Track: Modeling and Simulation

Abstract

This paper presents electricity forecasting by integration of Nested Long Short Term Memory (NLSTM) as the another type of LSTM, Attention Mechanism (AM) as the technique for Neural Network and optimized by Cuckoo Search Optimizer (CSO) as the metaheuristic. It is important because of the electricity forecasting issue. Electricity is the crucial energy, due to the many sectors which is require it. Estimation of electricity consumption can enhance the effectiveness of generation and supply. The shortage and outage of electricity is the thing that should be avoided through the electricity forecasting. In order to create better forecasting result, the forecasting by NLSTM-AM will be combined. NLSTM-AM has several parameters which influence the network performance. Those parameter will be optimized by CSO. To discover the performance of this proposed method, the RMSE of this method will be compared by other algorithms. In this paper there are two datasets which will be used, those are Taiwan electricity consumption and Turkey electricity consumption. It is proved that NLSTM-AM-CSO has the smallest RMSE rather than other algorithms.