This paper presents a systematic literature review (SLR) on the integration of renewable energy sources (RESs) into power grids through machine learning (ML) modeling. The review focuses on forecasting RES power outputs, particularly solar and wind energy, to address the challenges posed by the stochastic nature of RESs, such as solar radiation, temperature, and wind speed. Traditional methods, including physical and statistical models, are computationally expensive or unable to handle complex, non-linear data. In contrast, ML techniques, especially deep artificial neural networks (ANNs) and ensemble strategies, have shown superior performance in forecasting RES power outputs due to their ability to learn from historical data and handle large, noisy datasets. The SLR identifies deep ANNs, particularly long-short term memory (LSTM) networks, and ensemble methods as the most effective approaches. LSTM networks are effective in modeling the autoregressive nature of RES power outputs, while ensemble strategies can handle large amounts of fluctuating data. The paper also discusses the integration of forecasted RES outputs into decision-making problems, such as unit commitment, to address economic, operational, and managerial challenges in the power grid. The review covers the design of ML-based models, including data preprocessing, feature extraction, hyper-parameter optimization, and performance evaluation. It highlights the importance of selecting appropriate features and hyper-parameters to improve forecasting accuracy. The paper also discusses the current trends in ML-based approaches, the challenges in very short-term and long-term forecasting, and the use of different types of data, such as historical weather and power output data. Finally, the paper provides insights into the best practices for implementing ML-based RES power output forecasting and suggests future research directions to enhance the accuracy and applicability of these models in real-world scenarios.This paper presents a systematic literature review (SLR) on the integration of renewable energy sources (RESs) into power grids through machine learning (ML) modeling. The review focuses on forecasting RES power outputs, particularly solar and wind energy, to address the challenges posed by the stochastic nature of RESs, such as solar radiation, temperature, and wind speed. Traditional methods, including physical and statistical models, are computationally expensive or unable to handle complex, non-linear data. In contrast, ML techniques, especially deep artificial neural networks (ANNs) and ensemble strategies, have shown superior performance in forecasting RES power outputs due to their ability to learn from historical data and handle large, noisy datasets. The SLR identifies deep ANNs, particularly long-short term memory (LSTM) networks, and ensemble methods as the most effective approaches. LSTM networks are effective in modeling the autoregressive nature of RES power outputs, while ensemble strategies can handle large amounts of fluctuating data. The paper also discusses the integration of forecasted RES outputs into decision-making problems, such as unit commitment, to address economic, operational, and managerial challenges in the power grid. The review covers the design of ML-based models, including data preprocessing, feature extraction, hyper-parameter optimization, and performance evaluation. It highlights the importance of selecting appropriate features and hyper-parameters to improve forecasting accuracy. The paper also discusses the current trends in ML-based approaches, the challenges in very short-term and long-term forecasting, and the use of different types of data, such as historical weather and power output data. Finally, the paper provides insights into the best practices for implementing ML-based RES power output forecasting and suggests future research directions to enhance the accuracy and applicability of these models in real-world scenarios.