Matrix based Univariate and Multivariate Short Term Load Forecasting for Power System

Short term load forecasting (STLF) aims to predict system load over an interval of one day or one week. The major power system operations like unit commitment, scheduling, load flow calculation, and security assessments are mostly relies on the accuracy of STLF. An accurate electric load forecasting is an essential part of the smart grid for smart generation scheduling. Evolving smart grid reduces the dispatching time, earlier it was day ahead now it reduced to five minutes in most of the industries. This demands entire scheduling calculation within minutes, thereby ultra-fast forecasting is required. The conventional complex models need large quantum of training data thereby processing time. There are two ways to reduce the processing time; selection of models with less training data and ultra-fast models. This paper proposes matrix based linear regression which uses similar load curves for model formulation and simple matrix operations are using for forecasting, which increases speed of operation. The major challenge is the selection of similar load curves, this paper selects previous days data as similar load curves. Results obtained from these models shows that the proposed models have strong capability to predict the load in real-time short term duration and model accuracy can be further enhanced by considering factors affecting load.