摘要
短期电网负荷表现出非规律的波动性和非稳定的周期性,对其进行准确预测是一项挑战。采用时变滤波的经验模态分解,对一维观测信号进行分解,得到具有不同尺度特征的固有模态函数。通过随机重构构造一个新的观测值,形成二维矩阵。引入支持向量机来替代门控循环单元最终输出层中的归一化指数函数,并将交叉熵函数替换为基于边缘的函数,从而进行基于混合模型的短期电网负荷预测。试验结果表明,与大间隔最近邻算法、卷积神经网络以及融合门控循环单元的支持向量机相比,混合模型的计算成本虽然稍高,但均方根误差和平均绝对误差都是最小的。因此,混合模型具有最好的预测性能,可用于短期电网负荷预测。
It is a challenge to predict the short-term power grid load accurately because of its irregular fluctuations and unstable periodicity.Empirical mode decomposition(EMD)using time-varying filtering(TVF)was applied to decompose one-dimensional observation signals,and the intrinsic mode function(IMF)with different scale features was obtained.A new observation value was constructed through random reconstruction to form a two-dimensional matrix.The support vector machine was introduced to replace the Softmax in the final output layer of gated loop unit,and the cross-entropy function was replaced by the edge-based function,thus achieving short-term power grid load forecasting based on hybrid models.The experimental results show that compared with the large interval nearest neighbor algorithm,convolutional neural network,and support vector machine(SVM)with fused gated recurrent unit(GRU),the hybrid model has slightly higher computational costs,but the root-mean-square error and mean absolute error are the smallest.Therefore,the TVF-EMD-SVM-GRU hybrid model has the best predictive performance and can be used for short-term power grid load forecasting.
作者
唐啸
项诗娴
房宇娇
王钰楠
虞江
Tang Xiao;Xiang Shixian;Fang Yujiao;Wang Yunan;Yu Jiang(Shinan Power Supply Branch,State Grid Shanghai Electric Power Co.,Ltd.,Shanghai 201109,China)
出处
《电气自动化》
2024年第5期31-33,37,共4页
Electrical Automation
基金
国网上海市电力公司科技项目(B30932220002)。
关键词
短期电网负荷预测
时变滤波的经验模态分解
支持向量机
门控循环单元
混合模型
short-term power grid load forecasting
time-varying filtered empirical mode decomposition
support vector machine
gate recurrent unit
hybrid model
