16、Short-term electricity load forecasting method based on multilayered self-normalizing GRU network

多层自规范GRU

一种多层自正规门控递归算法

提出了短期电力负荷的单元(MS-GRU)模型。

预测

该模型引入比例指数线性单元。

(SELU)激活函数将隐藏状态压缩为

计算模型的输出。挤压状态

也有助于更新门、复位门的计算

GRU的候选状态,爆炸和消失

堆叠GRU神经网络可以克服梯度问题

网络采用这种自规范化方法。模糊聚类方法

(FCM)算法用于相似日的选择。

电力负荷。

 

传统的电力负荷预测方法包括:

回归分析[5-6],趋势外推[7-10],专家

系统〔11〕和时间序列法[12-14]等。

主要应用回归和趋势外推方法。

 A new

activation function called Scaled Exponential Linear Units

(SELU) proposed by Gunter Klambauer et al. [30] is used to

avoid exploding and vanishing gradients.This activation

function with self-normalizing properties will converge

towards zero mean and unit variance. even under the presence

of noise and perturbations.

模糊聚类相似日

FCM [31] is selected to conduct similarity analysis. It is a

soft algorithm fuzzy data in which an object is not only a

member of a cluster but member of many clusters in varying

degree of membership as well

缺失值异常值处理

 

最规范化GRU

In the traditional GRU architecture, there is no nonlinearity

transform for the output. However. when the layers of GRU

increase, there will be an exploding and vanishing gradient

problem. In order to overcome this. SELU[30] is introduced to

be an activation function for the output s

电力负荷预测如下:

1。选择相似的日子。用FCM确定相似日

集群。

2。数据标准化。以相似日负荷为输入数据,

并将输入数据标准化到[0,1]的范围内。

Nun max方法。

三。预测。使用MS-GRU进行输出

标准化投入。逆输出标准化

形式获得电力负荷。

R

猜你喜欢

转载自blog.csdn.net/qq_34514046/article/details/82811107