Wind Power Forecasting Enhancement Utilizing Adaptive Quantile Function and CNN-LSTM: A Probabilistic Approach

Wind power generation forecasting is a crucial aspect in renewable energy industry since accurate predictions of wind power generation can support the power grid operators and power plant owners to optimize their energy resources management. This will potentially reduce the purchase of extra electricity from other stakeholders or neighboring countries and will consequence significant cost savings and further reliability. With the increasing importance of wind power utilization in modern power systems, employing accurate forecasting model cannot be understated. In this study, a new probabilistic forecasting model based on the utilization of quantile functions is discussed. The proposed model integrates an adaptive optimal weighted continuous ranked probability score (CRPS) is presented to improve the computational efficiency of the prediction process and enabling more accurate forecast output. Also, an adaptive Adam optimization algorithm is presented for the CRPS loss minimization. A convolutional neural network based long short-Term memory (CNN-LSTM) is employed to evaluate the quantile function parameters. To validate the effectiveness of our approach, the specific forecasting models have been compared across a wide range of scenarios. The obtained results unequivocally reveal the superiority and improved accuracy of the proposed forecasting model.