The invention discloses a power system load prediction method based on singular spectrum analysis and a memory network. The method comprises the following steps of firstly, adopting singular spectrumanalysis to decompose a power load historical data time sequence, then grouping characteristic values and corresponding characteristic vectors and acquiring a plurality of reconstruction components through reconstruction; respectively constructing a long and short term memory network prediction model for each component; finally, adding the prediction results of each component to obtain a total load prediction result. According to the method, singular spectrum analysis is used as a preprocessing tool, the power load can be decomposed and reconstructed into components which reflect different fluctuation characteristics of the power load and are easier to predict, and therefore high prediction precision can be achieved.