Photovoltaic Power Forecasting Using a Hybrid CNN–Transformer–BiLSTM Neural Network

This dataset consists of high-frequency time-series operational data recorded at 15-minute intervals from January 1 to January 7, 2019, for a photovoltaic (PV) power generation system. The data encompasses two primary categories of variables: environmental and system performance metrics. Specifically, it includes environmental and meteorological parameters such as module temperature, ambient temperature, atmospheric pressure, humidity, global horizontal irradiance, direct normal irradiance, diffuse horizontal irradiance, along with the target output variable—actual power generation. In terms of data collection methodology, this is a field-measured dataset acquired automatically via a sensor network. The temporal continuity of timestamps and the synchronous recording of environmental conditions and power output indicate that the data were collected by a monitoring system installed at the PV plant, capable of real-time, automated acquisition of both system operational status and ambient environmental conditions. The relationship between this dataset and the present study lies in its role as a foundational resource for modeling in photovoltaic power forecasting or performance analysis. The environmental variables—particularly various components of solar irradiance and temperature—are critical predictors influencing PV power output, while the "actual power generation" serves as the target variable for model training or the ground truth for validation. Consequently, this dataset can be employed to train and evaluate machine learning or deep learning models—such as regression algorithms or time series models—to quantify the impact of environmental factors on power generation efficiency or to achieve accurate short-term forecasting of power output. Such capabilities are of significant research value for optimizing the operational efficiency of PV plants and facilitating stable integration into the power grid.