Energy system simulation tools are important to assess and plan the energy transition to future power systems with large integration of renewables. The calibration of power balance models, to assure that they accurately reproduce the energy system, is usually made by the reproduction of a single year past time series. This makes the calibration sensitive to that year's particular conditions, such as water availability to generate hydropower, which may jeopardise the adequacy of simulations of scenarios with different environmental conditions. This paper proposes a multiyear calibration method using linear regressions to obtain the relevant calibration parameters. To validate the method, calibrated EnergyPLAN simulations of two heterogeneous historical years in Portugal are used to compare the standard and the multiyear calibrations. Results show that the multiyear calibration allowed significant improvement in the model accuracy, showing up to 3.8 times lower errors and up to 11.2 times lower standard deviation, enhancing the confidence on the simulation of future scenarios. (C) 2018 Elsevier Ltd. All rights reserved.