A Short-Term Load Forecasting Scheme Based on Auto-Encoder and Random Forest
A Short-Term Load Forecasting Scheme Based on Auto-Encoder and Random Forest
复制标题
基于自动编码器和随机森林的短期负荷预测方案
DOI：
10.1007/978-3-030-21507-1_21
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
Eenjun Hwang
中科院分区：
文献类型：
--
作者：
Minjae Son;Jihoon Moon;Seung‐Won Jung;Eenjun Hwang
近年来，智能电网作为解决电力短缺问题的可行方案备受关注。准确预测短期电力负荷是提高其运行效率的关键问题之一。到目前为止，使用各种机器学习算法构建STLF（短期负荷预测）模型已经做了很多工作。通过考虑许多有影响的变量，他们给出了令人满意的结果预测整体电力负荷模式。但是，他们仍然缺乏预测微小的电力负荷模式。为了克服这一问题，本文提出了一种新的STLF模型，该模型结合了基于Auto-Encoder （AE）的特征提取和随机森林（Random Forest， RF），并通过对不同类型建筑集群的实际功耗数据进行了多次实验，展示了其性能。
Recently, the smart grid has been attracting much attention as a viable solution to the power shortage problem. One of critical issues for improving its operational efficiency is to predict the short-term electric load accurately. So far, many works have been done to construct STLF (Short-Term Load Forecasting) models using a variety of machine learning algorithms. By taking many influential variables into account, they gave satisfactory results in predicting overall electric load pattern. But, they are still lacking in predicting minute electric load patterns. To overcome this problem, in this paper, we propose a new STLF model that combines Auto-Encoder (AE) based feature extraction and Random Forest (RF) and show its performance by carrying out several experiments for the actual power consumption data collected from diverse types of building clusters.