IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 4, NO. 2, MARCH2014 659 Online Recording a PV Module Fingerprint P. L. Carotenuto, P. Manganiello, G. Petrone, and G. Spagnuolo Abstract—In many applications, the current versus voltage curve of a photovoltaic cell, module, string, or ﬁeld is acquired. A high number of samples are usually acquired, but the curve con- tains the main information in the open- and short-circuit points, as well as where it has a strong change in the slope. In this paper, these parts are called “the ﬁngerprint” of the photovoltaic genera- tor. The ﬁngerprint allows us to recognize the working conditions of the photovoltaic generator, e.g., if it is affected by a partial shad- owing or not. Saving the ﬁngerprint and discarding the other points of the original curve allows us to minimize the memory needs for storing the curve without losing the main information content. In this paper, a numerical technique for selecting, from among the samples of the acquired current versus voltage curve of any pho- tovoltaic generator, the ones to be included in the ﬁngerprint is proposed. The processing steps and the memory needed to achieve the result are minimized in order to allow an implementation of the algorithm also in a low-cost processor for on-ﬁeld real-time applications. The technique is validated through curves generated by using analytical models as well as by means of some curves acquired experimentally in outdoor conditions. Index Terms—Diagnostic, photovoltaic (PV), photovoltaic array reconﬁguration. I. INTRODUCTION T HE acquisition of the current versus voltage (I–V) or power versus voltage (P–V) curve of the generator is an opera- tion performed in many photovoltaic (PV) applications. The I–V curves of cells, modules, strings, and PV ﬁelds are measured for many different purposes. They are acquired by means of a cur- rent, voltage, or load sweep through dedicated instrumentation, or by using the power-processing systems whose main function is maximum power point tracking (MPPT) of the PV source itself. A good example is represented by many of the PV invert- ers in the market: they acquire the P–V curve of the string or of the whole ﬁeld in order to drive the inverter’s MPPT algorithm toward the absolute maximum power point. This function is run periodically but, to the best of the authors’ knowledge, the P–V curve is not stored at all for any post processing purpose. Moreover, many commercial products, performing the MPPT function at a string level or employing distributed MPPT (DMPPT) technologies, propose monitoring software tools Manuscript received October 3, 2013; accepted December 3, 2013. Date of publication January 2, 2014; date of current version February 17, 2014. P. L. Carotenuto, G. Petrone, and G. Spagnuolo are with the Department of In- formation Engineering, Electrical Engineering and Applied Mathematics, Uni- versity of Salerno, Fisciano Salerno 84084, Italy (e-mail: pcarotenuto@unisa.it; gpetrone@unisa.it; gspagnuolo@unisa.it). P. Manganiello is with the Department of Industrial and Information Engi- neering, Second University of Napoli, Aversa 81031, Italy, and also with the Department of Information Engineering, Electrical Engineering and Applied Mathematics, University of Salerno, Fisciano Salerno 84084, Italy (e-mail: patrizio.manganiello@unina2.it). Color versions of one or more of the ﬁgures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identiﬁer 10.1109/JPHOTOV.2013.2294759 aimed at visualizing the I–V curves of the PV sources, at differ- ent levels of granularity, for diagnostic purposes. In particular, in DMPPT architectures, one dc/dc converter, known as the “power optimizer,” or “microinverter” is dedicated to each PV module. Such devices often interact with a central unit that col- lects data concerning the modules operation and gives the user an idea of the produced power and energy. The availability of more detailed information, which is the whole I–V curve of all the PV modules in the ﬁeld, acquired with a predetermined periodicity and processed properly, would give the user fur- ther information about the sources of unexpected drops in the power and energy productivity of the PV ﬁeld (e.g., see [1], [2], and [3]). Some data about the actual operating conditions of the source are mandatory in those systems that reconﬁgure the PV modules of a ﬁeld in some strings in order to increase the electrical power produced in the presence of mismatching. They acquire the I–V curve of each module periodically and perform some calculations on these data in order to determine the best electrical connection among the modules that maximizes the produced power [4], [5]. Also, the curves are, or might be, processed for diagnostic and prognostic purposes. In all the aforementioned cases, the I–V curve of the PV gen- erator must be acquired through suitable hardware, at a suitable rate, in order to avoid undesired dynamical effects [6], and with a proper number of samples, in order to detect all the maxi- mum power points, and the voltage and current levels at which they occur. At the end of the acquisition process, it is evident that a high number of samples are useful where the knees of the I–V curve occur, but a very low number of samples would be enough to describe the almost-constant-current and almost- constant-voltage branches of the curve itself. Many samples in these parts of the curve are useless and increase the memory requirements for curve storage, e.g., aimed at the creation of a database that the diagnostic/prognostic function would be based on. Moreover, the transmission of a large number of samples from a module-dedicated unit to a central monitoring system requires a lot of time and expensive resources. The problem es- pecially arises when the I–V curves must be temporarily stored, thus, processed and transmitted by low-cost digital controllers, as in the case of reconﬁguration systems or module-dedicated power processing units for DMPPT applications. The problem described previously is trivial, if an I–V curve generated by a numerical PV model is analyzed by using a per- sonal computer: the P–V curve is obtained by doing as many multiplications as the number of the curve samples. Afterward, one of the many numerical procedures aimed at detecting the curve’s maxima can be used. Unfortunately, in the real applica- tions mentioned previously, the I–V curve is affected by the noise (e.g., due to quantization effects). Moreover, the implementa- tion into a digital controller requires algorithms performing the 2156-3381 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications standards/publications/rights/index.html for more information.