4.4: Alarms: Relative Performance Analytics

Setting Alarms with Relative Performance Analytics

This Analytic Tool is Available for All Users at No Extra Charge... Ideal for Systems with Granular Data!

Relative Performance Analytics (RPA) is ideal for larger PV systems with granular generation data. That includes systems with string inverters or multiple central inverters, as well as systems with sub-array or string level sensors on the DC side of the array. Relative Performance Analytics can be applied any time you have identical data types from two or more separate devices in one system.

RPA represents a shift in the traditional approach to solar power diagnostics that can help you identify under-performing devices in your PV system. This ‘smart software’ surveys historical system data to map the deltas among reporting values for separate data points (or relative performance). It actually learns the relative performance patterns of the entire system, so users can identify a change in performance of a single (or group of) components relative to the others. System managers will be able identify outlier data points and detect changing performance patterns with greater accuracy and fewer ‘nuisance’ alarms.

 “We developed Relative Performance Analytics as a direct response to requests from some of our most active software users” says company Principal Cosmos Corbin. “Users need actionable information from the monitoring system with a low time investment in configuration. When using monitoring systems we spend a great deal of time looking for a change in equipment performance on site. This tool simplifies that process.”

Getting Started

You can find Relative Performance Analytics in the Alarms Center for each system you monitor:

  • After logging in to your Admin Panel, choose your desired system from all monitored projects, then look for "Alarms" in the navigation tabs under the system name at the top of the page.
  • Once you have selected the Alarms Center for a system, click on "+Create Alarm" in the second row of navigation tabs in the upper left (as shown in image below).
  • Find "Relative Performance Alarm" in your list of available Alarm Types.




Alarm Settings

Selecting "Relative Performance Alarm" takes you to an Alarm Settings page. Here you can establish and adjust settings for an individual Relative Performance Alarm.

Please note that DECK software provides "tool tips" to help you understand settings within Relative Performance Analytics. Look for the small gray circles with a white question mark throughout these pages; clicking on a tool tip will prompt a pop-up window with explanations and advice.

The first field to fill in is Alarm Description. This is your opportunity to communicate functionality and goals of each alarm for current and future members of your team. Only individuals with Admin Panel access will be able to see your Alarm Description.

The next choice is a check box to receive Alarm Resolution Notifications. You can choose whether or not to receive email notifications when a triggered alarm sees data return to the acceptable reporting range (as defined by the user in Alarm Settings). In other words, after an alarm has been triggered, you may find that reporting data continues to fall outside your acceptable reporting range... or you may find that the reporting data returns to your approved range. If data returns to the acceptable range,do you want to receive email notifications? This is your opportunity to set that preference for this particular alarm (setting will not automatically extend to other individual system alarms).




The next group of options appears under the heading "Trigger Settings.

The first choice here is Data Type. DECK software will automatically populate this drop-down menu with every type of reporting data in your system. If your system has an irradiance sensor, for example, "irradiance" will appear as a Data Type option.

Note that each Relative Performance alarm looks at just one data type across two or more devices. Although irradiance may appear in your drop-down menu, you can only create a Relative Peformance alarm for that data type if you have two or more irradiance sensors in that installation.

Your next option lets you Select Devices to Include in the Group. DECK software will automatically look for all devices in your system that are providing your selected Data Type, then populate those devices into an Available Devices list. Select your desired devices from the list on the left, then click "Add" to include them in the Currently Selected Devices list on the right.

Please note that you must select at least 2 devices to create a valid alarm.




After you have selected the devices to include in your Relative Performance Alarm, you will move along to Results and Settings. Your first choice here is to decide between using Historic Weights or Manual Weights.

  • Historic Weights allow you to compare a set of recently reported data points against older historical data sets from your system. When using historic weights, you will select a Standard Deviation option as the trigger mechanism for the alarm.
  • Manual Weights allow you to compare a set of recently reported data points against a fixed set of data points that you will manually enter into this software interface. When using Manual Weights, you will select a Deviation Percentage as the trigger mechanism for the alarm.

(Please see more on working with Historic Weights and Manual Weights in the following section.)


The next choice to make can be accessed by clicking the Advanced Options button. Here you will enter values to determine how many Days of Historic Data and Days of Recent Data the alarm should look at. 

Note that the default values here will be 60 days of Historic Data and 3 days of Recent Data.

Note that if you are working with Manual Weights you will only have the option to enter a value for Days of Recent Data... Historic Data is not being considered when you are working with Manual Weights. 




The last step before saving your Relative Performance Alarm is familiar to anyone who has worked with DECK alarms before:  determine who will receive email notifications when an alarm is raised.

  • Registered Admin users from your company should appear in the drop-down menu under Company User Notifications. 
  • You may include additional individuals to receive email notifications by manually entering their email addresses into the Email Address field under External Email Notifications. 

