Predicting PV power generation from weather data

Abstract

In recent history, large‐scale solar photovoltaic (PV) systems have established themselves as an economically viable solution to offset fossil‐fuelled generated electricity and these are increasingly being integrated into district energy supply systems. However, solar generated electricity is intermittent by nature and is highly dependent on weather conditions, which results in an inconsistent supply. Therefore, it becomes pertinent to add another layer of energy generation planning and management of electrical power distribution systems through the integration of PV power generation forecasting tools and machine learning. This final year project analyses different machine learning approaches, and identifies the most appropriate one for forecasting solar electrical production using time‐series data of weather and PV power generation data. The tool is then further developed and applied for a case study to validate its prediction outcomes. We have identified several baseline machine learning and deep learning from literature and have investigated their applicability to this project. Machine learning and deep learning models can adapt to intense data fluctuations, highly dimensional and lowly correlated data. Research has shown that the most applied architectures for PV power generation forecasting is the Long Short‐Term Memory (LSTM), Convolutional Neural Networks (CNN) and a hybrid CNN‐LSTM. These architectures can learn the order dependency in time‐series data. We have also investigated the performance of these models and compared the results during the course of in this dissertation. In order to validate this task, we carried out a case study based on a publicly available Photovoltaic Output Database (PVOD). This database consists of several parameters such as date and time, global horizontal irradiance, direct normal irradiance, temperature, humidity, wind speed, wind direction, atmospheric pressure, and PV power generation. Certain parameters such as solar energy, cloudiness, temperature, and wind speed have a greater impact on PV power generation. Different scenarios of combined combinations of parameters were studied such as seasonal and diurnal variations in weather parameters and their impact on the accuracy of the developed prediction models. This project demonstrated the extent to which prediction models can be integrated as an additional layer in forecasting of energy supply and demand in large energy distribution systems, thus leading to better preparedness of the electricity distribution operator and contributing to optimal efficiency of the electrical distribution system.