Short-term wind power prediction method based on deep learning

Abstract

The invention discloses a short-term wind power prediction method based on deep learning, which comprises the steps of: 1) inputting historical wind power data and related weather characteristic factors such as wind speed and direction by a computer, and preprocessing the acquired data; 2) using a CNN to extract and mine the features of the preprocessed data to form feature map. 3) using a gated recurrent unit (GRU) neural network to train and model the feature map, and establishing a nonlinear relationship between the feature map and the wind farm power by continuous optimization and parameter adjustment to form a short-term wind power prediction model; 4) using the trained model to predict the wind power of the wind farm for a period of time and generate a wind power prediction result ofthe wind farm; 5) and outputting the wind power prediction result by the computer. The method has significant improvement in prediction accuracy and prediction efficiency, provides a basis for rational scheduling of a power grid, and has an industrial application value.

Description

technical field [0001] The present invention relates to the technical field of power system prediction and control, in particular to a short-term wind power prediction method based on deep learning. Background technique [0002] Wind power forecasting is a key technology in wind power generation systems. Accurate forecasting of the future wind power of wind farms can effectively reduce and avoid the impact of wind farms on the power system. Therefore, wind power prediction methods play an important role in the sustainable development of wind power generation. The current wind power prediction methods can be mainly divided into physical methods, statistical methods, learning methods and a mixture of the above methods, each of which has its own adaptive time scale and data type. [0003] The physical model is a method to indirectly predict wind power. First, the data in the NWP is used as the initial value, and the numerical method is used to solve the atmospheric dynamics an...

AI Technical Summary

Problems solved by technology

Compared with the traditional shallow neural network, deep learning has a series of hidden layers capable of nonlinear transformation, which can challenge more complex environments and problems

Images