André Mermoud

In PVsyst, the evaluation of the "Losses" of a PV array (as for the definition of the normalised performance ratios in Europe, see Help), takes as starting point the energy which would be produced if the system worked always at STC conditions (1000 W/m², 25°C, AM1.5). The first loss evaluated in the simulation process and the loss diagram is the "Irradiance loss", which is the difference between the STC efficiency and the effective efficiency under the instantaneous irradiance, at 25°C (see "How is evaluated the "Low-light" efficiency ?". Low-light efficiencies over STC may be obtained, mainly by the choice of the Rserie value. Values obtained by official flash-tests measurements for crystalline modules usually don't overcome 0.3 to 0.5%. Higher values in the model should be suspected of manipulation of the Rserie parameter. These efficiency performances measured outdoor are lower by 0.5 to 1%, probably due to the diffuse contents of the incident irradiance, which suffers of IAM loss effect. You can also obtain slightly positive irradiance loss in very sunny situations, when the system mainly operates above 400 to 600 W/m2.

The backtracking with axis not facing south is allowed and works if you turn the whole trackers set toward not-south azimuth (i.e. when the alignment basis of each tracker is perpendicular to the axis, misalign parameter=0). For axis with azimuths not perpendicular to the trackers aligment (i.e. not-null "misalign" parameter), the problem is much more complex, and I was not able to find an analytical algorithm for tilted axis. Now for horizontal axis in this situation, the algorithm should be the same as in the Misalign=0 case, but I have indeed not developed this special case.

Sorry, it is indeed an error in the version 5.58, that I have corrected in the version 5.59. It is related to the clipboard contents, and appears when the last selection and "copy" was not a piece of text.

For the temperatures: this is a modification of the format appeared in Firefox 13: see my answer import meteo data from PVgis with firefox 13 For the February error, the Leap Yera information is not well initialized. see my answer in Import Meteo PVGIS bug

You are right, I have forgotten initializing the Leap-year information when importing PVGIS data. Therefore the import process uses sometimes 28 and sometimes 29 days for February. This induces an error of 3.5% on the February data, but less than 0.15% on the yearly sum (for Europe). By the way the real error when considering a simulation as valid for set of years would be 3/4 of this values, as 1 year over 4 is leap... This error is random (depending on the last use of this variable in the program), not specific to Classic or SAF data. It will be corrected for the next issue 5.59 of PVsyst.

You should redefine your SMA inverter as 5 MPPT inputs, and use one SubArray with 5 inverters * 4 strings (MPPT), and another one by 5 inverters * 1 string (MPPT). See the FAQ How to use the inverters with very different MPPT inputs (Tripower of SMA) ?

What you call "Parasitic resistance" is named "Series resistance" in PVsyst. Please observe that the measured Rserie proposed by the manufacturer is not the Rserie as defined in the one-diode model. See ou FAQ I can't specify my measured Rserie. NB: With version 5, the best way of setting a reliable value to the Rserie of the One-diode model, is to try to reproduce the low-light performances if these are provided by the manufacturer. You can analyse this in PVsyst, using the graph "Efficiency as function of Irradiance", and proceed by trys and errors. There is now a tool for performing directly this adjustment in the version 6. Please also see our FAQ What explaind differences of yield between different modules ?

In the version 5, the soiling loss parameter (to be defined in the "Detailed losses" and appearing in the first page of the report) was defined as a percentage of the STC energy value. Now during the simulation, the corresponding energy loss is indeed calculated as a percentage of the STC energy as specified. But the result on the loss diagram is referenced as a percentage of the remaining energy just before the loss, therefore a slightly different percentage value. In the versionn 6, it is accounted in the optical losses, and is a loss by respect to the remaining irradiance (after shading and IAM losses). The value on the loss diagram is identical to the specified parameter.

The format of the data structure obtained by "Copy" has indeed been changed with Firefeox V 13. I will adapt the importing tool in the next version 5.58 of PVsyst. In the mean time you should use another browser.

