OICC Press International Journal of Energy and Environmental Engineering 2251-6832 13 1 (March 2022) 2021 07 01 A novel hybrid model based on weather variables relationships improving applied for wind speed forecasting 10.1007/s40095-021-00408-x EN Hervice Roméo Fogno Fotso Labotatory of Mechanics, Department of Physics, University of Yaoundé I, Yaoundé, CM Claude Vidal Aloyem Kazé Department of Renewable Energy, HTTTC Kumba, University of Buea, Buea, CM Germaine Djuidje Kenmoé Labotatory of Mechanics, Department of Physics, University of Yaoundé I, Yaoundé, CM Journal Article 2021 07 01 Abstract Accurate wind speed forecasting is imperative for producing wind power integration. Thus, this paper presents a novel combined ARIMA—artificial neural network (ANN) forecasting model based on improved relationships between the wind speed and other weather variables pursuing to forecast wind speed. The weather variables on which the wind speed depends are transformed for normalization and relationships improving by using a proposed approach. The ARIMA models are employed for each transformed variables modeling and the unknown values forecasting. The measured wind speed is normalized and used as the target variable and the transformed weather variables as input variables to train the ANN model. The predicted weather variables are employed as input variables of trained ANN to forecast the unknown wind speed values. The proposed forecasting model has been validated with five different ANN structures for multi-step ahead wind speed forecasting from two Datasets in Bapouh, Cameroon. The experimental results indicate that the proposed data preprocessing strategy is appropriate to enhance the relationships between two variables and decrease the seasonal variation. Furthermore, the proposed hybrid model in terms of forecasting accuracy outperforms other comparable models.