Load forecast anomaly detection under cyber attacks using a novel approach

In order to make essential and practical choices about the demand and supply of energy, power grid operators rely on load prediction data. Consequently, effective load forecasting is critical to achieving economic benefits. Cyber-attacked load prediction data may mislead power grid operators into making unwarranted choices about the distribution of electricity. This research has given a unique methodology for the detection and identification of cyber assaults on the electric grid's load prediction data. It has two stages. In the first phase, a benchmark is generated using historical real load data and an unsupervised machine learning model. The categorization of cyber threats using supervised machine learning models is the second phase. Finally, ensemble approaches are used to construct a novel hybrid model. The novel approach has been tested on a publicly available dataset and it produced impressive accuracy of 97.25 percentage.