Prediction of Transmission Line Power in a Scaled Down IEEE 5 Bus System Using Machine Learning Technique

Over the last three decades, liberalization has led to the reshape of global electric power industries. This swift has introduced competition among power producers and buyers, thereby achieving the aim of social benefit. Power trading is performed among the sellers and buyers through bilateral contracts or auctions. The new framework calls for enhanced communication networks and protocols, along with the deployment of Internet of Things, to share the system real-time parameter values. To maintain system stability and security, it is highly necessary to monitor the system parameters. Under hybrid market model that involves both auction and bilateral contract, the system operator regularly monitors the power flow through transmission line to avoid congestion. Generally, power flow through transmission line is analyzed based on the bus parameters. Machine learning models are found to exhibit good performance in multiple areas. There exists much research that adopts different models in power system applications especially classification of faults and prediction of fault severity. This paper attempts to develop a machine learning model that predicts the transmission line power based on system bus parameters. Multilayer neural network, decision tree, and random forest models are considered for the development and analysis. The models developed in Python are trained using the dataset obtained from scaled down hardware model of IEEE 5 bus network. The performance evaluation is done based on mean absolute error and mean square error and the model with least indices is selected.