基于遗传优化聚类的GRU无损电力监测数据压缩OA北大核心CSTPCD
GRU Neural Network Lossless Compression of Power Monitoring Data Based on Genetic Optimization Clustering
针对电力调度中心监测数据记录体量大、存储困难的问题,提出基于遗传优化K-means聚类的门控循环单元神经网络无损数据压缩方法.首先,搭建分布式集群,将多维原始电力数据聚类成相似性较高的数据块,并利用遗传算法对聚类进行寻优,提高数据聚类的效果;再通过门控循环单元神经网络训练数据编码的概率分布模型,结合算术编码对数据进行编码压缩;最后,以多个电力数据集为算例进行分析.经验证本文所提的压缩算法能实现数据的高比例压缩、优化集群性能.
Aimed at the problems of large volume and difficult storage of monitoring data records at power dispatching centers,a gated recurrent unit(GRU)neural network lossless data compression method based on genetic optimization K-means clustering is proposed.First,a distributed cluster is built to cluster the multi-dimensional raw power data into data blocks with a high similarity,in which the genetic algorithm is used to find the best cluster and improve the effect of data clustering.Then,the probability distribution model of data coding is trained by the GRU neural network,and the data is coded and compressed by combining with arithmetic coding.Finally,several power datasets are analyzed as examples to show that the proposed compression algorithm can achieve high proportional compression of data and opti-mize the clustering performance.
屈志坚;帅诚鹏;吴广龙;梁家敏;李迪
华东交通大学轨道交通基础设施性能监测与保障国家重点实验室,南昌 330013||华东交通大学电气与自动化工程学院,南昌 330013
动力与电气工程
电力数据遗传算法聚类分析循环神经网络分布式集群压缩
power datagenetic algorithmclustering analysisrecurrent neural network(RNN)distributed cluster compression
《电力系统及其自动化学报》 2024 (004)
1-8,18 / 9
江西省自然科学基金重点项目(20232ACB204025);江西省高层次高技能领军人才培养工程资助项目(202223323);轨道交通基础设施性能监测与保障国家重点实验室开放课题资助项目(HJGZ2022203)
评论