Forecasting of renewable energy using ANN, GPANN and ANFIS (A comparative study and performance analysis)

Abstract

Prediction and forecasting is preserved to be an important stage in diverse problems. The main aim of our manuscript is to forecast the wind speed and the temperature based on data collected months ago. The data and calculations we obtained for the temperatures in about 4 years ago from 2015 till 2018, whereas the statistics calculated for the wind speed were about 20 years from 1996 until 2015. The data of the wind speed and the temperature collected in different regions of Egypt East coast and Alsheihkzayid. The system used for prediction is based on three different methods which are Artificial Neural network (ANN), Genetic algorithm fused with artificial neural network (GPANN) and Adaptive Neuro-fuzzy inference system (ANFIS). They were used to forecast the future temperature and the future wind speed. The results proved that the system is robust, and it can be applicable during real time. The performance of ANFIS with the trapezoidal membership function proved to obtain the highest performance over all other methods. The optimal mean square error (MSE) reached for the wind speed was 7.2989 m/s and for the temperature is 3.8364 C°.