The use of dynamic wind farm flow models is beneficial for power reference following wind farm control. However, currently investigated flow models are non-linear and computationally expensive, while common control approaches require fast, linear models. This work presents a novel wind farm operation modelling approach named the Dynamic Flow Predictor. The Dynamic Flow Predictor was developed with the objective to provide predictions of wind speed and turbine power using a computationally effective, linear, dynamic state space model. The model estimates wind turbine aerodynamic interaction using a linearized engineering wake model in combination with a delay process. Simulations of two turbines and eight turbines in SimWindFarm show that the Dynamic Flow Predictor can provide accurate estimates and predictions of wind turbine rotor effective wind speed and power. Additionally, the Dynamic Flow Predictor is computationally effective as it requires only 5 % of the states of a comparable, dynamic 2D CFD model. The presented modelling approach is thus well suited for the use in wind farm control, while it is envisioned that the model can also be useful for wind turbine control and as a virtual wind turbine sensor.