In silico cardiac action potential (AP) models are increasingly used to describe the electrophysiology of action potentials and to predict the response of cardiomyocytes to various perturbations, especially to inhibition of ionic currents by drugs. Experimental data for model development can be recorded with sharp electrodes in adult human cardiac trabeculae with various perturbations, e.g., multiple drug concentrations. The stability of the electrode's position can not always be maintained over the entire duration of such experiments, leading to spontaneous changes in the recorded voltage. Therefore, subsequent data cannot be used for analysis and model development. In this study, we explored the normalisation (such that 0 < V < 1) of APs recorded with sharp electrodes to reduce the impact of electrode instability on data quality. Normalised APs remained unaltered despite spontaneous changes in the recorded voltage, thus they are still usable for model development. Furthermore, it was verified that normalising AP traces used for model development did not impact significantly the model's predictions, in the example case of the Ten Tusscher & Panfilov 2006 model fitted to a synthetic (simulated) dataset of normalised APs. As a conclusion, normalisation of APs increases the effective size of datasets obtained from sharp electrode measurements in human ventricular trabeculae, without adversely compromising the identifiability and accuracy of inferred model parameters. These findings also suggest that the electrophysiological activity of the recorded cardiac cell was not significantly affected by the electrode's instability.