Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Wind Power Prediction across Different Locations using Deep Domain Adaptive Learning

Md Saiful Islam Sajol, Md Shazid Islam, A S M Jahid Hasan, Md Saydur Rahman, Jubair Yusuf

TL;DR

This work addresses wind power forecasting across different geographic locations by leveraging a source-free, domain-adaptive deep learning framework. A pre-trained DNN from one climate is adapted to another with minimal updates, after selecting a compact set of weather features via random forest. A novel dataset combining EMHIRES wind data with Viscro meteorology from Germany, France, and the UK supports the evaluation, achieving significant accuracy gains (6.14% to 28.44%) over non-adaptive baselines. The approach offers faster convergence and reduced data needs, promising robust, location-agnostic wind power prediction in distributed grid settings.

Abstract

Accurate prediction of wind power is essential for the grid integration of this intermittent renewable source and aiding grid planners in forecasting available wind capacity. Spatial differences lead to discrepancies in climatological data distributions between two geographically dispersed regions, consequently making the prediction task more difficult. Thus, a prediction model that learns from the data of a particular climatic region can suffer from being less robust. A deep neural network (DNN) based domain adaptive approach is proposed to counter this drawback. Effective weather features from a large set of weather parameters are selected using a random forest approach. A pre-trained model from the source domain is utilized to perform the prediction task, assuming no source data is available during target domain prediction. The weights of only the last few layers of the DNN model are updated throughout the task, keeping the rest of the network unchanged, making the model faster compared to the traditional approaches. The proposed approach demonstrates higher accuracy ranging from 6.14% to even 28.44% compared to the traditional non-adaptive method.

Wind Power Prediction across Different Locations using Deep Domain Adaptive Learning

TL;DR

This work addresses wind power forecasting across different geographic locations by leveraging a source-free, domain-adaptive deep learning framework. A pre-trained DNN from one climate is adapted to another with minimal updates, after selecting a compact set of weather features via random forest. A novel dataset combining EMHIRES wind data with Viscro meteorology from Germany, France, and the UK supports the evaluation, achieving significant accuracy gains (6.14% to 28.44%) over non-adaptive baselines. The approach offers faster convergence and reduced data needs, promising robust, location-agnostic wind power prediction in distributed grid settings.

Abstract

Accurate prediction of wind power is essential for the grid integration of this intermittent renewable source and aiding grid planners in forecasting available wind capacity. Spatial differences lead to discrepancies in climatological data distributions between two geographically dispersed regions, consequently making the prediction task more difficult. Thus, a prediction model that learns from the data of a particular climatic region can suffer from being less robust. A deep neural network (DNN) based domain adaptive approach is proposed to counter this drawback. Effective weather features from a large set of weather parameters are selected using a random forest approach. A pre-trained model from the source domain is utilized to perform the prediction task, assuming no source data is available during target domain prediction. The weights of only the last few layers of the DNN model are updated throughout the task, keeping the rest of the network unchanged, making the model faster compared to the traditional approaches. The proposed approach demonstrates higher accuracy ranging from 6.14% to even 28.44% compared to the traditional non-adaptive method.
Paper Structure (14 sections, 4 figures, 4 tables)

This paper contains 14 sections, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Methodology
  4. Training the model in source domain’s wind dataset
  5. Adaptation in Target Domain’s Dataset
  6. Dataset
  7. Experiments and Results
  8. Experimental Setup
  9. Performance of Adaptation on Accuracy:
  10. Performance comparison between training from scratch and using adaptation :
  11. Aditional Studies
  12. Performance comparison between updating full network and partial network:
  13. Performance comparison between with and without ef- fective feature selection:
  14. Conclusion & Future Work

Figures (4)

  • Figure 1: Overview of the methodology is shown in the figure. At first, a model is trained from scratch on the source side using the source data $(X_s, Y_s)$. Then the pretrained model of the source side is transferred to the target side. On the target side, the weight of the last two layers (FC layers) of the model is adapted using target data $(X_t, Y_t)$. The rest of the network weights are kept as same as the pre-trained model.10407265
  • Figure 2: Histogram of wind power generation shown in 6 bins. Note, that the values along the x-axis are normalized by their maximum generated power.
  • Figure 3: Correlation matrix of the selected features
  • Figure 4: Comparing the temporal performance between training from scratch and using adaptation from a pre-trained model for three countries - Germany, United Kingdom and France. The graphs show that with domain adaptation, the network completes the training (reaches the saturation of accuracy) faster.