Electric load data characterising and forecasting based on trend index and auto-encoders

The Journal of Engineering

Huakun Que¹  Shixiang Lu²  Guoying Lin³ 

[1] College of Electrical Engineering, Zhejiang University , Hangzhou 310027 , Zhejiang Province , People'Electric Power Research Institute of Guangdong Power Grid Corporation , Guangzhou 510080 , Guangdong Province , People's Republic of China

关键词: load data volume;    stacked auto-encoders;    electric load data characterization;    trend-based method;    load description;    smart grids;    advanced market-based applications;    electricity consumption data;    stochastic oscillator;    high-quality meters;    moving average convergence;    trend-based load characterising approach;    dem;    response;    load service entities;    input historical trend indexes;   

DOI  :  10.1049/joe.2018.8350

学科分类：工程和技术（综合）

来源: IET

【 摘 要 】

Electricity consumption data are collected more frequently by high-quality meters in smart grids. Therefore, the load data volume and length increase dramatically. On the other hand, for advanced market-based applications, e.g. demand response, load service entities hope to identify or classify users better. In this study, a trend-based load characterising approach is proposed. Firstly, the concept of the candlestick chart is utilised as an innovative tool for load description. In addition, electricity trend indexes, e.g. stochastic oscillator and moving average convergence/divergence, are introduced as parameters for load characterising. Secondly, the stacked auto-encoders are utilised to forecast the future load based on the input historical trend indexes. Case studies in Guangdong province demonstrate that the proposed trend-based method is more applicable than existing approaches both in physical significance and accuracy.

【 授权许可】

CC BY