Real-Time Prediction on Power Efficiency of Photovoltaic Thermal System with Panel Cooling Technology using Artificial Neural Network

Ng, Cer Tat; Kamarudin, Muhammad Raihaan; Idris, Muhammad Idzdihar; Zainudin, Muhammad Noorazlan Shah; Muhammad, Sufry

doi:10.22059/jser.2026.403851.1648

Real-Time Prediction on Power Efficiency of Photovoltaic Thermal System with Panel Cooling Technology using Artificial Neural Network

Articles in Press

Document Type : Research Article

Authors

Cer Tat Ng ¹

Muhammad Raihaan Kamarudin ²

Muhammad Idzdihar Idris ³

Muhammad Noorazlan Shah Zainudin ²

Sufry Muhammad ⁴

¹ Fakulti Teknologi dan Kejuruteraan Elektronik dan Komputer, Universiti Teknikal Malaysia Melaka, Melaka

² Fakulti Kecerdasan Buatan dan Keselamatan Siber, Universiti Teknikal Malaysia Melaka, Malaysia

³ Fakulti Teknologi dan Kejuruteraan Elektronik dan Komputer, Universiti Teknikal Malaysia Melaka, Malaysia

⁴ Fakulti Sains Komputer dan Teknologi Maklumat, Universiti Putra Malaysia, Malaysia

10.22059/jser.2026.403851.1648

Abstract

Active cooling typically provides higher thermal management efficiency than passive methods. However, its continuous power consumption reduces the net energy output of photovoltaic (PV) systems. To address the limitations of traditional fixed-threshold cooling approaches, this work introduces an adaptive ANN-based hybrid cooling strategy capable of autonomously selecting the optimal cooling mode in real time. A hybrid PV cooling system integrated with Internet of Things (IoT) monitoring is developed, where a Feed Forward Neural Network Cooling System (FNNCS) is trained using real-time environmental and operational data to predict the required cooling power and intelligently choose between water- and air-based cooling. Experimental results show that the proposed FNNCS improves PV electrical performance by an average of 3.0% compared to an uncooled panel. The system achieves a maximum reduction of 14.1 °C in the backside temperature of the PV module. In addition, by dynamically adjusting cooling activation based on irradiance and temperature conditions, the FNNCS decreases cooling power consumption by 35.7% relative to a fixed cooling strategy. These findings demonstrate the effectiveness of the ANN-based hybrid cooling approach in enhancing PV performance while reducing auxiliary energy usage.

Keywords

Photovoltaic

Cooling System

Feed Forward Network

Real Time Prediction

Internet of Things