Improvement of perez daylight model on the base of adaptive network based fuzzy inference systems (ANFIS)

Nowadays, electrical energy is required for devices that we use from lighting to air-conditioning, from mobile phones to computers and from houses to factories. In other words, electrical energy is indispensable in our lives. Energy is also one of the most important factors in understanding the level of development and levels of welfare of countries. Moreover, as in other countries, it is an important factor in enhancing economic and social development as well as the life standards in Turkey. Energy production is mostly achieved by fossil fuel combustion. However, the burning of fossil fuels increases CO2, SO2 and NO2 emissions which cause global warming and climate change. Due to the unfavourable effects of fossil fuels, savings in energy consumption and use of natural resources for energy generation has gained greater importance. Sunlight is the best source for all living creatures. Daylight illumination is one of the most important parameters for comfort life and high level standards because comfort conditions mostly depend on the natural daylighting. Daylight optimization with the use of renewable energy sources contributes to the increase of the potential of economically active population due to technology transfer and the mobilization of internal resources of countries. Daylight thus plays an important role in preventing energy dependence on other countries. For these reasons, energy management has become a necessity. The day-to-day technology has revealed the necessity of minimum electricity consumption by taking advantage of daylight for the sustainability of energy. Unfortunately, in Turkey and especially in Istanbul this limitless source is not used efficiently for illumination in the buildings. Especially artificial illumination engineers should know outdoor properties of cities. Besides, electricity consumption in the buildings has a significant percentage. Instead of natural illumination, typically artificial illumination tools are used and this case causes far more electricity generation and consumption. Developing natural illumination technologies would provide saving energy and conjointly healthy conditions for the human. Moreover, recently natural daylighting has been applied to many buildings and related structures. Daylight illumination is not only important for building designers but also energy analyzers. Generally, global solar radiation measurements are performed habitually by meteorological units. It is necessary to know levels of illumination as well as global solar radiation. Unfortunately, illuminance measurements in outdoor are not as common as solar radiation measurements under realistic sky conditions. One of the important gaps for illuminance research is the lack of measured illumination data in other words especially in Turkey it is not easy to reach measured daylight illumination data not only for time scale but also for regional dependence. Besides, there is no daylighting measurements and modelling around Istanbul which is one of the biggest megacities in the World. The aim of this thesis was to fill this gap of no daylight database exists. Thus, a new research has been started related with daylight illumination. The study examines the data of global horizontal daylight illumination measured in Istanbul between the years 2015-2017 and the global horizontal solar radiation data measured in and around Istanbul. In this context, as a first step, an illumination level measurement sensor (luxmeter) has been installed at the meteorological observation park of Istanbul Technical University (ITU). One-minute time interval measured daylight illumination data are analyzed in detail and compared with the measured solar radiation values between 2015-2017. Along with the installed luxmeter, solar radiation measurements from the previously installed pyranometer at the same station were evaluated simultaneously. In the next stage, measurements are taken one the central meteorological station near Bosphorus as a reference station (ITU) and after that around this station, additional solar radiation data of 7 stations are considered and correlated. If there is a relationship between daylight illuminations and any other parameter, lacking daylight illumination data can be perfectible. There is an excellent relationship between the global solar radiation measured at all stations and the illumination levels measured at the ITU station, where R2 is seen very close to 1.0 from time to time. Cloudiness and air pollution are main components to reduce daylight illumination. The mentioned components take important roles modulating the level of illumination especially in winter, spring and autumn months. This strong correlation reflected from the results allows the comparison of the solar radiation values measured at other stations with the illumination levels at ITU. Then, by using the levels of illumination measured at ITU station, it was tried to calculate the illumination levels for other stations based on the relationships evaluated. The solar radiation values were interpolated by means of simple geostatistical algorithms such as semivariograms and Kriging for the measured area or distant areas. Semivariogram is a spatial model that is used to indicate the relationship between semivariance and the distance. As a result of these findings, daylight illumination maps were prepared for and around Istanbul. It is evaluated that especially at southern parts of Istanbul daylight illumination is moderate. As a result, daylight illumination is highly proportional to the solar radiation. Therefore, the levels of outdoor illumination can be calculated depending on the solar radiation values. Nowadays, the knowledge of daylight illumination is investigated by many countries around the world. Illumination is a basic term used in photometry. Therefore, some information about photometry is required. Some basic terms about photometry and atmospheric parameters which are used to calculate the daylight illumination are explained in the second chapter. International daylight measurement program studies have been carried out in many countries and daylight models have been developed by many researchers. Although many illumination models have been developed so far, one of the most widely known and frequently referenced models, Perez et al. daylight illumination model, is quite successful. In the next part of this study, Perez model results were compared with the illumination data which were measured and at first glance, consistent results were obtained. However, when the statistical errors were calculated, results were higher than expected from a successful model. Hence, this model which is compatible with the measurement values has been developed by some improvements. Knowing the amount of precipitable water vapor, which is one of the meteorological variables, is an important step for the calculation of the daylight illumination level. As mentioned above, one of the most widely used models in the literature, the Perez daylight model uses the amount of precipitable water vapor as one of the four atmospheric variables. Instead of the precipitable water vapor expression used in the model, the equation derived specifically for Istanbul in a previous study were used. In this way, the model could reflect the characteristics of the region and the results were closer to the measurements. Besides, the cloudiness factor was examined in eight categories according to the values of sky clearness, ε, calculated in Perez model. When the coefficients used in the model determined as a result of this discrete classification are jumped from one class to another, there is no transition between the previous and the next value. In the thesis study, Adaptive Network Based Fuzzy Inference Systems (ANFIS) approach, which is a fuzzy logic based artificial intelligence method, has been adopted to soften the sharp class transitions. As a result, instead of the constant coefficients used in the Perez model, coefficients varying according to the weight of neighborhood in each class were calculated and a fuzzy model containing the concept of continuity was formed. In addition to the eight discrete classes used in the Perez model, 15, 7 and 3 classed fuzzy models and 4 and 2 discrete classed non-fuzzy (classic) models were created. In this way, instead of the classic Perez model with fixed coefficients and fixed classes, the option of using the desired model is highlighted. Finally, the results obtained from all the models were compared with the measurement values and the classic Perez model and evaluations were made according to the model errors. In the comparison of errors, the most commonly used Percentage Relative Error (RE), Percentage Root Mean Square Error (RMSE) and Percentage Mean Bias Error (MBE) values were compared. Additionally, probability of detection (POD) and false alarm ratio (FAR) are evaluated as statistical approach. Along with the improvements to the model, successful results were achieved. It is advised that daylight illumination tools immediately should be applied for outdoor and indoor illuminance applications in Istanbul to improve the level of development. This study is aimed especially at researchers interested in energy efficiency, lighting engineers and building design practitioners allowing designers to consider daylight illumination in their projects. Knowing the values of daylight illumination levels is crucial for the use of advanced natural illumination techniques in building designs, improving people's daily life comfort, energy efficiency, global climate change and energy management strategies of countries. In order to develop these strategies, it is necessary to have a database that will allow to make any evaluation about the area of interest. With the method proposed in the thesis, it will contribute to the energy saving that is desired to be made today in the interest of our country.