基于信息融合的电力大数据可视化预处理方法

Visualization Pretreatment Method for Electric Power Big Data Based on Information Fusion

在研究电力大数据基本特征的基础上,分析了电力大数据可视化预处理中的效率与性能问题。通过引入信息融合方法,针对电力大数据"数量大"、"类型多"等特征,采用基于Map-Reduce与神经网络相结合的大数据集样例分类融合模型,解决大数据在结构化、半结构化、非结构化并存的异构数据融合问题。其中,利用神经网络的聚类和非线性映射能力,选用改进的Hermite正交多项式激励前向神经网络模型,结合Map-Reduce模型实现大规模数据集的并行处理,将属于同一对象的异源、异构数据融合在统一的描述视图中,完成多源异构数据的分类融合。最后,通过Hadoop平台进行了算例仿真。

On the basis of studying fundamental features of electric power big data, this paper analyzes efficiency and per- formance in visualization pretreatment for electric power big data. By introducing information fusion method and in allusion to features of large quantities and various types of electric power big data, it uses big data example classification and fusion model based on combination of Map-Reduce and neural network to solve the problem of heterogeneous data fusion in coexis- tence of structuring, semi-structuring and non-structuring of big data. By using clustering and non-linear mapping capability of neural network and forward neural network model based on improved Hermite orthogonal polynomial, and combining Map-Reduce model, it is able to realize parallel processing for large-scale dataset and blend heterogeneous data of the same target into a unified description view so as to finish classification and fusion for multi-source heterogeneous data. Finally, ex- ample simulation is finished on Hadoop platform.