Accurate harmonic estimation is the foundation to ensure a reliable power quality environment in a power system. This paper presents a new algorithm based on a Group Search Optimiser (GSO) to estimate the harmonic components presented in a voltage or current waveform. The structure of harmonic estimation is represented as linear in amplitude and non-linear in phase. The proposed algorithm takes advantage of this feature and estimates amplitudes and phases of harmonics by a linear Least Squared (LS) algorithm and a non-linear GSO-based method respectively. The improved estimation accuracy is demonstrated in this paper in comparison with that of the conventional Discrete Fourier Transform (DFT) and Genetic Algorithms (GAs). Moreover, the performance is still satisfactory even in simulations with the presence of inter-harmonics and frequency deviation.