Solar modules at many of Invenergy’s solar projects are mounted on trackers that follow the path of the sun throughout the day. Many of these sites use a control logic that dynamically adjusts individual rows to maximize captured sunlight.

This is particularly helpful on cloudy days. However, when groups of modules are moving on different trajectories it can be difficult to detect trackers that are deviating as intended from those that are behaving incorrectly. Overlooking these small inefficiencies can greatly affect a site’s productivity, making it critical to find a way to better identify underperforming trackers.

The Process:

Invenergy Services’ Operations Engineering team implemented an algorithm to detect problematic tracker angles. Dubbed the Tracker Anomaly Detection Algorithm (TADA), Data Scientist Axel De Mendoza adapted a density-based spatial clustering of applications with noise (DBSCAN) algorithm to analyze the performance datasets of a single utility-scale solar facility. 

TADA creates clusters for groups of trackers based on the angles at which they are operating, such as a group of trackers obstructed by the environment that have different tracking settings, a group of trackers on a tilted surface that have different tracking settings, or a group of trackers adjusting to clouds covering only that area.

TADA examines data taken the previous day of each tracker’s angle, which is taken at five-minute intervals every operating hour. Trackers not closely following any of the clusters of trackers are flagged as anomalous by DBSCAN. TADA awards a score based on how many times throughout the day each tracker was flagged as anomalous, and a report of the entire analysis is sent to the site’s manager, technicians, and Invenergy’s operations engineers in Chicago. The graph below shows how an anomalous tracker is compared to a tracker cluster

The Results:

TADA was operated for approximately one year at a utility-scale solar site, and the results were promising. As designed, TADA excelled at detecting tracker issues, which ranged from bad batteries to faulty gears.

The TADA results exceeded De Mendoza’s expectations. Identifying – and correcting – this number of trackers would have been much more difficult if TADA hadn’t been in place, and the algorithm was especially helpful in identifying anomalous trackers in areas where conventional methods of data collection might not be able to distinguish subtle angle discrepancies. TADA is being expanded to the rest of Invenergy Services’ portfolio, which is rapidly approaching 1-gigawatt(GW) under operations and scaling to 5 GW by the end of 2023.

“Using TADA was a unique approach to detecting tracker performance, and there was a bit of a challenge in summarizing and translating the algorithm’s outputs in a way that would be useful for the site technicians,” says De Mendoza.

But as any owner can attest, trackers off by even a slight degree can impact site productivity, but innovations like TADA will help boost the reliability, productivity, and financial viability of solar projects large and small.