摘要
精确高效的用电需求预测在现代电网建设和资源利用分配上具有重要的研究意义。为了提高资源利用率和电力负荷预测精度,本文提出一种采用长短期记忆网络(LSTM)与支持向量回归机(SVR)相结合的组合算法对用户用电需求进行预测,并利用改进粒子群算法(PSO)对SVR参数进行寻优。以福建某地区的用电数据为例,验证了LSTM-SVR组合模型的可行性与有效性。实验结果表明,使用LSTM-SVR模型可以有效改进单一模型(LSTM、SVR)预测精度较低的缺点,相比单一模型在MAPE、RMSE和拟合系数R;上更优。
Accurate and efficient power demand forecasting has important research significance in modern power grid construction and resource utilization and allocation.In order to improve resource utilization and power load forecasting accuracy, an algorithm combining LSTM and SVR is proposed to predict user power demand, and the improved PSO is used to optimize SVR parameters.Taking the power consumption data of a region in Fujian as an example, the feasibility and effectiveness of lstm-svr combined forecasting model are verified.The experimental results show that the lSTM-SVR model can effectively improve the low prediction accuracy of the single model(LSTM,SVR),and is better than the single model in MAPE,RMSE and fitting coefficient R;.
作者
雷升
徐启峰
林穿
LEI Sheng;XU Qi-feng;LIN Chuan(College of Electrical Engineering and Automation,Fuzhou University,Fuzhou 350108,China)
出处
《电气开关》
2022年第2期13-18,共6页
Electric Switchgear
关键词
用电需求预测
深度学习
支持向量回归
长短期记忆网络
power demand forecasting
deep learning
support vector regression
long short-term memory
