基于改进多种群遗传算法的尾矿坝形变预测

针对遗传神经网络(GA-BP)建立的尾矿坝形变预测模型易出现早熟现象、预测结果不稳定、容易陷入局部最优值的不足,引入一种具有混沌局部搜索的多种群自适应遗传算法。该算法以双种群寻优为基础,改进了遗传参数的计算方式,分别以种群进化中染色体适应度值的集中程度和空间距离的分布作为自适应交叉率、变异率的计算依据应用于不同种群中,提高了种群的多样性和遗传算法全局搜索的能力;同时引入混沌局部搜索技术(CLS),完善了遗传算法局部搜索能力的不足。采用改进的遗传神经网络模型对贵州省白岩尾矿坝三维变形数据进行预测,并与传统的GA-BP和AGA-BP模型预测结果进行比较。结果表明:改进后的模型预测精度更高,结果更加稳定,具有良好的预测效果。

Deformation Prediction of Tailings Dam Based on Improved Multi-population Genetic Algorithm

As the traditional GA-BP algorithm has low efficiency, instability and with failure of local search, a multi-group adaptive genetic algorithm with chaotic local search is proposed to improve the inadequacy of tailings dam deformation model established by GA-BP algorithm. Based on the double populations genetic algorithm, the calculation method of p c and p m is improved by the similarity of population fitness value and the spatial distance distribution of the individuals in population, and the diversity of the population and the global search ability of genetic algorithm are improved. And the CLS algorithm is introduced to improve the local search ability of genetic algorithm. The improved genetic neural network model is used to predict the 3D deformation data of the white rock tailings dam in Guizhou Province, and compared with the prediction results of the traditional GA-BP and AGA-BP models. The results show that the improved model has higher prediction accuracy, more stable results and better prediction effect.