With near shadings in PVsyst, you can define and perform simulations with 3 kinds of near shadings (to be applied to the beam component): - The shading loss due to the irradiance deficit on the modules (optical loss), which we call "Linear shadings". The shading factor is the ratio of the effective shaded area with respect to the total field area. - The Electrical losses, i.e. the effect of the partial shadings on the electrical response of your PV array. The calculation requires to define the position of each module, and specify its electrical connexion on the inverter input, in order to combine each I/V curve of each sub-module. This is done in the "Module Layout" part. The electrical loss will appear in the Array losses on the loss diagram. - Simplified Electrical losses, which we called shadings "according to module strings": this is intended to the evaluation of the maximum electrical effect on the array production, in a simplified and approximate way. Remember that in a string, the poorer cell determines the current of the whole string. That is, when one cell is shaded, the entire string is affected, and is supposed to become almost inefficient (for the beam component). Therefore in this mode, the array is partitioned into rectangles, each rectangle representing (approximately) the area of a full string of modules (to be defined within the shading scene, icon on the left). Then, when a part of this rectangle is shaded, the whole rectangle is considered inactive. The Global Shading factor is the ratio of the shaded rectangles (grey + yellow) to the total field area. The additional loss represented by the yellow parts (Global - Linear shading factor) gives indeed an evaluation of the upper limit for the electrical loss. Now some part of the string energy is recovered by the by-pass diodes (see How to evaluate the effect of by-pass diodes in shaded arrays?). This is the reason why we have defined the "Fraction for relectrical effect". However when about one third of "sub-modules" in a string are shaded (as for example in rows arrangement), the output for beam irradiance becomes almost null, and this fraction will be near to 100%. This evaluation of electrical losses "according to modules strings" was the only way of estimating the electrical losses up to the version 5. It should still be used with very big systems, when the "Module layout" becomes impracticable due to calculation time. In this case you can evaluate the "Fraction for electrical effect" by comparing the "real" electrical loss calculated by the Module Layout on a reduced sample of your system, to the calculation "according to module strings". And apply this last calculation to the full system.

When you compute shadings with the option "according to module strings", the full rectangle-string is considered electrically inactive as soon as hit by a shade. Nevertheless, some part of the string energy is recovered by the by-pass diodes; this parameter describes which fraction of the string production is really lost. In version 5: The version 5 calculation cannot give a reliable value for this "Fraction for electrical effect" parameter. This is dependent on the shades distribution on the field and the electrical array configuration. For a shed arrangement (where the shades are very "regular" and several sub-modules are shaded), it is near to 100%, because as soon as 1/3 of its sub-modules are shaded, a string doesn't participate to the electrical production anymore. When with more "distributed" shades like Chimneys, far buildings, it could probably be of the order of 60 to 80%, depending on the "regularity" of the shade (a diagonal-like shade has a lower impact as it concerns modules better distributed in the array). In version 6 The realistic calculation of the electrical effect of shadings implies the exact positioning of each module on the geometrical plane, using the "Module Layout" tool, the identification of each electrical string in the array. Coupled to the shading calculation, the "Module layout" tool evaluates the real I/V curve of the PV array (on one MPPT input), and provides a realistic evaluation of the loss due to the electrical mismatch. Now comparing the Electrical loss calculated by this "Module layout" option, and the "Electrical loss" evaluated by the approximated shading mode "according to module strings", you will be able to establish the "Fraction for electrical effect" to be applied in order to match the Module Layout calculation. As an example, if you get 4% electrical loss from the option "according to module strings", and 3% with the Module Layout, the fraction for electrical loss will be 75%.

In tracking arrays mutual shadings may be very important, as the gains are mainly waited when the sun is low on the horizon. The backtracking strategy tries to suppress the mutual shadings by reorienting the modules. But it is an illusion to think that you should obtain a much better yield with Backtracking. When the trackers perform a normal tracking (most perpendicular to sun as possible, with mutual shadings), or perform a Backtracking (deviating from the optimal orientation) they intercept about the same "Light tube" ! In one case you have shading losses, in the other one losses for mis-orientation (cos angle). It is not clear which configuration receives more irradiance: without backtracking, the shadings usually don't affect the full "length" of the modules (=> more "active" area), with backtracking you have additional IAM losses. The only decisive advantage of the backtracking – if any – is to avoid the electrical effect of shadings (i.e. when a part of a string is shaded, the full production of the string is affected). As an example, here is a comparison between "normal tracking" (with shades) and backtracking, for a N/S axis horizontal trackers array in Santiago (Chile). The phi angles limit is +/- 45°. Backtracking and Normal tracking comparison

