Predictive Model Analytics Using Data Mining and Machine Learning: A Case Study on Forecasting GCC Power Demand

The electricity demand forecast is a vital tool to be able to achieve energy sustainability and economic reliability as the world today runs on energy, thermal, renewable, nuclear, etc. Predicting electricity energy consumption is critical in any state’s decision-making, planning, and regulation of the electrical power sector. GCC states need to ensure the most accurate prediction and forecast for energy demand that will meet the ambitious economic plans and plans to face the challenges of climate change. In this paper, a predictive model will be introduced that will use data mining techniques as well as machine learning in creating a more accurate future electrical power demand forecast for GCC states: Kingdom of Bahrain, Kingdom of Saudi Arabia, Sultanate of Oman, United Arab Emirates, State of Kuwait, and States of Qatar. The timeline of data is from April 2015 to March 2019. The data for electrical demand used is from the national transmission level. There is a number of prediction algorithms that will be tested for the purpose of providing the best-fit model for GCC electrical demand for the years 2014 to 2019. The quality of the prediction algorithms used will be evaluated based on forecasting error parameters of mean absolute deviation (MAD), mean square error (MSE), and mean absolute percentage error (MAPE). The best forecasting performance was found from the adoption of the autoregressive integrated moving average (ARIMA) method after using K mean data clustering to drive the pattern between the independent and dependent variables. It was concluded that the use of data mining and autoregressive integrated moving average (ARIMA) could reduce the prediction error in GCC countries and thus lead to better resource optimization for generation expansion projects on the GCC level.