The Use of Additive Decomposition and Deep Neural Network for Photovoltaic Power Forecasting

Mohsen Parsa Moghaddam,M. Imani,Fariba Dehghan
DOI: https://doi.org/10.1109/ICEE59167.2023.10334838
2023-05-09
Abstract:Predict photovoltaic (PV) power production is indispensable for security and reliability of the grid. In this article, a short-term forecasting method, namely trend decomposition two-dimensional convolutional neural network based on additive decomposition and convolution neural network (CNN) is proposed. Firstly, the additive decomposition model is deployed to decompose the PV power generation series to the long-term trend (LT), the seasonal trend (ST), and the random component. Then, three independent two-dimensional convolutional neural networks are designed to extract daily and hourly dependencies among the decomposed components. Finally, the prediction results of these networks are summed for the final forecast. The one-day-ahead forecasting capability of the presented method is evaluated with two case studies using real datasets gathered from Limburg and Luxembourg, Belgium. Analysis of the prediction's results indicates that the proposed method has higher accuracy compared to individual multi-layer perceptron, two-dimensional convolutional neural network, long short-term memory (LSTM), gated recurrent unit, and bidirectional LSTM networks.
Computer Science,Environmental Science,Engineering
What problem does this paper attempt to address?