Elsevier

Applied Energy

Volume 238, 15 March 2019, Pages 249-257
Applied Energy

Scene learning: Deep convolutional networks for wind power prediction by embedding turbines into grid space

https://doi.org/10.1016/j.apenergy.2019.01.010Get rights and content

Highlights

  • The spatio-temporal feature is proposed, instead of traditional feature, time series.

  • Convolutional network is used to predict wind power based on spatio-temporal feature.

  • Much higher accuracy is achieved within much less training time than existing works.

Abstract

Wind power prediction is of vital importance in wind power utilization. There have been a lot of researches based on the time series of the wind power or speed. But in fact, these time series cannot express the temporal and spatial changes of wind, which fundamentally hinders the advance of wind power prediction. In this paper, a new kind of feature that can describe the process of temporal and spatial variation is proposed, namely, spatio-temporal feature. We first map the data collected at each moment from the wind turbines to the plane to form the state map, namely, the scene, according to the relative positions. The scene time series over a period of time is a multi-channel image, i.e. the spatio-temporal feature. Based on the spatio-temporal features, the deep convolutional network is applied to predict the wind power, achieving a far better accuracy than the existing methods. Compared with the state-of-the-art methods, the mean-square error in our method is reduced by 49.83%, and the average time cost for training models can be shortened by a factor of more than 150.

Introduction

Wind power has become a significant renewable resource that can be developed and utilized on a large scale [1]. Thanks to the mass production of equipment, wind power has turned to be the fastest growing renewable energy in the world. By 2017, the worldwide wind power installed capacity has reached 539 GW, and 52 GW was added in 2017 [2], thus making wind power expected to be one of the major power sources in the 21st century. However, due to the influence of wind speed and direction, randomness and volatility of wind turbines can not be avoided, bringing severe challenges to the safety and stability of the operation of power systems [3]. Accurate wind power prediction can enhance the controllability of wind power, ensure the stable operation of the power grid, and promote the ability of the grid to accept wind power.

Smart grid [4] is a topic of great concern in recent years [5], and wind power forecasting technology is conducive to smart grid. At present, scholars have done a lot of related researches, including physical methods [6], statistical methods [7] and machine learning methods. Among them, machine learning methods, including support vector machine regression (SVR) [8], k-nearest neighbor regression (kNN) [9] or multi-layer perceptron neural network (MLP) [10] are used to model wind speed time series or power time series to achieve prediction. Machine learning methods simplify the wind power forecasting problem, but the accuracy rate has failed to be improved in the past several years.

We think that the wind is temporal and spatial correlation process, however, the time series can only express the information at the time level, but say nothing at the space level, let alone the spatio-temporal process of air flow, thus fundamentally standing in the way of the progress of wind power prediction. Therefore, finding the features that can better express the state of the wind farm is the key to breaking through the bottleneck of accuracy.

Such being the case, this paper put forward a new feature that can express the spatio-temporal process of air flow, called spatio-temporal feature (STF). The scene time series over a period of time is a multi-channel image, in which each scene is a sample of the true distribution of physical data in space, expressing spatial-related information, as shown in Fig. 1. The scene sequence represents the change of wind farm state over time, expressing time information, so the multi-channel image is called spatio-temporal feature. Compared with wind speed or power series, STF implies factors such as wind speed, wind direction and air density, which greatly expands the ability to express wind-related information and lays a foundation for breaking through the bottleneck of wind power prediction accuracy.

Based on the STF, the spatio-temporal process of the wind farm is simulated and predicted by using the deep convolutional network, which has achieved good effects. The experimental results on two wind farms with 592 wind turbines (farm1) and 454 wind turbines (farm2) respectively show that, the proposed methods are better than the existing state-of-the-art series modeling methods, for the reason that the MSE of the proposed method decreases by an average of 26.69% and 49.83% at most in wind farm1, and by an average of 24.37% and 46.94% at most in wind farm2, and the time for training models is both shortened by more than 150 times.

The innovations of this paper are as follows:

  • The spatio-temporal feature in the form of the multichannel image is proposed for the first time by embedding wind turbines into the grid space, which fully expresses the spatio-temporal variation process of the air flow and can perfectly combine with the most advanced theory of deep learning at present.

  • The convolutional neural network is reasonably used to predict wind power for the first time based on the spatio-temporal feature, it can predict the wind power of a large number of turbines in parallel. And both the accuracy and time cost of the prediction have been greatly optimized.

Section snippets

Machine learning methods in WPP

Machine learning methods perform well in short term prediction. By means of the regression model or neural network, researchers map the time series to the wind power of the future moment, so as to make the prediction. The commonly used methods are SVR [8], kNN [9], Multilayer Perceptron Network (MLP) [10] and its variant [11] and Long and Short Term Memory Neural Network (LSTM) [12], etc., among which SVR and kNN are the representatives [13].

SVR has a perfect mathematical foundation in theory

Proposed method

The information related to wind such as wind speed or turbine’s output power can be strongly combined with convolutional networks. On one hand, convolutional networks are quite suitable to deal with data in tensor form, which can automatically extract features at different layers and realize the end-to-end learning. On the other hand, the wind turbines distributed on the plane is easy to be embedded into the grid to construct a two-dimensional tensor, namely, a matrix or grid. But the current

Data sets and evaluation criteria

The data set used in this paper is the wind data set from the NREL,1 which contains the output values of every 10 min of wind turbines in the United States from 2004 to 2006. To validate our method, two wind farms are selected. The longitude of wind farm1 is range from 105.00W to 105.34W and latitude is range from 41.40N to 41.90N (17.3km×38.7km), it is located in the central United States. Wind turbines are densely distributed reaching a number of 592. Wind farm2 contains

Conclusion

This paper proposes spatio-temporal feature for wind power prediction, and uses convolutional network to predict wind power. Compared with the existing methods, the proposed method greatly optimizes the prediction accuracy and the time cost for training models. In addition, this paper also proposes an approach to fuse various types of data by multi-spatio-temporal feature, which is then proved to be effective in the experiment.

In fact, spatio-temporal feature is modeling the spatio-temporal

Acknowledgments

This work is supported by National Natural Science Foundation of China (Grant No. 61877043), Major Scientific and Technological Projects for A New Generation of Artificial Intelligence of Tianjin (Grant No. 18ZXZNSY00300), and Key Project for Science and Technology Support from Key R&D Program of Tianjin (Grant No. 18YFZCGX00960).

References (37)

Cited by (0)

View full text