17. July 2018

Innovations in pattern recognition on energy data

Discover the essential automatically

Since 2016 an automatic pattern recognition is available in IngSoft InterWatt. Intelligent algorithms recognize typical processes in energy data and report unusual consumption processes to the energy manager for evaluation. The benefits are enormous: even without looking at the load curve of each individual meter every day, you can be sure that you will not miss any problems that would be apparent from the load curves.

The conventional alternative to this type of pattern recognition would be to model min/max envelopes manually and limit their validity to certain day types (e.g. Mo-Th, holidays etc.) and boundary conditions (e.g. at low outside temperatures).

In the past releases of IngSoft InterWatt IngSoft has extended the pattern recognition by helpful functions:

More sensitivity

Adaptive bandwidth: There are times during the day when the load curve hardly deviates from similar days, e.g. during the night when only the same devices are always running in standby. At other times, for example when production is running, there is a wider range of "normal" energy consumption. Through an improvement of the algorithm, the monitoring becomes more accurate in times with very regular energy consumption, without overrunning the monitoring with false alarms at other times.

More focus on the essentials

Many problems that are detected with the automatic pattern recognition are errors that only occur on one day, e.g. forgotten light at night. The information about such errors is helpful, for example, if these results are included in the training for the employees. Only if a switching error or defect lasts longer than one day and the results of pattern recognition are responded to promptly, however, energy and money can be saved immediately with the help of the automatic pattern recognition.

The pattern recognition in IngSoft InterWatt usually reduces the number of daily profiles to be viewed by a factor of 30, i.e. on average only one profile per month is reported as conspicuous per meter. In situations where hundreds or thousands of meters have to be monitored by very few persons, this can be too much.

In such cases, the pattern recognition can be configured in such a way that anomalies are only reported to the user if they are noticeable for at least n consecutive days. Thus, the energy manager is not bothered with errors that heal themselves after one day.

Less setup effort

Pattern recognition can be set up for meters or entire buildings with just a few mouse clicks. If it is too much effort to repeat these clicks for each meter or building, you can now determine for all elements in the system (or a defined part of it) that pattern recognitions are automatically created and activated. This even applies to elements that are created subsequently.

Faster information

Although the actual recognition algorithms are the core of IngSoft's innovation, the software support for the processes around determines the utility value for the daily work of the users. IngSoft InterWatt embeds pattern recognition in a well thought-out surface that supports the analysis, classification and standard-compliant documentation of anomalies. A continuous hierarchical "traffic light system" helps to keep an overview even in large organizations. Detected anomalies become visible immediately - when you are logged into the system. The release of IngSoft InterWatt, published in July 2018, offers an important innovation for occasional users: A user can be notified by e-mail as soon as an important event occurs.