Wind energy is one of the least cost methods of electricity generation, produces no carbon emissions, and is highly scaleable. However, the intermittency of wind and the passive reaction of wind turbines means spinning and scheduled reserves are required to ensure grid security. It is important for power system planning that temporally and spatially accurate models of wind power are formulated. While the simulation of wind speed time-series is well defined the transformation from wind speed to wind power time-series is less so. Wind power plants comprise arrays of wind turbines and the individual wind turbine power time-series are not independent; thus a temporally consistent model for the spatial correlation of the wind resource is required. The spatial correlation of the wind resource can be separated into steady state and dynamic factors. Here the steady state factors are modelled using the Measure Correlate Predict algorithm and the dynamic factors modelled using the Sandia method. However, the Sandia method does not accommodate the heteroskedasticity inherent in turbulence thus a method using Wavelet Multi-resolution Analysis is developed and validated against measurements made at a wind power plant in New Zealand. The Sandia and Wavelet Multi-resolution Analysis models are numerically complex and not suited to simulations involving large number of wind power plants hence the equivalence to a first order low pass filter is demonstrated.