Box-jenkins and genetic algorithm hybrid model for electricity forecasting system

Abstract

Energy is considered a prime agent in the generation of wealth and also a significant factor in economic development. There has been a strong relationship between the availability of energy to the economic activity, improvements in standards of living and the overall social well-being. In making a forecast for energy demand, accuracy is the primary criteria in selecting among forecasting techniques. Time Series method has always been used in a variety of forecasting applications. In this thesis, an approach that combines the Box-Jenkins methodology for SARIMA model and Genetic Algorithm (GA) will been introduced as a new approach in making a forecast. Data used in this study were collected from the year 1996 until year 2003 that has been classified into total monthly electricity generated in kWh unit. GA is widely known as a multi-purpose searching procedure commonly use in optimization and approximation field. The increasing popularity of GA is due to their adaptability and simplicity as a problem solution especially when they are applied into several complex problems. By adopting the GA blind search, the algorithm combines searching techniques and their capabilities to learn about the relationship of the pattern-recognition of the past data. This character helps GA to make a prediction of future values. This study proposed the possibility of using GA’s approach as one of the unique forecasting method. It also represents a preliminary work in the current research and practices of GA. The investigation is simulated using Intelligent Electricity Forecasting System (IEFS) developed in this research which written in Borland Delphi 7.0 programming.