Key points are not available for this paper at this time.
The development of renewable clean energy wind power is very rapid. However, the strong randomness and weak dispatchability of wind power bring significant challenges to wind power operation and maintenance, such as grid connection, power dispatch, and unit combination. At this stage, the combined model based on signal decomposition-prediction technology plays an essential role in ultra-short-term deterministic wind power prediction, promoting the development of the wind power industry. With the development of artificial intelligence technology, this paper discusses and summarizes the latest research progress on the critical technologies of the combined signal decomposition-prediction technology model to improve the prediction accuracy for the problem of high demand for ultra-short-term power prediction accuracy. Data decomposition from two perspectives of decomposition techniques reduces input data complexity and feature extraction by eliminating redundant information. The prediction techniques are from two perspectives: prediction methods to demonstrate specific laws and parameter search to determine the optimal parameters. Three perspectives evaluated prediction performance: vertical error evaluation, horizontal error evaluation, and prediction error evaluation by analyzing the above standard methods in detail, including the principle, idea, application examples, advantages and disadvantages, key points, and application notes of each technique. And the prediction performance of the current combined models is compared extensively. Finally, the future research trends of wind power ultra-short term are discussed and foreseen to provide a reference for the development of ultra-short-term power prediction technology, prediction performance improvement, and safe and stable grid operation.
Chen et al. (Tue,) studied this question.