Smart Forecasting Of Electricity Price For Cloud Computing Using Ml Techniques

Abstract

Accurate short-term electricity demand forecasting has become an essential requirement for modern cloud 
computing environments due to the increasing energy consumption of data centres and the need for efficient 
resource management. This paper presents a machine learning-based electricity demand forecasting framework 
that performs a comparative analysis of four forecasting models: Autoregressive Integrated Moving Average 
(ARIMA), Long Short-Term Memory (LSTM), Extreme Gradient Boosting (XGBoost), and a Hybrid LSTM + XG 
Boost model. The proposed framework utilizes a high-resolution 5-minute interval electricity demand dataset 
collected from Delhi, India, spanning the period from 2021 to 2024, along with weather-related features to 
improve forecasting accuracy. Data preprocessing and feature engineering techniques are applied to extract 
temporal and environmental characteristics before model training and evaluation. The forecasting models are 
assessed using standard performance metrics including Root Mean Square Error (RMSE), Mean Absolute Error 
(MAE), Mean Absolute Percentage Error (MAPE), and the coefficient of determination (R²). Experimental results 
demonstrate that the Hybrid LSTM + XGBoost model achieves the lowest prediction errors and the highest 
forecasting accuracy among the evaluated models by combining the temporal learning capability of LSTM with 
the nonlinear prediction strength of XGBoost. The forecasted electricity demand is further utilized to estimate 
electricity price using a simple demand-based linear relationship. The proposed framework provides an effective 
solution for demand management, cost optimization, infrastructure planning, and energy-efficient operation of 
cloud computing environments and smart power systems.