Wind power prediction method and system for optimizing deep transformer network

Patent metadata

Field	Value
Publication number	US-11874640-B2
Application number	US-202117496778-A
Country	US
Kind code	B2
Filing date	Oct 8, 2021
Priority date	Dec 18, 2020
Publication date	Jan 16, 2024
Grant date	Jan 16, 2024

Abstract

A wind power prediction method and system for optimizing a deep Transformer network by whale optimization algorithm are disclosed. The sequence data of wind power and related influence factors are taken as sample data which is divided into a training set and a test set, where the data is trained and predicted by a Transformer network model established according to values of the initialized hyper-parameters, and an average absolute error of wind power prediction is taken as a fitness value of each whale group. A local optimal position is determined according to the initial fitness value of individual whale group, and the current optimal position is updated by utilizing whale group optimization, and the best prediction effect is obtained by comparing the local optimal solution with the global optimal solution. An optimal hyper-parameter combination is obtained after multiple iterations of the whale optimization algorithm, and the wind power is predicted.

First claim

What is claimed is:

1. A wind power prediction method for optimizing a deep Transformer network, comprising:
taking a collected sequence data of wind power and related influencing factors as a sample data;
performing maximum and minimum normalization processing on all of the sample data, and dividing the normalized sample data into a training set and a test set;
initializing Transformer network parameters, setting a value range and a search range of hyper-parameters in the Transformer network to be optimized, and determining a maximum number of iteration and a population size of a whale group;
establishing a Transformer network model according to the values of the hyper-parameters in the initialized Transformer network, respectively training and predicting the data in the training set and the test set, and taking an average absolute error of wind power prediction as a fitness value of each of the whale groups;
determining a local optimal position according to the initial fitness value of the individual whale group, updating a current optimal position by utilising whale group optimization, and obtaining an optimal prediction effect by comparing a local optimal solution with a global optimal solution;
obtaining an optimal hyper-parameter combination in the Transformer network after a plurality of iterations of the whale optimization algorithm WOA, and predicting wind power by using the optimal parameters to construct a WOA-Transformer wind power prediction model,
wherein the step of obtaining the optimal hyper-parameter combination in the Transformer network after the plurality of iterations of the whale optimization algorithm WOA comprises:
taking the hyper-parameters in the Transformer network as individuals in the whale group, initializing the whale group, and using a random number generator to automatically generate an initial solution of the hyper-parameters in the Transformer network;
in response to a value of a random parameter p being less than a first preset value, then it is determined whether a coefficient vector |A| is less than a second preset value, and in response to the coefficient vector |A| being less than the second preset value, then a shrink-envelopment predation mechanism is selected, a position of the individual is updated according to A=2a·r−a,C=2r and

a
=

2
-

2
⁢
j

M

,

in response to the coefficient vector |A| being not less than the second preset value, then a search and predation mechanism is selected, and the position of the individual is updated according to

{

X
=

X

r
⁢
a
⁢
n
⁢
d

-

A
·
D

D
=

❘
"\[LeftBracketingBar]"

C
·

X

r
⁢
and

,
j

-
X

❘
"\[RightBracketingBar]"

,

in the expression, a is a constant corresponding to a current number of iterations j, and a is a matrix that is formed by the constant a during the iterations, M is the maximum number of iterations, r is a random vector and r∈[0,1], X rand represents a current random position vector of a humpback whale population, X rand,i represents the j-th data in X rand ;
in response to the value of the random parameter p being not less than the first preset value, then a spiral predation mechanism is selected, and the position of the individual is updated according to

{

X

j
+
1

=

D
′

⁢

e

b
⁢
l

⁢

cos
⁡
(

2
⁢
π
⁢
l

)

+

X
j
*

D
′

=

❘
"\[LeftBracketingBar]"

X
j
*

Assignees

• Univ Wuhan

Inventors

• He Yigang
• Wang Lei
• Zhao Yingying
• Xiang Ming
• Li Lie
• He Liulu

Classifications

• H02J2103/30

  Simulating, planning, modelling, reliability check or computer assisted design [CAD] of electric power networks · CPC title

• H02J2101/28

  Wind energy · CPC title

• G05B19/042 Primary

  using digital processors (G05B19/05 takes precedence) · CPC title

• H02J3/004 Primary

  Generation forecast, e.g. methods or systems for forecasting future energy generation · CPC title

• H02J3/381

  Dispersed generators · CPC title