Assessing the Demand Response Program in a Network with High Integration of Photovoltaic Plants using Machine Learning

The paper presents the procedure for assessing future demand response program in a network with high integration of photovoltaic plants using machine learning methods. By using scenario-based probabilistic forecasting, the load probability distributions are calculated for the future year. The results derived from probabilistic forecasts are used for assessing the future demand response activation time intervals. Instead of assessing the possible annual peak decrease by computing load duration curves for a past analyzed year, here the median load duration curve was forecasted with corresponding prediction intervals. The main advantage of this procedure is providing a range of load values that could be decreased before implementing the demand response system instead of using a single value, that was estimated from the past data. The proposed procedure in this paper is a follow up to the analyses that were carried out as a part of the demand response project in the scope of the Slovenian-Japanese NEDO project. The procedure can be used by distribution or transmission system operators in order to select an appropriate network for demand response system integration.