Title

Solar Irradiance Forecasting using Deep Recurrent Neural Networks

Abstract

Solar irradiance prediction has a significant impact on various aspects of power system generation. The predictive models can be deployed to improve the planning and operation of renewable systems and can improve the power purchase process and bring several advantages to the power utilities. The irradiance is affected by several factors, such as clouds and dust, and it becomes challenging for physical models to predict and capture the dynamics. The statistical methods are commonly used to predict the irradiance. These methods include autoregressive moving average, support vector machine, and artificial neural network. Deficiencies and challenges of existing methods include low prediction accuracy, low scalability for big data, and inability to capture long-term dependencies. In this paper, a deep recurrent neural network is used to predict the solar irradiance. Deep recurrent neural network (DRNN) is an artificial neural network with more hidden layers to improve the complexity of the model and enable the extraction of high-level features. The neural network is trained, tested, and validated using real data from the National Resources in Canada. The simulation and experimental results are compared to other methods to illustrate the advantages using the proposed approach.

Meeting Name

6th International Conference on Renewable Energy Research and Applications (2017: Nov. 5-8, San Diego, CA)

Department(s)

Electrical and Computer Engineering

Second Department

Engineering Management and Systems Engineering

Keywords and Phrases

Big Data; Deep Learning; Deep Neural Networks; Forecasting; Neural Networks; Purchasing; Recurrent Neural Networks; Solar Radiation; Autoregressive Moving Average; DRNN; High-Level Features; Irradiance; Long-Term Dependencies; LSTM; Prediction Accuracy; Solar; Long Short-Term Memory; Prediction; PV; RNN

International Standard Book Number (ISBN)

9781538620953; 9781538620960

International Standard Serial Number (ISSN)

2572-6013

Document Type

Article - Conference proceedings

Document Version

Citation

File Type

text

Language(s)

English

Rights

© 2017 Institute of Electrical and Electronics Engineers (IEEE), All rights reserved.

Share

 
COinS