The increase of distributed Photovoltaic (PV) generators in the power system changes the load demand pattern as some of them are located behind the meter (BtM). As known as BtM PV generators, the impact on netload demand pattern increases the uncertainty of scheduling and balancing of the power system. The distortion of load demand by BtM PV must be decoupled to make an accurate forecast model for the load. In this paper, the impact of BtM PV on stochastic forecast error (SFE) is analyzed and verified how they change the efficiency of microgrid operation. The net load pattern and solar irradiation data are collected and used to decouple BtM PV distortion by using K-mean clustering and support vector machine regression analysis. Then, the baseload demand and solar irradiation forecast model is built using recurrent neural networks with external input parameters. Using the forecast model, the correlation between distributions of forecast and forecast error is built as a joint probability distribution. The SFE is modeled as conditional forecast error in a form of a probability density function and used for microgrid scheduling optimization. Case studies are based on data acquired from global data assimilation and prediction system (GDAPS) and local transmission system operator (TSO) at Jeju island in Korea.
ASJC Scopus subject areas
- Renewable Energy, Sustainability and the Environment