The basics of photovoltaic system monitoring

The complexity and the most capital needed for a photovoltaic plant is concentrated at the beginning of its lifetime. From then onwards, it is mostly just gaining revenue from the electricity produced. The varying price of the electricity can be a factor on the revenue, as well as the business model chosen. Support frameworks for utility scale projects are moving towards the premium based model (more info on the different models in my other article), which means that system operators sell the electricity on the market and receive a bonus premium on top of the market price.

Leaving the business model and market fluctuations aside for a moment, the other risk is the efficiency and performance of the system during its lifetime, or to be more precise the deviation of performance from the optimal value. Many factors can affect the power yield of the system during its lifetime, some of them include: soiling of the modules, module degradation (photovoltaic modules loose efficiency over time and produce less electricity), inverter efficiency, system components’ faults. Even a single faulty PV module can substantially affect the whole string of modules and if not treated, cause more damage to the system and other components.

Good inverters have protection mechanisms in place, to limit the possibility of self-damage, but those mechanisms work in a way which reduces the power output.

This leads to a necessity to continuously monitor key parameters about the systems’ health. Modern inverters have mechanisms which gather and report data about the health and operation of the system, but the diversity of models and various protocols for communication and reporting used, make it difficult to create one solution that suits every manufacturer and model.

Perhaps, the most important parameter that one would like to keep an eye on is the power yield. Understandably, it is the parameter that influences the revenue generated from the system. One would like to detect deviations of the power yield from the nominal value. There are statistical methods that can prove with a given degree of certainty that there are deviations, but those require data over a longer period, typically one year. This means a significant delay for finding faults in the system that affect its performance, which is costly and reduces revenue.

There is another approach that can be used for the same purpose, which includes simulation of the system power yield by models. Those models require near real-time weather data such as irradiance, temperature, wind speed and similar. The models calculate expected power yield from the system, a value which can be then compared to the actual power yield. If there is a statistically significant discrepancy between these two values, the possibility of a fault in some system component is high.

This approach can detect performance degradation as fast as data about the necessary weather parameters is available. Satellite data can be used for this purpose and in the best case have a delay from several hours to a few days.

The problem with satellite data is that they are based on models, which have uncertainties themselves. That means that greater discrepancy between the actual and expected power yield is needed to be able to conclude that indeed there is a possible fault in the system. This can be solved by on-site measurements from sensors. Their data are available almost instantaneously, and have low uncertainties (2-7%).

The instruments need maintenance, appropriate data acquisition, processing and analysis processes need to be in place, in order to be able to get a valuable insight about the systems health. Then, how to perform all of it without tremendous cost that will outweigh the potential benefit?

Solar Data Collector provides a service which includes every aspect, from the installation and maintenance of the equipment needed to monitor performance, to data processing, PV yield simulation and automated alerts and reports.