The main problem that affects the accurate prediction is non-stationarity of wind power and photovoltaic power. The signal decomposition technique can decompose the non-stationary power sequence into several intrinsic mode component of different frequencies, smoothing the fluctuation degree of the sequence, extracting sequence features, establishing a prediction model with strong adaptability, and improving the prediction accuracy. This paper firstly classifies and divides the signal decomposition technology, and systematically reviews the application research status of domestic and foreign researchers on the signal decomposition technology from the perspectives of time domain decomposition and frequency domain decomposition. Secondly, the application research examples of the signal decomposition technology are reviewed in detail from the two aspects of wind power and photovoltaic power prediction, and the advantages and disadvantages of the decomposition technology are compared through the actual decomposition results. Finally, the application scenarios of the signal decomposition technology are summarized, and the application fields of the decomposition technology and the research direction of improving the prediction accuracy are prospected.