Efficient thermal performance prediction and optimization in HVAC/thermoelectric systems with artificial neural networks and non-dominated sorting genetic algorithm II

Abstract

This study aims to utilize an Artificial Neural Network (ANN) and a Non-dominated Sorting Genetic Algorithm II to predict and optimize the thermal performance of air conditioning/thermoelectric cooling systems that operate in tandem. The experimental findings reveal the lowest and highest values for the starting and target variables: the time required to run the system ranged from 0 to 180 min, and the thermoelectric cooling (TEC) operational state values ranged from 0 to 1, signifying an open or closed condition. Air temperature data ranges from 18 to 31 °C, COP values range from 0 to 2.78, and EER values range from 0 to 9.49. An essential first step in building AI models is to discover critical input elements through a correlation matrix analysis. The ANN research used the proposed architecture and made use of two hidden layers, one with sixteen neurons and the other with four. The Multi-Objective Genetic Algorithm (MOGA) and other evolutionary algorithms are very good at juggling many objectives. Air temperature should be 19.0 to 20.5 °C, COP should be 2.72 to 2.74, operating time should be 100 to 160 min, TEC should be turned on at all times, and air velocity should be adjusted between 4.0 and 4.5 m s⁻¹, according to the current study’s cooling and efficiency optimization conditions. Artificial Neural Networks (ANN) and Multi-Objective Genetic Algorithms (MOGA) facilitate machine learning and optimization. ANN is extremely useful in applications such as classification, regression, and time series prediction due to its ability to display intricate data linkages and patterns. They can also learn from large datasets, which improves performance over time.