Vulnerability Assessment of Machine Learning Based Short-Term Residential Load Forecast against Cyber Attacks on Smart Meters

Abstract

Short-term load forecast at residential house level plays a critical role in home energy management system. While a variety of machine learning based load forecasting methods have been proposed, their prediction performance have not been assessed against cyber threats on smart meters which have been increasingly reported. This thesis investigates the vulnerability of four extensively used machine learning algorithms for residential short-term load forecast against cyberattacks, including nonlinear auto regression with external input (NARX) neural network, support vector machine (SVM), decision tree (DT), and long-short-term memory (LSTM) deep learning. We use two datasets with different data sampling rates, including the REFIT dataset which collected whole-house aggregated loads at 8-second intervals continuously from 20 houses over a two-year period in the U.K, and the AMPds dataset which collected hourly load of one residential house in Canada for one year. The predication performances on these two datasets are compared by using NARX, SVM, DT, and LSTM. Four cyberattack models are investigated, including pulse attack, scale attack, ramp attack, and random attack. The vulnerability assessment results indicate the LSTM provides the most accurate prediction without cyberattacks. However, the prediction accuracy of the LSTM fluctuates when there are cyberattacks. Among the four cyberattacks, the random attack triggered the largest variations on the predication results.Keywords-Decision tree, Short-term Load prediction, Support vector machine, Time series forecasting, Nonlinear Auto Regression with external input (NARX) neural network.I would like to express my deepest appreciation to my supervisor, Dr Shichao Liu, for his precious patience and support. Dr Shichao's valuable beyond estimation encouragement and continuous feedback provided me with tremendous guidance throughout my master's journey.Throughout the harder times in the start of my journey during the outbreak of the Covid-19 crisis Dr. Shichao Liu was prominently present. He made it his priority to continuously check up on me and provided me with extensive personal and professional guidance. His critique, assistance and dedicated involvement throughout this process made this paper possible.I am grateful to all of those with whom I have had the pleasure to work with and learn from during this study.Most importantly, I could not have undertaken this journey without my parents and siblings. Their belief in me has maintained my high motivation and spirit.Lastly, I would like to extend my sincere thanks to Carleton University for their space, inclusivity, and resources.