Energy Consumption Forecasting Using Machine Learning and IoT System

Abstract

The objective of this research is to propose the development of a forecasting model for the electricity consumption of air conditioning systems in server rooms using machine learning and Internet of Things (IoT) systems. The study collects data on external temperature, humidity, and air conditioning power consumption through IoT technology. The methodology consists of three main parts: (1) data collection of temperature, humidity, and electricity consumption using IoT systems; (2) development of an electricity consumption forecasting model; and (3) performance evaluation of the forecasting methods. Four forecasting methods are selected for comparison: Long Short-Term Memory (LSTM), Bidirectional LSTM (Bi-LSTM), Artificial Neural Network (ANN), and Autoregressive Integrated Moving Average (ARIMA). The performance is assessed based upon Mean Square Error (MSE), Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). The results indicate that the IoT system can collect data accurately and efficiently. Among the forecasting methods, LSTM demonstrated the highest accuracy in terms of the lowest MSE and RMSE values, while ARIMA exhibited the lowest MAPE. Machine Learning enhances the accuracy of electricity consumption forecasting by efficiently learning patterns from large datasets collected through IoT systems. These findings suggest that LSTM is the most appropriate model for forecasting electricity consumption. However, the choice of a forecasting model depends on the specific problem and practical requirements. This forecasting model enables users to estimate energy consumption in advance and manage energy more effectively.