An Evaluation of Meteorological Variables Impact on Photovoltaic Power Generation Estimation Based on Deep Learning Model

Accurate forecasting of photovoltaic (PV) power generation is essential for efficient grid integration and operation of solar energy systems. This study assesses a PV station in Scotland, UK, by analyzing the correlation between meteorological variables and power output. Two methods are employed: Pearson Correlation Coefficient and deep learning forecasting using Long Short-Term Memory (LSTM) networks. The study uses four datasets, each incorporating a different meteorological variable alongside historical PV power data. Evaluation metrics include Mean Squared Error (MSE), Normalized Root Mean Squared Error (NRMSE), R-squared (R2), and Mean Absolute Percentage Error (MAPE). The analysis reveals that global total irradiance has the strongest positive correlation (0.92) with PV output, followed by temperature (0.66). Wind speed and relative humidity show negative correlations (−0.35 and -0.74, respectively). The LSTM model, trained with a 10-day input window, demonstrates the highest coefficient of determination with temperature range variable (99.48<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>) and the lowest NRMSE (1.77<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>). These findings highlight the significance of meteorological factors in optimizing PV power generation in Scotland.