Solar Power Forecasting Using Long Short-Term Memory Algorithm in Tamil Nadu State
Nallamotu Haritha, Jose Anand
Abstract
Solar power forecasting entails understanding of the sun’s path, the atmosphere’s condition, the scattering techniques and the traits of a solar plant which makes use of the sun light to generate solar energy. This system converts the solar energy into electrical energy. The output power relies upon at the incoming radiation and at the sun panel traits. Photovoltaic energy manufacturing is growing nowadays. Forecast data are vital for a green use, control of the power grid and for solar power trading. This challenge predicts sun irradiance for numerous hours for the future, primarily based totally on modern sun Irradiance and nearby climate situations by the usage of DL algorithms like LSTM model. Data supply is from TANGEDCO and additionally obtained from internet sources. Date-oriented and month-oriented data are saved in CSV files. Later the same is extracted using Pandas Library function and for the entire system Python software is used. From the ancient irradiance, information’s are collected in the form of data frames. Reading and cleaning of data is made by the support of Pandas Libraries. Pandas Library is a machine-learning library that is used to clear the data and to accept valuable and non-redundant data set. The newly derived data set is saved in the form of pickle format and the same is used for later use. The data preparation is done using Numpy and Pandas Center Library functions. All the prepared information is stored as data frames. From the obtained data frames, different varieties of manipulations are carried out to generate the results in the form of reports with diagrams and charts. The outcomes are related to the real power generated and are concluded with good prediction efficiency.