The feasibility of genetic programming and ANFIS in prediction energetic performance of a building integrated photovoltaic thermal (BIPVT) system
Introduction
Global energy needs grew by 2.1% in 2017, more than twice the previous year's rate, boosted by strong global economic growth, with oil, gas and coal meeting most of the increase in demand for energy (IEA, 2018). As a consequence, the worldwide emissions of climate-warming CO2 emissions rose by 1.4% in 2017 (IEA, 2018). This will cause warmer average global temperature which is now considered among the biggest challenges facing humanity. To overcome these issues, the renewable sources of energy are promising options. Renewables are growing worldwide, with a sixfold enhancement in non-hydro renewables over the last decade from 85 to 657 GW (GW) (IEA, 2017).
Solar energy is perhaps the most recognized renewable energy source. There are two types of solar systems including solar thermal and photovoltaic (PV). These systems harness radiant light from the sun, convert it to electricity and heat, respectively. Solar PV panels are the fastest-growing source of new energy worldwide (GUARDIAN, 2017). This can be attributed to their significant cost reduction over the past 35 years and the zero fuel cost. The main limit of PV systems is their poor conversion efficiency (up to 20%), which is considerably affected by their operating temperature. Besides, this poor efficiency is even lower by increasing the panel's temperature. This problem can be overcome by using a coolant over or under PV panels which not only the reduction of panel efficiency due to temperature rise can be prevented but also the heated fluid can be utilized in applications such as air conditioning or solar drying. This type of system is known as hybrid PVT systems. So far, several studies have been carried out to evaluate the electrical and thermal aspects of these systems by all means of analytical solutions (Khaki et al., 2017, Khaki et al., 2018, Shahsavar et al., 2011, Shahsavar and Khanmohammadi, 2019), experimental measurements (Pounraj et al., 2018, Otanicar et al., 2018, Kazemian et al., 2018), and numerical simulations (Hosseinzadeh et al., 2018, Mousavi et al., 2018, Shadmehri et al., 2018).
One of the attractive applications of the PVT systems is BIPVT system. Many research works have been conducted to analyze the performance of these systems. In a numerical investigation, Shahsavar et al. (2011) proposed a novel BIPVT system and investigated its thermal and electrical aspects. The system could reduce the temperature of PV panels using ambient/exhaust air during winter/summer and producing electricity. The findings revealed that the suggested system has a considerable energy recovery potential. Piratheepan and Anderson (2017) implemented a novel model to study the performance of a BIPVT system. They found that the tube spacing and thermal conductivity of PV panels have a remarkable influence on the system performance. Athienitis et al. (2018) developed a novel dynamic model to analyze both the active and passive performance of a BIPVT collector. The suggested model was based on a detailed transient finite difference thermal network. Shahsavar et al. (2018) proposed an innovative exhaust air heat recovery system consisting of a BIPVT system and a sensible rotary heat exchanger (SRHX). The novel system was capable of preheating/precooling the fresh ambient air in winter/summer and producing electric energy. They found that the heat recovery potential of the suggested system is higher than the individual BIPVT and SRHX systems.
Recently, several researchers have proposed the application of artificially intelligent systems in various fields, especially in energy systems. There are numerous successful applications of ANNs to predict the amount of energy in buildings (e.g., Asadi et al., 2014, Hong et al., 2015, Good et al., 2015, and Chae et al. (2016)). The main objective of the current study is to predict the energetic performance of a BIPVT system thorough ANN, GA, and ANFIS for the climatic conditions of Kermanshah (Iran). To the best of our knowledge, this is the first study focused on the use of artificial intelligence techniques to forecast the performance of BIPVT systems. The input parameters that implemented thorough the soft computing analysis include duct length (m), duct depth (m) duct width (m), and mass flow rate (kg/s) where the output is taken as the output.
