Electricity generation prediction of wind farm using machine learning algorithms with python

The rapid increase in the world's energy needs and the inadequacy of depleted resources to meet the energy demand cause countries to increase their interest in renewable energy sources. Among renewable energy sources, wind energy comes to the fore at this stage. Its ease of installation area, maintenance and continuity, combined with its regional suitability for wind, offers new opportunities for electricity generation. In recent years, wind power plants have come to the fore in Turkey with the rapid increase in their installation, both geographically and in terms of technological developments. In this thesis, the values measured at 10-minute intervals between 16.10.2020 and 16.04.2023, consisting of 392712 rows and 27 columns, belonging to the Wind Power Plant in Akkuş district of Ordu province in Turkey, are included and the produced active power, reactive power A data set containing column values such as, wind speed, humidity, wing positions, air pressure, etc. was used. The data set was made ready for machine learning by performing data preprocessing and outlier data analysis using data mining methods. Before applying machine learning models, the model framework was determined and the values in the data set were scaled to values between 0 and 1 using the normalization. In this way, the Python programmer is enabled to implement machine learning models faster and easier. 80% of the data is divided as train (learning) and 20% as test data. Since the data set consists of non-linear, complex and numerous data, the results were compared by applying Simple RNN, GRU and LSTM models, which are machine learning models. Prediction data were compared with real data and error rates were calculated using RMSE and MAE error methods. When the results were compared, it was determined that the Simple RNN model gave more realistic results. With this thesis study, it has been shown that the electricity production of a Wind Power Plant can be predicted realistically by using machine learning methods with the help of Python programmer and libraries. In this way, daily, monthly and annual electricity production planning can be made easier. Keywords: Machine learning, Python Programming, Wind Power Plant, Renewable Energy, Data Mining, Data Preprocessing.