Optimizing your O&M

Share

<grundtext_einzug>measured irradiance on a day with clear skies (see Figure 2 above) reveals that at least two of the sensors measure incorrectly. The calibration uncertainty of the pyranometers is roughly 1.5%. Measurement deviations between pyranometers (which are expected to report the same irradiance) of more than 3% indicate that something is incorrect. An on-site inspection revealed that the calibration coefficients in one of the systems were not correct, leading to incorrectly reported irradiance values. </grundtext_einzug>

Case study: PID

Potential-induced degradation (PID) refers to the degradation of some commercial PV modules when exposed to high voltages with respect to ground. For PV modules made of p-type silicon solar cells, the voltage has to be negative for this degradation to occur. PID affects the cells within the module individually and may cause considerable degradation of the fill factor (FF) and open-circuit voltage (V<zf_tiefergestellt>OC</zf_tiefergestellt>) of the module. Modules at the negative string end show more degradation than other modules in the strings due to being exposed to higher negative voltages. However, not all modules at the negative string end in one PV plant degrade at the same rate. Meanwhile, in some strings, modules with a performance reduction of 50% may be found, while in other strings modules may still show no measurable degradation. Note that this may be an indication that the small numbers of modules tested according to IEC TS 804 may not be representative of the millions of modules produced.
<grundtext_einzug>The fact that not all modules experience PID at the same rate makes it easier to detect the issue via monitoring if measurements at the string level are available. Even though the PID-affected strings show a reduction of FF and V<zf_tiefergestellt>OC</zf_tiefergestellt> due to the parallel connection with other unaffected strings, the affected strings are operating at lower currents. </grundtext_einzug><grundtext_einzug>Knowing this, comparing string currents versus time may help identify strings that are affected by PID. In order to exclude seasonal shading patterns that may be different from string to string, only string currents measured above a certain sun elevation should be considered. Figure 3 shows the comparison of string currents over a period of three months at a plant in southern Europe. Several strings show decreasing production during August and remained at around 80% of the production of the other strings during September. The results of the thermographic imaging and the IV-curve measurements undertaken during an inspection revealed considerable PID of some modules at the negative end of the strings. Figure 3 shows the temperature distribution on the module surface for a module at the negative string end, affected by PID (top), and the module at the positive end of the same string (bottom). The affected module shows the typical variation of temperature for individual cells due to being short-circuited. </grundtext_einzug>

Case study: tracking systems

In the boom years prior to 2008, lots of dual-axis tracking systems were installed in Spain. In the meantime, single-axis tracking systems have become the preferred solution in countries with high levels of irradiation. A potential disadvantage of a tracking system is that if the system erroneously remains in the morning or evening position, energy production suffers. Therefore, it is important to detect tracking failures immediately and correct the system operation. If the rotation angle of a single-axis tracking system is measured and stored in the data acquisition system then issues can be detected. With regard to a possible definition of tracking system availability in contractual agreements, a certain range around the theoretical curve may be defined (e.g. 10º), outside which the tracking system should be considered to be unavailable.
<grundtext_einzug>The described SDA approach of comparing measured data to expected performance is a valuable and efficient way to make decisions about whether a system is performing well and what corrective actions are appropriate. </grundtext_einzug> <grundtext_einzug>points around the straight line (center). Smaller time offsets (30 min, right) result in smaller circular shapes, which may be confused with the normal scattering of the system. It should be noted that there are numerous opportunities for observing similar patterns, i.e. in case the combiner box and inverter measurements are not connected to the same logger (and probably not synchronized) or if the azimuth orientation of the irradiance sensor is not the same as the azimuth orientation of the modules in a plant. </grundtext_einzug>

Case study: irradiance sensors

Production and irradiation are often key parameters when compliance with contractual agreements or with assumptions made in the financial model are checked. The energy production is almost always taken from a high quality utility-grade meter in the substation – the accuracy of which is almost never questioned. The situation for irradiation measurements is quite different. After deciding on the type of irradiation sensor, the number of sensors, their location, and orientation must be determined. With large systems that encompass diverse topography and shading conditions, multiple sensors are required to properly characterize the entire site. But even in moderately sized plants, redundant sensors may be very useful, as the following example outlines.
<grundtext_einzug>A total of six in-plane irradiance sensors were installed in a 10 MW plant in central Europe. The different sensors were connected to two different communication systems. The sensors were not installed on the module array structure but on separate poles. Visualization of three of the pyranometer measurements versus time (see Figure 2 above) seems to reveal a time offset between the measurements due to the fact that Sensor 1 is the first to measure increasing irradiance in the morning and the first to measure decreasing irradiance in the afternoon. </grundtext_einzug><grundtext_einzug>However, the fact that clouding at noon reduces irradiance measured by all three sensors contradicts this assumption. An on-site inspection revealed that the azimuth orientation of the different sensors differed by up to 4º. The impact on the daily measured irradiance is small (<0.5%), but analysis of the system during the course of the day is hindered as the Performance Ratio exhibits an error of several percent in the morning and afternoon. Note that two of the sensors were connected to a second communication system due to the fact that all input channels of the first system were occupied. The use of multiple communication systems increases the potential for synchronization problems. Comparing the </grundtext_einzug>
Monitoring systems are implemented in all utility-scale PV plants today. However, the data scope can vary considerably, from mere reporting of the technical assumptions in the financial model to the detection of malfunctions in individual components. Smart Data Analytics (SDA) is a method for delivering key figures for stakeholders, taking into account the design of the PV plant and the physical dependency between individual parameters. There are various SDA approaches and they can be examined using several case studies from operating PV plants where SDA has supported the rapid detection of underperformance and unavailability, prompting and supporting efficient O&M activities. These led to a reduction of downtime and an increase in energy production.

Data quality

Prior to analyzing the operational behavior of a PV plant, the quality of the monitoring data should be verified. This check should include the following steps as a minimum.
Data availability: Data availability should be checked for individual parameters and missing data points should be identified and flagged. If redundant sensors are available, signals from sensors may be synthesized to patch missing data points.
<grundtext_einzug> Data reasonableness: Ranges for reasonable allowed data should be defined for each monitored parameter, and values outside these ranges should be disregarded. However, analyzing these out-of-bound issues may identify required corrective actions related to sensor quality and maintenance. Due to “irradiance enhancement” or “cloud-edge effects,” reasonable irradiance data depend on data frequency. Higher irradiance values may be observed when sampled at higher frequencies when clouds are close by. </grundtext_einzug><grundtext_einzug> Synchronized data: All temporal data streams should be fully synchronized. Particular attention should be paid to sensors connected to different monitoring systems or to a change of system time from winter to summer time or vice versa. Changing the time in monitoring systems from winter to summer time is discouraged, as doing so will make data analysis more complicated. Once the data quality is verified, the information can be analyzed in a thorough manner to evaluate the operation of the PV plant and initiate corrective actions if issues are found. The SDA approach compares expected data to measured data. Problems can come from measurement or real performance issues. To illustrate SDA, several examples are presented. </grundtext_einzug><grundtext_einzug>The following example has been taken from a utility-scale PV plant in Europe where irradiance and inverter production data are stored in different databases. Before checking for synchronization issues, the architecture of the monitoring system was reviewed in order to identify parameters that may not be synchronized. The representation of production versus irradiance should always yield a relatively straight line (see Figure 1 below) for a well-performing system. The change from summer to winter time during the summer months causes a more symmetric circular distribution of the data </grundtext_einzug>