Abstract:

Aiming at the small scale,strong volatility and unpredictability of the park-level micro-grid load,less measurable and predictable factors like weather and motion sensing parameters,and greater impact of random starting load,the load prediction error of the traditional power system is large,a load prediction method based on long and short-term memory (LSTM) daytime prediction and particle swarm optimization algorithm (PSO) intra-day modification is proposed.The feature extraction ability and time series correlation learning ability of LSTM learning model is used to obtain the daytime forecast load curve.In order to further improve the prediction accuracy,PSO is used to revise the load curve twice on the date to be predicted.The example shows that the proposed method has high accuracy and can be applied to the short term load prediction of micro grid.

Key words: micro grid, short-term load forecasting, long and short term memory (LSTM), particle swarm optimization (PSO)