Accurate photovoltaic (PV) energy yield forecasting is an essential and cost-effective tool for the efficient integration and management of the electricity grid and for solar energy trading. The scope of this paper is to present a novel methodology for the implementation of a solar PV energy yield forecasting tool that provides hourly-averaged day-ahead PV power generation forecasts, based on data-driven machine learning techniques and statistical upscaling and aggregation algorithms. More specifically, the implemented tool in the form of an online software platform, comprises of a data quality validation functional block, a data-driven power output model and an upscaling aggregation clustering stage. Finally, the accuracy of the implemented tool was validated against simulated actual PV generation estimates for the entire installed capacity PV systems at the distribution system of Cyprus. The tool demonstrated high forecasting performance accuracies by yielding normalised root mean square error (nRMSE) of 8.34% and a mean absolute percentage error (MAPE) of 6.93%, over an evaluation period of 90 days.