Predictive analytics for power transformers

O.A. Zakharov, Lead Specialist of the PRANA Expert Group, a project by Rotec JSC, EnergyExpert journal

The natural desire of the owner of a technical facility, as a business unit, is to find out the minimum financial investment needed to ensure the equipment is in operable condition, and as a result, receive the maximum profit during operation of the facility. To do this, up-to-date and accurate information must be obtained on how the facility is operated. For power transformers, this information can be provided by PRANA, a diagnostics software complex and the first patented industrial IT solution in Russia for diagnostics and forecasting of the condition of industrial equipment.

At the very least, the owner of the technical facility wants to be informed of the following:

• how is the facility serviced by technical personnel?
• how are preemptive and corrective actions planned to maintain the equipment in an operable and functional condition?
• how are scheduled and unscheduled operations carried out (various types of repairs and upgrades)?

The information can be received from various sources:

• process violation investigation reports
• reports on completed work and tests conducted
• diagnostic data, conclusion by the engineering supervisor etc.

At present, monitoring and diagnostics data serve as a basis for determining the technical condition of equipment and creating a repairs schedule.

With the modern development of materials, technologies and control systems, as well as developments in key industries such as instrument engineering, power engineering, and hardware and software, there are many options for connecting various sensors and systems to any object subject to observation. These provide an output of a vast array of information on control panels, dispatch centers and situation centers.

Figure 1 shows an example of a "connected" power transformer, which allows for online monitoring of practically all significant defects and malfunctions in the system's nodes and elements.

Moreover, the specific producers and suppliers (both domestic and foreign) of the primary sensors and monitoring and diagnostics systems are shown.

So it would seem that there is more than enough information on the technical condition of the equipment. But there are two key points to take into account here:

• the variety of sensors and systems for processing and issuing signals, and often the inability to combine them, force the owner to incur additional financial costs due to the restrictions on using them together (either on purchasing a software-hardware “needle” or on radically replacing the existing hardware during upgrades)
• operational staff are not able to continuously record and analyze the flow of incoming information, and it is physically impossible to quickly assess the occurrence and degree of development of degradation because parameters trends may not be critical for a long time, and at a certain point the process will begin to snowball and it may be late to do something. As a compensatory measure, the main emphasis in operational and technical maintenance is placed on warning alarms and all sorts of technological and other protection, as well as emergency control automation.

This is immediately perceived as a huge disadvantage, because when an alarm and protection is triggered and equipment is shut down, there are two possible outcomes:

• the need for unnecessarily expensive repairs due to long-term development of degradation and the corresponding deterioration of the technical condition of adjacent nodes and elements
• targeted effects on the disconnected equipment, and including it in work with unidentified and non-eliminated degradations.

Is there a way out, and what does it involve?

Global experience has shown that, from the view point of economics and intellectual property, it is most viable to implement remote monitoring of objects carried out by a group of experts, in which each specialist is responsible for the equipment in their area (evaluates the presence and extent of degradation, gives recommendations to operating personnel), consisting of several different remote objects. At the same time, the experts make use of software and hardware tools for statistical data processing to carry out this task.

The PRANA system developed by ROTEC is a human-machine expert system (both fully domestic and patented), which includes:

• a Remote Monitoring Center (RMC), consisting of a dispatch center and expert group.
• the PRANA universal software-hardware “shell”, which is based on MSET (Multivariate State Estimation Technique), a well-known mathematical statistics device, and Hotelling's T² criterion (multivariate generalization of any number of independent variables with different influence coefficient).

The PRANA system is undeniably universal; any number of any signals at the input gives only one integral parameter at the output, which detects changes in the technical condition of the object by comparing the actual and T² model at any time. The main condition for correct matching is the correspondence between the number and name of the parameters of the compared cross-section (data set) and the model one. Sources of input signals can be both archived and current values of controlled parameters.

The applied method in the PRANA system is highly efficient as it automatically detects changes in the technical condition and is capable of doing so in any give time period within the equipment's life cycle:

• past (technical genetics)
• present (technical diagnostics)
• future (technical forecast)

It is important to note that, in addition to the method, the correct selection of input parameters is needed, as well as correctly constructed operating mode models.

Proceeding directly to the power transformers, the maximum information content is provided:

• by all key nodes on the object: OLTC, windings, high-voltage bushings, tank, core, cooling system
• by registration of all critical parameters associated with the characteristics of the electromagnetic, chemical and mechanical processes occurring in the transformer
• the group of parameters “partial discharges in isolation” and “vibration characteristics”, which clearly show the origin and development of degradation by temporary trends.

Conclusion

In order to conclusively prove the universality and efficiency of the PRANA system, the "connection" options for the power transformers are given: minimum, optimal, and maximum (Table 1)

Table1. Transformer "connection" options

Based on economic feasibility, the owner may choose any of the options. Even if the minimum option is chosen, the technical condition of the monitored transformer will be assessed sufficiently and in a timely manner, thanks to correctly selected parameters of primary information which optimally cover all the nodes directly or indirectly, as well as due to the method, which is unique both in terms of its transparency and objectivity.

It should be added that, at present, the corresponding mathematical models are being created and power transformers are being connected to the PRANA system for subsequent commissioning and commercial operation as objects of remote monitoring and diagnostics of technical condition.