Indonesian Journal of Electrical Engineering and Computer Science Vol. 17, No. 2, February 2020, pp. 671~679 ISSN: 2502-4752, DOI: 10.11591/ijeecs.v17.i2.pp671-679  671 Journal homepage: http://ijeecs.iaescore.com An early fault detection approach in grid-connected photovoltaic (GCPV) system N. Muhammad 1 , H. Zainuddin 2 , E. Jaaper 3 , Z. Idrus 4 1,2,3 Faculty of Applied Sciences, Universiti Teknologi MARA, Malaysia 4 Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA, Malaysia Article Info ABSTRACT Article history: Received May 26, 2019 Revised Jun 27, 2019 Accepted Jul 11, 2019 Faults in any components of PV system shall lead to performance degradation and if prolonged, it can leads to fire hazard. This paper presents an approach of early fault detection via acquired historical data sets of grid- connected PV (GCPV) systems. The approach is a developed algorithm comprises of failure detection on AC power by using Acceptance Ratio (AR) determination. Specifically, the implemented failure detection stage was based on the algorithm that detected differences between the actual and predicted AC power of PV system. Furthermore, the identified alarm of system failure was a decision stage which performed a process based on developed logic and decision trees. The results obtained by comparing two types of GCPV system (polycrystalline and monocrystalline silicon PV system), showed that the developed algorithm could perceive the early faults upon their occurrence. Finally, when applying AR to the PV systems, the faulty PV system demonstrated 93.38% of AR below 0.9, while the fault free PV system showed only 31.4% of AR below 0.9. Keywords: Acceptance ratio AC power Fault detection Grid-connected Photovoltaic system Copyright © 2020 Institute of Advanced Engineering and Science. All rights reserved. Corresponding Author: H. Zainuddin, Faculty of Applied Sciences, Universiti Teknologi MARA, 40450 Shah Alam, Selangor Darul Ehsan, Malaysia. Email: zainuddinhedzlin@gmail.com 1. INTRODUCTION The solar industry has grown rapidly over the past few years. The momentum of the growth is represented by the number and capacity of Photovoltaic (PV) system installations all over the world [1]. The effectiveness of PV system operational can be influenced by several circumstances which may results in power loss and waste [2]. Fault detection methods are significant to increase the performance, reliability and avoiding loss of income generation. Faults or abnormalities that presence in the system could be the factor that led to the low performance of the PV system. The faults which could be originated from AC or DC side should be identified in order to clarify the actual or exact positions of faults and could avoid the equipment damage and consequently the labor’s safety [3]. The ability to detect and diagnose potential failures at an early stage or before occurrence is also crucial to reduce costs associated with operation, maintenance and system downtime. Various studies of fault detection were seriously focused. These studies include the fault finding by using mathematical method diagnosis [4], evaluating performance ratio (PR), capture losses, array and grid power losses analysis [1] and also artificial neural network [5]. Numerous fault detection techniques on DC side of PV system have been applied; such as climatic data independent technique (CDI) [6], electrical current-voltage (I-V) measurement (EM) technique [7], measured and modeled PV system outputs (CMM) technique [8], power loss analysis (PLA) technique [9], Machine learning (ML) techniques [10, 11],