Abstract: Traditional Particle Swarm Optimization（PSO） algorithm used in training the neural network has certain drawbacks such as high time-consuming, slow learning speed and easily falling into local optimum. In order to overcome these disadvantages, a new model is presented in this paper, which is generated from Takagi-Sugeno fuzzy neural network based on adaptive quantum-behaved particle. The new adaptive quantum-behaved particle swarm optimization algorithm can adaptively adjust the contraction-expansion coefficient in the iteration by introducing the concept of the aggregation degree in the algorithm, which makes the algorithm more dynamically adaptive. The new model combines the advantages of quantum-behaved particle swarm optimization with Takagi-Sugeno fuzzy neural network, and improves the generalization ability of the model. The hydrological data of Dongjiang Lake watershed stations in Hunan province of China from 2002 to 2013 are used in the experiments. The experimental results demonstrate that the proposed model is more efficient than other 4 neural network models and can be used to the daily water quality evaluation.

Key words: quantum-behaved particle swarm optimization, aggregation degree, contraction-expansion coefficient, fuzzy neural network, water quality evaluation

摘要： 传统的粒子群算法训练神经网络的水质评价模型有学习速度慢，容易陷入局部最优和精确性不高的缺点。为了克服模型的缺点，提出了利用改进的自适应量子粒子群算法训练T-S模糊神经网络的新模型，新的自适应量子粒子群算法通过在算法中引入聚集度的概念，使得算法可以在迭代中自适应地调整收缩扩张系数，让算法更具动态自适应性。新的模型结合了量子粒子群算法和T-S模糊神经网络的优点，提高了模型的泛化能力。通过对东江湖流域站点2002到2013年的水文数据进行实验，结果显示，该模型比其他神经网络模型的评价结果具有更高的效率，适合被用于日常水质评价工作。

关键词: 量子粒子群算法, 聚合度, 收缩扩张系数, 模糊神经网络, 水质评价