2016 International Conference on Emerging Technological Trends [ICETT] Big Data Analytics of Smart Meter Data using Adaptive Neuro Fuzzy Inference System (ANFIS) S.M. Sulaiman Department of Computer Science and Engineering Kalasalingam University Anand Nagar, Krishnankoil, Tamil Nadu, India 626 126 Email: sul sm@yahoo.com P. Aruna Jeyanthy and D. Devaraj Department of Electrical and Electronics Engineering Kalasalingam University Anand Nagar, Krishnankoil, Tamil Nadu, India 626 126 Email: arunadarwin@yahoo.com, deva230@yahoo.com Abstract—The ever increasing human population and the as- sociated demand for electricity have challenged the power sector to modernize its equipment and operations. This renovation activity has made the existing grid to incorporate Information and Communication Technologies (ICT). Installation of Smart Meter is one of the significant changes due to developments in the power sector that establishes two-way communication between the Utility and the consumers. The Smart meters collect data at high velocity leading to tremendously huge volume of data and have been classified as Big Data. Uncovering useful information from these Smart Meter data is a Big Data challenge. In this paper, Smart Meter data is used to forecast the average electricity load for every hour on daily basis. The proposed method uses Adaptive Neuro Fuzzy Inference System (ANFIS) to predict the load ahead of 24 hours from present day meter readings. The experimental results are promising with the overall prediction accuracy of 84.02%. Keywords—Smart Meter, Big Data, Load Forecasting, ANFIS I. I NTRODUCTION E Lectricity consumption is increasing day by day due to growing global population. In order to meet the rising electricity demand, power plants are now using conventional energy sources such as coal, gas, oil and nuclear power. Most of these energy sources produce carbon dioxide (CO 2 ) as a by product. According to [1], 25.9% of the carbon emission is due to the use of high carbon fuels in these power plants. With an effort to reduce (CO 2 ) emission, many countries impose strict regulations to the power sector to generate cleaner and cheaper energy. One way to produce clean energy is the use of renewable energy sources like solar and wind that requires advanced power infrastructure which the existing grid is lagging due to constructional deficiencies. Smart Grid is the solution to solve the above mentioned problems. Addition of ICT into the existing grid transforms the current grid to Smart Grid, one that functions more cooperatively, responsively and organically. The major driving factors for today’s Smart Grid are: • Capacity: Providing sufficient power supply to meet the growing demand. • Reliability: Continuous delivery of high-quality electrical energy without any block-out. • Efficiency: Ensuring less power generation loss, transmis- sion loss and distribution loss in the entire grid system • Sustainability: Incorporating renewable energy sources into existing power grid One of the important components of the Smart Grid is the Advanced Metering Infrastructure (AMI) an integrated system of smart meters, communications networks, and data manage- ment systems which enables two-way communication between Utilities and consumers. Smart Meters offer collection of fine- grained (usually every seconds or few minutes) energy con- sumption data in an automated manner. The data accumulated at very high velocity from these smart meters evolves into Big Data. Retrieving useful information from these mountain of data is a real challenge. In fact, a new area named Meter Data Analytics (MDA) has received significant interest in the research community [2], [3]. One of the significant features in Meter Data Analytics is Load forecasting which predicts future expected power demand based on past history of electrical power consumption. Load forecasting adds intelligence to the Smart Grid which enables the Utilities to schedule the operations of power generators. The consumers also benefit from forecasts to plan their loads appropriately in the case of variable tariff based on time of power consumption. The Smart Meter technology thus leads to understanding of one’s energy consumption, demand flexibility and better choices on tariff plans at the consumer side. In order to prevent load shedding, Utilities can control the operation of high power heating and cooling loads if granted access to them by the customers. The problem of load forecasting has received wide atten- tion and there are many attempts to solve this issue since 1990 [4]. Several soft-computing techniques have been used such as artificial neural network (ANN) [5]–[8], neuro-fuzzy method [9] and fuzzy logic [10] to name a few. In addition, some researchers have also used other techniques like time series analysis [11] and support vector regression (SVR) [12]. However most of the earlier works had no access to fine grained electrical consumption data from residential buildings due to device limitations and the less frequent manual reading habits.This paper works on a high volume, high resolution (two 978-1-5090-3751-3/16/$31.00 ©2016 IEEE