摘要
针对传统单一的预测方法精度不高,为进一步提高预测精度,提出VMD-SE与BiLSTM结合的形式建立预测模型,分析和利用气象参数、日期类型等辅助参数之间的相关性。首先对一段时间的用电负荷运用VMD分解,分解形成一系列子序列;然后运用样本熵对各个子序列分析其复杂度,并进行重构减少预测的规模;最后对得到的新的子序列与天气参数、日期类型通过BiLSTM模型进行预测,将得到各个子序列的预测值进行相加处理,再进行反归一化最终得到预测的结果。通过仿真验证并与其他模型进行比较,该模型的预测精度为98.63%,预测精度较高,具有较好的理论和实践意义。
In view of the low accuracy of the traditional single prediction method,a prediction model is built by a combination of VMD-SE and BiLSTM to further improve the prediction accuracy.The correlation between auxiliary parameters including meteorological parameters and date types are analyzed and used with this model.The VMD(variational mode decomposition)is used to decompose the electricity load of a period of time to form a series of sub-sequences.The sample entropy is used to analyze the complexity of each sub-sequence and carry out reconstruction to reduce the scale of prediction.The obtained new sub-sequences,weather parameters and date types are predicted by the BiLSTM model,and then the predicted values of each sub-sequence are added together for denormalization to finally get the predicted result.Simulation verification was performed.This model was compared with other models.The prediction accuracy of this model is higher(reaches 98.63%),which has good theoretical and practical significance.
作者
张栋栋
陈洁
李洋
ZHANG Dongdong;CHEN Jie;LI Yang(School of Electrical Engineering,Xinjiang University,Urumqi 830046,China)
出处
《现代电子技术》
2021年第23期155-159,共5页
Modern Electronics Technique
基金
新疆维吾尔自治区自然科学基金(2019D01C078)。
关键词
负荷预测
预测精度
变分模态分解
样本熵
复杂度
长短期记忆网络
归一化
反归一化
load forecasting
forecasting accuracy
VMD
sample entropy
complexity
long-and short-term memory network
normalization
denormalization
作者简介
张栋栋(1994-),硕士研究生,主要研究方向为风功率和负荷预测;陈洁(1975-),副教授,硕士研究生导师,主要研究方向为风力发电;李洋(1992-),硕士研究生,主要研究方向为风电储能平抑风功率。
