引用本文: | 陈光宇,黄越辉,张仰飞,郝思鹏,张友泉,吕干云.历史数据驱动下基于粗糙集的AVC系统关键参数挖掘方法[J].电力自动化设备,2020,40(6): |
| CHEN Guangyu,HUANG Yuehui,ZHANG Yangfei,HAO Sipeng,ZHANG Youquan,Lü Ganyun.Key parameter mining method of AVC system based on rough set driven by historical data[J].Electric Power Automation Equipment,2020,40(6): |
|
摘要: |
自动电压控制(AVC)系统中的参数设置过程繁琐且设置结果无依据,以历史大数据为基础,通过对历史数据的挖掘指导系统关键参数的设置。首先,给出一种基于强化正域的属性综合约简策略对关联属性进行约简;然后,采用基于最优分类的属性变换策略将连续属性离散化,并给出一种基于数据预处理的集合近似匹配策略,用于计算不同曲线间的相似度;最后,提出一种基于粗糙集的AVC系统关键参数辨识框架,对历史大数据进行挖掘。基于真实电网数据进行算例分析,挖掘结果表明所提辨识框架能自动给出合理的参数设置结果;实际应用结果表明,相比于传统方法,基于历史大数据的挖掘结果取得了更好的控制效果。 |
关键词: AVC 粗糙集 数据挖掘 数据预处理 相似度量 数据驱动 |
DOI:10.16081/j.epae.202006026 |
分类号:TM732 |
基金项目:新能源与储能运行控制国家重点实验室开放基金资助项目(NYB51201901205);国家自然科学基金资助项目(51607086);江苏省高校自然科学研究重大项目(17KJA-470003);江苏省配电网智能技术与装备协同创新中心开放基金资助项目 |
|
Key parameter mining method of AVC system based on rough set driven by historical data |
CHEN Guangyu1, HUANG Yuehui2, ZHANG Yangfei1, HAO Sipeng1, ZHANG Youquan3, Lü Ganyun1
|
1.School of Electric Power Engineering, Nanjing Institute of Technology, Nanjing 211167, China;2.State Key Laboratory of Operation and Control of Renewable Energy & Storage Systems, China Electric Power Research Institute, Beijing 100192, China;3.State Grid Shandong Electric Power Co.,Ltd.,Jinan 250001, China
|
Abstract: |
The parameter setting process in the AVC(Automatic Voltage Control) system is tedious and the setting results are unsubstantiated. Based on the historical big data, the setting of system key parameters is guided by mining the historical data. Firstly, an attribute synthesis reduction strategy based on enhanced positive domain is presented to reduce the associated attributes. Then, the optimal classification-based attribute transformation strategy is adopted to discretize the continuous attributes and a set approximate matching strategy based on data preprocessing is presented to calculate the similarity between different curves. Finally, the rough set-based key parameter identification framework of AVC system is proposed for mining the historical big data. Based on real power grid data, the mining results show that the proposed identification framework can automatically provide reasonable parameter setting results. The practical application results show that, compared with the traditional method, the mining results based on historical big data have better control effect. |
Key words: AVC rough set data mining data preprocessing similarity degree measurement data-driven |