Research on short-term load forecasting based on Adam-SGD-LSTM

Aiming at the problem of low prediction certainty of existing load forecasting methods, this paper proposes a short-term load forecasting system that is based on the LSTM neural network. The original dataset is first processed by the median elimination method, Markov-Monte Carlo algorithm and improved hierarchical clustering algorithm. Then, the data are input into the LSTM prediction model optimized by the Adam method and SGD method for prediction. The example analysis proves that the method proposed in this paper has better prediction accuracy, which is helpful for the planning and movement of the power system.