With increasing worldwide wind generation capacity, efficient wind power integration into the electrical grid becomes important. The intermittent nature of wind generation makes it challenging, and transmission capacity constraints add a major level of complexity since with congestion, wind generation at one node may not be the same as wind generation at another node. When multiple wind farms are located at different nodes in the transmission network, the complexity increases drastically. In this paper, wind generation is formulated at the node level using discrete Markov processes, and is integrated into the nodal demand. The Markov property reduces the number of realizations of wind uncertainty over time, compared with the stochastic programming based on scenarios. To overcome the complexity because of multiple states of multiple wind farms, power flows are formulated using voltage phase angles, assuming that the power flows are dominated by the states of the two nodes connecting the line. Then the resulting flow imbalance at each node is handled by setting aside generation and transmission capacities. Numerical results of two examples demonstrate the efficiency and scalability of the new method.