Data-Driven Modelling And Optimal Scheduling Of Building Interior Light Environments

The amount of energy consumed annually by office buildings in China is rising in proportion to economic growth and urbanization,but the amount of environmental resources available is steadily declining.This significantly reduces the potential for energy development and utilization.Lighting systems,one of the most significant energy-consuming services in office buildings,frequently lead to issues like an imbalance between energy consumption and visual comfort in office buildings because it is challenging to model the interior light environment and because lighting control systems are not in sync with the spatial and temporal distribution of individuals.To achieve intelligent scheduling of lighting system,we concentrate on the lighting system in office buildings,starting from single lamp illuminance analysis and modelling,data-based interior light environment modelling method,simulation experiment platform construction,and optimal illuminance scheduling strategy based on individuals,etc.In order to fulfill occupant demand and save energy consumption,the goal is to accomplish intelligent scheduling of lighting systems based on individuals distribution.This thesis’ s core research work consists of the following:(1)An approach for modeling the interior lighting environment of buildings that is datadriven.To create a single-light illuminance prediction model for daylight-free situations,this thesis suggests a combination prediction technique based on Gaussian fitting and error compensation.On the basis of non-continuously adjustable luminance lighting devices,the method investigates the effects of light at various dimming levels.It then suggests a "multilight-multiple" illuminance prediction method based on individuals distribution without desktop illuminance sensors,allowing quick and precise acquisition of light levels at the location of individuals.The outcomes of the comparison trials demonstrate that the technique may be used to forecast an individual’s brightness.Comparative experimental findings demonstrate that,within the useful range of interior illumination,the approach successfully achieves high accuracy illuminance prediction.The DIALux evo daylight simulation program is used to create a simulation model of the interior lighting environment.To gather the data base needed for the research,over 100 trials are then simulated using the simulation platform.The validation experiment findings show that the approach is accurate and generalizable.(2)An approach for modeling interior daylight settings that is data-driven.This thesis proposes a Bayesian optimized gradient boosting regression tree for predicting interior daylight levels to address the issues of non-linear variation of natural light and machine learning techniques that heavily rely on expert tuning experience and human intervention.This method offers significant improvements in adaptive tuning,high accuracy prediction,and ease of application,and can significantly reduce sample bias values.The method investigates the effects of different meteorological conditions(sunny,cloudy),as well as time granularity attributes(unbroken hours from 08:00 to 17:00),on the light environment under typical dates of four seasons(equinox,summer solstice,autumn equinox,and winter solstice),while making sure that all of the equipment is turned off.It then builds a prediction model for daily illuminance under no lighting conditions.Unlit Comparative experimental findings demonstrate that the approach outperforms several conventional algorithms(RF,GBRT,and ANN,etc.).And it can predict the luminance at location points for various working conditions with various test samples,which can help with indirect energy savings and intelligent dimming of future lighting.(3)A distribution-based scheduling approach for interior lighting in buildings.Using a combined prediction model of daylight and lighting,this thesis presents an improved teaching and learning optimization algorithm based on Lévy flight and hybrid variational operators as well as an optimal scheduling strategy for interior illuminance based on individuals distribution.Comparative experimental findings demonstrate that,when lighting is assisted,the strategy seeks out and obtains the ideal dimming brightness combination(i.e.,the optimal decision value for the brightness adjustment amount of each luminaire)for all interior lighting devices while satisfying the uniformity of illumination requirements of both the working(surrounding)area and the personnel.As a consequence,the operation of the interior lighting system is optimized for low energy consumption,multi-temporal performance,and high efficiency.