Theory-guided Output Feedback Neural Network (Tg-OFNN) for Short-term Wind Power Forecasting

Under the dual stimulation of the grid requirements for renewable energy and the demand to reduce the cost of wind power, the wind power forecasting begins to be in a position that cannot be ignored. In order to overcome the challenge of high randomness of wind, a theory-guided output feedback neural network (Tg-OFNN) is proposed for short-term wind power forecasting. According to the known characteristics of the wind turbine, the wind power is decomposed into a fixed part and an uncertain part, in which the first part is expressed by a known mathematical model, and the second part is approximated by OFNN because of its strong ability to fit the nonlinear relationship. Part of data from the Global Energy Forecasting Competition 2012 is used to verify the performance of the proposed model. The results of the case study show that the introduction of domain knowledge improves the forecasting performance of the model, and the memory function is significant to deal with time series such as wind power. Coefficient of determination (R
²
) and Root Mean Square Error (RMSE) are chosen as the index to evaluate the performance of models.