Power Load Forecasting in Station Areas bBased on ASGMD-MLR-CLSTM

Accurately predicting the power load in a station area can encourage power companies to arrange dispatching plans reasonably, ensuring the safety and stable economic operation of the station area. To fully exploit the characteristics of power load data and improve the accuracy of forecasting, a power load prediction method based on adaptive symplectic geometry mode decomposition (abbr. ASGMD), multiple linear regression (abbr. MLR) and convolutional long short-term memory (abbr. CLSTM) was proposed. Firstly, ASGMD was applied to decompose the station load data into two components: weakly correlated and strongly correlated. Secondly, MLR and LSTM were adopted to forecast the above two components, respectively. Finally, the load forecast value was obtained by combining the results of each model. The experiments show that the proposed method obtains higher forecasting accuracy than other models.