Please note that you may add as many individuals as you like to receive email notifications without incurring any additional monitoring costs.


After you have completed these settings, click the "Save" button to finalize your new Relative Performance Alarm.



Working with Historic Weights

Any system with at least one or two months of historical monitored data will be able to immediately start using Historic Weights to create insightful Relative Performance Alarms. To understand what DECK software is looking for with these alarms, it will be helpful to understand the Standard Deviation trigger mechanism.

  • The Standard Deviation calculation begins by finding the mathematical mean among the designated data sets. 
  • The Standard Deviation calculation next determines the degree of variance from the mean across the data sets. A low Variation value indicates that the data points tend to be very close to the mean. High Variation indicates that the data points are spread out over a large range of values.


The Relative Performance Alarm maps reported data for multiple devices, then looks for the Standard Deviation among the device data sets (not within each data set). The key metric here is variance over time in relative performance, i.e. "inverter 1 historically produces 10% of the total AC production relative to the other inverters; over the recent sample period, however, inverter 1 has only produced 8% of the total system generation."

In other words, a low Variation value indicates that all surveyed devices are maintaining consistent performance relative to one another. A high Variation value indicates that the relative performance of the surveyed devices is shifting over time.

Note that this metric accounts for varying sunlight conditions, because that variable presumably impacts all device data sets equally. 


Choosing Your Standard Deviation Option

The following example shows the different results you can generate by adjusting the number of Standard Deviations. The first image shows raised alarms when the Standard Deviation value is set very low, just 0.5 Standard Deviations. This sensitive trigger raises an alarm for 3 devices (shown highlighted in red and green). 

Note that green highlighting indicates a device that is reporting data values that are higher than normal relative to other surveyed devices, and red highlighting indicates a device that is reporting data values that are lower than normal relative to other surveyed devices.




The second image (below) shows the same data sets, but in this case the number of Standard Deviations has been set at 1. This metric still raises an alarm, but in this case only 1 device is out of range, rather than 3 devices in the previous example. 




Where should I set my Standard Deviation Option?

DECK recommends starting your Relative Performance alarms with a Standard Deviation value of 2. This number of Standard Deviations is reasonably sensitive in most applications, but will avoid "noisy" email notifications for "nuisance alarms."

One of the great features of the Relative Performance alarm is the ability to see results immediately after you have finalized the alarm. This allows you to employ 'trial and error' methodology to find the right number of Standard Deviations for each alarm you create.

  • After the alarm has been created, click on the "System Status" tab (in Alarms section) to see the full list of alarms you have created for this system.
  • Find your new Relative Performance alarm in this list, and click on the alarm name to visit the "home page" for that Relative Performance alarm.
  • Look for the Trigger Settings section on this page, and click the prompt to Expand/Collapse Settings.

This will allow you to see the Historic Ratio and the Recent Ratio of relative performance for each surveyed device. You can also see the Variation for each device between the Historic Ratio and the Recent Ratio. As in the images above, raised alarms will be highlighted with either green or red.

You may begin by selecting 2 Standard Deviations, then immediately check results to gauge your desired level of sensitivity based on the number of raised alarms for each Standard Deviation option you try.



Working with Manual Weights

Manual Weights have been added to our Relative Performance Analytics to serve customers who do not yet have available historical data to create Historic Weights, or who have historical data that is not desirable for comparison purposes for one reason or another. If this applies to you, you can enter expected performance values as Manual Weights until you have enough good historical data to switch to Historic Weights.

Manual Weights can also reveal some performance metrics that are not available using Historic Weights:

  • For example, it can be useful to measure Relative Performance among inverters against the metric of DC Nameplate Ratings for each inverter. Enter the Nameplate Ratings for each device as a Manual Weight to quickly diagnose whether each inverter is producing its expected share of power relative to the other units.
  • You can achieve the same effect using percentages of total generation as Manual Weights; for example, a system that evenly splits production among 20 string inverters could enter "5" as a manual weight for each device, signifying 5% of total system output.
  • Of course, you can fine-tune your Relative Performance Analytics by adjusting these Manual Weight values to reflect your own understanding of relative performance expectations across your various devices.

Using Relative Performance Analytics in this way is different from traditional performance analytics. In this case you are not comparing actual production values against baseline production values; instead, you are comparing the relative deltas among various device performance values against baseline relative deltas. This calculation filters out the "noise" of fluctuating weather conditions because all device performance levels move up and down by equal measures as sunlight conditions change.

Note that any numeric values can be entered as manual weights... it is not the numeric value that matters, but rather the relation among the various values that are entered. So a system manager with ten 30 kW DC string inverters could either enter "30" as a manual weight for each inverter, or simply enter "1" in each Manual Weight field, since the relation among the values is identical in both cases. 




Was this article helpful?
2 out of 2 found this helpful
Have more questions? Submit a request


Article is closed for comments.
Powered by Zendesk