Abstract :
High-energy consumption is a major challenge most sectors face, including data centers. the data center sector accounts for about 3% of the world's total energy consumption, which has been predicted to keep increasing. Most data centers are run for profit-making, and the high-energy usage makes them expensive to operate. This high-energy consumption also causes environmental pollution due to the emission of greenhouse gases. Forecasting energy consumption for data centers is important in decision-making for effective energy saving. This study considered statistical, machine learning, and deep learning algorithms. Dataset was obtained from the EnergyPlus simulation platform. the simulation was informed by the information gathered from one of the leading data centers in Lagos, Nigeria. the algorithms considered were ARIMA, SVR, and LSTM. These algorithms were compared to determine an optimal algorithm using five (5) performance evaluation metrics: MSE, RMSE, MAE, MAPE, and Accuracy. the optimal algorithm was modified and utilized to develop a model. This is a step towards producing an accurate energy consumption forecast tool for data centers.
Keyword :
Algorithms, Data Centre, Energy, Machine Learning, Model.