The article presents a new model, E-BLSTM, to forecast short-term electricity prices in the market more accurately. By using a two-way LSTM model with ELU activation function, the model addresses the issue of large errors in current forecasting methods. Experiments on the PJM power market database show that the E-BLSTM model outperforms traditional ARIMA and ARMA models, providing more accurate predictions with lower loss rates. This improved forecasting can help better predict fluctuations in electricity prices on the supply side of the market.