An integrated wind power prediction method based on heterogeneous clustering and DALSTM

Currently, the power grid is greatly impacted by wind power’s intermittent and fluctuating problems. If wind power forecasting can be accurately implemented, it will be crucial for the power grid to develop coordinated power dispatch. It is found that the traditional methods have problems such as insufficient data mining and low accuracy of prediction algorithms. Therefore, an integrated wind power prediction method based on heterogeneous clustering and DALSTM is proposed. Firstly, an integrated clustering framework with multiple heterogeneous algorithms is constructed, and wind power sequences under different fluctuation states are obtained. Secondly, the LSTM prediction model is established, and the feature space and temporal attention mechanism are introduced to dynamically mine the patterns of wind power sequences and input features. Finally, a wind farm in China is used as an example to demonstrate the validity and accuracy of the method.

I. Introduction

In recent years, wind power technology has achieved rapid development due to its many advantages of cleanliness, wide distribution, and rich resources [1], At present, wind power generation has become the main form of renewable energy utilization. The global new installed capacity of wind power reached 93.6 GW in 2021, with China, the United States, and Brazil occupying the top three [2], However, the intermittent and fluctuating nature of wind power has greatly impacted the grid and greatly increased the complexity of grid dispatch. It is evident that accurate wind power forecasting is of great importance [3]: it can provide forecast information for safe grid operation and reduce the complexity of grid operation.