基于区域划分的分布式短期负荷预测方案设计

Design of Distributed Short-term Load Forecasting Scheme Based on Region Division

针对电网台区分布特征,本文提出基于负荷分解的台区大数据负荷预测方案,首先结合Hadoop集群构建了面向用电信息采集系统大数据的处理架构和计算服务架构,其次使用BP神经网络进行台区停电时刻负荷预测,最后基于Map/Reduce函数式程序处理大数据的"分而治之"思想将大区域台区负荷进行分层分区,在数据分析基础上建立负荷预测模型,设计台区负荷预测方案。测试结果表明:该方法可实现大区内负荷精细化预测,与传统的系统负荷预测方式相比,实例证明该方法具有明显优势。

Against the distribution characteristics of the grid station, this paper proposes a zone large data load lorecasting scheme based on load decomposition. Firstly, the Hadoop cluster is used to construct the processing architecture and computing service architecture for the large data of the electricity-consumption information collection system, secondly, BP neural network is used to forecast the load in the zone at the power outage time,finally,based on the idea of ＂divide and conquer＂ which is used by Map/Reduce function program to deal with large data, large regional zone load is hierarchical partition, the load forecasting model is established on the basis of data analysis, design of zone load forecasting scheme. The test results show： this method can realize the fine load forecasting in the large region, compared with the traditional system load forecasting method, the example proves that the methdod has obvious advantages.