In the smart grid era, distributed algorithms are paving their way to solve optimal power flow (OPF) problems in lesser computation complexity. In distributed algorithms, the original centralized problem is decomposed into multiple subproblems and they are solved in parallel by enabling coordination among them. In this context, the alternating direction method of multiplier (ADMM) has proved itself superior over other methods and is used extensively for power networks. However, vanilla ADMM completes a large number of macro-iterations $(\approx 10^{2})$ before convergence. Therefore, to reduce the number of iterations and the solution time, in this article the concept of predictor-corrector acceleration, proposed by Nesterov, is merged with the vanilla ADMM for solving three phase OPF problem of unbalanced distribution networks. The convergence speed is further improved by utilizing the concept of adaptive penalty. The efficacy of the proposed Nesterov-type accelerated ADMM (N-ADMM) with adaptive penalty is established by implementing on IEEE 123 bus test system. Again, two non-ideal data transfer scenarios, viz. bad and noisy data transfer, are modeled to show their impacts on the N-ADMM algorithm.