In the present time the Grid-systems and the Stand-alone systems are quite distinct in PVsyst. Grid systems with a battery storage are more and more used now in the reality. However their simulation is a very difficult problem, not yet implemented in PVsyst. I don't see any possibility for approaching a solution with the existing Grid-connected and Stand-alone systems presently available in the software. The objective of such hybrid systems may be quite different from case to case: - For "purists" of the PV energy, consuming a minimum of energy coming from the grid, whatever the price, - According to consuming and feed-in tariffs, optimizing the costs of electricity, - For the optimization of the grid management, injecting power during the "best" periods of the day, - Grid management: peak shaving, - Grid management: short term grid stabilization (for example for clouds), - In "rich" countries, ensuring a secure back-up in case of (rare) grid failure, - In countries where the grid is weak or intermittent, ensuring electrical availability during the whole day, - Mini-grids for the electrification of whole villages or islands, - etc... Each of these uses of the PV energy will involve different sizings, different constraints, and quite different control strategies. On the one hand, the control will depend on the self-consumption profile and the grid characteristics (availability, overload, etc), On the other hand, when should the PV array charge the batteries? When they are not full ? When the consumers are low? When the foreseen weather of the next day is bad ? And how to optimize the size of the batteries ? As an order of magnitude: for an household consumption of 15 kWh/day (a standard in Europe), storing one only day consumption would represent about 700 Ah for a 24V battery bank, i.e. about 600 kg of Lead-acid batteries. Remember that the price of the stored energy is very high. It can be evaluated by the price of the battery pack, divided by the total energy stored along the battery lifetime, i.e. Capacity (in kWh) x DOD x Max. nb. of cycles. If you assume a full storage/destorage every day, a battery pack of 1'500 cycles should be replaced every 4 years. For household systems connected to the grid, this price of kWh should be compared to the difference between the buying and the selling prices. Now components manufacturers propose a great variety of devices, configurations, usually specifically suited for one or the other of these uses. In this evaluation, we should also define the prices of the injection, consumption, stored energy (taking the limited lifetime in terms of number of cycles into account). Many people ask for such systems and strategies, without understanding that the problematics is very complex and the realization expensive. We intend to study these systems during the next months. We will probably propose simulations for a selection of these kinds of systems, to be extended progressively.

When sizing a PV stand-alone system, the basic constraints are the availability of solar energy during the year, and the satisfaction of the user's needs. The problem to be solved is the optimisation of the size of the photovoltaic generator and the storage capacity, in order to meet the energy requirements, eventually accepting a specified "Loss-of-load" probability that the energy will be missing. The first requirement is therefore the evaluation of the user's needs along the year, and the choice of an autonomy period for the battery bank (which should cover the worst sequences of days without sun). A usual choice is 4 days. PVsyst pre-sizing tool : a random process In the PVsyst pre-sizing tool, the optimization of the required PV power is performed by a fast simulation of the whole system for different PV array sizes. This simulation uses an approximate calculation based on monthly meteo values, and a series of 365 days constructed by the Synthetic Generation of the Collares-Pereira model. The Loss-of-load evaluation is of course dependent on the day's series: it may be different if you construct another weather series. And there is no clear limit for determining the "Fully safe" limit (in battery capacity or PV array), i.e. which will contain the worst weather time-series possible for this site. The system sizing highly depends on the time-series, and therefore will be different from one execution to another one of the presizing tool, as well as for the detailed simulation. NB: In a future version, we will provide a more refined statistical évaluation over several random years. PV array size and back-up The PV array size is dependent on the specified risk of "loss-of-load". Especially in middle-latitudes, implementing a back-up genset for covering - say - 5% of the time drastically reduces the PV array. This is less the case for sub-tropical climates where the insolation is rather well distributed over the year. An oversized array will of course lead to overload losses (i.e. unuseable available energy when the battery is full). But with the decrease of the price of the modules, this may be now a good option.