PREDICTING ELECTRICITY CONSUMPTION IN NORWAY: A COMPARISON OF MACHINE LEARNING MODELS

Abstract

This paper proposes a model for the estimation of the consumption of electricity in Norway, which can accurately predict the next 24 h of load with and estimation of load for 1 week to 1 month using in one-hour intervals. It shows the utilization of one timeseries model; ARIMA, two regression models; Linear regression and Ridge Regression Models, two Ensemble; XGBoost and Random Forest, and one Neural network; MLP. In addition, the present paper shows the way to significantly improve the accuracy of the prediction through feature engineering, ensemble machine learning and neural network process.

Upon rigorous evaluation, the ARIMA model demonstrated exceptional predictive accuracy across all zones, as evidenced by the close alignment between the actual and predicted values of electricity consumption. Furthermore, meticulous parameter tuning, and model optimization techniques were employed to enhance the model's predictive performance and generalization capabilities.

The rest of the models performed equally well with XGBoost leading with 99% accurate predictions.

Ensemble methods like XGBoost and Random Forest excel in capturing complex relationships and achieving high accuracy, linear regression-based approaches offer simplicity and interpretability. MLPs provide a flexible and powerful modelling framework that can capture intricate patterns in the data. Understanding the trade-offs and strengths of each algorithm is crucial for making informed decisions in predictive modelling tasks, especially in domains like energy consumption forecasting where accurate predictions are essential for effective resource management and decision-making.

In conclusion, the findings of this study highlight the efficacy of ARIMA as a valuable tool for forecasting electricity consumption across diverse geographical zones.