GA-BP based daylight prediction and sensor placement in residential areas

Abstract

This paper conducts a comprehensive analysis of the light environment in different areas of residential houses located in Beijing’s Guangyuanli neighborhood. The main objective of this study is to assess the contribution of natural daylight to residential illuminance, with the aim of providing a foundation for reducing energy consumption and enhancing light comfort. While previous research predominantly focused on sensor deployment in single scenes such as offices and classrooms, these approaches are inadequate for multi-scene analysis of daylight in residential areas. Moreover, traditional deployment methods can negatively impact the quality of life and everyday living experience. To overcome these limitations, a prediction model is developed to estimate ceiling and window illuminance, as well as work surface illuminance in various scenes within residential buildings. The Dialux software is employed for data simulation, and a GA-optimized BP neural network is utilized to train the model, establishing the mapping relationship between input variables and corresponding illuminance outputs. By inputting sensor values from different scenarios into the model, the predicted daylight illuminance values can be obtained. This predictive approach facilitates the evaluation of light comfort and energy consumption reduction in different functional areas of residential buildings.