A Review on Daylighting Prediction by Using Artificial Neural Network Techniques

Abstract

Daylighting is a key renewable strategy in sustainable building design, reducing energy use while improving visual comfort and productivity. However, predicting daylight performance is complex due to nonlinear interactions among many variables, limiting traditional methods such as physical models, equations, and simulations. Artificial Neural Networks (ANNs) have emerged as an effective alternative. This paper reviews ANN applications in daylight prediction and proposes a framework for algorithm selection, evaluation, and optimization, along with future research directions. The aim is to support more effective use of Artificial Neural Networks in building energy efficiency and luminous environment design. The advantages of using artificial neural networks for daylighting prediction include three aspects: Firstly, the prediction accuracy of ANNs is significantly better than that of traditional empirical models, enabling effective handling of complex nonlinear relationships within daylighting systems. Secondly, ANNs have high computational efficiency, which allows for a rapid real-time response after training, making them suitable for dynamic daylighting control and large-scale design optimization. Finally, the strong coupling capability of ANNs facilitates the integration of multi-dimensional variables including building geometry, meteorological conditions and occupant behaviours, thereby supporting the coordinated optimization of lighting and HVAC systems. In this paper, the state-of-the-art of the application of ANNs in predicting daylighting performance, which covers solar luminance and illuminance, daylighting control scheme and energy saving strategies, is presented. Strengths by using ANNs are highlighted and evaluated. Moreover, the research gaps have been identified and discussed. Further improvement for accuracy of ANNs and its application in daylighting study are suggested.