Power Transformer Summer Peak Load Prediction Using SCADA and Supervised Learning

Load forecasting is used to meet the demand and supply equilibrium in any distribution system. It is basically multi-variable and multi-dimension estimation problem. It helps electric utilities to make advance decisions about purchase and generation of electric power, switching of loads and infrastructure expansion, etc. This paper presents peak load prediction of a power transformer from the given datasets. Peak load data are obtained using supervisory control and data acquisition (SCADA). Curve fitting prediction is devised to calculate hourly load forecasting of the week days. However, this method does not provide accurate results as compared to machine learning (ML) algorithms. ML algorithms not only able to review large volumes of data but also provide accurate results with percentage error of only up to 10%. The load data are acquired from the DF 8000 software. To obtain the predictions, multiple regression model is applied to the dataset. A python data science platform is used for simulation work. The results obtained are presented.