A Day-Ahead Power Demand Prediction for Distribution-Side Peak Load Management

Abstract

In this paper, a day-ahead load management system is proposed using the household power demand and photovoltaic (PV) power generation prediction. The prediction is made using an artificial neural network. A power demand management algorithm is developed to process these predicted values considering the boundary conditions of battery storage to flatten the peaks in a load curve. The proposed system is tested in a real power distribution network under realistic load pattern, and power demand and PV power generation uncertainties. The study found that strategic use of battery energy storage and PV, and a time-ahead prediction of power demand can substantially reduce the peaks and improve the load factor.