the chaotic strategy

针对非线性Volterra泛函级数的参数辨识问题,提出了一种基于组合混沌策略自适应量子微粒群算法(CCSAQPSO算法)的Volterra时域核辨识方法.该方法在量子微粒群算法(QPSO)的基础上,采用混沌策略分两个阶段对QPSO进行优化,在初始化时以混沌序列初始化种群,在搜索过程中则引入混沌变异机制,利用混沌变异算子空间遍历特性对个体进行变异操作,同时按照各微粒适应度的优劣程度对其进化过程中的收缩扩张系数进行自适应调节,有效避免了早熟收敛现象的发生,提高了算法的全局寻优能力,保证了算法的准确性和精度.最后将该Volterra核辨识方法与基于标准微粒群算法(PSO算法)和QPSO算法的Volterra核辨识方法进行了对比分析.仿真结果表明,提出的方法具有参数辨识精度高、抗噪声能力强等优点,且在全局优化能力和快速收敛能力上都有较大提高.

A Volterra kernel identification method based on Adaptive Quantum-behaved Particle Swarm Opti-mization with the combination of the chaotic strategy （CCSAQPSO） was proposed for parameter identification of nonlinear Volterra series. The chaotic strategy was adopted to optimize the Quantum-behaved Particle Swarm Optimization （QPSO） algorithm in the two stages of search process, the population was initialized with chaotic sequences during initialization, and chaos mutation mechanism was used in the search process. The ergodicity of the chaotic mutation operator was used to finish the mutation operation of the selected individual. At the same time, the contraction expansion coefficient of the algorithm was adjusted adaptively in the evolutionary process according to the fitness of each particle, so that the premature convergence was efficiently avoided, the global convergence ability of the algorithm was improved, and the accuracy and precision of the algorithm was guaranteed. The proposed method was compared with Volterra kernel identification methods based on standard Particle Swarm Optimization （PSO） and QPSO, and the simulation results showed that the proposed method has some advantages, such as high identification accuracy, good anti-noise performance, etc. The global conver-gence ability and the convergence speed of the proposed method were improved greatly.