Module temperature is an important parameter of photovoltaic energy systems since their performance is affected by its variation. Several cooling controllers require a precise estimation of module temperature to reduce excessive heating and power losses. In this work, an adaptive neuro fuzzy inference system technique is developed for temperature estimation of photovoltaic systems. For the learning process, experimental measurements comprising six environmental variables (temperature,

more »

... , wind velocity, wind direction, relative humidity, and atmospheric pressure) and one operational variable (photovoltaic power output) were used as training parameters. The proposed predictive model comprises a zero-order Sugeno neuro fuzzy system with two generalized bell-shaped membership functions per input and 128 fuzzy rules. The model is validated with experimental information from an instrumented photovoltaic system with a fitness correlation parameter of R = 95%. The obtained results indicate that the proposed methodology provides a reliable tool for estimation of modules temperature based on environmental variables. The developed algorithm can be implemented as part of a cooling control system of photovoltaic modules to reduce the efficiency losses. requires measurement of the PV temperature using sensors that are usually imprecise. Moreover, the measurement system requires maintenance and calibration, and they are sensitive to climate variations. These factors make them unreliable when the measurement systems are used on a long-term basis [11, 12] . An alternative to solve this issue is the use of mathematical and computational models to indirectly estimate the system variables. Several expressions that correlate PV module temperature as a function of diode PV model electronic parameters (resistance, current, voltage, band gap, etc.) [13] , material, and system-dependent properties (glazing-cover transmittance, plate absorptance, etc.) [14] , can be found in the relevant literature. In practice, it is complicated to know or obtain the majority of these parameters under long periods of operation. Moreover, it is widely documented that external weather parameters such as environmental temperature, wind speed, wind direction, atmospheric pressure, and relative humidity [15] [16] [17] [18] , influence the behavior of PV temperature with high complexity. Therefore, the aim of the present work is the development of an adaptive mathematical model of the PV temperature considering the variability and nonlinear behavior of the environmental parameters [19, 20] . In recent years, artificial intelligence (AI) methods have proven to be a powerful tool for nonlinear complex engineering applications. The main advantages of these computational tools are their versatility, robustness, fast computing process, and optimization achieved through learning processes [21, 22] . AI methods have been successfully applied in many renewable energy problems [23] [24] [25] [26] [27] [28] [29] . In [23] artificial neural networks (ANN) and genetic algorithms (GA) have been used to predict the thermal efficiency of steam generation plants. Yaïci et al. employed ANN and adaptive neuro fuzzy inference system (ANFIS) for modeling the performance of solar thermal systems [24, 25] . Wind speed and wind direction have been forecasted using artificial intelligence algorithms as particle swarm optimization (PSO) and GA [26, 27] . Bassam et al. [28] used ANN to estimate the static formation temperatures in geothermal wells. Fuzzy logic algorithms have been employed to design strategies for wind farm efficiency estimation [29] . Among AI modeling methods, adaptive neuro fuzzy inference system (ANFIS) is considered one of the most feasible tools to predict energy systems performance [25] . Several studies on PV technology have been developed using ANFIS. Mellit et al. [30] developed an ANFIS model to predict the optimal sizing coefficient of PV systems based only on geographical coordinates. The results obtained were compared and analyzed with artificial neural networks (ANN) proving that ANFIS model presents the most accurate results. Mohanty [31] generated an ANFIS model to predict monthly solar global radiation for PV system sizing using sunshine duration, ambient temperature, humidity, and clearness index as inputs. The model exhibited an absolute percentage error of 0.48, which yielded better results compared to other methods such as ANN and support vector machine (SVM). Mellit and Kalogirou [32] developed ANFIS models for different components of PV systems such as the generator, battery, and regulator. They designed a global model that combined different ANFIS models relative to each component of the PV system. The components of the global model were trained using various input data. The developed model predicts electrical variables of each component in the PV system using measurements of environmental temperature, irradiation, and clearness index. On the other hand, several studies related to PV maximum power point tracking (MPPT) using ANFIS have been reported in the literature [33] [34] [35] . These techniques improve the operation efficiency of the systems using input variables as short circuit current (I sc ), open circuit voltage (V oc ), environmental temperature, irradiation, and others. In the present work, a new model based on ANFIS methodology to estimate the operation temperature of a PV array is proposed. The rest of this paper is organized as follows: the second section describes the environmental variables that influence the temperature of PV modules, the experimental PV array monitoring, and ANFIS theory. The third section develops the methodology employed in the modeling process by ANFIS technique. The fourth section presents the obtained results, and the last section contains the conclusions.