Solar power prediction approach using data augmented deep learning technique

Articles in Press

Document Type : Article

Authors

1 - Dept. of Computer Science and Engineering, Nit Meghalaya, Meghalaya, India - 1School of Computer Engineering, Kalinga Institute of Industrial Technology Deemed to be University, Bhubaneswar, Odisha-751024, India

2 Dept. of Computer Science and Engineering, Nit Meghalaya, Meghalaya, India

3 Department of CSE, Parala Maharaja Engineering College,Odisha, India

Abstract

Solar power prediction holds a significant impact for future renewable energy scenario. To achieve a more accurate predicted output a novel prediction technique has been included in this paper for short and medium term solar power prediction. Initially the original solar power is decomposed into a set of subseries using VMD based decomposition technique. Data augmentation technique is applied for generating more training data thus avoiding the problem of over fitting. Then a novel prediction model based on LSTM (long short term memory) MKRVFLN (multi kernel based random vector functional link network) is proposed for point prediction of short term and medium term solar power. A fuzzy entropy based strategy is implemented for partitioning the subseries and assigning it to LSTM and MKRVFLN. For further improvement in prediction accuracy WCO technique is used for obtaining optimised kernel parameters. The performance of the proposed technique is compared with seven other prediction techniques. Solar power data from two different data sites are considered for comparison purpose. The experiment is performed on two aspects: short term and medium term point prediction, the result analysis shows that the proposed solar power prediction model shows excellent result as compared to other traditional methods.

Keywords

Main Subjects

Scientia Iranica

Articles in Press, Accepted Manuscript
Available Online from 24 July 2024

Files

History

  • Receive Date: 25 April 2023
  • Revise Date: 22 June 2023
  • Accept Date: 24 July 2024

Share

How to cite

Statistics