Cyber Risk Identification and Classification-Based Load Forecasting Tool for Pandemic Situations

Abstract

Smart grid operators use load forecasting algorithms to predict energy load for the reliable and economical operation of the electricity grid. COVID-19 pandemic-like situations (PLS) can significantly impact energy load demand due to uncertainties in factors such as regulatory orders, pandemic severity and human behavioural patterns. Additionally, in a smart grid, cyberattacks can manipulate forecasted load data, leading to suboptimal decisions, economic losses and potential blackouts. Forecasting load during these situations is challenging for traditional load forecasting tools, as they struggle to identify cyberattacks amidst uncertain load demand, where cyberattacks may mimic pandemic-like load patterns. Traditional forecasting methods do not incorporate factors related to pandemics and cyberattacks. Recent studies have focused on forecasting by considering factors such as COVID-19 cases, social distancing, weather, and temperature but fail to account for the impact of regulatory orders and pandemic severity. They also lack the ability to differentiate between normal and anomalous forecasts and classify the type of attack in anomalous data. This paper presents a tool for short-term load forecasting, anomaly detection and cyberattack classification for pandemic-like situations (PLS). The proposed short-term load forecasting algorithm uses a weighted moving average and an adjustment factor incorporating regulatory orders and pandemic severity, making it computationally efficient and deterministic. Additionally, the proposed anomaly detection and cyberattack classification algorithm provides robust options for detecting anomalies and classifying various types of cyberattacks. The proposed tool has been evaluated using K-Fold cross-validation to improve generalisability and reduce overfitting. The mean squared error (MSE) was used to measure prediction accuracy and detect discrepancies. It has been analysed and tested on real-load data from the State Load Dispatch Centre (SLDC), Delhi, of the Indian National Grid.