Photovoltaic Power Generation Prediction Using Artificial Neural Network Method And Its Prediction Error Control

Distributed generation such as photovoltaic(PV)in power generation plant is helpful to reduce pollution in power generation cause by using the traditional power generation plants which seriously have negative impact to environment and human’s life.In process of generating power,Solar photovoltaic panel provide clean energy(called green energy),no noise at all and they are a good solution for urban areas which far from power grid and for residential applications where there is sunlight.Typically,PV power fluctuations are counter-balanced by the use of battery storage systems.The ability to accurately predict and prevent power fluctuations is of considerable importance to solar PV power plant operators in terms of sustaining profitability,estimating revenue returns and ensuring customer quality of service.Variations in solar irradiance can cause rapid fluctuations in power generation,reducing the quality and reliability of the power generated by large grid-connected PV plants.Forecasts are an important method for managing the variability and the uncertainty of PV and should be incorporated into system planning and operations.Most studies do not consider possible power deviation between the day-ahead schedule and the actual power,which might cause a failure of the day-ahead plan.First,PV power output need to be predicted in day-ahead which achieve by using Nonlinear Auto-Regressive based Neural Network.Artificial Neural Network(ANN)method is a powerful tool amount order in field of prediction.So many researchers have presented the advantage of various type of ANN model.Nonlinear Auto-Regressive Neural Network(NARNN)is one type of ANN method which uses the past values of the time series to predict future values.The hyperbolic tangent transfer function is performed well in NARNN training.Second,a model predictive control(MPC)based real-time dispatching model(mixed integer quadratic programming)is generated to keep tracking of PV power output in the day-ahead based on active power control of battery.The rolling optimization method based on MPC is adopted in the real-time dispatching.Finally,three cases studies is presented to evaluate the performance of NARNN in PV power generation prediction and its prediction error control which can be achieved by using model predictive control by controlling the active power control of BESS.