IOT based Online Load Forecasting using Machine Learning Algorithms

Abstract

Abstract This paper presents a novel online load forecasting using supervised Machine Learning (ML) algorithms in Internet of Things (IOT) environment. Short Term Load Forecasting (STLF) is an essential aspect for smart grid operations such as power dispatch and load management. IOT is an emerging Technology breaching into every segment of science and engineering. This work presents the possibility of STLF online with accurate prediction models by using ML algorithms. Electrical load consumption data and weather data at a research Lab, JNTUH, Hyderabad is used to train ML algorithms in order to implement STLF. ML algorithms based forecasting models are developed using MATLAB code through cloud computing. Online forecasting is more sophisticated and effective because of its ability to use recent data logs for training and forecasting online. Online forecasting is useful in Online Home Energy Management Systems (OHEMS) for effective energy management. ML algorithms such as Linear Regression (LR), Support Vector Machines (SVM) for regression, Ensemble Bagged (EB) regression, Ensemble Boosted (EBo) regression, Gaussian Process Regression (GPR) and Fine Tree (FT) regression are implemented on the cloud to forecast the power consumption. Performance parameters such as RMSE, MSE and MAE are derived to evaluate the effectiveness of the ML algorithms implemented. Cost effective Arduino Uno, Node MCU/ESP8266, PZEM 004T and DHT 11 sensors are used to fabricate the hardware model in order to acquire the load data for the proposed load forecasting approach. Best suited ML algorithm is suggested for the proposed online forecasting with supporting results.