Tropical cyclones are severe weather events which have massive human and economic effect, so it is important to be able to understand how their location, frequency and structure might change in future climate. Here, a lightweight Deep Learning model is presented which is intended for detecting the presence or absence of tropical cyclones in running numerical simulations. This model has been developed to investigate the avoidance of saving vast amounts of data for analysis by filtering data during simulations so as to save only relevant data. Subsequent analysis workflow can target that data, avoiding the need to save all simulation outputs for cyclone analysis. The model was trained on ERA-Interim reanalysis data from 1979 to 2017 and the training concentrated on delivering the highest possible recall rate (successful detection of cyclones) while rejecting enough data to make a difference in outputs. When tested using data from the two subsequent years, the recall rate was 92% and the precision was 36%. For the desired filtration application, if the desired target included relevant meteorological events, the effective precision was 85%. The recall rate compares favourably with other methods of cyclone identification having the best Area Under Curve for the Precision/Recall (AUC-PR) and using the smallest number of parameters for both training and inference.