Long Term Load Forecasting Based on Hybrid Model of Feed Forward Net and Modified Grey Wolf Optimization

Abstract

This paper presents a hybrid modified grey wolf optimization (MGWO) algorithm with the feed forward net (FFN), named MGWO-FFN, for solving electrical load forecasting. The proposed model is implemented with two stages: firstly, MGWO algorithm estimates the optimum variables of the FFN through the pre-determined training samples. Then the adapted FFN is tested with the remaining other samples and is utilized to predict the electrical peak load (PL). The proposed algorithm is investigated on two real cases (i.e. predicting the annual total electrical load consumption of Beijing's city and the annual PL consumed in Egypt). To prove the superiority of the proposed algorithm, MGWO is validated by comparing with algorithm including classical GWO and PSO algorithms. Both of Beijing's and Egypt's cases results indicate that the proposed MGWO-FFN algorithm outperforms the others where less mean square error (MSE) and more accuracy are obtained compared to the error that yields using the other two algorithms.