Power Consumption Prediction in Urban Areas using Machine Learning as a Strategy towards Smart Cities

Abstract

The bedrock of every smart city is the continuous electricity supply. It is an important factor in the growth of smart cities. Smart transmission and distribution of power are crucial for the constant power supply in a smart city. The ability to correctly predict power consumption is a key factor in achieving demand-driven output through careful planning, and prompt usage of electricity generated. This paper applies Machine learning techniques to predict power consumption using historical data of power consumed in an urban area. Dataset obtained from Kaggle was used for the experiments in this paper. The paper used four machine learning models, namely, Bagging (BAG), Stochastic Gradient Boosting (GBM), Model Averaged Neural Network (MAN), and K-Nearest Neighbours (KNN). Experimental results showed that GBM outperforms other machine learning models used for the study with a prediction accuracy of 96%. This is an indication that GBM is a promising algorithm for predicting power consumption in a smart city if adopted.