面向函数型数据的快速特征选择方法*

函数型数据的特征选择是从庞大的函数信息中选出那些相关性小、代表性强的少部分特征,以简化后期分类器的计算,提高泛化能力.由于特征选择方法用于函数数据分类效果并不理想,文中提出面向函数型数据的结合主成分分析法和最小凸包法的快速特征选择(FFS)方法,可以快速获得稳定的特征子集.此外,考虑到特征之间可能存在相关性,将FFS的结果作为其它方法的初始特征子集,故融合FFS与条件互信息方法.在UCR数据集上的实验证明FFS的有效性,并通过对比实验给出在不同时间代价和分类精度需求下的方法选择策略.     

计及复杂气象影响的含光伏电源的母线峰值负荷预测

光伏电源并网后,配电网母线负荷波动会更复杂,峰值负荷预测更加困难。为提高母线峰值负荷预测精度,文章提出了计及复杂气象影响的母线峰值负荷预测方法。首先,根据不同气象日下累积的历史数据,通过条件互信息分析母线峰荷数据与高维气象、社会等特征间相关性,获得特征重要度排序;其次,在条件互信息降低潜在特征集合特征间冗余性基础上,针对不同气象日,以改进粒子群优化极限学习机预测精度为决策变量,开展针对性前向特征选择,确定不同最优特征子集;最后,根据最优特征子集,针对性建立不同气象日下母线峰值负荷最优预测模型。以某地区实际含高渗透率光伏电源母线负荷开展实验,证明所提方法可有效提高母线峰荷预测精度。     

羧甲基菊粉的制备及其抗氧化活性研究

采用氢氧化钠-氯乙酸反应体系,以异丙醇作为溶剂制备羧甲基菊粉（CMI）,并研究其抗氧化活性。以取代度为指标,通过单因素及正交实验优化CMI制备条件,并对其·OH清除活性和红外光谱进行研究。结果表明,CMI制备的最佳条件为碱化时间3min,氯乙酸用量6.75g,异丙醇用量4mL,醚化时间15min,该条件下取代度为1.3327。抗氧化活性研究表明,CMI具有一定的·OH清除能力,且与浓度呈正相关效应,但与取代度之间却无确定关系。红外光谱分析表明CMI的制备工艺可行,应用前景广阔。     

Driving control of an AGV for an automated container terminal using an immunized PID controller based on cell-mediated immunity

In this article, an immunized PID (I-PID) controller using a cell-mediated immune algorithm (CMIA) is proposed to control an autonomous guided vehicle (AGV) more effectively. The proposed controller is based on the specific immune response of the biological immune system that is cell-mediated immunity. Moreover, we designed and manufactured an AGV system for an automated container terminal (ACT), and some experiments are shown to verify the performance of the proposed controller. The tracking error of the AGV driving was investigated for this purpose. As a result, the capabilities of realization and reliability were proved by comparing the response characteristics of the proposed I-PID controllers with those of conventional PID and neural network PID (NNPID) controllers.This work was presented, in part, at the 9th International Symposium on Artificial Life and Robotics, Oita, Japan, January 28–30, 2004