Prediction of short-term electricity production amount of solar power plant with 550 kwp installed capacity using machine learning

Today, the use of renewable energy sources such as sun and wind has become an unavoidable necessity due to the fact that they are sustainable, environmentally friendly, and renewable. Renewable energy methods, whose efficiency is increasing with the advancement of technology, have become more prevalent in recent years, bringing with them some issues. Increasing the amount of energy produced alone is insufficient, and one of the most serious of these issues is inefficient energy use. As a result, monitoring the produced energy and forecasting the production amount have begun to gain importance. For a more efficient and productive use of energy, it is crucial to forecast the amount of energy that solar power plants will produce. Forecasting the power to be produced is critical for investment accuracy, determining the optimum production amount, determining whether the required power can be reached, planning the income targets of the companies investing in the sector, and avoiding energy waste. Electricity producers in Turkey have the option of selling their energy to the government under a contract or at the price set the previous day on the spot market. Because of this, forecasting the amount of energy produced is critical for commercial purposes. In this thesis, the amount of energy expected to be produced in the short term by a solar power plant in Denizli province was forecasted using machine learning-based linear regression, random forest, k nearest neighbor, support vector machines, and artificial neural networks and compared to actual production values. Furthermore, the study employed the coefficient of determination, root mean square error, and mean absolute percentage error algorithms to assess the performance of the forecasting methods. In the thesis, it has been demonstrated that machine learning methods applied to data obtained from the solar energy power plant used for forecasting achieved high performance and made forecasting close to actual values. Furthermore, the forecasting methods developed were compared to one another. This study demonstrated that it is possible to forecast the amount of energy produced by solar power plants with a high degree of accuracy.