PV Generation Short-term Forecasting Using Grey Theory

With the increase in demand for grid-connected PV systems,need of energy system forecast is also increasing for better planning,operations and management of the electric grid.This research concentrates to achieve high accuracy for short-term PV generation forecasting using grey theory.As in the past many scholars have implemented different types of models but achieving accurate forecasting results has always been a challenge in forecasting field.Initially different types of models are analyzed in order to select correct and suitable one.As grey model GM(1,1)being the foundation,it is then merged with Markov Chain model called Grey-Markov model.To increase forecast consistency,the results of grey forecasting were altered and implemented in forecasting series which comprises of time period and probability.Although it has been used in the past for PV generation forecasting,but it has never been implied for short-term forecasting.Hence,an adequate predictive model for photovoltaic power generation was established.In this dissertation,Grey forecasting results were directly applied in proposed algorithm's relative deviation probability according to given time interval.The results from the sequence were then used to establish transfer probability matrix in order to achieve transfer probability of the relative sequence.To test and evaluate the model's forecast accuracy,a 250kW grid connected PV power plant has been used which is modeled on of the modeling tool provided by NREL called System Advisor Model(SAM)which uses Sandia modeling.And the location chosen for this paper is the city of Houston,Texas as this city benefits from all types of weathers throughout the year.In this paper real time data has been harvested directly from National Renewable Energy Laboratory(NREL)of USA.The values of data includes real time irradiance levels,sunshine hours,temperature,wind speeds,and PV generation output which in the end of the paper will be used for comparison with forecasted result to evaluate accuracy rate.Three typical cases are selected to inspect the tendency of the forecast engine to work while forecasting accurate results.After achieving forecasting results,two types of error index are implied on final output to check the stability of Grey-Markov forecasting accuracy.Proposed model showed better performance than other methods(RMSE%<2%)and(MAE%<1.5%).The results did prove higher accuracy than typical Grey theory making Grey-Markov model in favor of short-term forecasting.