Designing solar power generation output forecasting methods using time series algorithms.

• This work focuses on the PV power output forecasting using time series algorithms. • The input datasets are selected PV data from north region of South Korea. • Feature selection for time series algorithms is based on different seasons and climate changes. • PV power generation forecasting by different AI models has carried out on an hourly basis to test efficacy. • LSTM model shows the lowest error rate for quick PV power generation forecasting. The present photovoltaic (PV) power generation systems are globally facing the irregularity problem in the distribution of PV generation. In particular, the exact PV power forecasting is critical for grid-connected photovoltaic (PV) systems under unwanted changes in environmental circumstances. The grid energy management, grid operation and scheduling are important factors to forecast the PV power output. Time series analysis is one of the most important aspects of PV output prediction, especially in places (in South Korea) where past solar radiation data or other weather parameters have not been recorded. In this paper, a variety of time-series methods including deep-learning algorithm and machine learning algorithms was used to predict the PV power generation output for quick respond to equipment and panel defects. For designing AI models, the input data were characterized by dividing seasons and choosing the multiple parameters from seasons. In this study, the photovoltaic power generation data was collected from Ansan city, South Korea during January 2017 to June 2021 and the weather data was collected from Suwon city, South Korea during January 2017 to June 2021. In this work, approx. 40,000 hours of operation data from 1.5 MW grid-connected PV system in South Korea was used. PV power generation forecasting was carried out on an hourly basis to test efficacy of various models. Among all models (Holt-Winters, Multivariate Linear Regression, ARIMA, SARIMA, ARIMAX, SARIMAX), LSTM model presented the lowest error rate as compared to other models for quick PV power generation forecasting. [ABSTRACT FROM AUTHOR]