<p>The quality of the models used in dynamic simulation has fundamental importance for studies on the operation and planning of electric power systems (EPS). A typical problem is the identification of incorrect parameters. This work proposes a methodology to identify errors in model parameters, using a systemic validation approach. It uses machine learning (ML) as its main tool. Although the choice of systemic validation approach is due to its comprehensive strategy due to its ability to evaluate models in a single process, the choice of using ML is appropriate because it emulates the same analyses performed by a human expert with the advantage of processing, comparing and classifying a much larger amount of data. From the generation of cases using the electromechanical transients analysis software (ETAS), two databases were created for later training of the ML: one containing errors and the other without any type of parameter errors. With appropriate data processing steps, such as signal filtering and dimensionality reduction, experiments on a fictitious 68-bus system demonstrate that the application of the new technique allows not only to check for errors in physical model parameters with great accuracy, but also to identify which of them are incorrect and the region with problems. In this way, the methodology allows to complement the techniques present in the literature, as an initial step in selecting the equipment or subsystems that require attention, especially for large-scale EPS.</p>