With the increasing wind power penetration stringent grid codes related to frequency, power quality, reactive power support and fault ride through conditions are imposed. Optimal Reactive Power Dispatch (ORPD) in wind farm level is essential for meeting the grid code related steady state reactive power requirements at the Point of Common Coupling (PCC). In this process the On-Load Tap Changer (OLTC) tap adjustments are required, to maintain the collector grid voltage profile within the limits. Due to highly stochastic nature of wind power generation, continuous regulation of OLTC tap will increase operational and maintenance cost of transformer. To address this issue, this paper propose a predictive short-term ORPD within a wind farm using a Hybrid Particle Swarm Optimization (HPSO). The current time interval reactive power settings are decided based the optimization carried over multiple time intervals, considering the corresponding wind power forecasts and Unit Adjustment Cost (UAC) of the OLTC tap movements. Analysis is carried simulating a large size practical offshore wind farm. The results show that the OLTC action can be minimized with a reasonably accurate wind power forecasts for few time step intervals.