부산대학교 도서관

In recent years solar energy penetration in local grids is increasing, resulting in a reduction in reliability, so smart grid planning is required to improve grid reliability and leverage the grid's capabilities. Due to the increasing, no of solar power plants, day by day in the energy sector, and reduction of dependency on fossil fuels, prediction of solar power generation is necessary for future planning for smart grid integration. Forecasting techniques are employed to tackle a wide range of problems, such as renewable energy generation, electric load and price, demand-side management, the financial & banking sector, healthcare, stock market, and cyberspace security prediction, etc. In the time-series dataset, there is information related to the time that can be used to predict and analyze data. Machine Learning algorithms such as Facebook (FB) Prophet and Extreme Gradient Boost (XGB) are used for predicting solar energy generation on a monthly and weekly basis. From this proposed research, it concluded that the XGB model is efficient to forecast in terms of better prediction and better fitting than the FB prophet model. RMSE, MAPE, and MAE parameters are calculated to check the performance of the time series model.