Prediction of Electrical Power Consumption in Jordan

Abstract

Modeling and predicting electricity power consumption is a crucial task in managing energy performance in present and future. For successful planning, it is important for decision-makers to have a precise idea about the needed electrical power in any period of time in the future. This paper tackles this task of predicting future power consumption in Jordan using a number of machine learning algorithms. Neural network, support vector machine (SVM), and random forest are used to build our prediction models. Two datasets are use: one small and another large. The results show that the random forest performed best when trained on the large data with and without data normalization. After normalizing the data, the performance of the neural networks becomes similar to that of random forest. The performance of the support vector machine was high before and after normalizing the data. Regarding the small dataset and before normalizing the data, the performance of the random forest was better than the other two algorithms. However, after normalizing the data, the neural network performed the best and the random forest comes next. Finally, in terms of feature importance, the experimental results show that the price feature was the most important feature in the large dataset. The price and the renewal energy projects were the most important features in the small dataset.