论文部分内容阅读
针对在线数据校正效果差的问题,提出了一种基于改进万有引力和支持向量机的数据校正方法.首先为了减小计算量,对万有引力算法中的适应度函数进行改进,利用改进的万有引力算法对影响支持向量机性能的重要参数进行优化.然后利用少数准确的离线试验数据对支持向量机回归模型进行训练,当在线监测的历史或实时数据不在回归模型允许偏差范围内时,通过回归模型对异常数据进行校正.最后通过实际数据对提出的方法进行验证,结果证明了该方法的可行性和有效性.
Aiming at the problem of poor correction of on-line data, a data correction method based on improved gravitation and support vector machine is proposed.Firstly, in order to reduce the computational complexity, the fitness function in gravitation algorithm is improved, and the improved gravitation algorithm Support vector machine performance optimization of the important parameters.And then use a small number of off-line experimental data to support SVM regression model training, when the online monitoring of historical or real-time data is not within the allowable deviation of the regression model, the regression model of abnormal data Finally, the proposed method is validated by the actual data, and the result proves the feasibility and effectiveness of this method.