This paper proposes a hybrid model that integrates Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) autoencoder networks for short-term photovoltaic (PV) power generation forecasting. The model aims to address the randomness and intermittency issues of solar energy, which affect power grid stability. By combining CNN and LSTM, the model extracts spatial features from time series data and temporal features from the data's dynamics, respectively. The model is tested on a dataset from a 5MW solar farm in southern UK, using both PV power generation data and weather data (temperature and solar radiation). The performance of the proposed model is evaluated using Root-Mean-Square Error (RMSE) and Mean Absolute Error (MAE), and compared with other models such as LSTM, GRU, and hybrid models like CNN-LSTM and CNN-GRU. The results show that the proposed model achieves better performance, with an average improvement of 5% to 25% in RMSE and MAE, and a significant reduction in training time by 70%. The model's effectiveness is further demonstrated by its ability to handle larger data sets and improve prediction accuracy when weather data is included. The study concludes that the proposed hybrid model is a promising approach for short-term PV power generation forecasting, offering enhanced accuracy and efficiency compared to existing methods.This paper proposes a hybrid model that integrates Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) autoencoder networks for short-term photovoltaic (PV) power generation forecasting. The model aims to address the randomness and intermittency issues of solar energy, which affect power grid stability. By combining CNN and LSTM, the model extracts spatial features from time series data and temporal features from the data's dynamics, respectively. The model is tested on a dataset from a 5MW solar farm in southern UK, using both PV power generation data and weather data (temperature and solar radiation). The performance of the proposed model is evaluated using Root-Mean-Square Error (RMSE) and Mean Absolute Error (MAE), and compared with other models such as LSTM, GRU, and hybrid models like CNN-LSTM and CNN-GRU. The results show that the proposed model achieves better performance, with an average improvement of 5% to 25% in RMSE and MAE, and a significant reduction in training time by 70%. The model's effectiveness is further demonstrated by its ability to handle larger data sets and improve prediction accuracy when weather data is included. The study concludes that the proposed hybrid model is a promising approach for short-term PV power generation forecasting, offering enhanced accuracy and efficiency compared to existing methods.