Predicting Household Energy Consumption in Smart Grid Based on Seasonality Using Stochastic Markov Chain J.Jasmine Christina Magdalene 1 , Smitha Evelin Zoraida 2 1 Asst.Professor, CA,Bishop Heber College, Tiruchirappalli, India 2 Asst.Professor, CSE, Bharathidasan University, Tiruchirappalli, India jjasminebhc@gmail.com 1 ,b.s.e.zoraida@gmail.com 2 Abstract- In today’s scenario the consumption of electricity is ever increasing and the electric grids are being upgraded to smart grids. A smart grid is an enhanced version of electrical grid where digital technology is used to communicate between utilities and consumers. This helps the users to manage the consumption of energy in a better way. Seasons play an important role in the consumption of energy as the home appliances are used based upon seasons. To achieve better energy management there arises the need for prediction. In this paper an attempt is made to use Markov Chain for the prediction of energy consumption during various seasons in a smart grid. Markov chain is one of the stochastic process which is used when random variables are considered. Since the energy consumption during various seasons are random, Markov Chain is used for prediction. The dataset is taken from Pecan Street, Austin, USA and the energy consumed from the grid is combined based on various seasons. The seasons in Austin namely winter, spring, summer and autumn are taken as the state space and a transition matrix is built based on the energy consumed during these seasons. After building the transition matrix, the future data is simulated and by using Markov Chain the energy consumption during various seasons for the forthcoming year is predicted.In this paper prediction is done for five years based on the present state and the RMSE is calculated. The loglikelihood of using Markov Chain is compared with Markov Bayesian and Markov Bootstrap models. Markov Chain has higher loglikelihood indicating that this gives better results for the dataset taken. Keywords- Smart Grid, Energy Management, Stochastic Process, Transition Matrix, MarkovChain. I. INTRODUCTION Electric energy has become one of the highly prioritized requirements in today’s world. All the latest technologies work only with the assistance of electric energy. The consumption of energy is increasing day by day as the usage of smart devices increases in a sturdy pace. Electrical energy can be obtained from various sources like water, air, sun and wind. Usually houses get electricity regularly from the electrical grid and lately smart grids are preferred as an advanced and upgraded grid to effectively endure the task. A smart grid is a digital technology that allows two- way communication between the utility and its customer [1]. This two-way communication helps the consumers to know their exact usage of current and the amount of energy consumed by each smart appliance in their home. So, Energy management has become more vital as it helps in reducing price and to get continuous energy supply at the time of outages. The excess energy can be given back to the grid and the price for the energy given can be obtained. Hence there arises the need for prediction. Energy prediction depends on various parameters like seasons, usage of energy by various appliances in the home, infrastructure of a home, renewable energy resources etc., Since most of the parameters are fluctuating and cannot be deterministic prediction becomes a challenge. Many researches show that traditional forecasting techniques are used for prediction. This paper aims in predicting the consumption of energy during various seasons for the forthcoming year. Since there is uncertainty in the consumption of energy during various seasons, a stochastic process will prove much better than traditional prediction algorithms like neural- network, SVM and ARIMA. The traditional ARIMA algorithms lack incremental learning mechanism. The parameters are learnt by the model and used as the same for all the future data [2]. A stochastic model represents a situation where uncertainty is present [3]. The Markov-Chain technique is used to predict the future based on the Studia Rosenthaliana (Journal for the Study of Research) Volume XII, Issue II, February-2020 ISSN NO: 1781-7838 Page No:93