基于迁移学习的中长期电力负荷预测

Medium and long-term load forecasting based on transfer learning

由于时间跨度大并受多种复杂因素影响,电力系统中长期负荷预测需要从大量数据中挖掘负荷特性及影响因素,实际应用中常常面临数据不足的问题。提出基于迁移学习的中长期负荷预测以降低样本不足对预测精度的影响,模型将源地区电力负荷及社会经济因素数据样本进行迁移以扩充目标地区数据集,通过隐变量描述不同源地区的特征,继而对目标地区和源地区建立集成模型进行预测。通过实际算例进行验证,所提出的模型能有效降低中长期负荷预测的误差。

Due to long time span and various complex factors,it is necessary to unearth the load characteristics and influencing factors from massive data for medium and long-term load forecasting,while there is a lack of available data in practice.In view of this,this paper proposes a medium and long-term load forecasting model based on transfer learning to reduce the impact of sample shortage on accuracy.The proposed model migrates the load data and socio-economic data of the source area to expand the samples of the target area,describes the characteristics of different areas’databy hidden variable and then establishes an ensemble model to predict for the target area and the source areas.Experiments show that the proposed model can effectively reduce the prediction error of medium and long-term load forecasting.