A novel harmonic current detection method based on the adaptive neural networks is proposed in this paper to improve the quality of the power station. The harmonic parameters are estimated through the adaptive measurement theorem. It can predict the future time of harmonic currents according to the current data and the former historical data, achieving the harmonic in real-time and with fast detection. Simulations are conducted on the PV Power stations in a certain area. The results show that the algorithm achieves high accuracy and rapid speed in convergence and is a good candidate for measuring the harmonics with asynchronous sampling and short data in a grid-connected power plant.