IMPROVED LOAD MANAGEMENT ALGORITHM FOR EMU/ HEMS USING MACHINE LEARNING ALGORITHMS

This paper introduces a Machine Learning (ML) based load management algorithm for EMU/ HEMS, at a domestic load center in smart micro grid environment. The algorithm proposes prediction of threshold power (PTh) for the hour by implementing Demand Side Management (DSM) Techniques. Demand Side Management Techniques are very renowned techniques used in energy management. In this work peak clipping and load priority DSM Techniques are adapted on historical power consumption data of a domestic load center. Objective function for threshold power (PTh) is formulated by applying load priority and peak clipping techniques alternatively. Simulated historical load data of residential house is considered for the implementation of algorithm, from Birmingham municipality, Alabama, USA. ML based prediction model is implemented for one-year load data to predict threshold power PTh for the hour. The predicted/scaled values of PTh data set is used for the implementation of load management algorithm. The main goal of load management algorithm, is to transfer the loads between conventional and nonconventional power supply at domestic load center. Which is effectively implemented by using the ML based predicted PTh values of the hour. This paper presents two ML algorithms such as Bayesian linear regression and Neural Network regression in predicting the threshold power (PTh). These two ML algorithms are compared with their detailed performance indexes. Increased accuracy and reduced absolute error of ML algorithms based PTh values, improves the accuracy of EMU/HEMS systems at residential load center to increase the penetration of non-conventional energy sources in smart grid environment.