Evaluation of A Data Analytic Based Anomaly Detection Method for Load Forecasting Data

Grid operation relies on accurate short-term load forecast, and therefore, can become vulnerable to cybersecurity issues. The cyber adversary, once breaches the forecasting systems, may launch coordinated cyberattacks to covertly tamper with essential forecasting data including time series load and meteorological data and/or forecasting models. Detection and mitigation of data anomalies induced by such cyberattacks are more difficult. A previously introduced data analytics based method (DABM) using the so-called "Symbolic Aggregation approximation" (SAX) to detect abnormal patterns is further developed and its detailed evaluation is performed for forecast data compromised using different cyberattack templates. It is also demonstrated that attacks on weather data such as temperature may also be detected indirectly by applying the DAMB. A mitigation strategy is developed upon the detection of anomalies for a cybersecure forecasting scheme.