Research And Application Of Improved Ant Colony Clustering Algorithm To Optimize The RBF Neural Network

Radial basis function (RBF) neural network has simple structure, and can approximate any nonlinear system, etc, make it become a kind of effective tool which is not dependent on the control object model, suitable to control this class of nonlinear uncertain systems such as the water treatment system. However, the shortcomings of RBF neural network is also existing, on the one hand, setting of initial parameters of RBF neural network model has a vital role. On the other hand traditional RBF neural network learning algorithm, such as K-means algorithm depends on the given initial cluster number and cluster centers of sample data vector, and K-means algorithm is only a cursory search process and easy to fall into the local minimum value, this finally has directly affected the training results of the RBF neural network. This paper will adopt the improved ant colony clustering algorithm to optimize the k-means algorithm, and use the optimized learning algorithm to train the RBF neural network, then will apply the based on optimization learning algorithm of RBF neural network PID controller to control the water PH value, the experimental results show the feasibility and effectiveness of this control method.At first, this paper introduces the research of the RBF neural network and ant clustering colony algorithm, the model principle of RBF neural network, and its advantages or disadvantages, and the basic thinking and implementation steps of the k-means learning algorithms. Then focus on to introduce two class under the principle of Ant Colony Clustering Algorithm and BM model and AM model, then based on AM model to use an Improved Ant Colony Clustering Algorithm (Improved Ant Colony Clustering, IACC). It improves the AAC algorithm which based on the AM model, adds an "off"state in the original based on the model of the AM model, reducing the consumed time which because of the artificial ants couldn’t find a suitable place to sleep and blindly walk, speed up the speed of the cluster formation, improve the quality of the clustering. Through the three data sets of simulation experiment proves that the fast clustering and clustering quality good characteristic of the IACC algorithm. Use IACC algorithm to optimize the K-means learning algorithm for RBF neural network, and use the optimized K-means learning algorithm of RBF neural network to model, based on the nonlinear function fitting and approximation optimization experiment to verify the superiority of RBF neural network. Final design based on optimized RBF neural network PID controller and use it control the water treatment PH value control system, through the SIMULINK module of MATLAB simulation, the simulation results show that the optimization of the RBF-PID controller in robustness and steady-state accuracy, anti-jamming, etc have obvious advantages, not only can make the system has smaller overshoot, PH value reached the national standard but also short response time to reach steady state speed and have the high controlling precision, good self-healing capability.