A novel intelligent scheme using the wavelet packet transform (WPT) and extreme learning machine (ELM) is proposed for fault event classification in the grid-connected photovoltaic (PV) system. The WPT is applied for preprocessing the cycle of the post-fault voltage samples at the point of common coupling (PCC) measurement to get the normalized logarithmic energy entropy (NLEE). The ELM is applied to classify the different fault cases. To enhance the performance of ELM for faults classification, a hybrid optimization mechanism based on an equilibrium optimization algorithm (EOA) is proposed to optimize the selection of input feature subset and the number of ELM hidden nodes. Furthermore, to evaluate the proposed scheme's performance, a comprehensive evaluation was conducted on a 250 kW grid-connected photovoltaic system. From simulation, the classification accuracy is recorded to be 100% under the no-noise condition, while at the signal-to-noise ratios (SNR) of 30, 35, and 40 dB, the accuracies are 98.96, 99.04, and 99.36%, respectively. Moreover, the practical performance of the EOA-ELM classifier is validated using IEEE 34 bus system. The obtained results validate the effectiveness of the proposed scheme in terms of robustness against measurement noise, computation time, and detection accuracy.