Section snippets
System layout
Fig. 1 displays the schematic of the studied BIPVT system. It is an all-air heating, ventilation and air conditioning system with one air handling unit. This system has two modes of operation: winter mode and summer mode. During the winter mode (October to March), the cold ambient air is preheated by passing through the duct underneath the PV panels. Besides, the PV panels are cooled and their efficiency increases. In the summer mode (April to September), the building exhaust air, which is
Energy balance equations
Fig. 2 depicts the cross-sectional view of the studied BIPVT system with heat transfer coefficients. The energy balance equations for each components are (Shahsavar and Ameri, 2010):
For the PV panels,
For air flowing inside the air duct,
For the back insulation surface,
By solving the energy balance equations for various components of the BIPVT system along with the
Evaluation of performance parameters
The rate of heat required to increase the ambient air temperature to the considered indoor design temperature (i.e., 296 K, based on the comfort zone) is determined using the following equation:
The rate of useful thermal energy available from the BIPVT system during the winter operation mode can be expressed as:
The rate of electrical energy produced by the PV panels:
The fan power can be obtained by the following equation (Shahsavar
Artificial neural network
The artificial neural network was first represented by McCulloch and Pitts (1943), and first trained by Hebb (1949). Many scholars have implemented several types of this tool that are aspired from the biological neural network. Due to their ability in solving the engineering problems, that is carried out through establishing the non-linear equations between inputs and desired outputs (i.e., target dataset), they have been widely applied to various fields (Moayedi and Armaghani, 2018, Moayedi
Data collection and methodology
In this work, the performance of the studied BIPVT system is simulated using a MATLAB home-made code. The constant design parameters of the system are reported in Table 2. The climatic conditions of typical days of each month for Kermanshah were listed in Khaki et al. (2017).
Examining the used equations of the suggested BIPVT system shows that the energetic performance of the system is mainly affected by the following factors: the length, width and depth of the air duct underneath the PV
Results and discussion
This study aimed to calculate the energetic performance of a BIPVT system using artificial intelligent based predictive tools. In this regard, an MLP neural network and a hybrid GA model are applied to approximate the of the considered BIPVT system. The information regarding dimensional properties (i.e., length, width, and the depth) of the duct below the PV panels and the air mass flow rate are provided to train the networks. The collected database was randomly separated into two
Model presentation in the prediction of energy performance
The results for all three proposed models of ANN, ANFIS, and GP in predicting energy performance of the considered BIPVT system are presented in the form of regression charts in Fig. 11, Fig. 12, Fig. 13, respectively. Also, the obtained values for R2 and RMSE indices are available in Table 3. As these tables indicate, the R2 (i.e., accumulated coefficient of determination) for both training and testing datasets are equal to 0.9996 and 0.9994, 0.9948 and 0.9952, and 0.9990 and 0.9990,
Conclusions
This work outlined the applicability of three artificial intelligence tools in the field of BIPVT systems. In this way, ANN, GP and ANFIS predictive techniques were applied. The suggested BIPVT system has the ability of cooling PV panels by ventilation/exhaust air in winter/summer and generating electricity. To train the mentioned networks, a dataset containing four effective parameters including the length, width, and depth of the duct underneath the PV panels and the air mass flow rate were
References (41)
- et al.
Modeling deformation modulus of a stratified sedimentary rock mass using neural network, fuzzy inference and genetic programming
Eng. Geol.
(2016) - et al.
Prediction of energetic performance of a building integrated photovoltaic/thermal system thorough artificial neural network and hybrid particle swarm optimization models
Energy Convers. Manage.
(2019) - et al.
Electro-osmotic permeability coefficient of peat with different degree of humification
Int. J. Electrochem. Sci.
(2011) - et al.
Prediction of zeta potential for tropical peat in the presence of different cations using artificial neural networks
Int. J. Electrochem. Sci.
(2011) - et al.
Artificial neural networks approach for electrochemical resistivity of highly organic soil
Int. J. Electrochem. Sci.
(2011) - et al.
Multi-objective optimization for building retrofit: a model using genetic algorithm and artificial neural network and an application
Energy Build.
(2014) - et al.
Assessing active and passive effects of façade building integrated photovoltaics/thermal systems: Dynamic modelling and simulation
Appl. Energy
(2018) - et al.
Artificial neural network model for forecasting sub-hourly electricity usage in commercial buildings
Energy Build.
(2016) - et al.
Genetic programming and gene expression programming for flyrock assessment due to mine blasting
Int. J. Rock Mech. Min. Sci.
(2016) - et al.
Solar energy for net zero energy buildings–A comparison between solar thermal, PV and photovoltaic–thermal (PV/T) systems
Sol. Energy
(2015)