Fault classification and identification through machine learning approaches for a solar PV – battery based water pumping system

Sowmmiya U,Abinaya K,Rishik Jhunjhunwala,Prateek Satpathy,Arul R
DOI: https://doi.org/10.1007/s11042-024-19593-y
IF: 2.577
2024-06-13
Multimedia Tools and Applications
Abstract:The world progresses towards enabling renewable sources into the mainstream supply of energy and it is imperative to develop systems that can handle new challenges and disturbances. This paper aims at machine learning model-based fault identification and classification of an islanded Solar PV – battery integrated system feeding a water pumping system. The system is powered through solar PV and battery (discharge mode) which are controlled through Semi Pilot Cell based Fractional Open Circuit Voltage (SFOCV) method and conventional voltage control method respectively. The water pumping system is simulated using MATLAB which is interfaced with RT-LAB software for extracting the real-time data from the OPAL – RT Controller (OP4510) through Software-in-Loop (SiL) approach. All possible faults associated with a stand-alone solar PV-battery system are aggregated to execute the machine learning model. This data is pre-processed and the distribution plots are observed to describe the statistical nature of the data. The processed data is then used to train a Logistic Regression (LR) Model and Artificial Neural Network (ANN) Model to identify whether there is a fault or non-fault condition (Binary Classification) and then classify which fault case has occurred (Multi-class Classification). The models are trained and validated in the controller. The resulting comparative parameters such as accuracy and kappa values indicate that ANN model proves to be a better solution for fault identification and classification. Also, the cost functions (error) seem to be less in ANN model and the comparative results are discussed in this work. The proposed intelligent, fault identification and classification model being implemented with real-time data ensures reliability and immediate indication to the stakeholders on the status of the fault in the system.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering
What problem does this paper attempt to address?