A Novel Hybrid Model Based on an Optimized Multilayer Perceptron Neural Network and the Cuckoo Search Algorithm for Forecasting Monthly Household Electricity Consumption: A Case Study of Ilam County

Abstract

The objective of this study is to propose a novel method for forecasting the monthly electricity consumption of subscribers in Ilam County. The results of this study can assist energy-supply companies in obtaining an overall understanding of household electricity consumption patterns. For forecasting purposes, a combination of a multilayer perceptron (MLP) neural network and the Cuckoo Search Optimization Algorithm is employed. In the proposed method, the estimation process is performed in three stages: “preprocessing,” “feature selection,” and “estimation.” The minimum redundancy–maximum relevance (MRMR) algorithm is used for feature selection. This algorithm identifies the optimal features for forecasting electricity consumption such that they exhibit minimum redundancy and maximum relevance to the target variable. In the estimation stage, a combination of three multilayer perceptron neural networks is employed, and their configurations are adjusted using the Cuckoo Search Optimization Algorithm so as to minimize the estimation error. The results of the performance evaluation of the proposed method indicated that the accuracy of forecasting subscribers’ monthly electricity consumption can be maximized using six features. Furthermore, comparison of the proposed method with other existing techniques showed that the proposed hybrid system can perform the forecasting process with a mean absolute error of 5.66 kWh, representing an error reduction of at least 29.4% compared with previous methods.