Economic Dispatch Considering Renewable Energy Generation and Load Forecasting

Abstract

Accurate short-term load forecasting (STLF) and optimal generation scheduling are essential for reliable and economic power system operation. This paper proposes a hybrid deep learning model combining Gated Recurrent Unit (GRU) and Long Short-Term Memory (LSTM) networks for STLF, integrated with a Genetic Algorithm (GA)-based Economic Dispatch (ED) framework. The model is trained and validated using the publicly available Panama Short-Term Load Forecasting dataset, which provides hourly electricity demand along with meteorological variables. Input features include temperature, wind speed, time of day, day of the week, and historical demand. Forecasting performance is evaluated using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and the Coefficient of Determination (R²). Simulations on the IEEE 9-bus system demonstrate that the proposed GRU+LSTM model achieves superior performance compared to standalone GRU and LSTM, with an MSE of 0.0310, RMSE of 0.1760, and R² of 0.9670. Furthermore, the forecasted demand is used in the ED problem, formulated as a multi-objective function incorporating fuel cost, emission cost, and wind generation cost. Results confirm that the proposed integrated approach enhances forecasting accuracy and reduces operational costs, making it effective for data-driven power system operation.