You can train a simple classifier to check if an Excel file matches the expected structure. Put valid files in one folder and invalid ones in another, then extract structural features (e.g., number of sheets, column names, row counts, missing values) using pandas or openpyxl, and train a model (e.g., RandomForest or XGBoost).
Try tools like PyCaret, AutoGluon, or scikit-learn for quick prototyping.