The models for photovoltaic (PV) systems currently in ESP-r prove very useful in estimating the electrical and thermal impact of building-integrated photovoltaics. However, while they represent well the impact of photovoltaics on the building’s thermal energy balance, they may lack in accuracy in the prediction of the system’s energy production. To achieve both goals at once it is suggested to improve the PV models in ESP-r, taking into account all phenomena affecting the power output of PV modules: solar radiation intensity, cell temperature, angle of incidence, spectral distribution, uncertainty in manufacturer’s ratings, ageing, mismatch, soil and dirt, snow, partial shading, diodes and wiring. This would provide a more realistic estimate of the probable output of the PV system over its lifetime. It is suggested to implement three models: a simple model based on constant efficiency, a one-diode equivalent model with explicit temperature dependency of the parameters, and the Sandia model for cases when detailed modeling is required.