Daily average load forecasting using dynamic linear regression

Load forecasting plays a vital role in demand management. The primary goal of demand management strategy is to shave the peak load in order to reduce the dependency on the peaking plants and to avoid the overloading of the transmission and distribution equipment. Battery storage can also be utilized for peak shaving by storing excess energy during the off-peak and consuming battery energy during peak hours. For effective battery use, the battery management system must have the accurate forecast of the load demand. This paper proposes a dynamic regression scheme to predict the average daily load of a feeder so that the battery management system can decide the amount of charging and discharging required at each instant. Forecasting of average daily load rather than point forecast of load demand at every hour avoids the complexity of battery scheduling and reduces the computational effort. This paper uses Perth solar city data to showcase the effectiveness of dynamic regression for forecasting future loads.