基于NSPSO算法的微调模糊宽度学习系统用于非线性系统辨识

NSPSO Algorithm Based Fine-tuned Fuzzy Broad Learning System for Nonlinear System Identification

非线性系统的辨识涉及对对象的控制输入和输出之间的关系进行近似。其中,模糊宽度学习系统是一种新颖的且具有潜力的非线性系统辨识方法。然而,该方法在辨识过程中存在性能不足且超参数敏感的问题。为此,本文提出了一种基于节点敏感性粒子群算法的微调模糊宽度学习系统,用于求解非线性系统辨识问题。首先,本文在原始模糊宽度学习系统结构基础上使用稀疏自编码器对后件参数进行微调,构建了一种新型的微调模糊宽度学习系统。然后,提出了一种节点敏感性PSO算法对模糊宽度学习系统的模糊规则、TSK模糊子系统和增强节点三个参数进行搜索。使用模拟的非线性系统生成的数据验证模型的有效性,并利用经过节点敏感性PSO优化搜索的最优微调模糊宽度学习系统结构进行预测。该模型计算过程简单高效,能够同时用于神经网络逼近和模糊推理,具有快速确定最优模型结构的优点,可以有效提升模型的非线性系统辨识能力,在非线性系统辨识任务中具有较强的应用价值。

The identification of nonlinear systems involves approximating the relationship between the control inputs and outputs of an object.One of the novel and promising methods for recognizing nonlinear systems is the fuzzy broad learning system.However,the method suffers from insufficient performance and hyperparameter sensitivity in the identification process.For this reason,this paper proposes a fine-tuned fuzzy broad learning system based on a node-sensitive particle swarm algorithm for solving nonlinear system identification problems.Firstly,this paper constructs a novel fine-tuned fuzzy broad learning system based on the structure of the original fuzzy broad learning system using a sparse auto-encoder to fine-tune the posterior parameters.Then,a node-sensitive PSO algorithm is proposed to search the three parameters of the fuzzy rule,TSK fuzzy subsystem,and enhancement node of the fuzzy broad learning system.The validity of the model is verified using the data generated from the simulated nonlinear system.The optimal structure of the fine-tuned fuzzy broad learning system is predicted using searches by the node-sensitive PSO.The model is simple and efficient in the computational process,can be used for both neural network approximation and fuzzy inference and has the advantage of quickly determining the optimal model structure,which can effectively improve the model's ability to recognize nonlinear systems.It has strong application value in nonlinear system recognition tasks.