This paper explores the methodologies and results of an experiment to transform Vietnamese natural language queries into SQL statements. The paper overviews existing Text2SQL models, including state-of-the-art architectures such as T5, GPT, and BERT. These models have demonstrated the ability to transform natural language into SQL with high accuracy, but still face some challenges in handling the semantics and context of the query. This study focuses on developing an effective transformation model and analysing the unique challenges of Vietnamese, a language with a different grammatical and syntactic structure than other languages. The paper also proposes a specific transformation model, combining language preprocessing techniques, a T5-based core model, and postprocessing methods to optimise the generated SQL statements. The transformation process is detailed, from input analysis to generating the final SQL statement. Experimental results and evaluation of the test model show that the proposed model can convert Vietnamese queries to SQL with high accuracy and point out future development directions, including expanding the dataset and improving the ability to handle complex cases in the future.