An investigation into visualisation and forecasting of real-time electrical consumption based on smart grid data
The smart grid, and in particular smart meters, is a growing world-wide phenomenon which has allowed for the availability of detailed real time usage data to the user in ways that were not possible in the past. South Africa has been slow-moving in adapting smart meters, but in the past two years this has changed and smart meters are becoming the new standard. This has given rise to the need for software applications to help both the South African consumer and local power utilities get the most out of the smart meter data. The purpose of this research is to investigate the possibilities offered by smart grid data obtained from advanced metering infrastructures, with particular emphasis on real time energy usage visualisation and peak load forecasting. Previously, detailed energy usage data has not been available to consumers hence there has not been much research focusing on utilising this data for direct consumer benefit. The focus of most research has mainly been on the power utilities supply side where attention has been on visualising their consumers’ usage and forecasting consumer demand in order to supply them with electricity continuously and efficiently. In this dissertation a benchmarking model for developing smart grid data visualisation dashboards is proposed and this model is used to present and prototype a consumer side dashboard. The prototype implements real time data visualisation techniques, as well as a Multiple Linear Regression model based forecasting algorithm for half hourly peak load forecasting using data collected from the University of the Witwatersrand’s advanced metering infrastructure. In this study the Multiple Linear Regression model is built through a comprehensive analysis of 2 years’ worth of energy usage data from the University of the Witwatersrand and 3 years’ worth of hourly temperature data from the South African Weather Services. The prototype’s performance is evaluated with reference to the proposed benchmark and a user technology acceptance evaluation done by the University’s Property and Infrastructure Management division. The dashboard is found to be a useful and acceptable tool in energy monitoring at the University. The forecasting model performs well with a mean absolute percentage error of 3.69%. The inclusion of a forecasting functionality within the energy management dashboard is shown to have the ability to help the university reduce its electricity bill by being able to shave their peak loads. The analysis highlights the importance of better data archiving and smart meter monitoring thereby ensuring that the meters are always online and no data goes missing which is vital for accurate forecasting results.