The calorific value of solid waste is a vital parameter for the process design and techno-economic analysis of waste-to-energy thermal systems. However, existing predictive models are often constrained by limited dataset sizes and restricted geographical applicability. In this study, we developed calorific value prediction models using an experimental dataset from China, employing linear regression, random forest, support vector regression, extreme gradient boosting, and a novel automated machine learning (AutoML) approach. The results indicate that the AutoML model demonstrates superior robustness, achieving a minimum mean squared error (MSE) of 0.934. To further enhance the model's accuracy and generalization capabilities, we incorporated coal calorific data from various Chinese regions and applied data augmentation techniques. This strategic expansion of the dataset significantly improved predictive performance, reducing the MSE to 0.562. Additionally, we utilized the feature importance metrics from the AutoGluon model alongside Pearson correlation coefficients to analyze the relationships between proximate and ultimate analysis constituents and their corresponding calorific values. The findings reveal that carbon content exerts the most significant impact on the calorific value of solid waste. Ultimately, the highly accurate prediction models developed in this work serve as valuable tools for facilitating clean energy production from solid waste.