Research on short-term wind speed prediction based on deep learning model in multi-fan scenario of distributed generation. (November 2022)

• Record Type: Journal Article
• Authors: Long, Hongyu; He, Yongsheng; Cui, Hui; Li, Qionghui; Tan, Hao; Tang, Bangrui
• Is Part Of: Energy reports. Volume 8(2022)
• Page Start: 14183
• Page End: 14199
• Publication Date: 2022-11
• Subjects: Short-term wind speed prediction -- Distributed generation -- Convolutional neural network (CNN) -- Modified sparrow search algorithm (MSSA) -- Deep learning model; Power resources -- Periodicals; Energy industries -- Periodicals; Power resources; Periodicals; Electronic journals; 621.04205
• DOI: 10.1016/j.egyr.2022.10.399
• Languages: English
• ISSNs: 2352-4847

Abstract

Abstract: The climate and environmental pollution problems caused by carbon dioxide and other harmful gases emitted from traditional fossil fuel thermal power plants are increasingly threatening the living environment of mankind. In September, 2020, the Chinese government clearly put forward the national strategic goal of "Carbon Peak and Carbon Neutrality". Distributed generation is the main means to effectively reduce carbon emissions, especially the rapid development of wind power generation. Accurate and stable wind speed prediction can reasonably formulate power generation plans and optimize power dispatching, which is an effective means to reduce the overall carbon emissions of the power system. This paper designs and proposes a hybrid wind speed prediction model based on convolutional neural network and long short-term memory network deep learning model. Based on the historical wind speed data set collected at the location of multi-fan in the same wind farm, high-precision and stable short-term wind speed prediction is realized. Firstly, singular spectrum analysis aims to remove the noise components in the wind speed series and reduce the impact of noise on the prediction performance of the model. Secondly, convolutional neural network (CNN) is introduced to extract the features of the de-noised wind speed sequence, which provides more effective information for the training of the prediction network. Then, a prediction network based on sparrow search algorithm and long short-term memory network is constructed, and the modified sparrow search algorithm is used to optimize the selection of long short-term memory (LSTM) Hyper-parameters. Eventually, to verify the superiority of the proposed model, an evaluation system based on accuracy, stability and complexity indicators is constructed. The experimental results show that the index values of the wind speed prediction model proposed in this paper based on dataset 1 in the one-step prediction simulation experiment are the smallest among all comparison models, and the same is true for dataset 2.